id	sid	tid	token	lemma	pos
dad-10683	1	1	dialogue	dialogue	PROPN
dad-10683	1	2	&	&	CCONJ
dad-10683	1	3	discourse	discourse	PROPN
dad-10683	1	4	8(1	8(1	NOUN
dad-10683	1	5	)	)	PUNCT
dad-10683	1	6	(	(	PUNCT
dad-10683	1	7	2017	2017	NUM
dad-10683	1	8	)	)	PUNCT
dad-10683	1	9	31–65	31–65	NUM
dad-10683	1	10	doi	doi	NOUN
dad-10683	1	11	:	:	PUNCT
dad-10683	1	12	10.5087	10.5087	NUM
dad-10683	1	13	/	/	SYM
dad-10683	1	14	dad.2017.102	dad.2017.102	NOUN
dad-10683	1	15	training	training	NOUN
dad-10683	1	16	end	end	NOUN
dad-10683	1	17	-	-	PUNCT
dad-10683	1	18	to	to	ADP
dad-10683	1	19	-	-	PUNCT
dad-10683	1	20	end	end	NOUN
dad-10683	1	21	dialogue	dialogue	NOUN
dad-10683	1	22	systems	system	NOUN
dad-10683	1	23	with	with	ADP
dad-10683	1	24	the	the	DET
dad-10683	1	25	ubuntu	ubuntu	ADJ
dad-10683	1	26	dialogue	dialogue	PROPN
dad-10683	1	27	corpus	corpus	PROPN
dad-10683	1	28	ryan	ryan	PROPN
dad-10683	1	29	lowe	lowe	PROPN
dad-10683	1	30	ryan.lowe@cs.mcgill.ca	ryan.lowe@cs.mcgill.ca	PROPN
dad-10683	1	31	school	school	NOUN
dad-10683	1	32	of	of	ADP
dad-10683	1	33	computer	computer	NOUN
dad-10683	1	34	science	science	NOUN
dad-10683	1	35	,	,	PUNCT
dad-10683	1	36	mcgill	mcgill	PROPN
dad-10683	1	37	university	university	PROPN
dad-10683	1	38	nissan	nissan	PROPN
dad-10683	1	39	pow	pow	PROPN
dad-10683	1	40	nissan.pow@mail.mcgill.com	nissan.pow@mail.mcgill.com	PROPN
dad-10683	1	41	school	school	PROPN
dad-10683	1	42	of	of	ADP
dad-10683	1	43	computer	computer	NOUN
dad-10683	1	44	science	science	NOUN
dad-10683	1	45	,	,	PUNCT
dad-10683	1	46	mcgill	mcgill	PROPN
dad-10683	1	47	university	university	PROPN
dad-10683	1	48	iulian	iulian	PROPN
dad-10683	1	49	vlad	vlad	PROPN
dad-10683	2	1	serban	serban	PROPN
dad-10683	2	2	julianserban@gmail.com	julianserban@gmail.com	PROPN
dad-10683	2	3	diro	diro	PROPN
dad-10683	2	4	,	,	PUNCT
dad-10683	2	5	université	université	ADJ
dad-10683	2	6	de	de	ADP
dad-10683	2	7	montréal	montréal	PROPN
dad-10683	2	8	laurent	laurent	PROPN
dad-10683	2	9	charlin	charlin	PROPN
dad-10683	2	10	lcharlin@gmail.com	lcharlin@gmail.com	PROPN
dad-10683	3	1	school	school	NOUN
dad-10683	3	2	of	of	ADP
dad-10683	3	3	computer	computer	NOUN
dad-10683	3	4	science	science	NOUN
dad-10683	3	5	,	,	PUNCT
dad-10683	3	6	mcgill	mcgill	PROPN
dad-10683	3	7	university	university	PROPN
dad-10683	3	8	chia	chia	PROPN
dad-10683	3	9	-	-	PUNCT
dad-10683	3	10	wei	wei	PROPN
dad-10683	3	11	liu	liu	PROPN
dad-10683	3	12	chia-wei.liu@mail.mcgill.ca	chia-wei.liu@mail.mcgill.ca	PROPN
dad-10683	3	13	school	school	NOUN
dad-10683	3	14	of	of	ADP
dad-10683	3	15	computer	computer	NOUN
dad-10683	3	16	science	science	NOUN
dad-10683	3	17	,	,	PUNCT
dad-10683	3	18	mcgill	mcgill	PROPN
dad-10683	3	19	university	university	PROPN
dad-10683	3	20	joelle	joelle	PROPN
dad-10683	3	21	pineau	pineau	PROPN
dad-10683	3	22	jpineau@cs.mcgill.ca	jpineau@cs.mcgill.ca	PROPN
dad-10683	3	23	school	school	NOUN
dad-10683	3	24	of	of	ADP
dad-10683	3	25	computer	computer	NOUN
dad-10683	3	26	science	science	NOUN
dad-10683	3	27	,	,	PUNCT
dad-10683	3	28	mcgill	mcgill	PROPN
dad-10683	3	29	university	university	PROPN
dad-10683	3	30	editor	editor	NOUN
dad-10683	3	31	:	:	PUNCT
dad-10683	3	32	amanda	amanda	PROPN
dad-10683	3	33	stent	stent	NOUN
dad-10683	3	34	submitted	submit	VERB
dad-10683	3	35	04/2016	04/2016	PROPN
dad-10683	3	36	;	;	PUNCT
dad-10683	3	37	accepted	accept	VERB
dad-10683	3	38	12/2016	12/2016	NUM
dad-10683	3	39	;	;	PUNCT
dad-10683	3	40	published	publish	VERB
dad-10683	3	41	online	online	ADV
dad-10683	3	42	01/2017	01/2017	NUM
dad-10683	3	43	abstract	abstract	ADJ
dad-10683	3	44	in	in	ADP
dad-10683	3	45	this	this	DET
dad-10683	3	46	paper	paper	NOUN
dad-10683	3	47	,	,	PUNCT
dad-10683	3	48	we	we	PRON
dad-10683	3	49	analyze	analyze	VERB
dad-10683	3	50	neural	neural	ADJ
dad-10683	3	51	network	network	NOUN
dad-10683	3	52	-	-	PUNCT
dad-10683	3	53	based	base	VERB
dad-10683	3	54	dialogue	dialogue	NOUN
dad-10683	3	55	systems	system	NOUN
dad-10683	3	56	trained	train	VERB
dad-10683	3	57	in	in	ADP
dad-10683	3	58	an	an	DET
dad-10683	3	59	end	end	NOUN
dad-10683	3	60	-	-	PUNCT
dad-10683	3	61	to	to	ADP
dad-10683	3	62	-	-	PUNCT
dad-10683	3	63	end	end	NOUN
dad-10683	3	64	manner	manner	NOUN
dad-10683	3	65	using	use	VERB
dad-10683	3	66	an	an	DET
dad-10683	3	67	updated	update	VERB
dad-10683	3	68	version	version	NOUN
dad-10683	3	69	of	of	ADP
dad-10683	3	70	the	the	DET
dad-10683	3	71	recent	recent	ADJ
dad-10683	3	72	ubuntu	ubuntu	ADJ
dad-10683	3	73	dialogue	dialogue	NOUN
dad-10683	3	74	corpus	corpus	NOUN
dad-10683	3	75	,	,	PUNCT
dad-10683	3	76	a	a	DET
dad-10683	3	77	dataset	dataset	NOUN
dad-10683	3	78	containing	contain	VERB
dad-10683	3	79	almost	almost	ADV
dad-10683	3	80	1	1	NUM
dad-10683	3	81	million	million	NUM
dad-10683	3	82	multi	multi	ADJ
dad-10683	3	83	-	-	ADJ
dad-10683	3	84	turn	turn	ADJ
dad-10683	3	85	dialogues	dialogue	NOUN
dad-10683	3	86	,	,	PUNCT
dad-10683	3	87	with	with	ADP
dad-10683	3	88	a	a	DET
dad-10683	3	89	total	total	NOUN
dad-10683	3	90	of	of	ADP
dad-10683	3	91	over	over	ADP
dad-10683	3	92	7	7	NUM
dad-10683	3	93	million	million	NUM
dad-10683	3	94	utterances	utterance	NOUN
dad-10683	3	95	and	and	CCONJ
dad-10683	3	96	100	100	NUM
dad-10683	3	97	million	million	NUM
dad-10683	3	98	words1	words1	NOUN
dad-10683	3	99	.	.	PUNCT
dad-10683	4	1	this	this	DET
dad-10683	4	2	dataset	dataset	NOUN
dad-10683	4	3	is	be	AUX
dad-10683	4	4	interesting	interesting	ADJ
dad-10683	4	5	because	because	SCONJ
dad-10683	4	6	of	of	ADP
dad-10683	4	7	its	its	PRON
dad-10683	4	8	size	size	NOUN
dad-10683	4	9	,	,	PUNCT
dad-10683	4	10	long	long	ADJ
dad-10683	4	11	context	context	NOUN
dad-10683	4	12	lengths	length	NOUN
dad-10683	4	13	,	,	PUNCT
dad-10683	4	14	and	and	CCONJ
dad-10683	4	15	technical	technical	ADJ
dad-10683	4	16	nature	nature	NOUN
dad-10683	4	17	;	;	PUNCT
dad-10683	4	18	thus	thus	ADV
dad-10683	4	19	,	,	PUNCT
dad-10683	4	20	it	it	PRON
dad-10683	4	21	can	can	AUX
dad-10683	4	22	be	be	AUX
dad-10683	4	23	used	use	VERB
dad-10683	4	24	to	to	PART
dad-10683	4	25	train	train	VERB
dad-10683	4	26	large	large	ADJ
dad-10683	4	27	models	model	NOUN
dad-10683	4	28	directly	directly	ADV
dad-10683	4	29	from	from	ADP
dad-10683	4	30	data	datum	NOUN
dad-10683	4	31	with	with	ADP
dad-10683	4	32	minimal	minimal	ADJ
dad-10683	4	33	feature	feature	NOUN
dad-10683	4	34	engineering	engineering	NOUN
dad-10683	4	35	.	.	PUNCT
dad-10683	5	1	we	we	PRON
dad-10683	5	2	provide	provide	VERB
dad-10683	5	3	baselines	baseline	NOUN
dad-10683	5	4	in	in	ADP
dad-10683	5	5	two	two	NUM
dad-10683	5	6	different	different	ADJ
dad-10683	5	7	environments	environment	NOUN
dad-10683	5	8	:	:	PUNCT
dad-10683	5	9	one	one	NUM
dad-10683	5	10	where	where	SCONJ
dad-10683	5	11	models	model	NOUN
dad-10683	5	12	are	be	AUX
dad-10683	5	13	trained	train	VERB
dad-10683	5	14	to	to	PART
dad-10683	5	15	select	select	VERB
dad-10683	5	16	the	the	DET
dad-10683	5	17	correct	correct	ADJ
dad-10683	5	18	next	next	ADJ
dad-10683	5	19	response	response	NOUN
dad-10683	5	20	from	from	ADP
dad-10683	5	21	a	a	DET
dad-10683	5	22	list	list	NOUN
dad-10683	5	23	of	of	ADP
dad-10683	5	24	candidate	candidate	NOUN
dad-10683	5	25	responses	response	NOUN
dad-10683	5	26	,	,	PUNCT
dad-10683	5	27	and	and	CCONJ
dad-10683	5	28	one	one	NUM
dad-10683	5	29	where	where	SCONJ
dad-10683	5	30	models	model	NOUN
dad-10683	5	31	are	be	AUX
dad-10683	5	32	trained	train	VERB
dad-10683	5	33	to	to	PART
dad-10683	5	34	maximize	maximize	VERB
dad-10683	5	35	the	the	DET
dad-10683	5	36	loglikelihood	loglikelihood	NOUN
dad-10683	5	37	of	of	ADP
dad-10683	5	38	a	a	DET
dad-10683	5	39	generated	generate	VERB
dad-10683	5	40	utterance	utterance	NOUN
dad-10683	5	41	conditioned	condition	VERB
dad-10683	5	42	on	on	ADP
dad-10683	5	43	the	the	DET
dad-10683	5	44	context	context	NOUN
dad-10683	5	45	of	of	ADP
dad-10683	5	46	the	the	DET
dad-10683	5	47	conversation	conversation	NOUN
dad-10683	5	48	.	.	PUNCT
dad-10683	6	1	these	these	PRON
dad-10683	6	2	are	be	AUX
dad-10683	6	3	both	both	PRON
dad-10683	6	4	evaluated	evaluate	VERB
dad-10683	6	5	on	on	ADP
dad-10683	6	6	a	a	DET
dad-10683	6	7	recall	recall	NOUN
dad-10683	6	8	task	task	NOUN
dad-10683	6	9	that	that	PRON
dad-10683	6	10	we	we	PRON
dad-10683	6	11	call	call	VERB
dad-10683	6	12	next	next	ADJ
dad-10683	6	13	utterance	utterance	NOUN
dad-10683	6	14	classification	classification	NOUN
dad-10683	6	15	(	(	PUNCT
dad-10683	6	16	nuc	nuc	NOUN
dad-10683	6	17	)	)	PUNCT
dad-10683	6	18	,	,	PUNCT
dad-10683	6	19	and	and	CCONJ
dad-10683	6	20	using	use	VERB
dad-10683	6	21	vector	vector	NOUN
dad-10683	6	22	-	-	PUNCT
dad-10683	6	23	based	base	VERB
dad-10683	6	24	metrics	metric	NOUN
dad-10683	6	25	that	that	PRON
dad-10683	6	26	capture	capture	VERB
dad-10683	6	27	the	the	DET
dad-10683	6	28	topicality	topicality	NOUN
dad-10683	6	29	of	of	ADP
dad-10683	6	30	the	the	DET
dad-10683	6	31	responses	response	NOUN
dad-10683	6	32	.	.	PUNCT
dad-10683	7	1	we	we	PRON
dad-10683	7	2	observe	observe	VERB
dad-10683	7	3	that	that	SCONJ
dad-10683	7	4	current	current	ADJ
dad-10683	7	5	end	end	NOUN
dad-10683	7	6	-	-	PUNCT
dad-10683	7	7	to	to	ADP
dad-10683	7	8	-	-	PUNCT
dad-10683	7	9	end	end	NOUN
dad-10683	7	10	models	model	NOUN
dad-10683	7	11	are	be	AUX
dad-10683	7	12	1	1	NUM
dad-10683	7	13	.	.	PUNCT
dad-10683	8	1	this	this	DET
dad-10683	8	2	work	work	NOUN
dad-10683	8	3	is	be	AUX
dad-10683	8	4	an	an	DET
dad-10683	8	5	extension	extension	NOUN
dad-10683	8	6	of	of	ADP
dad-10683	8	7	a	a	DET
dad-10683	8	8	paper	paper	NOUN
dad-10683	8	9	appearing	appear	VERB
dad-10683	8	10	in	in	ADP
dad-10683	8	11	sigdial	sigdial	ADJ
dad-10683	8	12	(	(	PUNCT
dad-10683	8	13	lowe	lowe	PROPN
dad-10683	8	14	et	et	PROPN
dad-10683	8	15	al	al	PROPN
dad-10683	8	16	.	.	PROPN
dad-10683	8	17	,	,	PUNCT
dad-10683	8	18	2015	2015	NUM
dad-10683	8	19	)	)	PUNCT
dad-10683	8	20	.	.	PUNCT
dad-10683	9	1	this	this	DET
dad-10683	9	2	paper	paper	NOUN
dad-10683	9	3	further	far	ADV
dad-10683	9	4	includes	include	VERB
dad-10683	9	5	results	result	NOUN
dad-10683	9	6	on	on	ADP
dad-10683	9	7	generative	generative	ADJ
dad-10683	9	8	dialogue	dialogue	NOUN
dad-10683	9	9	models	model	NOUN
dad-10683	9	10	,	,	PUNCT
dad-10683	9	11	more	more	ADV
dad-10683	9	12	extensive	extensive	ADJ
dad-10683	9	13	evaluation	evaluation	NOUN
dad-10683	9	14	of	of	ADP
dad-10683	9	15	the	the	DET
dad-10683	9	16	retrieval	retrieval	NOUN
dad-10683	9	17	models	model	NOUN
dad-10683	9	18	using	use	VERB
dad-10683	9	19	vector	vector	NOUN
dad-10683	9	20	-	-	PUNCT
dad-10683	9	21	based	base	VERB
dad-10683	9	22	generative	generative	ADJ
dad-10683	9	23	metrics	metric	NOUN
dad-10683	9	24	,	,	PUNCT
dad-10683	9	25	and	and	CCONJ
dad-10683	9	26	a	a	DET
dad-10683	9	27	qualitative	qualitative	ADJ
dad-10683	9	28	examination	examination	NOUN
dad-10683	9	29	of	of	ADP
dad-10683	9	30	responses	response	NOUN
dad-10683	9	31	from	from	ADP
dad-10683	9	32	the	the	DET
dad-10683	9	33	generative	generative	ADJ
dad-10683	9	34	models	model	NOUN
dad-10683	9	35	and	and	CCONJ
dad-10683	9	36	classification	classification	NOUN
dad-10683	9	37	errors	error	NOUN
dad-10683	9	38	made	make	VERB
dad-10683	9	39	by	by	ADP
dad-10683	9	40	the	the	DET
dad-10683	9	41	dual	dual	ADJ
dad-10683	9	42	encoder	encoder	NOUN
dad-10683	9	43	model	model	NOUN
dad-10683	9	44	.	.	PUNCT
dad-10683	10	1	experiments	experiment	NOUN
dad-10683	10	2	are	be	AUX
dad-10683	10	3	performed	perform	VERB
dad-10683	10	4	on	on	ADP
dad-10683	10	5	a	a	DET
dad-10683	10	6	new	new	ADJ
dad-10683	10	7	version	version	NOUN
dad-10683	10	8	of	of	ADP
dad-10683	10	9	the	the	DET
dad-10683	10	10	corpus	corpus	NOUN
dad-10683	10	11	,	,	PUNCT
dad-10683	10	12	the	the	DET
dad-10683	10	13	ubuntu	ubuntu	ADJ
dad-10683	10	14	dialogue	dialogue	NOUN
dad-10683	10	15	corpus	corpus	PROPN
dad-10683	10	16	v2	v2	PROPN
dad-10683	10	17	,	,	PUNCT
dad-10683	10	18	which	which	PRON
dad-10683	10	19	is	be	AUX
dad-10683	10	20	publicly	publicly	ADV
dad-10683	10	21	available	available	ADJ
dad-10683	10	22	:	:	PUNCT
dad-10683	11	1	https://github.com/rkadlec/ubuntu-ranking-dataset-creator	https://github.com/rkadlec/ubuntu-ranking-dataset-creator	X
dad-10683	11	2	.	.	PUNCT
dad-10683	12	1	the	the	DET
dad-10683	12	2	early	early	ADJ
dad-10683	12	3	dataset	dataset	NOUN
dad-10683	12	4	has	have	AUX
dad-10683	12	5	been	be	AUX
dad-10683	12	6	updated	update	VERB
dad-10683	12	7	to	to	PART
dad-10683	12	8	add	add	VERB
dad-10683	12	9	features	feature	NOUN
dad-10683	12	10	and	and	CCONJ
dad-10683	12	11	fix	fix	VERB
dad-10683	12	12	bugs	bug	NOUN
dad-10683	12	13	,	,	PUNCT
dad-10683	12	14	which	which	PRON
dad-10683	12	15	are	be	AUX
dad-10683	12	16	detailed	detail	VERB
dad-10683	12	17	in	in	ADP
dad-10683	12	18	section	section	NOUN
dad-10683	12	19	3	3	NUM
dad-10683	12	20	.	.	PUNCT
dad-10683	13	1	c	c	PROPN
dad-10683	13	2	©	©	PROPN
dad-10683	13	3	2017	2017	NUM
dad-10683	13	4	ryan	ryan	PROPN
dad-10683	13	5	lowe	lowe	PROPN
dad-10683	13	6	,	,	PUNCT
dad-10683	13	7	nissan	nissan	PROPN
dad-10683	13	8	pow	pow	PROPN
dad-10683	13	9	,	,	PUNCT
dad-10683	13	10	iulian	iulian	PROPN
dad-10683	13	11	vlad	vlad	PROPN
dad-10683	13	12	serban	serban	PROPN
dad-10683	13	13	,	,	PUNCT
dad-10683	13	14	laurent	laurent	PROPN
dad-10683	13	15	charlinn	charlinn	PROPN
dad-10683	13	16	,	,	PUNCT
dad-10683	13	17	chia	chia	PROPN
dad-10683	13	18	-	-	PUNCT
dad-10683	13	19	wei	wei	PROPN
dad-10683	13	20	liu	liu	PROPN
dad-10683	13	21	and	and	CCONJ
dad-10683	13	22	joelle	joelle	PROPN
dad-10683	13	23	pineau	pineau	PROPN
dad-10683	13	24	this	this	PRON
dad-10683	13	25	is	be	AUX
dad-10683	13	26	an	an	DET
dad-10683	13	27	open	open	ADJ
dad-10683	13	28	-	-	PUNCT
dad-10683	13	29	access	access	NOUN
dad-10683	13	30	article	article	NOUN
dad-10683	13	31	distributed	distribute	VERB
dad-10683	13	32	under	under	ADP
dad-10683	13	33	the	the	DET
dad-10683	13	34	terms	term	NOUN
dad-10683	13	35	of	of	ADP
dad-10683	13	36	a	a	DET
dad-10683	13	37	creative	creative	ADJ
dad-10683	13	38	commons	common	NOUN
dad-10683	13	39	attribution	attribution	NOUN
dad-10683	13	40	license	license	NOUN
dad-10683	13	41	(	(	PUNCT
dad-10683	13	42	http	http	NOUN
dad-10683	13	43	:	:	PUNCT
dad-10683	13	44	//creativecommons.org	//creativecommons.org	SYM
dad-10683	13	45	/	/	SYM
dad-10683	13	46	licenses	license	NOUN
dad-10683	13	47	/	/	SYM
dad-10683	13	48	by/3.0/	by/3.0/	NOUN
dad-10683	13	49	)	)	PUNCT
dad-10683	13	50	.	.	PUNCT
dad-10683	14	1	lowe	lowe	PROPN
dad-10683	14	2	,	,	PUNCT
dad-10683	14	3	pow	pow	PROPN
dad-10683	14	4	,	,	PUNCT
dad-10683	14	5	serban	serban	PROPN
dad-10683	14	6	,	,	PUNCT
dad-10683	14	7	charlinn	charlinn	PROPN
dad-10683	14	8	,	,	PUNCT
dad-10683	14	9	liu	liu	PROPN
dad-10683	14	10	and	and	CCONJ
dad-10683	14	11	pineau	pineau	PROPN
dad-10683	14	12	unable	unable	ADJ
dad-10683	14	13	to	to	PART
dad-10683	14	14	completely	completely	ADV
dad-10683	14	15	solve	solve	VERB
dad-10683	14	16	these	these	DET
dad-10683	14	17	tasks	task	NOUN
dad-10683	14	18	;	;	PUNCT
dad-10683	14	19	thus	thus	ADV
dad-10683	14	20	,	,	PUNCT
dad-10683	14	21	we	we	PRON
dad-10683	14	22	provide	provide	VERB
dad-10683	14	23	a	a	DET
dad-10683	14	24	qualitative	qualitative	ADJ
dad-10683	14	25	error	error	NOUN
dad-10683	14	26	analysis	analysis	NOUN
dad-10683	14	27	to	to	PART
dad-10683	14	28	determine	determine	VERB
dad-10683	14	29	the	the	DET
dad-10683	14	30	primary	primary	ADJ
dad-10683	14	31	causes	cause	NOUN
dad-10683	14	32	of	of	ADP
dad-10683	14	33	error	error	NOUN
dad-10683	14	34	for	for	ADP
dad-10683	14	35	end	end	NOUN
dad-10683	14	36	-	-	PUNCT
dad-10683	14	37	to	to	ADP
dad-10683	14	38	-	-	PUNCT
dad-10683	14	39	end	end	NOUN
dad-10683	14	40	models	model	NOUN
dad-10683	14	41	evaluated	evaluate	VERB
dad-10683	14	42	on	on	ADP
dad-10683	14	43	nuc	nuc	PROPN
dad-10683	14	44	,	,	PUNCT
dad-10683	14	45	and	and	CCONJ
dad-10683	14	46	examine	examine	VERB
dad-10683	14	47	sample	sample	NOUN
dad-10683	14	48	utterances	utterance	NOUN
dad-10683	14	49	from	from	ADP
dad-10683	14	50	the	the	DET
dad-10683	14	51	generative	generative	ADJ
dad-10683	14	52	models	model	NOUN
dad-10683	14	53	.	.	PUNCT
dad-10683	15	1	as	as	ADP
dad-10683	15	2	a	a	DET
dad-10683	15	3	result	result	NOUN
dad-10683	15	4	of	of	ADP
dad-10683	15	5	this	this	DET
dad-10683	15	6	analysis	analysis	NOUN
dad-10683	15	7	,	,	PUNCT
dad-10683	15	8	we	we	PRON
dad-10683	15	9	suggest	suggest	VERB
dad-10683	15	10	some	some	DET
dad-10683	15	11	promising	promising	ADJ
dad-10683	15	12	directions	direction	NOUN
dad-10683	15	13	for	for	ADP
dad-10683	15	14	future	future	ADJ
dad-10683	15	15	research	research	NOUN
dad-10683	15	16	on	on	ADP
dad-10683	15	17	the	the	DET
dad-10683	15	18	ubuntu	ubuntu	ADJ
dad-10683	15	19	dialogue	dialogue	NOUN
dad-10683	15	20	corpus	corpus	PROPN
dad-10683	15	21	,	,	PUNCT
dad-10683	15	22	which	which	PRON
dad-10683	15	23	can	can	AUX
dad-10683	15	24	also	also	ADV
dad-10683	15	25	be	be	AUX
dad-10683	15	26	applied	apply	VERB
dad-10683	15	27	to	to	ADP
dad-10683	15	28	end	end	NOUN
dad-10683	15	29	-	-	PUNCT
dad-10683	15	30	to	to	ADP
dad-10683	15	31	-	-	PUNCT
dad-10683	15	32	end	end	NOUN
dad-10683	15	33	dialogue	dialogue	NOUN
dad-10683	15	34	systems	system	NOUN
dad-10683	15	35	in	in	ADP
dad-10683	15	36	general	general	ADJ
dad-10683	15	37	.	.	PUNCT
dad-10683	16	1	1	1	X
dad-10683	16	2	.	.	X
dad-10683	16	3	introduction	introduction	NOUN
dad-10683	16	4	deriving	derive	VERB
dad-10683	16	5	statistical	statistical	ADJ
dad-10683	16	6	models	model	NOUN
dad-10683	16	7	that	that	PRON
dad-10683	16	8	can	can	AUX
dad-10683	16	9	naturally	naturally	ADV
dad-10683	16	10	and	and	CCONJ
dad-10683	16	11	coherently	coherently	ADV
dad-10683	16	12	converse	converse	NOUN
dad-10683	16	13	with	with	ADP
dad-10683	16	14	humans	human	NOUN
dad-10683	16	15	is	be	AUX
dad-10683	16	16	one	one	NUM
dad-10683	16	17	of	of	ADP
dad-10683	16	18	the	the	DET
dad-10683	16	19	cornerstone	cornerstone	NOUN
dad-10683	16	20	problems	problem	NOUN
dad-10683	16	21	of	of	ADP
dad-10683	16	22	artificial	artificial	ADJ
dad-10683	16	23	intelligence	intelligence	NOUN
dad-10683	16	24	.	.	PUNCT
dad-10683	17	1	until	until	ADP
dad-10683	17	2	recently	recently	ADV
dad-10683	17	3	,	,	PUNCT
dad-10683	17	4	such	such	ADJ
dad-10683	17	5	models	model	NOUN
dad-10683	17	6	required	require	VERB
dad-10683	17	7	significant	significant	ADJ
dad-10683	17	8	handengineering	handengineering	NOUN
dad-10683	17	9	of	of	ADP
dad-10683	17	10	features	feature	NOUN
dad-10683	17	11	,	,	PUNCT
dad-10683	17	12	and	and	CCONJ
dad-10683	17	13	thus	thus	ADV
dad-10683	17	14	could	could	AUX
dad-10683	17	15	only	only	ADV
dad-10683	17	16	generate	generate	VERB
dad-10683	17	17	a	a	DET
dad-10683	17	18	limited	limited	ADJ
dad-10683	17	19	number	number	NOUN
dad-10683	17	20	of	of	ADP
dad-10683	17	21	responses	response	NOUN
dad-10683	17	22	and	and	CCONJ
dad-10683	17	23	be	be	AUX
dad-10683	17	24	deployed	deploy	VERB
dad-10683	17	25	in	in	ADP
dad-10683	17	26	constrained	constrained	ADJ
dad-10683	17	27	situations	situation	NOUN
dad-10683	17	28	.	.	PUNCT
dad-10683	18	1	recent	recent	ADJ
dad-10683	18	2	advances	advance	NOUN
dad-10683	18	3	in	in	ADP
dad-10683	18	4	neural	neural	ADJ
dad-10683	18	5	network	network	NOUN
dad-10683	18	6	-	-	PUNCT
dad-10683	18	7	based	base	VERB
dad-10683	18	8	language	language	NOUN
dad-10683	18	9	models	model	NOUN
dad-10683	18	10	have	have	AUX
dad-10683	18	11	begun	begin	VERB
dad-10683	18	12	to	to	PART
dad-10683	18	13	make	make	VERB
dad-10683	18	14	feasible	feasible	ADJ
dad-10683	18	15	the	the	DET
dad-10683	18	16	idea	idea	NOUN
dad-10683	18	17	of	of	ADP
dad-10683	18	18	learning	learn	VERB
dad-10683	18	19	an	an	DET
dad-10683	18	20	entire	entire	ADJ
dad-10683	18	21	dialogue	dialogue	NOUN
dad-10683	18	22	model	model	NOUN
dad-10683	18	23	directly	directly	ADV
dad-10683	18	24	from	from	ADP
dad-10683	18	25	conversational	conversational	ADJ
dad-10683	18	26	data	datum	NOUN
dad-10683	18	27	,	,	PUNCT
dad-10683	18	28	with	with	ADP
dad-10683	18	29	humans	human	NOUN
dad-10683	18	30	only	only	ADV
dad-10683	18	31	specifying	specify	VERB
dad-10683	18	32	the	the	DET
dad-10683	18	33	model	model	NOUN
dad-10683	18	34	hyper	hyper	NOUN
dad-10683	18	35	-	-	NOUN
dad-10683	18	36	parameters	parameter	NOUN
dad-10683	18	37	.	.	PUNCT
dad-10683	19	1	however	however	ADV
dad-10683	19	2	,	,	PUNCT
dad-10683	19	3	significant	significant	ADJ
dad-10683	19	4	work	work	NOUN
dad-10683	19	5	needs	need	VERB
dad-10683	19	6	to	to	PART
dad-10683	19	7	be	be	AUX
dad-10683	19	8	done	do	VERB
dad-10683	19	9	before	before	SCONJ
dad-10683	19	10	these	these	DET
dad-10683	19	11	models	model	NOUN
dad-10683	19	12	can	can	AUX
dad-10683	19	13	be	be	AUX
dad-10683	19	14	implemented	implement	VERB
dad-10683	19	15	in	in	ADP
dad-10683	19	16	practice	practice	NOUN
dad-10683	19	17	with	with	ADP
dad-10683	19	18	high	high	ADJ
dad-10683	19	19	confidence	confidence	NOUN
dad-10683	19	20	.	.	PUNCT
dad-10683	20	1	in	in	ADP
dad-10683	20	2	this	this	DET
dad-10683	20	3	paper	paper	NOUN
dad-10683	20	4	we	we	PRON
dad-10683	20	5	consider	consider	VERB
dad-10683	20	6	the	the	DET
dad-10683	20	7	problem	problem	NOUN
dad-10683	20	8	of	of	ADP
dad-10683	20	9	building	build	VERB
dad-10683	20	10	dialogue	dialogue	NOUN
dad-10683	20	11	agents	agent	NOUN
dad-10683	20	12	in	in	ADP
dad-10683	20	13	an	an	DET
dad-10683	20	14	end	end	NOUN
dad-10683	20	15	-	-	PUNCT
dad-10683	20	16	to	to	ADP
dad-10683	20	17	-	-	PUNCT
dad-10683	20	18	end	end	NOUN
dad-10683	20	19	manner	manner	NOUN
dad-10683	20	20	.	.	PUNCT
dad-10683	21	1	we	we	PRON
dad-10683	21	2	define	define	VERB
dad-10683	21	3	end	end	NOUN
dad-10683	21	4	-	-	PUNCT
dad-10683	21	5	to	to	ADP
dad-10683	21	6	-	-	PUNCT
dad-10683	21	7	end	end	NOUN
dad-10683	21	8	systems	system	NOUN
dad-10683	21	9	,	,	PUNCT
dad-10683	21	10	contrary	contrary	ADJ
dad-10683	21	11	to	to	ADP
dad-10683	21	12	modular	modular	ADJ
dad-10683	21	13	systems	system	NOUN
dad-10683	21	14	,	,	PUNCT
dad-10683	21	15	as	as	ADP
dad-10683	21	16	those	those	PRON
dad-10683	21	17	that	that	PRON
dad-10683	21	18	are	be	AUX
dad-10683	21	19	trained	train	VERB
dad-10683	21	20	directly	directly	ADV
dad-10683	21	21	from	from	ADP
dad-10683	21	22	conversational	conversational	ADJ
dad-10683	21	23	data	datum	NOUN
dad-10683	21	24	to	to	PART
dad-10683	21	25	optimize	optimize	VERB
dad-10683	21	26	a	a	DET
dad-10683	21	27	single	single	ADJ
dad-10683	21	28	objective	objective	ADJ
dad-10683	21	29	function	function	NOUN
dad-10683	21	30	(	(	PUNCT
dad-10683	21	31	see	see	VERB
dad-10683	21	32	section	section	NOUN
dad-10683	21	33	1.1	1.1	NUM
dad-10683	21	34	)	)	PUNCT
dad-10683	21	35	.	.	PUNCT
dad-10683	22	1	we	we	PRON
dad-10683	22	2	use	use	VERB
dad-10683	22	3	the	the	DET
dad-10683	22	4	recently	recently	ADV
dad-10683	22	5	released	release	VERB
dad-10683	22	6	ubuntu	ubuntu	ADJ
dad-10683	22	7	dialogue	dialogue	NOUN
dad-10683	22	8	corpus	corpus	NOUN
dad-10683	22	9	,	,	PUNCT
dad-10683	22	10	which	which	PRON
dad-10683	22	11	consists	consist	VERB
dad-10683	22	12	of	of	ADP
dad-10683	22	13	almost	almost	ADV
dad-10683	22	14	one	one	NUM
dad-10683	22	15	million	million	NUM
dad-10683	22	16	two	two	NUM
dad-10683	22	17	-	-	PUNCT
dad-10683	22	18	person	person	NOUN
dad-10683	22	19	(	(	PUNCT
dad-10683	22	20	dyadic	dyadic	ADJ
dad-10683	22	21	)	)	PUNCT
dad-10683	22	22	conversations	conversation	NOUN
dad-10683	22	23	extracted	extract	VERB
dad-10683	22	24	from	from	ADP
dad-10683	22	25	the	the	DET
dad-10683	22	26	ubuntu	ubuntu	NOUN
dad-10683	22	27	chat	chat	NOUN
dad-10683	22	28	logs	log	NOUN
dad-10683	22	29	,	,	PUNCT
dad-10683	22	30	which	which	PRON
dad-10683	22	31	provide	provide	VERB
dad-10683	22	32	technical	technical	ADJ
dad-10683	22	33	support	support	NOUN
dad-10683	22	34	for	for	ADP
dad-10683	22	35	various	various	ADJ
dad-10683	22	36	ubunturelated	ubunturelated	ADJ
dad-10683	22	37	problems	problem	NOUN
dad-10683	22	38	.	.	PUNCT
dad-10683	23	1	dialogues	dialogue	NOUN
dad-10683	23	2	in	in	ADP
dad-10683	23	3	the	the	DET
dad-10683	23	4	corpus	corpus	NOUN
dad-10683	23	5	are	be	AUX
dad-10683	23	6	multi	multi	ADJ
dad-10683	23	7	-	-	ADJ
dad-10683	23	8	turn	turn	VERB
dad-10683	23	9	and	and	CCONJ
dad-10683	23	10	unstructured	unstructure	VERB
dad-10683	23	11	,	,	PUNCT
dad-10683	23	12	as	as	SCONJ
dad-10683	23	13	there	there	PRON
dad-10683	23	14	is	be	VERB
dad-10683	23	15	no	no	DET
dad-10683	23	16	a	a	DET
dad-10683	23	17	priori	priori	ADJ
dad-10683	23	18	logical	logical	ADJ
dad-10683	23	19	representation	representation	NOUN
dad-10683	23	20	for	for	ADP
dad-10683	23	21	the	the	DET
dad-10683	23	22	information	information	NOUN
dad-10683	23	23	exchanged	exchange	VERB
dad-10683	23	24	during	during	ADP
dad-10683	23	25	the	the	DET
dad-10683	23	26	conversation	conversation	NOUN
dad-10683	23	27	.	.	PUNCT
dad-10683	24	1	this	this	PRON
dad-10683	24	2	is	be	AUX
dad-10683	24	3	in	in	ADP
dad-10683	24	4	contrast	contrast	NOUN
dad-10683	24	5	to	to	ADP
dad-10683	24	6	recent	recent	ADJ
dad-10683	24	7	systems	system	NOUN
dad-10683	24	8	which	which	PRON
dad-10683	24	9	focus	focus	VERB
dad-10683	24	10	on	on	ADP
dad-10683	24	11	structured	structured	ADJ
dad-10683	24	12	dialogue	dialogue	NOUN
dad-10683	24	13	tasks	task	NOUN
dad-10683	24	14	,	,	PUNCT
dad-10683	24	15	using	use	VERB
dad-10683	24	16	slot	slot	NOUN
dad-10683	24	17	-	-	PUNCT
dad-10683	24	18	filling	fill	VERB
dad-10683	24	19	representations	representation	NOUN
dad-10683	24	20	(	(	PUNCT
dad-10683	24	21	williams	williams	PROPN
dad-10683	24	22	et	et	PROPN
dad-10683	24	23	al	al	PROPN
dad-10683	24	24	.	.	PROPN
dad-10683	24	25	,	,	PUNCT
dad-10683	24	26	2013	2013	NUM
dad-10683	24	27	;	;	PUNCT
dad-10683	24	28	henderson	henderson	PROPN
dad-10683	24	29	et	et	PROPN
dad-10683	24	30	al	al	PROPN
dad-10683	24	31	.	.	PROPN
dad-10683	24	32	,	,	PUNCT
dad-10683	24	33	2014a	2014a	NUM
dad-10683	24	34	;	;	PUNCT
dad-10683	24	35	singh	singh	PROPN
dad-10683	24	36	et	et	PROPN
dad-10683	24	37	al	al	PROPN
dad-10683	24	38	.	.	PROPN
dad-10683	24	39	,	,	PUNCT
dad-10683	24	40	2002	2002	NUM
dad-10683	24	41	)	)	PUNCT
dad-10683	24	42	.	.	PUNCT
dad-10683	25	1	the	the	DET
dad-10683	25	2	creation	creation	NOUN
dad-10683	25	3	of	of	ADP
dad-10683	25	4	such	such	DET
dad-10683	25	5	a	a	DET
dad-10683	25	6	large	large	ADJ
dad-10683	25	7	,	,	PUNCT
dad-10683	25	8	unstructured	unstructured	ADJ
dad-10683	25	9	dialogue	dialogue	NOUN
dad-10683	25	10	dataset	dataset	NOUN
dad-10683	25	11	was	be	AUX
dad-10683	25	12	motivated	motivate	VERB
dad-10683	25	13	by	by	ADP
dad-10683	25	14	observations	observation	NOUN
dad-10683	25	15	of	of	ADP
dad-10683	25	16	progress	progress	NOUN
dad-10683	25	17	in	in	ADP
dad-10683	25	18	various	various	ADJ
dad-10683	25	19	sub	sub	NOUN
dad-10683	25	20	-	-	NOUN
dad-10683	25	21	fields	field	NOUN
dad-10683	25	22	of	of	ADP
dad-10683	25	23	ai	ai	NOUN
dad-10683	25	24	.	.	PUNCT
dad-10683	26	1	in	in	ADP
dad-10683	26	2	particular	particular	ADJ
dad-10683	26	3	,	,	PUNCT
dad-10683	26	4	it	it	PRON
dad-10683	26	5	has	have	AUX
dad-10683	26	6	been	be	AUX
dad-10683	26	7	argued	argue	VERB
dad-10683	26	8	that	that	SCONJ
dad-10683	26	9	this	this	DET
dad-10683	26	10	progress	progress	NOUN
dad-10683	26	11	can	can	AUX
dad-10683	26	12	be	be	AUX
dad-10683	26	13	attributed	attribute	VERB
dad-10683	26	14	to	to	ADP
dad-10683	26	15	three	three	NUM
dad-10683	26	16	major	major	ADJ
dad-10683	26	17	factors	factor	NOUN
dad-10683	26	18	:	:	PUNCT
dad-10683	26	19	1	1	X
dad-10683	26	20	)	)	PUNCT
dad-10683	26	21	the	the	DET
dad-10683	26	22	public	public	ADJ
dad-10683	26	23	distribution	distribution	NOUN
dad-10683	26	24	of	of	ADP
dad-10683	26	25	very	very	ADV
dad-10683	26	26	large	large	ADJ
dad-10683	26	27	rich	rich	ADJ
dad-10683	26	28	datasets	dataset	NOUN
dad-10683	26	29	(	(	PUNCT
dad-10683	26	30	deng	deng	PROPN
dad-10683	26	31	et	et	PROPN
dad-10683	26	32	al	al	PROPN
dad-10683	26	33	.	.	PROPN
dad-10683	26	34	,	,	PUNCT
dad-10683	26	35	2009	2009	NUM
dad-10683	26	36	)	)	PUNCT
dad-10683	26	37	,	,	PUNCT
dad-10683	26	38	2	2	X
dad-10683	26	39	)	)	PUNCT
dad-10683	26	40	the	the	DET
dad-10683	26	41	availability	availability	NOUN
dad-10683	26	42	of	of	ADP
dad-10683	26	43	substantial	substantial	ADJ
dad-10683	26	44	computing	computing	NOUN
dad-10683	26	45	power	power	NOUN
dad-10683	26	46	,	,	PUNCT
dad-10683	26	47	and	and	CCONJ
dad-10683	26	48	3	3	X
dad-10683	26	49	)	)	PUNCT
dad-10683	26	50	the	the	DET
dad-10683	26	51	development	development	NOUN
dad-10683	26	52	of	of	ADP
dad-10683	26	53	new	new	ADJ
dad-10683	26	54	training	training	NOUN
dad-10683	26	55	methods	method	NOUN
dad-10683	26	56	for	for	ADP
dad-10683	26	57	neural	neural	ADJ
dad-10683	26	58	architectures	architecture	NOUN
dad-10683	26	59	,	,	PUNCT
dad-10683	26	60	in	in	ADP
dad-10683	26	61	particular	particular	ADJ
dad-10683	26	62	leveraging	leverage	VERB
dad-10683	26	63	unlabeled	unlabeled	ADJ
dad-10683	26	64	data	datum	NOUN
dad-10683	26	65	.	.	PUNCT
dad-10683	27	1	we	we	PRON
dad-10683	27	2	conduct	conduct	VERB
dad-10683	27	3	an	an	DET
dad-10683	27	4	analysis	analysis	NOUN
dad-10683	27	5	of	of	ADP
dad-10683	27	6	several	several	ADJ
dad-10683	27	7	dialogue	dialogue	NOUN
dad-10683	27	8	models	model	NOUN
dad-10683	27	9	that	that	PRON
dad-10683	27	10	can	can	AUX
dad-10683	27	11	be	be	AUX
dad-10683	27	12	used	use	VERB
dad-10683	27	13	in	in	ADP
dad-10683	27	14	conjunction	conjunction	NOUN
dad-10683	27	15	with	with	ADP
dad-10683	27	16	the	the	DET
dad-10683	27	17	ubuntu	ubuntu	ADJ
dad-10683	27	18	dialogue	dialogue	NOUN
dad-10683	27	19	corpus	corpus	NOUN
dad-10683	27	20	.	.	PUNCT
dad-10683	28	1	we	we	PRON
dad-10683	28	2	first	first	ADV
dad-10683	28	3	consider	consider	VERB
dad-10683	28	4	classification	classification	NOUN
dad-10683	28	5	models	model	NOUN
dad-10683	28	6	,	,	PUNCT
dad-10683	28	7	which	which	PRON
dad-10683	28	8	are	be	AUX
dad-10683	28	9	trained	train	VERB
dad-10683	28	10	to	to	PART
dad-10683	28	11	select	select	VERB
dad-10683	28	12	the	the	DET
dad-10683	28	13	correct	correct	ADJ
dad-10683	28	14	next	next	ADJ
dad-10683	28	15	response	response	NOUN
dad-10683	28	16	of	of	ADP
dad-10683	28	17	a	a	DET
dad-10683	28	18	conversation	conversation	NOUN
dad-10683	28	19	from	from	ADP
dad-10683	28	20	a	a	DET
dad-10683	28	21	list	list	NOUN
dad-10683	28	22	of	of	ADP
dad-10683	28	23	candidate	candidate	NOUN
dad-10683	28	24	responses	response	NOUN
dad-10683	28	25	.	.	PUNCT
dad-10683	29	1	we	we	PRON
dad-10683	29	2	use	use	VERB
dad-10683	29	3	a	a	DET
dad-10683	29	4	baseline	baseline	ADJ
dad-10683	29	5	model	model	NOUN
dad-10683	29	6	that	that	PRON
dad-10683	29	7	calculates	calculate	VERB
dad-10683	29	8	term	term	NOUN
dad-10683	29	9	frequency	frequency	NOUN
dad-10683	29	10	-	-	PUNCT
dad-10683	29	11	inverse	inverse	NOUN
dad-10683	29	12	document	document	NOUN
dad-10683	29	13	frequency	frequency	NOUN
dad-10683	29	14	(	(	PUNCT
dad-10683	29	15	tf	tf	PROPN
dad-10683	29	16	-	-	PUNCT
dad-10683	29	17	idf	idf	NOUN
dad-10683	29	18	)	)	PUNCT
dad-10683	29	19	between	between	ADP
dad-10683	29	20	the	the	DET
dad-10683	29	21	context	context	NOUN
dad-10683	29	22	and	and	CCONJ
dad-10683	29	23	each	each	DET
dad-10683	29	24	response	response	NOUN
dad-10683	29	25	,	,	PUNCT
dad-10683	29	26	and	and	CCONJ
dad-10683	29	27	compare	compare	VERB
dad-10683	29	28	it	it	PRON
dad-10683	29	29	to	to	ADP
dad-10683	29	30	a	a	DET
dad-10683	29	31	dual	dual	ADJ
dad-10683	29	32	encoder	encoder	NOUN
dad-10683	29	33	(	(	PUNCT
dad-10683	29	34	de	de	ADJ
dad-10683	29	35	)	)	PUNCT
dad-10683	29	36	model	model	NOUN
dad-10683	29	37	using	use	VERB
dad-10683	29	38	both	both	CCONJ
dad-10683	29	39	recurrent	recurrent	ADJ
dad-10683	29	40	neural	neural	ADJ
dad-10683	29	41	networks	network	NOUN
dad-10683	29	42	(	(	PUNCT
dad-10683	29	43	rnns	rnns	PROPN
dad-10683	29	44	)	)	PUNCT
dad-10683	29	45	and	and	CCONJ
dad-10683	29	46	long	long	ADJ
dad-10683	29	47	short	short	ADJ
dad-10683	29	48	-	-	PUNCT
dad-10683	29	49	term	term	NOUN
dad-10683	29	50	memory	memory	NOUN
dad-10683	29	51	(	(	PUNCT
dad-10683	29	52	lstms	lstms	ADJ
dad-10683	29	53	)	)	PUNCT
dad-10683	29	54	.	.	PUNCT
dad-10683	30	1	next	next	ADV
dad-10683	30	2	,	,	PUNCT
dad-10683	30	3	we	we	PRON
dad-10683	30	4	present	present	VERB
dad-10683	30	5	encoder	encoder	NOUN
dad-10683	30	6	-	-	PUNCT
dad-10683	30	7	decoder	decoder	NOUN
dad-10683	30	8	models	model	NOUN
dad-10683	30	9	that	that	PRON
dad-10683	30	10	are	be	AUX
dad-10683	30	11	trained	train	VERB
dad-10683	30	12	to	to	PART
dad-10683	30	13	generate	generate	VERB
dad-10683	30	14	an	an	DET
dad-10683	30	15	utterance	utterance	NOUN
dad-10683	30	16	given	give	VERB
dad-10683	30	17	the	the	DET
dad-10683	30	18	context	context	NOUN
dad-10683	30	19	.	.	PUNCT
dad-10683	31	1	we	we	PRON
dad-10683	31	2	consider	consider	VERB
dad-10683	31	3	both	both	CCONJ
dad-10683	31	4	the	the	DET
dad-10683	31	5	traditional	traditional	ADJ
dad-10683	31	6	lstm	lstm	ADJ
dad-10683	31	7	language	language	NOUN
dad-10683	31	8	model	model	NOUN
dad-10683	31	9	,	,	PUNCT
dad-10683	31	10	which	which	PRON
dad-10683	31	11	corresponds	correspond	VERB
dad-10683	31	12	to	to	ADP
dad-10683	31	13	the	the	DET
dad-10683	31	14	encoder	encoder	NOUN
dad-10683	31	15	and	and	CCONJ
dad-10683	31	16	decoder	decoder	NOUN
dad-10683	31	17	having	having	AUX
dad-10683	31	18	tied	tie	VERB
dad-10683	31	19	weights	weight	NOUN
dad-10683	31	20	,	,	PUNCT
dad-10683	31	21	and	and	CCONJ
dad-10683	31	22	the	the	DET
dad-10683	31	23	recently	recently	ADV
dad-10683	31	24	proposed	propose	VERB
dad-10683	31	25	hierarchical	hierarchical	ADJ
dad-10683	31	26	recurrent	recurrent	ADJ
dad-10683	31	27	encoder	encoder	NOUN
dad-10683	31	28	-	-	PUNCT
dad-10683	31	29	decoder	decoder	NOUN
dad-10683	31	30	(	(	PUNCT
dad-10683	31	31	hred	hre	VERB
dad-10683	31	32	)	)	PUNCT
dad-10683	31	33	(	(	PUNCT
dad-10683	31	34	serban	serban	PROPN
dad-10683	31	35	et	et	PROPN
dad-10683	31	36	al	al	PROPN
dad-10683	31	37	.	.	PROPN
dad-10683	31	38	,	,	PUNCT
dad-10683	31	39	2016	2016	NUM
dad-10683	31	40	)	)	PUNCT
dad-10683	31	41	,	,	PUNCT
dad-10683	31	42	which	which	PRON
dad-10683	31	43	has	have	VERB
dad-10683	31	44	a	a	DET
dad-10683	31	45	second	second	ADJ
dad-10683	31	46	recurrent	recurrent	ADJ
dad-10683	31	47	network	network	NOUN
dad-10683	31	48	that	that	PRON
dad-10683	31	49	encodes	encode	VERB
dad-10683	31	50	utterance	utterance	ADV
dad-10683	31	51	-	-	PUNCT
dad-10683	31	52	specific	specific	ADJ
dad-10683	31	53	information	information	NOUN
dad-10683	31	54	,	,	PUNCT
dad-10683	31	55	and	and	CCONJ
dad-10683	31	56	is	be	AUX
dad-10683	31	57	thus	thus	ADV
dad-10683	31	58	able	able	ADJ
dad-10683	31	59	to	to	PART
dad-10683	31	60	model	model	VERB
dad-10683	31	61	longer	long	ADJ
dad-10683	31	62	-	-	PUNCT
dad-10683	31	63	term	term	NOUN
dad-10683	31	64	dependencies	dependency	NOUN
dad-10683	31	65	in	in	ADP
dad-10683	31	66	the	the	DET
dad-10683	31	67	context	context	NOUN
dad-10683	31	68	.	.	PUNCT
dad-10683	32	1	we	we	PRON
dad-10683	32	2	evaluate	evaluate	VERB
dad-10683	32	3	these	these	DET
dad-10683	32	4	models	model	NOUN
dad-10683	32	5	on	on	ADP
dad-10683	32	6	the	the	DET
dad-10683	32	7	task	task	NOUN
dad-10683	32	8	of	of	ADP
dad-10683	32	9	next	next	ADJ
dad-10683	32	10	utterance	utterance	NOUN
dad-10683	32	11	classification	classification	NOUN
dad-10683	32	12	(	(	PUNCT
dad-10683	32	13	nuc	nuc	NOUN
dad-10683	32	14	)	)	PUNCT
dad-10683	32	15	,	,	PUNCT
dad-10683	32	16	where	where	SCONJ
dad-10683	32	17	the	the	DET
dad-10683	32	18	model	model	NOUN
dad-10683	32	19	ranks	rank	VERB
dad-10683	32	20	a	a	DET
dad-10683	32	21	list	list	NOUN
dad-10683	32	22	of	of	ADP
dad-10683	32	23	candidate	candidate	NOUN
dad-10683	32	24	responses	response	NOUN
dad-10683	32	25	by	by	ADP
dad-10683	32	26	how	how	SCONJ
dad-10683	32	27	likely	likely	ADJ
dad-10683	32	28	they	they	PRON
dad-10683	32	29	are	be	AUX
dad-10683	32	30	to	to	PART
dad-10683	32	31	have	have	AUX
dad-10683	32	32	followed	follow	VERB
dad-10683	32	33	the	the	DET
dad-10683	32	34	context	context	NOUN
dad-10683	32	35	.	.	PUNCT
dad-10683	33	1	we	we	PRON
dad-10683	33	2	also	also	ADV
dad-10683	33	3	evaluate	evaluate	VERB
dad-10683	33	4	using	use	VERB
dad-10683	33	5	vector	vector	NOUN
dad-10683	33	6	-	-	PUNCT
dad-10683	33	7	based	base	VERB
dad-10683	33	8	metrics	metric	NOUN
dad-10683	33	9	to	to	PART
dad-10683	33	10	determine	determine	VERB
dad-10683	33	11	the	the	DET
dad-10683	33	12	quality	quality	NOUN
dad-10683	33	13	of	of	ADP
dad-10683	33	14	generated	generate	VERB
dad-10683	33	15	responses	response	NOUN
dad-10683	33	16	,	,	PUNCT
dad-10683	33	17	in	in	ADP
dad-10683	33	18	terms	term	NOUN
dad-10683	33	19	of	of	ADP
dad-10683	33	20	semantic	semantic	ADJ
dad-10683	33	21	similarity	similarity	NOUN
dad-10683	33	22	to	to	ADP
dad-10683	33	23	the	the	DET
dad-10683	33	24	ground	ground	NOUN
dad-10683	33	25	-	-	PUNCT
dad-10683	33	26	truth	truth	NOUN
dad-10683	33	27	next	next	ADP
dad-10683	33	28	utterance	utterance	ADV
dad-10683	33	29	.	.	PUNCT
dad-10683	34	1	we	we	PRON
dad-10683	34	2	observe	observe	VERB
dad-10683	34	3	that	that	SCONJ
dad-10683	34	4	the	the	DET
dad-10683	34	5	state	state	NOUN
dad-10683	34	6	-	-	PUNCT
dad-10683	34	7	of	of	ADP
dad-10683	34	8	-	-	PUNCT
dad-10683	34	9	the	the	DET
dad-10683	34	10	-	-	PUNCT
dad-10683	34	11	art	art	NOUN
dad-10683	34	12	models	model	NOUN
dad-10683	34	13	,	,	PUNCT
dad-10683	34	14	the	the	DET
dad-10683	34	15	lstm	lstm	ADJ
dad-10683	34	16	dual	dual	ADJ
dad-10683	34	17	encoder	encoder	NOUN
dad-10683	34	18	and	and	CCONJ
dad-10683	34	19	hred	hre	VERB
dad-10683	34	20	,	,	PUNCT
dad-10683	34	21	outperform	outperform	VERB
dad-10683	34	22	the	the	DET
dad-10683	34	23	baselines	baseline	NOUN
dad-10683	34	24	on	on	ADP
dad-10683	34	25	all	all	DET
dad-10683	34	26	metrics	metric	NOUN
dad-10683	34	27	.	.	PUNCT
dad-10683	35	1	32	32	NUM
dad-10683	35	2	training	training	NOUN
dad-10683	35	3	end	end	NOUN
dad-10683	35	4	-	-	PUNCT
dad-10683	35	5	to	to	ADP
dad-10683	35	6	-	-	PUNCT
dad-10683	35	7	end	end	NOUN
dad-10683	35	8	dialogue	dialogue	NOUN
dad-10683	35	9	systems	system	NOUN
dad-10683	35	10	finally	finally	ADV
dad-10683	35	11	,	,	PUNCT
dad-10683	35	12	we	we	PRON
dad-10683	35	13	conduct	conduct	VERB
dad-10683	35	14	a	a	DET
dad-10683	35	15	qualitative	qualitative	ADJ
dad-10683	35	16	analysis	analysis	NOUN
dad-10683	35	17	to	to	PART
dad-10683	35	18	determine	determine	VERB
dad-10683	35	19	the	the	DET
dad-10683	35	20	main	main	ADJ
dad-10683	35	21	sources	source	NOUN
dad-10683	35	22	of	of	ADP
dad-10683	35	23	error	error	NOUN
dad-10683	35	24	for	for	ADP
dad-10683	35	25	the	the	DET
dad-10683	35	26	de	de	PROPN
dad-10683	35	27	model	model	NOUN
dad-10683	35	28	.	.	PUNCT
dad-10683	36	1	we	we	PRON
dad-10683	36	2	find	find	VERB
dad-10683	36	3	that	that	SCONJ
dad-10683	36	4	the	the	DET
dad-10683	36	5	most	most	ADV
dad-10683	36	6	common	common	ADJ
dad-10683	36	7	errors	error	NOUN
dad-10683	36	8	are	be	AUX
dad-10683	36	9	a	a	DET
dad-10683	36	10	lack	lack	NOUN
dad-10683	36	11	of	of	ADP
dad-10683	36	12	understanding	understanding	NOUN
dad-10683	36	13	of	of	ADP
dad-10683	36	14	the	the	DET
dad-10683	36	15	semantics	semantic	NOUN
dad-10683	36	16	of	of	ADP
dad-10683	36	17	the	the	DET
dad-10683	36	18	responses	response	NOUN
dad-10683	36	19	,	,	PUNCT
dad-10683	36	20	which	which	PRON
dad-10683	36	21	includes	include	VERB
dad-10683	36	22	missing	miss	VERB
dad-10683	36	23	key	key	ADJ
dad-10683	36	24	words	word	NOUN
dad-10683	36	25	that	that	PRON
dad-10683	36	26	are	be	AUX
dad-10683	36	27	copied	copy	VERB
dad-10683	36	28	between	between	ADP
dad-10683	36	29	the	the	DET
dad-10683	36	30	context	context	NOUN
dad-10683	36	31	and	and	CCONJ
dad-10683	36	32	target	target	NOUN
dad-10683	36	33	response	response	NOUN
dad-10683	36	34	,	,	PUNCT
dad-10683	36	35	and	and	CCONJ
dad-10683	36	36	a	a	DET
dad-10683	36	37	lack	lack	NOUN
dad-10683	36	38	of	of	ADP
dad-10683	36	39	higher	high	ADJ
dad-10683	36	40	-	-	PUNCT
dad-10683	36	41	level	level	NOUN
dad-10683	36	42	inference	inference	NOUN
dad-10683	36	43	.	.	PUNCT
dad-10683	37	1	there	there	PRON
dad-10683	37	2	are	be	VERB
dad-10683	37	3	also	also	ADV
dad-10683	37	4	a	a	DET
dad-10683	37	5	number	number	NOUN
dad-10683	37	6	of	of	ADP
dad-10683	37	7	cases	case	NOUN
dad-10683	37	8	where	where	SCONJ
dad-10683	37	9	the	the	DET
dad-10683	37	10	model	model	NOUN
dad-10683	37	11	would	would	AUX
dad-10683	37	12	benefit	benefit	VERB
dad-10683	37	13	from	from	ADP
dad-10683	37	14	explicitly	explicitly	ADV
dad-10683	37	15	incorporating	incorporate	VERB
dad-10683	37	16	the	the	DET
dad-10683	37	17	turn	turn	NOUN
dad-10683	37	18	-	-	PUNCT
dad-10683	37	19	taking	take	VERB
dad-10683	37	20	structure	structure	NOUN
dad-10683	37	21	of	of	ADP
dad-10683	37	22	dialogue	dialogue	NOUN
dad-10683	37	23	,	,	PUNCT
dad-10683	37	24	and	and	CCONJ
dad-10683	37	25	using	use	VERB
dad-10683	37	26	some	some	DET
dad-10683	37	27	source	source	NOUN
dad-10683	37	28	of	of	ADP
dad-10683	37	29	external	external	ADJ
dad-10683	37	30	knowledge	knowledge	NOUN
dad-10683	37	31	for	for	ADP
dad-10683	37	32	ubuntu	ubuntu	ADJ
dad-10683	37	33	terminology	terminology	NOUN
dad-10683	37	34	.	.	PUNCT
dad-10683	38	1	an	an	DET
dad-10683	38	2	examination	examination	NOUN
dad-10683	38	3	of	of	ADP
dad-10683	38	4	the	the	DET
dad-10683	38	5	responses	response	NOUN
dad-10683	38	6	produced	produce	VERB
dad-10683	38	7	by	by	ADP
dad-10683	38	8	the	the	DET
dad-10683	38	9	generative	generative	ADJ
dad-10683	38	10	models	model	NOUN
dad-10683	38	11	reveals	reveal	VERB
dad-10683	38	12	similar	similar	ADJ
dad-10683	38	13	shortcomings	shortcoming	NOUN
dad-10683	38	14	;	;	PUNCT
dad-10683	38	15	while	while	SCONJ
dad-10683	38	16	the	the	DET
dad-10683	38	17	models	model	NOUN
dad-10683	38	18	are	be	AUX
dad-10683	38	19	able	able	ADJ
dad-10683	38	20	to	to	PART
dad-10683	38	21	generate	generate	VERB
dad-10683	38	22	reasonable	reasonable	ADJ
dad-10683	38	23	responses	response	NOUN
dad-10683	38	24	,	,	PUNCT
dad-10683	38	25	they	they	PRON
dad-10683	38	26	are	be	AUX
dad-10683	38	27	often	often	ADV
dad-10683	38	28	generic	generic	ADJ
dad-10683	38	29	or	or	CCONJ
dad-10683	38	30	lack	lack	VERB
dad-10683	38	31	a	a	DET
dad-10683	38	32	semantic	semantic	ADJ
dad-10683	38	33	understanding	understanding	NOUN
dad-10683	38	34	of	of	ADP
dad-10683	38	35	the	the	DET
dad-10683	38	36	context	context	NOUN
dad-10683	38	37	.	.	PUNCT
dad-10683	39	1	it	it	PRON
dad-10683	39	2	is	be	AUX
dad-10683	39	3	clear	clear	ADJ
dad-10683	39	4	that	that	SCONJ
dad-10683	39	5	end	end	NOUN
dad-10683	39	6	-	-	PUNCT
dad-10683	39	7	to	to	ADP
dad-10683	39	8	-	-	PUNCT
dad-10683	39	9	end	end	NOUN
dad-10683	39	10	systems	system	NOUN
dad-10683	39	11	are	be	AUX
dad-10683	39	12	not	not	PART
dad-10683	39	13	close	close	ADJ
dad-10683	39	14	to	to	ADP
dad-10683	39	15	solving	solve	VERB
dad-10683	39	16	a	a	DET
dad-10683	39	17	domain	domain	NOUN
dad-10683	39	18	as	as	ADV
dad-10683	39	19	complex	complex	ADJ
dad-10683	39	20	as	as	ADP
dad-10683	39	21	ubuntu	ubuntu	ADJ
dad-10683	39	22	.	.	PUNCT
dad-10683	40	1	we	we	PRON
dad-10683	40	2	hope	hope	VERB
dad-10683	40	3	that	that	SCONJ
dad-10683	40	4	this	this	DET
dad-10683	40	5	analysis	analysis	NOUN
dad-10683	40	6	can	can	AUX
dad-10683	40	7	help	help	AUX
dad-10683	40	8	guide	guide	VERB
dad-10683	40	9	future	future	ADJ
dad-10683	40	10	research	research	NOUN
dad-10683	40	11	on	on	ADP
dad-10683	40	12	the	the	DET
dad-10683	40	13	ubuntu	ubuntu	ADJ
dad-10683	40	14	dialogue	dialogue	NOUN
dad-10683	40	15	corpus	corpus	NOUN
dad-10683	40	16	,	,	PUNCT
dad-10683	40	17	and	and	CCONJ
dad-10683	40	18	the	the	DET
dad-10683	40	19	development	development	NOUN
dad-10683	40	20	of	of	ADP
dad-10683	40	21	end	end	NOUN
dad-10683	40	22	-	-	PUNCT
dad-10683	40	23	to	to	ADP
dad-10683	40	24	-	-	PUNCT
dad-10683	40	25	end	end	NOUN
dad-10683	40	26	dialogue	dialogue	NOUN
dad-10683	40	27	systems	system	NOUN
dad-10683	40	28	.	.	PUNCT
dad-10683	41	1	1.1	1.1	NUM
dad-10683	41	2	motivation	motivation	NOUN
dad-10683	41	3	for	for	ADP
dad-10683	41	4	end	end	NOUN
dad-10683	41	5	-	-	PUNCT
dad-10683	41	6	to	to	ADP
dad-10683	41	7	-	-	PUNCT
dad-10683	41	8	end	end	NOUN
dad-10683	41	9	dialogue	dialogue	NOUN
dad-10683	41	10	systems	system	NOUN
dad-10683	41	11	it	it	PRON
dad-10683	41	12	is	be	AUX
dad-10683	41	13	important	important	ADJ
dad-10683	41	14	to	to	PART
dad-10683	41	15	define	define	VERB
dad-10683	41	16	specifically	specifically	ADV
dad-10683	41	17	what	what	PRON
dad-10683	41	18	is	be	AUX
dad-10683	41	19	meant	mean	VERB
dad-10683	41	20	by	by	ADP
dad-10683	41	21	an	an	DET
dad-10683	41	22	‘	'	PUNCT
dad-10683	41	23	end	end	NOUN
dad-10683	41	24	-	-	PUNCT
dad-10683	41	25	to	to	ADP
dad-10683	41	26	-	-	PUNCT
dad-10683	41	27	end	end	VERB
dad-10683	41	28	’	'	PUNCT
dad-10683	41	29	dialogue	dialogue	NOUN
dad-10683	41	30	system	system	NOUN
dad-10683	41	31	.	.	PUNCT
dad-10683	42	1	we	we	PRON
dad-10683	42	2	begin	begin	VERB
dad-10683	42	3	with	with	ADP
dad-10683	42	4	the	the	DET
dad-10683	42	5	standard	standard	ADJ
dad-10683	42	6	architecture	architecture	NOUN
dad-10683	42	7	for	for	ADP
dad-10683	42	8	a	a	DET
dad-10683	42	9	dialogue	dialogue	NOUN
dad-10683	42	10	system	system	NOUN
dad-10683	42	11	,	,	PUNCT
dad-10683	42	12	which	which	PRON
dad-10683	42	13	incorporates	incorporate	VERB
dad-10683	42	14	a	a	DET
dad-10683	42	15	speech	speech	NOUN
dad-10683	42	16	recognizer	recognizer	NOUN
dad-10683	42	17	,	,	PUNCT
dad-10683	42	18	language	language	NOUN
dad-10683	42	19	interpreter	interpreter	NOUN
dad-10683	42	20	,	,	PUNCT
dad-10683	42	21	state	state	NOUN
dad-10683	42	22	tracker	tracker	NOUN
dad-10683	42	23	,	,	PUNCT
dad-10683	42	24	response	response	NOUN
dad-10683	42	25	generator	generator	PROPN
dad-10683	42	26	,	,	PUNCT
dad-10683	42	27	natural	natural	ADJ
dad-10683	42	28	language	language	NOUN
dad-10683	42	29	generator	generator	NOUN
dad-10683	42	30	,	,	PUNCT
dad-10683	42	31	and	and	CCONJ
dad-10683	42	32	speech	speech	NOUN
dad-10683	42	33	synthesizer	synthesizer	NOUN
dad-10683	42	34	.	.	PUNCT
dad-10683	43	1	in	in	ADP
dad-10683	43	2	the	the	DET
dad-10683	43	3	case	case	NOUN
dad-10683	43	4	of	of	ADP
dad-10683	43	5	text	text	NOUN
dad-10683	43	6	-	-	PUNCT
dad-10683	43	7	based	base	VERB
dad-10683	43	8	(	(	PUNCT
dad-10683	43	9	written	write	VERB
dad-10683	43	10	)	)	PUNCT
dad-10683	43	11	dialogues	dialogue	NOUN
dad-10683	43	12	,	,	PUNCT
dad-10683	43	13	the	the	DET
dad-10683	43	14	speech	speech	NOUN
dad-10683	43	15	recognizer	recognizer	NOUN
dad-10683	43	16	and	and	CCONJ
dad-10683	43	17	speech	speech	NOUN
dad-10683	43	18	synthesizer	synthesizer	NOUN
dad-10683	43	19	can	can	AUX
dad-10683	43	20	be	be	AUX
dad-10683	43	21	left	leave	VERB
dad-10683	43	22	out	out	ADP
dad-10683	43	23	.	.	PUNCT
dad-10683	44	1	although	although	SCONJ
dad-10683	44	2	some	some	DET
dad-10683	44	3	previous	previous	ADJ
dad-10683	44	4	literature	literature	NOUN
dad-10683	44	5	on	on	ADP
dad-10683	44	6	dialogue	dialogue	NOUN
dad-10683	44	7	systems	system	NOUN
dad-10683	44	8	identifies	identify	VERB
dad-10683	44	9	only	only	ADV
dad-10683	44	10	the	the	DET
dad-10683	44	11	state	state	NOUN
dad-10683	44	12	tracker	tracker	NOUN
dad-10683	44	13	and	and	CCONJ
dad-10683	44	14	response	response	NOUN
dad-10683	44	15	selection	selection	NOUN
dad-10683	44	16	components	component	NOUN
dad-10683	44	17	as	as	ADP
dad-10683	44	18	belonging	belong	VERB
dad-10683	44	19	inside	inside	ADP
dad-10683	44	20	the	the	DET
dad-10683	44	21	dialogue	dialogue	NOUN
dad-10683	44	22	manager	manager	NOUN
dad-10683	44	23	(	(	PUNCT
dad-10683	44	24	young	young	ADJ
dad-10683	44	25	,	,	PUNCT
dad-10683	44	26	2000	2000	NUM
dad-10683	44	27	)	)	PUNCT
dad-10683	44	28	,	,	PUNCT
dad-10683	44	29	we	we	PRON
dad-10683	44	30	adopt	adopt	VERB
dad-10683	44	31	a	a	DET
dad-10683	44	32	broader	broad	ADJ
dad-10683	44	33	view	view	NOUN
dad-10683	44	34	where	where	SCONJ
dad-10683	44	35	the	the	DET
dad-10683	44	36	language	language	NOUN
dad-10683	44	37	interpreter	interpreter	NOUN
dad-10683	44	38	and	and	CCONJ
dad-10683	44	39	natural	natural	ADJ
dad-10683	44	40	language	language	NOUN
dad-10683	44	41	generator	generator	NOUN
dad-10683	44	42	are	be	AUX
dad-10683	44	43	also	also	ADV
dad-10683	44	44	part	part	NOUN
dad-10683	44	45	of	of	ADP
dad-10683	44	46	the	the	DET
dad-10683	44	47	dialogue	dialogue	NOUN
dad-10683	44	48	manager	manager	NOUN
dad-10683	44	49	.	.	PUNCT
dad-10683	45	1	figure	figure	NOUN
dad-10683	45	2	1	1	NUM
dad-10683	45	3	:	:	PUNCT
dad-10683	45	4	an	an	DET
dad-10683	45	5	end	end	NOUN
dad-10683	45	6	-	-	PUNCT
dad-10683	45	7	to	to	ADP
dad-10683	45	8	-	-	PUNCT
dad-10683	45	9	end	end	NOUN
dad-10683	45	10	dialogue	dialogue	NOUN
dad-10683	45	11	system	system	NOUN
dad-10683	45	12	replaces	replace	VERB
dad-10683	45	13	the	the	DET
dad-10683	45	14	traditional	traditional	ADJ
dad-10683	45	15	components	component	NOUN
dad-10683	45	16	of	of	ADP
dad-10683	45	17	a	a	DET
dad-10683	45	18	dialogue	dialogue	NOUN
dad-10683	45	19	system	system	NOUN
dad-10683	45	20	with	with	ADP
dad-10683	45	21	a	a	DET
dad-10683	45	22	single	single	ADJ
dad-10683	45	23	statistical	statistical	ADJ
dad-10683	45	24	model	model	NOUN
dad-10683	45	25	.	.	PUNCT
dad-10683	46	1	when	when	SCONJ
dad-10683	46	2	we	we	PRON
dad-10683	46	3	speak	speak	VERB
dad-10683	46	4	of	of	ADP
dad-10683	46	5	an	an	DET
dad-10683	46	6	‘	'	PUNCT
dad-10683	46	7	end	end	NOUN
dad-10683	46	8	-	-	PUNCT
dad-10683	46	9	to	to	ADP
dad-10683	46	10	-	-	PUNCT
dad-10683	46	11	end	end	VERB
dad-10683	46	12	’	'	PUNCT
dad-10683	46	13	dialogue	dialogue	NOUN
dad-10683	46	14	system	system	NOUN
dad-10683	46	15	,	,	PUNCT
dad-10683	46	16	we	we	PRON
dad-10683	46	17	mean	mean	VERB
dad-10683	46	18	a	a	DET
dad-10683	46	19	single	single	ADJ
dad-10683	46	20	system	system	NOUN
dad-10683	46	21	that	that	PRON
dad-10683	46	22	can	can	AUX
dad-10683	46	23	be	be	AUX
dad-10683	46	24	used	use	VERB
dad-10683	46	25	to	to	PART
dad-10683	46	26	solve	solve	VERB
dad-10683	46	27	each	each	PRON
dad-10683	46	28	of	of	ADP
dad-10683	46	29	these	these	DET
dad-10683	46	30	four	four	NUM
dad-10683	46	31	aspects	aspect	NOUN
dad-10683	46	32	simultaneously	simultaneously	ADV
dad-10683	46	33	(	(	PUNCT
dad-10683	46	34	see	see	VERB
dad-10683	46	35	figure	figure	NOUN
dad-10683	46	36	1	1	NUM
dad-10683	46	37	)	)	PUNCT
dad-10683	46	38	.	.	PUNCT
dad-10683	47	1	typically	typically	ADV
dad-10683	47	2	this	this	PRON
dad-10683	47	3	is	be	AUX
dad-10683	47	4	a	a	DET
dad-10683	47	5	system	system	NOUN
dad-10683	47	6	that	that	PRON
dad-10683	47	7	takes	take	VERB
dad-10683	47	8	as	as	ADP
dad-10683	47	9	input	input	NOUN
dad-10683	47	10	the	the	DET
dad-10683	47	11	history	history	NOUN
dad-10683	47	12	of	of	ADP
dad-10683	47	13	the	the	DET
dad-10683	47	14	conversation	conversation	NOUN
dad-10683	47	15	and	and	CCONJ
dad-10683	47	16	is	be	AUX
dad-10683	47	17	trained	train	VERB
dad-10683	47	18	to	to	PART
dad-10683	47	19	optimize	optimize	VERB
dad-10683	47	20	a	a	DET
dad-10683	47	21	single	single	ADJ
dad-10683	47	22	objective	objective	NOUN
dad-10683	47	23	,	,	PUNCT
dad-10683	47	24	which	which	PRON
dad-10683	47	25	is	be	AUX
dad-10683	47	26	a	a	DET
dad-10683	47	27	function	function	NOUN
dad-10683	47	28	of	of	ADP
dad-10683	47	29	the	the	DET
dad-10683	47	30	textual	textual	ADJ
dad-10683	47	31	output	output	NOUN
dad-10683	47	32	produced	produce	VERB
dad-10683	47	33	by	by	ADP
dad-10683	47	34	the	the	DET
dad-10683	47	35	system	system	NOUN
dad-10683	47	36	and	and	CCONJ
dad-10683	47	37	the	the	DET
dad-10683	47	38	correct	correct	ADJ
dad-10683	47	39	(	(	PUNCT
dad-10683	47	40	ground	ground	NOUN
dad-10683	47	41	truth	truth	NOUN
dad-10683	47	42	)	)	PUNCT
dad-10683	47	43	response	response	NOUN
dad-10683	47	44	.	.	PUNCT
dad-10683	48	1	this	this	PRON
dad-10683	48	2	is	be	AUX
dad-10683	48	3	in	in	ADP
dad-10683	48	4	contrast	contrast	NOUN
dad-10683	48	5	to	to	ADP
dad-10683	48	6	the	the	DET
dad-10683	48	7	‘	'	PUNCT
dad-10683	48	8	modular	modular	ADJ
dad-10683	48	9	’	'	PUNCT
dad-10683	48	10	system	system	NOUN
dad-10683	48	11	approach	approach	NOUN
dad-10683	48	12	to	to	ADP
dad-10683	48	13	dialogue	dialogue	NOUN
dad-10683	48	14	systems	system	NOUN
dad-10683	48	15	,	,	PUNCT
dad-10683	48	16	where	where	SCONJ
dad-10683	48	17	each	each	DET
dad-10683	48	18	component	component	NOUN
dad-10683	48	19	of	of	ADP
dad-10683	48	20	figure	figure	NOUN
dad-10683	48	21	1	1	NUM
dad-10683	48	22	is	be	AUX
dad-10683	48	23	trained	train	VERB
dad-10683	48	24	separately	separately	ADV
dad-10683	48	25	,	,	PUNCT
dad-10683	48	26	and	and	CCONJ
dad-10683	48	27	either	either	CCONJ
dad-10683	48	28	takes	take	VERB
dad-10683	48	29	a	a	DET
dad-10683	48	30	more	more	ADV
dad-10683	48	31	structured	structured	ADJ
dad-10683	48	32	input	input	NOUN
dad-10683	48	33	,	,	PUNCT
dad-10683	48	34	such	such	ADJ
dad-10683	48	35	as	as	ADP
dad-10683	48	36	a	a	DET
dad-10683	48	37	set	set	NOUN
dad-10683	48	38	of	of	ADP
dad-10683	48	39	dialogue	dialogue	NOUN
dad-10683	48	40	acts	act	NOUN
dad-10683	48	41	,	,	PUNCT
dad-10683	48	42	or	or	CCONJ
dad-10683	48	43	is	be	AUX
dad-10683	48	44	trained	train	VERB
dad-10683	48	45	to	to	PART
dad-10683	48	46	maximize	maximize	VERB
dad-10683	48	47	an	an	DET
dad-10683	48	48	intermediary	intermediary	ADJ
dad-10683	48	49	objective	objective	NOUN
dad-10683	48	50	,	,	PUNCT
dad-10683	48	51	such	such	ADJ
dad-10683	48	52	as	as	ADP
dad-10683	48	53	slot	slot	NOUN
dad-10683	48	54	-	-	PUNCT
dad-10683	48	55	filling	filling	NOUN
dad-10683	48	56	.	.	PUNCT
dad-10683	49	1	more	more	ADV
dad-10683	49	2	formally	formally	ADV
dad-10683	49	3	,	,	PUNCT
dad-10683	49	4	we	we	PRON
dad-10683	49	5	define	define	VERB
dad-10683	49	6	33	33	NUM
dad-10683	49	7	lowe	lowe	NOUN
dad-10683	49	8	,	,	PUNCT
dad-10683	49	9	pow	pow	PROPN
dad-10683	49	10	,	,	PUNCT
dad-10683	49	11	serban	serban	PROPN
dad-10683	49	12	,	,	PUNCT
dad-10683	49	13	charlinn	charlinn	PROPN
dad-10683	49	14	,	,	PUNCT
dad-10683	49	15	liu	liu	PROPN
dad-10683	49	16	and	and	CCONJ
dad-10683	49	17	pineau	pineau	PROPN
dad-10683	49	18	a	a	DET
dad-10683	49	19	modular	modular	ADJ
dad-10683	49	20	dialogue	dialogue	NOUN
dad-10683	49	21	system	system	NOUN
dad-10683	49	22	as	as	ADP
dad-10683	49	23	a	a	DET
dad-10683	49	24	system	system	NOUN
dad-10683	49	25	where	where	SCONJ
dad-10683	49	26	two	two	NUM
dad-10683	49	27	or	or	CCONJ
dad-10683	49	28	more	more	ADJ
dad-10683	49	29	elements	element	NOUN
dad-10683	49	30	(	(	PUNCT
dad-10683	49	31	sub	sub	NOUN
dad-10683	49	32	-	-	NOUN
dad-10683	49	33	components	component	NOUN
dad-10683	49	34	or	or	CCONJ
dad-10683	49	35	system	system	NOUN
dad-10683	49	36	parameters	parameter	NOUN
dad-10683	49	37	)	)	PUNCT
dad-10683	49	38	are	be	AUX
dad-10683	49	39	optimized	optimize	VERB
dad-10683	49	40	with	with	ADP
dad-10683	49	41	respect	respect	NOUN
dad-10683	49	42	to	to	ADP
dad-10683	49	43	two	two	NUM
dad-10683	49	44	or	or	CCONJ
dad-10683	49	45	more	more	ADV
dad-10683	49	46	different	different	ADJ
dad-10683	49	47	objective	objective	ADJ
dad-10683	49	48	functions	function	NOUN
dad-10683	49	49	(	(	PUNCT
dad-10683	49	50	e.g.	e.g.	ADV
dad-10683	49	51	where	where	SCONJ
dad-10683	49	52	the	the	DET
dad-10683	49	53	state	state	NOUN
dad-10683	49	54	tracker	tracker	NOUN
dad-10683	49	55	is	be	AUX
dad-10683	49	56	trained	train	VERB
dad-10683	49	57	to	to	PART
dad-10683	49	58	minimize	minimize	VERB
dad-10683	49	59	the	the	DET
dad-10683	49	60	cross	cross	ADJ
dad-10683	49	61	-	-	ADJ
dad-10683	49	62	entropy	entropy	ADJ
dad-10683	49	63	error	error	NOUN
dad-10683	49	64	of	of	ADP
dad-10683	49	65	predicting	predict	VERB
dad-10683	49	66	the	the	DET
dad-10683	49	67	slot	slot	NOUN
dad-10683	49	68	-	-	PUNCT
dad-10683	49	69	value	value	NOUN
dad-10683	49	70	pairs	pair	NOUN
dad-10683	49	71	,	,	PUNCT
dad-10683	49	72	and	and	CCONJ
dad-10683	49	73	where	where	SCONJ
dad-10683	49	74	the	the	DET
dad-10683	49	75	response	response	NOUN
dad-10683	49	76	generator	generator	NOUN
dad-10683	49	77	is	be	AUX
dad-10683	49	78	trained	train	VERB
dad-10683	49	79	to	to	PART
dad-10683	49	80	maximize	maximize	VERB
dad-10683	49	81	the	the	DET
dad-10683	49	82	conditional	conditional	ADJ
dad-10683	49	83	log	log	NOUN
dad-10683	49	84	-	-	PUNCT
dad-10683	49	85	likelihood	likelihood	NOUN
dad-10683	49	86	of	of	ADP
dad-10683	49	87	the	the	DET
dad-10683	49	88	correct	correct	ADJ
dad-10683	49	89	response	response	NOUN
dad-10683	49	90	given	give	VERB
dad-10683	49	91	the	the	DET
dad-10683	49	92	slot	slot	NOUN
dad-10683	49	93	-	-	PUNCT
dad-10683	49	94	value	value	NOUN
dad-10683	49	95	pairs	pair	NOUN
dad-10683	49	96	)	)	PUNCT
dad-10683	49	97	.	.	PUNCT
dad-10683	50	1	thus	thus	ADV
dad-10683	50	2	,	,	PUNCT
dad-10683	50	3	any	any	DET
dad-10683	50	4	machine	machine	NOUN
dad-10683	50	5	learning	learning	NOUN
dad-10683	50	6	-	-	PUNCT
dad-10683	50	7	based	base	VERB
dad-10683	50	8	dialogue	dialogue	NOUN
dad-10683	50	9	system	system	NOUN
dad-10683	50	10	which	which	PRON
dad-10683	50	11	is	be	AUX
dad-10683	50	12	not	not	PART
dad-10683	50	13	a	a	DET
dad-10683	50	14	modular	modular	ADJ
dad-10683	50	15	dialogue	dialogue	NOUN
dad-10683	50	16	system	system	NOUN
dad-10683	50	17	is	be	AUX
dad-10683	50	18	an	an	DET
dad-10683	50	19	end	end	NOUN
dad-10683	50	20	-	-	PUNCT
dad-10683	50	21	to	to	ADP
dad-10683	50	22	-	-	PUNCT
dad-10683	50	23	end	end	NOUN
dad-10683	50	24	dialogue	dialogue	NOUN
dad-10683	50	25	system	system	NOUN
dad-10683	50	26	.	.	PUNCT
dad-10683	51	1	note	note	VERB
dad-10683	51	2	that	that	SCONJ
dad-10683	51	3	,	,	PUNCT
dad-10683	51	4	according	accord	VERB
dad-10683	51	5	to	to	ADP
dad-10683	51	6	this	this	DET
dad-10683	51	7	definition	definition	NOUN
dad-10683	51	8	,	,	PUNCT
dad-10683	51	9	whether	whether	SCONJ
dad-10683	51	10	a	a	DET
dad-10683	51	11	system	system	NOUN
dad-10683	51	12	is	be	AUX
dad-10683	51	13	end	end	NOUN
dad-10683	51	14	-	-	PUNCT
dad-10683	51	15	to	to	ADP
dad-10683	51	16	-	-	PUNCT
dad-10683	51	17	end	end	NOUN
dad-10683	51	18	is	be	AUX
dad-10683	51	19	independent	independent	ADJ
dad-10683	51	20	of	of	ADP
dad-10683	51	21	how	how	SCONJ
dad-10683	51	22	it	it	PRON
dad-10683	51	23	is	be	AUX
dad-10683	51	24	evaluated	evaluate	VERB
dad-10683	51	25	.	.	PUNCT
dad-10683	52	1	both	both	DET
dad-10683	52	2	retrieval	retrieval	ADJ
dad-10683	52	3	and	and	CCONJ
dad-10683	52	4	generative	generative	ADJ
dad-10683	52	5	models	model	NOUN
dad-10683	52	6	can	can	AUX
dad-10683	52	7	be	be	AUX
dad-10683	52	8	end	end	NOUN
dad-10683	52	9	-	-	PUNCT
dad-10683	52	10	to	to	ADP
dad-10683	52	11	-	-	PUNCT
dad-10683	52	12	end	end	NOUN
dad-10683	52	13	so	so	ADV
dad-10683	52	14	long	long	ADV
dad-10683	52	15	as	as	SCONJ
dad-10683	52	16	they	they	PRON
dad-10683	52	17	are	be	AUX
dad-10683	52	18	trained	train	VERB
dad-10683	52	19	using	use	VERB
dad-10683	52	20	a	a	DET
dad-10683	52	21	single	single	ADJ
dad-10683	52	22	objective	objective	ADJ
dad-10683	52	23	function	function	NOUN
dad-10683	52	24	.	.	PUNCT
dad-10683	53	1	similarly	similarly	ADV
dad-10683	53	2	,	,	PUNCT
dad-10683	53	3	end	end	NOUN
dad-10683	53	4	-	-	PUNCT
dad-10683	53	5	to	to	ADP
dad-10683	53	6	-	-	PUNCT
dad-10683	53	7	end	end	NOUN
dad-10683	53	8	models	model	NOUN
dad-10683	53	9	can	can	AUX
dad-10683	53	10	be	be	AUX
dad-10683	53	11	evaluated	evaluate	VERB
dad-10683	53	12	using	use	VERB
dad-10683	53	13	intermediary	intermediary	ADJ
dad-10683	53	14	tasks	task	NOUN
dad-10683	53	15	such	such	ADJ
dad-10683	53	16	as	as	ADP
dad-10683	53	17	nuc	nuc	PROPN
dad-10683	53	18	,	,	PUNCT
dad-10683	53	19	which	which	PRON
dad-10683	53	20	do	do	AUX
dad-10683	53	21	not	not	PART
dad-10683	53	22	evaluate	evaluate	VERB
dad-10683	53	23	the	the	DET
dad-10683	53	24	ability	ability	NOUN
dad-10683	53	25	of	of	ADP
dad-10683	53	26	the	the	DET
dad-10683	53	27	models	model	NOUN
dad-10683	53	28	to	to	PART
dad-10683	53	29	generate	generate	VERB
dad-10683	53	30	new	new	ADJ
dad-10683	53	31	utterances	utterance	NOUN
dad-10683	53	32	unseen	unseen	ADJ
dad-10683	53	33	in	in	ADP
dad-10683	53	34	the	the	DET
dad-10683	53	35	training	training	NOUN
dad-10683	53	36	set	set	NOUN
dad-10683	53	37	.	.	PUNCT
dad-10683	54	1	of	of	ADP
dad-10683	54	2	course	course	NOUN
dad-10683	54	3	,	,	PUNCT
dad-10683	54	4	in	in	ADP
dad-10683	54	5	order	order	NOUN
dad-10683	54	6	to	to	PART
dad-10683	54	7	evaluate	evaluate	VERB
dad-10683	54	8	the	the	DET
dad-10683	54	9	full	full	ADJ
dad-10683	54	10	capability	capability	NOUN
dad-10683	54	11	of	of	ADP
dad-10683	54	12	the	the	DET
dad-10683	54	13	models	model	NOUN
dad-10683	54	14	it	it	PRON
dad-10683	54	15	is	be	AUX
dad-10683	54	16	best	good	ADJ
dad-10683	54	17	to	to	PART
dad-10683	54	18	evaluate	evaluate	VERB
dad-10683	54	19	their	their	PRON
dad-10683	54	20	outputs	output	NOUN
dad-10683	54	21	in	in	ADP
dad-10683	54	22	a	a	DET
dad-10683	54	23	setting	setting	NOUN
dad-10683	54	24	as	as	ADV
dad-10683	54	25	realistic	realistic	ADJ
dad-10683	54	26	as	as	ADP
dad-10683	54	27	possible	possible	ADJ
dad-10683	54	28	;	;	PUNCT
dad-10683	54	29	however	however	ADV
dad-10683	54	30	,	,	PUNCT
dad-10683	54	31	this	this	PRON
dad-10683	54	32	this	this	PRON
dad-10683	54	33	difficult	difficult	ADJ
dad-10683	54	34	to	to	PART
dad-10683	54	35	do	do	AUX
dad-10683	54	36	automatically	automatically	ADV
dad-10683	54	37	when	when	SCONJ
dad-10683	54	38	there	there	PRON
dad-10683	54	39	is	be	VERB
dad-10683	54	40	no	no	DET
dad-10683	54	41	notion	notion	NOUN
dad-10683	54	42	of	of	ADP
dad-10683	54	43	task	task	NOUN
dad-10683	54	44	completion	completion	NOUN
dad-10683	54	45	(	(	PUNCT
dad-10683	54	46	liu	liu	PROPN
dad-10683	54	47	et	et	PROPN
dad-10683	54	48	al	al	PROPN
dad-10683	54	49	.	.	PROPN
dad-10683	54	50	,	,	PUNCT
dad-10683	54	51	2016	2016	NUM
dad-10683	54	52	)	)	PUNCT
dad-10683	54	53	.	.	PUNCT
dad-10683	55	1	examples	example	NOUN
dad-10683	55	2	of	of	ADP
dad-10683	55	3	end	end	NOUN
dad-10683	55	4	-	-	PUNCT
dad-10683	55	5	to	to	ADP
dad-10683	55	6	-	-	PUNCT
dad-10683	55	7	end	end	NOUN
dad-10683	55	8	dialogue	dialogue	NOUN
dad-10683	55	9	systems	system	NOUN
dad-10683	55	10	in	in	ADP
dad-10683	55	11	the	the	DET
dad-10683	55	12	recent	recent	ADJ
dad-10683	55	13	literature	literature	NOUN
dad-10683	55	14	involve	involve	VERB
dad-10683	55	15	neural	neural	ADJ
dad-10683	55	16	network	network	NOUN
dad-10683	55	17	-	-	PUNCT
dad-10683	55	18	based	base	VERB
dad-10683	55	19	approaches	approach	NOUN
dad-10683	55	20	that	that	PRON
dad-10683	55	21	are	be	AUX
dad-10683	55	22	fully	fully	ADV
dad-10683	55	23	differentiable	differentiable	ADJ
dad-10683	55	24	,	,	PUNCT
dad-10683	55	25	and	and	CCONJ
dad-10683	55	26	are	be	AUX
dad-10683	55	27	usually	usually	ADV
dad-10683	55	28	trained	train	VERB
dad-10683	55	29	to	to	PART
dad-10683	55	30	maximize	maximize	VERB
dad-10683	55	31	the	the	DET
dad-10683	55	32	log	log	NOUN
dad-10683	55	33	-	-	PUNCT
dad-10683	55	34	likelihood	likelihood	NOUN
dad-10683	55	35	of	of	ADP
dad-10683	55	36	the	the	DET
dad-10683	55	37	generated	generate	VERB
dad-10683	55	38	utterance	utterance	NOUN
dad-10683	55	39	conditioned	condition	VERB
dad-10683	55	40	on	on	ADP
dad-10683	55	41	some	some	DET
dad-10683	55	42	conversational	conversational	ADJ
dad-10683	55	43	context	context	NOUN
dad-10683	55	44	(	(	PUNCT
dad-10683	55	45	serban	serban	PROPN
dad-10683	55	46	et	et	PROPN
dad-10683	55	47	al	al	PROPN
dad-10683	55	48	.	.	PROPN
dad-10683	55	49	,	,	PUNCT
dad-10683	55	50	2016	2016	NUM
dad-10683	55	51	;	;	PUNCT
dad-10683	55	52	vinyals	vinyal	NOUN
dad-10683	55	53	and	and	CCONJ
dad-10683	55	54	le	le	NOUN
dad-10683	55	55	,	,	PUNCT
dad-10683	55	56	2015	2015	NUM
dad-10683	55	57	)	)	PUNCT
dad-10683	55	58	.	.	PUNCT
dad-10683	56	1	these	these	DET
dad-10683	56	2	systems	system	NOUN
dad-10683	56	3	learn	learn	VERB
dad-10683	56	4	off	off	ADP
dad-10683	56	5	-	-	PUNCT
dad-10683	56	6	line	line	NOUN
dad-10683	56	7	through	through	ADP
dad-10683	56	8	examples	example	NOUN
dad-10683	56	9	of	of	ADP
dad-10683	56	10	human	human	ADJ
dad-10683	56	11	-	-	PUNCT
dad-10683	56	12	human	human	NOUN
dad-10683	56	13	dialogues	dialogue	NOUN
dad-10683	56	14	,	,	PUNCT
dad-10683	56	15	and	and	CCONJ
dad-10683	56	16	thus	thus	ADV
dad-10683	56	17	learn	learn	VERB
dad-10683	56	18	to	to	PART
dad-10683	56	19	emulate	emulate	VERB
dad-10683	56	20	the	the	DET
dad-10683	56	21	behaviour	behaviour	NOUN
dad-10683	56	22	of	of	ADP
dad-10683	56	23	agents	agent	NOUN
dad-10683	56	24	in	in	ADP
dad-10683	56	25	the	the	DET
dad-10683	56	26	training	training	NOUN
dad-10683	56	27	corpus	corpus	NOUN
dad-10683	56	28	.	.	PUNCT
dad-10683	57	1	however	however	ADV
dad-10683	57	2	,	,	PUNCT
dad-10683	57	3	differentiability	differentiability	NOUN
dad-10683	57	4	and	and	CCONJ
dad-10683	57	5	offline	offline	ADJ
dad-10683	57	6	learning	learning	NOUN
dad-10683	57	7	are	be	AUX
dad-10683	57	8	not	not	PART
dad-10683	57	9	strict	strict	ADJ
dad-10683	57	10	prerequisites	prerequisite	NOUN
dad-10683	57	11	for	for	ADP
dad-10683	57	12	end	end	NOUN
dad-10683	57	13	-	-	PUNCT
dad-10683	57	14	to	to	ADP
dad-10683	57	15	-	-	PUNCT
dad-10683	57	16	end	end	NOUN
dad-10683	57	17	dialogue	dialogue	NOUN
dad-10683	57	18	systems	system	NOUN
dad-10683	57	19	,	,	PUNCT
dad-10683	57	20	and	and	CCONJ
dad-10683	57	21	other	other	ADJ
dad-10683	57	22	methods	method	NOUN
dad-10683	57	23	could	could	AUX
dad-10683	57	24	be	be	AUX
dad-10683	57	25	devised	devise	VERB
dad-10683	57	26	.	.	PUNCT
dad-10683	58	1	modular	modular	ADJ
dad-10683	58	2	dialogue	dialogue	NOUN
dad-10683	58	3	systems	system	NOUN
dad-10683	58	4	have	have	AUX
dad-10683	58	5	been	be	AUX
dad-10683	58	6	historically	historically	ADV
dad-10683	58	7	preferred	prefer	VERB
dad-10683	58	8	over	over	ADP
dad-10683	58	9	end	end	NOUN
dad-10683	58	10	-	-	PUNCT
dad-10683	58	11	to	to	ADP
dad-10683	58	12	-	-	PUNCT
dad-10683	58	13	end	end	NOUN
dad-10683	58	14	systems	system	NOUN
dad-10683	58	15	.	.	PUNCT
dad-10683	59	1	this	this	PRON
dad-10683	59	2	is	be	AUX
dad-10683	59	3	because	because	SCONJ
dad-10683	59	4	such	such	ADJ
dad-10683	59	5	modular	modular	ADJ
dad-10683	59	6	systems	system	NOUN
dad-10683	59	7	are	be	AUX
dad-10683	59	8	easier	easy	ADJ
dad-10683	59	9	to	to	PART
dad-10683	59	10	train	train	VERB
dad-10683	59	11	,	,	PUNCT
dad-10683	59	12	require	require	VERB
dad-10683	59	13	less	less	ADJ
dad-10683	59	14	data	datum	NOUN
dad-10683	59	15	,	,	PUNCT
dad-10683	59	16	and	and	CCONJ
dad-10683	59	17	so	so	ADV
dad-10683	59	18	far	far	ADV
dad-10683	59	19	have	have	AUX
dad-10683	59	20	been	be	AUX
dad-10683	59	21	shown	show	VERB
dad-10683	59	22	to	to	PART
dad-10683	59	23	achieve	achieve	VERB
dad-10683	59	24	better	well	ADJ
dad-10683	59	25	results	result	NOUN
dad-10683	59	26	in	in	ADP
dad-10683	59	27	practice	practice	NOUN
dad-10683	59	28	,	,	PUNCT
dad-10683	59	29	albeit	albeit	SCONJ
dad-10683	59	30	typically	typically	ADV
dad-10683	59	31	for	for	ADP
dad-10683	59	32	highly	highly	ADV
dad-10683	59	33	structured	structured	ADJ
dad-10683	59	34	tasks	task	NOUN
dad-10683	59	35	.	.	PUNCT
dad-10683	60	1	it	it	PRON
dad-10683	60	2	is	be	AUX
dad-10683	60	3	also	also	ADV
dad-10683	60	4	easier	easy	ADJ
dad-10683	60	5	to	to	PART
dad-10683	60	6	manually	manually	ADV
dad-10683	60	7	program	program	VERB
dad-10683	60	8	each	each	DET
dad-10683	60	9	component	component	NOUN
dad-10683	60	10	specifically	specifically	ADV
dad-10683	60	11	to	to	PART
dad-10683	60	12	obey	obey	VERB
dad-10683	60	13	certain	certain	ADJ
dad-10683	60	14	task	task	NOUN
dad-10683	60	15	constraints	constraint	NOUN
dad-10683	60	16	or	or	CCONJ
dad-10683	60	17	to	to	PART
dad-10683	60	18	solve	solve	VERB
dad-10683	60	19	one	one	NUM
dad-10683	60	20	or	or	CCONJ
dad-10683	60	21	more	more	ADV
dad-10683	60	22	isolated	isolated	ADJ
dad-10683	60	23	tasks	task	NOUN
dad-10683	60	24	.	.	PUNCT
dad-10683	61	1	however	however	ADV
dad-10683	61	2	,	,	PUNCT
dad-10683	61	3	there	there	PRON
dad-10683	61	4	are	be	VERB
dad-10683	61	5	significant	significant	ADJ
dad-10683	61	6	advantages	advantage	NOUN
dad-10683	61	7	to	to	ADP
dad-10683	61	8	end	end	VERB
dad-10683	61	9	-	-	PUNCT
dad-10683	61	10	to	to	ADP
dad-10683	61	11	-	-	PUNCT
dad-10683	61	12	end	end	NOUN
dad-10683	61	13	dialogue	dialogue	NOUN
dad-10683	61	14	systems	system	NOUN
dad-10683	61	15	that	that	PRON
dad-10683	61	16	make	make	VERB
dad-10683	61	17	investigating	investigate	VERB
dad-10683	61	18	them	they	PRON
dad-10683	61	19	worthwhile	worthwhile	ADJ
dad-10683	61	20	.	.	PUNCT
dad-10683	62	1	in	in	ADP
dad-10683	62	2	particular	particular	ADJ
dad-10683	62	3	,	,	PUNCT
dad-10683	62	4	modular	modular	ADJ
dad-10683	62	5	dialogue	dialogue	NOUN
dad-10683	62	6	systems	system	NOUN
dad-10683	62	7	are	be	AUX
dad-10683	62	8	restricted	restrict	VERB
dad-10683	62	9	to	to	ADP
dad-10683	62	10	taskspecific	taskspecific	VERB
dad-10683	62	11	domains	domain	NOUN
dad-10683	62	12	,	,	PUNCT
dad-10683	62	13	and	and	CCONJ
dad-10683	62	14	often	often	ADV
dad-10683	62	15	require	require	VERB
dad-10683	62	16	significant	significant	ADJ
dad-10683	62	17	human	human	ADJ
dad-10683	62	18	feature	feature	NOUN
dad-10683	62	19	engineering	engineering	NOUN
dad-10683	62	20	,	,	PUNCT
dad-10683	62	21	including	include	VERB
dad-10683	62	22	pre	pre	VERB
dad-10683	62	23	-	-	ADJ
dad-10683	62	24	defining	define	VERB
dad-10683	62	25	the	the	DET
dad-10683	62	26	state	state	NOUN
dad-10683	62	27	and	and	CCONJ
dad-10683	62	28	action	action	NOUN
dad-10683	62	29	spaces	space	NOUN
dad-10683	62	30	of	of	ADP
dad-10683	62	31	the	the	DET
dad-10683	62	32	model	model	NOUN
dad-10683	62	33	.	.	PUNCT
dad-10683	63	1	although	although	SCONJ
dad-10683	63	2	this	this	PRON
dad-10683	63	3	can	can	AUX
dad-10683	63	4	work	work	VERB
dad-10683	63	5	well	well	ADV
dad-10683	63	6	for	for	ADP
dad-10683	63	7	narrow	narrow	ADJ
dad-10683	63	8	domains	domain	NOUN
dad-10683	63	9	,	,	PUNCT
dad-10683	63	10	it	it	PRON
dad-10683	63	11	does	do	AUX
dad-10683	63	12	not	not	PART
dad-10683	63	13	necessarily	necessarily	ADV
dad-10683	63	14	generalize	generalize	VERB
dad-10683	63	15	to	to	ADP
dad-10683	63	16	general	general	ADJ
dad-10683	63	17	-	-	PUNCT
dad-10683	63	18	purpose	purpose	NOUN
dad-10683	63	19	dialogue	dialogue	NOUN
dad-10683	63	20	.	.	PUNCT
dad-10683	64	1	on	on	ADP
dad-10683	64	2	the	the	DET
dad-10683	64	3	other	other	ADJ
dad-10683	64	4	hand	hand	NOUN
dad-10683	64	5	,	,	PUNCT
dad-10683	64	6	end	end	NOUN
dad-10683	64	7	-	-	PUNCT
dad-10683	64	8	to	to	ADP
dad-10683	64	9	-	-	PUNCT
dad-10683	64	10	end	end	NOUN
dad-10683	64	11	models	model	NOUN
dad-10683	64	12	do	do	AUX
dad-10683	64	13	not	not	PART
dad-10683	64	14	require	require	VERB
dad-10683	64	15	a	a	DET
dad-10683	64	16	pre	pre	ADJ
dad-10683	64	17	-	-	ADJ
dad-10683	64	18	defined	define	VERB
dad-10683	64	19	state	state	NOUN
dad-10683	64	20	or	or	CCONJ
dad-10683	64	21	action	action	NOUN
dad-10683	64	22	space	space	NOUN
dad-10683	64	23	representation	representation	NOUN
dad-10683	64	24	;	;	PUNCT
dad-10683	64	25	instead	instead	ADV
dad-10683	64	26	,	,	PUNCT
dad-10683	64	27	these	these	DET
dad-10683	64	28	representations	representation	NOUN
dad-10683	64	29	are	be	AUX
dad-10683	64	30	learned	learn	VERB
dad-10683	64	31	directly	directly	ADV
dad-10683	64	32	from	from	ADP
dad-10683	64	33	conversational	conversational	ADJ
dad-10683	64	34	data	datum	NOUN
dad-10683	64	35	.	.	PUNCT
dad-10683	65	1	once	once	ADV
dad-10683	65	2	an	an	DET
dad-10683	65	3	end	end	NOUN
dad-10683	65	4	-	-	PUNCT
dad-10683	65	5	to	to	ADP
dad-10683	65	6	-	-	PUNCT
dad-10683	65	7	end	end	NOUN
dad-10683	65	8	model	model	NOUN
dad-10683	65	9	architecture	architecture	NOUN
dad-10683	65	10	is	be	AUX
dad-10683	65	11	specified	specify	VERB
dad-10683	65	12	,	,	PUNCT
dad-10683	65	13	all	all	PRON
dad-10683	65	14	that	that	PRON
dad-10683	65	15	is	be	AUX
dad-10683	65	16	needed	need	VERB
dad-10683	65	17	to	to	PART
dad-10683	65	18	have	have	VERB
dad-10683	65	19	the	the	DET
dad-10683	65	20	system	system	NOUN
dad-10683	65	21	learn	learn	VERB
dad-10683	65	22	to	to	PART
dad-10683	65	23	converse	converse	VERB
dad-10683	65	24	about	about	ADP
dad-10683	65	25	another	another	DET
dad-10683	65	26	domain	domain	NOUN
dad-10683	65	27	is	be	AUX
dad-10683	65	28	to	to	PART
dad-10683	65	29	provide	provide	VERB
dad-10683	65	30	new	new	ADJ
dad-10683	65	31	training	training	NOUN
dad-10683	65	32	data	datum	NOUN
dad-10683	65	33	for	for	ADP
dad-10683	65	34	that	that	DET
dad-10683	65	35	domain	domain	NOUN
dad-10683	65	36	.	.	PUNCT
dad-10683	66	1	as	as	SCONJ
dad-10683	66	2	the	the	DET
dad-10683	66	3	amount	amount	NOUN
dad-10683	66	4	of	of	ADP
dad-10683	66	5	available	available	ADJ
dad-10683	66	6	dialogue	dialogue	NOUN
dad-10683	66	7	data	datum	NOUN
dad-10683	66	8	grows	grow	VERB
dad-10683	66	9	and	and	CCONJ
dad-10683	66	10	more	more	ADJ
dad-10683	66	11	general	general	ADJ
dad-10683	66	12	-	-	PUNCT
dad-10683	66	13	purpose	purpose	NOUN
dad-10683	66	14	conversational	conversational	ADJ
dad-10683	66	15	systems	system	NOUN
dad-10683	66	16	are	be	AUX
dad-10683	66	17	desired	desire	VERB
dad-10683	66	18	,	,	PUNCT
dad-10683	66	19	we	we	PRON
dad-10683	66	20	believe	believe	VERB
dad-10683	66	21	that	that	SCONJ
dad-10683	66	22	training	training	NOUN
dad-10683	66	23	end	end	NOUN
dad-10683	66	24	-	-	PUNCT
dad-10683	66	25	to	to	ADP
dad-10683	66	26	-	-	PUNCT
dad-10683	66	27	end	end	NOUN
dad-10683	66	28	models	model	NOUN
dad-10683	66	29	without	without	ADP
dad-10683	66	30	hand	hand	NOUN
dad-10683	66	31	-	-	PUNCT
dad-10683	66	32	crafted	craft	VERB
dad-10683	66	33	features	feature	NOUN
dad-10683	66	34	will	will	AUX
dad-10683	66	35	yield	yield	VERB
dad-10683	66	36	better	well	ADJ
dad-10683	66	37	performance	performance	NOUN
dad-10683	66	38	.	.	PUNCT
dad-10683	67	1	however	however	ADV
dad-10683	67	2	,	,	PUNCT
dad-10683	67	3	despite	despite	SCONJ
dad-10683	67	4	these	these	DET
dad-10683	67	5	advantages	advantage	NOUN
dad-10683	67	6	it	it	PRON
dad-10683	67	7	is	be	AUX
dad-10683	67	8	not	not	PART
dad-10683	67	9	clear	clear	ADJ
dad-10683	67	10	whether	whether	SCONJ
dad-10683	67	11	end	end	NOUN
dad-10683	67	12	-	-	PUNCT
dad-10683	67	13	to	to	ADP
dad-10683	67	14	-	-	PUNCT
dad-10683	67	15	end	end	NOUN
dad-10683	67	16	approaches	approach	NOUN
dad-10683	67	17	will	will	AUX
dad-10683	67	18	work	work	VERB
dad-10683	67	19	for	for	ADP
dad-10683	67	20	any	any	DET
dad-10683	67	21	dialogue	dialogue	NOUN
dad-10683	67	22	domain	domain	NOUN
dad-10683	67	23	.	.	PUNCT
dad-10683	68	1	for	for	ADP
dad-10683	68	2	example	example	NOUN
dad-10683	68	3	,	,	PUNCT
dad-10683	68	4	in	in	ADP
dad-10683	68	5	complex	complex	ADJ
dad-10683	68	6	negotiation	negotiation	NOUN
dad-10683	68	7	domains	domain	NOUN
dad-10683	68	8	where	where	SCONJ
dad-10683	68	9	thorough	thorough	ADJ
dad-10683	68	10	analysis	analysis	NOUN
dad-10683	68	11	is	be	AUX
dad-10683	68	12	required	require	VERB
dad-10683	68	13	and	and	CCONJ
dad-10683	68	14	little	little	ADJ
dad-10683	68	15	data	datum	NOUN
dad-10683	68	16	is	be	AUX
dad-10683	68	17	available	available	ADJ
dad-10683	68	18	,	,	PUNCT
dad-10683	68	19	end	end	NOUN
dad-10683	68	20	-	-	PUNCT
dad-10683	68	21	to	to	ADP
dad-10683	68	22	-	-	PUNCT
dad-10683	68	23	end	end	NOUN
dad-10683	68	24	systems	system	NOUN
dad-10683	68	25	may	may	AUX
dad-10683	68	26	not	not	PART
dad-10683	68	27	be	be	AUX
dad-10683	68	28	able	able	ADJ
dad-10683	68	29	to	to	PART
dad-10683	68	30	learn	learn	VERB
dad-10683	68	31	successful	successful	ADJ
dad-10683	68	32	strategies	strategy	NOUN
dad-10683	68	33	for	for	ADP
dad-10683	68	34	every	every	DET
dad-10683	68	35	combination	combination	NOUN
dad-10683	68	36	of	of	ADP
dad-10683	68	37	preferences	preference	NOUN
dad-10683	68	38	and	and	CCONJ
dad-10683	68	39	goals	goal	NOUN
dad-10683	68	40	.	.	PUNCT
dad-10683	69	1	thus	thus	ADV
dad-10683	69	2	,	,	PUNCT
dad-10683	69	3	it	it	PRON
dad-10683	69	4	is	be	AUX
dad-10683	69	5	crucial	crucial	ADJ
dad-10683	69	6	to	to	PART
dad-10683	69	7	conduct	conduct	VERB
dad-10683	69	8	further	further	ADJ
dad-10683	69	9	research	research	NOUN
dad-10683	69	10	on	on	ADP
dad-10683	69	11	end	end	NOUN
dad-10683	69	12	-	-	PUNCT
dad-10683	69	13	to	to	ADP
dad-10683	69	14	-	-	PUNCT
dad-10683	69	15	end	end	NOUN
dad-10683	69	16	dialogue	dialogue	NOUN
dad-10683	69	17	systems	system	NOUN
dad-10683	69	18	in	in	ADP
dad-10683	69	19	order	order	NOUN
dad-10683	69	20	to	to	PART
dad-10683	69	21	determine	determine	VERB
dad-10683	69	22	the	the	DET
dad-10683	69	23	domains	domain	NOUN
dad-10683	69	24	in	in	ADP
dad-10683	69	25	which	which	PRON
dad-10683	69	26	they	they	PRON
dad-10683	69	27	are	be	AUX
dad-10683	69	28	most	most	ADV
dad-10683	69	29	effective	effective	ADJ
dad-10683	69	30	.	.	PUNCT
dad-10683	70	1	1.2	1.2	NUM
dad-10683	70	2	paper	paper	NOUN
dad-10683	70	3	outline	outline	NOUN
dad-10683	70	4	the	the	DET
dad-10683	70	5	paper	paper	NOUN
dad-10683	70	6	is	be	AUX
dad-10683	70	7	structured	structure	VERB
dad-10683	70	8	as	as	SCONJ
dad-10683	70	9	follows	follow	VERB
dad-10683	70	10	.	.	PUNCT
dad-10683	71	1	in	in	ADP
dad-10683	71	2	section	section	NOUN
dad-10683	71	3	2	2	NUM
dad-10683	71	4	,	,	PUNCT
dad-10683	71	5	we	we	PRON
dad-10683	71	6	detail	detail	VERB
dad-10683	71	7	some	some	DET
dad-10683	71	8	relevant	relevant	ADJ
dad-10683	71	9	dialogue	dialogue	NOUN
dad-10683	71	10	datasets	dataset	NOUN
dad-10683	71	11	that	that	PRON
dad-10683	71	12	are	be	AUX
dad-10683	71	13	available	available	ADJ
dad-10683	71	14	,	,	PUNCT
dad-10683	71	15	and	and	CCONJ
dad-10683	71	16	give	give	VERB
dad-10683	71	17	an	an	DET
dad-10683	71	18	overview	overview	NOUN
dad-10683	71	19	of	of	ADP
dad-10683	71	20	existing	exist	VERB
dad-10683	71	21	end	end	NOUN
dad-10683	71	22	-	-	PUNCT
dad-10683	71	23	to	to	ADP
dad-10683	71	24	-	-	PUNCT
dad-10683	71	25	end	end	NOUN
dad-10683	71	26	dialogue	dialogue	NOUN
dad-10683	71	27	systems	system	NOUN
dad-10683	71	28	.	.	PUNCT
dad-10683	72	1	in	in	ADP
dad-10683	72	2	section	section	NOUN
dad-10683	72	3	3	3	NUM
dad-10683	72	4	,	,	PUNCT
dad-10683	72	5	we	we	PRON
dad-10683	72	6	describe	describe	VERB
dad-10683	72	7	34	34	NUM
dad-10683	72	8	training	training	NOUN
dad-10683	72	9	end	end	NOUN
dad-10683	72	10	-	-	PUNCT
dad-10683	72	11	to	to	ADP
dad-10683	72	12	-	-	PUNCT
dad-10683	72	13	end	end	NOUN
dad-10683	72	14	dialogue	dialogue	NOUN
dad-10683	72	15	systems	system	NOUN
dad-10683	72	16	the	the	DET
dad-10683	72	17	ubuntu	ubuntu	ADJ
dad-10683	72	18	dialogue	dialogue	NOUN
dad-10683	72	19	corpus	corpus	NOUN
dad-10683	72	20	,	,	PUNCT
dad-10683	72	21	including	include	VERB
dad-10683	72	22	corpus	corpus	NOUN
dad-10683	72	23	statistics	statistic	NOUN
dad-10683	72	24	,	,	PUNCT
dad-10683	72	25	how	how	SCONJ
dad-10683	72	26	it	it	PRON
dad-10683	72	27	was	be	AUX
dad-10683	72	28	collected	collect	VERB
dad-10683	72	29	,	,	PUNCT
dad-10683	72	30	and	and	CCONJ
dad-10683	72	31	any	any	DET
dad-10683	72	32	changes	change	NOUN
dad-10683	72	33	made	make	VERB
dad-10683	72	34	in	in	ADP
dad-10683	72	35	the	the	DET
dad-10683	72	36	updated	update	VERB
dad-10683	72	37	version	version	NOUN
dad-10683	72	38	.	.	PUNCT
dad-10683	73	1	we	we	PRON
dad-10683	73	2	then	then	ADV
dad-10683	73	3	go	go	VERB
dad-10683	73	4	on	on	ADP
dad-10683	73	5	to	to	PART
dad-10683	73	6	describe	describe	VERB
dad-10683	73	7	the	the	DET
dad-10683	73	8	response	response	NOUN
dad-10683	73	9	ranking	ranking	NOUN
dad-10683	73	10	models	model	NOUN
dad-10683	73	11	on	on	ADP
dad-10683	73	12	the	the	DET
dad-10683	73	13	ubuntu	ubuntu	ADJ
dad-10683	73	14	dialogue	dialogue	NOUN
dad-10683	73	15	corpus	corpus	NOUN
dad-10683	73	16	in	in	ADP
dad-10683	73	17	section	section	NOUN
dad-10683	73	18	4	4	NUM
dad-10683	73	19	,	,	PUNCT
dad-10683	73	20	and	and	CCONJ
dad-10683	73	21	the	the	DET
dad-10683	73	22	response	response	NOUN
dad-10683	73	23	generation	generation	NOUN
dad-10683	73	24	models	model	NOUN
dad-10683	73	25	in	in	ADP
dad-10683	73	26	section	section	NOUN
dad-10683	73	27	5	5	NUM
dad-10683	73	28	.	.	PUNCT
dad-10683	74	1	we	we	PRON
dad-10683	74	2	give	give	VERB
dad-10683	74	3	our	our	PRON
dad-10683	74	4	results	result	NOUN
dad-10683	74	5	in	in	ADP
dad-10683	74	6	each	each	PRON
dad-10683	74	7	of	of	ADP
dad-10683	74	8	these	these	DET
dad-10683	74	9	sections	section	NOUN
dad-10683	74	10	,	,	PUNCT
dad-10683	74	11	including	include	VERB
dad-10683	74	12	our	our	PRON
dad-10683	74	13	models	model	NOUN
dad-10683	74	14	’	'	PUNCT
dad-10683	74	15	performance	performance	NOUN
dad-10683	74	16	on	on	ADP
dad-10683	74	17	next	next	ADJ
dad-10683	74	18	utterance	utterance	NOUN
dad-10683	74	19	classification	classification	NOUN
dad-10683	74	20	and	and	CCONJ
dad-10683	74	21	embedding	embed	VERB
dad-10683	74	22	similarity	similarity	NOUN
dad-10683	74	23	.	.	PUNCT
dad-10683	75	1	we	we	PRON
dad-10683	75	2	also	also	ADV
dad-10683	75	3	provide	provide	VERB
dad-10683	75	4	a	a	DET
dad-10683	75	5	qualitative	qualitative	ADJ
dad-10683	75	6	error	error	NOUN
dad-10683	75	7	analysis	analysis	NOUN
dad-10683	75	8	of	of	ADP
dad-10683	75	9	the	the	DET
dad-10683	75	10	mistakes	mistake	NOUN
dad-10683	75	11	made	make	VERB
dad-10683	75	12	by	by	ADP
dad-10683	75	13	the	the	DET
dad-10683	75	14	classification	classification	NOUN
dad-10683	75	15	model	model	NOUN
dad-10683	75	16	in	in	ADP
dad-10683	75	17	section	section	NOUN
dad-10683	75	18	4.6	4.6	NUM
dad-10683	75	19	.	.	PUNCT
dad-10683	76	1	finally	finally	ADV
dad-10683	76	2	,	,	PUNCT
dad-10683	76	3	we	we	PRON
dad-10683	76	4	conclude	conclude	VERB
dad-10683	76	5	in	in	ADP
dad-10683	76	6	section	section	NOUN
dad-10683	76	7	6	6	NUM
dad-10683	76	8	,	,	PUNCT
dad-10683	76	9	and	and	CCONJ
dad-10683	76	10	discuss	discuss	VERB
dad-10683	76	11	potential	potential	ADJ
dad-10683	76	12	extensions	extension	NOUN
dad-10683	76	13	and	and	CCONJ
dad-10683	76	14	limitations	limitation	NOUN
dad-10683	76	15	of	of	ADP
dad-10683	76	16	the	the	DET
dad-10683	76	17	ubuntu	ubuntu	ADJ
dad-10683	76	18	dialogue	dialogue	NOUN
dad-10683	76	19	corpus	corpus	PROPN
dad-10683	76	20	,	,	PUNCT
dad-10683	76	21	evaluation	evaluation	NOUN
dad-10683	76	22	metrics	metric	NOUN
dad-10683	76	23	,	,	PUNCT
dad-10683	76	24	and	and	CCONJ
dad-10683	76	25	future	future	ADJ
dad-10683	76	26	directions	direction	NOUN
dad-10683	76	27	for	for	ADP
dad-10683	76	28	end	end	NOUN
dad-10683	76	29	-	-	PUNCT
dad-10683	76	30	to	to	ADP
dad-10683	76	31	-	-	PUNCT
dad-10683	76	32	end	end	NOUN
dad-10683	76	33	dialogue	dialogue	NOUN
dad-10683	76	34	systems	system	NOUN
dad-10683	76	35	.	.	PUNCT
dad-10683	77	1	2	2	X
dad-10683	77	2	.	.	X
dad-10683	77	3	related	relate	VERB
dad-10683	77	4	work	work	NOUN
dad-10683	77	5	we	we	PRON
dad-10683	77	6	briefly	briefly	ADV
dad-10683	77	7	review	review	VERB
dad-10683	77	8	existing	exist	VERB
dad-10683	77	9	dialogue	dialogue	NOUN
dad-10683	77	10	datasets	dataset	NOUN
dad-10683	77	11	,	,	PUNCT
dad-10683	77	12	and	and	CCONJ
dad-10683	77	13	some	some	PRON
dad-10683	77	14	of	of	ADP
dad-10683	77	15	the	the	DET
dad-10683	77	16	more	more	ADV
dad-10683	77	17	recent	recent	ADJ
dad-10683	77	18	learning	learning	NOUN
dad-10683	77	19	architectures	architecture	NOUN
dad-10683	77	20	used	use	VERB
dad-10683	77	21	for	for	ADP
dad-10683	77	22	both	both	CCONJ
dad-10683	77	23	structured	structured	ADJ
dad-10683	77	24	and	and	CCONJ
dad-10683	77	25	unstructured	unstructured	ADJ
dad-10683	77	26	dialogues	dialogue	NOUN
dad-10683	77	27	.	.	PUNCT
dad-10683	78	1	this	this	PRON
dad-10683	78	2	is	be	AUX
dad-10683	78	3	by	by	ADP
dad-10683	78	4	no	no	DET
dad-10683	78	5	means	means	NOUN
dad-10683	78	6	an	an	DET
dad-10683	78	7	exhaustive	exhaustive	ADJ
dad-10683	78	8	list	list	NOUN
dad-10683	78	9	,	,	PUNCT
dad-10683	78	10	but	but	CCONJ
dad-10683	78	11	surveys	survey	NOUN
dad-10683	78	12	resources	resource	NOUN
dad-10683	78	13	most	most	ADV
dad-10683	78	14	related	related	ADJ
dad-10683	78	15	to	to	ADP
dad-10683	78	16	our	our	PRON
dad-10683	78	17	contributions	contribution	NOUN
dad-10683	78	18	.	.	PUNCT
dad-10683	79	1	2.1	2.1	NUM
dad-10683	79	2	dialogue	dialogue	NOUN
dad-10683	79	3	datasets	dataset	VERB
dad-10683	79	4	the	the	DET
dad-10683	79	5	switchboard	switchboard	NOUN
dad-10683	79	6	dataset	dataset	NOUN
dad-10683	79	7	(	(	PUNCT
dad-10683	79	8	godfrey	godfrey	PROPN
dad-10683	79	9	et	et	PROPN
dad-10683	79	10	al	al	PROPN
dad-10683	79	11	.	.	PROPN
dad-10683	79	12	,	,	PUNCT
dad-10683	79	13	1992	1992	NUM
dad-10683	79	14	)	)	PUNCT
dad-10683	79	15	,	,	PUNCT
dad-10683	79	16	and	and	CCONJ
dad-10683	79	17	the	the	DET
dad-10683	79	18	dialogue	dialogue	NOUN
dad-10683	79	19	state	state	NOUN
dad-10683	79	20	tracking	tracking	NOUN
dad-10683	79	21	challenge	challenge	NOUN
dad-10683	79	22	(	(	PUNCT
dad-10683	79	23	dstc	dstc	NOUN
dad-10683	79	24	)	)	PUNCT
dad-10683	79	25	datasets	dataset	NOUN
dad-10683	79	26	(	(	PUNCT
dad-10683	79	27	williams	williams	PROPN
dad-10683	79	28	et	et	PROPN
dad-10683	79	29	al	al	PROPN
dad-10683	79	30	.	.	PROPN
dad-10683	79	31	,	,	PUNCT
dad-10683	79	32	2013	2013	NUM
dad-10683	79	33	,	,	PUNCT
dad-10683	79	34	2016	2016	NUM
dad-10683	79	35	)	)	PUNCT
dad-10683	79	36	have	have	AUX
dad-10683	79	37	been	be	AUX
dad-10683	79	38	used	use	VERB
dad-10683	79	39	to	to	PART
dad-10683	79	40	train	train	VERB
dad-10683	79	41	and	and	CCONJ
dad-10683	79	42	validate	validate	VERB
dad-10683	79	43	dialogue	dialogue	NOUN
dad-10683	79	44	management	management	NOUN
dad-10683	79	45	systems	system	NOUN
dad-10683	79	46	for	for	ADP
dad-10683	79	47	interactive	interactive	ADJ
dad-10683	79	48	information	information	NOUN
dad-10683	79	49	retrieval	retrieval	NOUN
dad-10683	79	50	.	.	PUNCT
dad-10683	80	1	in	in	ADP
dad-10683	80	2	the	the	DET
dad-10683	80	3	case	case	NOUN
dad-10683	80	4	of	of	ADP
dad-10683	80	5	the	the	DET
dad-10683	80	6	dstc	dstc	NOUN
dad-10683	80	7	,	,	PUNCT
dad-10683	80	8	the	the	DET
dad-10683	80	9	problem	problem	NOUN
dad-10683	80	10	is	be	AUX
dad-10683	80	11	typically	typically	ADV
dad-10683	80	12	formalized	formalize	VERB
dad-10683	80	13	as	as	ADP
dad-10683	80	14	a	a	DET
dad-10683	80	15	slot	slot	NOUN
dad-10683	80	16	filling	filling	NOUN
dad-10683	80	17	task	task	NOUN
dad-10683	80	18	,	,	PUNCT
dad-10683	80	19	where	where	SCONJ
dad-10683	80	20	agents	agent	NOUN
dad-10683	80	21	attempt	attempt	VERB
dad-10683	80	22	to	to	PART
dad-10683	80	23	predict	predict	VERB
dad-10683	80	24	the	the	DET
dad-10683	80	25	goal	goal	NOUN
dad-10683	80	26	of	of	ADP
dad-10683	80	27	a	a	DET
dad-10683	80	28	user	user	NOUN
dad-10683	80	29	during	during	ADP
dad-10683	80	30	the	the	DET
dad-10683	80	31	conversation	conversation	NOUN
dad-10683	80	32	.	.	PUNCT
dad-10683	81	1	these	these	DET
dad-10683	81	2	datasets	dataset	NOUN
dad-10683	81	3	have	have	AUX
dad-10683	81	4	been	be	AUX
dad-10683	81	5	significant	significant	ADJ
dad-10683	81	6	resources	resource	NOUN
dad-10683	81	7	for	for	ADP
dad-10683	81	8	structured	structured	ADJ
dad-10683	81	9	dialogues	dialogue	NOUN
dad-10683	81	10	,	,	PUNCT
dad-10683	81	11	and	and	CCONJ
dad-10683	81	12	have	have	AUX
dad-10683	81	13	allowed	allow	VERB
dad-10683	81	14	major	major	ADJ
dad-10683	81	15	progress	progress	NOUN
dad-10683	81	16	in	in	ADP
dad-10683	81	17	this	this	DET
dad-10683	81	18	field	field	NOUN
dad-10683	81	19	,	,	PUNCT
dad-10683	81	20	though	though	SCONJ
dad-10683	81	21	they	they	PRON
dad-10683	81	22	are	be	AUX
dad-10683	81	23	relatively	relatively	ADV
dad-10683	81	24	small	small	ADJ
dad-10683	81	25	compared	compare	VERB
dad-10683	81	26	to	to	ADP
dad-10683	81	27	datasets	dataset	NOUN
dad-10683	81	28	currently	currently	ADV
dad-10683	81	29	used	use	VERB
dad-10683	81	30	for	for	ADP
dad-10683	81	31	training	train	VERB
dad-10683	81	32	neural	neural	ADJ
dad-10683	81	33	architectures	architecture	NOUN
dad-10683	81	34	for	for	ADP
dad-10683	81	35	language	language	NOUN
dad-10683	81	36	-	-	PUNCT
dad-10683	81	37	related	relate	VERB
dad-10683	81	38	tasks	task	NOUN
dad-10683	81	39	.	.	PUNCT
dad-10683	82	1	recently	recently	ADV
dad-10683	82	2	,	,	PUNCT
dad-10683	82	3	a	a	DET
dad-10683	82	4	few	few	ADJ
dad-10683	82	5	datasets	dataset	NOUN
dad-10683	82	6	have	have	AUX
dad-10683	82	7	been	be	AUX
dad-10683	82	8	used	use	VERB
dad-10683	82	9	containing	contain	VERB
dad-10683	82	10	unstructured	unstructured	ADJ
dad-10683	82	11	dialogues	dialogue	NOUN
dad-10683	82	12	extracted	extract	VERB
dad-10683	82	13	from	from	ADP
dad-10683	82	14	twitter2	twitter2	PROPN
dad-10683	82	15	.	.	PUNCT
dad-10683	83	1	ritter	ritter	PROPN
dad-10683	83	2	et	et	PROPN
dad-10683	83	3	al	al	PROPN
dad-10683	83	4	.	.	PROPN
dad-10683	84	1	(	(	PUNCT
dad-10683	84	2	2010	2010	NUM
dad-10683	84	3	)	)	PUNCT
dad-10683	84	4	collected	collect	VERB
dad-10683	84	5	1.3	1.3	NUM
dad-10683	84	6	million	million	NUM
dad-10683	84	7	conversations	conversation	NOUN
dad-10683	84	8	;	;	PUNCT
dad-10683	84	9	this	this	PRON
dad-10683	84	10	was	be	AUX
dad-10683	84	11	extended	extend	VERB
dad-10683	84	12	in	in	ADP
dad-10683	84	13	sordoni	sordoni	PROPN
dad-10683	84	14	et	et	PROPN
dad-10683	84	15	al	al	PROPN
dad-10683	84	16	.	.	PROPN
dad-10683	84	17	(	(	PUNCT
dad-10683	84	18	2015b	2015b	NUM
dad-10683	84	19	)	)	PUNCT
dad-10683	84	20	to	to	PART
dad-10683	84	21	take	take	VERB
dad-10683	84	22	advantage	advantage	NOUN
dad-10683	84	23	of	of	ADP
dad-10683	84	24	longer	long	ADJ
dad-10683	84	25	contexts	context	NOUN
dad-10683	84	26	by	by	ADP
dad-10683	84	27	using	use	VERB
dad-10683	84	28	a	a	DET
dad-10683	84	29	-	-	PUNCT
dad-10683	84	30	b	b	NOUN
dad-10683	84	31	-	-	PUNCT
dad-10683	84	32	a	a	DET
dad-10683	84	33	triples	triple	NOUN
dad-10683	84	34	.	.	PUNCT
dad-10683	85	1	shang	shang	PROPN
dad-10683	85	2	et	et	PROPN
dad-10683	85	3	al	al	PROPN
dad-10683	85	4	.	.	PROPN
dad-10683	86	1	(	(	PUNCT
dad-10683	86	2	2015	2015	NUM
dad-10683	86	3	)	)	PUNCT
dad-10683	86	4	used	use	VERB
dad-10683	86	5	data	datum	NOUN
dad-10683	86	6	from	from	ADP
dad-10683	86	7	a	a	DET
dad-10683	86	8	similar	similar	ADJ
dad-10683	86	9	chinese	chinese	ADJ
dad-10683	86	10	website	website	NOUN
dad-10683	86	11	called	call	VERB
dad-10683	86	12	weibo3	weibo3	PROPN
dad-10683	86	13	.	.	PUNCT
dad-10683	87	1	however	however	ADV
dad-10683	87	2	to	to	ADP
dad-10683	87	3	our	our	PRON
dad-10683	87	4	knowledge	knowledge	NOUN
dad-10683	87	5	,	,	PUNCT
dad-10683	87	6	these	these	DET
dad-10683	87	7	datasets	dataset	NOUN
dad-10683	87	8	have	have	AUX
dad-10683	87	9	not	not	PART
dad-10683	87	10	been	be	AUX
dad-10683	87	11	made	make	VERB
dad-10683	87	12	public	public	ADJ
dad-10683	87	13	,	,	PUNCT
dad-10683	87	14	and	and	CCONJ
dad-10683	87	15	furthermore	furthermore	ADV
dad-10683	87	16	,	,	PUNCT
dad-10683	87	17	the	the	DET
dad-10683	87	18	post	post	ADJ
dad-10683	87	19	-	-	ADJ
dad-10683	87	20	reply	reply	ADJ
dad-10683	87	21	format	format	NOUN
dad-10683	87	22	of	of	ADP
dad-10683	87	23	such	such	ADJ
dad-10683	87	24	microblogging	microblogging	NOUN
dad-10683	87	25	services	service	NOUN
dad-10683	87	26	is	be	AUX
dad-10683	87	27	perhaps	perhaps	ADV
dad-10683	87	28	not	not	PART
dad-10683	87	29	as	as	ADV
dad-10683	87	30	representative	representative	NOUN
dad-10683	87	31	of	of	ADP
dad-10683	87	32	natural	natural	ADJ
dad-10683	87	33	dialogue	dialogue	NOUN
dad-10683	87	34	between	between	ADP
dad-10683	87	35	humans	human	NOUN
dad-10683	87	36	as	as	ADP
dad-10683	87	37	the	the	DET
dad-10683	87	38	continuous	continuous	ADJ
dad-10683	87	39	stream	stream	NOUN
dad-10683	87	40	of	of	ADP
dad-10683	87	41	messages	message	NOUN
dad-10683	87	42	in	in	ADP
dad-10683	87	43	a	a	DET
dad-10683	87	44	chat	chat	NOUN
dad-10683	87	45	room	room	NOUN
dad-10683	87	46	.	.	PUNCT
dad-10683	88	1	in	in	ADP
dad-10683	88	2	fact	fact	NOUN
dad-10683	88	3	,	,	PUNCT
dad-10683	88	4	ritter	ritter	PROPN
dad-10683	88	5	et	et	PROPN
dad-10683	88	6	al	al	PROPN
dad-10683	88	7	.	.	PROPN
dad-10683	89	1	(	(	PUNCT
dad-10683	89	2	2010	2010	NUM
dad-10683	89	3	)	)	PUNCT
dad-10683	89	4	estimate	estimate	VERB
dad-10683	89	5	that	that	SCONJ
dad-10683	89	6	only	only	ADV
dad-10683	89	7	37	37	NUM
dad-10683	89	8	%	%	NOUN
dad-10683	89	9	of	of	ADP
dad-10683	89	10	posts	post	NOUN
dad-10683	89	11	on	on	ADP
dad-10683	89	12	twitter	twitter	NOUN
dad-10683	89	13	are	be	AUX
dad-10683	89	14	‘	'	PUNCT
dad-10683	89	15	conversational	conversational	ADJ
dad-10683	89	16	in	in	ADP
dad-10683	89	17	nature	nature	NOUN
dad-10683	89	18	’	'	PUNCT
dad-10683	89	19	,	,	PUNCT
dad-10683	89	20	and	and	CCONJ
dad-10683	89	21	69	69	NUM
dad-10683	89	22	%	%	NOUN
dad-10683	89	23	of	of	ADP
dad-10683	89	24	their	their	PRON
dad-10683	89	25	collected	collect	VERB
dad-10683	89	26	data	data	NOUN
dad-10683	89	27	contains	contain	VERB
dad-10683	89	28	exchanges	exchange	NOUN
dad-10683	89	29	of	of	ADP
dad-10683	89	30	only	only	ADJ
dad-10683	89	31	length	length	NOUN
dad-10683	89	32	2	2	NUM
dad-10683	89	33	(	(	PUNCT
dad-10683	89	34	ritter	ritter	PROPN
dad-10683	89	35	et	et	PROPN
dad-10683	89	36	al	al	PROPN
dad-10683	89	37	.	.	PROPN
dad-10683	89	38	,	,	PUNCT
dad-10683	89	39	2010	2010	NUM
dad-10683	89	40	)	)	PUNCT
dad-10683	89	41	.	.	PUNCT
dad-10683	90	1	we	we	PRON
dad-10683	90	2	hypothesize	hypothesize	VERB
dad-10683	90	3	that	that	SCONJ
dad-10683	90	4	the	the	DET
dad-10683	90	5	interaction	interaction	NOUN
dad-10683	90	6	patterns	pattern	NOUN
dad-10683	90	7	of	of	ADP
dad-10683	90	8	chat	chat	NOUN
dad-10683	90	9	-	-	PUNCT
dad-10683	90	10	room	room	NOUN
dad-10683	90	11	style	style	NOUN
dad-10683	90	12	messaging	messaging	NOUN
dad-10683	90	13	are	be	AUX
dad-10683	90	14	more	more	ADV
dad-10683	90	15	closely	closely	ADV
dad-10683	90	16	correlated	correlate	VERB
dad-10683	90	17	to	to	ADP
dad-10683	90	18	human	human	ADJ
dad-10683	90	19	-	-	PUNCT
dad-10683	90	20	human	human	ADJ
dad-10683	90	21	dialogue	dialogue	NOUN
dad-10683	90	22	than	than	ADP
dad-10683	90	23	micro	micro	ADJ
dad-10683	90	24	-	-	ADJ
dad-10683	90	25	blogging	blogging	ADJ
dad-10683	90	26	websites	website	NOUN
dad-10683	90	27	,	,	PUNCT
dad-10683	90	28	or	or	CCONJ
dad-10683	90	29	forum	forum	NOUN
dad-10683	90	30	-	-	PUNCT
dad-10683	90	31	based	base	VERB
dad-10683	90	32	sites	site	NOUN
dad-10683	90	33	such	such	ADJ
dad-10683	90	34	as	as	ADP
dad-10683	90	35	reddit	reddit	NOUN
dad-10683	90	36	.	.	PUNCT
dad-10683	91	1	part	part	NOUN
dad-10683	91	2	of	of	ADP
dad-10683	91	3	the	the	DET
dad-10683	91	4	ubuntu	ubuntu	NOUN
dad-10683	91	5	chat	chat	NOUN
dad-10683	91	6	logs	log	NOUN
dad-10683	91	7	have	have	AUX
dad-10683	91	8	previously	previously	ADV
dad-10683	91	9	been	be	AUX
dad-10683	91	10	aggregated	aggregate	VERB
dad-10683	91	11	into	into	ADP
dad-10683	91	12	a	a	DET
dad-10683	91	13	dataset	dataset	NOUN
dad-10683	91	14	,	,	PUNCT
dad-10683	91	15	called	call	VERB
dad-10683	91	16	the	the	DET
dad-10683	91	17	ubuntu	ubuntu	ADJ
dad-10683	91	18	chat	chat	PROPN
dad-10683	91	19	corpus	corpus	NOUN
dad-10683	91	20	(	(	PUNCT
dad-10683	91	21	uthus	uthus	PROPN
dad-10683	91	22	and	and	CCONJ
dad-10683	91	23	aha	aha	INTJ
dad-10683	91	24	,	,	PUNCT
dad-10683	91	25	2013b	2013b	NUM
dad-10683	91	26	)	)	PUNCT
dad-10683	91	27	.	.	PUNCT
dad-10683	92	1	however	however	ADV
dad-10683	92	2	that	that	DET
dad-10683	92	3	resource	resource	NOUN
dad-10683	92	4	preserves	preserve	VERB
dad-10683	92	5	the	the	DET
dad-10683	92	6	multi	multi	ADJ
dad-10683	92	7	-	-	ADJ
dad-10683	92	8	participant	participant	ADJ
dad-10683	92	9	structure	structure	NOUN
dad-10683	92	10	and	and	CCONJ
dad-10683	92	11	thus	thus	ADV
dad-10683	92	12	is	be	AUX
dad-10683	92	13	less	less	ADV
dad-10683	92	14	amenable	amenable	ADJ
dad-10683	92	15	to	to	ADP
dad-10683	92	16	the	the	DET
dad-10683	92	17	investigation	investigation	NOUN
dad-10683	92	18	of	of	ADP
dad-10683	92	19	more	more	ADV
dad-10683	92	20	traditional	traditional	ADJ
dad-10683	92	21	two	two	NUM
dad-10683	92	22	-	-	PUNCT
dad-10683	92	23	party	party	NOUN
dad-10683	92	24	conversations	conversation	NOUN
dad-10683	92	25	.	.	PUNCT
dad-10683	93	1	also	also	ADV
dad-10683	93	2	weakly	weakly	ADV
dad-10683	93	3	related	related	ADJ
dad-10683	93	4	to	to	ADP
dad-10683	93	5	our	our	PRON
dad-10683	93	6	contribution	contribution	NOUN
dad-10683	93	7	is	be	AUX
dad-10683	93	8	the	the	DET
dad-10683	93	9	problem	problem	NOUN
dad-10683	93	10	of	of	ADP
dad-10683	93	11	question	question	NOUN
dad-10683	93	12	-	-	PUNCT
dad-10683	93	13	answer	answer	NOUN
dad-10683	93	14	systems	system	NOUN
dad-10683	93	15	.	.	PUNCT
dad-10683	94	1	several	several	ADJ
dad-10683	94	2	datasets	dataset	NOUN
dad-10683	94	3	of	of	ADP
dad-10683	94	4	question	question	NOUN
dad-10683	94	5	-	-	PUNCT
dad-10683	94	6	answer	answer	NOUN
dad-10683	94	7	pairs	pair	NOUN
dad-10683	94	8	are	be	AUX
dad-10683	94	9	available	available	ADJ
dad-10683	94	10	(	(	PUNCT
dad-10683	94	11	boyd	boyd	NOUN
dad-10683	94	12	-	-	PUNCT
dad-10683	94	13	graber	graber	PROPN
dad-10683	94	14	et	et	NOUN
dad-10683	94	15	al	al	PROPN
dad-10683	94	16	.	.	PROPN
dad-10683	94	17	,	,	PUNCT
dad-10683	94	18	2012	2012	NUM
dad-10683	94	19	)	)	PUNCT
dad-10683	94	20	,	,	PUNCT
dad-10683	94	21	however	however	ADV
dad-10683	94	22	these	these	DET
dad-10683	94	23	interactions	interaction	NOUN
dad-10683	94	24	are	be	AUX
dad-10683	94	25	much	much	ADV
dad-10683	94	26	shorter	short	ADJ
dad-10683	94	27	than	than	ADP
dad-10683	94	28	what	what	PRON
dad-10683	94	29	we	we	PRON
dad-10683	94	30	seek	seek	VERB
dad-10683	94	31	to	to	PART
dad-10683	94	32	study	study	VERB
dad-10683	94	33	.	.	PUNCT
dad-10683	95	1	for	for	ADP
dad-10683	95	2	a	a	DET
dad-10683	95	3	comprehensive	comprehensive	ADJ
dad-10683	95	4	survey	survey	NOUN
dad-10683	95	5	of	of	ADP
dad-10683	95	6	both	both	CCONJ
dad-10683	95	7	available	available	ADJ
dad-10683	95	8	dialogue	dialogue	NOUN
dad-10683	95	9	datasets	dataset	NOUN
dad-10683	95	10	and	and	CCONJ
dad-10683	95	11	prevalent	prevalent	ADJ
dad-10683	95	12	models	model	NOUN
dad-10683	95	13	,	,	PUNCT
dad-10683	95	14	see	see	VERB
dad-10683	95	15	serban	serban	PROPN
dad-10683	95	16	et	et	PROPN
dad-10683	95	17	al	al	PROPN
dad-10683	95	18	.	.	PROPN
dad-10683	96	1	(	(	PUNCT
dad-10683	96	2	2015	2015	NUM
dad-10683	96	3	)	)	PUNCT
dad-10683	96	4	.	.	PUNCT
dad-10683	97	1	2	2	X
dad-10683	97	2	.	.	X
dad-10683	97	3	https://twitter.com/	https://twitter.com/	PROPN
dad-10683	97	4	3	3	NUM
dad-10683	97	5	.	.	PUNCT
dad-10683	97	6	http://www.weibo.com/	http://www.weibo.com/	PROPN
dad-10683	97	7	35	35	NUM
dad-10683	97	8	lowe	lowe	NOUN
dad-10683	97	9	,	,	PUNCT
dad-10683	97	10	pow	pow	PROPN
dad-10683	97	11	,	,	PUNCT
dad-10683	97	12	serban	serban	PROPN
dad-10683	97	13	,	,	PUNCT
dad-10683	97	14	charlinn	charlinn	PROPN
dad-10683	97	15	,	,	PUNCT
dad-10683	97	16	liu	liu	PROPN
dad-10683	97	17	and	and	CCONJ
dad-10683	97	18	pineau	pineau	PROPN
dad-10683	97	19	2.2	2.2	NUM
dad-10683	97	20	learning	learning	NOUN
dad-10683	97	21	architectures	architecture	NOUN
dad-10683	97	22	for	for	ADP
dad-10683	97	23	end	end	NOUN
dad-10683	97	24	-	-	PUNCT
dad-10683	97	25	to	to	ADP
dad-10683	97	26	-	-	PUNCT
dad-10683	97	27	end	end	NOUN
dad-10683	97	28	dialogue	dialogue	NOUN
dad-10683	97	29	systems	system	NOUN
dad-10683	97	30	most	most	ADJ
dad-10683	97	31	dialogue	dialogue	NOUN
dad-10683	97	32	research	research	NOUN
dad-10683	97	33	has	have	AUX
dad-10683	97	34	historically	historically	ADV
dad-10683	97	35	focused	focus	VERB
dad-10683	97	36	on	on	ADP
dad-10683	97	37	structured	structured	ADJ
dad-10683	97	38	slot	slot	NOUN
dad-10683	97	39	-	-	PUNCT
dad-10683	97	40	filling	fill	VERB
dad-10683	97	41	tasks	task	NOUN
dad-10683	97	42	(	(	PUNCT
dad-10683	97	43	schatzmann	schatzmann	X
dad-10683	97	44	et	et	PROPN
dad-10683	97	45	al	al	PROPN
dad-10683	97	46	.	.	PROPN
dad-10683	97	47	,	,	PUNCT
dad-10683	97	48	2005	2005	NUM
dad-10683	97	49	)	)	PUNCT
dad-10683	97	50	.	.	PUNCT
dad-10683	98	1	various	various	ADJ
dad-10683	98	2	approaches	approach	NOUN
dad-10683	98	3	were	be	AUX
dad-10683	98	4	proposed	propose	VERB
dad-10683	98	5	,	,	PUNCT
dad-10683	98	6	yet	yet	CCONJ
dad-10683	98	7	few	few	ADJ
dad-10683	98	8	attempts	attempt	NOUN
dad-10683	98	9	leverage	leverage	VERB
dad-10683	98	10	more	more	ADV
dad-10683	98	11	recent	recent	ADJ
dad-10683	98	12	developments	development	NOUN
dad-10683	98	13	in	in	ADP
dad-10683	98	14	neural	neural	ADJ
dad-10683	98	15	learning	learning	NOUN
dad-10683	98	16	architectures	architecture	NOUN
dad-10683	98	17	.	.	PUNCT
dad-10683	99	1	a	a	DET
dad-10683	99	2	notable	notable	ADJ
dad-10683	99	3	exception	exception	NOUN
dad-10683	99	4	is	be	AUX
dad-10683	99	5	the	the	DET
dad-10683	99	6	work	work	NOUN
dad-10683	99	7	of	of	ADP
dad-10683	99	8	henderson	henderson	PROPN
dad-10683	99	9	et	et	PROPN
dad-10683	99	10	al	al	PROPN
dad-10683	99	11	.	.	PROPN
dad-10683	100	1	(	(	PUNCT
dad-10683	100	2	2014b	2014b	NUM
dad-10683	100	3	)	)	PUNCT
dad-10683	100	4	,	,	PUNCT
dad-10683	100	5	which	which	PRON
dad-10683	100	6	proposes	propose	VERB
dad-10683	100	7	an	an	DET
dad-10683	100	8	rnn	rnn	NOUN
dad-10683	100	9	structure	structure	NOUN
dad-10683	100	10	,	,	PUNCT
dad-10683	100	11	initialized	initialize	VERB
dad-10683	100	12	with	with	ADP
dad-10683	100	13	a	a	DET
dad-10683	100	14	denoising	denoise	VERB
dad-10683	100	15	autoencoder	autoencoder	NOUN
dad-10683	100	16	,	,	PUNCT
dad-10683	100	17	to	to	PART
dad-10683	100	18	tackle	tackle	VERB
dad-10683	100	19	the	the	DET
dad-10683	100	20	dstc	dstc	NOUN
dad-10683	100	21	3	3	NUM
dad-10683	100	22	domain	domain	NOUN
dad-10683	100	23	.	.	PUNCT
dad-10683	101	1	work	work	NOUN
dad-10683	101	2	on	on	ADP
dad-10683	101	3	end	end	NOUN
dad-10683	101	4	-	-	PUNCT
dad-10683	101	5	to	to	ADP
dad-10683	101	6	-	-	PUNCT
dad-10683	101	7	end	end	NOUN
dad-10683	101	8	dialogue	dialogue	NOUN
dad-10683	101	9	systems	system	NOUN
dad-10683	101	10	was	be	AUX
dad-10683	101	11	recently	recently	ADV
dad-10683	101	12	pioneered	pioneer	VERB
dad-10683	101	13	by	by	ADP
dad-10683	101	14	ritter	ritter	PROPN
dad-10683	101	15	et	et	PROPN
dad-10683	101	16	al	al	PROPN
dad-10683	101	17	.	.	PROPN
dad-10683	102	1	(	(	PUNCT
dad-10683	102	2	2011	2011	NUM
dad-10683	102	3	)	)	PUNCT
dad-10683	102	4	,	,	PUNCT
dad-10683	102	5	who	who	PRON
dad-10683	102	6	proposed	propose	VERB
dad-10683	102	7	a	a	DET
dad-10683	102	8	response	response	NOUN
dad-10683	102	9	generation	generation	NOUN
dad-10683	102	10	model	model	NOUN
dad-10683	102	11	for	for	ADP
dad-10683	102	12	twitter	twitter	NOUN
dad-10683	102	13	data	datum	NOUN
dad-10683	102	14	based	base	VERB
dad-10683	102	15	on	on	ADP
dad-10683	102	16	ideas	idea	NOUN
dad-10683	102	17	from	from	ADP
dad-10683	102	18	statistical	statistical	ADJ
dad-10683	102	19	machine	machine	NOUN
dad-10683	102	20	translation	translation	NOUN
dad-10683	102	21	.	.	PUNCT
dad-10683	103	1	in	in	ADP
dad-10683	103	2	particular	particular	ADJ
dad-10683	103	3	,	,	PUNCT
dad-10683	103	4	they	they	PRON
dad-10683	103	5	consider	consider	VERB
dad-10683	103	6	a	a	DET
dad-10683	103	7	model	model	NOUN
dad-10683	103	8	that	that	SCONJ
dad-10683	103	9	‘	'	PUNCT
dad-10683	103	10	translates	translate	VERB
dad-10683	103	11	’	'	PUNCT
dad-10683	103	12	from	from	ADP
dad-10683	103	13	the	the	DET
dad-10683	103	14	context	context	NOUN
dad-10683	103	15	of	of	ADP
dad-10683	103	16	a	a	DET
dad-10683	103	17	conversation	conversation	NOUN
dad-10683	103	18	to	to	ADP
dad-10683	103	19	the	the	DET
dad-10683	103	20	associated	associated	ADJ
dad-10683	103	21	response	response	NOUN
dad-10683	103	22	.	.	PUNCT
dad-10683	104	1	this	this	PRON
dad-10683	104	2	is	be	AUX
dad-10683	104	3	shown	show	VERB
dad-10683	104	4	to	to	PART
dad-10683	104	5	give	give	VERB
dad-10683	104	6	superior	superior	ADJ
dad-10683	104	7	performance	performance	NOUN
dad-10683	104	8	to	to	ADP
dad-10683	104	9	previous	previous	ADJ
dad-10683	104	10	information	information	NOUN
dad-10683	104	11	retrieval	retrieval	NOUN
dad-10683	104	12	(	(	PUNCT
dad-10683	104	13	e.g.	e.g.	ADV
dad-10683	104	14	nearest	near	ADJ
dad-10683	104	15	neighbour	neighbour	NOUN
dad-10683	104	16	)	)	PUNCT
dad-10683	104	17	approaches	approach	NOUN
dad-10683	104	18	(	(	PUNCT
dad-10683	104	19	jafarpour	jafarpour	PROPN
dad-10683	104	20	et	et	PROPN
dad-10683	104	21	al	al	PROPN
dad-10683	104	22	.	.	PROPN
dad-10683	104	23	,	,	PUNCT
dad-10683	104	24	2010	2010	NUM
dad-10683	104	25	)	)	PUNCT
dad-10683	104	26	.	.	PUNCT
dad-10683	105	1	this	this	DET
dad-10683	105	2	idea	idea	NOUN
dad-10683	105	3	was	be	AUX
dad-10683	105	4	further	far	ADV
dad-10683	105	5	developed	develop	VERB
dad-10683	105	6	by	by	ADP
dad-10683	105	7	sordoni	sordoni	PROPN
dad-10683	105	8	et	et	PROPN
dad-10683	105	9	al	al	PROPN
dad-10683	105	10	.	.	PROPN
dad-10683	105	11	(	(	PUNCT
dad-10683	105	12	2015b	2015b	NUM
dad-10683	105	13	)	)	PUNCT
dad-10683	105	14	to	to	PART
dad-10683	105	15	exploit	exploit	VERB
dad-10683	105	16	information	information	NOUN
dad-10683	105	17	from	from	ADP
dad-10683	105	18	a	a	DET
dad-10683	105	19	longer	long	ADJ
dad-10683	105	20	context	context	NOUN
dad-10683	105	21	,	,	PUNCT
dad-10683	105	22	using	use	VERB
dad-10683	105	23	a	a	DET
dad-10683	105	24	structure	structure	NOUN
dad-10683	105	25	similar	similar	ADJ
dad-10683	105	26	to	to	ADP
dad-10683	105	27	the	the	DET
dad-10683	105	28	recurrent	recurrent	ADJ
dad-10683	105	29	neural	neural	ADJ
dad-10683	105	30	network	network	NOUN
dad-10683	105	31	encoder	encoder	NOUN
dad-10683	105	32	-	-	PUNCT
dad-10683	105	33	decoder	decoder	NOUN
dad-10683	105	34	model	model	NOUN
dad-10683	105	35	(	(	PUNCT
dad-10683	105	36	cho	cho	PROPN
dad-10683	105	37	et	et	PROPN
dad-10683	105	38	al	al	PROPN
dad-10683	105	39	.	.	PROPN
dad-10683	105	40	,	,	PUNCT
dad-10683	105	41	2014	2014	NUM
dad-10683	105	42	)	)	PUNCT
dad-10683	105	43	.	.	PUNCT
dad-10683	106	1	this	this	PRON
dad-10683	106	2	achieves	achieve	VERB
dad-10683	106	3	rather	rather	ADV
dad-10683	106	4	poor	poor	ADJ
dad-10683	106	5	performance	performance	NOUN
dad-10683	106	6	on	on	ADP
dad-10683	106	7	a	a	DET
dad-10683	106	8	-	-	PUNCT
dad-10683	106	9	b	b	NOUN
dad-10683	106	10	-	-	PUNCT
dad-10683	106	11	a	a	PRON
dad-10683	106	12	twitter	twitter	NOUN
dad-10683	106	13	triples	triple	NOUN
dad-10683	106	14	when	when	SCONJ
dad-10683	106	15	measured	measure	VERB
dad-10683	106	16	by	by	ADP
dad-10683	106	17	the	the	DET
dad-10683	106	18	bleu	bleu	NOUN
dad-10683	106	19	score	score	NOUN
dad-10683	106	20	(	(	PUNCT
dad-10683	106	21	a	a	DET
dad-10683	106	22	standard	standard	NOUN
dad-10683	106	23	for	for	ADP
dad-10683	106	24	machine	machine	NOUN
dad-10683	106	25	translation	translation	NOUN
dad-10683	106	26	)	)	PUNCT
dad-10683	106	27	,	,	PUNCT
dad-10683	106	28	yet	yet	CCONJ
dad-10683	106	29	performs	perform	VERB
dad-10683	106	30	comparatively	comparatively	ADV
dad-10683	106	31	better	well	ADJ
dad-10683	106	32	than	than	ADP
dad-10683	106	33	the	the	DET
dad-10683	106	34	model	model	NOUN
dad-10683	106	35	of	of	ADP
dad-10683	106	36	ritter	ritter	PROPN
dad-10683	106	37	et	et	PROPN
dad-10683	106	38	al	al	PROPN
dad-10683	106	39	.	.	PROPN
dad-10683	107	1	(	(	PUNCT
dad-10683	107	2	2011	2011	NUM
dad-10683	107	3	)	)	PUNCT
dad-10683	107	4	.	.	PUNCT
dad-10683	108	1	their	their	PRON
dad-10683	108	2	results	result	NOUN
dad-10683	108	3	were	be	AUX
dad-10683	108	4	also	also	ADV
dad-10683	108	5	verified	verify	VERB
dad-10683	108	6	with	with	ADP
dad-10683	108	7	a	a	DET
dad-10683	108	8	human	human	ADJ
dad-10683	108	9	-	-	PUNCT
dad-10683	108	10	subject	subject	NOUN
dad-10683	108	11	study	study	NOUN
dad-10683	108	12	.	.	PUNCT
dad-10683	109	1	a	a	DET
dad-10683	109	2	similar	similar	ADJ
dad-10683	109	3	encoder	encoder	NOUN
dad-10683	109	4	-	-	PUNCT
dad-10683	109	5	decoder	decoder	NOUN
dad-10683	109	6	framework	framework	NOUN
dad-10683	109	7	for	for	ADP
dad-10683	109	8	dialogue	dialogue	NOUN
dad-10683	109	9	is	be	AUX
dad-10683	109	10	presented	present	VERB
dad-10683	109	11	by	by	ADP
dad-10683	109	12	shang	shang	PROPN
dad-10683	109	13	et	et	PROPN
dad-10683	109	14	al	al	PROPN
dad-10683	109	15	.	.	PROPN
dad-10683	110	1	(	(	PUNCT
dad-10683	110	2	2015	2015	NUM
dad-10683	110	3	)	)	PUNCT
dad-10683	110	4	and	and	CCONJ
dad-10683	110	5	vinyals	vinyal	NOUN
dad-10683	110	6	and	and	CCONJ
dad-10683	110	7	le	le	X
dad-10683	110	8	(	(	PUNCT
dad-10683	110	9	2015	2015	NUM
dad-10683	110	10	)	)	PUNCT
dad-10683	110	11	.	.	PUNCT
dad-10683	111	1	this	this	DET
dad-10683	111	2	model	model	NOUN
dad-10683	111	3	also	also	ADV
dad-10683	111	4	uses	use	VERB
dad-10683	111	5	one	one	NUM
dad-10683	111	6	rnn	rnn	NOUN
dad-10683	111	7	to	to	PART
dad-10683	111	8	transform	transform	VERB
dad-10683	111	9	the	the	DET
dad-10683	111	10	input	input	NOUN
dad-10683	111	11	to	to	ADP
dad-10683	111	12	some	some	DET
dad-10683	111	13	vector	vector	NOUN
dad-10683	111	14	representation	representation	NOUN
dad-10683	111	15	,	,	PUNCT
dad-10683	111	16	and	and	CCONJ
dad-10683	111	17	another	another	PRON
dad-10683	111	18	rnn	rnn	VERB
dad-10683	111	19	to	to	ADP
dad-10683	111	20	‘	'	PUNCT
dad-10683	111	21	decode	decode	VERB
dad-10683	111	22	’	'	PUNCT
dad-10683	111	23	this	this	DET
dad-10683	111	24	representation	representation	NOUN
dad-10683	111	25	to	to	ADP
dad-10683	111	26	a	a	DET
dad-10683	111	27	response	response	NOUN
dad-10683	111	28	by	by	ADP
dad-10683	111	29	generating	generate	VERB
dad-10683	111	30	one	one	NUM
dad-10683	111	31	word	word	NOUN
dad-10683	111	32	at	at	ADP
dad-10683	111	33	a	a	DET
dad-10683	111	34	time	time	NOUN
dad-10683	111	35	.	.	PUNCT
dad-10683	112	1	the	the	DET
dad-10683	112	2	model	model	NOUN
dad-10683	112	3	from	from	ADP
dad-10683	112	4	shang	shang	PROPN
dad-10683	112	5	et	et	PROPN
dad-10683	112	6	al	al	PROPN
dad-10683	112	7	.	.	PROPN
dad-10683	112	8	(	(	PUNCT
dad-10683	112	9	2015	2015	NUM
dad-10683	112	10	)	)	PUNCT
dad-10683	112	11	was	be	AUX
dad-10683	112	12	also	also	ADV
dad-10683	112	13	evaluated	evaluate	VERB
dad-10683	112	14	in	in	ADP
dad-10683	112	15	a	a	DET
dad-10683	112	16	human	human	ADJ
dad-10683	112	17	-	-	PUNCT
dad-10683	112	18	subject	subject	NOUN
dad-10683	112	19	study	study	NOUN
dad-10683	112	20	,	,	PUNCT
dad-10683	112	21	although	although	SCONJ
dad-10683	112	22	on	on	ADP
dad-10683	112	23	a	a	DET
dad-10683	112	24	smaller	small	ADJ
dad-10683	112	25	scale	scale	NOUN
dad-10683	112	26	compared	compare	VERB
dad-10683	112	27	to	to	PART
dad-10683	112	28	sordoni	sordoni	VERB
dad-10683	112	29	et	et	PROPN
dad-10683	112	30	al	al	PROPN
dad-10683	112	31	.	.	PROPN
dad-10683	113	1	(	(	PUNCT
dad-10683	113	2	2015b	2015b	NUM
dad-10683	113	3	)	)	PUNCT
dad-10683	113	4	.	.	PUNCT
dad-10683	114	1	a	a	DET
dad-10683	114	2	hierarchical	hierarchical	ADJ
dad-10683	114	3	version	version	NOUN
dad-10683	114	4	of	of	ADP
dad-10683	114	5	the	the	DET
dad-10683	114	6	encoder	encoder	NOUN
dad-10683	114	7	-	-	PUNCT
dad-10683	114	8	decoder	decoder	NOUN
dad-10683	114	9	framework	framework	NOUN
dad-10683	114	10	has	have	AUX
dad-10683	114	11	also	also	ADV
dad-10683	114	12	recently	recently	ADV
dad-10683	114	13	been	be	AUX
dad-10683	114	14	proposed	propose	VERB
dad-10683	114	15	(	(	PUNCT
dad-10683	114	16	serban	serban	PROPN
dad-10683	114	17	et	et	PROPN
dad-10683	114	18	al	al	PROPN
dad-10683	114	19	.	.	PROPN
dad-10683	114	20	,	,	PUNCT
dad-10683	114	21	2016	2016	NUM
dad-10683	114	22	)	)	PUNCT
dad-10683	114	23	.	.	PUNCT
dad-10683	115	1	this	this	DET
dad-10683	115	2	model	model	NOUN
dad-10683	115	3	consists	consist	VERB
dad-10683	115	4	of	of	ADP
dad-10683	115	5	two	two	NUM
dad-10683	115	6	rnns	rnn	NOUN
dad-10683	115	7	stacked	stack	VERB
dad-10683	115	8	on	on	ADP
dad-10683	115	9	top	top	NOUN
dad-10683	115	10	of	of	ADP
dad-10683	115	11	each	each	DET
dad-10683	115	12	other	other	ADJ
dad-10683	115	13	:	:	PUNCT
dad-10683	115	14	one	one	NUM
dad-10683	115	15	‘	'	PUNCT
dad-10683	115	16	sentencelevel	sentencelevel	NOUN
dad-10683	115	17	’	'	PUNCT
dad-10683	115	18	rnn	rnn	VERB
dad-10683	115	19	encodes	encode	NOUN
dad-10683	115	20	each	each	DET
dad-10683	115	21	utterance	utterance	NOUN
dad-10683	115	22	into	into	ADP
dad-10683	115	23	a	a	DET
dad-10683	115	24	fixed	fix	VERB
dad-10683	115	25	length	length	NOUN
dad-10683	115	26	vector	vector	NOUN
dad-10683	115	27	,	,	PUNCT
dad-10683	115	28	while	while	SCONJ
dad-10683	115	29	a	a	DET
dad-10683	115	30	‘	'	PUNCT
dad-10683	115	31	conversation	conversation	ADJ
dad-10683	115	32	-	-	PUNCT
dad-10683	115	33	level	level	NOUN
dad-10683	115	34	’	'	PUNCT
dad-10683	115	35	rnn	rnn	NOUN
dad-10683	115	36	takes	take	VERB
dad-10683	115	37	as	as	ADP
dad-10683	115	38	input	input	NOUN
dad-10683	115	39	each	each	DET
dad-10683	115	40	utterance	utterance	NOUN
dad-10683	115	41	vector	vector	NOUN
dad-10683	115	42	and	and	CCONJ
dad-10683	115	43	outputs	output	NOUN
dad-10683	115	44	a	a	DET
dad-10683	115	45	vector	vector	NOUN
dad-10683	115	46	that	that	PRON
dad-10683	115	47	summarizes	summarize	VERB
dad-10683	115	48	the	the	DET
dad-10683	115	49	conversation	conversation	NOUN
dad-10683	115	50	so	so	ADV
dad-10683	115	51	far	far	ADV
dad-10683	115	52	.	.	PUNCT
dad-10683	116	1	this	this	PRON
dad-10683	116	2	is	be	AUX
dad-10683	116	3	mapped	map	VERB
dad-10683	116	4	back	back	ADV
dad-10683	116	5	to	to	ADP
dad-10683	116	6	text	text	NOUN
dad-10683	116	7	using	use	VERB
dad-10683	116	8	a	a	DET
dad-10683	116	9	recurrent	recurrent	ADJ
dad-10683	116	10	decoder	decoder	NOUN
dad-10683	116	11	.	.	PUNCT
dad-10683	117	1	this	this	PRON
dad-10683	117	2	improves	improve	VERB
dad-10683	117	3	over	over	ADP
dad-10683	117	4	the	the	DET
dad-10683	117	5	traditional	traditional	ADJ
dad-10683	117	6	encoderdecoder	encoderdecoder	NOUN
dad-10683	117	7	frameworks	framework	NOUN
dad-10683	117	8	in	in	ADP
dad-10683	117	9	both	both	PRON
dad-10683	117	10	word	word	NOUN
dad-10683	117	11	perplexity	perplexity	NOUN
dad-10683	117	12	and	and	CCONJ
dad-10683	117	13	word	word	NOUN
dad-10683	117	14	error	error	NOUN
dad-10683	117	15	rate	rate	NOUN
dad-10683	117	16	,	,	PUNCT
dad-10683	117	17	particularly	particularly	ADV
dad-10683	117	18	when	when	SCONJ
dad-10683	117	19	bootstrapped	bootstrappe	VERB
dad-10683	117	20	with	with	ADP
dad-10683	117	21	word	word	NOUN
dad-10683	117	22	embeddings	embedding	NOUN
dad-10683	117	23	derived	derive	VERB
dad-10683	117	24	from	from	ADP
dad-10683	117	25	distributional	distributional	ADJ
dad-10683	117	26	semantics	semantic	NOUN
dad-10683	117	27	.	.	PUNCT
dad-10683	118	1	however	however	ADV
dad-10683	118	2	,	,	PUNCT
dad-10683	118	3	the	the	DET
dad-10683	118	4	model	model	NOUN
dad-10683	118	5	has	have	AUX
dad-10683	118	6	not	not	PART
dad-10683	118	7	been	be	AUX
dad-10683	118	8	evaluated	evaluate	VERB
dad-10683	118	9	in	in	ADP
dad-10683	118	10	any	any	DET
dad-10683	118	11	human	human	ADJ
dad-10683	118	12	-	-	PUNCT
dad-10683	118	13	subject	subject	NOUN
dad-10683	118	14	studies	study	NOUN
dad-10683	118	15	.	.	PUNCT
dad-10683	119	1	another	another	DET
dad-10683	119	2	approach	approach	NOUN
dad-10683	119	3	,	,	PUNCT
dad-10683	119	4	taken	take	VERB
dad-10683	119	5	in	in	ADP
dad-10683	119	6	traum	traum	PROPN
dad-10683	119	7	et	et	PROPN
dad-10683	119	8	al	al	PROPN
dad-10683	119	9	.	.	PROPN
dad-10683	119	10	(	(	PUNCT
dad-10683	119	11	2015	2015	NUM
dad-10683	119	12	)	)	PUNCT
dad-10683	119	13	,	,	PUNCT
dad-10683	119	14	uses	use	VERB
dad-10683	119	15	information	information	NOUN
dad-10683	119	16	retrieval	retrieval	NOUN
dad-10683	119	17	techniques	technique	NOUN
dad-10683	119	18	to	to	PART
dad-10683	119	19	map	map	VERB
dad-10683	119	20	user	user	NOUN
dad-10683	119	21	questions	question	NOUN
dad-10683	119	22	to	to	ADP
dad-10683	119	23	systems	system	NOUN
dad-10683	119	24	responses	response	NOUN
dad-10683	119	25	in	in	ADP
dad-10683	119	26	the	the	DET
dad-10683	119	27	domain	domain	NOUN
dad-10683	119	28	of	of	ADP
dad-10683	119	29	time	time	NOUN
dad-10683	119	30	-	-	PUNCT
dad-10683	119	31	offset	offset	NOUN
dad-10683	119	32	interaction	interaction	NOUN
dad-10683	119	33	.	.	PUNCT
dad-10683	120	1	since	since	SCONJ
dad-10683	120	2	the	the	DET
dad-10683	120	3	natural	natural	ADJ
dad-10683	120	4	language	language	NOUN
dad-10683	120	5	interpreter	interpreter	NOUN
dad-10683	120	6	,	,	PUNCT
dad-10683	120	7	dialogue	dialogue	NOUN
dad-10683	120	8	response	response	NOUN
dad-10683	120	9	selection	selection	NOUN
dad-10683	120	10	,	,	PUNCT
dad-10683	120	11	and	and	CCONJ
dad-10683	120	12	natural	natural	ADJ
dad-10683	120	13	language	language	NOUN
dad-10683	120	14	generator	generator	NOUN
dad-10683	120	15	model	model	NOUN
dad-10683	120	16	are	be	AUX
dad-10683	120	17	all	all	PRON
dad-10683	120	18	combined	combine	VERB
dad-10683	120	19	,	,	PUNCT
dad-10683	120	20	this	this	PRON
dad-10683	120	21	can	can	AUX
dad-10683	120	22	also	also	ADV
dad-10683	120	23	be	be	AUX
dad-10683	120	24	seen	see	VERB
dad-10683	120	25	as	as	ADP
dad-10683	120	26	a	a	DET
dad-10683	120	27	form	form	NOUN
dad-10683	120	28	of	of	ADP
dad-10683	120	29	end	end	NOUN
dad-10683	120	30	-	-	PUNCT
dad-10683	120	31	to	to	ADP
dad-10683	120	32	-	-	PUNCT
dad-10683	120	33	end	end	NOUN
dad-10683	120	34	dialogue	dialogue	NOUN
dad-10683	120	35	system	system	NOUN
dad-10683	120	36	.	.	PUNCT
dad-10683	121	1	inaba	inaba	PROPN
dad-10683	121	2	and	and	CCONJ
dad-10683	121	3	takahashi	takahashi	PROPN
dad-10683	121	4	(	(	PUNCT
dad-10683	121	5	2016	2016	NUM
dad-10683	121	6	)	)	PUNCT
dad-10683	121	7	also	also	ADV
dad-10683	121	8	propose	propose	VERB
dad-10683	121	9	an	an	DET
dad-10683	121	10	end	end	NOUN
dad-10683	121	11	-	-	PUNCT
dad-10683	121	12	to	to	ADP
dad-10683	121	13	-	-	PUNCT
dad-10683	121	14	end	end	NOUN
dad-10683	121	15	retrieval	retrieval	NOUN
dad-10683	121	16	model	model	NOUN
dad-10683	121	17	,	,	PUNCT
dad-10683	121	18	however	however	ADV
dad-10683	121	19	they	they	PRON
dad-10683	121	20	use	use	VERB
dad-10683	121	21	neural	neural	ADJ
dad-10683	121	22	networks	network	NOUN
dad-10683	121	23	to	to	PART
dad-10683	121	24	select	select	VERB
dad-10683	121	25	a	a	DET
dad-10683	121	26	response	response	NOUN
dad-10683	121	27	from	from	ADP
dad-10683	121	28	a	a	DET
dad-10683	121	29	fixed	fix	VERB
dad-10683	121	30	dataset	dataset	NOUN
dad-10683	121	31	.	.	PUNCT
dad-10683	122	1	this	this	PRON
dad-10683	122	2	is	be	AUX
dad-10683	122	3	similar	similar	ADJ
dad-10683	122	4	to	to	ADP
dad-10683	122	5	the	the	DET
dad-10683	122	6	model	model	NOUN
dad-10683	122	7	used	use	VERB
dad-10683	122	8	by	by	ADP
dad-10683	122	9	lowe	lowe	PROPN
dad-10683	122	10	et	et	PROPN
dad-10683	122	11	al	al	PROPN
dad-10683	122	12	.	.	PROPN
dad-10683	123	1	(	(	PUNCT
dad-10683	123	2	2015	2015	NUM
dad-10683	123	3	)	)	PUNCT
dad-10683	123	4	.	.	PUNCT
dad-10683	124	1	our	our	PRON
dad-10683	124	2	work	work	NOUN
dad-10683	124	3	is	be	AUX
dad-10683	124	4	also	also	ADV
dad-10683	124	5	inspired	inspire	VERB
dad-10683	124	6	by	by	ADP
dad-10683	124	7	nio	nio	PROPN
dad-10683	124	8	et	et	PROPN
dad-10683	124	9	al	al	PROPN
dad-10683	124	10	.	.	PROPN
dad-10683	125	1	(	(	PUNCT
dad-10683	125	2	2014	2014	NUM
dad-10683	125	3	)	)	PUNCT
dad-10683	125	4	whose	whose	DET
dad-10683	125	5	model	model	NOUN
dad-10683	125	6	,	,	PUNCT
dad-10683	125	7	although	although	SCONJ
dad-10683	125	8	rule	rule	NOUN
dad-10683	125	9	-	-	PUNCT
dad-10683	125	10	based	base	VERB
dad-10683	125	11	,	,	PUNCT
dad-10683	125	12	is	be	AUX
dad-10683	125	13	not	not	PART
dad-10683	125	14	composed	compose	VERB
dad-10683	125	15	of	of	ADP
dad-10683	125	16	modules	module	NOUN
dad-10683	125	17	,	,	PUNCT
dad-10683	125	18	as	as	SCONJ
dad-10683	125	19	it	it	PRON
dad-10683	125	20	retrieves	retrieve	VERB
dad-10683	125	21	a	a	DET
dad-10683	125	22	response	response	NOUN
dad-10683	125	23	to	to	ADP
dad-10683	125	24	the	the	DET
dad-10683	125	25	context	context	NOUN
dad-10683	125	26	based	base	VERB
dad-10683	125	27	on	on	ADP
dad-10683	125	28	cosine	cosine	NOUN
dad-10683	125	29	similarity	similarity	NOUN
dad-10683	125	30	.	.	PUNCT
dad-10683	126	1	this	this	PRON
dad-10683	126	2	is	be	AUX
dad-10683	126	3	in	in	ADP
dad-10683	126	4	turn	turn	NOUN
dad-10683	126	5	related	relate	VERB
dad-10683	126	6	to	to	ADP
dad-10683	126	7	the	the	DET
dad-10683	126	8	work	work	NOUN
dad-10683	126	9	on	on	ADP
dad-10683	126	10	example	example	NOUN
dad-10683	126	11	-	-	PUNCT
dad-10683	126	12	based	base	VERB
dad-10683	126	13	dialogue	dialogue	NOUN
dad-10683	126	14	modeling	modeling	NOUN
dad-10683	126	15	(	(	PUNCT
dad-10683	126	16	lee	lee	PROPN
dad-10683	126	17	et	et	PROPN
dad-10683	126	18	al	al	PROPN
dad-10683	126	19	.	.	PROPN
dad-10683	126	20	,	,	PUNCT
dad-10683	126	21	2009	2009	NUM
dad-10683	126	22	)	)	PUNCT
dad-10683	126	23	.	.	PUNCT
dad-10683	127	1	there	there	PRON
dad-10683	127	2	has	have	AUX
dad-10683	127	3	been	be	AUX
dad-10683	127	4	some	some	DET
dad-10683	127	5	work	work	NOUN
dad-10683	127	6	on	on	ADP
dad-10683	127	7	combining	combine	VERB
dad-10683	127	8	end	end	NOUN
dad-10683	127	9	-	-	PUNCT
dad-10683	127	10	to	to	ADP
dad-10683	127	11	-	-	PUNCT
dad-10683	127	12	end	end	NOUN
dad-10683	127	13	dialogue	dialogue	NOUN
dad-10683	127	14	models	model	NOUN
dad-10683	127	15	with	with	ADP
dad-10683	127	16	auxiliary	auxiliary	ADJ
dad-10683	127	17	information	information	NOUN
dad-10683	127	18	regarding	regard	VERB
dad-10683	127	19	the	the	DET
dad-10683	127	20	persona	persona	NOUN
dad-10683	127	21	or	or	CCONJ
dad-10683	127	22	participant	participant	ADJ
dad-10683	127	23	role	role	NOUN
dad-10683	127	24	of	of	ADP
dad-10683	127	25	each	each	DET
dad-10683	127	26	person	person	NOUN
dad-10683	127	27	in	in	ADP
dad-10683	127	28	the	the	DET
dad-10683	127	29	dialogue	dialogue	NOUN
dad-10683	127	30	.	.	PUNCT
dad-10683	128	1	luan	luan	PROPN
dad-10683	128	2	et	et	PROPN
dad-10683	128	3	al	al	PROPN
dad-10683	128	4	.	.	PROPN
dad-10683	129	1	(	(	PUNCT
dad-10683	129	2	2016	2016	NUM
dad-10683	129	3	)	)	PUNCT
dad-10683	129	4	investigate	investigate	VERB
dad-10683	129	5	several	several	ADJ
dad-10683	129	6	models	model	NOUN
dad-10683	129	7	that	that	PRON
dad-10683	129	8	incorporate	incorporate	VERB
dad-10683	129	9	participant	participant	ADJ
dad-10683	129	10	roles	role	NOUN
dad-10683	129	11	,	,	PUNCT
dad-10683	129	12	using	use	VERB
dad-10683	129	13	topic	topic	NOUN
dad-10683	129	14	-	-	PUNCT
dad-10683	129	15	modelling	modelling	NOUN
dad-10683	129	16	based	base	VERB
dad-10683	129	17	approaches	approach	NOUN
dad-10683	129	18	with	with	ADP
dad-10683	129	19	lda	lda	PROPN
dad-10683	129	20	.	.	PUNCT
dad-10683	130	1	li	li	PROPN
dad-10683	130	2	et	et	PROPN
dad-10683	130	3	al	al	PROPN
dad-10683	130	4	.	.	PROPN
dad-10683	130	5	(	(	PUNCT
dad-10683	130	6	2016a	2016a	NUM
dad-10683	130	7	)	)	PUNCT
dad-10683	130	8	use	use	VERB
dad-10683	130	9	an	an	DET
dad-10683	130	10	embedding	embedding	NOUN
dad-10683	130	11	for	for	ADP
dad-10683	130	12	each	each	DET
dad-10683	130	13	separate	separate	ADJ
dad-10683	130	14	speaker	speaker	NOUN
dad-10683	130	15	in	in	ADP
dad-10683	130	16	the	the	DET
dad-10683	130	17	conversation	conversation	NOUN
dad-10683	130	18	,	,	PUNCT
dad-10683	130	19	which	which	PRON
dad-10683	130	20	is	be	AUX
dad-10683	130	21	used	use	VERB
dad-10683	130	22	to	to	PART
dad-10683	130	23	condition	condition	VERB
dad-10683	130	24	the	the	DET
dad-10683	130	25	decoder	decoder	NOUN
dad-10683	130	26	in	in	ADP
dad-10683	130	27	an	an	DET
dad-10683	130	28	lstm	lstm	NOUN
dad-10683	130	29	model	model	NOUN
dad-10683	130	30	.	.	PUNCT
dad-10683	131	1	they	they	PRON
dad-10683	131	2	achieve	achieve	VERB
dad-10683	131	3	improvements	improvement	NOUN
dad-10683	131	4	in	in	ADP
dad-10683	131	5	both	both	DET
dad-10683	131	6	perplexity	perplexity	NOUN
dad-10683	131	7	and	and	CCONJ
dad-10683	131	8	bleu	bleu	NOUN
dad-10683	131	9	on	on	ADP
dad-10683	131	10	a	a	DET
dad-10683	131	11	twitter	twitter	NOUN
dad-10683	131	12	dataset	dataset	VERB
dad-10683	131	13	.	.	PUNCT
dad-10683	132	1	36	36	NUM
dad-10683	132	2	training	training	NOUN
dad-10683	132	3	end	end	NOUN
dad-10683	132	4	-	-	PUNCT
dad-10683	132	5	to	to	ADP
dad-10683	132	6	-	-	PUNCT
dad-10683	132	7	end	end	NOUN
dad-10683	132	8	dialogue	dialogue	NOUN
dad-10683	132	9	systems	system	NOUN
dad-10683	132	10	there	there	ADV
dad-10683	132	11	has	have	AUX
dad-10683	132	12	also	also	ADV
dad-10683	132	13	been	be	AUX
dad-10683	132	14	interesting	interesting	ADJ
dad-10683	132	15	work	work	NOUN
dad-10683	132	16	using	use	VERB
dad-10683	132	17	deep	deep	ADJ
dad-10683	132	18	reinforcement	reinforcement	NOUN
dad-10683	132	19	learning	learning	NOUN
dad-10683	132	20	for	for	ADP
dad-10683	132	21	end	end	NOUN
dad-10683	132	22	-	-	PUNCT
dad-10683	132	23	to	to	ADP
dad-10683	132	24	-	-	PUNCT
dad-10683	132	25	end	end	NOUN
dad-10683	132	26	dialogue	dialogue	NOUN
dad-10683	132	27	generation	generation	NOUN
dad-10683	132	28	.	.	PUNCT
dad-10683	133	1	li	li	PROPN
dad-10683	133	2	et	et	PROPN
dad-10683	133	3	al	al	PROPN
dad-10683	133	4	.	.	PROPN
dad-10683	133	5	(	(	PUNCT
dad-10683	133	6	2016b	2016b	NUM
dad-10683	133	7	)	)	PUNCT
dad-10683	133	8	propose	propose	VERB
dad-10683	133	9	using	use	VERB
dad-10683	133	10	a	a	DET
dad-10683	133	11	deep	deep	ADJ
dad-10683	133	12	q	q	NOUN
dad-10683	133	13	-	-	PUNCT
dad-10683	133	14	network	network	NOUN
dad-10683	133	15	(	(	PUNCT
dad-10683	133	16	dqn	dqn	PROPN
dad-10683	133	17	)	)	PUNCT
dad-10683	133	18	for	for	ADP
dad-10683	133	19	dialogue	dialogue	NOUN
dad-10683	133	20	generation	generation	NOUN
dad-10683	133	21	,	,	PUNCT
dad-10683	133	22	using	use	VERB
dad-10683	133	23	a	a	DET
dad-10683	133	24	set	set	NOUN
dad-10683	133	25	of	of	ADP
dad-10683	133	26	reward	reward	NOUN
dad-10683	133	27	functions	function	NOUN
dad-10683	133	28	designed	design	VERB
dad-10683	133	29	to	to	PART
dad-10683	133	30	increase	increase	VERB
dad-10683	133	31	the	the	DET
dad-10683	133	32	diversity	diversity	NOUN
dad-10683	133	33	of	of	ADP
dad-10683	133	34	generated	generate	VERB
dad-10683	133	35	responses	response	NOUN
dad-10683	133	36	.	.	PUNCT
dad-10683	134	1	zhao	zhao	NOUN
dad-10683	134	2	and	and	CCONJ
dad-10683	134	3	eskenazi	eskenazi	ADJ
dad-10683	134	4	(	(	PUNCT
dad-10683	134	5	2016	2016	NUM
dad-10683	134	6	)	)	PUNCT
dad-10683	134	7	similarly	similarly	ADV
dad-10683	134	8	use	use	VERB
dad-10683	134	9	a	a	DET
dad-10683	134	10	deep	deep	ADJ
dad-10683	134	11	recurrent	recurrent	ADJ
dad-10683	134	12	q	q	NOUN
dad-10683	134	13	-	-	PUNCT
dad-10683	134	14	network	network	NOUN
dad-10683	134	15	(	(	PUNCT
dad-10683	134	16	drqn	drqn	NOUN
dad-10683	134	17	)	)	PUNCT
dad-10683	134	18	to	to	PART
dad-10683	134	19	replace	replace	VERB
dad-10683	134	20	the	the	DET
dad-10683	134	21	conventional	conventional	ADJ
dad-10683	134	22	nlu	nlu	NOUN
dad-10683	134	23	,	,	PUNCT
dad-10683	134	24	state	state	NOUN
dad-10683	134	25	tracking	tracking	NOUN
dad-10683	134	26	,	,	PUNCT
dad-10683	134	27	and	and	CCONJ
dad-10683	134	28	dialogue	dialogue	NOUN
dad-10683	134	29	policy	policy	NOUN
dad-10683	134	30	modules	module	NOUN
dad-10683	134	31	for	for	ADP
dad-10683	134	32	task	task	NOUN
dad-10683	134	33	-	-	PUNCT
dad-10683	134	34	oriented	orient	VERB
dad-10683	134	35	dialogue	dialogue	NOUN
dad-10683	134	36	.	.	PUNCT
dad-10683	135	1	one	one	NUM
dad-10683	135	2	of	of	ADP
dad-10683	135	3	the	the	DET
dad-10683	135	4	most	most	ADV
dad-10683	135	5	effective	effective	ADJ
dad-10683	135	6	task	task	NOUN
dad-10683	135	7	-	-	PUNCT
dad-10683	135	8	oriented	orient	VERB
dad-10683	135	9	end	end	NOUN
dad-10683	135	10	-	-	PUNCT
dad-10683	135	11	to	to	ADP
dad-10683	135	12	-	-	PUNCT
dad-10683	135	13	end	end	NOUN
dad-10683	135	14	systems	system	NOUN
dad-10683	135	15	is	be	AUX
dad-10683	135	16	presented	present	VERB
dad-10683	135	17	by	by	ADP
dad-10683	135	18	wen	wen	PROPN
dad-10683	135	19	et	et	PROPN
dad-10683	135	20	al	al	PROPN
dad-10683	135	21	.	.	PROPN
dad-10683	136	1	(	(	PUNCT
dad-10683	136	2	2016	2016	NUM
dad-10683	136	3	)	)	PUNCT
dad-10683	136	4	,	,	PUNCT
dad-10683	136	5	who	who	PRON
dad-10683	136	6	train	train	VERB
dad-10683	136	7	an	an	DET
dad-10683	136	8	end	end	NOUN
dad-10683	136	9	-	-	PUNCT
dad-10683	136	10	to	to	ADP
dad-10683	136	11	-	-	PUNCT
dad-10683	136	12	end	end	NOUN
dad-10683	136	13	system	system	NOUN
dad-10683	136	14	on	on	ADP
dad-10683	136	15	a	a	DET
dad-10683	136	16	small	small	ADJ
dad-10683	136	17	dataset	dataset	NOUN
dad-10683	136	18	of	of	ADP
dad-10683	136	19	restaurant	restaurant	NOUN
dad-10683	136	20	recommendations	recommendation	NOUN
dad-10683	136	21	.	.	PUNCT
dad-10683	137	1	they	they	PRON
dad-10683	137	2	show	show	VERB
dad-10683	137	3	that	that	SCONJ
dad-10683	137	4	they	they	PRON
dad-10683	137	5	are	be	AUX
dad-10683	137	6	able	able	ADJ
dad-10683	137	7	to	to	PART
dad-10683	137	8	achieve	achieve	VERB
dad-10683	137	9	a	a	DET
dad-10683	137	10	higher	high	ADJ
dad-10683	137	11	task	task	NOUN
dad-10683	137	12	completion	completion	NOUN
dad-10683	137	13	rate	rate	NOUN
dad-10683	137	14	than	than	ADP
dad-10683	137	15	a	a	DET
dad-10683	137	16	modular	modular	ADJ
dad-10683	137	17	baseline	baseline	NOUN
dad-10683	137	18	,	,	PUNCT
dad-10683	137	19	and	and	CCONJ
dad-10683	137	20	have	have	VERB
dad-10683	137	21	significantly	significantly	ADV
dad-10683	137	22	higher	high	ADJ
dad-10683	137	23	scores	score	NOUN
dad-10683	137	24	in	in	ADP
dad-10683	137	25	naturalness	naturalness	NOUN
dad-10683	137	26	,	,	PUNCT
dad-10683	137	27	comprehension	comprehension	NOUN
dad-10683	137	28	,	,	PUNCT
dad-10683	137	29	preference	preference	NOUN
dad-10683	137	30	,	,	PUNCT
dad-10683	137	31	and	and	CCONJ
dad-10683	137	32	performance	performance	NOUN
dad-10683	137	33	.	.	PUNCT
dad-10683	138	1	overall	overall	ADV
dad-10683	138	2	,	,	PUNCT
dad-10683	138	3	these	these	DET
dad-10683	138	4	models	model	NOUN
dad-10683	138	5	highlight	highlight	VERB
dad-10683	138	6	the	the	DET
dad-10683	138	7	potential	potential	NOUN
dad-10683	138	8	of	of	ADP
dad-10683	138	9	end	end	NOUN
dad-10683	138	10	-	-	PUNCT
dad-10683	138	11	to	to	ADP
dad-10683	138	12	-	-	PUNCT
dad-10683	138	13	end	end	NOUN
dad-10683	138	14	learning	learning	NOUN
dad-10683	138	15	architectures	architecture	NOUN
dad-10683	138	16	for	for	ADP
dad-10683	138	17	interactive	interactive	ADJ
dad-10683	138	18	systems	system	NOUN
dad-10683	138	19	.	.	PUNCT
dad-10683	139	1	however	however	ADV
dad-10683	139	2	,	,	PUNCT
dad-10683	139	3	much	much	ADJ
dad-10683	139	4	work	work	NOUN
dad-10683	139	5	remains	remain	VERB
dad-10683	139	6	before	before	SCONJ
dad-10683	139	7	these	these	PRON
dad-10683	139	8	can	can	AUX
dad-10683	139	9	be	be	AUX
dad-10683	139	10	implemented	implement	VERB
dad-10683	139	11	with	with	ADP
dad-10683	139	12	confidence	confidence	NOUN
dad-10683	139	13	in	in	ADP
dad-10683	139	14	a	a	DET
dad-10683	139	15	variety	variety	NOUN
dad-10683	139	16	of	of	ADP
dad-10683	139	17	settings	setting	NOUN
dad-10683	139	18	.	.	PUNCT
dad-10683	140	1	3	3	X
dad-10683	140	2	.	.	X
dad-10683	140	3	the	the	DET
dad-10683	140	4	ubuntu	ubuntu	ADJ
dad-10683	140	5	dialogue	dialogue	NOUN
dad-10683	140	6	corpus	corpus	NOUN
dad-10683	140	7	there	there	PRON
dad-10683	140	8	are	be	VERB
dad-10683	140	9	several	several	ADJ
dad-10683	140	10	factors	factor	NOUN
dad-10683	140	11	that	that	PRON
dad-10683	140	12	motivated	motivate	VERB
dad-10683	140	13	the	the	DET
dad-10683	140	14	creation	creation	NOUN
dad-10683	140	15	of	of	ADP
dad-10683	140	16	the	the	DET
dad-10683	140	17	ubuntu	ubuntu	ADJ
dad-10683	140	18	dialogue	dialogue	NOUN
dad-10683	140	19	corpus	corpus	NOUN
dad-10683	140	20	.	.	PUNCT
dad-10683	141	1	in	in	ADP
dad-10683	141	2	particular	particular	ADJ
dad-10683	141	3	,	,	PUNCT
dad-10683	141	4	there	there	PRON
dad-10683	141	5	was	be	VERB
dad-10683	141	6	a	a	DET
dad-10683	141	7	lack	lack	NOUN
dad-10683	141	8	of	of	ADP
dad-10683	141	9	large	large	ADJ
dad-10683	141	10	,	,	PUNCT
dad-10683	141	11	multi	multi	ADJ
dad-10683	141	12	-	-	NOUN
dad-10683	141	13	turn	turn	ADJ
dad-10683	141	14	,	,	PUNCT
dad-10683	141	15	publicly	publicly	ADV
dad-10683	141	16	available	available	ADJ
dad-10683	141	17	dialogue	dialogue	NOUN
dad-10683	141	18	datasets	dataset	NOUN
dad-10683	141	19	.	.	PUNCT
dad-10683	142	1	in	in	ADP
dad-10683	142	2	addition	addition	NOUN
dad-10683	142	3	to	to	ADP
dad-10683	142	4	providing	provide	VERB
dad-10683	142	5	a	a	DET
dad-10683	142	6	dataset	dataset	NOUN
dad-10683	142	7	that	that	PRON
dad-10683	142	8	satisfied	satisfy	VERB
dad-10683	142	9	these	these	DET
dad-10683	142	10	constraints	constraint	NOUN
dad-10683	142	11	,	,	PUNCT
dad-10683	142	12	we	we	PRON
dad-10683	142	13	wanted	want	VERB
dad-10683	142	14	the	the	DET
dad-10683	142	15	dataset	dataset	NOUN
dad-10683	142	16	to	to	PART
dad-10683	142	17	be	be	AUX
dad-10683	142	18	two	two	NUM
dad-10683	142	19	-	-	PUNCT
dad-10683	142	20	way	way	NOUN
dad-10683	142	21	(	(	PUNCT
dad-10683	142	22	dyadic	dyadic	NOUN
dad-10683	142	23	)	)	PUNCT
dad-10683	142	24	,	,	PUNCT
dad-10683	142	25	as	as	SCONJ
dad-10683	142	26	opposed	oppose	VERB
dad-10683	142	27	to	to	ADP
dad-10683	142	28	multi	multi	ADJ
dad-10683	142	29	-	-	ADJ
dad-10683	142	30	participant	participant	ADJ
dad-10683	142	31	chat	chat	NOUN
dad-10683	142	32	,	,	PUNCT
dad-10683	142	33	and	and	CCONJ
dad-10683	142	34	we	we	PRON
dad-10683	142	35	desired	desire	VERB
dad-10683	142	36	a	a	DET
dad-10683	142	37	task	task	NOUN
dad-10683	142	38	-	-	PUNCT
dad-10683	142	39	specific	specific	ADJ
dad-10683	142	40	domain	domain	NOUN
dad-10683	142	41	.	.	PUNCT
dad-10683	143	1	all	all	PRON
dad-10683	143	2	of	of	ADP
dad-10683	143	3	these	these	DET
dad-10683	143	4	characteristics	characteristic	NOUN
dad-10683	143	5	are	be	AUX
dad-10683	143	6	satisfied	satisfied	ADJ
dad-10683	143	7	by	by	ADP
dad-10683	143	8	the	the	DET
dad-10683	143	9	ubuntu	ubuntu	ADJ
dad-10683	143	10	dialogue	dialogue	NOUN
dad-10683	143	11	corpus	corpus	PROPN
dad-10683	143	12	.	.	PUNCT
dad-10683	144	1	3.1	3.1	NUM
dad-10683	144	2	ubuntu	ubuntu	NOUN
dad-10683	144	3	chat	chat	NOUN
dad-10683	144	4	logs	log	VERB
dad-10683	144	5	the	the	DET
dad-10683	144	6	ubuntu	ubuntu	NOUN
dad-10683	144	7	chat	chat	NOUN
dad-10683	144	8	logs	log	NOUN
dad-10683	144	9	refer	refer	VERB
dad-10683	144	10	to	to	ADP
dad-10683	144	11	a	a	DET
dad-10683	144	12	collection	collection	NOUN
dad-10683	144	13	of	of	ADP
dad-10683	144	14	logs	log	NOUN
dad-10683	144	15	from	from	ADP
dad-10683	144	16	ubuntu	ubuntu	NOUN
dad-10683	144	17	-	-	PUNCT
dad-10683	144	18	related	relate	VERB
dad-10683	144	19	chat	chat	NOUN
dad-10683	144	20	rooms	room	NOUN
dad-10683	144	21	on	on	ADP
dad-10683	144	22	the	the	DET
dad-10683	144	23	freenode	freenode	NOUN
dad-10683	144	24	internet	internet	NOUN
dad-10683	144	25	relay	relay	NOUN
dad-10683	144	26	chat	chat	NOUN
dad-10683	144	27	(	(	PUNCT
dad-10683	144	28	irc	irc	NOUN
dad-10683	144	29	)	)	PUNCT
dad-10683	144	30	network	network	NOUN
dad-10683	144	31	.	.	PUNCT
dad-10683	145	1	this	this	DET
dad-10683	145	2	protocol	protocol	NOUN
dad-10683	145	3	allows	allow	VERB
dad-10683	145	4	for	for	ADP
dad-10683	145	5	real	real	ADJ
dad-10683	145	6	-	-	PUNCT
dad-10683	145	7	time	time	NOUN
dad-10683	145	8	chat	chat	NOUN
dad-10683	145	9	between	between	ADP
dad-10683	145	10	a	a	DET
dad-10683	145	11	large	large	ADJ
dad-10683	145	12	number	number	NOUN
dad-10683	145	13	of	of	ADP
dad-10683	145	14	participants	participant	NOUN
dad-10683	145	15	.	.	PUNCT
dad-10683	146	1	each	each	DET
dad-10683	146	2	chat	chat	NOUN
dad-10683	146	3	room	room	NOUN
dad-10683	146	4	,	,	PUNCT
dad-10683	146	5	or	or	CCONJ
dad-10683	146	6	channel	channel	NOUN
dad-10683	146	7	,	,	PUNCT
dad-10683	146	8	has	have	VERB
dad-10683	146	9	a	a	DET
dad-10683	146	10	particular	particular	ADJ
dad-10683	146	11	topic	topic	NOUN
dad-10683	146	12	,	,	PUNCT
dad-10683	146	13	and	and	CCONJ
dad-10683	146	14	every	every	DET
dad-10683	146	15	channel	channel	NOUN
dad-10683	146	16	participant	participant	NOUN
dad-10683	146	17	can	can	AUX
dad-10683	146	18	see	see	VERB
dad-10683	146	19	all	all	DET
dad-10683	146	20	the	the	DET
dad-10683	146	21	messages	message	NOUN
dad-10683	146	22	posted	post	VERB
dad-10683	146	23	in	in	ADP
dad-10683	146	24	a	a	DET
dad-10683	146	25	given	give	VERB
dad-10683	146	26	channel	channel	NOUN
dad-10683	146	27	.	.	PUNCT
dad-10683	147	1	many	many	ADJ
dad-10683	147	2	of	of	ADP
dad-10683	147	3	these	these	DET
dad-10683	147	4	channels	channel	NOUN
dad-10683	147	5	are	be	AUX
dad-10683	147	6	used	use	VERB
dad-10683	147	7	for	for	ADP
dad-10683	147	8	obtaining	obtain	VERB
dad-10683	147	9	technical	technical	ADJ
dad-10683	147	10	support	support	NOUN
dad-10683	147	11	with	with	ADP
dad-10683	147	12	various	various	ADJ
dad-10683	147	13	ubuntu	ubuntu	ADJ
dad-10683	147	14	issues	issue	NOUN
dad-10683	147	15	.	.	PUNCT
dad-10683	148	1	as	as	SCONJ
dad-10683	148	2	the	the	DET
dad-10683	148	3	contents	content	NOUN
dad-10683	148	4	of	of	ADP
dad-10683	148	5	each	each	DET
dad-10683	148	6	channel	channel	NOUN
dad-10683	148	7	are	be	AUX
dad-10683	148	8	moderated	moderate	VERB
dad-10683	148	9	,	,	PUNCT
dad-10683	148	10	most	most	ADJ
dad-10683	148	11	interactions	interaction	NOUN
dad-10683	148	12	follow	follow	VERB
dad-10683	148	13	a	a	DET
dad-10683	148	14	similar	similar	ADJ
dad-10683	148	15	pattern	pattern	NOUN
dad-10683	148	16	.	.	PUNCT
dad-10683	149	1	a	a	DET
dad-10683	149	2	new	new	ADJ
dad-10683	149	3	user	user	NOUN
dad-10683	149	4	joins	join	VERB
dad-10683	149	5	the	the	DET
dad-10683	149	6	channel	channel	NOUN
dad-10683	149	7	,	,	PUNCT
dad-10683	149	8	and	and	CCONJ
dad-10683	149	9	asks	ask	VERB
dad-10683	149	10	a	a	DET
dad-10683	149	11	general	general	ADJ
dad-10683	149	12	question	question	NOUN
dad-10683	149	13	about	about	ADP
dad-10683	149	14	a	a	DET
dad-10683	149	15	problem	problem	NOUN
dad-10683	149	16	they	they	PRON
dad-10683	149	17	are	be	AUX
dad-10683	149	18	having	have	VERB
dad-10683	149	19	with	with	ADP
dad-10683	149	20	ubuntu	ubuntu	NOUN
dad-10683	149	21	.	.	PUNCT
dad-10683	150	1	then	then	ADV
dad-10683	150	2	,	,	PUNCT
dad-10683	150	3	another	another	DET
dad-10683	150	4	more	more	ADV
dad-10683	150	5	experienced	experienced	ADJ
dad-10683	150	6	user	user	NOUN
dad-10683	150	7	replies	reply	NOUN
dad-10683	150	8	with	with	ADP
dad-10683	150	9	a	a	DET
dad-10683	150	10	potential	potential	ADJ
dad-10683	150	11	solution	solution	NOUN
dad-10683	150	12	,	,	PUNCT
dad-10683	150	13	after	after	ADP
dad-10683	150	14	first	first	ADV
dad-10683	150	15	addressing	address	VERB
dad-10683	150	16	the	the	DET
dad-10683	150	17	‘	'	PUNCT
dad-10683	150	18	username	username	NOUN
dad-10683	150	19	’	'	PUNCT
dad-10683	150	20	of	of	ADP
dad-10683	150	21	the	the	DET
dad-10683	150	22	first	first	ADJ
dad-10683	150	23	user	user	NOUN
dad-10683	150	24	.	.	PUNCT
dad-10683	151	1	this	this	PRON
dad-10683	151	2	is	be	AUX
dad-10683	151	3	called	call	VERB
dad-10683	151	4	a	a	DET
dad-10683	151	5	name	name	NOUN
dad-10683	151	6	mention	mention	NOUN
dad-10683	151	7	(	(	PUNCT
dad-10683	151	8	uthus	uthus	ADJ
dad-10683	151	9	and	and	CCONJ
dad-10683	151	10	aha	aha	INTJ
dad-10683	151	11	,	,	PUNCT
dad-10683	151	12	2013a	2013a	NUM
dad-10683	151	13	)	)	PUNCT
dad-10683	151	14	,	,	PUNCT
dad-10683	151	15	and	and	CCONJ
dad-10683	151	16	is	be	AUX
dad-10683	151	17	done	do	VERB
dad-10683	151	18	to	to	PART
dad-10683	151	19	avoid	avoid	VERB
dad-10683	151	20	confusion	confusion	NOUN
dad-10683	151	21	in	in	ADP
dad-10683	151	22	the	the	DET
dad-10683	151	23	channel	channel	NOUN
dad-10683	151	24	—	—	PUNCT
dad-10683	151	25	at	at	ADP
dad-10683	151	26	any	any	DET
dad-10683	151	27	given	give	VERB
dad-10683	151	28	time	time	NOUN
dad-10683	151	29	during	during	ADP
dad-10683	151	30	the	the	DET
dad-10683	151	31	day	day	NOUN
dad-10683	151	32	,	,	PUNCT
dad-10683	151	33	there	there	PRON
dad-10683	151	34	can	can	AUX
dad-10683	151	35	be	be	AUX
dad-10683	151	36	between	between	ADP
dad-10683	151	37	1	1	NUM
dad-10683	151	38	and	and	CCONJ
dad-10683	151	39	20	20	NUM
dad-10683	151	40	simultaneous	simultaneous	ADJ
dad-10683	151	41	conversations	conversation	NOUN
dad-10683	151	42	happening	happen	VERB
dad-10683	151	43	in	in	ADP
dad-10683	151	44	some	some	DET
dad-10683	151	45	channels	channel	NOUN
dad-10683	151	46	.	.	PUNCT
dad-10683	152	1	in	in	ADP
dad-10683	152	2	the	the	DET
dad-10683	152	3	most	most	ADV
dad-10683	152	4	popular	popular	ADJ
dad-10683	152	5	channels	channel	NOUN
dad-10683	152	6	,	,	PUNCT
dad-10683	152	7	there	there	PRON
dad-10683	152	8	is	be	VERB
dad-10683	152	9	almost	almost	ADV
dad-10683	152	10	never	never	ADV
dad-10683	152	11	a	a	DET
dad-10683	152	12	time	time	NOUN
dad-10683	152	13	when	when	SCONJ
dad-10683	152	14	only	only	ADV
dad-10683	152	15	one	one	NUM
dad-10683	152	16	conversation	conversation	NOUN
dad-10683	152	17	is	be	AUX
dad-10683	152	18	occurring	occur	VERB
dad-10683	152	19	;	;	PUNCT
dad-10683	152	20	this	this	PRON
dad-10683	152	21	renders	render	VERB
dad-10683	152	22	it	it	PRON
dad-10683	152	23	particularly	particularly	ADV
dad-10683	152	24	problematic	problematic	ADJ
dad-10683	152	25	to	to	PART
dad-10683	152	26	extract	extract	VERB
dad-10683	152	27	dyadic	dyadic	ADJ
dad-10683	152	28	dialogues	dialogue	NOUN
dad-10683	152	29	.	.	PUNCT
dad-10683	153	1	a	a	DET
dad-10683	153	2	conversation	conversation	NOUN
dad-10683	153	3	between	between	ADP
dad-10683	153	4	a	a	DET
dad-10683	153	5	pair	pair	NOUN
dad-10683	153	6	of	of	ADP
dad-10683	153	7	users	user	NOUN
dad-10683	153	8	generally	generally	ADV
dad-10683	153	9	stops	stop	VERB
dad-10683	153	10	when	when	SCONJ
dad-10683	153	11	the	the	DET
dad-10683	153	12	problem	problem	NOUN
dad-10683	153	13	has	have	AUX
dad-10683	153	14	been	be	AUX
dad-10683	153	15	solved	solve	VERB
dad-10683	153	16	,	,	PUNCT
dad-10683	153	17	though	though	SCONJ
dad-10683	153	18	some	some	DET
dad-10683	153	19	users	user	NOUN
dad-10683	153	20	occasionally	occasionally	ADV
dad-10683	153	21	continue	continue	VERB
dad-10683	153	22	to	to	PART
dad-10683	153	23	discuss	discuss	VERB
dad-10683	153	24	a	a	DET
dad-10683	153	25	topic	topic	NOUN
dad-10683	153	26	not	not	PART
dad-10683	153	27	related	relate	VERB
dad-10683	153	28	to	to	ADP
dad-10683	153	29	ubuntu	ubuntu	NOUN
dad-10683	153	30	.	.	PUNCT
dad-10683	154	1	despite	despite	SCONJ
dad-10683	154	2	the	the	DET
dad-10683	154	3	nature	nature	NOUN
dad-10683	154	4	of	of	ADP
dad-10683	154	5	the	the	DET
dad-10683	154	6	chat	chat	NOUN
dad-10683	154	7	room	room	NOUN
dad-10683	154	8	being	be	AUX
dad-10683	154	9	a	a	DET
dad-10683	154	10	constant	constant	ADJ
dad-10683	154	11	stream	stream	NOUN
dad-10683	154	12	of	of	ADP
dad-10683	154	13	messages	message	NOUN
dad-10683	154	14	from	from	ADP
dad-10683	154	15	multiple	multiple	ADJ
dad-10683	154	16	users	user	NOUN
dad-10683	154	17	,	,	PUNCT
dad-10683	154	18	it	it	PRON
dad-10683	154	19	is	be	AUX
dad-10683	154	20	through	through	ADP
dad-10683	154	21	the	the	DET
dad-10683	154	22	fairly	fairly	ADV
dad-10683	154	23	rigid	rigid	ADJ
dad-10683	154	24	structure	structure	NOUN
dad-10683	154	25	in	in	ADP
dad-10683	154	26	the	the	DET
dad-10683	154	27	messages	message	NOUN
dad-10683	154	28	that	that	PRON
dad-10683	154	29	we	we	PRON
dad-10683	154	30	can	can	AUX
dad-10683	154	31	extract	extract	VERB
dad-10683	154	32	the	the	DET
dad-10683	154	33	dialogues	dialogue	NOUN
dad-10683	154	34	between	between	ADP
dad-10683	154	35	users	user	NOUN
dad-10683	154	36	.	.	PUNCT
dad-10683	155	1	figures	figure	NOUN
dad-10683	155	2	2	2	NUM
dad-10683	155	3	and	and	CCONJ
dad-10683	155	4	3	3	NUM
dad-10683	155	5	show	show	VERB
dad-10683	155	6	an	an	DET
dad-10683	155	7	example	example	NOUN
dad-10683	155	8	chat	chat	NOUN
dad-10683	155	9	room	room	NOUN
dad-10683	155	10	conversation	conversation	NOUN
dad-10683	155	11	from	from	ADP
dad-10683	155	12	the	the	DET
dad-10683	155	13	#	#	NOUN
dad-10683	155	14	ubuntu	ubuntu	ADJ
dad-10683	155	15	channel	channel	NOUN
dad-10683	155	16	as	as	ADV
dad-10683	155	17	well	well	ADV
dad-10683	155	18	as	as	ADP
dad-10683	155	19	the	the	DET
dad-10683	155	20	extracted	extract	VERB
dad-10683	155	21	dialogues	dialogue	NOUN
dad-10683	155	22	,	,	PUNCT
dad-10683	155	23	which	which	PRON
dad-10683	155	24	illustrates	illustrate	VERB
dad-10683	155	25	how	how	SCONJ
dad-10683	155	26	users	user	NOUN
dad-10683	155	27	usually	usually	ADV
dad-10683	155	28	state	state	VERB
dad-10683	155	29	the	the	DET
dad-10683	155	30	username	username	NOUN
dad-10683	155	31	of	of	ADP
dad-10683	155	32	the	the	DET
dad-10683	155	33	intended	intend	VERB
dad-10683	155	34	message	message	NOUN
dad-10683	155	35	recipient	recipient	NOUN
dad-10683	155	36	before	before	ADP
dad-10683	155	37	writing	write	VERB
dad-10683	155	38	their	their	PRON
dad-10683	155	39	reply	reply	NOUN
dad-10683	155	40	(	(	PUNCT
dad-10683	155	41	we	we	PRON
dad-10683	155	42	refer	refer	VERB
dad-10683	155	43	to	to	ADP
dad-10683	155	44	all	all	DET
dad-10683	155	45	initial	initial	ADJ
dad-10683	155	46	questions	question	NOUN
dad-10683	155	47	and	and	CCONJ
dad-10683	155	48	replies	reply	NOUN
dad-10683	155	49	as	as	ADP
dad-10683	155	50	‘	'	PUNCT
dad-10683	155	51	utterances	utterance	NOUN
dad-10683	155	52	’	'	PUNCT
dad-10683	155	53	)	)	PUNCT
dad-10683	155	54	.	.	PUNCT
dad-10683	156	1	for	for	ADP
dad-10683	156	2	example	example	NOUN
dad-10683	156	3	,	,	PUNCT
dad-10683	156	4	it	it	PRON
dad-10683	156	5	is	be	AUX
dad-10683	156	6	clear	clear	ADJ
dad-10683	156	7	that	that	SCONJ
dad-10683	156	8	users	user	NOUN
dad-10683	156	9	‘	'	PUNCT
dad-10683	156	10	taru	taru	NOUN
dad-10683	156	11	’	'	PUNCT
dad-10683	156	12	and	and	CCONJ
dad-10683	156	13	‘	'	PUNCT
dad-10683	156	14	kuja	kuja	NOUN
dad-10683	156	15	’	'	PUNCT
dad-10683	156	16	are	be	AUX
dad-10683	156	17	engaged	engage	VERB
dad-10683	156	18	in	in	ADP
dad-10683	156	19	a	a	DET
dad-10683	156	20	dialogue	dialogue	NOUN
dad-10683	156	21	,	,	PUNCT
dad-10683	156	22	as	as	SCONJ
dad-10683	156	23	are	be	AUX
dad-10683	156	24	users	user	NOUN
dad-10683	156	25	‘	'	PUNCT
dad-10683	156	26	old	old	ADJ
dad-10683	156	27	’	'	PUNCT
dad-10683	156	28	and	and	CCONJ
dad-10683	156	29	‘	'	PUNCT
dad-10683	156	30	bur[n]er	bur[n]er	NOUN
dad-10683	156	31	’	'	PUNCT
dad-10683	156	32	,	,	PUNCT
dad-10683	156	33	while	while	SCONJ
dad-10683	156	34	user	user	NOUN
dad-10683	156	35	‘	'	PUNCT
dad-10683	156	36	pm	pm	NOUN
dad-10683	156	37	’	'	PUNCT
dad-10683	156	38	is	be	AUX
dad-10683	156	39	asking	ask	VERB
dad-10683	156	40	an	an	DET
dad-10683	156	41	initial	initial	ADJ
dad-10683	156	42	question	question	NOUN
dad-10683	156	43	,	,	PUNCT
dad-10683	156	44	and	and	CCONJ
dad-10683	156	45	‘	'	PUNCT
dad-10683	156	46	livecd	livecd	NOUN
dad-10683	156	47	’	'	PUNCT
dad-10683	156	48	is	be	AUX
dad-10683	156	49	perhaps	perhaps	ADV
dad-10683	156	50	elaborating	elaborate	VERB
dad-10683	156	51	on	on	ADP
dad-10683	156	52	a	a	DET
dad-10683	156	53	previous	previous	ADJ
dad-10683	156	54	comment	comment	NOUN
dad-10683	156	55	.	.	PUNCT
dad-10683	157	1	37	37	NUM
dad-10683	157	2	lowe	lowe	PROPN
dad-10683	157	3	,	,	PUNCT
dad-10683	157	4	pow	pow	PROPN
dad-10683	157	5	,	,	PUNCT
dad-10683	157	6	serban	serban	PROPN
dad-10683	157	7	,	,	PUNCT
dad-10683	157	8	charlinn	charlinn	PROPN
dad-10683	157	9	,	,	PUNCT
dad-10683	157	10	liu	liu	PROPN
dad-10683	157	11	and	and	CCONJ
dad-10683	157	12	pineau	pineau	PROPN
dad-10683	157	13	time	time	PROPN
dad-10683	157	14	user	user	NOUN
dad-10683	157	15	utterance	utterance	VERB
dad-10683	157	16	03:44	03:44	NUM
dad-10683	157	17	old	old	ADJ
dad-10683	157	18	i	i	PRON
dad-10683	157	19	do	do	AUX
dad-10683	157	20	nt	not	PART
dad-10683	157	21	run	run	VERB
dad-10683	157	22	graphical	graphical	ADJ
dad-10683	157	23	ubuntu	ubuntu	NOUN
dad-10683	157	24	,	,	PUNCT
dad-10683	157	25	i	i	PRON
dad-10683	157	26	run	run	VERB
dad-10683	157	27	ubuntu	ubuntu	ADJ
dad-10683	157	28	server	server	NOUN
dad-10683	157	29	.	.	PROPN
dad-10683	157	30	03:45	03:45	NUM
dad-10683	157	31	kuja	kuja	PROPN
dad-10683	157	32	taru	taru	PROPN
dad-10683	157	33	:	:	PUNCT
dad-10683	157	34	haha	haha	NOUN
dad-10683	157	35	sucker	sucker	NOUN
dad-10683	157	36	.	.	PUNCT
dad-10683	158	1	03:45	03:45	NUM
dad-10683	158	2	taru	taru	PROPN
dad-10683	158	3	kuja	kuja	PROPN
dad-10683	158	4	:	:	PUNCT
dad-10683	158	5	?	?	PUNCT
dad-10683	159	1	03:45	03:45	NUM
dad-10683	159	2	bur[n]er	bur[n]er	NOUN
dad-10683	159	3	old	old	ADJ
dad-10683	159	4	:	:	PUNCT
dad-10683	159	5	you	you	PRON
dad-10683	159	6	can	can	AUX
dad-10683	159	7	use	use	VERB
dad-10683	159	8	“	"	PUNCT
dad-10683	159	9	ps	ps	NOUN
dad-10683	159	10	ax	ax	NOUN
dad-10683	159	11	”	"	PUNCT
dad-10683	159	12	and	and	CCONJ
dad-10683	159	13	“	"	PUNCT
dad-10683	159	14	kill	kill	VERB
dad-10683	159	15	(	(	PUNCT
dad-10683	159	16	pid	pid	NOUN
dad-10683	159	17	#	#	NOUN
dad-10683	159	18	)	)	PUNCT
dad-10683	159	19	”	"	PUNCT
dad-10683	159	20	03:45	03:45	NUM
dad-10683	159	21	kuja	kuja	PROPN
dad-10683	159	22	taru	taru	PROPN
dad-10683	159	23	:	:	PUNCT
dad-10683	159	24	anyways	anyways	ADV
dad-10683	159	25	,	,	PUNCT
dad-10683	159	26	you	you	PRON
dad-10683	159	27	made	make	VERB
dad-10683	159	28	the	the	DET
dad-10683	159	29	changes	change	NOUN
dad-10683	159	30	right	right	ADJ
dad-10683	159	31	?	?	PUNCT
dad-10683	160	1	03:45	03:45	NUM
dad-10683	161	1	taru	taru	NOUN
dad-10683	161	2	kuja	kuja	PROPN
dad-10683	161	3	:	:	PUNCT
dad-10683	162	1	yes	yes	INTJ
dad-10683	162	2	.	.	PUNCT
dad-10683	162	3	03:45	03:45	NUM
dad-10683	162	4	livecd	livecd	NOUN
dad-10683	162	5	or	or	CCONJ
dad-10683	162	6	killall	killall	PROPN
dad-10683	162	7	speedlink	speedlink	PROPN
dad-10683	162	8	03:45	03:45	PROPN
dad-10683	162	9	kuja	kuja	PROPN
dad-10683	162	10	taru	taru	PROPN
dad-10683	162	11	:	:	PUNCT
dad-10683	162	12	then	then	ADV
dad-10683	162	13	from	from	ADP
dad-10683	162	14	the	the	DET
dad-10683	162	15	terminal	terminal	ADJ
dad-10683	162	16	type	type	NOUN
dad-10683	162	17	:	:	PUNCT
dad-10683	162	18	sudo	sudo	X
dad-10683	162	19	apt	apt	ADJ
dad-10683	162	20	-	-	PUNCT
dad-10683	162	21	get	get	NOUN
dad-10683	162	22	update	update	NOUN
dad-10683	162	23	03:46	03:46	NUM
dad-10683	162	24	pm	pm	NOUN
dad-10683	162	25	if	if	SCONJ
dad-10683	162	26	i	i	PRON
dad-10683	162	27	install	install	VERB
dad-10683	162	28	the	the	DET
dad-10683	162	29	beta	beta	ADJ
dad-10683	162	30	version	version	NOUN
dad-10683	162	31	,	,	PUNCT
dad-10683	162	32	how	how	SCONJ
dad-10683	162	33	can	can	AUX
dad-10683	162	34	i	i	PRON
dad-10683	162	35	update	update	VERB
dad-10683	162	36	it	it	PRON
dad-10683	162	37	when	when	SCONJ
dad-10683	162	38	the	the	DET
dad-10683	162	39	final	final	ADJ
dad-10683	162	40	version	version	NOUN
dad-10683	162	41	comes	come	VERB
dad-10683	162	42	out	out	ADP
dad-10683	162	43	?	?	PUNCT
dad-10683	163	1	03:46	03:46	NUM
dad-10683	163	2	taru	taru	PROPN
dad-10683	163	3	kuja	kuja	PROPN
dad-10683	163	4	:	:	PUNCT
dad-10683	164	1	i	i	PRON
dad-10683	164	2	did	do	VERB
dad-10683	164	3	.	.	PUNCT
dad-10683	165	1	sender	sender	NOUN
dad-10683	165	2	recipient	recipient	NOUN
dad-10683	165	3	utterance	utterance	VERB
dad-10683	165	4	old	old	ADJ
dad-10683	166	1	i	i	PRON
dad-10683	166	2	do	do	AUX
dad-10683	166	3	nt	not	PART
dad-10683	166	4	run	run	VERB
dad-10683	166	5	graphical	graphical	ADJ
dad-10683	166	6	ubuntu	ubuntu	NOUN
dad-10683	166	7	,	,	PUNCT
dad-10683	166	8	i	i	PRON
dad-10683	166	9	run	run	VERB
dad-10683	166	10	ubuntu	ubuntu	ADJ
dad-10683	166	11	server	server	NOUN
dad-10683	166	12	.	.	PUNCT
dad-10683	167	1	bur[n]er	bur[n]er	PROPN
dad-10683	167	2	old	old	ADJ
dad-10683	167	3	you	you	PRON
dad-10683	167	4	can	can	AUX
dad-10683	167	5	use	use	VERB
dad-10683	167	6	“	"	PUNCT
dad-10683	167	7	ps	ps	NOUN
dad-10683	167	8	ax	ax	NOUN
dad-10683	167	9	”	"	PUNCT
dad-10683	167	10	and	and	CCONJ
dad-10683	167	11	“	"	PUNCT
dad-10683	167	12	kill	kill	VERB
dad-10683	167	13	(	(	PUNCT
dad-10683	167	14	pid	pid	NOUN
dad-10683	167	15	#	#	NOUN
dad-10683	167	16	)	)	PUNCT
dad-10683	167	17	”	"	PUNCT
dad-10683	168	1	kuja	kuja	PROPN
dad-10683	168	2	taru	taru	PROPN
dad-10683	168	3	haha	haha	PROPN
dad-10683	168	4	sucker	sucker	NOUN
dad-10683	168	5	.	.	PUNCT
dad-10683	169	1	taru	taru	PROPN
dad-10683	169	2	kuja	kuja	PROPN
dad-10683	169	3	?	?	PUNCT
dad-10683	170	1	kuja	kuja	PROPN
dad-10683	170	2	taru	taru	PROPN
dad-10683	170	3	anyways	anyways	ADV
dad-10683	170	4	,	,	PUNCT
dad-10683	170	5	you	you	PRON
dad-10683	170	6	made	make	VERB
dad-10683	170	7	the	the	DET
dad-10683	170	8	changes	change	NOUN
dad-10683	170	9	right	right	ADJ
dad-10683	170	10	?	?	PUNCT
dad-10683	171	1	taru	taru	PROPN
dad-10683	171	2	kuja	kuja	PROPN
dad-10683	171	3	yes	yes	INTJ
dad-10683	171	4	.	.	PUNCT
dad-10683	172	1	kuja	kuja	PROPN
dad-10683	172	2	taru	taru	PROPN
dad-10683	172	3	then	then	ADV
dad-10683	172	4	from	from	ADP
dad-10683	172	5	the	the	DET
dad-10683	172	6	terminal	terminal	ADJ
dad-10683	172	7	type	type	NOUN
dad-10683	172	8	:	:	PUNCT
dad-10683	172	9	sudo	sudo	X
dad-10683	172	10	apt	apt	ADJ
dad-10683	172	11	-	-	PUNCT
dad-10683	172	12	get	get	NOUN
dad-10683	172	13	update	update	NOUN
dad-10683	172	14	taru	taru	NOUN
dad-10683	172	15	kuja	kuja	PROPN
dad-10683	172	16	i	i	PRON
dad-10683	172	17	did	do	VERB
dad-10683	172	18	.	.	PUNCT
dad-10683	173	1	figure	figure	VERB
dad-10683	173	2	2	2	NUM
dad-10683	173	3	:	:	PUNCT
dad-10683	173	4	example	example	NOUN
dad-10683	173	5	chat	chat	NOUN
dad-10683	173	6	room	room	NOUN
dad-10683	173	7	conversation	conversation	NOUN
dad-10683	173	8	from	from	ADP
dad-10683	173	9	the	the	DET
dad-10683	173	10	#	#	NOUN
dad-10683	173	11	ubuntu	ubuntu	ADJ
dad-10683	173	12	channel	channel	NOUN
dad-10683	173	13	of	of	ADP
dad-10683	173	14	the	the	DET
dad-10683	173	15	ubuntu	ubuntu	NOUN
dad-10683	173	16	chat	chat	PROPN
dad-10683	173	17	logs	log	NOUN
dad-10683	173	18	(	(	PUNCT
dad-10683	173	19	left	left	ADJ
dad-10683	173	20	)	)	PUNCT
dad-10683	173	21	,	,	PUNCT
dad-10683	173	22	with	with	ADP
dad-10683	173	23	the	the	DET
dad-10683	173	24	disentangled	disentangle	VERB
dad-10683	173	25	conversations	conversation	NOUN
dad-10683	173	26	for	for	ADP
dad-10683	173	27	the	the	DET
dad-10683	173	28	ubuntu	ubuntu	ADJ
dad-10683	173	29	dialogue	dialogue	NOUN
dad-10683	173	30	corpus	corpus	NOUN
dad-10683	173	31	(	(	PUNCT
dad-10683	173	32	right	right	ADJ
dad-10683	173	33	)	)	PUNCT
dad-10683	173	34	.	.	PUNCT
dad-10683	174	1	time	time	NOUN
dad-10683	174	2	user	user	NOUN
dad-10683	174	3	utterance	utterance	NOUN
dad-10683	174	4	[	[	X
dad-10683	174	5	12:21	12:21	NUM
dad-10683	174	6	]	]	PUNCT
dad-10683	174	7	dell	dell	NOUN
dad-10683	174	8	well	well	ADV
dad-10683	174	9	,	,	PUNCT
dad-10683	174	10	can	can	AUX
dad-10683	174	11	i	i	PRON
dad-10683	174	12	move	move	VERB
dad-10683	174	13	the	the	DET
dad-10683	174	14	drives	drive	NOUN
dad-10683	174	15	?	?	PUNCT
dad-10683	175	1	[	[	X
dad-10683	175	2	12:21	12:21	NUM
dad-10683	175	3	]	]	X
dad-10683	175	4	cucho	cucho	PROPN
dad-10683	175	5	dell	dell	PROPN
dad-10683	175	6	:	:	PUNCT
dad-10683	175	7	ah	ah	INTJ
dad-10683	175	8	not	not	PART
dad-10683	175	9	like	like	ADP
dad-10683	175	10	that	that	PRON
dad-10683	175	11	[	[	X
dad-10683	175	12	12:21	12:21	NUM
dad-10683	175	13	]	]	X
dad-10683	175	14	rc	rc	PROPN
dad-10683	175	15	dell	dell	PROPN
dad-10683	175	16	:	:	PUNCT
dad-10683	175	17	you	you	PRON
dad-10683	175	18	ca	can	AUX
dad-10683	175	19	n’t	not	PART
dad-10683	175	20	move	move	VERB
dad-10683	175	21	the	the	DET
dad-10683	175	22	drives	drive	NOUN
dad-10683	175	23	[	[	X
dad-10683	175	24	12:21	12:21	NUM
dad-10683	175	25	]	]	X
dad-10683	175	26	rc	rc	PROPN
dad-10683	175	27	dell	dell	PROPN
dad-10683	175	28	:	:	PUNCT
dad-10683	175	29	definitely	definitely	ADV
dad-10683	175	30	not	not	PART
dad-10683	175	31	[	[	X
dad-10683	175	32	12:21	12:21	NUM
dad-10683	175	33	]	]	PUNCT
dad-10683	175	34	dell	dell	NOUN
dad-10683	175	35	ok	ok	INTJ
dad-10683	176	1	[	[	X
dad-10683	176	2	12:21	12:21	NUM
dad-10683	176	3	]	]	PUNCT
dad-10683	176	4	dell	dell	NOUN
dad-10683	176	5	lol	lol	VERB
dad-10683	177	1	[	[	X
dad-10683	177	2	12:21	12:21	NUM
dad-10683	177	3	]	]	PUNCT
dad-10683	177	4	rc	rc	PROPN
dad-10683	177	5	this	this	PRON
dad-10683	177	6	is	be	AUX
dad-10683	177	7	the	the	DET
dad-10683	177	8	problem	problem	NOUN
dad-10683	177	9	with	with	ADP
dad-10683	177	10	raid	raid	NOUN
dad-10683	177	11	:)	:)	PUNCT
dad-10683	178	1	[	[	X
dad-10683	178	2	12:21	12:21	NUM
dad-10683	178	3	]	]	X
dad-10683	178	4	dell	dell	PROPN
dad-10683	178	5	rc	rc	PROPN
dad-10683	178	6	haha	haha	INTJ
dad-10683	179	1	yeah	yeah	INTJ
dad-10683	180	1	[	[	X
dad-10683	180	2	12:22	12:22	NUM
dad-10683	180	3	]	]	X
dad-10683	180	4	dell	dell	PROPN
dad-10683	180	5	cucho	cucho	PROPN
dad-10683	180	6	,	,	PUNCT
dad-10683	180	7	i	i	PRON
dad-10683	180	8	guess	guess	VERB
dad-10683	180	9	i	i	PRON
dad-10683	180	10	could	could	AUX
dad-10683	180	11	just	just	ADV
dad-10683	180	12	get	get	VERB
dad-10683	180	13	an	an	DET
dad-10683	180	14	enclosure	enclosure	NOUN
dad-10683	180	15	and	and	CCONJ
dad-10683	180	16	copy	copy	VERB
dad-10683	180	17	via	via	ADP
dad-10683	180	18	usb	usb	PRON
dad-10683	180	19	...	...	PUNCT
dad-10683	181	1	[	[	X
dad-10683	181	2	12:22	12:22	NUM
dad-10683	181	3	]	]	X
dad-10683	181	4	cucho	cucho	PROPN
dad-10683	181	5	dell	dell	PROPN
dad-10683	181	6	:	:	PUNCT
dad-10683	181	7	i	i	PRON
dad-10683	181	8	would	would	AUX
dad-10683	181	9	advise	advise	VERB
dad-10683	181	10	you	you	PRON
dad-10683	181	11	to	to	PART
dad-10683	181	12	get	get	VERB
dad-10683	181	13	the	the	DET
dad-10683	181	14	disk	disk	NOUN
dad-10683	181	15	sender	sender	NOUN
dad-10683	181	16	recipient	recipient	NOUN
dad-10683	181	17	utterance	utterance	VERB
dad-10683	181	18	dell	dell	NOUN
dad-10683	181	19	well	well	ADV
dad-10683	181	20	,	,	PUNCT
dad-10683	181	21	can	can	AUX
dad-10683	181	22	i	i	PRON
dad-10683	181	23	move	move	VERB
dad-10683	181	24	the	the	DET
dad-10683	181	25	drives	drive	NOUN
dad-10683	181	26	?	?	PUNCT
dad-10683	182	1	cucho	cucho	ADJ
dad-10683	182	2	dell	dell	PROPN
dad-10683	182	3	ah	ah	INTJ
dad-10683	182	4	not	not	PART
dad-10683	182	5	like	like	ADP
dad-10683	182	6	that	that	DET
dad-10683	182	7	dell	dell	NOUN
dad-10683	183	1	cucho	cucho	PROPN
dad-10683	183	2	i	i	PRON
dad-10683	183	3	guess	guess	VERB
dad-10683	183	4	i	i	PRON
dad-10683	183	5	could	could	AUX
dad-10683	183	6	just	just	ADV
dad-10683	183	7	get	get	VERB
dad-10683	183	8	an	an	DET
dad-10683	183	9	enclosure	enclosure	NOUN
dad-10683	183	10	and	and	CCONJ
dad-10683	183	11	copy	copy	VERB
dad-10683	183	12	via	via	ADP
dad-10683	183	13	usb	usb	PROPN
dad-10683	183	14	cucho	cucho	ADJ
dad-10683	183	15	dell	dell	NOUN
dad-10683	184	1	i	i	PRON
dad-10683	184	2	would	would	AUX
dad-10683	184	3	advise	advise	VERB
dad-10683	184	4	you	you	PRON
dad-10683	184	5	to	to	PART
dad-10683	184	6	get	get	VERB
dad-10683	184	7	the	the	DET
dad-10683	184	8	disk	disk	NOUN
dad-10683	184	9	dell	dell	NOUN
dad-10683	184	10	well	well	ADV
dad-10683	184	11	,	,	PUNCT
dad-10683	184	12	can	can	AUX
dad-10683	184	13	i	i	PRON
dad-10683	184	14	move	move	VERB
dad-10683	184	15	the	the	DET
dad-10683	184	16	drives	drive	NOUN
dad-10683	184	17	?	?	PUNCT
dad-10683	185	1	rc	rc	PROPN
dad-10683	185	2	dell	dell	NOUN
dad-10683	185	3	you	you	PRON
dad-10683	185	4	ca	can	AUX
dad-10683	185	5	n’t	not	PART
dad-10683	185	6	move	move	VERB
dad-10683	185	7	the	the	DET
dad-10683	185	8	drives	drive	NOUN
dad-10683	185	9	.	.	PUNCT
dad-10683	186	1	definitely	definitely	ADV
dad-10683	186	2	not	not	PART
dad-10683	186	3	.	.	PUNCT
dad-10683	187	1	this	this	PRON
dad-10683	187	2	is	be	AUX
dad-10683	187	3	the	the	DET
dad-10683	187	4	problem	problem	NOUN
dad-10683	187	5	with	with	ADP
dad-10683	187	6	raid	raid	NOUN
dad-10683	187	7	:)	:)	INTJ
dad-10683	188	1	dell	dell	PROPN
dad-10683	188	2	rc	rc	PROPN
dad-10683	188	3	haha	haha	INTJ
dad-10683	188	4	yeah	yeah	INTJ
dad-10683	188	5	figure	figure	NOUN
dad-10683	188	6	3	3	NUM
dad-10683	188	7	:	:	PUNCT
dad-10683	188	8	example	example	NOUN
dad-10683	188	9	of	of	ADP
dad-10683	188	10	before	before	ADV
dad-10683	188	11	(	(	PUNCT
dad-10683	188	12	left	left	ADJ
dad-10683	188	13	)	)	PUNCT
dad-10683	188	14	and	and	CCONJ
dad-10683	188	15	after	after	ADP
dad-10683	188	16	(	(	PUNCT
dad-10683	188	17	right	right	INTJ
dad-10683	188	18	)	)	PUNCT
dad-10683	188	19	the	the	DET
dad-10683	188	20	algorithm	algorithm	NOUN
dad-10683	188	21	adds	add	VERB
dad-10683	188	22	and	and	CCONJ
dad-10683	188	23	concatenates	concatenate	VERB
dad-10683	188	24	utterances	utterance	NOUN
dad-10683	188	25	in	in	ADP
dad-10683	188	26	dialogue	dialogue	NOUN
dad-10683	188	27	extraction	extraction	NOUN
dad-10683	188	28	.	.	PUNCT
dad-10683	189	1	since	since	SCONJ
dad-10683	189	2	rc	rc	PROPN
dad-10683	189	3	only	only	ADV
dad-10683	189	4	addresses	address	NOUN
dad-10683	189	5	dell	dell	VERB
dad-10683	189	6	,	,	PUNCT
dad-10683	189	7	all	all	PRON
dad-10683	189	8	of	of	ADP
dad-10683	189	9	his	his	PRON
dad-10683	189	10	utterances	utterance	NOUN
dad-10683	189	11	are	be	AUX
dad-10683	189	12	added	add	VERB
dad-10683	189	13	,	,	PUNCT
dad-10683	189	14	however	however	ADV
dad-10683	189	15	this	this	PRON
dad-10683	189	16	is	be	AUX
dad-10683	189	17	not	not	PART
dad-10683	189	18	done	do	VERB
dad-10683	189	19	for	for	ADP
dad-10683	189	20	dell	dell	NOUN
dad-10683	189	21	as	as	SCONJ
dad-10683	189	22	he	he	PRON
dad-10683	189	23	addresses	address	VERB
dad-10683	189	24	both	both	CCONJ
dad-10683	189	25	rc	rc	PROPN
dad-10683	189	26	and	and	CCONJ
dad-10683	189	27	cucho	cucho	PROPN
dad-10683	189	28	.	.	PUNCT
dad-10683	190	1	3.2	3.2	NUM
dad-10683	190	2	dataset	dataset	NOUN
dad-10683	190	3	creation	creation	NOUN
dad-10683	190	4	in	in	ADP
dad-10683	190	5	order	order	NOUN
dad-10683	190	6	to	to	PART
dad-10683	190	7	create	create	VERB
dad-10683	190	8	the	the	DET
dad-10683	190	9	ubuntu	ubuntu	ADJ
dad-10683	190	10	dialogue	dialogue	NOUN
dad-10683	190	11	corpus	corpus	PROPN
dad-10683	190	12	,	,	PUNCT
dad-10683	190	13	first	first	ADV
dad-10683	190	14	a	a	DET
dad-10683	190	15	method	method	NOUN
dad-10683	190	16	had	have	VERB
dad-10683	190	17	to	to	PART
dad-10683	190	18	be	be	AUX
dad-10683	190	19	devised	devise	VERB
dad-10683	190	20	to	to	PART
dad-10683	190	21	extract	extract	VERB
dad-10683	190	22	dyadic	dyadic	ADJ
dad-10683	190	23	dialogues	dialogue	NOUN
dad-10683	190	24	from	from	ADP
dad-10683	190	25	the	the	DET
dad-10683	190	26	chat	chat	PROPN
dad-10683	190	27	room	room	NOUN
dad-10683	190	28	multi	multi	ADJ
dad-10683	190	29	-	-	ADJ
dad-10683	190	30	party	party	NOUN
dad-10683	190	31	conversations	conversation	NOUN
dad-10683	190	32	.	.	PUNCT
dad-10683	191	1	the	the	DET
dad-10683	191	2	first	first	ADJ
dad-10683	191	3	step	step	NOUN
dad-10683	191	4	was	be	AUX
dad-10683	191	5	to	to	PART
dad-10683	191	6	separate	separate	VERB
dad-10683	191	7	every	every	DET
dad-10683	191	8	message	message	NOUN
dad-10683	191	9	into	into	ADP
dad-10683	191	10	4	4	NUM
dad-10683	191	11	-	-	PUNCT
dad-10683	191	12	tuples	tuple	NOUN
dad-10683	191	13	of	of	ADP
dad-10683	191	14	(	(	PUNCT
dad-10683	191	15	time	time	NOUN
dad-10683	191	16	,	,	PUNCT
dad-10683	191	17	sender	sender	NOUN
dad-10683	191	18	,	,	PUNCT
dad-10683	191	19	recipient	recipient	NOUN
dad-10683	191	20	,	,	PUNCT
dad-10683	191	21	utterance	utterance	ADJ
dad-10683	191	22	)	)	PUNCT
dad-10683	191	23	.	.	PUNCT
dad-10683	192	1	given	give	VERB
dad-10683	192	2	these	these	DET
dad-10683	192	3	4	4	NUM
dad-10683	192	4	-	-	PUNCT
dad-10683	192	5	tuples	tuple	NOUN
dad-10683	192	6	,	,	PUNCT
dad-10683	192	7	it	it	PRON
dad-10683	192	8	is	be	AUX
dad-10683	192	9	straightforward	straightforward	ADJ
dad-10683	192	10	38	38	NUM
dad-10683	192	11	training	training	NOUN
dad-10683	192	12	end	end	NOUN
dad-10683	192	13	-	-	PUNCT
dad-10683	192	14	to	to	ADP
dad-10683	192	15	-	-	PUNCT
dad-10683	192	16	end	end	NOUN
dad-10683	192	17	dialogue	dialogue	NOUN
dad-10683	192	18	systems	system	NOUN
dad-10683	192	19	figure	figure	VERB
dad-10683	192	20	4	4	NUM
dad-10683	192	21	:	:	PUNCT
dad-10683	192	22	plot	plot	NOUN
dad-10683	192	23	of	of	ADP
dad-10683	192	24	number	number	NOUN
dad-10683	192	25	of	of	ADP
dad-10683	192	26	conversations	conversation	NOUN
dad-10683	192	27	with	with	ADP
dad-10683	192	28	a	a	DET
dad-10683	192	29	given	give	VERB
dad-10683	192	30	number	number	NOUN
dad-10683	192	31	of	of	ADP
dad-10683	192	32	turns	turn	NOUN
dad-10683	192	33	.	.	PUNCT
dad-10683	193	1	both	both	DET
dad-10683	193	2	axes	axis	NOUN
dad-10683	193	3	use	use	VERB
dad-10683	193	4	a	a	DET
dad-10683	193	5	log	log	NOUN
dad-10683	193	6	scale	scale	NOUN
dad-10683	193	7	.	.	PUNCT
dad-10683	194	1	to	to	PART
dad-10683	194	2	group	group	VERB
dad-10683	194	3	all	all	DET
dad-10683	194	4	tuples	tuple	NOUN
dad-10683	194	5	where	where	SCONJ
dad-10683	194	6	there	there	PRON
dad-10683	194	7	is	be	VERB
dad-10683	194	8	a	a	DET
dad-10683	194	9	matching	matching	NOUN
dad-10683	194	10	sender	sender	NOUN
dad-10683	194	11	and	and	CCONJ
dad-10683	194	12	recipient	recipient	NOUN
dad-10683	194	13	.	.	PUNCT
dad-10683	195	1	although	although	SCONJ
dad-10683	195	2	it	it	PRON
dad-10683	195	3	is	be	AUX
dad-10683	195	4	easy	easy	ADJ
dad-10683	195	5	to	to	PART
dad-10683	195	6	separate	separate	VERB
dad-10683	195	7	the	the	DET
dad-10683	195	8	time	time	NOUN
dad-10683	195	9	and	and	CCONJ
dad-10683	195	10	the	the	DET
dad-10683	195	11	sender	sender	NOUN
dad-10683	195	12	from	from	ADP
dad-10683	195	13	the	the	DET
dad-10683	195	14	rest	rest	NOUN
dad-10683	195	15	,	,	PUNCT
dad-10683	195	16	finding	find	VERB
dad-10683	195	17	the	the	DET
dad-10683	195	18	intended	intend	VERB
dad-10683	195	19	recipient	recipient	NOUN
dad-10683	195	20	of	of	ADP
dad-10683	195	21	the	the	DET
dad-10683	195	22	message	message	NOUN
dad-10683	195	23	is	be	AUX
dad-10683	195	24	not	not	PART
dad-10683	195	25	always	always	ADV
dad-10683	195	26	trivial	trivial	ADJ
dad-10683	195	27	.	.	PUNCT
dad-10683	196	1	3.2.1	3.2.1	NUM
dad-10683	196	2	recipient	recipient	NOUN
dad-10683	196	3	identification	identification	NOUN
dad-10683	196	4	while	while	SCONJ
dad-10683	196	5	in	in	ADP
dad-10683	196	6	most	most	ADJ
dad-10683	196	7	cases	case	NOUN
dad-10683	196	8	the	the	DET
dad-10683	196	9	recipient	recipient	NOUN
dad-10683	196	10	is	be	AUX
dad-10683	196	11	the	the	DET
dad-10683	196	12	first	first	ADJ
dad-10683	196	13	word	word	NOUN
dad-10683	196	14	of	of	ADP
dad-10683	196	15	the	the	DET
dad-10683	196	16	utterance	utterance	NOUN
dad-10683	196	17	,	,	PUNCT
dad-10683	196	18	it	it	PRON
dad-10683	196	19	is	be	AUX
dad-10683	196	20	sometimes	sometimes	ADV
dad-10683	196	21	located	locate	VERB
dad-10683	196	22	at	at	ADP
dad-10683	196	23	the	the	DET
dad-10683	196	24	end	end	NOUN
dad-10683	196	25	,	,	PUNCT
dad-10683	196	26	or	or	CCONJ
dad-10683	196	27	not	not	PART
dad-10683	196	28	at	at	ADV
dad-10683	196	29	all	all	ADV
dad-10683	196	30	in	in	ADP
dad-10683	196	31	the	the	DET
dad-10683	196	32	case	case	NOUN
dad-10683	196	33	of	of	ADP
dad-10683	196	34	initial	initial	ADJ
dad-10683	196	35	questions	question	NOUN
dad-10683	196	36	.	.	PUNCT
dad-10683	197	1	furthermore	furthermore	ADV
dad-10683	197	2	,	,	PUNCT
dad-10683	197	3	some	some	DET
dad-10683	197	4	users	user	NOUN
dad-10683	197	5	choose	choose	VERB
dad-10683	197	6	names	name	NOUN
dad-10683	197	7	corresponding	correspond	VERB
dad-10683	197	8	to	to	ADP
dad-10683	197	9	common	common	ADJ
dad-10683	197	10	english	english	ADJ
dad-10683	197	11	words	word	NOUN
dad-10683	197	12	,	,	PUNCT
dad-10683	197	13	such	such	ADJ
dad-10683	197	14	as	as	ADP
dad-10683	197	15	‘	'	PUNCT
dad-10683	197	16	the	the	PRON
dad-10683	197	17	’	'	PUNCT
dad-10683	197	18	or	or	CCONJ
dad-10683	197	19	‘	'	PUNCT
dad-10683	197	20	stop	stop	VERB
dad-10683	197	21	’	'	PUNCT
dad-10683	197	22	,	,	PUNCT
dad-10683	197	23	which	which	PRON
dad-10683	197	24	could	could	AUX
dad-10683	197	25	lead	lead	VERB
dad-10683	197	26	to	to	ADP
dad-10683	197	27	many	many	ADJ
dad-10683	197	28	false	false	ADJ
dad-10683	197	29	positives	positive	NOUN
dad-10683	197	30	.	.	PUNCT
dad-10683	198	1	in	in	ADP
dad-10683	198	2	order	order	NOUN
dad-10683	198	3	to	to	PART
dad-10683	198	4	solve	solve	VERB
dad-10683	198	5	this	this	DET
dad-10683	198	6	issue	issue	NOUN
dad-10683	198	7	,	,	PUNCT
dad-10683	198	8	we	we	PRON
dad-10683	198	9	create	create	VERB
dad-10683	198	10	a	a	DET
dad-10683	198	11	dictionary	dictionary	NOUN
dad-10683	198	12	of	of	ADP
dad-10683	198	13	usernames	username	NOUN
dad-10683	198	14	from	from	ADP
dad-10683	198	15	the	the	DET
dad-10683	198	16	current	current	ADJ
dad-10683	198	17	and	and	CCONJ
dad-10683	198	18	previous	previous	ADJ
dad-10683	198	19	days	day	NOUN
dad-10683	198	20	,	,	PUNCT
dad-10683	198	21	and	and	CCONJ
dad-10683	198	22	compare	compare	VERB
dad-10683	198	23	the	the	DET
dad-10683	198	24	first	first	ADJ
dad-10683	198	25	word	word	NOUN
dad-10683	198	26	of	of	ADP
dad-10683	198	27	each	each	DET
dad-10683	198	28	utterance	utterance	NOUN
dad-10683	198	29	to	to	ADP
dad-10683	198	30	its	its	PRON
dad-10683	198	31	entries	entry	NOUN
dad-10683	198	32	.	.	PUNCT
dad-10683	199	1	if	if	SCONJ
dad-10683	199	2	a	a	DET
dad-10683	199	3	match	match	NOUN
dad-10683	199	4	is	be	AUX
dad-10683	199	5	found	find	VERB
dad-10683	199	6	,	,	PUNCT
dad-10683	199	7	and	and	CCONJ
dad-10683	199	8	the	the	DET
dad-10683	199	9	word	word	NOUN
dad-10683	199	10	does	do	AUX
dad-10683	199	11	not	not	PART
dad-10683	199	12	correspond	correspond	VERB
dad-10683	199	13	to	to	ADP
dad-10683	199	14	a	a	DET
dad-10683	199	15	very	very	ADV
dad-10683	199	16	common	common	ADJ
dad-10683	199	17	english	english	ADJ
dad-10683	199	18	word4	word4	PROPN
dad-10683	199	19	,	,	PUNCT
dad-10683	199	20	it	it	PRON
dad-10683	199	21	is	be	AUX
dad-10683	199	22	assumed	assume	VERB
dad-10683	199	23	that	that	SCONJ
dad-10683	199	24	this	this	DET
dad-10683	199	25	user	user	NOUN
dad-10683	199	26	was	be	AUX
dad-10683	199	27	the	the	DET
dad-10683	199	28	intended	intend	VERB
dad-10683	199	29	recipient	recipient	NOUN
dad-10683	199	30	of	of	ADP
dad-10683	199	31	the	the	DET
dad-10683	199	32	message	message	NOUN
dad-10683	199	33	.	.	PUNCT
dad-10683	200	1	if	if	SCONJ
dad-10683	200	2	no	no	DET
dad-10683	200	3	matches	match	NOUN
dad-10683	200	4	are	be	AUX
dad-10683	200	5	found	find	VERB
dad-10683	200	6	,	,	PUNCT
dad-10683	200	7	it	it	PRON
dad-10683	200	8	is	be	AUX
dad-10683	200	9	assumed	assume	VERB
dad-10683	200	10	that	that	SCONJ
dad-10683	200	11	the	the	DET
dad-10683	200	12	message	message	NOUN
dad-10683	200	13	was	be	AUX
dad-10683	200	14	an	an	DET
dad-10683	200	15	initial	initial	ADJ
dad-10683	200	16	question	question	NOUN
dad-10683	200	17	,	,	PUNCT
dad-10683	200	18	and	and	CCONJ
dad-10683	200	19	the	the	DET
dad-10683	200	20	recipient	recipient	NOUN
dad-10683	200	21	value	value	NOUN
dad-10683	200	22	is	be	AUX
dad-10683	200	23	left	leave	VERB
dad-10683	200	24	empty	empty	ADJ
dad-10683	200	25	.	.	PUNCT
dad-10683	201	1	3.2.2	3.2.2	NUM
dad-10683	201	2	utterance	utterance	NOUN
dad-10683	201	3	creation	creation	NOUN
dad-10683	201	4	the	the	DET
dad-10683	201	5	dialogue	dialogue	NOUN
dad-10683	201	6	extraction	extraction	NOUN
dad-10683	201	7	algorithm	algorithm	NOUN
dad-10683	201	8	works	work	VERB
dad-10683	201	9	backwards	backwards	ADV
dad-10683	201	10	from	from	ADP
dad-10683	201	11	the	the	DET
dad-10683	201	12	first	first	ADJ
dad-10683	201	13	response	response	NOUN
dad-10683	201	14	to	to	PART
dad-10683	201	15	find	find	VERB
dad-10683	201	16	the	the	DET
dad-10683	201	17	initial	initial	ADJ
dad-10683	201	18	question	question	NOUN
dad-10683	201	19	that	that	PRON
dad-10683	201	20	was	be	AUX
dad-10683	201	21	replied	reply	VERB
dad-10683	201	22	to	to	ADP
dad-10683	201	23	,	,	PUNCT
dad-10683	201	24	within	within	ADP
dad-10683	201	25	a	a	DET
dad-10683	201	26	time	time	NOUN
dad-10683	201	27	frame	frame	NOUN
dad-10683	201	28	of	of	ADP
dad-10683	201	29	3	3	NUM
dad-10683	201	30	minutes	minute	NOUN
dad-10683	201	31	.	.	PUNCT
dad-10683	202	1	a	a	DET
dad-10683	202	2	first	first	ADJ
dad-10683	202	3	response	response	NOUN
dad-10683	202	4	is	be	AUX
dad-10683	202	5	identified	identify	VERB
dad-10683	202	6	by	by	ADP
dad-10683	202	7	the	the	DET
dad-10683	202	8	presence	presence	NOUN
dad-10683	202	9	of	of	ADP
dad-10683	202	10	a	a	DET
dad-10683	202	11	recipient	recipient	NOUN
dad-10683	202	12	name	name	NOUN
dad-10683	202	13	(	(	PUNCT
dad-10683	202	14	someone	someone	PRON
dad-10683	202	15	from	from	ADP
dad-10683	202	16	the	the	DET
dad-10683	202	17	recent	recent	ADJ
dad-10683	202	18	conversation	conversation	NOUN
dad-10683	202	19	history	history	NOUN
dad-10683	202	20	)	)	PUNCT
dad-10683	202	21	.	.	PUNCT
dad-10683	203	1	the	the	DET
dad-10683	203	2	initial	initial	ADJ
dad-10683	203	3	question	question	NOUN
dad-10683	203	4	is	be	AUX
dad-10683	203	5	identified	identify	VERB
dad-10683	203	6	to	to	PART
dad-10683	203	7	be	be	AUX
dad-10683	203	8	the	the	DET
dad-10683	203	9	most	most	ADV
dad-10683	203	10	recent	recent	ADJ
dad-10683	203	11	utterance	utterance	NOUN
dad-10683	203	12	by	by	ADP
dad-10683	203	13	the	the	DET
dad-10683	203	14	recipient	recipient	NOUN
dad-10683	203	15	identified	identify	VERB
dad-10683	203	16	in	in	ADP
dad-10683	203	17	the	the	DET
dad-10683	203	18	first	first	ADJ
dad-10683	203	19	response	response	NOUN
dad-10683	203	20	.	.	PUNCT
dad-10683	204	1	all	all	DET
dad-10683	204	2	utterances	utterance	NOUN
dad-10683	204	3	that	that	PRON
dad-10683	204	4	do	do	AUX
dad-10683	204	5	not	not	PART
dad-10683	204	6	qualify	qualify	VERB
dad-10683	204	7	as	as	ADP
dad-10683	204	8	a	a	DET
dad-10683	204	9	first	first	ADJ
dad-10683	204	10	response	response	NOUN
dad-10683	204	11	or	or	CCONJ
dad-10683	204	12	an	an	DET
dad-10683	204	13	initial	initial	ADJ
dad-10683	204	14	question	question	NOUN
dad-10683	204	15	are	be	AUX
dad-10683	204	16	discarded	discard	VERB
dad-10683	204	17	;	;	PUNCT
dad-10683	204	18	initial	initial	ADJ
dad-10683	204	19	questions	question	NOUN
dad-10683	204	20	that	that	PRON
dad-10683	204	21	do	do	AUX
dad-10683	204	22	not	not	PART
dad-10683	204	23	generate	generate	VERB
dad-10683	204	24	any	any	DET
dad-10683	204	25	response	response	NOUN
dad-10683	204	26	are	be	AUX
dad-10683	204	27	also	also	ADV
dad-10683	204	28	discarded	discard	VERB
dad-10683	204	29	.	.	PUNCT
dad-10683	205	1	we	we	PRON
dad-10683	205	2	additionally	additionally	ADV
dad-10683	205	3	discard	discard	VERB
dad-10683	205	4	conversations	conversation	NOUN
dad-10683	205	5	longer	long	ADV
dad-10683	205	6	than	than	ADP
dad-10683	205	7	five	five	NUM
dad-10683	205	8	utterances	utterance	NOUN
dad-10683	205	9	where	where	SCONJ
dad-10683	205	10	one	one	NUM
dad-10683	205	11	user	user	NOUN
dad-10683	205	12	says	say	VERB
dad-10683	205	13	more	more	ADJ
dad-10683	205	14	than	than	ADP
dad-10683	205	15	80	80	NUM
dad-10683	205	16	%	%	NOUN
dad-10683	205	17	of	of	ADP
dad-10683	205	18	the	the	DET
dad-10683	205	19	utterances	utterance	NOUN
dad-10683	205	20	,	,	PUNCT
dad-10683	205	21	as	as	SCONJ
dad-10683	205	22	these	these	PRON
dad-10683	205	23	are	be	AUX
dad-10683	205	24	4	4	NUM
dad-10683	205	25	.	.	PUNCT
dad-10683	206	1	we	we	PRON
dad-10683	206	2	use	use	VERB
dad-10683	206	3	the	the	DET
dad-10683	206	4	gnu	gnu	PROPN
dad-10683	206	5	aspell	aspell	PROPN
dad-10683	206	6	spell	spell	PROPN
dad-10683	206	7	checking	checking	PROPN
dad-10683	206	8	dictionary	dictionary	PROPN
dad-10683	206	9	.	.	PUNCT
dad-10683	207	1	39	39	NUM
dad-10683	207	2	lowe	lowe	PROPN
dad-10683	207	3	,	,	PUNCT
dad-10683	207	4	pow	pow	PROPN
dad-10683	207	5	,	,	PUNCT
dad-10683	207	6	serban	serban	PROPN
dad-10683	207	7	,	,	PUNCT
dad-10683	207	8	charlinn	charlinn	PROPN
dad-10683	207	9	,	,	PUNCT
dad-10683	207	10	liu	liu	PROPN
dad-10683	207	11	and	and	CCONJ
dad-10683	207	12	pineau	pineau	PROPN
dad-10683	207	13	#	#	NOUN
dad-10683	207	14	dialogues	dialogue	NOUN
dad-10683	207	15	(	(	PUNCT
dad-10683	207	16	human	human	NOUN
dad-10683	207	17	-	-	PUNCT
dad-10683	207	18	human	human	NOUN
dad-10683	207	19	)	)	PUNCT
dad-10683	207	20	936,000	936,000	NUM
dad-10683	207	21	#	#	NOUN
dad-10683	207	22	utterances	utterance	NOUN
dad-10683	207	23	(	(	PUNCT
dad-10683	207	24	in	in	ADP
dad-10683	207	25	total	total	ADJ
dad-10683	207	26	)	)	PUNCT
dad-10683	207	27	7,100,000	7,100,000	NUM
dad-10683	207	28	#	#	NOUN
dad-10683	207	29	words	word	NOUN
dad-10683	207	30	(	(	PUNCT
dad-10683	207	31	in	in	ADP
dad-10683	207	32	total	total	ADJ
dad-10683	207	33	)	)	PUNCT
dad-10683	207	34	100,000,000	100,000,000	NUM
dad-10683	207	35	min	min	NOUN
dad-10683	207	36	.	.	PROPN
dad-10683	207	37	#	#	NOUN
dad-10683	207	38	turns	turn	VERB
dad-10683	207	39	per	per	ADP
dad-10683	207	40	dialogue	dialogue	NOUN
dad-10683	207	41	3	3	NUM
dad-10683	207	42	avg	avg	NOUN
dad-10683	207	43	.	.	PUNCT
dad-10683	208	1	#	#	NOUN
dad-10683	208	2	turns	turn	VERB
dad-10683	208	3	per	per	ADP
dad-10683	208	4	dialogue	dialogue	NOUN
dad-10683	208	5	7.71	7.71	NUM
dad-10683	208	6	avg	avg	NOUN
dad-10683	208	7	.	.	PUNCT
dad-10683	209	1	#	#	NOUN
dad-10683	209	2	words	word	NOUN
dad-10683	209	3	per	per	ADP
dad-10683	209	4	utterance	utterance	NOUN
dad-10683	209	5	10.34	10.34	NUM
dad-10683	209	6	median	median	ADJ
dad-10683	209	7	conversation	conversation	NOUN
dad-10683	209	8	length	length	NOUN
dad-10683	209	9	(	(	PUNCT
dad-10683	209	10	min	min	NOUN
dad-10683	209	11	)	)	PUNCT
dad-10683	209	12	6	6	NUM
dad-10683	209	13	training	training	NOUN
dad-10683	209	14	set	set	VERB
dad-10683	209	15	dialogues	dialogue	NOUN
dad-10683	209	16	898,000	898,000	NUM
dad-10683	209	17	validation	validation	NOUN
dad-10683	209	18	/	/	SYM
dad-10683	209	19	test	test	NOUN
dad-10683	209	20	set	set	VERB
dad-10683	209	21	dialogues	dialogue	NOUN
dad-10683	209	22	19,000	19,000	NUM
dad-10683	209	23	training	training	NOUN
dad-10683	209	24	set	set	VERB
dad-10683	209	25	examples	example	NOUN
dad-10683	209	26	unspecified	unspecified	ADJ
dad-10683	209	27	table	table	NOUN
dad-10683	209	28	1	1	NUM
dad-10683	209	29	:	:	PUNCT
dad-10683	209	30	properties	property	NOUN
dad-10683	209	31	of	of	ADP
dad-10683	209	32	ubuntu	ubuntu	ADJ
dad-10683	209	33	dialogue	dialogue	NOUN
dad-10683	209	34	corpus	corpus	NOUN
dad-10683	209	35	.	.	PUNCT
dad-10683	210	1	note	note	VERB
dad-10683	210	2	that	that	SCONJ
dad-10683	210	3	any	any	DET
dad-10683	210	4	number	number	NOUN
dad-10683	210	5	of	of	ADP
dad-10683	210	6	training	training	NOUN
dad-10683	210	7	examples	example	NOUN
dad-10683	210	8	can	can	AUX
dad-10683	210	9	be	be	AUX
dad-10683	210	10	specified	specify	VERB
dad-10683	210	11	during	during	ADP
dad-10683	210	12	creation	creation	NOUN
dad-10683	210	13	of	of	ADP
dad-10683	210	14	the	the	DET
dad-10683	210	15	training	training	NOUN
dad-10683	210	16	set	set	NOUN
dad-10683	210	17	.	.	PUNCT
dad-10683	211	1	depending	depend	VERB
dad-10683	211	2	on	on	ADP
dad-10683	211	3	the	the	DET
dad-10683	211	4	desired	desire	VERB
dad-10683	211	5	number	number	NOUN
dad-10683	211	6	of	of	ADP
dad-10683	211	7	examples	example	NOUN
dad-10683	211	8	,	,	PUNCT
dad-10683	211	9	multiple	multiple	ADJ
dad-10683	211	10	passes	pass	NOUN
dad-10683	211	11	are	be	AUX
dad-10683	211	12	made	make	VERB
dad-10683	211	13	through	through	ADP
dad-10683	211	14	the	the	DET
dad-10683	211	15	dataset	dataset	NOUN
dad-10683	211	16	,	,	PUNCT
dad-10683	211	17	where	where	SCONJ
dad-10683	211	18	each	each	PRON
dad-10683	211	19	pass	pass	NOUN
dad-10683	211	20	samples	sample	NOUN
dad-10683	211	21	a	a	DET
dad-10683	211	22	new	new	ADJ
dad-10683	211	23	context	context	NOUN
dad-10683	211	24	stochastically	stochastically	ADV
dad-10683	211	25	from	from	ADP
dad-10683	211	26	each	each	DET
dad-10683	211	27	dialogue	dialogue	NOUN
dad-10683	211	28	.	.	PUNCT
dad-10683	212	1	very	very	ADV
dad-10683	212	2	large	large	ADJ
dad-10683	212	3	training	training	NOUN
dad-10683	212	4	sets	set	NOUN
dad-10683	212	5	are	be	AUX
dad-10683	212	6	possible	possible	ADJ
dad-10683	212	7	,	,	PUNCT
dad-10683	212	8	yet	yet	CCONJ
dad-10683	212	9	they	they	PRON
dad-10683	212	10	will	will	AUX
dad-10683	212	11	have	have	VERB
dad-10683	212	12	overlapping	overlap	VERB
dad-10683	212	13	examples	example	NOUN
dad-10683	212	14	.	.	PUNCT
dad-10683	213	1	typically	typically	ADV
dad-10683	213	2	not	not	PART
dad-10683	213	3	representative	representative	NOUN
dad-10683	213	4	of	of	ADP
dad-10683	213	5	real	real	ADJ
dad-10683	213	6	chat	chat	NOUN
dad-10683	213	7	dialogues	dialogue	NOUN
dad-10683	213	8	.	.	PUNCT
dad-10683	214	1	finally	finally	ADV
dad-10683	214	2	,	,	PUNCT
dad-10683	214	3	we	we	PRON
dad-10683	214	4	consider	consider	VERB
dad-10683	214	5	only	only	ADV
dad-10683	214	6	extracted	extract	VERB
dad-10683	214	7	dialogues	dialogue	NOUN
dad-10683	214	8	that	that	PRON
dad-10683	214	9	consist	consist	VERB
dad-10683	214	10	of	of	ADP
dad-10683	214	11	3	3	NUM
dad-10683	214	12	turns	turn	NOUN
dad-10683	214	13	or	or	CCONJ
dad-10683	214	14	more	more	ADJ
dad-10683	214	15	to	to	PART
dad-10683	214	16	encourage	encourage	VERB
dad-10683	214	17	the	the	DET
dad-10683	214	18	modeling	modeling	NOUN
dad-10683	214	19	of	of	ADP
dad-10683	214	20	longer	long	ADJ
dad-10683	214	21	-	-	PUNCT
dad-10683	214	22	term	term	NOUN
dad-10683	214	23	dependencies	dependency	NOUN
dad-10683	214	24	.	.	PUNCT
dad-10683	215	1	to	to	PART
dad-10683	215	2	alleviate	alleviate	VERB
dad-10683	215	3	the	the	DET
dad-10683	215	4	problem	problem	NOUN
dad-10683	215	5	of	of	ADP
dad-10683	215	6	‘	'	PUNCT
dad-10683	215	7	holes	hole	NOUN
dad-10683	215	8	’	'	PUNCT
dad-10683	215	9	in	in	ADP
dad-10683	215	10	the	the	DET
dad-10683	215	11	dialogue	dialogue	NOUN
dad-10683	215	12	,	,	PUNCT
dad-10683	215	13	where	where	SCONJ
dad-10683	215	14	one	one	NUM
dad-10683	215	15	user	user	NOUN
dad-10683	215	16	does	do	AUX
dad-10683	215	17	not	not	PART
dad-10683	215	18	address	address	VERB
dad-10683	215	19	the	the	DET
dad-10683	215	20	other	other	ADJ
dad-10683	215	21	explicitly	explicitly	ADV
dad-10683	215	22	,	,	PUNCT
dad-10683	215	23	as	as	ADP
dad-10683	215	24	in	in	ADP
dad-10683	215	25	figure	figure	NOUN
dad-10683	215	26	3	3	NUM
dad-10683	215	27	,	,	PUNCT
dad-10683	215	28	we	we	PRON
dad-10683	215	29	check	check	VERB
dad-10683	215	30	whether	whether	SCONJ
dad-10683	215	31	each	each	DET
dad-10683	215	32	user	user	NOUN
dad-10683	215	33	talks	talk	VERB
dad-10683	215	34	to	to	ADP
dad-10683	215	35	someone	someone	PRON
dad-10683	215	36	else	else	ADV
dad-10683	215	37	for	for	ADP
dad-10683	215	38	the	the	DET
dad-10683	215	39	duration	duration	NOUN
dad-10683	215	40	of	of	ADP
dad-10683	215	41	their	their	PRON
dad-10683	215	42	conversation	conversation	NOUN
dad-10683	215	43	.	.	PUNCT
dad-10683	216	1	if	if	SCONJ
dad-10683	216	2	not	not	PART
dad-10683	216	3	,	,	PUNCT
dad-10683	216	4	all	all	DET
dad-10683	216	5	non	non	ADJ
dad-10683	216	6	-	-	ADJ
dad-10683	216	7	addressed	address	VERB
dad-10683	216	8	utterances	utterance	NOUN
dad-10683	216	9	are	be	AUX
dad-10683	216	10	added	add	VERB
dad-10683	216	11	to	to	ADP
dad-10683	216	12	the	the	DET
dad-10683	216	13	dialogue	dialogue	NOUN
dad-10683	216	14	.	.	PUNCT
dad-10683	217	1	an	an	DET
dad-10683	217	2	example	example	NOUN
dad-10683	217	3	conversation	conversation	NOUN
dad-10683	217	4	along	along	ADP
dad-10683	217	5	with	with	ADP
dad-10683	217	6	the	the	DET
dad-10683	217	7	extracted	extract	VERB
dad-10683	217	8	dialogues	dialogue	NOUN
dad-10683	217	9	is	be	AUX
dad-10683	217	10	shown	show	VERB
dad-10683	217	11	in	in	ADP
dad-10683	217	12	figure	figure	NOUN
dad-10683	217	13	3	3	NUM
dad-10683	217	14	.	.	PUNCT
dad-10683	217	15	note	note	VERB
dad-10683	217	16	that	that	SCONJ
dad-10683	217	17	we	we	PRON
dad-10683	217	18	also	also	ADV
dad-10683	217	19	concatenate	concatenate	VERB
dad-10683	217	20	all	all	DET
dad-10683	217	21	consecutive	consecutive	ADJ
dad-10683	217	22	utterances	utterance	NOUN
dad-10683	217	23	from	from	ADP
dad-10683	217	24	a	a	DET
dad-10683	217	25	given	give	VERB
dad-10683	217	26	user	user	NOUN
dad-10683	217	27	.	.	PUNCT
dad-10683	218	1	we	we	PRON
dad-10683	218	2	do	do	AUX
dad-10683	218	3	not	not	PART
dad-10683	218	4	apply	apply	VERB
dad-10683	218	5	any	any	DET
dad-10683	218	6	further	further	ADJ
dad-10683	218	7	pre	pre	ADJ
dad-10683	218	8	-	-	ADJ
dad-10683	218	9	processing	processing	ADJ
dad-10683	218	10	(	(	PUNCT
dad-10683	218	11	e.g.	e.g.	ADV
dad-10683	218	12	tokenization	tokenization	NOUN
dad-10683	218	13	,	,	PUNCT
dad-10683	218	14	stemming	stemming	NOUN
dad-10683	218	15	)	)	PUNCT
dad-10683	218	16	to	to	ADP
dad-10683	218	17	the	the	DET
dad-10683	218	18	data	datum	NOUN
dad-10683	218	19	as	as	SCONJ
dad-10683	218	20	released	release	VERB
dad-10683	218	21	in	in	ADP
dad-10683	218	22	the	the	DET
dad-10683	218	23	ubuntu	ubuntu	ADJ
dad-10683	218	24	dialogue	dialogue	NOUN
dad-10683	218	25	corpus	corpus	NOUN
dad-10683	218	26	.	.	PUNCT
dad-10683	219	1	however	however	ADV
dad-10683	219	2	the	the	DET
dad-10683	219	3	use	use	NOUN
dad-10683	219	4	of	of	ADP
dad-10683	219	5	pre	pre	ADJ
dad-10683	219	6	-	-	ADJ
dad-10683	219	7	processing	processing	NOUN
dad-10683	219	8	is	be	AUX
dad-10683	219	9	standard	standard	ADJ
dad-10683	219	10	for	for	ADP
dad-10683	219	11	most	most	ADJ
dad-10683	219	12	nlp	nlp	NOUN
dad-10683	219	13	systems	system	NOUN
dad-10683	219	14	,	,	PUNCT
dad-10683	219	15	and	and	CCONJ
dad-10683	219	16	was	be	AUX
dad-10683	219	17	also	also	ADV
dad-10683	219	18	used	use	VERB
dad-10683	219	19	in	in	ADP
dad-10683	219	20	our	our	PRON
dad-10683	219	21	analysis	analysis	NOUN
dad-10683	219	22	(	(	PUNCT
dad-10683	219	23	see	see	VERB
dad-10683	219	24	section	section	NOUN
dad-10683	219	25	4	4	NUM
dad-10683	219	26	)	)	PUNCT
dad-10683	219	27	.	.	PUNCT
dad-10683	220	1	3.2.3	3.2.3	NUM
dad-10683	220	2	special	special	ADJ
dad-10683	220	3	cases	case	NOUN
dad-10683	220	4	and	and	CCONJ
dad-10683	220	5	limitations	limitation	NOUN
dad-10683	220	6	it	it	PRON
dad-10683	220	7	is	be	AUX
dad-10683	220	8	often	often	ADV
dad-10683	220	9	the	the	DET
dad-10683	220	10	case	case	NOUN
dad-10683	220	11	that	that	SCONJ
dad-10683	220	12	a	a	DET
dad-10683	220	13	user	user	NOUN
dad-10683	220	14	will	will	AUX
dad-10683	220	15	post	post	VERB
dad-10683	220	16	an	an	DET
dad-10683	220	17	initial	initial	ADJ
dad-10683	220	18	question	question	NOUN
dad-10683	220	19	,	,	PUNCT
dad-10683	220	20	and	and	CCONJ
dad-10683	220	21	multiple	multiple	ADJ
dad-10683	220	22	people	people	NOUN
dad-10683	220	23	will	will	AUX
dad-10683	220	24	respond	respond	VERB
dad-10683	220	25	to	to	ADP
dad-10683	220	26	it	it	PRON
dad-10683	220	27	with	with	ADP
dad-10683	220	28	different	different	ADJ
dad-10683	220	29	answers	answer	NOUN
dad-10683	220	30	.	.	PUNCT
dad-10683	221	1	in	in	ADP
dad-10683	221	2	this	this	DET
dad-10683	221	3	instance	instance	NOUN
dad-10683	221	4	,	,	PUNCT
dad-10683	221	5	each	each	DET
dad-10683	221	6	conversation	conversation	NOUN
dad-10683	221	7	between	between	ADP
dad-10683	221	8	the	the	DET
dad-10683	221	9	first	first	ADJ
dad-10683	221	10	user	user	NOUN
dad-10683	221	11	and	and	CCONJ
dad-10683	221	12	the	the	DET
dad-10683	221	13	user	user	NOUN
dad-10683	221	14	who	who	PRON
dad-10683	221	15	replied	reply	VERB
dad-10683	221	16	is	be	AUX
dad-10683	221	17	treated	treat	VERB
dad-10683	221	18	as	as	ADP
dad-10683	221	19	a	a	DET
dad-10683	221	20	separate	separate	ADJ
dad-10683	221	21	dialogue	dialogue	NOUN
dad-10683	221	22	.	.	PUNCT
dad-10683	222	1	this	this	PRON
dad-10683	222	2	has	have	VERB
dad-10683	222	3	the	the	DET
dad-10683	222	4	unfortunate	unfortunate	ADJ
dad-10683	222	5	side	side	NOUN
dad-10683	222	6	-	-	PUNCT
dad-10683	222	7	effect	effect	NOUN
dad-10683	222	8	of	of	ADP
dad-10683	222	9	having	have	VERB
dad-10683	222	10	the	the	DET
dad-10683	222	11	initial	initial	ADJ
dad-10683	222	12	question	question	NOUN
dad-10683	222	13	appear	appear	VERB
dad-10683	222	14	multiple	multiple	ADJ
dad-10683	222	15	times	time	NOUN
dad-10683	222	16	in	in	ADP
dad-10683	222	17	several	several	ADJ
dad-10683	222	18	dialogues	dialogue	NOUN
dad-10683	222	19	.	.	PUNCT
dad-10683	223	1	however	however	ADV
dad-10683	223	2	the	the	DET
dad-10683	223	3	number	number	NOUN
dad-10683	223	4	of	of	ADP
dad-10683	223	5	such	such	ADJ
dad-10683	223	6	cases	case	NOUN
dad-10683	223	7	is	be	AUX
dad-10683	223	8	sufficiently	sufficiently	ADV
dad-10683	223	9	small	small	ADJ
dad-10683	223	10	compared	compare	VERB
dad-10683	223	11	to	to	ADP
dad-10683	223	12	the	the	DET
dad-10683	223	13	size	size	NOUN
dad-10683	223	14	of	of	ADP
dad-10683	223	15	the	the	DET
dad-10683	223	16	dataset	dataset	NOUN
dad-10683	223	17	.	.	PUNCT
dad-10683	224	1	another	another	DET
dad-10683	224	2	issue	issue	NOUN
dad-10683	224	3	to	to	PART
dad-10683	224	4	note	note	VERB
dad-10683	224	5	is	be	AUX
dad-10683	224	6	that	that	SCONJ
dad-10683	224	7	the	the	DET
dad-10683	224	8	utterance	utterance	NOUN
dad-10683	224	9	posting	posting	NOUN
dad-10683	224	10	time	time	NOUN
dad-10683	224	11	is	be	AUX
dad-10683	224	12	not	not	PART
dad-10683	224	13	considered	consider	VERB
dad-10683	224	14	for	for	ADP
dad-10683	224	15	segmenting	segment	VERB
dad-10683	224	16	conversations	conversation	NOUN
dad-10683	224	17	between	between	ADP
dad-10683	224	18	two	two	NUM
dad-10683	224	19	users	user	NOUN
dad-10683	224	20	.	.	PUNCT
dad-10683	225	1	even	even	ADV
dad-10683	225	2	if	if	SCONJ
dad-10683	225	3	two	two	NUM
dad-10683	225	4	users	user	NOUN
dad-10683	225	5	have	have	VERB
dad-10683	225	6	a	a	DET
dad-10683	225	7	conversation	conversation	NOUN
dad-10683	225	8	that	that	PRON
dad-10683	225	9	spans	span	VERB
dad-10683	225	10	multiple	multiple	ADJ
dad-10683	225	11	hours	hour	NOUN
dad-10683	225	12	,	,	PUNCT
dad-10683	225	13	or	or	CCONJ
dad-10683	225	14	even	even	ADV
dad-10683	225	15	days	day	NOUN
dad-10683	225	16	,	,	PUNCT
dad-10683	225	17	this	this	PRON
dad-10683	225	18	is	be	AUX
dad-10683	225	19	treated	treat	VERB
dad-10683	225	20	as	as	ADP
dad-10683	225	21	a	a	DET
dad-10683	225	22	single	single	ADJ
dad-10683	225	23	dialogue	dialogue	NOUN
dad-10683	225	24	.	.	PUNCT
dad-10683	226	1	however	however	ADV
dad-10683	226	2	,	,	PUNCT
dad-10683	226	3	such	such	ADJ
dad-10683	226	4	dialogues	dialogue	NOUN
dad-10683	226	5	are	be	AUX
dad-10683	226	6	rare	rare	ADJ
dad-10683	226	7	.	.	PUNCT
dad-10683	227	1	we	we	PRON
dad-10683	227	2	include	include	VERB
dad-10683	227	3	the	the	DET
dad-10683	227	4	posting	posting	NOUN
dad-10683	227	5	time	time	NOUN
dad-10683	227	6	in	in	ADP
dad-10683	227	7	the	the	DET
dad-10683	227	8	corpus	corpus	NOUN
dad-10683	227	9	so	so	SCONJ
dad-10683	227	10	that	that	SCONJ
dad-10683	227	11	other	other	ADJ
dad-10683	227	12	researchers	researcher	NOUN
dad-10683	227	13	may	may	AUX
dad-10683	227	14	filter	filter	VERB
dad-10683	227	15	as	as	ADP
dad-10683	227	16	desired	desire	VERB
dad-10683	227	17	.	.	PUNCT
dad-10683	228	1	3.3	3.3	NUM
dad-10683	228	2	dataset	dataset	NOUN
dad-10683	228	3	statistics	statistic	NOUN
dad-10683	228	4	table	table	NOUN
dad-10683	228	5	1	1	NUM
dad-10683	228	6	summarizes	summarize	NOUN
dad-10683	228	7	properties	property	NOUN
dad-10683	228	8	of	of	ADP
dad-10683	228	9	the	the	DET
dad-10683	228	10	ubuntu	ubuntu	ADJ
dad-10683	228	11	dialogue	dialogue	NOUN
dad-10683	228	12	corpus	corpus	NOUN
dad-10683	228	13	.	.	PUNCT
dad-10683	229	1	one	one	NUM
dad-10683	229	2	of	of	ADP
dad-10683	229	3	the	the	DET
dad-10683	229	4	most	most	ADV
dad-10683	229	5	important	important	ADJ
dad-10683	229	6	features	feature	NOUN
dad-10683	229	7	of	of	ADP
dad-10683	229	8	the	the	DET
dad-10683	229	9	ubuntu	ubuntu	NOUN
dad-10683	229	10	chat	chat	NOUN
dad-10683	229	11	logs	log	NOUN
dad-10683	229	12	is	be	AUX
dad-10683	229	13	its	its	PRON
dad-10683	229	14	size	size	NOUN
dad-10683	229	15	.	.	PUNCT
dad-10683	230	1	this	this	PRON
dad-10683	230	2	is	be	AUX
dad-10683	230	3	crucial	crucial	ADJ
dad-10683	230	4	for	for	ADP
dad-10683	230	5	research	research	NOUN
dad-10683	230	6	into	into	ADP
dad-10683	230	7	building	building	NOUN
dad-10683	230	8	dialogue	dialogue	NOUN
dad-10683	230	9	managers	manager	NOUN
dad-10683	230	10	based	base	VERB
dad-10683	230	11	40	40	NUM
dad-10683	230	12	training	training	NOUN
dad-10683	230	13	end	end	NOUN
dad-10683	230	14	-	-	PUNCT
dad-10683	230	15	to	to	ADP
dad-10683	230	16	-	-	PUNCT
dad-10683	230	17	end	end	NOUN
dad-10683	230	18	dialogue	dialogue	NOUN
dad-10683	230	19	systems	system	NOUN
dad-10683	230	20	on	on	ADP
dad-10683	230	21	neural	neural	ADJ
dad-10683	230	22	architectures	architecture	NOUN
dad-10683	230	23	.	.	PUNCT
dad-10683	231	1	another	another	DET
dad-10683	231	2	important	important	ADJ
dad-10683	231	3	characteristic	characteristic	NOUN
dad-10683	231	4	is	be	AUX
dad-10683	231	5	the	the	DET
dad-10683	231	6	number	number	NOUN
dad-10683	231	7	of	of	ADP
dad-10683	231	8	turns	turn	NOUN
dad-10683	231	9	in	in	ADP
dad-10683	231	10	these	these	DET
dad-10683	231	11	dialogues	dialogue	NOUN
dad-10683	231	12	.	.	PUNCT
dad-10683	232	1	the	the	DET
dad-10683	232	2	distribution	distribution	NOUN
dad-10683	232	3	of	of	ADP
dad-10683	232	4	the	the	DET
dad-10683	232	5	number	number	NOUN
dad-10683	232	6	of	of	ADP
dad-10683	232	7	turns	turn	NOUN
dad-10683	232	8	is	be	AUX
dad-10683	232	9	shown	show	VERB
dad-10683	232	10	in	in	ADP
dad-10683	232	11	figure	figure	NOUN
dad-10683	232	12	4	4	NUM
dad-10683	232	13	.	.	PUNCT
dad-10683	233	1	it	it	PRON
dad-10683	233	2	can	can	AUX
dad-10683	233	3	be	be	AUX
dad-10683	233	4	seen	see	VERB
dad-10683	233	5	that	that	SCONJ
dad-10683	233	6	the	the	DET
dad-10683	233	7	number	number	NOUN
dad-10683	233	8	of	of	ADP
dad-10683	233	9	dialogues	dialogue	NOUN
dad-10683	233	10	and	and	CCONJ
dad-10683	233	11	turns	turn	VERB
dad-10683	233	12	per	per	ADP
dad-10683	233	13	dialogue	dialogue	NOUN
dad-10683	233	14	follow	follow	VERB
dad-10683	233	15	an	an	DET
dad-10683	233	16	approximate	approximate	ADJ
dad-10683	233	17	power	power	NOUN
dad-10683	233	18	law	law	NOUN
dad-10683	233	19	relationship	relationship	NOUN
dad-10683	233	20	.	.	PUNCT
dad-10683	234	1	3.4	3.4	NUM
dad-10683	234	2	test	test	NOUN
dad-10683	234	3	set	set	NOUN
dad-10683	234	4	generation	generation	NOUN
dad-10683	234	5	we	we	PRON
dad-10683	234	6	set	set	VERB
dad-10683	234	7	aside	aside	ADV
dad-10683	234	8	2	2	NUM
dad-10683	234	9	%	%	NOUN
dad-10683	234	10	of	of	ADP
dad-10683	234	11	the	the	DET
dad-10683	234	12	ubuntu	ubuntu	ADJ
dad-10683	234	13	dialogue	dialogue	NOUN
dad-10683	234	14	corpus	corpus	NOUN
dad-10683	234	15	conversations	conversation	NOUN
dad-10683	234	16	to	to	PART
dad-10683	234	17	form	form	VERB
dad-10683	234	18	a	a	DET
dad-10683	234	19	test	test	NOUN
dad-10683	234	20	set	set	NOUN
dad-10683	234	21	that	that	PRON
dad-10683	234	22	can	can	AUX
dad-10683	234	23	be	be	AUX
dad-10683	234	24	used	use	VERB
dad-10683	234	25	for	for	ADP
dad-10683	234	26	evaluation	evaluation	NOUN
dad-10683	234	27	of	of	ADP
dad-10683	234	28	response	response	NOUN
dad-10683	234	29	selection	selection	NOUN
dad-10683	234	30	algorithms5	algorithms5	NOUN
dad-10683	234	31	.	.	PUNCT
dad-10683	235	1	compared	compare	VERB
dad-10683	235	2	to	to	ADP
dad-10683	235	3	the	the	DET
dad-10683	235	4	rest	rest	NOUN
dad-10683	235	5	of	of	ADP
dad-10683	235	6	the	the	DET
dad-10683	235	7	corpus	corpus	NOUN
dad-10683	235	8	,	,	PUNCT
dad-10683	235	9	this	this	DET
dad-10683	235	10	test	test	NOUN
dad-10683	235	11	set	set	NOUN
dad-10683	235	12	has	have	AUX
dad-10683	235	13	been	be	AUX
dad-10683	235	14	further	far	ADV
dad-10683	235	15	processed	process	VERB
dad-10683	235	16	to	to	PART
dad-10683	235	17	extract	extract	VERB
dad-10683	235	18	a	a	DET
dad-10683	235	19	pair	pair	NOUN
dad-10683	235	20	of	of	ADP
dad-10683	235	21	(	(	PUNCT
dad-10683	235	22	context	context	NOUN
dad-10683	235	23	,	,	PUNCT
dad-10683	235	24	response	response	NOUN
dad-10683	235	25	,	,	PUNCT
dad-10683	235	26	flag	flag	NOUN
dad-10683	235	27	)	)	PUNCT
dad-10683	235	28	triples	triple	NOUN
dad-10683	235	29	from	from	ADP
dad-10683	235	30	each	each	DET
dad-10683	235	31	dialogue	dialogue	NOUN
dad-10683	235	32	.	.	PUNCT
dad-10683	236	1	the	the	DET
dad-10683	236	2	flag	flag	NOUN
dad-10683	236	3	is	be	AUX
dad-10683	236	4	a	a	DET
dad-10683	236	5	boolean	boolean	ADJ
dad-10683	236	6	variable	variable	NOUN
dad-10683	236	7	indicating	indicate	VERB
dad-10683	236	8	whether	whether	SCONJ
dad-10683	236	9	or	or	CCONJ
dad-10683	236	10	not	not	PART
dad-10683	236	11	the	the	DET
dad-10683	236	12	response	response	NOUN
dad-10683	236	13	was	be	AUX
dad-10683	236	14	the	the	DET
dad-10683	236	15	actual	actual	ADJ
dad-10683	236	16	next	next	ADJ
dad-10683	236	17	utterance	utterance	NOUN
dad-10683	236	18	after	after	ADP
dad-10683	236	19	the	the	DET
dad-10683	236	20	given	give	VERB
dad-10683	236	21	context	context	NOUN
dad-10683	236	22	.	.	PUNCT
dad-10683	237	1	the	the	DET
dad-10683	237	2	response	response	NOUN
dad-10683	237	3	is	be	AUX
dad-10683	237	4	a	a	DET
dad-10683	237	5	target	target	NOUN
dad-10683	237	6	(	(	PUNCT
dad-10683	237	7	output	output	NOUN
dad-10683	237	8	)	)	PUNCT
dad-10683	237	9	utterance	utterance	ADP
dad-10683	237	10	which	which	PRON
dad-10683	237	11	we	we	PRON
dad-10683	237	12	aim	aim	VERB
dad-10683	237	13	to	to	PART
dad-10683	237	14	correctly	correctly	ADV
dad-10683	237	15	identify	identify	VERB
dad-10683	237	16	.	.	PUNCT
dad-10683	238	1	the	the	DET
dad-10683	238	2	context	context	NOUN
dad-10683	238	3	consists	consist	VERB
dad-10683	238	4	of	of	ADP
dad-10683	238	5	the	the	DET
dad-10683	238	6	sequence	sequence	NOUN
dad-10683	238	7	of	of	ADP
dad-10683	238	8	utterances	utterance	NOUN
dad-10683	238	9	appearing	appear	VERB
dad-10683	238	10	in	in	ADP
dad-10683	238	11	the	the	DET
dad-10683	238	12	conversation	conversation	NOUN
dad-10683	238	13	prior	prior	ADV
dad-10683	238	14	to	to	ADP
dad-10683	238	15	the	the	DET
dad-10683	238	16	response	response	NOUN
dad-10683	238	17	.	.	PUNCT
dad-10683	239	1	we	we	PRON
dad-10683	239	2	create	create	VERB
dad-10683	239	3	a	a	DET
dad-10683	239	4	pair	pair	NOUN
dad-10683	239	5	of	of	ADP
dad-10683	239	6	triples	triple	NOUN
dad-10683	239	7	,	,	PUNCT
dad-10683	239	8	where	where	SCONJ
dad-10683	239	9	one	one	NUM
dad-10683	239	10	triple	triple	NOUN
dad-10683	239	11	contains	contain	VERB
dad-10683	239	12	the	the	DET
dad-10683	239	13	correct	correct	ADJ
dad-10683	239	14	response	response	NOUN
dad-10683	239	15	(	(	PUNCT
dad-10683	239	16	i.e.	i.e.	X
dad-10683	239	17	the	the	DET
dad-10683	239	18	actual	actual	ADJ
dad-10683	239	19	next	next	ADJ
dad-10683	239	20	utterance	utterance	NOUN
dad-10683	239	21	in	in	ADP
dad-10683	239	22	the	the	DET
dad-10683	239	23	dialogue	dialogue	NOUN
dad-10683	239	24	)	)	PUNCT
dad-10683	239	25	,	,	PUNCT
dad-10683	239	26	and	and	CCONJ
dad-10683	239	27	the	the	DET
dad-10683	239	28	other	other	ADJ
dad-10683	239	29	triple	triple	NOUN
dad-10683	239	30	contains	contain	VERB
dad-10683	239	31	a	a	DET
dad-10683	239	32	false	false	ADJ
dad-10683	239	33	response	response	NOUN
dad-10683	239	34	,	,	PUNCT
dad-10683	239	35	sampled	sample	VERB
dad-10683	239	36	randomly	randomly	ADV
dad-10683	239	37	from	from	ADP
dad-10683	239	38	elsewhere	elsewhere	ADV
dad-10683	239	39	within	within	ADP
dad-10683	239	40	the	the	DET
dad-10683	239	41	test	test	NOUN
dad-10683	239	42	set	set	NOUN
dad-10683	239	43	.	.	PUNCT
dad-10683	240	1	the	the	DET
dad-10683	240	2	flag	flag	NOUN
dad-10683	240	3	is	be	AUX
dad-10683	240	4	set	set	VERB
dad-10683	240	5	to	to	ADP
dad-10683	240	6	1	1	NUM
dad-10683	240	7	in	in	ADP
dad-10683	240	8	the	the	DET
dad-10683	240	9	first	first	ADJ
dad-10683	240	10	case	case	NOUN
dad-10683	240	11	and	and	CCONJ
dad-10683	240	12	to	to	ADP
dad-10683	240	13	0	0	NUM
dad-10683	240	14	in	in	ADP
dad-10683	240	15	the	the	DET
dad-10683	240	16	second	second	ADJ
dad-10683	240	17	case	case	NOUN
dad-10683	240	18	.	.	PUNCT
dad-10683	241	1	an	an	DET
dad-10683	241	2	example	example	NOUN
dad-10683	241	3	pair	pair	NOUN
dad-10683	241	4	is	be	AUX
dad-10683	241	5	shown	show	VERB
dad-10683	241	6	in	in	ADP
dad-10683	241	7	table	table	NOUN
dad-10683	241	8	2	2	NUM
dad-10683	241	9	.	.	PUNCT
dad-10683	241	10	to	to	PART
dad-10683	241	11	make	make	VERB
dad-10683	241	12	the	the	DET
dad-10683	241	13	task	task	NOUN
dad-10683	241	14	harder	hard	ADV
dad-10683	241	15	,	,	PUNCT
dad-10683	241	16	we	we	PRON
dad-10683	241	17	can	can	AUX
dad-10683	241	18	move	move	VERB
dad-10683	241	19	from	from	ADP
dad-10683	241	20	pairs	pair	NOUN
dad-10683	241	21	of	of	ADP
dad-10683	241	22	responses	response	NOUN
dad-10683	241	23	(	(	PUNCT
dad-10683	241	24	one	one	NUM
dad-10683	241	25	correct	correct	ADJ
dad-10683	241	26	,	,	PUNCT
dad-10683	241	27	one	one	NUM
dad-10683	241	28	incorrect	incorrect	ADJ
dad-10683	241	29	)	)	PUNCT
dad-10683	241	30	to	to	ADP
dad-10683	241	31	a	a	DET
dad-10683	241	32	larger	large	ADJ
dad-10683	241	33	set	set	NOUN
dad-10683	241	34	of	of	ADP
dad-10683	241	35	wrong	wrong	ADJ
dad-10683	241	36	responses	response	NOUN
dad-10683	241	37	(	(	PUNCT
dad-10683	241	38	all	all	PRON
dad-10683	241	39	with	with	ADP
dad-10683	241	40	flag=0	flag=0	NUM
dad-10683	241	41	)	)	PUNCT
dad-10683	241	42	.	.	PUNCT
dad-10683	242	1	in	in	ADP
dad-10683	242	2	our	our	PRON
dad-10683	242	3	experiments	experiment	NOUN
dad-10683	242	4	below	below	ADP
dad-10683	242	5	,	,	PUNCT
dad-10683	242	6	we	we	PRON
dad-10683	242	7	consider	consider	VERB
dad-10683	242	8	both	both	PRON
dad-10683	242	9	the	the	DET
dad-10683	242	10	case	case	NOUN
dad-10683	242	11	of	of	ADP
dad-10683	242	12	1	1	NUM
dad-10683	242	13	wrong	wrong	ADJ
dad-10683	242	14	response	response	NOUN
dad-10683	242	15	and	and	CCONJ
dad-10683	242	16	10	10	NUM
dad-10683	242	17	wrong	wrong	ADJ
dad-10683	242	18	responses	response	NOUN
dad-10683	242	19	.	.	PUNCT
dad-10683	243	1	context	context	NOUN
dad-10683	243	2	response	response	NOUN
dad-10683	243	3	flag	flag	NOUN
dad-10683	243	4	well	well	ADV
dad-10683	243	5	,	,	PUNCT
dad-10683	243	6	can	can	AUX
dad-10683	243	7	i	i	PRON
dad-10683	243	8	move	move	VERB
dad-10683	243	9	the	the	DET
dad-10683	243	10	drives	drive	NOUN
dad-10683	243	11	?	?	PUNCT
dad-10683	244	1	i	i	PRON
dad-10683	244	2	guess	guess	VERB
dad-10683	244	3	i	i	PRON
dad-10683	244	4	could	could	AUX
dad-10683	244	5	just	just	ADV
dad-10683	244	6	1	1	NUM
dad-10683	244	7	eot	eot	NOUN
dad-10683	244	8	ah	ah	INTJ
dad-10683	244	9	not	not	PART
dad-10683	244	10	like	like	ADP
dad-10683	244	11	that	that	PRON
dad-10683	244	12	get	get	VERB
dad-10683	244	13	an	an	DET
dad-10683	244	14	enclosure	enclosure	NOUN
dad-10683	244	15	and	and	CCONJ
dad-10683	244	16	copy	copy	VERB
dad-10683	244	17	via	via	ADP
dad-10683	244	18	usb	usb	PRON
dad-10683	244	19	well	well	ADV
dad-10683	244	20	,	,	PUNCT
dad-10683	244	21	can	can	AUX
dad-10683	244	22	i	i	PRON
dad-10683	244	23	move	move	VERB
dad-10683	244	24	the	the	DET
dad-10683	244	25	drives	drive	NOUN
dad-10683	244	26	?	?	PUNCT
dad-10683	245	1	you	you	PRON
dad-10683	245	2	can	can	AUX
dad-10683	245	3	use	use	VERB
dad-10683	245	4	“	"	PUNCT
dad-10683	245	5	ps	ps	NOUN
dad-10683	245	6	ax	ax	NOUN
dad-10683	245	7	”	"	PUNCT
dad-10683	245	8	0	0	NUM
dad-10683	245	9	eot	eot	PROPN
dad-10683	245	10	ah	ah	INTJ
dad-10683	245	11	not	not	PART
dad-10683	245	12	like	like	ADP
dad-10683	245	13	that	that	PRON
dad-10683	245	14	and	and	CCONJ
dad-10683	245	15	“	"	PUNCT
dad-10683	245	16	kill	kill	VERB
dad-10683	245	17	(	(	PUNCT
dad-10683	245	18	pid	pid	NOUN
dad-10683	245	19	#	#	NOUN
dad-10683	245	20	)	)	PUNCT
dad-10683	245	21	”	"	PUNCT
dad-10683	245	22	table	table	NOUN
dad-10683	245	23	2	2	NUM
dad-10683	245	24	:	:	PUNCT
dad-10683	245	25	test	test	NOUN
dad-10683	245	26	set	set	VERB
dad-10683	245	27	example	example	NOUN
dad-10683	245	28	with	with	ADP
dad-10683	245	29	(	(	PUNCT
dad-10683	245	30	context	context	NOUN
dad-10683	245	31	,	,	PUNCT
dad-10683	245	32	reply	reply	NOUN
dad-10683	245	33	,	,	PUNCT
dad-10683	245	34	flag	flag	NOUN
dad-10683	245	35	)	)	PUNCT
dad-10683	245	36	format	format	NOUN
dad-10683	245	37	.	.	PUNCT
dad-10683	246	1	the	the	DET
dad-10683	246	2	‘	'	PUNCT
dad-10683	246	3	eot	eot	NOUN
dad-10683	246	4	’	'	PUNCT
dad-10683	246	5	tag	tag	NOUN
dad-10683	246	6	is	be	AUX
dad-10683	246	7	used	use	VERB
dad-10683	246	8	to	to	PART
dad-10683	246	9	denote	denote	VERB
dad-10683	246	10	the	the	DET
dad-10683	246	11	end	end	NOUN
dad-10683	246	12	of	of	ADP
dad-10683	246	13	a	a	DET
dad-10683	246	14	user	user	NOUN
dad-10683	246	15	’s	’s	PART
dad-10683	246	16	turn	turn	NOUN
dad-10683	246	17	within	within	ADP
dad-10683	246	18	the	the	DET
dad-10683	246	19	context	context	NOUN
dad-10683	246	20	,	,	PUNCT
dad-10683	246	21	and	and	CCONJ
dad-10683	246	22	the	the	DET
dad-10683	246	23	‘	'	PUNCT
dad-10683	246	24	eou	eou	NOUN
dad-10683	246	25	’	'	PUNCT
dad-10683	246	26	tag	tag	NOUN
dad-10683	246	27	is	be	AUX
dad-10683	246	28	used	use	VERB
dad-10683	246	29	to	to	PART
dad-10683	246	30	denote	denote	VERB
dad-10683	246	31	the	the	DET
dad-10683	246	32	end	end	NOUN
dad-10683	246	33	of	of	ADP
dad-10683	246	34	a	a	DET
dad-10683	246	35	user	user	NOUN
dad-10683	246	36	utterance	utterance	NOUN
dad-10683	246	37	without	without	ADP
dad-10683	246	38	a	a	DET
dad-10683	246	39	change	change	NOUN
dad-10683	246	40	of	of	ADP
dad-10683	246	41	turn	turn	NOUN
dad-10683	246	42	.	.	PUNCT
dad-10683	247	1	since	since	SCONJ
dad-10683	247	2	we	we	PRON
dad-10683	247	3	want	want	VERB
dad-10683	247	4	to	to	PART
dad-10683	247	5	learn	learn	VERB
dad-10683	247	6	to	to	PART
dad-10683	247	7	predict	predict	VERB
dad-10683	247	8	all	all	DET
dad-10683	247	9	parts	part	NOUN
dad-10683	247	10	of	of	ADP
dad-10683	247	11	a	a	DET
dad-10683	247	12	conversation	conversation	NOUN
dad-10683	247	13	,	,	PUNCT
dad-10683	247	14	as	as	SCONJ
dad-10683	247	15	opposed	oppose	VERB
dad-10683	247	16	to	to	ADP
dad-10683	247	17	only	only	ADV
dad-10683	247	18	the	the	DET
dad-10683	247	19	closing	closing	NOUN
dad-10683	247	20	statement	statement	NOUN
dad-10683	247	21	,	,	PUNCT
dad-10683	247	22	we	we	PRON
dad-10683	247	23	consider	consider	VERB
dad-10683	247	24	various	various	ADJ
dad-10683	247	25	portions	portion	NOUN
dad-10683	247	26	of	of	ADP
dad-10683	247	27	context	context	NOUN
dad-10683	247	28	for	for	ADP
dad-10683	247	29	the	the	DET
dad-10683	247	30	conversations	conversation	NOUN
dad-10683	247	31	in	in	ADP
dad-10683	247	32	the	the	DET
dad-10683	247	33	test	test	NOUN
dad-10683	247	34	set	set	NOUN
dad-10683	247	35	.	.	PUNCT
dad-10683	248	1	the	the	DET
dad-10683	248	2	context	context	NOUN
dad-10683	248	3	size	size	NOUN
dad-10683	248	4	is	be	AUX
dad-10683	248	5	determined	determine	VERB
dad-10683	248	6	stochastically	stochastically	ADV
dad-10683	248	7	by	by	ADP
dad-10683	248	8	uniform	uniform	ADJ
dad-10683	248	9	sampling6	sampling6	PROPN
dad-10683	248	10	:	:	PUNCT
dad-10683	249	1	c	c	NOUN
dad-10683	249	2	=	=	NOUN
dad-10683	249	3	∼	∼	NOUN
dad-10683	249	4	unif(2	unif(2	ADJ
dad-10683	249	5	,	,	PUNCT
dad-10683	249	6	t−	t−	PROPN
dad-10683	249	7	1	1	NUM
dad-10683	249	8	)	)	PUNCT
dad-10683	249	9	.	.	PUNCT
dad-10683	250	1	here	here	ADV
dad-10683	250	2	,	,	PUNCT
dad-10683	250	3	parameter	parameter	PROPN
dad-10683	250	4	t	t	PROPN
dad-10683	250	5	is	be	AUX
dad-10683	250	6	the	the	DET
dad-10683	250	7	actual	actual	ADJ
dad-10683	250	8	length	length	NOUN
dad-10683	250	9	of	of	ADP
dad-10683	250	10	that	that	DET
dad-10683	250	11	dialogue	dialogue	NOUN
dad-10683	250	12	(	(	PUNCT
dad-10683	250	13	thus	thus	ADV
dad-10683	250	14	the	the	DET
dad-10683	250	15	constraint	constraint	NOUN
dad-10683	250	16	that	that	PRON
dad-10683	250	17	c	c	PROPN
dad-10683	250	18	≤	≤	PROPN
dad-10683	250	19	t	t	PROPN
dad-10683	250	20	−	−	NOUN
dad-10683	250	21	1	1	NUM
dad-10683	250	22	)	)	PUNCT
dad-10683	250	23	.	.	PUNCT
dad-10683	251	1	in	in	ADP
dad-10683	251	2	practice	practice	NOUN
dad-10683	251	3	,	,	PUNCT
dad-10683	251	4	this	this	PRON
dad-10683	251	5	leads	lead	VERB
dad-10683	251	6	to	to	ADP
dad-10683	251	7	short	short	ADJ
dad-10683	251	8	test	test	NOUN
dad-10683	251	9	dialogues	dialogue	NOUN
dad-10683	251	10	having	have	VERB
dad-10683	251	11	short	short	ADJ
dad-10683	251	12	contexts	context	NOUN
dad-10683	251	13	,	,	PUNCT
dad-10683	251	14	while	while	SCONJ
dad-10683	251	15	longer	long	ADJ
dad-10683	251	16	dialogues	dialogue	NOUN
dad-10683	251	17	are	be	AUX
dad-10683	251	18	often	often	ADV
dad-10683	251	19	broken	break	VERB
dad-10683	251	20	into	into	ADP
dad-10683	251	21	a	a	DET
dad-10683	251	22	combination	combination	NOUN
dad-10683	251	23	of	of	ADP
dad-10683	251	24	short	short	ADJ
dad-10683	251	25	,	,	PUNCT
dad-10683	251	26	medium	medium	ADJ
dad-10683	251	27	,	,	PUNCT
dad-10683	251	28	and	and	CCONJ
dad-10683	251	29	long	long	ADJ
dad-10683	251	30	contexts	contexts	NOUN
dad-10683	251	31	.	.	PUNCT
dad-10683	252	1	5	5	X
dad-10683	252	2	.	.	X
dad-10683	252	3	note	note	VERB
dad-10683	252	4	that	that	SCONJ
dad-10683	252	5	,	,	PUNCT
dad-10683	252	6	contrary	contrary	ADV
dad-10683	252	7	to	to	ADP
dad-10683	252	8	the	the	DET
dad-10683	252	9	original	original	ADJ
dad-10683	252	10	ubuntu	ubuntu	ADJ
dad-10683	252	11	dialogue	dialogue	NOUN
dad-10683	252	12	corpus	corpus	NOUN
dad-10683	252	13	,	,	PUNCT
dad-10683	252	14	the	the	DET
dad-10683	252	15	updated	update	VERB
dad-10683	252	16	version	version	NOUN
dad-10683	252	17	separates	separate	VERB
dad-10683	252	18	the	the	DET
dad-10683	252	19	training	training	NOUN
dad-10683	252	20	,	,	PUNCT
dad-10683	252	21	validation	validation	NOUN
dad-10683	252	22	,	,	PUNCT
dad-10683	252	23	and	and	CCONJ
dad-10683	252	24	test	test	NOUN
dad-10683	252	25	sets	set	NOUN
dad-10683	252	26	by	by	ADP
dad-10683	252	27	time	time	NOUN
dad-10683	252	28	.	.	PUNCT
dad-10683	253	1	that	that	PRON
dad-10683	253	2	is	is	ADV
dad-10683	253	3	,	,	PUNCT
dad-10683	253	4	the	the	DET
dad-10683	253	5	training	training	NOUN
dad-10683	253	6	set	set	NOUN
dad-10683	253	7	consists	consist	VERB
dad-10683	253	8	of	of	ADP
dad-10683	253	9	conversations	conversation	NOUN
dad-10683	253	10	that	that	PRON
dad-10683	253	11	started	start	VERB
dad-10683	253	12	from	from	ADP
dad-10683	253	13	2004	2004	NUM
dad-10683	253	14	to	to	ADP
dad-10683	253	15	approximately	approximately	ADV
dad-10683	253	16	april	april	PROPN
dad-10683	253	17	27	27	NUM
dad-10683	253	18	,	,	PUNCT
dad-10683	253	19	2012	2012	NUM
dad-10683	253	20	;	;	PUNCT
dad-10683	253	21	the	the	DET
dad-10683	253	22	validation	validation	NOUN
dad-10683	253	23	set	set	VERB
dad-10683	253	24	consists	consist	VERB
dad-10683	253	25	of	of	ADP
dad-10683	253	26	dialogues	dialogue	NOUN
dad-10683	253	27	starting	start	VERB
dad-10683	253	28	from	from	ADP
dad-10683	253	29	april	april	PROPN
dad-10683	253	30	27	27	NUM
dad-10683	253	31	to	to	ADP
dad-10683	253	32	august	august	PROPN
dad-10683	253	33	7	7	NUM
dad-10683	253	34	,	,	PUNCT
dad-10683	253	35	2012	2012	NUM
dad-10683	253	36	;	;	PUNCT
dad-10683	253	37	and	and	CCONJ
dad-10683	253	38	the	the	DET
dad-10683	253	39	test	test	NOUN
dad-10683	253	40	set	set	NOUN
dad-10683	253	41	has	have	VERB
dad-10683	253	42	dialogues	dialogue	NOUN
dad-10683	253	43	from	from	ADP
dad-10683	253	44	august	august	PROPN
dad-10683	253	45	7	7	NUM
dad-10683	253	46	to	to	ADP
dad-10683	253	47	december	december	PROPN
dad-10683	253	48	1	1	NUM
dad-10683	253	49	,	,	PUNCT
dad-10683	253	50	2012	2012	NUM
dad-10683	253	51	.	.	PUNCT
dad-10683	254	1	this	this	PRON
dad-10683	254	2	mimics	mimic	VERB
dad-10683	254	3	the	the	DET
dad-10683	254	4	training	training	NOUN
dad-10683	254	5	of	of	ADP
dad-10683	254	6	dialogue	dialogue	NOUN
dad-10683	254	7	systems	system	NOUN
dad-10683	254	8	in	in	ADP
dad-10683	254	9	practice	practice	NOUN
dad-10683	254	10	,	,	PUNCT
dad-10683	254	11	where	where	SCONJ
dad-10683	254	12	we	we	PRON
dad-10683	254	13	only	only	ADV
dad-10683	254	14	have	have	VERB
dad-10683	254	15	access	access	NOUN
dad-10683	254	16	to	to	ADP
dad-10683	254	17	data	datum	NOUN
dad-10683	254	18	in	in	ADP
dad-10683	254	19	the	the	DET
dad-10683	254	20	past	past	NOUN
dad-10683	254	21	,	,	PUNCT
dad-10683	254	22	and	and	CCONJ
dad-10683	254	23	want	want	VERB
dad-10683	254	24	to	to	PART
dad-10683	254	25	answer	answer	VERB
dad-10683	254	26	user	user	NOUN
dad-10683	254	27	queries	query	NOUN
dad-10683	254	28	in	in	ADP
dad-10683	254	29	the	the	DET
dad-10683	254	30	future	future	NOUN
dad-10683	254	31	.	.	PUNCT
dad-10683	255	1	6	6	X
dad-10683	255	2	.	.	X
dad-10683	255	3	note	note	VERB
dad-10683	255	4	that	that	SCONJ
dad-10683	255	5	this	this	PRON
dad-10683	255	6	is	be	AUX
dad-10683	255	7	a	a	DET
dad-10683	255	8	different	different	ADJ
dad-10683	255	9	formula	formula	NOUN
dad-10683	255	10	than	than	ADP
dad-10683	255	11	the	the	DET
dad-10683	255	12	original	original	ADJ
dad-10683	255	13	ubuntu	ubuntu	ADJ
dad-10683	255	14	dialogue	dialogue	NOUN
dad-10683	255	15	corpus	corpus	NOUN
dad-10683	255	16	,	,	PUNCT
dad-10683	255	17	which	which	PRON
dad-10683	255	18	sampled	sample	VERB
dad-10683	255	19	from	from	ADP
dad-10683	255	20	a	a	DET
dad-10683	255	21	decreasing	decrease	VERB
dad-10683	255	22	distribution	distribution	NOUN
dad-10683	255	23	.	.	PUNCT
dad-10683	256	1	the	the	DET
dad-10683	256	2	new	new	ADJ
dad-10683	256	3	formula	formula	NOUN
dad-10683	256	4	is	be	AUX
dad-10683	256	5	simpler	simple	ADJ
dad-10683	256	6	and	and	CCONJ
dad-10683	256	7	leads	lead	VERB
dad-10683	256	8	to	to	ADP
dad-10683	256	9	longer	long	ADV
dad-10683	256	10	sampled	sample	VERB
dad-10683	256	11	contexts	context	NOUN
dad-10683	256	12	,	,	PUNCT
dad-10683	256	13	which	which	PRON
dad-10683	256	14	we	we	PRON
dad-10683	256	15	consider	consider	VERB
dad-10683	256	16	desirable	desirable	ADJ
dad-10683	256	17	.	.	PUNCT
dad-10683	257	1	41	41	NUM
dad-10683	257	2	lowe	lowe	PROPN
dad-10683	257	3	,	,	PUNCT
dad-10683	257	4	pow	pow	PROPN
dad-10683	257	5	,	,	PUNCT
dad-10683	257	6	serban	serban	PROPN
dad-10683	257	7	,	,	PUNCT
dad-10683	257	8	charlinn	charlinn	PROPN
dad-10683	257	9	,	,	PUNCT
dad-10683	257	10	liu	liu	PROPN
dad-10683	257	11	and	and	CCONJ
dad-10683	257	12	pineau	pineau	PROPN
dad-10683	257	13	note	note	VERB
dad-10683	257	14	that	that	SCONJ
dad-10683	257	15	,	,	PUNCT
dad-10683	257	16	except	except	SCONJ
dad-10683	257	17	for	for	ADP
dad-10683	257	18	the	the	DET
dad-10683	257	19	plot	plot	NOUN
dad-10683	257	20	in	in	ADP
dad-10683	257	21	figure	figure	NOUN
dad-10683	257	22	7	7	NUM
dad-10683	257	23	,	,	PUNCT
dad-10683	257	24	all	all	DET
dad-10683	257	25	experiments	experiment	NOUN
dad-10683	257	26	,	,	PUNCT
dad-10683	257	27	results	result	NOUN
dad-10683	257	28	,	,	PUNCT
dad-10683	257	29	and	and	CCONJ
dad-10683	257	30	analysis	analysis	NOUN
dad-10683	257	31	in	in	ADP
dad-10683	257	32	this	this	DET
dad-10683	257	33	paper	paper	NOUN
dad-10683	257	34	will	will	AUX
dad-10683	257	35	refer	refer	VERB
dad-10683	257	36	to	to	ADP
dad-10683	257	37	the	the	DET
dad-10683	257	38	updated	update	VERB
dad-10683	257	39	ubuntu	ubuntu	ADJ
dad-10683	257	40	dialogue	dialogue	NOUN
dad-10683	257	41	corpus	corpus	PROPN
dad-10683	257	42	v2	v2	PROPN
dad-10683	257	43	.	.	PUNCT
dad-10683	258	1	4	4	X
dad-10683	258	2	.	.	X
dad-10683	258	3	response	response	NOUN
dad-10683	258	4	classification	classification	NOUN
dad-10683	258	5	architectures	architecture	NOUN
dad-10683	258	6	to	to	PART
dad-10683	258	7	provide	provide	VERB
dad-10683	258	8	further	further	ADJ
dad-10683	258	9	evidence	evidence	NOUN
dad-10683	258	10	of	of	ADP
dad-10683	258	11	the	the	DET
dad-10683	258	12	value	value	NOUN
dad-10683	258	13	of	of	ADP
dad-10683	258	14	our	our	PRON
dad-10683	258	15	dataset	dataset	NOUN
dad-10683	258	16	for	for	ADP
dad-10683	258	17	research	research	NOUN
dad-10683	258	18	into	into	ADP
dad-10683	258	19	neural	neural	ADJ
dad-10683	258	20	architectures	architecture	NOUN
dad-10683	258	21	for	for	ADP
dad-10683	258	22	dialogue	dialogue	NOUN
dad-10683	258	23	managers	manager	NOUN
dad-10683	258	24	,	,	PUNCT
dad-10683	258	25	we	we	PRON
dad-10683	258	26	provide	provide	VERB
dad-10683	258	27	performance	performance	NOUN
dad-10683	258	28	benchmarks	benchmark	NOUN
dad-10683	258	29	using	use	VERB
dad-10683	258	30	two	two	NUM
dad-10683	258	31	different	different	ADJ
dad-10683	258	32	training	training	NOUN
dad-10683	258	33	and	and	CCONJ
dad-10683	258	34	evaluation	evaluation	NOUN
dad-10683	258	35	criteria	criterion	NOUN
dad-10683	258	36	:	:	PUNCT
dad-10683	258	37	response	response	NOUN
dad-10683	258	38	classification	classification	NOUN
dad-10683	258	39	,	,	PUNCT
dad-10683	258	40	and	and	CCONJ
dad-10683	258	41	response	response	NOUN
dad-10683	258	42	generation	generation	NOUN
dad-10683	258	43	.	.	PUNCT
dad-10683	259	1	we	we	PRON
dad-10683	259	2	first	first	ADV
dad-10683	259	3	consider	consider	VERB
dad-10683	259	4	response	response	NOUN
dad-10683	259	5	classification	classification	NOUN
dad-10683	259	6	architectures	architecture	NOUN
dad-10683	259	7	,	,	PUNCT
dad-10683	259	8	which	which	PRON
dad-10683	259	9	attempt	attempt	VERB
dad-10683	259	10	to	to	PART
dad-10683	259	11	distinguish	distinguish	VERB
dad-10683	259	12	between	between	ADP
dad-10683	259	13	valid	valid	ADJ
dad-10683	259	14	and	and	CCONJ
dad-10683	259	15	invalid	invalid	ADJ
dad-10683	259	16	next	next	ADJ
dad-10683	259	17	responses	response	NOUN
dad-10683	259	18	to	to	ADP
dad-10683	259	19	the	the	DET
dad-10683	259	20	context	context	NOUN
dad-10683	259	21	of	of	ADP
dad-10683	259	22	a	a	DET
dad-10683	259	23	conversation	conversation	NOUN
dad-10683	259	24	.	.	PUNCT
dad-10683	260	1	these	these	PRON
dad-10683	260	2	are	be	AUX
dad-10683	260	3	trained	train	VERB
dad-10683	260	4	on	on	ADP
dad-10683	260	5	the	the	DET
dad-10683	260	6	task	task	NOUN
dad-10683	260	7	of	of	ADP
dad-10683	260	8	best	good	ADJ
dad-10683	260	9	response	response	NOUN
dad-10683	260	10	selection	selection	NOUN
dad-10683	260	11	,	,	PUNCT
dad-10683	260	12	which	which	PRON
dad-10683	260	13	we	we	PRON
dad-10683	260	14	call	call	VERB
dad-10683	260	15	next	next	ADJ
dad-10683	260	16	utterance	utterance	NOUN
dad-10683	260	17	classification	classification	NOUN
dad-10683	260	18	(	(	PUNCT
dad-10683	260	19	nuc	nuc	NOUN
dad-10683	260	20	)	)	PUNCT
dad-10683	260	21	.	.	PUNCT
dad-10683	261	1	this	this	PRON
dad-10683	261	2	can	can	AUX
dad-10683	261	3	be	be	AUX
dad-10683	261	4	achieved	achieve	VERB
dad-10683	261	5	by	by	ADP
dad-10683	261	6	processing	process	VERB
dad-10683	261	7	the	the	DET
dad-10683	261	8	data	datum	NOUN
dad-10683	261	9	as	as	SCONJ
dad-10683	261	10	described	describe	VERB
dad-10683	261	11	in	in	ADP
dad-10683	261	12	section	section	NOUN
dad-10683	261	13	3.4	3.4	NUM
dad-10683	261	14	,	,	PUNCT
dad-10683	261	15	without	without	ADP
dad-10683	261	16	requiring	require	VERB
dad-10683	261	17	any	any	DET
dad-10683	261	18	human	human	ADJ
dad-10683	261	19	labels	label	NOUN
dad-10683	261	20	.	.	PUNCT
dad-10683	262	1	this	this	DET
dad-10683	262	2	classification	classification	NOUN
dad-10683	262	3	task	task	NOUN
dad-10683	262	4	is	be	AUX
dad-10683	262	5	an	an	DET
dad-10683	262	6	adaptation	adaptation	NOUN
dad-10683	262	7	of	of	ADP
dad-10683	262	8	the	the	DET
dad-10683	262	9	recall	recall	NOUN
dad-10683	262	10	and	and	CCONJ
dad-10683	262	11	precision	precision	NOUN
dad-10683	262	12	metrics	metric	NOUN
dad-10683	262	13	previously	previously	ADV
dad-10683	262	14	applied	apply	VERB
dad-10683	262	15	to	to	ADP
dad-10683	262	16	dialogue	dialogue	NOUN
dad-10683	262	17	datasets	dataset	NOUN
dad-10683	262	18	(	(	PUNCT
dad-10683	262	19	schatzmann	schatzmann	X
dad-10683	262	20	et	et	PROPN
dad-10683	262	21	al	al	PROPN
dad-10683	262	22	.	.	PROPN
dad-10683	262	23	,	,	PUNCT
dad-10683	262	24	2005	2005	NUM
dad-10683	262	25	)	)	PUNCT
dad-10683	262	26	.	.	PUNCT
dad-10683	263	1	note	note	VERB
dad-10683	263	2	that	that	SCONJ
dad-10683	263	3	retrieval	retrieval	NOUN
dad-10683	263	4	models	model	NOUN
dad-10683	263	5	trained	train	VERB
dad-10683	263	6	on	on	ADP
dad-10683	263	7	the	the	DET
dad-10683	263	8	task	task	NOUN
dad-10683	263	9	of	of	ADP
dad-10683	263	10	nuc	nuc	PROPN
dad-10683	263	11	are	be	AUX
dad-10683	263	12	still	still	ADV
dad-10683	263	13	end	end	NOUN
dad-10683	263	14	-	-	PUNCT
dad-10683	263	15	to	to	ADP
dad-10683	263	16	-	-	PUNCT
dad-10683	263	17	end	end	NOUN
dad-10683	263	18	,	,	PUNCT
dad-10683	263	19	as	as	SCONJ
dad-10683	263	20	the	the	DET
dad-10683	263	21	natural	natural	ADJ
dad-10683	263	22	language	language	NOUN
dad-10683	263	23	understanding	understanding	NOUN
dad-10683	263	24	,	,	PUNCT
dad-10683	263	25	dialogue	dialogue	NOUN
dad-10683	263	26	planning	planning	NOUN
dad-10683	263	27	,	,	PUNCT
dad-10683	263	28	and	and	CCONJ
dad-10683	263	29	generation	generation	NOUN
dad-10683	263	30	modules	module	NOUN
dad-10683	263	31	are	be	AUX
dad-10683	263	32	combined	combine	VERB
dad-10683	263	33	,	,	PUNCT
dad-10683	263	34	and	and	CCONJ
dad-10683	263	35	the	the	DET
dad-10683	263	36	system	system	NOUN
dad-10683	263	37	is	be	AUX
dad-10683	263	38	learned	learn	VERB
dad-10683	263	39	with	with	ADP
dad-10683	263	40	a	a	DET
dad-10683	263	41	single	single	ADJ
dad-10683	263	42	supervision	supervision	NOUN
dad-10683	263	43	signal	signal	NOUN
dad-10683	263	44	.	.	PUNCT
dad-10683	264	1	these	these	DET
dad-10683	264	2	models	model	NOUN
dad-10683	264	3	can	can	AUX
dad-10683	264	4	be	be	AUX
dad-10683	264	5	used	use	VERB
dad-10683	264	6	to	to	PART
dad-10683	264	7	‘	'	PUNCT
dad-10683	264	8	generate	generate	VERB
dad-10683	264	9	’	'	PUNCT
dad-10683	264	10	the	the	DET
dad-10683	264	11	next	next	ADJ
dad-10683	264	12	utterance	utterance	NOUN
dad-10683	264	13	in	in	ADP
dad-10683	264	14	a	a	DET
dad-10683	264	15	conversation	conversation	NOUN
dad-10683	264	16	by	by	ADP
dad-10683	264	17	retrieving	retrieve	VERB
dad-10683	264	18	the	the	DET
dad-10683	264	19	most	most	ADV
dad-10683	264	20	probable	probable	ADJ
dad-10683	264	21	next	next	ADJ
dad-10683	264	22	utterance	utterance	NOUN
dad-10683	264	23	from	from	ADP
dad-10683	264	24	the	the	DET
dad-10683	264	25	entire	entire	ADJ
dad-10683	264	26	training	training	NOUN
dad-10683	264	27	set	set	NOUN
dad-10683	264	28	,	,	PUNCT
dad-10683	264	29	given	give	VERB
dad-10683	264	30	the	the	DET
dad-10683	264	31	context	context	NOUN
dad-10683	264	32	.	.	PUNCT
dad-10683	265	1	thus	thus	ADV
dad-10683	265	2	,	,	PUNCT
dad-10683	265	3	we	we	PRON
dad-10683	265	4	can	can	AUX
dad-10683	265	5	also	also	ADV
dad-10683	265	6	evaluate	evaluate	VERB
dad-10683	265	7	these	these	DET
dad-10683	265	8	models	model	NOUN
dad-10683	265	9	using	use	VERB
dad-10683	265	10	several	several	ADJ
dad-10683	265	11	generative	generative	ADJ
dad-10683	265	12	metrics	metric	NOUN
dad-10683	265	13	,	,	PUNCT
dad-10683	265	14	that	that	PRON
dad-10683	265	15	compare	compare	VERB
dad-10683	265	16	the	the	DET
dad-10683	265	17	selected	select	VERB
dad-10683	265	18	response	response	NOUN
dad-10683	265	19	to	to	ADP
dad-10683	265	20	the	the	DET
dad-10683	265	21	ground	ground	NOUN
dad-10683	265	22	-	-	PUNCT
dad-10683	265	23	truth	truth	NOUN
dad-10683	265	24	response	response	NOUN
dad-10683	265	25	.	.	PUNCT
dad-10683	266	1	we	we	PRON
dad-10683	266	2	carry	carry	VERB
dad-10683	266	3	this	this	PRON
dad-10683	266	4	out	out	ADP
dad-10683	266	5	in	in	ADP
dad-10683	266	6	section	section	NOUN
dad-10683	266	7	4.5	4.5	NUM
dad-10683	266	8	.	.	PUNCT
dad-10683	267	1	we	we	PRON
dad-10683	267	2	consider	consider	VERB
dad-10683	267	3	one	one	NUM
dad-10683	267	4	naive	naive	ADJ
dad-10683	267	5	model	model	NOUN
dad-10683	267	6	and	and	CCONJ
dad-10683	267	7	two	two	NUM
dad-10683	267	8	neural	neural	ADJ
dad-10683	267	9	network	network	NOUN
dad-10683	267	10	-	-	PUNCT
dad-10683	267	11	based	base	VERB
dad-10683	267	12	retrieval	retrieval	NOUN
dad-10683	267	13	models	model	NOUN
dad-10683	267	14	.	.	PUNCT
dad-10683	268	1	the	the	DET
dad-10683	268	2	approaches	approach	NOUN
dad-10683	268	3	considered	consider	VERB
dad-10683	268	4	are	be	AUX
dad-10683	268	5	:	:	PUNCT
dad-10683	268	6	tf	tf	ADJ
dad-10683	268	7	-	-	PUNCT
dad-10683	268	8	idf	idf	PROPN
dad-10683	268	9	,	,	PUNCT
dad-10683	268	10	and	and	CCONJ
dad-10683	268	11	models	model	NOUN
dad-10683	268	12	using	use	VERB
dad-10683	268	13	recurrent	recurrent	ADJ
dad-10683	268	14	neural	neural	ADJ
dad-10683	268	15	networks	network	NOUN
dad-10683	268	16	(	(	PUNCT
dad-10683	268	17	rnn	rnn	PROPN
dad-10683	268	18	)	)	PUNCT
dad-10683	268	19	and	and	CCONJ
dad-10683	268	20	long	long	ADJ
dad-10683	268	21	short	short	ADJ
dad-10683	268	22	-	-	PUNCT
dad-10683	268	23	term	term	NOUN
dad-10683	268	24	memory	memory	NOUN
dad-10683	268	25	(	(	PUNCT
dad-10683	268	26	lstm	lstm	NOUN
dad-10683	268	27	)	)	PUNCT
dad-10683	268	28	.	.	PUNCT
dad-10683	269	1	prior	prior	ADV
dad-10683	269	2	to	to	ADP
dad-10683	269	3	applying	apply	VERB
dad-10683	269	4	each	each	DET
dad-10683	269	5	method	method	NOUN
dad-10683	269	6	,	,	PUNCT
dad-10683	269	7	we	we	PRON
dad-10683	269	8	perform	perform	VERB
dad-10683	269	9	standard	standard	ADJ
dad-10683	269	10	pre	pre	ADJ
dad-10683	269	11	-	-	ADJ
dad-10683	269	12	processing	processing	NOUN
dad-10683	269	13	of	of	ADP
dad-10683	269	14	the	the	DET
dad-10683	269	15	data	datum	NOUN
dad-10683	269	16	using	use	VERB
dad-10683	269	17	the	the	DET
dad-10683	269	18	nltk7	nltk7	NOUN
dad-10683	269	19	library	library	NOUN
dad-10683	269	20	and	and	CCONJ
dad-10683	269	21	twitter	twitter	NOUN
dad-10683	269	22	tokenizer8	tokenizer8	NOUN
dad-10683	269	23	to	to	PART
dad-10683	269	24	parse	parse	VERB
dad-10683	269	25	each	each	DET
dad-10683	269	26	utterance	utterance	NOUN
dad-10683	269	27	.	.	PUNCT
dad-10683	270	1	we	we	PRON
dad-10683	270	2	use	use	VERB
dad-10683	270	3	generic	generic	ADJ
dad-10683	270	4	tags	tag	NOUN
dad-10683	270	5	for	for	ADP
dad-10683	270	6	various	various	ADJ
dad-10683	270	7	word	word	NOUN
dad-10683	270	8	categories	category	NOUN
dad-10683	270	9	,	,	PUNCT
dad-10683	270	10	such	such	ADJ
dad-10683	270	11	as	as	ADP
dad-10683	270	12	names	name	NOUN
dad-10683	270	13	,	,	PUNCT
dad-10683	270	14	locations	location	NOUN
dad-10683	270	15	,	,	PUNCT
dad-10683	270	16	organizations	organization	NOUN
dad-10683	270	17	,	,	PUNCT
dad-10683	270	18	urls	url	NOUN
dad-10683	270	19	,	,	PUNCT
dad-10683	270	20	and	and	CCONJ
dad-10683	270	21	system	system	NOUN
dad-10683	270	22	paths	path	NOUN
dad-10683	270	23	.	.	PUNCT
dad-10683	271	1	to	to	PART
dad-10683	271	2	train	train	VERB
dad-10683	271	3	the	the	DET
dad-10683	271	4	rnn	rnn	NOUN
dad-10683	271	5	and	and	CCONJ
dad-10683	271	6	lstm	lstm	PROPN
dad-10683	271	7	architectures	architecture	NOUN
dad-10683	271	8	,	,	PUNCT
dad-10683	271	9	we	we	PRON
dad-10683	271	10	process	process	VERB
dad-10683	271	11	the	the	DET
dad-10683	271	12	full	full	ADJ
dad-10683	271	13	training	training	NOUN
dad-10683	271	14	ubuntu	ubuntu	ADJ
dad-10683	271	15	dialogue	dialogue	NOUN
dad-10683	271	16	corpus	corpus	NOUN
dad-10683	271	17	into	into	ADP
dad-10683	271	18	the	the	DET
dad-10683	271	19	same	same	ADJ
dad-10683	271	20	format	format	NOUN
dad-10683	271	21	as	as	SCONJ
dad-10683	271	22	the	the	DET
dad-10683	271	23	test	test	NOUN
dad-10683	271	24	set	set	NOUN
dad-10683	271	25	described	describe	VERB
dad-10683	271	26	in	in	ADP
dad-10683	271	27	section	section	NOUN
dad-10683	271	28	3.4	3.4	NUM
dad-10683	271	29	,	,	PUNCT
dad-10683	271	30	extracting	extract	VERB
dad-10683	271	31	(	(	PUNCT
dad-10683	271	32	context	context	NOUN
dad-10683	271	33	,	,	PUNCT
dad-10683	271	34	response	response	NOUN
dad-10683	271	35	,	,	PUNCT
dad-10683	271	36	flag	flag	NOUN
dad-10683	271	37	)	)	PUNCT
dad-10683	271	38	triples	triple	NOUN
dad-10683	271	39	from	from	ADP
dad-10683	271	40	dialogues	dialogue	NOUN
dad-10683	271	41	.	.	PUNCT
dad-10683	272	1	for	for	ADP
dad-10683	272	2	the	the	DET
dad-10683	272	3	training	training	NOUN
dad-10683	272	4	set	set	NOUN
dad-10683	272	5	,	,	PUNCT
dad-10683	272	6	we	we	PRON
dad-10683	272	7	sample	sample	VERB
dad-10683	272	8	the	the	DET
dad-10683	272	9	responses	response	NOUN
dad-10683	272	10	in	in	ADP
dad-10683	272	11	the	the	DET
dad-10683	272	12	same	same	ADJ
dad-10683	272	13	way	way	NOUN
dad-10683	272	14	described	describe	VERB
dad-10683	272	15	in	in	ADP
dad-10683	272	16	section	section	NOUN
dad-10683	272	17	3.4	3.4	NUM
dad-10683	272	18	.	.	PUNCT
dad-10683	273	1	one	one	PRON
dad-10683	273	2	can	can	AUX
dad-10683	273	3	generate	generate	VERB
dad-10683	273	4	any	any	DET
dad-10683	273	5	number	number	NOUN
dad-10683	273	6	of	of	ADP
dad-10683	273	7	training	training	NOUN
dad-10683	273	8	examples	example	NOUN
dad-10683	273	9	by	by	ADP
dad-10683	273	10	iterating	iterate	VERB
dad-10683	273	11	several	several	ADJ
dad-10683	273	12	times	time	NOUN
dad-10683	273	13	through	through	ADP
dad-10683	273	14	the	the	DET
dad-10683	273	15	training	training	NOUN
dad-10683	273	16	data	datum	NOUN
dad-10683	273	17	.	.	PUNCT
dad-10683	274	1	negative	negative	ADJ
dad-10683	274	2	responses	response	NOUN
dad-10683	274	3	are	be	AUX
dad-10683	274	4	selected	select	VERB
dad-10683	274	5	at	at	ADP
dad-10683	274	6	random	random	ADJ
dad-10683	274	7	from	from	ADP
dad-10683	274	8	the	the	DET
dad-10683	274	9	rest	rest	NOUN
dad-10683	274	10	of	of	ADP
dad-10683	274	11	the	the	DET
dad-10683	274	12	training	training	NOUN
dad-10683	274	13	data	datum	NOUN
dad-10683	274	14	.	.	PUNCT
dad-10683	275	1	we	we	PRON
dad-10683	275	2	note	note	VERB
dad-10683	275	3	that	that	SCONJ
dad-10683	275	4	for	for	ADP
dad-10683	275	5	all	all	DET
dad-10683	275	6	models	model	NOUN
dad-10683	275	7	presented	present	VERB
dad-10683	275	8	in	in	ADP
dad-10683	275	9	this	this	DET
dad-10683	275	10	paper	paper	NOUN
dad-10683	275	11	,	,	PUNCT
dad-10683	275	12	the	the	DET
dad-10683	275	13	entire	entire	ADJ
dad-10683	275	14	context	context	NOUN
dad-10683	275	15	of	of	ADP
dad-10683	275	16	the	the	DET
dad-10683	275	17	dialogue	dialogue	NOUN
dad-10683	275	18	that	that	PRON
dad-10683	275	19	is	be	AUX
dad-10683	275	20	available	available	ADJ
dad-10683	275	21	(	(	PUNCT
dad-10683	275	22	i.e.	i.e.	X
dad-10683	275	23	after	after	SCONJ
dad-10683	275	24	the	the	DET
dad-10683	275	25	context	context	NOUN
dad-10683	275	26	length	length	NOUN
dad-10683	275	27	sampling	sample	VERB
dad-10683	275	28	procedure	procedure	NOUN
dad-10683	275	29	in	in	ADP
dad-10683	275	30	section	section	NOUN
dad-10683	275	31	3.4	3.4	NUM
dad-10683	275	32	used	use	VERB
dad-10683	275	33	to	to	PART
dad-10683	275	34	create	create	VERB
dad-10683	275	35	the	the	DET
dad-10683	275	36	dataset	dataset	NOUN
dad-10683	275	37	)	)	PUNCT
dad-10683	275	38	is	be	AUX
dad-10683	275	39	taken	take	VERB
dad-10683	275	40	into	into	ADP
dad-10683	275	41	account	account	NOUN
dad-10683	275	42	,	,	PUNCT
dad-10683	275	43	and	and	CCONJ
dad-10683	275	44	not	not	PART
dad-10683	275	45	just	just	ADV
dad-10683	275	46	the	the	DET
dad-10683	275	47	most	most	ADV
dad-10683	275	48	recent	recent	ADJ
dad-10683	275	49	utterance	utterance	NOUN
dad-10683	275	50	.	.	PUNCT
dad-10683	276	1	for	for	ADP
dad-10683	276	2	the	the	DET
dad-10683	276	3	response	response	NOUN
dad-10683	276	4	classification	classification	NOUN
dad-10683	276	5	architectures	architecture	NOUN
dad-10683	276	6	,	,	PUNCT
dad-10683	276	7	this	this	PRON
dad-10683	276	8	is	be	AUX
dad-10683	276	9	done	do	VERB
dad-10683	276	10	by	by	ADP
dad-10683	276	11	concatenating	concatenate	VERB
dad-10683	276	12	all	all	DET
dad-10683	276	13	context	context	NOUN
dad-10683	276	14	utterances	utterance	NOUN
dad-10683	276	15	together	together	ADV
dad-10683	276	16	.	.	PUNCT
dad-10683	277	1	we	we	PRON
dad-10683	277	2	note	note	VERB
dad-10683	277	3	that	that	SCONJ
dad-10683	277	4	the	the	DET
dad-10683	277	5	models	model	NOUN
dad-10683	277	6	proposed	propose	VERB
dad-10683	277	7	below	below	ADP
dad-10683	277	8	do	do	AUX
dad-10683	277	9	not	not	PART
dad-10683	277	10	explicitly	explicitly	ADV
dad-10683	277	11	take	take	VERB
dad-10683	277	12	into	into	ADP
dad-10683	277	13	account	account	NOUN
dad-10683	277	14	ordinal	ordinal	ADJ
dad-10683	277	15	information	information	NOUN
dad-10683	277	16	.	.	PUNCT
dad-10683	278	1	the	the	DET
dad-10683	278	2	reasons	reason	NOUN
dad-10683	278	3	for	for	ADP
dad-10683	278	4	doing	do	VERB
dad-10683	278	5	this	this	PRON
dad-10683	278	6	are	be	AUX
dad-10683	278	7	two	two	NUM
dad-10683	278	8	-	-	NUM
dad-10683	278	9	fold	fold	ADJ
dad-10683	278	10	.	.	PUNCT
dad-10683	279	1	first	first	ADV
dad-10683	279	2	,	,	PUNCT
dad-10683	279	3	training	train	VERB
dad-10683	279	4	neural	neural	ADJ
dad-10683	279	5	networks	network	NOUN
dad-10683	279	6	using	use	VERB
dad-10683	279	7	classification	classification	NOUN
dad-10683	279	8	for	for	ADP
dad-10683	279	9	ranking	rank	VERB
dad-10683	279	10	tasks	task	NOUN
dad-10683	279	11	is	be	AUX
dad-10683	279	12	well	well	ADV
dad-10683	279	13	-	-	PUNCT
dad-10683	279	14	established	establish	VERB
dad-10683	279	15	in	in	ADP
dad-10683	279	16	the	the	DET
dad-10683	279	17	literature	literature	NOUN
dad-10683	279	18	(	(	PUNCT
dad-10683	279	19	bordes	bordes	PROPN
dad-10683	279	20	et	et	PROPN
dad-10683	279	21	al	al	PROPN
dad-10683	279	22	.	.	PROPN
dad-10683	279	23	,	,	PUNCT
dad-10683	279	24	2014	2014	NUM
dad-10683	279	25	)	)	PUNCT
dad-10683	279	26	,	,	PUNCT
dad-10683	279	27	and	and	CCONJ
dad-10683	279	28	is	be	AUX
dad-10683	279	29	both	both	PRON
dad-10683	279	30	simple	simple	ADJ
dad-10683	279	31	to	to	PART
dad-10683	279	32	implement	implement	VERB
dad-10683	279	33	and	and	CCONJ
dad-10683	279	34	effective	effective	ADJ
dad-10683	279	35	in	in	ADP
dad-10683	279	36	practice	practice	NOUN
dad-10683	279	37	.	.	PUNCT
dad-10683	280	1	second	second	ADJ
dad-10683	280	2	,	,	PUNCT
dad-10683	280	3	in	in	ADP
dad-10683	280	4	the	the	DET
dad-10683	280	5	ubuntu	ubuntu	ADJ
dad-10683	280	6	dialogue	dialogue	NOUN
dad-10683	280	7	corpus	corpus	NOUN
dad-10683	280	8	we	we	PRON
dad-10683	280	9	do	do	AUX
dad-10683	280	10	not	not	PART
dad-10683	280	11	have	have	AUX
dad-10683	280	12	supervised	supervise	VERB
dad-10683	280	13	ordinal	ordinal	ADJ
dad-10683	280	14	data	datum	NOUN
dad-10683	280	15	for	for	ADP
dad-10683	280	16	the	the	DET
dad-10683	280	17	relative	relative	ADJ
dad-10683	280	18	quality	quality	NOUN
dad-10683	280	19	of	of	ADP
dad-10683	280	20	next	next	ADJ
dad-10683	280	21	responses	response	NOUN
dad-10683	280	22	given	give	VERB
dad-10683	280	23	a	a	DET
dad-10683	280	24	context	context	NOUN
dad-10683	280	25	.	.	PUNCT
dad-10683	281	1	more	more	ADV
dad-10683	281	2	advanced	advanced	ADJ
dad-10683	281	3	methods	method	NOUN
dad-10683	281	4	could	could	AUX
dad-10683	281	5	consider	consider	VERB
dad-10683	281	6	some	some	DET
dad-10683	281	7	way	way	NOUN
dad-10683	281	8	to	to	PART
dad-10683	281	9	approximate	approximate	VERB
dad-10683	281	10	this	this	DET
dad-10683	281	11	ordinal	ordinal	ADJ
dad-10683	281	12	information	information	NOUN
dad-10683	281	13	,	,	PUNCT
dad-10683	281	14	such	such	ADJ
dad-10683	281	15	that	that	SCONJ
dad-10683	281	16	a	a	DET
dad-10683	281	17	neural	neural	ADJ
dad-10683	281	18	network	network	NOUN
dad-10683	281	19	model	model	NOUN
dad-10683	281	20	could	could	AUX
dad-10683	281	21	be	be	AUX
dad-10683	281	22	explicitly	explicitly	ADV
dad-10683	281	23	trained	train	VERB
dad-10683	281	24	as	as	ADP
dad-10683	281	25	a	a	DET
dad-10683	281	26	ranking	ranking	ADJ
dad-10683	281	27	system	system	NOUN
dad-10683	281	28	;	;	PUNCT
dad-10683	281	29	however	however	ADV
dad-10683	281	30	,	,	PUNCT
dad-10683	281	31	this	this	PRON
dad-10683	281	32	is	be	AUX
dad-10683	281	33	beyond	beyond	ADP
dad-10683	281	34	the	the	DET
dad-10683	281	35	scope	scope	NOUN
dad-10683	281	36	of	of	ADP
dad-10683	281	37	this	this	DET
dad-10683	281	38	paper	paper	NOUN
dad-10683	281	39	.	.	PUNCT
dad-10683	282	1	7	7	X
dad-10683	282	2	.	.	X
dad-10683	282	3	www.nltk.org/	www.nltk.org/	ADJ
dad-10683	282	4	8	8	NUM
dad-10683	282	5	.	.	PUNCT
dad-10683	282	6	http://www.ark.cs.cmu.edu/tweetnlp/	http://www.ark.cs.cmu.edu/tweetnlp/	PROPN
dad-10683	283	1	42	42	NUM
dad-10683	283	2	training	training	NOUN
dad-10683	283	3	end	end	NOUN
dad-10683	283	4	-	-	PUNCT
dad-10683	283	5	to	to	ADP
dad-10683	283	6	-	-	PUNCT
dad-10683	283	7	end	end	NOUN
dad-10683	283	8	dialogue	dialogue	NOUN
dad-10683	283	9	systems	system	NOUN
dad-10683	283	10	4.1	4.1	NUM
dad-10683	283	11	tf	tf	ADJ
dad-10683	283	12	-	-	PUNCT
dad-10683	283	13	idf	idf	PROPN
dad-10683	283	14	term	term	NOUN
dad-10683	283	15	frequency	frequency	NOUN
dad-10683	283	16	-	-	PUNCT
dad-10683	283	17	inverse	inverse	NOUN
dad-10683	283	18	document	document	NOUN
dad-10683	283	19	frequency	frequency	NOUN
dad-10683	283	20	is	be	AUX
dad-10683	283	21	a	a	DET
dad-10683	283	22	statistic	statistic	NOUN
dad-10683	283	23	that	that	PRON
dad-10683	283	24	intends	intend	VERB
dad-10683	283	25	to	to	PART
dad-10683	283	26	capture	capture	VERB
dad-10683	283	27	how	how	SCONJ
dad-10683	283	28	important	important	ADJ
dad-10683	283	29	a	a	DET
dad-10683	283	30	given	give	VERB
dad-10683	283	31	word	word	NOUN
dad-10683	283	32	is	be	AUX
dad-10683	283	33	to	to	ADP
dad-10683	283	34	some	some	DET
dad-10683	283	35	document	document	NOUN
dad-10683	283	36	,	,	PUNCT
dad-10683	283	37	which	which	PRON
dad-10683	283	38	in	in	ADP
dad-10683	283	39	our	our	PRON
dad-10683	283	40	case	case	NOUN
dad-10683	283	41	is	be	AUX
dad-10683	283	42	the	the	DET
dad-10683	283	43	context	context	NOUN
dad-10683	283	44	(	(	PUNCT
dad-10683	283	45	ramos	ramos	PROPN
dad-10683	283	46	,	,	PUNCT
dad-10683	283	47	2003	2003	NUM
dad-10683	283	48	)	)	PUNCT
dad-10683	283	49	.	.	PUNCT
dad-10683	284	1	it	it	PRON
dad-10683	284	2	is	be	AUX
dad-10683	284	3	a	a	DET
dad-10683	284	4	technique	technique	NOUN
dad-10683	284	5	often	often	ADV
dad-10683	284	6	used	use	VERB
dad-10683	284	7	in	in	ADP
dad-10683	284	8	document	document	NOUN
dad-10683	284	9	classification	classification	NOUN
dad-10683	284	10	and	and	CCONJ
dad-10683	284	11	information	information	NOUN
dad-10683	284	12	retrieval	retrieval	NOUN
dad-10683	284	13	.	.	PUNCT
dad-10683	285	1	the	the	DET
dad-10683	285	2	‘	'	PUNCT
dad-10683	285	3	term	term	NOUN
dad-10683	285	4	-	-	PUNCT
dad-10683	285	5	frequency	frequency	NOUN
dad-10683	285	6	’	'	PUNCT
dad-10683	285	7	term	term	NOUN
dad-10683	285	8	is	be	AUX
dad-10683	285	9	simply	simply	ADV
dad-10683	285	10	a	a	DET
dad-10683	285	11	count	count	NOUN
dad-10683	285	12	of	of	ADP
dad-10683	285	13	the	the	DET
dad-10683	285	14	number	number	NOUN
dad-10683	285	15	of	of	ADP
dad-10683	285	16	times	time	NOUN
dad-10683	285	17	a	a	DET
dad-10683	285	18	word	word	NOUN
dad-10683	285	19	appears	appear	VERB
dad-10683	285	20	in	in	ADP
dad-10683	285	21	a	a	DET
dad-10683	285	22	given	give	VERB
dad-10683	285	23	context	context	NOUN
dad-10683	285	24	,	,	PUNCT
dad-10683	285	25	while	while	SCONJ
dad-10683	285	26	the	the	DET
dad-10683	285	27	‘	'	PUNCT
dad-10683	285	28	inverse	inverse	ADJ
dad-10683	285	29	document	document	NOUN
dad-10683	285	30	frequency	frequency	NOUN
dad-10683	285	31	’	'	PUNCT
dad-10683	285	32	term	term	NOUN
dad-10683	285	33	puts	put	VERB
dad-10683	285	34	a	a	DET
dad-10683	285	35	penalty	penalty	NOUN
dad-10683	285	36	on	on	ADP
dad-10683	285	37	how	how	SCONJ
dad-10683	285	38	often	often	ADV
dad-10683	285	39	the	the	DET
dad-10683	285	40	word	word	NOUN
dad-10683	285	41	appears	appear	VERB
dad-10683	285	42	elsewhere	elsewhere	ADV
dad-10683	285	43	in	in	ADP
dad-10683	285	44	the	the	DET
dad-10683	285	45	corpus	corpus	NOUN
dad-10683	285	46	.	.	PUNCT
dad-10683	286	1	the	the	DET
dad-10683	286	2	final	final	ADJ
dad-10683	286	3	score	score	NOUN
dad-10683	286	4	is	be	AUX
dad-10683	286	5	calculated	calculate	VERB
dad-10683	286	6	as	as	ADP
dad-10683	286	7	the	the	DET
dad-10683	286	8	product	product	NOUN
dad-10683	286	9	of	of	ADP
dad-10683	286	10	these	these	DET
dad-10683	286	11	two	two	NUM
dad-10683	286	12	terms	term	NOUN
dad-10683	286	13	,	,	PUNCT
dad-10683	286	14	and	and	CCONJ
dad-10683	286	15	has	have	VERB
dad-10683	286	16	the	the	DET
dad-10683	286	17	form	form	NOUN
dad-10683	286	18	:	:	PUNCT
dad-10683	286	19	tfidf(w	tfidf(w	NOUN
dad-10683	286	20	,	,	PUNCT
dad-10683	286	21	d	d	NOUN
dad-10683	286	22	,	,	PUNCT
dad-10683	286	23	d	d	NOUN
dad-10683	286	24	)	)	PUNCT
dad-10683	286	25	=	=	SYM
dad-10683	286	26	f(w	f(w	NOUN
dad-10683	286	27	,	,	PUNCT
dad-10683	286	28	d)×	d)×	X
dad-10683	286	29	log	log	NOUN
dad-10683	286	30	|d|	|d|	PROPN
dad-10683	286	31	|{d	|{d	NOUN
dad-10683	286	32	∈	∈	PROPN
dad-10683	287	1	d	d	X
dad-10683	287	2	:	:	PUNCT
dad-10683	287	3	w	w	PROPN
dad-10683	287	4	∈	∈	PROPN
dad-10683	287	5	d}|	d}|	NOUN
dad-10683	287	6	,	,	PUNCT
dad-10683	287	7	(	(	PUNCT
dad-10683	287	8	1	1	X
dad-10683	287	9	)	)	PUNCT
dad-10683	287	10	where	where	SCONJ
dad-10683	287	11	f(w	f(w	PROPN
dad-10683	287	12	,	,	PUNCT
dad-10683	287	13	d	d	NOUN
dad-10683	287	14	)	)	PUNCT
dad-10683	287	15	indicates	indicate	VERB
dad-10683	287	16	the	the	DET
dad-10683	287	17	number	number	NOUN
dad-10683	287	18	of	of	ADP
dad-10683	287	19	times	time	NOUN
dad-10683	287	20	word	word	NOUN
dad-10683	287	21	w	w	NOUN
dad-10683	287	22	appeared	appear	VERB
dad-10683	287	23	in	in	ADP
dad-10683	287	24	context	context	NOUN
dad-10683	287	25	d	d	NOUN
dad-10683	287	26	and	and	CCONJ
dad-10683	287	27	the	the	DET
dad-10683	287	28	denominator	denominator	NOUN
dad-10683	287	29	represents	represent	VERB
dad-10683	287	30	the	the	DET
dad-10683	287	31	number	number	NOUN
dad-10683	287	32	of	of	ADP
dad-10683	287	33	dialogues	dialogue	NOUN
dad-10683	287	34	in	in	ADP
dad-10683	287	35	which	which	PRON
dad-10683	287	36	the	the	DET
dad-10683	287	37	word	word	NOUN
dad-10683	287	38	w	w	NOUN
dad-10683	287	39	appears	appear	VERB
dad-10683	287	40	.	.	PUNCT
dad-10683	288	1	for	for	ADP
dad-10683	288	2	classification	classification	NOUN
dad-10683	288	3	,	,	PUNCT
dad-10683	288	4	the	the	DET
dad-10683	288	5	tf	tf	PROPN
dad-10683	288	6	-	-	PUNCT
dad-10683	288	7	idf	idf	PROPN
dad-10683	288	8	vectors	vector	NOUN
dad-10683	288	9	are	be	AUX
dad-10683	288	10	first	first	ADV
dad-10683	288	11	calculated	calculate	VERB
dad-10683	288	12	for	for	ADP
dad-10683	288	13	the	the	DET
dad-10683	288	14	context	context	NOUN
dad-10683	288	15	and	and	CCONJ
dad-10683	288	16	each	each	PRON
dad-10683	288	17	of	of	ADP
dad-10683	288	18	the	the	DET
dad-10683	288	19	candidate	candidate	NOUN
dad-10683	288	20	responses	response	NOUN
dad-10683	288	21	.	.	PUNCT
dad-10683	289	1	given	give	VERB
dad-10683	289	2	a	a	DET
dad-10683	289	3	set	set	NOUN
dad-10683	289	4	of	of	ADP
dad-10683	289	5	candidate	candidate	NOUN
dad-10683	289	6	response	response	NOUN
dad-10683	289	7	vectors	vector	NOUN
dad-10683	289	8	,	,	PUNCT
dad-10683	289	9	the	the	DET
dad-10683	289	10	one	one	NOUN
dad-10683	289	11	with	with	ADP
dad-10683	289	12	the	the	DET
dad-10683	289	13	highest	high	ADJ
dad-10683	289	14	cosine	cosine	NOUN
dad-10683	289	15	similarity	similarity	NOUN
dad-10683	289	16	to	to	ADP
dad-10683	289	17	the	the	DET
dad-10683	289	18	context	context	NOUN
dad-10683	289	19	vector	vector	NOUN
dad-10683	289	20	is	be	AUX
dad-10683	289	21	selected	select	VERB
dad-10683	289	22	as	as	ADP
dad-10683	289	23	the	the	DET
dad-10683	289	24	output	output	NOUN
dad-10683	289	25	.	.	PUNCT
dad-10683	290	1	for	for	ADP
dad-10683	290	2	recall@k	recall@k	NOUN
dad-10683	290	3	,	,	PUNCT
dad-10683	290	4	the	the	DET
dad-10683	290	5	top	top	ADJ
dad-10683	290	6	k	k	PROPN
dad-10683	290	7	responses	response	NOUN
dad-10683	290	8	are	be	AUX
dad-10683	290	9	returned	return	VERB
dad-10683	290	10	.	.	PUNCT
dad-10683	291	1	4.2	4.2	NUM
dad-10683	291	2	rnn	rnn	VERB
dad-10683	291	3	dual	dual	ADJ
dad-10683	291	4	encoder	encoder	NOUN
dad-10683	291	5	recurrent	recurrent	ADJ
dad-10683	291	6	neural	neural	ADJ
dad-10683	291	7	networks	network	NOUN
dad-10683	291	8	are	be	AUX
dad-10683	291	9	a	a	DET
dad-10683	291	10	variant	variant	NOUN
dad-10683	291	11	of	of	ADP
dad-10683	291	12	neural	neural	ADJ
dad-10683	291	13	networks	network	NOUN
dad-10683	291	14	that	that	PRON
dad-10683	291	15	allows	allow	VERB
dad-10683	291	16	for	for	ADP
dad-10683	291	17	time	time	NOUN
dad-10683	291	18	-	-	PUNCT
dad-10683	291	19	delayed	delay	VERB
dad-10683	291	20	directed	direct	VERB
dad-10683	291	21	cycles	cycle	NOUN
dad-10683	291	22	between	between	ADP
dad-10683	291	23	units	unit	NOUN
dad-10683	291	24	(	(	PUNCT
dad-10683	291	25	elman	elman	NOUN
dad-10683	291	26	,	,	PUNCT
dad-10683	291	27	1990	1990	NUM
dad-10683	291	28	)	)	PUNCT
dad-10683	291	29	.	.	PUNCT
dad-10683	292	1	this	this	PRON
dad-10683	292	2	leads	lead	VERB
dad-10683	292	3	to	to	ADP
dad-10683	292	4	the	the	DET
dad-10683	292	5	formation	formation	NOUN
dad-10683	292	6	of	of	ADP
dad-10683	292	7	an	an	DET
dad-10683	292	8	internal	internal	ADJ
dad-10683	292	9	state	state	NOUN
dad-10683	292	10	of	of	ADP
dad-10683	292	11	the	the	DET
dad-10683	292	12	network	network	NOUN
dad-10683	292	13	,	,	PUNCT
dad-10683	292	14	ht	ht	PROPN
dad-10683	292	15	at	at	ADP
dad-10683	292	16	time	time	NOUN
dad-10683	292	17	step	step	NOUN
dad-10683	292	18	(	(	PUNCT
dad-10683	292	19	word	word	NOUN
dad-10683	292	20	index	index	PROPN
dad-10683	292	21	)	)	PUNCT
dad-10683	292	22	t	t	PROPN
dad-10683	292	23	,	,	PUNCT
dad-10683	292	24	which	which	PRON
dad-10683	292	25	allows	allow	VERB
dad-10683	292	26	it	it	PRON
dad-10683	292	27	to	to	PART
dad-10683	292	28	model	model	VERB
dad-10683	292	29	time	time	NOUN
dad-10683	292	30	-	-	PUNCT
dad-10683	292	31	dependent	dependent	ADJ
dad-10683	292	32	data	datum	NOUN
dad-10683	292	33	.	.	PUNCT
dad-10683	293	1	the	the	DET
dad-10683	293	2	internal	internal	ADJ
dad-10683	293	3	state	state	NOUN
dad-10683	293	4	is	be	AUX
dad-10683	293	5	updated	update	VERB
dad-10683	293	6	at	at	ADP
dad-10683	293	7	each	each	DET
dad-10683	293	8	time	time	NOUN
dad-10683	293	9	step	step	NOUN
dad-10683	293	10	as	as	ADP
dad-10683	293	11	some	some	DET
dad-10683	293	12	function	function	NOUN
dad-10683	293	13	of	of	ADP
dad-10683	293	14	the	the	DET
dad-10683	293	15	observed	observe	VERB
dad-10683	293	16	variables	variable	NOUN
dad-10683	293	17	xt	xt	ADP
dad-10683	293	18	,	,	PUNCT
dad-10683	293	19	and	and	CCONJ
dad-10683	293	20	the	the	DET
dad-10683	293	21	hidden	hidden	ADJ
dad-10683	293	22	state	state	NOUN
dad-10683	293	23	at	at	ADP
dad-10683	293	24	the	the	DET
dad-10683	293	25	previous	previous	ADJ
dad-10683	293	26	time	time	NOUN
dad-10683	293	27	step	step	NOUN
dad-10683	293	28	ht−1	ht−1	PROPN
dad-10683	293	29	,	,	PUNCT
dad-10683	293	30	with	with	ADP
dad-10683	293	31	w	w	PROPN
dad-10683	293	32	x	x	X
dad-10683	293	33	and	and	CCONJ
dad-10683	293	34	w	w	PROPN
dad-10683	293	35	h	h	NOUN
dad-10683	293	36	matrices	matrix	NOUN
dad-10683	293	37	associated	associate	VERB
dad-10683	293	38	with	with	ADP
dad-10683	293	39	the	the	DET
dad-10683	293	40	input	input	NOUN
dad-10683	293	41	and	and	CCONJ
dad-10683	293	42	hidden	hidden	ADJ
dad-10683	293	43	state	state	NOUN
dad-10683	293	44	:	:	PUNCT
dad-10683	293	45	ht	ht	PROPN
dad-10683	293	46	=	=	PUNCT
dad-10683	294	1	f(w	f(w	PROPN
dad-10683	294	2	hht−1	hht−1	PROPN
dad-10683	295	1	+	+	PROPN
dad-10683	295	2	w	w	PROPN
dad-10683	295	3	xxt	xxt	PROPN
dad-10683	295	4	)	)	PUNCT
dad-10683	295	5	.	.	PUNCT
dad-10683	296	1	(	(	PUNCT
dad-10683	296	2	2	2	X
dad-10683	296	3	)	)	PUNCT
dad-10683	296	4	rnns	rnn	NOUN
dad-10683	296	5	have	have	AUX
dad-10683	296	6	been	be	AUX
dad-10683	296	7	the	the	DET
dad-10683	296	8	primary	primary	ADJ
dad-10683	296	9	building	building	NOUN
dad-10683	296	10	block	block	NOUN
dad-10683	296	11	of	of	ADP
dad-10683	296	12	many	many	ADJ
dad-10683	296	13	current	current	ADJ
dad-10683	296	14	neural	neural	ADJ
dad-10683	296	15	models	model	NOUN
dad-10683	296	16	for	for	ADP
dad-10683	296	17	language	language	NOUN
dad-10683	296	18	-	-	PUNCT
dad-10683	296	19	related	relate	VERB
dad-10683	296	20	tasks	task	NOUN
dad-10683	296	21	(	(	PUNCT
dad-10683	296	22	sutskever	sutskever	PROPN
dad-10683	296	23	et	et	PROPN
dad-10683	296	24	al	al	PROPN
dad-10683	296	25	.	.	PROPN
dad-10683	296	26	,	,	PUNCT
dad-10683	296	27	2014	2014	NUM
dad-10683	296	28	;	;	PUNCT
dad-10683	296	29	sordoni	sordoni	PROPN
dad-10683	296	30	et	et	PROPN
dad-10683	296	31	al	al	PROPN
dad-10683	296	32	.	.	PROPN
dad-10683	296	33	,	,	PUNCT
dad-10683	296	34	2015b	2015b	NUM
dad-10683	296	35	)	)	PUNCT
dad-10683	296	36	,	,	PUNCT
dad-10683	296	37	which	which	PRON
dad-10683	296	38	use	use	VERB
dad-10683	296	39	rnns	rnn	NOUN
dad-10683	296	40	as	as	ADP
dad-10683	296	41	encoders	encoder	NOUN
dad-10683	296	42	and	and	CCONJ
dad-10683	296	43	decoders	decoder	NOUN
dad-10683	296	44	;	;	PUNCT
dad-10683	296	45	in	in	ADP
dad-10683	296	46	this	this	DET
dad-10683	296	47	case	case	NOUN
dad-10683	296	48	,	,	PUNCT
dad-10683	296	49	the	the	DET
dad-10683	296	50	first	first	ADJ
dad-10683	296	51	rnn	rnn	NOUN
dad-10683	296	52	is	be	AUX
dad-10683	296	53	used	use	VERB
dad-10683	296	54	to	to	PART
dad-10683	296	55	encode	encode	VERB
dad-10683	296	56	the	the	DET
dad-10683	296	57	given	give	VERB
dad-10683	296	58	context	context	NOUN
dad-10683	296	59	,	,	PUNCT
dad-10683	296	60	and	and	CCONJ
dad-10683	296	61	the	the	DET
dad-10683	296	62	second	second	ADJ
dad-10683	296	63	rnn	rnn	NOUN
dad-10683	296	64	generates	generate	VERB
dad-10683	296	65	a	a	DET
dad-10683	296	66	response	response	NOUN
dad-10683	296	67	by	by	ADP
dad-10683	296	68	using	use	VERB
dad-10683	296	69	beam	beam	NOUN
dad-10683	296	70	-	-	PUNCT
dad-10683	296	71	search	search	NOUN
dad-10683	296	72	,	,	PUNCT
dad-10683	296	73	where	where	SCONJ
dad-10683	296	74	its	its	PRON
dad-10683	296	75	initial	initial	ADJ
dad-10683	296	76	hidden	hide	VERB
dad-10683	296	77	state	state	NOUN
dad-10683	296	78	is	be	AUX
dad-10683	296	79	biased	bias	VERB
dad-10683	296	80	using	use	VERB
dad-10683	296	81	the	the	DET
dad-10683	296	82	final	final	ADJ
dad-10683	296	83	hidden	hidden	ADJ
dad-10683	296	84	state	state	NOUN
dad-10683	296	85	from	from	ADP
dad-10683	296	86	the	the	DET
dad-10683	296	87	first	first	ADJ
dad-10683	296	88	rnn	rnn	NOUN
dad-10683	296	89	.	.	PUNCT
dad-10683	297	1	we	we	PRON
dad-10683	297	2	detail	detail	VERB
dad-10683	297	3	such	such	ADJ
dad-10683	297	4	models	model	NOUN
dad-10683	297	5	in	in	ADP
dad-10683	297	6	section	section	NOUN
dad-10683	297	7	5	5	NUM
dad-10683	297	8	.	.	PUNCT
dad-10683	298	1	however	however	ADV
dad-10683	298	2	,	,	PUNCT
dad-10683	298	3	in	in	ADP
dad-10683	298	4	this	this	DET
dad-10683	298	5	section	section	NOUN
dad-10683	298	6	,	,	PUNCT
dad-10683	298	7	we	we	PRON
dad-10683	298	8	are	be	AUX
dad-10683	298	9	concerned	concerned	ADJ
dad-10683	298	10	with	with	ADP
dad-10683	298	11	classification	classification	NOUN
dad-10683	298	12	of	of	ADP
dad-10683	298	13	responses	response	NOUN
dad-10683	298	14	,	,	PUNCT
dad-10683	298	15	and	and	CCONJ
dad-10683	298	16	thus	thus	ADV
dad-10683	298	17	using	use	VERB
dad-10683	298	18	a	a	DET
dad-10683	298	19	decoder	decoder	NOUN
dad-10683	298	20	rnn	rnn	NOUN
dad-10683	298	21	for	for	ADP
dad-10683	298	22	generation	generation	NOUN
dad-10683	298	23	is	be	AUX
dad-10683	298	24	not	not	PART
dad-10683	298	25	strictly	strictly	ADV
dad-10683	298	26	necessary	necessary	ADJ
dad-10683	298	27	(	(	PUNCT
dad-10683	298	28	and	and	CCONJ
dad-10683	298	29	thus	thus	ADV
dad-10683	298	30	is	be	AUX
dad-10683	298	31	not	not	PART
dad-10683	298	32	used	use	VERB
dad-10683	298	33	in	in	ADP
dad-10683	298	34	the	the	DET
dad-10683	298	35	model	model	NOUN
dad-10683	298	36	shown	show	VERB
dad-10683	298	37	in	in	ADP
dad-10683	298	38	figure	figure	NOUN
dad-10683	298	39	5	5	NUM
dad-10683	298	40	)	)	PUNCT
dad-10683	298	41	.	.	PUNCT
dad-10683	299	1	in	in	ADP
dad-10683	299	2	this	this	DET
dad-10683	299	3	section	section	NOUN
dad-10683	299	4	we	we	PRON
dad-10683	299	5	build	build	VERB
dad-10683	299	6	upon	upon	SCONJ
dad-10683	299	7	the	the	DET
dad-10683	299	8	approach	approach	NOUN
dad-10683	299	9	in	in	ADP
dad-10683	299	10	(	(	PUNCT
dad-10683	299	11	bordes	borde	NOUN
dad-10683	299	12	et	et	PROPN
dad-10683	299	13	al	al	PROPN
dad-10683	299	14	.	.	PROPN
dad-10683	299	15	,	,	PUNCT
dad-10683	299	16	2014	2014	NUM
dad-10683	299	17	)	)	PUNCT
dad-10683	299	18	,	,	PUNCT
dad-10683	299	19	which	which	PRON
dad-10683	299	20	has	have	AUX
dad-10683	299	21	also	also	ADV
dad-10683	299	22	been	be	AUX
dad-10683	299	23	recently	recently	ADV
dad-10683	299	24	applied	apply	VERB
dad-10683	299	25	to	to	ADP
dad-10683	299	26	the	the	DET
dad-10683	299	27	problem	problem	NOUN
dad-10683	299	28	of	of	ADP
dad-10683	299	29	question	question	NOUN
dad-10683	299	30	answering	answer	VERB
dad-10683	299	31	(	(	PUNCT
dad-10683	299	32	yu	yu	PROPN
dad-10683	299	33	et	et	PROPN
dad-10683	299	34	al	al	PROPN
dad-10683	299	35	.	.	PROPN
dad-10683	299	36	,	,	PUNCT
dad-10683	299	37	2014	2014	NUM
dad-10683	299	38	)	)	PUNCT
dad-10683	299	39	,	,	PUNCT
dad-10683	299	40	and	and	CCONJ
dad-10683	299	41	use	use	VERB
dad-10683	299	42	rnns	rnn	NOUN
dad-10683	299	43	for	for	ADP
dad-10683	299	44	classification	classification	NOUN
dad-10683	299	45	rather	rather	ADV
dad-10683	299	46	then	then	ADV
dad-10683	299	47	generation	generation	NOUN
dad-10683	299	48	.	.	PUNCT
dad-10683	300	1	we	we	PRON
dad-10683	300	2	use	use	VERB
dad-10683	300	3	a	a	DET
dad-10683	300	4	siamese	siamese	ADJ
dad-10683	300	5	network	network	NOUN
dad-10683	300	6	consisting	consist	VERB
dad-10683	300	7	of	of	ADP
dad-10683	300	8	two	two	NUM
dad-10683	300	9	rnns	rnn	NOUN
dad-10683	300	10	with	with	ADP
dad-10683	300	11	tied	tie	VERB
dad-10683	300	12	weights	weight	NOUN
dad-10683	300	13	to	to	PART
dad-10683	300	14	produce	produce	VERB
dad-10683	300	15	the	the	DET
dad-10683	300	16	embeddings	embedding	NOUN
dad-10683	300	17	for	for	ADP
dad-10683	300	18	the	the	DET
dad-10683	300	19	context	context	NOUN
dad-10683	300	20	and	and	CCONJ
dad-10683	300	21	response	response	NOUN
dad-10683	300	22	,	,	PUNCT
dad-10683	300	23	that	that	SCONJ
dad-10683	300	24	we	we	PRON
dad-10683	300	25	call	call	VERB
dad-10683	300	26	the	the	DET
dad-10683	300	27	dual	dual	ADJ
dad-10683	300	28	-	-	PUNCT
dad-10683	300	29	encoder	encoder	NOUN
dad-10683	300	30	(	(	PUNCT
dad-10683	300	31	de	de	NOUN
dad-10683	300	32	)	)	PUNCT
dad-10683	300	33	model	model	NOUN
dad-10683	300	34	.	.	PUNCT
dad-10683	301	1	given	give	VERB
dad-10683	301	2	some	some	DET
dad-10683	301	3	input	input	NOUN
dad-10683	301	4	context	context	NOUN
dad-10683	301	5	and	and	CCONJ
dad-10683	301	6	response	response	NOUN
dad-10683	301	7	,	,	PUNCT
dad-10683	301	8	we	we	PRON
dad-10683	301	9	compute	compute	VERB
dad-10683	301	10	their	their	PRON
dad-10683	301	11	embeddings	embedding	NOUN
dad-10683	301	12	—	—	PUNCT
dad-10683	301	13	c	c	X
dad-10683	301	14	,	,	PUNCT
dad-10683	301	15	r	r	PROPN
dad-10683	301	16	∈	∈	PROPN
dad-10683	301	17	rd	rd	PROPN
dad-10683	301	18	,	,	PUNCT
dad-10683	301	19	respectively	respectively	ADV
dad-10683	301	20	—	—	PUNCT
dad-10683	301	21	by	by	ADP
dad-10683	301	22	feeding	feed	VERB
dad-10683	301	23	the	the	DET
dad-10683	301	24	word	word	NOUN
dad-10683	301	25	embeddings	embedding	VERB
dad-10683	301	26	one	one	NUM
dad-10683	301	27	at	at	ADP
dad-10683	301	28	a	a	DET
dad-10683	301	29	time	time	NOUN
dad-10683	301	30	into	into	ADP
dad-10683	301	31	its	its	PRON
dad-10683	301	32	respective	respective	ADJ
dad-10683	301	33	rnn	rnn	NOUN
dad-10683	301	34	.	.	PUNCT
dad-10683	302	1	word	word	NOUN
dad-10683	302	2	embeddings	embedding	NOUN
dad-10683	302	3	are	be	AUX
dad-10683	302	4	initialized	initialize	VERB
dad-10683	302	5	using	use	VERB
dad-10683	302	6	the	the	DET
dad-10683	302	7	pretrained	pretraine	VERB
dad-10683	302	8	vectors	vector	NOUN
dad-10683	302	9	(	(	PUNCT
dad-10683	302	10	common	common	ADJ
dad-10683	302	11	crawl	crawl	NOUN
dad-10683	302	12	,	,	PUNCT
dad-10683	302	13	840b	840b	NUM
dad-10683	302	14	tokens	token	NOUN
dad-10683	302	15	from	from	ADP
dad-10683	302	16	(	(	PUNCT
dad-10683	302	17	pennington	pennington	PROPN
dad-10683	302	18	et	et	PROPN
dad-10683	302	19	al	al	PROPN
dad-10683	302	20	.	.	PROPN
dad-10683	302	21	,	,	PUNCT
dad-10683	302	22	2014	2014	NUM
dad-10683	302	23	)	)	PUNCT
dad-10683	302	24	)	)	PUNCT
dad-10683	302	25	,	,	PUNCT
dad-10683	302	26	and	and	CCONJ
dad-10683	302	27	fine	fine	ADV
dad-10683	302	28	-	-	PUNCT
dad-10683	302	29	tuned	tune	VERB
dad-10683	302	30	during	during	ADP
dad-10683	302	31	training	training	NOUN
dad-10683	302	32	.	.	PUNCT
dad-10683	303	1	the	the	DET
dad-10683	303	2	hidden	hidden	ADJ
dad-10683	303	3	state	state	NOUN
dad-10683	303	4	of	of	ADP
dad-10683	303	5	the	the	DET
dad-10683	303	6	rnn	rnn	NOUN
dad-10683	303	7	is	be	AUX
dad-10683	303	8	updated	update	VERB
dad-10683	303	9	at	at	ADP
dad-10683	303	10	each	each	DET
dad-10683	303	11	step	step	NOUN
dad-10683	303	12	,	,	PUNCT
dad-10683	303	13	and	and	CCONJ
dad-10683	303	14	the	the	DET
dad-10683	303	15	final	final	ADJ
dad-10683	303	16	hidden	hidden	ADJ
dad-10683	303	17	state	state	NOUN
dad-10683	303	18	represents	represent	VERB
dad-10683	303	19	a	a	DET
dad-10683	303	20	summary	summary	NOUN
dad-10683	303	21	of	of	ADP
dad-10683	303	22	the	the	DET
dad-10683	303	23	input	input	NOUN
dad-10683	303	24	utterance	utterance	ADV
dad-10683	303	25	.	.	PUNCT
dad-10683	304	1	using	use	VERB
dad-10683	304	2	the	the	DET
dad-10683	304	3	final	final	ADJ
dad-10683	304	4	hidden	hide	VERB
dad-10683	304	5	states	state	NOUN
dad-10683	304	6	from	from	ADP
dad-10683	304	7	both	both	DET
dad-10683	304	8	rnns	rnn	NOUN
dad-10683	304	9	,	,	PUNCT
dad-10683	304	10	we	we	PRON
dad-10683	304	11	then	then	ADV
dad-10683	304	12	calculate	calculate	VERB
dad-10683	304	13	the	the	DET
dad-10683	304	14	probability	probability	NOUN
dad-10683	304	15	that	that	SCONJ
dad-10683	304	16	this	this	PRON
dad-10683	304	17	is	be	AUX
dad-10683	304	18	a	a	DET
dad-10683	304	19	valid	valid	ADJ
dad-10683	304	20	pair	pair	NOUN
dad-10683	304	21	:	:	PUNCT
dad-10683	304	22	p(flag	p(flag	NOUN
dad-10683	304	23	=	=	SYM
dad-10683	304	24	1|c	1|c	PROPN
dad-10683	304	25	,	,	PUNCT
dad-10683	304	26	r	r	NOUN
dad-10683	304	27	,	,	PUNCT
dad-10683	304	28	m	m	NOUN
dad-10683	304	29	)	)	PUNCT
dad-10683	304	30	=	=	SYM
dad-10683	305	1	σ(ctmr	σ(ctmr	NOUN
dad-10683	306	1	+	+	CCONJ
dad-10683	307	1	b	b	X
dad-10683	307	2	)	)	PUNCT
dad-10683	307	3	,	,	PUNCT
dad-10683	307	4	(	(	PUNCT
dad-10683	307	5	3	3	X
dad-10683	307	6	)	)	SYM
dad-10683	307	7	43	43	NUM
dad-10683	307	8	lowe	lowe	NOUN
dad-10683	307	9	,	,	PUNCT
dad-10683	307	10	pow	pow	PROPN
dad-10683	307	11	,	,	PUNCT
dad-10683	307	12	serban	serban	PROPN
dad-10683	307	13	,	,	PUNCT
dad-10683	307	14	charlinn	charlinn	PROPN
dad-10683	307	15	,	,	PUNCT
dad-10683	307	16	liu	liu	PROPN
dad-10683	307	17	and	and	CCONJ
dad-10683	307	18	pineau	pineau	PROPN
dad-10683	307	19	figure	figure	NOUN
dad-10683	307	20	5	5	NUM
dad-10683	307	21	:	:	PUNCT
dad-10683	307	22	diagram	diagram	NOUN
dad-10683	307	23	of	of	ADP
dad-10683	307	24	the	the	DET
dad-10683	307	25	dual	dual	ADJ
dad-10683	307	26	encoder	encoder	NOUN
dad-10683	307	27	(	(	PUNCT
dad-10683	307	28	de	de	NOUN
dad-10683	307	29	)	)	PUNCT
dad-10683	307	30	model	model	NOUN
dad-10683	307	31	.	.	PUNCT
dad-10683	308	1	the	the	DET
dad-10683	308	2	rnns	rnn	NOUN
dad-10683	308	3	have	have	AUX
dad-10683	308	4	tied	tie	VERB
dad-10683	308	5	weights	weight	NOUN
dad-10683	308	6	.	.	PUNCT
dad-10683	309	1	c	c	X
dad-10683	309	2	,	,	PUNCT
dad-10683	309	3	r	r	NOUN
dad-10683	309	4	are	be	AUX
dad-10683	309	5	the	the	DET
dad-10683	309	6	last	last	ADJ
dad-10683	309	7	hidden	hide	VERB
dad-10683	309	8	states	state	NOUN
dad-10683	309	9	from	from	ADP
dad-10683	309	10	the	the	DET
dad-10683	309	11	rnns	rnn	NOUN
dad-10683	309	12	.	.	PUNCT
dad-10683	310	1	ci	ci	PROPN
dad-10683	310	2	,	,	PUNCT
dad-10683	310	3	ri	ri	PROPN
dad-10683	310	4	are	be	AUX
dad-10683	310	5	word	word	NOUN
dad-10683	310	6	vectors	vector	NOUN
dad-10683	310	7	for	for	ADP
dad-10683	310	8	the	the	DET
dad-10683	310	9	context	context	NOUN
dad-10683	310	10	and	and	CCONJ
dad-10683	310	11	response	response	NOUN
dad-10683	310	12	,	,	PUNCT
dad-10683	310	13	i	i	PRON
dad-10683	310	14	<	<	X
dad-10683	310	15	t.	t.	X
dad-10683	310	16	we	we	PRON
dad-10683	310	17	consider	consider	VERB
dad-10683	310	18	contexts	context	NOUN
dad-10683	310	19	up	up	ADP
dad-10683	310	20	to	to	ADP
dad-10683	310	21	a	a	DET
dad-10683	310	22	maximum	maximum	NOUN
dad-10683	310	23	of	of	ADP
dad-10683	310	24	t	t	NOUN
dad-10683	310	25	=	=	SYM
dad-10683	310	26	160	160	NUM
dad-10683	310	27	.	.	PUNCT
dad-10683	310	28	where	where	SCONJ
dad-10683	310	29	the	the	DET
dad-10683	310	30	bias	bias	NOUN
dad-10683	310	31	b	b	PROPN
dad-10683	310	32	and	and	CCONJ
dad-10683	310	33	the	the	DET
dad-10683	310	34	matrix	matrix	NOUN
dad-10683	310	35	m	m	NOUN
dad-10683	310	36	∈	∈	PROPN
dad-10683	310	37	rd×d	rd×d	PROPN
dad-10683	310	38	are	be	AUX
dad-10683	310	39	learned	learn	VERB
dad-10683	310	40	model	model	NOUN
dad-10683	310	41	parameters	parameter	NOUN
dad-10683	310	42	.	.	PUNCT
dad-10683	311	1	this	this	PRON
dad-10683	311	2	can	can	AUX
dad-10683	311	3	be	be	AUX
dad-10683	311	4	thought	think	VERB
dad-10683	311	5	of	of	ADP
dad-10683	311	6	as	as	ADP
dad-10683	311	7	a	a	DET
dad-10683	311	8	generative	generative	ADJ
dad-10683	311	9	approach	approach	NOUN
dad-10683	311	10	;	;	PUNCT
dad-10683	311	11	given	give	VERB
dad-10683	311	12	some	some	DET
dad-10683	311	13	input	input	NOUN
dad-10683	311	14	response	response	NOUN
dad-10683	311	15	,	,	PUNCT
dad-10683	311	16	we	we	PRON
dad-10683	311	17	generate	generate	VERB
dad-10683	311	18	a	a	DET
dad-10683	311	19	context	context	NOUN
dad-10683	311	20	with	with	ADP
dad-10683	311	21	the	the	DET
dad-10683	311	22	product	product	NOUN
dad-10683	311	23	c′	c′	NOUN
dad-10683	311	24	=	=	SYM
dad-10683	311	25	mr	mr	PROPN
dad-10683	311	26	,	,	PUNCT
dad-10683	311	27	and	and	CCONJ
dad-10683	311	28	measure	measure	VERB
dad-10683	311	29	the	the	DET
dad-10683	311	30	similarity	similarity	NOUN
dad-10683	311	31	to	to	ADP
dad-10683	311	32	the	the	DET
dad-10683	311	33	actual	actual	ADJ
dad-10683	311	34	context	context	NOUN
dad-10683	311	35	using	use	VERB
dad-10683	311	36	the	the	DET
dad-10683	311	37	dot	dot	NOUN
dad-10683	311	38	product	product	NOUN
dad-10683	311	39	.	.	PUNCT
dad-10683	312	1	this	this	PRON
dad-10683	312	2	is	be	AUX
dad-10683	312	3	converted	convert	VERB
dad-10683	312	4	to	to	ADP
dad-10683	312	5	a	a	DET
dad-10683	312	6	probability	probability	NOUN
dad-10683	312	7	with	with	ADP
dad-10683	312	8	the	the	DET
dad-10683	312	9	sigmoid	sigmoid	NOUN
dad-10683	312	10	function	function	NOUN
dad-10683	312	11	.	.	PUNCT
dad-10683	313	1	the	the	DET
dad-10683	313	2	model	model	NOUN
dad-10683	313	3	is	be	AUX
dad-10683	313	4	trained	train	VERB
dad-10683	313	5	by	by	ADP
dad-10683	313	6	minimizing	minimize	VERB
dad-10683	313	7	the	the	DET
dad-10683	313	8	cross	cross	NOUN
dad-10683	313	9	entropy	entropy	NOUN
dad-10683	313	10	of	of	ADP
dad-10683	313	11	all	all	DET
dad-10683	313	12	labeled	label	VERB
dad-10683	313	13	(	(	PUNCT
dad-10683	313	14	context	context	ADJ
dad-10683	313	15	,	,	PUNCT
dad-10683	313	16	response	response	NOUN
dad-10683	313	17	)	)	PUNCT
dad-10683	313	18	pairs	pair	NOUN
dad-10683	313	19	(	(	PUNCT
dad-10683	313	20	yu	yu	PROPN
dad-10683	313	21	et	et	PROPN
dad-10683	313	22	al	al	PROPN
dad-10683	313	23	.	.	PROPN
dad-10683	313	24	,	,	PUNCT
dad-10683	313	25	2014	2014	NUM
dad-10683	313	26	):	):	PUNCT
dad-10683	313	27	l	l	NOUN
dad-10683	313	28	=	=	SYM
dad-10683	314	1	−	−	PROPN
dad-10683	314	2	∑	∑	PUNCT
dad-10683	314	3	n	n	ADP
dad-10683	314	4	log	log	NOUN
dad-10683	314	5	p(flagn|cn	p(flagn|cn	NOUN
dad-10683	314	6	,	,	PUNCT
dad-10683	314	7	rn	rn	PROPN
dad-10683	314	8	,	,	PUNCT
dad-10683	314	9	m	m	PROPN
dad-10683	314	10	)	)	PUNCT
dad-10683	314	11	(	(	PUNCT
dad-10683	314	12	4	4	X
dad-10683	314	13	)	)	PUNCT
dad-10683	314	14	where	where	SCONJ
dad-10683	314	15	||θ||2f	||θ||2f	NOUN
dad-10683	314	16	is	be	AUX
dad-10683	314	17	the	the	DET
dad-10683	314	18	frobenius	frobenius	ADJ
dad-10683	314	19	norm	norm	NOUN
dad-10683	314	20	of	of	ADP
dad-10683	314	21	θ	θ	PROPN
dad-10683	314	22	=	=	PUNCT
dad-10683	314	23	{	{	PUNCT
dad-10683	314	24	m	m	PROPN
dad-10683	314	25	,	,	PUNCT
dad-10683	314	26	b	b	NOUN
dad-10683	314	27	}	}	PUNCT
dad-10683	314	28	.	.	PUNCT
dad-10683	315	1	a	a	DET
dad-10683	315	2	diagram	diagram	NOUN
dad-10683	315	3	of	of	ADP
dad-10683	315	4	the	the	DET
dad-10683	315	5	de	de	PROPN
dad-10683	315	6	model	model	NOUN
dad-10683	315	7	can	can	AUX
dad-10683	315	8	be	be	AUX
dad-10683	315	9	seen	see	VERB
dad-10683	315	10	in	in	ADP
dad-10683	315	11	figure	figure	NOUN
dad-10683	315	12	5	5	NUM
dad-10683	315	13	.	.	PUNCT
dad-10683	316	1	for	for	ADP
dad-10683	316	2	training	training	NOUN
dad-10683	316	3	,	,	PUNCT
dad-10683	316	4	we	we	PRON
dad-10683	316	5	used	use	VERB
dad-10683	316	6	a	a	DET
dad-10683	316	7	1:1	1:1	NUM
dad-10683	316	8	ratio	ratio	NOUN
dad-10683	316	9	between	between	ADP
dad-10683	316	10	true	true	ADJ
dad-10683	316	11	responses	response	NOUN
dad-10683	316	12	(	(	PUNCT
dad-10683	316	13	flag	flag	NOUN
dad-10683	316	14	=	=	SYM
dad-10683	316	15	1	1	NUM
dad-10683	316	16	)	)	PUNCT
dad-10683	316	17	,	,	PUNCT
dad-10683	316	18	and	and	CCONJ
dad-10683	316	19	negative	negative	ADJ
dad-10683	316	20	responses	response	NOUN
dad-10683	316	21	(	(	PUNCT
dad-10683	316	22	flag=0	flag=0	NOUN
dad-10683	316	23	)	)	PUNCT
dad-10683	316	24	drawn	draw	VERB
dad-10683	316	25	randomly	randomly	ADV
dad-10683	316	26	from	from	ADP
dad-10683	316	27	elsewhere	elsewhere	ADV
dad-10683	316	28	in	in	ADP
dad-10683	316	29	the	the	DET
dad-10683	316	30	training	training	NOUN
dad-10683	316	31	set	set	NOUN
dad-10683	316	32	.	.	PUNCT
dad-10683	317	1	the	the	DET
dad-10683	317	2	rnn	rnn	NOUN
dad-10683	317	3	architecture	architecture	NOUN
dad-10683	317	4	is	be	AUX
dad-10683	317	5	set	set	VERB
dad-10683	317	6	to	to	ADP
dad-10683	317	7	1	1	NUM
dad-10683	317	8	hidden	hidden	ADJ
dad-10683	317	9	layer	layer	NOUN
dad-10683	317	10	with	with	ADP
dad-10683	317	11	100	100	NUM
dad-10683	317	12	neurons	neuron	NOUN
dad-10683	317	13	(	(	PUNCT
dad-10683	317	14	optimized	optimize	VERB
dad-10683	317	15	over	over	ADP
dad-10683	317	16	{	{	PUNCT
dad-10683	317	17	10	10	NUM
dad-10683	317	18	,	,	PUNCT
dad-10683	317	19	50	50	NUM
dad-10683	317	20	,	,	PUNCT
dad-10683	317	21	100	100	NUM
dad-10683	317	22	,	,	PUNCT
dad-10683	317	23	200	200	NUM
dad-10683	317	24	,	,	PUNCT
dad-10683	317	25	300	300	NUM
dad-10683	317	26	}	}	PUNCT
dad-10683	317	27	)	)	PUNCT
dad-10683	317	28	,	,	PUNCT
dad-10683	317	29	and	and	CCONJ
dad-10683	317	30	a	a	DET
dad-10683	317	31	learning	learning	NOUN
dad-10683	317	32	rate	rate	NOUN
dad-10683	317	33	of	of	ADP
dad-10683	317	34	0.0001	0.0001	NUM
dad-10683	317	35	(	(	PUNCT
dad-10683	317	36	optimized	optimize	VERB
dad-10683	317	37	over	over	ADP
dad-10683	317	38	{	{	PUNCT
dad-10683	317	39	0.1	0.1	NUM
dad-10683	317	40	,	,	PUNCT
dad-10683	317	41	0.01	0.01	NUM
dad-10683	317	42	,	,	PUNCT
dad-10683	317	43	0.001	0.001	NUM
dad-10683	317	44	,	,	PUNCT
dad-10683	317	45	0.0001	0.0001	NUM
dad-10683	317	46	}	}	PUNCT
dad-10683	317	47	)	)	PUNCT
dad-10683	317	48	.	.	PUNCT
dad-10683	318	1	the	the	DET
dad-10683	318	2	w	w	PROPN
dad-10683	318	3	h	h	NOUN
dad-10683	318	4	matrix	matrix	NOUN
dad-10683	318	5	is	be	AUX
dad-10683	318	6	initialized	initialize	VERB
dad-10683	318	7	using	use	VERB
dad-10683	318	8	orthogonal	orthogonal	ADJ
dad-10683	318	9	weights	weight	NOUN
dad-10683	318	10	(	(	PUNCT
dad-10683	318	11	saxe	saxe	NOUN
dad-10683	318	12	et	et	PROPN
dad-10683	318	13	al	al	PROPN
dad-10683	318	14	.	.	PROPN
dad-10683	318	15	,	,	PUNCT
dad-10683	318	16	2013	2013	NUM
dad-10683	318	17	)	)	PUNCT
dad-10683	318	18	,	,	PUNCT
dad-10683	318	19	while	while	SCONJ
dad-10683	318	20	w	w	NOUN
dad-10683	318	21	x	x	VERB
dad-10683	318	22	is	be	AUX
dad-10683	318	23	initialized	initialize	VERB
dad-10683	318	24	using	use	VERB
dad-10683	318	25	a	a	DET
dad-10683	318	26	uniform	uniform	ADJ
dad-10683	318	27	distribution	distribution	NOUN
dad-10683	318	28	with	with	ADP
dad-10683	318	29	values	value	NOUN
dad-10683	318	30	between	between	ADP
dad-10683	318	31	-0.01	-0.01	NUM
dad-10683	318	32	and	and	CCONJ
dad-10683	318	33	0.01	0.01	NUM
dad-10683	318	34	.	.	PUNCT
dad-10683	319	1	we	we	PRON
dad-10683	319	2	use	use	VERB
dad-10683	319	3	the	the	DET
dad-10683	319	4	first	first	ADJ
dad-10683	319	5	-	-	PUNCT
dad-10683	319	6	order	order	NOUN
dad-10683	319	7	stochastic	stochastic	ADJ
dad-10683	319	8	gradient	gradient	NOUN
dad-10683	319	9	optimization	optimization	NOUN
dad-10683	319	10	procedure	procedure	NOUN
dad-10683	319	11	adam	adam	PROPN
dad-10683	319	12	(	(	PUNCT
dad-10683	319	13	kingma	kingma	PROPN
dad-10683	319	14	and	and	CCONJ
dad-10683	319	15	ba	ba	PROPN
dad-10683	319	16	,	,	PUNCT
dad-10683	319	17	2014	2014	NUM
dad-10683	319	18	)	)	PUNCT
dad-10683	319	19	with	with	ADP
dad-10683	319	20	the	the	DET
dad-10683	319	21	default	default	NOUN
dad-10683	319	22	parameters	parameter	NOUN
dad-10683	319	23	,	,	PUNCT
dad-10683	319	24	using	use	VERB
dad-10683	319	25	gradients	gradient	NOUN
dad-10683	319	26	clipped	clip	VERB
dad-10683	319	27	to	to	ADP
dad-10683	319	28	10	10	NUM
dad-10683	319	29	and	and	CCONJ
dad-10683	319	30	a	a	DET
dad-10683	319	31	batch	batch	NOUN
dad-10683	319	32	size	size	NOUN
dad-10683	319	33	of	of	ADP
dad-10683	319	34	512	512	NUM
dad-10683	319	35	(	(	PUNCT
dad-10683	319	36	optimized	optimize	VERB
dad-10683	319	37	over	over	ADP
dad-10683	319	38	{	{	PUNCT
dad-10683	319	39	128	128	NUM
dad-10683	319	40	,	,	PUNCT
dad-10683	319	41	256	256	NUM
dad-10683	319	42	,	,	PUNCT
dad-10683	319	43	512	512	NUM
dad-10683	319	44	}	}	PUNCT
dad-10683	319	45	)	)	PUNCT
dad-10683	319	46	.	.	PUNCT
dad-10683	320	1	we	we	PRON
dad-10683	320	2	found	find	VERB
dad-10683	320	3	that	that	SCONJ
dad-10683	320	4	weight	weight	NOUN
dad-10683	320	5	initialization	initialization	NOUN
dad-10683	320	6	as	as	ADV
dad-10683	320	7	well	well	ADV
dad-10683	320	8	as	as	ADP
dad-10683	320	9	the	the	DET
dad-10683	320	10	choice	choice	NOUN
dad-10683	320	11	of	of	ADP
dad-10683	320	12	optimizer	optimizer	NOUN
dad-10683	320	13	were	be	AUX
dad-10683	320	14	critical	critical	ADJ
dad-10683	320	15	for	for	ADP
dad-10683	320	16	training	train	VERB
dad-10683	320	17	the	the	DET
dad-10683	320	18	rnns	rnn	NOUN
dad-10683	320	19	.	.	PUNCT
dad-10683	321	1	4.3	4.3	NUM
dad-10683	321	2	lstm	lstm	ADJ
dad-10683	321	3	dual	dual	ADJ
dad-10683	321	4	encoder	encoder	NOUN
dad-10683	321	5	in	in	ADP
dad-10683	321	6	addition	addition	NOUN
dad-10683	321	7	to	to	ADP
dad-10683	321	8	the	the	DET
dad-10683	321	9	rnn	rnn	NOUN
dad-10683	321	10	model	model	NOUN
dad-10683	321	11	,	,	PUNCT
dad-10683	321	12	we	we	PRON
dad-10683	321	13	consider	consider	VERB
dad-10683	321	14	the	the	DET
dad-10683	321	15	same	same	ADJ
dad-10683	321	16	architecture	architecture	NOUN
dad-10683	321	17	but	but	CCONJ
dad-10683	321	18	change	change	VERB
dad-10683	321	19	the	the	DET
dad-10683	321	20	hidden	hide	VERB
dad-10683	321	21	units	unit	NOUN
dad-10683	321	22	to	to	ADP
dad-10683	321	23	long	long	ADJ
dad-10683	321	24	-	-	PUNCT
dad-10683	321	25	short	short	ADJ
dad-10683	321	26	term	term	NOUN
dad-10683	321	27	memory	memory	NOUN
dad-10683	321	28	(	(	PUNCT
dad-10683	321	29	lstm	lstm	NOUN
dad-10683	321	30	)	)	PUNCT
dad-10683	321	31	units	unit	NOUN
dad-10683	321	32	(	(	PUNCT
dad-10683	321	33	hochreiter	hochreiter	PROPN
dad-10683	321	34	and	and	CCONJ
dad-10683	321	35	schmidhuber	schmidhuber	NOUN
dad-10683	321	36	,	,	PUNCT
dad-10683	321	37	1997	1997	NUM
dad-10683	321	38	)	)	PUNCT
dad-10683	321	39	.	.	PUNCT
dad-10683	322	1	lstms	lstms	PROPN
dad-10683	322	2	were	be	AUX
dad-10683	322	3	introduced	introduce	VERB
dad-10683	322	4	in	in	ADP
dad-10683	322	5	order	order	NOUN
dad-10683	322	6	to	to	PART
dad-10683	322	7	model	model	VERB
dad-10683	322	8	longer	long	ADJ
dad-10683	322	9	-	-	PUNCT
dad-10683	322	10	term	term	NOUN
dad-10683	322	11	dependencies	dependency	NOUN
dad-10683	322	12	.	.	PUNCT
dad-10683	323	1	this	this	PRON
dad-10683	323	2	is	be	AUX
dad-10683	323	3	accomplished	accomplish	VERB
dad-10683	323	4	using	use	VERB
dad-10683	323	5	a	a	DET
dad-10683	323	6	series	series	NOUN
dad-10683	323	7	of	of	ADP
dad-10683	323	8	gates	gate	NOUN
dad-10683	323	9	that	that	PRON
dad-10683	323	10	determine	determine	VERB
dad-10683	323	11	whether	whether	SCONJ
dad-10683	323	12	a	a	DET
dad-10683	323	13	new	new	ADJ
dad-10683	323	14	input	input	NOUN
dad-10683	323	15	should	should	AUX
dad-10683	323	16	be	be	AUX
dad-10683	323	17	remembered	remember	VERB
dad-10683	323	18	,	,	PUNCT
dad-10683	323	19	forgotten	forget	VERB
dad-10683	323	20	(	(	PUNCT
dad-10683	323	21	and	and	CCONJ
dad-10683	323	22	the	the	DET
dad-10683	323	23	old	old	ADJ
dad-10683	323	24	value	value	NOUN
dad-10683	323	25	retained	retain	VERB
dad-10683	323	26	)	)	PUNCT
dad-10683	323	27	,	,	PUNCT
dad-10683	323	28	or	or	CCONJ
dad-10683	323	29	used	use	VERB
dad-10683	323	30	as	as	ADP
dad-10683	323	31	output	output	NOUN
dad-10683	323	32	.	.	PUNCT
dad-10683	324	1	the	the	DET
dad-10683	324	2	error	error	NOUN
dad-10683	324	3	signal	signal	NOUN
dad-10683	324	4	can	can	AUX
dad-10683	324	5	now	now	ADV
dad-10683	324	6	be	be	AUX
dad-10683	324	7	propagated	propagate	VERB
dad-10683	324	8	back	back	ADV
dad-10683	324	9	much	much	ADV
dad-10683	324	10	further	far	ADV
dad-10683	324	11	using	use	VERB
dad-10683	324	12	the	the	DET
dad-10683	324	13	gates	gate	NOUN
dad-10683	324	14	of	of	ADP
dad-10683	324	15	the	the	DET
dad-10683	324	16	lstm	lstm	PROPN
dad-10683	324	17	unit	unit	NOUN
dad-10683	324	18	.	.	PUNCT
dad-10683	325	1	this	this	PRON
dad-10683	325	2	helps	help	VERB
dad-10683	325	3	overcome	overcome	VERB
dad-10683	325	4	the	the	DET
dad-10683	325	5	vanishing	vanish	VERB
dad-10683	325	6	gradient	gradient	NOUN
dad-10683	325	7	and	and	CCONJ
dad-10683	325	8	exploding	explode	VERB
dad-10683	325	9	gradient	gradient	NOUN
dad-10683	325	10	problems	problem	NOUN
dad-10683	325	11	in	in	ADP
dad-10683	325	12	standard	standard	ADJ
dad-10683	325	13	rnns	rnn	NOUN
dad-10683	325	14	,	,	PUNCT
dad-10683	325	15	where	where	SCONJ
dad-10683	325	16	the	the	DET
dad-10683	325	17	error	error	NOUN
dad-10683	325	18	gradients	gradient	NOUN
dad-10683	325	19	would	would	AUX
dad-10683	325	20	otherwise	otherwise	ADV
dad-10683	325	21	decrease	decrease	VERB
dad-10683	325	22	or	or	CCONJ
dad-10683	325	23	increase	increase	VERB
dad-10683	325	24	at	at	ADP
dad-10683	325	25	an	an	DET
dad-10683	325	26	exponential	exponential	ADJ
dad-10683	325	27	rate	rate	NOUN
dad-10683	325	28	.	.	PUNCT
dad-10683	326	1	for	for	ADP
dad-10683	326	2	this	this	DET
dad-10683	326	3	model	model	NOUN
dad-10683	326	4	,	,	PUNCT
dad-10683	326	5	we	we	PRON
dad-10683	326	6	used	use	VERB
dad-10683	326	7	1	1	NUM
dad-10683	326	8	hidden	hidden	ADJ
dad-10683	326	9	layer	layer	NOUN
dad-10683	326	10	with	with	ADP
dad-10683	326	11	200	200	NUM
dad-10683	326	12	neurons	neuron	NOUN
dad-10683	326	13	,	,	PUNCT
dad-10683	326	14	a	a	DET
dad-10683	326	15	learning	learning	NOUN
dad-10683	326	16	rate	rate	NOUN
dad-10683	326	17	of	of	ADP
dad-10683	326	18	0.001	0.001	NUM
dad-10683	326	19	,	,	PUNCT
dad-10683	326	20	and	and	CCONJ
dad-10683	326	21	a	a	DET
dad-10683	326	22	batch	batch	NOUN
dad-10683	326	23	size	size	NOUN
dad-10683	326	24	of	of	ADP
dad-10683	326	25	256	256	NUM
dad-10683	326	26	(	(	PUNCT
dad-10683	326	27	optimized	optimize	VERB
dad-10683	326	28	over	over	ADP
dad-10683	326	29	the	the	DET
dad-10683	326	30	same	same	ADJ
dad-10683	326	31	values	value	NOUN
dad-10683	326	32	as	as	ADP
dad-10683	326	33	the	the	DET
dad-10683	326	34	rnn	rnn	NOUN
dad-10683	326	35	)	)	PUNCT
dad-10683	326	36	.	.	PUNCT
dad-10683	327	1	we	we	PRON
dad-10683	327	2	again	again	ADV
dad-10683	327	3	use	use	VERB
dad-10683	327	4	the	the	DET
dad-10683	327	5	default	default	NOUN
dad-10683	327	6	adam	adam	PROPN
dad-10683	327	7	settings	setting	NOUN
dad-10683	327	8	,	,	PUNCT
dad-10683	327	9	44	44	NUM
dad-10683	327	10	training	training	NOUN
dad-10683	327	11	end	end	NOUN
dad-10683	327	12	-	-	PUNCT
dad-10683	327	13	to	to	ADP
dad-10683	327	14	-	-	PUNCT
dad-10683	327	15	end	end	NOUN
dad-10683	327	16	dialogue	dialogue	NOUN
dad-10683	327	17	systems	system	NOUN
dad-10683	327	18	and	and	CCONJ
dad-10683	327	19	initialize	initialize	VERB
dad-10683	327	20	the	the	DET
dad-10683	327	21	forget	forget	NOUN
dad-10683	327	22	gate	gate	NOUN
dad-10683	327	23	bias	bias	NOUN
dad-10683	327	24	of	of	ADP
dad-10683	327	25	the	the	DET
dad-10683	327	26	lstm	lstm	NOUN
dad-10683	327	27	to	to	ADP
dad-10683	327	28	2.0	2.0	NUM
dad-10683	327	29	.	.	PUNCT
dad-10683	328	1	the	the	DET
dad-10683	328	2	hyper	hyper	ADJ
dad-10683	328	3	-	-	ADJ
dad-10683	328	4	parameter	parameter	NOUN
dad-10683	328	5	configuration	configuration	NOUN
dad-10683	328	6	(	(	PUNCT
dad-10683	328	7	including	include	VERB
dad-10683	328	8	number	number	NOUN
dad-10683	328	9	of	of	ADP
dad-10683	328	10	neurons	neuron	NOUN
dad-10683	328	11	)	)	PUNCT
dad-10683	328	12	was	be	AUX
dad-10683	328	13	optimized	optimize	VERB
dad-10683	328	14	independently	independently	ADV
dad-10683	328	15	for	for	ADP
dad-10683	328	16	rnns	rnns	NOUN
dad-10683	328	17	and	and	CCONJ
dad-10683	328	18	lstms	lstms	NOUN
dad-10683	328	19	using	use	VERB
dad-10683	328	20	a	a	DET
dad-10683	328	21	separate	separate	ADJ
dad-10683	328	22	validation	validation	NOUN
dad-10683	328	23	set	set	NOUN
dad-10683	328	24	extracted	extract	VERB
dad-10683	328	25	from	from	ADP
dad-10683	328	26	the	the	DET
dad-10683	328	27	training	training	NOUN
dad-10683	328	28	data	datum	NOUN
dad-10683	328	29	.	.	PUNCT
dad-10683	329	1	4.4	4.4	NUM
dad-10683	329	2	evaluation	evaluation	NOUN
dad-10683	329	3	metrics	metric	NOUN
dad-10683	329	4	we	we	PRON
dad-10683	329	5	consider	consider	VERB
dad-10683	329	6	two	two	NUM
dad-10683	329	7	types	type	NOUN
dad-10683	329	8	of	of	ADP
dad-10683	329	9	evaluation	evaluation	NOUN
dad-10683	329	10	metrics	metric	NOUN
dad-10683	329	11	:	:	PUNCT
dad-10683	329	12	retrieval	retrieval	NOUN
dad-10683	329	13	metrics	metric	NOUN
dad-10683	329	14	,	,	PUNCT
dad-10683	329	15	and	and	CCONJ
dad-10683	329	16	generative	generative	ADJ
dad-10683	329	17	metrics	metric	NOUN
dad-10683	329	18	.	.	PUNCT
dad-10683	330	1	these	these	DET
dad-10683	330	2	metrics	metric	NOUN
dad-10683	330	3	are	be	AUX
dad-10683	330	4	applicable	applicable	ADJ
dad-10683	330	5	to	to	ADP
dad-10683	330	6	both	both	DET
dad-10683	330	7	models	model	NOUN
dad-10683	330	8	trained	train	VERB
dad-10683	330	9	on	on	ADP
dad-10683	330	10	the	the	DET
dad-10683	330	11	task	task	NOUN
dad-10683	330	12	of	of	ADP
dad-10683	330	13	nuc	nuc	NOUN
dad-10683	330	14	,	,	PUNCT
dad-10683	330	15	detailed	detail	VERB
dad-10683	330	16	in	in	ADP
dad-10683	330	17	this	this	DET
dad-10683	330	18	section	section	NOUN
dad-10683	330	19	,	,	PUNCT
dad-10683	330	20	and	and	CCONJ
dad-10683	330	21	the	the	DET
dad-10683	330	22	generative	generative	ADJ
dad-10683	330	23	models	model	NOUN
dad-10683	330	24	introduced	introduce	VERB
dad-10683	330	25	in	in	ADP
dad-10683	330	26	section	section	NOUN
dad-10683	330	27	5	5	NUM
dad-10683	330	28	.	.	PUNCT
dad-10683	331	1	in	in	ADP
dad-10683	331	2	particular	particular	ADJ
dad-10683	331	3	,	,	PUNCT
dad-10683	331	4	they	they	PRON
dad-10683	331	5	offer	offer	VERB
dad-10683	331	6	two	two	NUM
dad-10683	331	7	ways	way	NOUN
dad-10683	331	8	of	of	ADP
dad-10683	331	9	automatically	automatically	ADV
dad-10683	331	10	evaluating	evaluate	VERB
dad-10683	331	11	dialogue	dialogue	NOUN
dad-10683	331	12	systems	system	NOUN
dad-10683	331	13	trained	train	VERB
dad-10683	331	14	in	in	ADP
dad-10683	331	15	an	an	DET
dad-10683	331	16	end	end	NOUN
dad-10683	331	17	-	-	PUNCT
dad-10683	331	18	to	to	ADP
dad-10683	331	19	-	-	PUNCT
dad-10683	331	20	end	end	NOUN
dad-10683	331	21	manner	manner	NOUN
dad-10683	331	22	.	.	PUNCT
dad-10683	332	1	for	for	ADP
dad-10683	332	2	retrieval	retrieval	NOUN
dad-10683	332	3	,	,	PUNCT
dad-10683	332	4	we	we	PRON
dad-10683	332	5	evaluate	evaluate	VERB
dad-10683	332	6	using	use	VERB
dad-10683	332	7	recall@k	recall@k	NOUN
dad-10683	332	8	(	(	PUNCT
dad-10683	332	9	denoted	denote	VERB
dad-10683	332	10	r@1	r@1	PROPN
dad-10683	332	11	r@2	r@2	PROPN
dad-10683	332	12	,	,	PUNCT
dad-10683	332	13	r@5	r@5	PROPN
dad-10683	332	14	below	below	ADV
dad-10683	332	15	)	)	PUNCT
dad-10683	332	16	,	,	PUNCT
dad-10683	332	17	which	which	PRON
dad-10683	332	18	has	have	AUX
dad-10683	332	19	often	often	ADV
dad-10683	332	20	been	be	AUX
dad-10683	332	21	used	use	VERB
dad-10683	332	22	in	in	ADP
dad-10683	332	23	language	language	NOUN
dad-10683	332	24	tasks	task	NOUN
dad-10683	332	25	.	.	PUNCT
dad-10683	333	1	here	here	ADV
dad-10683	333	2	the	the	DET
dad-10683	333	3	agent	agent	NOUN
dad-10683	333	4	is	be	AUX
dad-10683	333	5	asked	ask	VERB
dad-10683	333	6	to	to	PART
dad-10683	333	7	select	select	VERB
dad-10683	333	8	the	the	DET
dad-10683	333	9	k	k	PROPN
dad-10683	333	10	most	most	ADV
dad-10683	333	11	likely	likely	ADJ
dad-10683	333	12	responses	response	NOUN
dad-10683	333	13	,	,	PUNCT
dad-10683	333	14	and	and	CCONJ
dad-10683	333	15	it	it	PRON
dad-10683	333	16	is	be	AUX
dad-10683	333	17	correct	correct	ADJ
dad-10683	333	18	if	if	SCONJ
dad-10683	333	19	the	the	DET
dad-10683	333	20	true	true	ADJ
dad-10683	333	21	response	response	NOUN
dad-10683	333	22	is	be	AUX
dad-10683	333	23	among	among	ADP
dad-10683	333	24	these	these	DET
dad-10683	333	25	k	k	PROPN
dad-10683	333	26	candidates	candidate	NOUN
dad-10683	333	27	.	.	PUNCT
dad-10683	334	1	only	only	ADV
dad-10683	334	2	the	the	DET
dad-10683	334	3	r@1	r@1	PROPN
dad-10683	334	4	metric	metric	NOUN
dad-10683	334	5	is	be	AUX
dad-10683	334	6	relevant	relevant	ADJ
dad-10683	334	7	in	in	ADP
dad-10683	334	8	the	the	DET
dad-10683	334	9	case	case	NOUN
dad-10683	334	10	of	of	ADP
dad-10683	334	11	binary	binary	ADJ
dad-10683	334	12	classification	classification	NOUN
dad-10683	334	13	(	(	PUNCT
dad-10683	334	14	as	as	ADP
dad-10683	334	15	in	in	ADP
dad-10683	334	16	the	the	DET
dad-10683	334	17	table	table	NOUN
dad-10683	334	18	2	2	NUM
dad-10683	334	19	example	example	NOUN
dad-10683	334	20	)	)	PUNCT
dad-10683	334	21	.	.	PUNCT
dad-10683	335	1	although	although	SCONJ
dad-10683	335	2	a	a	DET
dad-10683	335	3	language	language	NOUN
dad-10683	335	4	model	model	NOUN
dad-10683	335	5	that	that	PRON
dad-10683	335	6	performs	perform	VERB
dad-10683	335	7	well	well	ADV
dad-10683	335	8	on	on	ADP
dad-10683	335	9	these	these	DET
dad-10683	335	10	retrieval	retrieval	NOUN
dad-10683	335	11	metrics	metric	NOUN
dad-10683	335	12	is	be	AUX
dad-10683	335	13	not	not	PART
dad-10683	335	14	guaranteed	guarantee	VERB
dad-10683	335	15	to	to	PART
dad-10683	335	16	achieve	achieve	VERB
dad-10683	335	17	good	good	ADJ
dad-10683	335	18	performance	performance	NOUN
dad-10683	335	19	on	on	ADP
dad-10683	335	20	utterance	utterance	NOUN
dad-10683	335	21	generation	generation	NOUN
dad-10683	335	22	,	,	PUNCT
dad-10683	335	23	we	we	PRON
dad-10683	335	24	hypothesize	hypothesize	VERB
dad-10683	335	25	that	that	SCONJ
dad-10683	335	26	improvements	improvement	NOUN
dad-10683	335	27	on	on	ADP
dad-10683	335	28	a	a	DET
dad-10683	335	29	model	model	NOUN
dad-10683	335	30	with	with	ADP
dad-10683	335	31	regards	regard	NOUN
dad-10683	335	32	to	to	ADP
dad-10683	335	33	the	the	DET
dad-10683	335	34	classification	classification	NOUN
dad-10683	335	35	task	task	NOUN
dad-10683	335	36	will	will	AUX
dad-10683	335	37	eventually	eventually	ADV
dad-10683	335	38	lead	lead	VERB
dad-10683	335	39	to	to	ADP
dad-10683	335	40	improvements	improvement	NOUN
dad-10683	335	41	for	for	ADP
dad-10683	335	42	the	the	DET
dad-10683	335	43	generation	generation	NOUN
dad-10683	335	44	task	task	NOUN
dad-10683	335	45	.	.	PUNCT
dad-10683	336	1	see	see	VERB
dad-10683	336	2	section	section	NOUN
dad-10683	336	3	6	6	NUM
dad-10683	336	4	for	for	ADP
dad-10683	336	5	further	further	ADJ
dad-10683	336	6	discussion	discussion	NOUN
dad-10683	336	7	of	of	ADP
dad-10683	336	8	this	this	DET
dad-10683	336	9	point	point	NOUN
dad-10683	336	10	.	.	PUNCT
dad-10683	337	1	we	we	PRON
dad-10683	337	2	also	also	ADV
dad-10683	337	3	consider	consider	VERB
dad-10683	337	4	generative	generative	ADJ
dad-10683	337	5	metrics	metric	NOUN
dad-10683	337	6	that	that	PRON
dad-10683	337	7	compare	compare	VERB
dad-10683	337	8	the	the	DET
dad-10683	337	9	generated	generate	VERB
dad-10683	337	10	or	or	CCONJ
dad-10683	337	11	retrieved	retrieve	VERB
dad-10683	337	12	utterance	utterance	NOUN
dad-10683	337	13	to	to	ADP
dad-10683	337	14	the	the	DET
dad-10683	337	15	ground	ground	NOUN
dad-10683	337	16	-	-	PUNCT
dad-10683	337	17	truth	truth	NOUN
dad-10683	337	18	next	next	ADP
dad-10683	337	19	utterance	utterance	NOUN
dad-10683	337	20	.	.	PUNCT
dad-10683	338	1	in	in	ADP
dad-10683	338	2	general	general	ADJ
dad-10683	338	3	,	,	PUNCT
dad-10683	338	4	this	this	PRON
dad-10683	338	5	is	be	AUX
dad-10683	338	6	a	a	DET
dad-10683	338	7	hard	hard	ADJ
dad-10683	338	8	open	open	ADJ
dad-10683	338	9	problem	problem	NOUN
dad-10683	338	10	(	(	PUNCT
dad-10683	338	11	liu	liu	PROPN
dad-10683	338	12	et	et	PROPN
dad-10683	338	13	al	al	PROPN
dad-10683	338	14	.	.	PROPN
dad-10683	338	15	,	,	PUNCT
dad-10683	338	16	2016	2016	NUM
dad-10683	338	17	)	)	PUNCT
dad-10683	338	18	.	.	PUNCT
dad-10683	339	1	we	we	PRON
dad-10683	339	2	use	use	VERB
dad-10683	339	3	methods	method	NOUN
dad-10683	339	4	based	base	VERB
dad-10683	339	5	on	on	ADP
dad-10683	339	6	word	word	NOUN
dad-10683	339	7	embeddings	embedding	NOUN
dad-10683	339	8	that	that	PRON
dad-10683	339	9	have	have	AUX
dad-10683	339	10	recently	recently	ADV
dad-10683	339	11	been	be	AUX
dad-10683	339	12	proposed	propose	VERB
dad-10683	339	13	for	for	ADP
dad-10683	339	14	use	use	NOUN
dad-10683	339	15	in	in	ADP
dad-10683	339	16	evaluating	evaluate	VERB
dad-10683	339	17	nontask	nontask	NOUN
dad-10683	339	18	oriented	orient	VERB
dad-10683	339	19	dialogue	dialogue	NOUN
dad-10683	339	20	systems	system	NOUN
dad-10683	339	21	,	,	PUNCT
dad-10683	339	22	when	when	SCONJ
dad-10683	339	23	no	no	DET
dad-10683	339	24	task	task	NOUN
dad-10683	339	25	completion	completion	NOUN
dad-10683	339	26	signal	signal	NOUN
dad-10683	339	27	is	be	AUX
dad-10683	339	28	available	available	ADJ
dad-10683	339	29	.	.	PUNCT
dad-10683	340	1	these	these	DET
dad-10683	340	2	metrics	metric	NOUN
dad-10683	340	3	use	use	VERB
dad-10683	340	4	external	external	ADJ
dad-10683	340	5	word	word	NOUN
dad-10683	340	6	embeddings	embedding	NOUN
dad-10683	340	7	trained	train	VERB
dad-10683	340	8	via	via	ADP
dad-10683	340	9	distributional	distributional	ADJ
dad-10683	340	10	semantics	semantic	NOUN
dad-10683	340	11	,	,	PUNCT
dad-10683	340	12	such	such	ADJ
dad-10683	340	13	as	as	ADP
dad-10683	340	14	glove	glove	NOUN
dad-10683	340	15	,	,	PUNCT
dad-10683	340	16	to	to	PART
dad-10683	340	17	determine	determine	VERB
dad-10683	340	18	how	how	SCONJ
dad-10683	340	19	close	close	ADJ
dad-10683	340	20	the	the	DET
dad-10683	340	21	generated	generate	VERB
dad-10683	340	22	utterance	utterance	NOUN
dad-10683	340	23	is	be	AUX
dad-10683	340	24	to	to	ADP
dad-10683	340	25	the	the	DET
dad-10683	340	26	ground	ground	NOUN
dad-10683	340	27	truth	truth	NOUN
dad-10683	340	28	next	next	ADP
dad-10683	340	29	utterance	utterance	ADV
dad-10683	340	30	.	.	PUNCT
dad-10683	341	1	we	we	PRON
dad-10683	341	2	note	note	VERB
dad-10683	341	3	that	that	SCONJ
dad-10683	341	4	these	these	DET
dad-10683	341	5	metrics	metric	NOUN
dad-10683	341	6	do	do	AUX
dad-10683	341	7	not	not	PART
dad-10683	341	8	necessarily	necessarily	ADV
dad-10683	341	9	correlate	correlate	VERB
dad-10683	341	10	strongly	strongly	ADV
dad-10683	341	11	with	with	ADP
dad-10683	341	12	human	human	ADJ
dad-10683	341	13	judgement	judgement	NOUN
dad-10683	341	14	(	(	PUNCT
dad-10683	341	15	liu	liu	PROPN
dad-10683	341	16	et	et	PROPN
dad-10683	341	17	al	al	PROPN
dad-10683	341	18	.	.	PROPN
dad-10683	341	19	,	,	PUNCT
dad-10683	341	20	2016	2016	NUM
dad-10683	341	21	)	)	PUNCT
dad-10683	341	22	;	;	PUNCT
dad-10683	341	23	here	here	ADV
dad-10683	341	24	,	,	PUNCT
dad-10683	341	25	we	we	PRON
dad-10683	341	26	consider	consider	VERB
dad-10683	341	27	them	they	PRON
dad-10683	341	28	to	to	PART
dad-10683	341	29	be	be	AUX
dad-10683	341	30	measures	measure	NOUN
dad-10683	341	31	of	of	ADP
dad-10683	341	32	the	the	DET
dad-10683	341	33	topicality	topicality	NOUN
dad-10683	341	34	of	of	ADP
dad-10683	341	35	the	the	DET
dad-10683	341	36	retrieved	retrieve	VERB
dad-10683	341	37	responses	response	NOUN
dad-10683	341	38	.	.	PUNCT
dad-10683	342	1	if	if	SCONJ
dad-10683	342	2	the	the	DET
dad-10683	342	3	generated	generate	VERB
dad-10683	342	4	response	response	NOUN
dad-10683	342	5	and	and	CCONJ
dad-10683	342	6	groundtruth	groundtruth	NOUN
dad-10683	342	7	response	response	NOUN
dad-10683	342	8	are	be	AUX
dad-10683	342	9	semantically	semantically	ADV
dad-10683	342	10	similar	similar	ADJ
dad-10683	342	11	,	,	PUNCT
dad-10683	342	12	then	then	ADV
dad-10683	342	13	the	the	DET
dad-10683	342	14	vector	vector	NOUN
dad-10683	342	15	-	-	PUNCT
dad-10683	342	16	based	base	VERB
dad-10683	342	17	metrics	metric	NOUN
dad-10683	342	18	should	should	AUX
dad-10683	342	19	be	be	AUX
dad-10683	342	20	higher	high	ADJ
dad-10683	342	21	,	,	PUNCT
dad-10683	342	22	as	as	SCONJ
dad-10683	342	23	word	word	NOUN
dad-10683	342	24	embeddings	embedding	VERB
dad-10683	342	25	themselves	themselves	PRON
dad-10683	342	26	contain	contain	VERB
dad-10683	342	27	semantic	semantic	ADJ
dad-10683	342	28	information	information	NOUN
dad-10683	342	29	(	(	PUNCT
dad-10683	342	30	mikolov	mikolov	NOUN
dad-10683	342	31	et	et	PROPN
dad-10683	342	32	al	al	PROPN
dad-10683	342	33	.	.	PROPN
dad-10683	342	34	,	,	PUNCT
dad-10683	342	35	2013	2013	NUM
dad-10683	342	36	)	)	PUNCT
dad-10683	342	37	.	.	PUNCT
dad-10683	343	1	it	it	PRON
dad-10683	343	2	is	be	AUX
dad-10683	343	3	because	because	SCONJ
dad-10683	343	4	of	of	ADP
dad-10683	343	5	this	this	DET
dad-10683	343	6	interpretation	interpretation	NOUN
dad-10683	343	7	that	that	SCONJ
dad-10683	343	8	we	we	PRON
dad-10683	343	9	prefer	prefer	VERB
dad-10683	343	10	the	the	DET
dad-10683	343	11	vector	vector	NOUN
dad-10683	343	12	-	-	PUNCT
dad-10683	343	13	based	base	VERB
dad-10683	343	14	metrics	metric	NOUN
dad-10683	343	15	over	over	ADP
dad-10683	343	16	word	word	NOUN
dad-10683	343	17	-	-	PUNCT
dad-10683	343	18	overlap	overlap	NOUN
dad-10683	343	19	metrics	metric	NOUN
dad-10683	343	20	such	such	ADJ
dad-10683	343	21	as	as	ADP
dad-10683	343	22	bleu	bleu	NOUN
dad-10683	343	23	.	.	PUNCT
dad-10683	344	1	the	the	DET
dad-10683	344	2	embedding	embed	VERB
dad-10683	344	3	average	average	ADJ
dad-10683	344	4	score	score	NOUN
dad-10683	344	5	approximates	approximate	VERB
dad-10683	344	6	the	the	DET
dad-10683	344	7	compositional	compositional	ADJ
dad-10683	344	8	embedding	embedding	NOUN
dad-10683	344	9	of	of	ADP
dad-10683	344	10	each	each	DET
dad-10683	344	11	utterance	utterance	NOUN
dad-10683	344	12	by	by	ADP
dad-10683	344	13	taking	take	VERB
dad-10683	344	14	an	an	DET
dad-10683	344	15	average	average	NOUN
dad-10683	344	16	of	of	ADP
dad-10683	344	17	the	the	DET
dad-10683	344	18	word	word	NOUN
dad-10683	344	19	vectors	vector	NOUN
dad-10683	344	20	that	that	PRON
dad-10683	344	21	compose	compose	VERB
dad-10683	344	22	the	the	DET
dad-10683	344	23	utterance	utterance	NOUN
dad-10683	344	24	.	.	PUNCT
dad-10683	345	1	the	the	DET
dad-10683	345	2	utterance	utterance	NOUN
dad-10683	345	3	embedding	embed	VERB
dad-10683	345	4	similarities	similarity	NOUN
dad-10683	345	5	are	be	AUX
dad-10683	345	6	then	then	ADV
dad-10683	345	7	compared	compare	VERB
dad-10683	345	8	using	use	VERB
dad-10683	345	9	cosine	cosine	NOUN
dad-10683	345	10	similarity	similarity	NOUN
dad-10683	345	11	.	.	PUNCT
dad-10683	346	1	the	the	DET
dad-10683	346	2	greedy	greedy	ADJ
dad-10683	346	3	matching	matching	NOUN
dad-10683	346	4	score	score	NOUN
dad-10683	346	5	,	,	PUNCT
dad-10683	346	6	originally	originally	ADV
dad-10683	346	7	used	use	VERB
dad-10683	346	8	to	to	PART
dad-10683	346	9	analyze	analyze	VERB
dad-10683	346	10	semantic	semantic	ADJ
dad-10683	346	11	similarity	similarity	NOUN
dad-10683	346	12	between	between	ADP
dad-10683	346	13	sentences	sentence	NOUN
dad-10683	346	14	in	in	ADP
dad-10683	346	15	intelligent	intelligent	ADJ
dad-10683	346	16	tutoring	tutoring	NOUN
dad-10683	346	17	systems	system	NOUN
dad-10683	346	18	(	(	PUNCT
dad-10683	346	19	rus	rus	NOUN
dad-10683	346	20	and	and	CCONJ
dad-10683	346	21	lintean	lintean	ADJ
dad-10683	346	22	,	,	PUNCT
dad-10683	346	23	2012	2012	NUM
dad-10683	346	24	)	)	PUNCT
dad-10683	346	25	,	,	PUNCT
dad-10683	346	26	matches	match	VERB
dad-10683	346	27	the	the	DET
dad-10683	346	28	most	most	ADV
dad-10683	346	29	similar	similar	ADJ
dad-10683	346	30	word	word	NOUN
dad-10683	346	31	in	in	ADP
dad-10683	346	32	the	the	DET
dad-10683	346	33	generated	generate	VERB
dad-10683	346	34	utterance	utterance	NOUN
dad-10683	346	35	to	to	ADP
dad-10683	346	36	the	the	DET
dad-10683	346	37	actual	actual	ADJ
dad-10683	346	38	utterance	utterance	NOUN
dad-10683	346	39	using	use	VERB
dad-10683	346	40	cosine	cosine	NOUN
dad-10683	346	41	similarity	similarity	NOUN
dad-10683	346	42	of	of	ADP
dad-10683	346	43	the	the	DET
dad-10683	346	44	word	word	NOUN
dad-10683	346	45	embeddings	embedding	VERB
dad-10683	346	46	.	.	PUNCT
dad-10683	347	1	the	the	DET
dad-10683	347	2	vector	vector	NOUN
dad-10683	347	3	extrema	extrema	NOUN
dad-10683	347	4	score	score	NOUN
dad-10683	347	5	was	be	AUX
dad-10683	347	6	proposed	propose	VERB
dad-10683	347	7	by	by	ADP
dad-10683	347	8	(	(	PUNCT
dad-10683	347	9	forgues	forgue	NOUN
dad-10683	347	10	et	et	PROPN
dad-10683	347	11	al	al	PROPN
dad-10683	347	12	.	.	PROPN
dad-10683	347	13	,	,	PUNCT
dad-10683	347	14	2014	2014	NUM
dad-10683	347	15	)	)	PUNCT
dad-10683	347	16	for	for	ADP
dad-10683	347	17	dialogue	dialogue	NOUN
dad-10683	347	18	systems	system	NOUN
dad-10683	347	19	.	.	PUNCT
dad-10683	348	1	instead	instead	ADV
dad-10683	348	2	of	of	ADP
dad-10683	348	3	averaging	average	VERB
dad-10683	348	4	each	each	DET
dad-10683	348	5	word	word	NOUN
dad-10683	348	6	embedding	embed	VERB
dad-10683	348	7	,	,	PUNCT
dad-10683	348	8	this	this	DET
dad-10683	348	9	approach	approach	NOUN
dad-10683	348	10	takes	take	VERB
dad-10683	348	11	the	the	DET
dad-10683	348	12	elementwise	elementwise	ADJ
dad-10683	348	13	maximum	maximum	ADJ
dad-10683	348	14	(	(	PUNCT
dad-10683	348	15	or	or	CCONJ
dad-10683	348	16	minimum	minimum	NOUN
dad-10683	348	17	)	)	PUNCT
dad-10683	348	18	of	of	ADP
dad-10683	348	19	each	each	DET
dad-10683	348	20	component	component	NOUN
dad-10683	348	21	in	in	ADP
dad-10683	348	22	the	the	DET
dad-10683	348	23	word	word	NOUN
dad-10683	348	24	vectors	vector	NOUN
dad-10683	348	25	composing	compose	VERB
dad-10683	348	26	an	an	DET
dad-10683	348	27	utterance	utterance	NOUN
dad-10683	348	28	.	.	PUNCT
dad-10683	349	1	this	this	PRON
dad-10683	349	2	results	result	VERB
dad-10683	349	3	in	in	ADP
dad-10683	349	4	utterance	utterance	NOUN
dad-10683	349	5	embeddings	embedding	NOUN
dad-10683	349	6	of	of	ADP
dad-10683	349	7	the	the	DET
dad-10683	349	8	same	same	ADJ
dad-10683	349	9	size	size	NOUN
dad-10683	349	10	of	of	ADP
dad-10683	349	11	each	each	DET
dad-10683	349	12	word	word	NOUN
dad-10683	349	13	vector	vector	NOUN
dad-10683	349	14	,	,	PUNCT
dad-10683	349	15	which	which	PRON
dad-10683	349	16	can	can	AUX
dad-10683	349	17	again	again	ADV
dad-10683	349	18	be	be	AUX
dad-10683	349	19	compared	compare	VERB
dad-10683	349	20	using	use	VERB
dad-10683	349	21	cosine	cosine	NOUN
dad-10683	349	22	similarity	similarity	NOUN
dad-10683	349	23	.	.	PUNCT
dad-10683	350	1	these	these	DET
dad-10683	350	2	metrics	metric	NOUN
dad-10683	350	3	are	be	AUX
dad-10683	350	4	able	able	ADJ
dad-10683	350	5	to	to	PART
dad-10683	350	6	capture	capture	VERB
dad-10683	350	7	some	some	DET
dad-10683	350	8	aspects	aspect	NOUN
dad-10683	350	9	of	of	ADP
dad-10683	350	10	dialogue	dialogue	NOUN
dad-10683	350	11	that	that	PRON
dad-10683	350	12	are	be	AUX
dad-10683	350	13	not	not	PART
dad-10683	350	14	present	present	ADJ
dad-10683	350	15	in	in	ADP
dad-10683	350	16	bleu	bleu	NOUN
dad-10683	350	17	score	score	NOUN
dad-10683	350	18	.	.	PUNCT
dad-10683	351	1	we	we	PRON
dad-10683	351	2	note	note	VERB
dad-10683	351	3	that	that	SCONJ
dad-10683	351	4	,	,	PUNCT
dad-10683	351	5	to	to	PART
dad-10683	351	6	keep	keep	VERB
dad-10683	351	7	the	the	DET
dad-10683	351	8	assumptions	assumption	NOUN
dad-10683	351	9	of	of	ADP
dad-10683	351	10	independent	independent	ADJ
dad-10683	351	11	and	and	CCONJ
dad-10683	351	12	identically	identically	ADV
dad-10683	351	13	distributed	distribute	VERB
dad-10683	351	14	(	(	PUNCT
dad-10683	351	15	i.i.d	i.i.d	ADJ
dad-10683	351	16	.	.	PUNCT
dad-10683	351	17	)	)	PUNCT
dad-10683	351	18	training	training	NOUN
dad-10683	351	19	and	and	CCONJ
dad-10683	351	20	test	test	NOUN
dad-10683	351	21	data	datum	NOUN
dad-10683	351	22	examples	example	NOUN
dad-10683	351	23	valid	valid	ADJ
dad-10683	351	24	,	,	PUNCT
dad-10683	351	25	it	it	PRON
dad-10683	351	26	is	be	AUX
dad-10683	351	27	important	important	ADJ
dad-10683	351	28	to	to	PART
dad-10683	351	29	have	have	VERB
dad-10683	351	30	the	the	DET
dad-10683	351	31	word	word	NOUN
dad-10683	351	32	embeddings	embedding	NOUN
dad-10683	351	33	used	use	VERB
dad-10683	351	34	for	for	ADP
dad-10683	351	35	the	the	DET
dad-10683	351	36	metrics	metric	NOUN
dad-10683	351	37	trained	train	VERB
dad-10683	351	38	on	on	ADP
dad-10683	351	39	a	a	DET
dad-10683	351	40	different	different	ADJ
dad-10683	351	41	corpus	corpus	NOUN
dad-10683	351	42	than	than	ADP
dad-10683	351	43	the	the	DET
dad-10683	351	44	task	task	NOUN
dad-10683	351	45	corpus	corpus	NOUN
dad-10683	351	46	,	,	PUNCT
dad-10683	351	47	as	as	SCONJ
dad-10683	351	48	argued	argue	VERB
dad-10683	351	49	by	by	ADP
dad-10683	351	50	liu	liu	PROPN
dad-10683	351	51	et	et	PROPN
dad-10683	351	52	al	al	PROPN
dad-10683	351	53	.	.	PROPN
dad-10683	352	1	(	(	PUNCT
dad-10683	352	2	2016	2016	NUM
dad-10683	352	3	)	)	PUNCT
dad-10683	352	4	.	.	PUNCT
dad-10683	353	1	this	this	PRON
dad-10683	353	2	preserves	preserve	VERB
dad-10683	353	3	the	the	DET
dad-10683	353	4	statistical	statistical	ADJ
dad-10683	353	5	independence	independence	NOUN
dad-10683	353	6	between	between	ADP
dad-10683	353	7	the	the	DET
dad-10683	353	8	task	task	NOUN
dad-10683	353	9	and	and	CCONJ
dad-10683	353	10	each	each	DET
dad-10683	353	11	performance	performance	NOUN
dad-10683	353	12	metric	metric	ADJ
dad-10683	353	13	,	,	PUNCT
dad-10683	353	14	and	and	CCONJ
dad-10683	353	15	alleviates	alleviate	VERB
dad-10683	353	16	the	the	DET
dad-10683	353	17	possibility	possibility	NOUN
dad-10683	353	18	of	of	ADP
dad-10683	353	19	spurious	spurious	ADJ
dad-10683	353	20	and	and	CCONJ
dad-10683	353	21	potentially	potentially	ADV
dad-10683	353	22	misleading	misleading	ADJ
dad-10683	353	23	correlations	correlation	NOUN
dad-10683	353	24	between	between	ADP
dad-10683	353	25	data	datum	NOUN
dad-10683	353	26	examples	example	NOUN
dad-10683	353	27	.	.	PUNCT
dad-10683	354	1	45	45	NUM
dad-10683	354	2	lowe	lowe	PROPN
dad-10683	354	3	,	,	PUNCT
dad-10683	354	4	pow	pow	PROPN
dad-10683	354	5	,	,	PUNCT
dad-10683	354	6	serban	serban	PROPN
dad-10683	354	7	,	,	PUNCT
dad-10683	354	8	charlinn	charlinn	PROPN
dad-10683	354	9	,	,	PUNCT
dad-10683	354	10	liu	liu	PROPN
dad-10683	354	11	and	and	CCONJ
dad-10683	354	12	pineau	pineau	PROPN
dad-10683	354	13	4.5	4.5	NUM
dad-10683	354	14	experimental	experimental	ADJ
dad-10683	354	15	results	result	NOUN
dad-10683	354	16	we	we	PRON
dad-10683	354	17	examine	examine	VERB
dad-10683	354	18	the	the	DET
dad-10683	354	19	performance	performance	NOUN
dad-10683	354	20	of	of	ADP
dad-10683	354	21	the	the	DET
dad-10683	354	22	models	model	NOUN
dad-10683	354	23	using	use	VERB
dad-10683	354	24	both	both	CCONJ
dad-10683	354	25	retrieval	retrieval	NOUN
dad-10683	354	26	and	and	CCONJ
dad-10683	354	27	vector	vector	NOUN
dad-10683	354	28	based	base	VERB
dad-10683	354	29	metrics	metric	NOUN
dad-10683	354	30	,	,	PUNCT
dad-10683	354	31	as	as	SCONJ
dad-10683	354	32	shown	show	VERB
dad-10683	354	33	in	in	ADP
dad-10683	354	34	tables	table	NOUN
dad-10683	354	35	3	3	NUM
dad-10683	354	36	and	and	CCONJ
dad-10683	354	37	4	4	NUM
dad-10683	354	38	.	.	X
dad-10683	354	39	for	for	ADP
dad-10683	354	40	nuc	nuc	PROPN
dad-10683	354	41	,	,	PUNCT
dad-10683	354	42	the	the	DET
dad-10683	354	43	models	model	NOUN
dad-10683	354	44	were	be	AUX
dad-10683	354	45	evaluated	evaluate	VERB
dad-10683	354	46	using	use	VERB
dad-10683	354	47	both	both	PRON
dad-10683	354	48	1	1	NUM
dad-10683	354	49	(	(	PUNCT
dad-10683	354	50	1	1	NUM
dad-10683	354	51	in	in	ADP
dad-10683	354	52	2	2	NUM
dad-10683	354	53	)	)	PUNCT
dad-10683	354	54	and	and	CCONJ
dad-10683	354	55	9	9	NUM
dad-10683	354	56	(	(	PUNCT
dad-10683	354	57	1	1	NUM
dad-10683	354	58	in	in	ADP
dad-10683	354	59	10	10	NUM
dad-10683	354	60	)	)	PUNCT
dad-10683	354	61	false	false	ADJ
dad-10683	354	62	examples.9	examples.9	PROPN
dad-10683	354	63	retrieval	retrieval	NOUN
dad-10683	354	64	metrics	metric	NOUN
dad-10683	354	65	method	method	NOUN
dad-10683	354	66	1	1	NUM
dad-10683	354	67	in	in	ADP
dad-10683	354	68	2	2	NUM
dad-10683	354	69	r@1	r@1	PROPN
dad-10683	354	70	1	1	NUM
dad-10683	354	71	in	in	ADP
dad-10683	354	72	10	10	NUM
dad-10683	354	73	r@1	r@1	PROPN
dad-10683	354	74	1	1	NUM
dad-10683	354	75	in	in	ADP
dad-10683	354	76	10	10	NUM
dad-10683	354	77	r@2	r@2	VERB
dad-10683	354	78	1	1	NUM
dad-10683	354	79	in	in	ADP
dad-10683	354	80	10	10	NUM
dad-10683	354	81	r@5	r@5	PROPN
dad-10683	354	82	tf	tf	PROPN
dad-10683	354	83	-	-	PUNCT
dad-10683	354	84	idf	idf	PROPN
dad-10683	354	85	74.9	74.9	NUM
dad-10683	354	86	%	%	NOUN
dad-10683	354	87	48.8	48.8	NUM
dad-10683	354	88	%	%	NOUN
dad-10683	354	89	58.7	58.7	NUM
dad-10683	354	90	%	%	NOUN
dad-10683	354	91	76.3	76.3	NUM
dad-10683	354	92	%	%	NOUN
dad-10683	354	93	dual	dual	ADJ
dad-10683	354	94	encoder	encoder	NOUN
dad-10683	354	95	w	w	PROPN
dad-10683	354	96	/	/	SYM
dad-10683	354	97	rnn	rnn	NOUN
dad-10683	354	98	units	unit	NOUN
dad-10683	354	99	77.7	77.7	NUM
dad-10683	354	100	%	%	NOUN
dad-10683	354	101	37.9	37.9	NUM
dad-10683	354	102	%	%	NOUN
dad-10683	354	103	56.1	56.1	NUM
dad-10683	354	104	%	%	NOUN
dad-10683	354	105	83.6	83.6	NUM
dad-10683	354	106	%	%	NOUN
dad-10683	354	107	dual	dual	ADJ
dad-10683	354	108	encoder	encoder	NOUN
dad-10683	354	109	w	w	PROPN
dad-10683	354	110	/	/	SYM
dad-10683	354	111	lstm	lstm	ADJ
dad-10683	354	112	units	unit	NOUN
dad-10683	354	113	86.9	86.9	NUM
dad-10683	354	114	%	%	NOUN
dad-10683	354	115	55.2	55.2	NUM
dad-10683	354	116	%	%	NOUN
dad-10683	354	117	72.1	72.1	NUM
dad-10683	354	118	%	%	NOUN
dad-10683	354	119	92.4	92.4	NUM
dad-10683	354	120	%	%	NOUN
dad-10683	354	121	table	table	NOUN
dad-10683	354	122	3	3	NUM
dad-10683	354	123	:	:	PUNCT
dad-10683	354	124	results	result	NOUN
dad-10683	354	125	for	for	ADP
dad-10683	354	126	the	the	DET
dad-10683	354	127	three	three	NUM
dad-10683	354	128	algorithms	algorithm	NOUN
dad-10683	354	129	using	use	VERB
dad-10683	354	130	various	various	ADJ
dad-10683	354	131	recall	recall	NOUN
dad-10683	354	132	measures	measure	NOUN
dad-10683	354	133	for	for	ADP
dad-10683	354	134	binary	binary	NOUN
dad-10683	354	135	(	(	PUNCT
dad-10683	354	136	1	1	NUM
dad-10683	354	137	in	in	ADP
dad-10683	354	138	2	2	NUM
dad-10683	354	139	)	)	PUNCT
dad-10683	354	140	and	and	CCONJ
dad-10683	354	141	1	1	NUM
dad-10683	354	142	in	in	ADP
dad-10683	354	143	10	10	NUM
dad-10683	354	144	(	(	PUNCT
dad-10683	354	145	1	1	NUM
dad-10683	354	146	in	in	ADP
dad-10683	354	147	10	10	NUM
dad-10683	354	148	)	)	PUNCT
dad-10683	354	149	next	next	ADJ
dad-10683	354	150	utterance	utterance	NOUN
dad-10683	354	151	classification	classification	NOUN
dad-10683	354	152	%	%	NOUN
dad-10683	354	153	.	.	PUNCT
dad-10683	355	1	generative	generative	ADJ
dad-10683	355	2	metrics	metric	NOUN
dad-10683	355	3	method	method	NOUN
dad-10683	355	4	embedding	embed	VERB
dad-10683	355	5	average	average	ADJ
dad-10683	355	6	greedy	greedy	ADJ
dad-10683	355	7	matching	matching	NOUN
dad-10683	355	8	vector	vector	NOUN
dad-10683	355	9	extrema	extrema	NOUN
dad-10683	355	10	tf	tf	ADJ
dad-10683	355	11	-	-	PUNCT
dad-10683	355	12	idf	idf	NOUN
dad-10683	355	13	0.536	0.536	NUM
dad-10683	355	14	0.370	0.370	NUM
dad-10683	355	15	0.342	0.342	NUM
dad-10683	355	16	dual	dual	ADJ
dad-10683	355	17	encoder	encoder	NOUN
dad-10683	355	18	w/	w/	NOUN
dad-10683	355	19	lstm	lstm	PROPN
dad-10683	355	20	units	unit	NOUN
dad-10683	355	21	0.650	0.650	NUM
dad-10683	355	22	0.413	0.413	NUM
dad-10683	355	23	0.376	0.376	NUM
dad-10683	355	24	table	table	NOUN
dad-10683	355	25	4	4	NUM
dad-10683	355	26	:	:	PUNCT
dad-10683	355	27	results	result	NOUN
dad-10683	355	28	for	for	ADP
dad-10683	355	29	tf	tf	PROPN
dad-10683	355	30	-	-	PUNCT
dad-10683	355	31	idf	idf	PROPN
dad-10683	355	32	and	and	CCONJ
dad-10683	355	33	the	the	DET
dad-10683	355	34	de	de	PROPN
dad-10683	355	35	model	model	NOUN
dad-10683	355	36	with	with	ADP
dad-10683	355	37	lstm	lstm	ADJ
dad-10683	355	38	units	unit	NOUN
dad-10683	355	39	on	on	ADP
dad-10683	355	40	the	the	DET
dad-10683	355	41	embedding	embed	VERB
dad-10683	355	42	average	average	ADJ
dad-10683	355	43	,	,	PUNCT
dad-10683	355	44	greedy	greedy	ADJ
dad-10683	355	45	matching	matching	NOUN
dad-10683	355	46	,	,	PUNCT
dad-10683	355	47	and	and	CCONJ
dad-10683	355	48	vector	vector	NOUN
dad-10683	355	49	extrema	extrema	NOUN
dad-10683	355	50	scores	score	NOUN
dad-10683	355	51	.	.	PUNCT
dad-10683	356	1	these	these	DET
dad-10683	356	2	scores	score	NOUN
dad-10683	356	3	provide	provide	VERB
dad-10683	356	4	an	an	DET
dad-10683	356	5	estimate	estimate	NOUN
dad-10683	356	6	of	of	ADP
dad-10683	356	7	the	the	DET
dad-10683	356	8	topic	topic	NOUN
dad-10683	356	9	consistency	consistency	NOUN
dad-10683	356	10	of	of	ADP
dad-10683	356	11	the	the	DET
dad-10683	356	12	generated	generate	VERB
dad-10683	356	13	responses	response	NOUN
dad-10683	356	14	.	.	PUNCT
dad-10683	357	1	context	context	NOUN
dad-10683	357	2	“	"	PUNCT
dad-10683	357	3	any	any	DET
dad-10683	357	4	apache	apache	NOUN
dad-10683	357	5	hax	hax	VERB
dad-10683	357	6	around	around	ADV
dad-10683	357	7	?	?	PUNCT
dad-10683	358	1	i	i	PRON
dad-10683	358	2	just	just	ADV
dad-10683	358	3	deleted	delete	VERB
dad-10683	358	4	all	all	PRON
dad-10683	358	5	of	of	ADP
dad-10683	358	6	path	path	NOUN
dad-10683	358	7	which	which	DET
dad-10683	358	8	package	package	NOUN
dad-10683	358	9	provides	provide	VERB
dad-10683	358	10	it	it	PRON
dad-10683	358	11	?	?	PUNCT
dad-10683	358	12	”	"	PUNCT
dad-10683	358	13	,	,	PUNCT
dad-10683	358	14	“	"	PUNCT
dad-10683	358	15	reconfiguring	reconfigure	VERB
dad-10683	358	16	apache	apache	NOUN
dad-10683	358	17	do	do	AUX
dad-10683	358	18	n’t	n’t	NOUN
dad-10683	358	19	solve	solve	VERB
dad-10683	358	20	it	it	PRON
dad-10683	358	21	?	?	PUNCT
dad-10683	358	22	”	"	PUNCT
dad-10683	358	23	ranked	rank	VERB
dad-10683	358	24	responses	response	NOUN
dad-10683	358	25	flag	flag	NOUN
dad-10683	358	26	1	1	NUM
dad-10683	358	27	.	.	PUNCT
dad-10683	359	1	“	"	PUNCT
dad-10683	359	2	does	do	AUX
dad-10683	359	3	n’t	n’t	NOUN
dad-10683	359	4	seem	seem	VERB
dad-10683	359	5	to	to	PART
dad-10683	359	6	,	,	PUNCT
dad-10683	359	7	no	no	INTJ
dad-10683	359	8	”	"	PUNCT
dad-10683	359	9	1	1	NUM
dad-10683	359	10	2	2	NUM
dad-10683	359	11	.	.	PUNCT
dad-10683	360	1	“	"	PUNCT
dad-10683	360	2	you	you	PRON
dad-10683	360	3	can	can	AUX
dad-10683	360	4	log	log	VERB
dad-10683	360	5	in	in	ADP
dad-10683	360	6	but	but	CCONJ
dad-10683	360	7	not	not	PART
dad-10683	360	8	transfer	transfer	VERB
dad-10683	360	9	files	file	NOUN
dad-10683	360	10	?	?	PUNCT
dad-10683	360	11	”	"	PUNCT
dad-10683	361	1	0	0	NUM
dad-10683	361	2	figure	figure	NOUN
dad-10683	361	3	6	6	NUM
dad-10683	361	4	:	:	PUNCT
dad-10683	361	5	example	example	NOUN
dad-10683	361	6	showing	show	VERB
dad-10683	361	7	the	the	DET
dad-10683	361	8	ranked	rank	VERB
dad-10683	361	9	responses	response	NOUN
dad-10683	361	10	from	from	ADP
dad-10683	361	11	the	the	DET
dad-10683	361	12	lstm	lstm	NOUN
dad-10683	361	13	.	.	PUNCT
dad-10683	362	1	each	each	DET
dad-10683	362	2	utterance	utterance	NOUN
dad-10683	362	3	is	be	AUX
dad-10683	362	4	shown	show	VERB
dad-10683	362	5	after	after	ADP
dad-10683	362	6	pre	pre	ADJ
dad-10683	362	7	-	-	ADJ
dad-10683	362	8	processing	processing	ADJ
dad-10683	362	9	steps	step	NOUN
dad-10683	362	10	.	.	PUNCT
dad-10683	363	1	we	we	PRON
dad-10683	363	2	observe	observe	VERB
dad-10683	363	3	that	that	SCONJ
dad-10683	363	4	the	the	DET
dad-10683	363	5	dual	dual	ADJ
dad-10683	363	6	encoder	encoder	NOUN
dad-10683	363	7	with	with	ADP
dad-10683	363	8	lstm	lstm	ADJ
dad-10683	363	9	units	unit	NOUN
dad-10683	363	10	outperforms	outperform	VERB
dad-10683	363	11	both	both	CCONJ
dad-10683	363	12	the	the	DET
dad-10683	363	13	dual	dual	ADJ
dad-10683	363	14	encoder	encoder	NOUN
dad-10683	363	15	with	with	ADP
dad-10683	363	16	rnn	rnn	NOUN
dad-10683	363	17	units	unit	NOUN
dad-10683	363	18	and	and	CCONJ
dad-10683	363	19	tf	tf	NOUN
dad-10683	363	20	-	-	PUNCT
dad-10683	363	21	idf	idf	PROPN
dad-10683	363	22	on	on	ADP
dad-10683	363	23	all	all	DET
dad-10683	363	24	evaluation	evaluation	NOUN
dad-10683	363	25	metrics	metric	NOUN
dad-10683	363	26	.	.	PUNCT
dad-10683	364	1	it	it	PRON
dad-10683	364	2	is	be	AUX
dad-10683	364	3	interesting	interesting	ADJ
dad-10683	364	4	to	to	PART
dad-10683	364	5	note	note	VERB
dad-10683	364	6	that	that	SCONJ
dad-10683	364	7	tf	tf	PROPN
dad-10683	364	8	-	-	PUNCT
dad-10683	364	9	idf	idf	PROPN
dad-10683	364	10	actually	actually	ADV
dad-10683	364	11	outperforms	outperform	VERB
dad-10683	364	12	the	the	DET
dad-10683	364	13	rnn	rnn	NOUN
dad-10683	364	14	on	on	ADP
dad-10683	364	15	the	the	DET
dad-10683	364	16	recall@1	recall@1	ADJ
dad-10683	364	17	case	case	NOUN
dad-10683	364	18	for	for	ADP
dad-10683	364	19	the	the	DET
dad-10683	364	20	1	1	NUM
dad-10683	364	21	in	in	ADP
dad-10683	364	22	10	10	NUM
dad-10683	364	23	classification	classification	NOUN
dad-10683	364	24	.	.	PUNCT
dad-10683	365	1	this	this	PRON
dad-10683	365	2	is	be	AUX
dad-10683	365	3	most	most	ADV
dad-10683	365	4	likely	likely	ADJ
dad-10683	365	5	due	due	ADP
dad-10683	365	6	to	to	ADP
dad-10683	365	7	the	the	DET
dad-10683	365	8	limited	limited	ADJ
dad-10683	365	9	ability	ability	NOUN
dad-10683	365	10	of	of	ADP
dad-10683	365	11	the	the	DET
dad-10683	365	12	rnn	rnn	NOUN
dad-10683	365	13	to	to	PART
dad-10683	365	14	take	take	VERB
dad-10683	365	15	into	into	ADP
dad-10683	365	16	account	account	NOUN
dad-10683	365	17	long	long	ADJ
dad-10683	365	18	contexts	contexts	NOUN
dad-10683	365	19	,	,	PUNCT
dad-10683	365	20	which	which	PRON
dad-10683	365	21	can	can	AUX
dad-10683	365	22	be	be	AUX
dad-10683	365	23	overcome	overcome	VERB
dad-10683	365	24	by	by	ADP
dad-10683	365	25	using	use	VERB
dad-10683	365	26	the	the	DET
dad-10683	365	27	lstm	lstm	NOUN
dad-10683	365	28	.	.	PUNCT
dad-10683	366	1	an	an	DET
dad-10683	366	2	example	example	NOUN
dad-10683	366	3	output	output	NOUN
dad-10683	366	4	of	of	ADP
dad-10683	366	5	the	the	DET
dad-10683	366	6	lstm	lstm	NOUN
dad-10683	366	7	where	where	SCONJ
dad-10683	366	8	the	the	DET
dad-10683	366	9	response	response	NOUN
dad-10683	366	10	is	be	AUX
dad-10683	366	11	correctly	correctly	ADV
dad-10683	366	12	classified	classified	ADJ
dad-10683	366	13	is	be	AUX
dad-10683	366	14	shown	show	VERB
dad-10683	366	15	in	in	ADP
dad-10683	366	16	figure	figure	NOUN
dad-10683	366	17	6	6	NUM
dad-10683	366	18	.	.	PUNCT
dad-10683	367	1	we	we	PRON
dad-10683	367	2	also	also	ADV
dad-10683	367	3	show	show	VERB
dad-10683	367	4	,	,	PUNCT
dad-10683	367	5	in	in	ADP
dad-10683	367	6	figure	figure	NOUN
dad-10683	367	7	7	7	NUM
dad-10683	367	8	,	,	PUNCT
dad-10683	367	9	the	the	DET
dad-10683	367	10	increase	increase	NOUN
dad-10683	367	11	in	in	ADP
dad-10683	367	12	performance	performance	NOUN
dad-10683	367	13	of	of	ADP
dad-10683	367	14	the	the	DET
dad-10683	367	15	lstm	lstm	NOUN
dad-10683	367	16	as	as	ADP
dad-10683	367	17	the	the	DET
dad-10683	367	18	amount	amount	NOUN
dad-10683	367	19	of	of	ADP
dad-10683	367	20	data	datum	NOUN
dad-10683	367	21	used	use	VERB
dad-10683	367	22	for	for	ADP
dad-10683	367	23	training	training	NOUN
dad-10683	367	24	increases	increase	NOUN
dad-10683	367	25	.	.	PUNCT
dad-10683	368	1	this	this	PRON
dad-10683	368	2	confirms	confirm	VERB
dad-10683	368	3	the	the	DET
dad-10683	368	4	importance	importance	NOUN
dad-10683	368	5	of	of	ADP
dad-10683	368	6	having	have	VERB
dad-10683	368	7	a	a	DET
dad-10683	368	8	large	large	ADJ
dad-10683	368	9	training	training	NOUN
dad-10683	368	10	set	set	NOUN
dad-10683	368	11	.	.	PUNCT
dad-10683	369	1	9	9	X
dad-10683	369	2	.	.	X
dad-10683	370	1	the	the	DET
dad-10683	370	2	performance	performance	NOUN
dad-10683	370	3	metrics	metric	NOUN
dad-10683	370	4	recall@2	recall@2	ADV
dad-10683	370	5	and	and	CCONJ
dad-10683	370	6	recall@5	recall@5	PRON
dad-10683	370	7	are	be	AUX
dad-10683	370	8	not	not	PART
dad-10683	370	9	relevant	relevant	ADJ
dad-10683	370	10	in	in	ADP
dad-10683	370	11	the	the	DET
dad-10683	370	12	binary	binary	ADJ
dad-10683	370	13	classification	classification	NOUN
dad-10683	370	14	case	case	NOUN
dad-10683	370	15	.	.	PUNCT
dad-10683	371	1	46	46	NUM
dad-10683	371	2	training	training	NOUN
dad-10683	371	3	end	end	NOUN
dad-10683	371	4	-	-	PUNCT
dad-10683	371	5	to	to	ADP
dad-10683	371	6	-	-	PUNCT
dad-10683	371	7	end	end	NOUN
dad-10683	371	8	dialogue	dialogue	NOUN
dad-10683	371	9	systems	system	NOUN
dad-10683	371	10	figure	figure	VERB
dad-10683	371	11	7	7	NUM
dad-10683	371	12	:	:	PUNCT
dad-10683	371	13	the	the	DET
dad-10683	371	14	lstm	lstm	NOUN
dad-10683	371	15	(	(	PUNCT
dad-10683	371	16	with	with	ADP
dad-10683	371	17	200	200	NUM
dad-10683	371	18	hidden	hide	VERB
dad-10683	371	19	units	unit	NOUN
dad-10683	371	20	)	)	PUNCT
dad-10683	371	21	,	,	PUNCT
dad-10683	371	22	showing	show	VERB
dad-10683	371	23	recall@1	recall@1	NOUN
dad-10683	371	24	for	for	ADP
dad-10683	371	25	the	the	DET
dad-10683	371	26	1	1	NUM
dad-10683	371	27	in	in	ADP
dad-10683	371	28	10	10	NUM
dad-10683	371	29	classification	classification	NOUN
dad-10683	371	30	,	,	PUNCT
dad-10683	371	31	with	with	ADP
dad-10683	371	32	increasing	increase	VERB
dad-10683	371	33	dataset	dataset	ADJ
dad-10683	371	34	sizes	size	NOUN
dad-10683	371	35	up	up	ADP
dad-10683	371	36	to	to	ADP
dad-10683	371	37	120k	120k	NUM
dad-10683	371	38	dialogues	dialogue	NOUN
dad-10683	371	39	.	.	PUNCT
dad-10683	372	1	note	note	VERB
dad-10683	372	2	that	that	SCONJ
dad-10683	372	3	this	this	PRON
dad-10683	372	4	was	be	AUX
dad-10683	372	5	calculated	calculate	VERB
dad-10683	372	6	using	use	VERB
dad-10683	372	7	the	the	DET
dad-10683	372	8	old	old	ADJ
dad-10683	372	9	version	version	NOUN
dad-10683	372	10	of	of	ADP
dad-10683	372	11	the	the	DET
dad-10683	372	12	ubuntu	ubuntu	ADJ
dad-10683	372	13	dialogue	dialogue	NOUN
dad-10683	372	14	corpus	corpus	NOUN
dad-10683	372	15	,	,	PUNCT
dad-10683	372	16	and	and	CCONJ
dad-10683	372	17	thus	thus	ADV
dad-10683	372	18	the	the	DET
dad-10683	372	19	recall@1	recall@1	ADJ
dad-10683	372	20	values	value	NOUN
dad-10683	372	21	are	be	AUX
dad-10683	372	22	higher	high	ADJ
dad-10683	372	23	than	than	ADP
dad-10683	372	24	those	those	PRON
dad-10683	372	25	in	in	ADP
dad-10683	372	26	table	table	NOUN
dad-10683	372	27	3	3	NUM
dad-10683	372	28	.	.	NOUN
dad-10683	372	29	4.6	4.6	NUM
dad-10683	372	30	qualitative	qualitative	ADJ
dad-10683	372	31	error	error	NOUN
dad-10683	372	32	analysis	analysis	NOUN
dad-10683	372	33	there	there	PRON
dad-10683	372	34	are	be	VERB
dad-10683	372	35	a	a	DET
dad-10683	372	36	large	large	ADJ
dad-10683	372	37	number	number	NOUN
dad-10683	372	38	of	of	ADP
dad-10683	372	39	technical	technical	ADJ
dad-10683	372	40	challenges	challenge	NOUN
dad-10683	372	41	that	that	PRON
dad-10683	372	42	must	must	AUX
dad-10683	372	43	be	be	AUX
dad-10683	372	44	solved	solve	VERB
dad-10683	372	45	in	in	ADP
dad-10683	372	46	order	order	NOUN
dad-10683	372	47	to	to	PART
dad-10683	372	48	construct	construct	VERB
dad-10683	372	49	a	a	DET
dad-10683	372	50	system	system	NOUN
dad-10683	372	51	that	that	PRON
dad-10683	372	52	can	can	AUX
dad-10683	372	53	provide	provide	VERB
dad-10683	372	54	adequate	adequate	ADJ
dad-10683	372	55	responses	response	NOUN
dad-10683	372	56	in	in	ADP
dad-10683	372	57	a	a	DET
dad-10683	372	58	dialogue	dialogue	NOUN
dad-10683	372	59	.	.	PUNCT
dad-10683	373	1	in	in	ADP
dad-10683	373	2	fact	fact	NOUN
dad-10683	373	3	,	,	PUNCT
dad-10683	373	4	almost	almost	ADV
dad-10683	373	5	all	all	PRON
dad-10683	373	6	common	common	ADJ
dad-10683	373	7	challenges	challenge	NOUN
dad-10683	373	8	in	in	ADP
dad-10683	373	9	natural	natural	ADJ
dad-10683	373	10	language	language	NOUN
dad-10683	373	11	processing	processing	NOUN
dad-10683	373	12	are	be	AUX
dad-10683	373	13	present	present	ADJ
dad-10683	373	14	in	in	ADP
dad-10683	373	15	some	some	DET
dad-10683	373	16	form	form	NOUN
dad-10683	373	17	or	or	CCONJ
dad-10683	373	18	another	another	PRON
dad-10683	373	19	in	in	ADP
dad-10683	373	20	the	the	DET
dad-10683	373	21	dialogue	dialogue	NOUN
dad-10683	373	22	problem	problem	NOUN
dad-10683	373	23	.	.	PUNCT
dad-10683	374	1	these	these	PRON
dad-10683	374	2	include	include	VERB
dad-10683	374	3	,	,	PUNCT
dad-10683	374	4	but	but	CCONJ
dad-10683	374	5	are	be	AUX
dad-10683	374	6	not	not	PART
dad-10683	374	7	limited	limit	VERB
dad-10683	374	8	to	to	PART
dad-10683	374	9	:	:	PUNCT
dad-10683	374	10	coreference	coreference	NOUN
dad-10683	374	11	resolution	resolution	NOUN
dad-10683	374	12	,	,	PUNCT
dad-10683	374	13	lexical	lexical	ADJ
dad-10683	374	14	semantics	semantic	NOUN
dad-10683	374	15	,	,	PUNCT
dad-10683	374	16	discourse	discourse	NOUN
dad-10683	374	17	coherence	coherence	NOUN
dad-10683	374	18	and	and	CCONJ
dad-10683	374	19	cohesion	cohesion	NOUN
dad-10683	374	20	,	,	PUNCT
dad-10683	374	21	natural	natural	ADJ
dad-10683	374	22	language	language	NOUN
dad-10683	374	23	understanding	understanding	NOUN
dad-10683	374	24	,	,	PUNCT
dad-10683	374	25	natural	natural	ADJ
dad-10683	374	26	language	language	NOUN
dad-10683	374	27	generation	generation	NOUN
dad-10683	374	28	,	,	PUNCT
dad-10683	374	29	compositional	compositional	ADJ
dad-10683	374	30	semantics	semantic	NOUN
dad-10683	374	31	,	,	PUNCT
dad-10683	374	32	the	the	DET
dad-10683	374	33	turn	turn	NOUN
dad-10683	374	34	taking	take	VERB
dad-10683	374	35	structure	structure	NOUN
dad-10683	374	36	of	of	ADP
dad-10683	374	37	dialogue	dialogue	NOUN
dad-10683	374	38	,	,	PUNCT
dad-10683	374	39	and	and	CCONJ
dad-10683	374	40	more	more	ADJ
dad-10683	374	41	.	.	PUNCT
dad-10683	375	1	further	far	ADV
dad-10683	375	2	,	,	PUNCT
dad-10683	375	3	it	it	PRON
dad-10683	375	4	is	be	AUX
dad-10683	375	5	often	often	ADV
dad-10683	375	6	necessary	necessary	ADJ
dad-10683	375	7	to	to	PART
dad-10683	375	8	have	have	VERB
dad-10683	375	9	some	some	DET
dad-10683	375	10	technical	technical	ADJ
dad-10683	375	11	knowledge	knowledge	NOUN
dad-10683	375	12	about	about	ADP
dad-10683	375	13	the	the	DET
dad-10683	375	14	subject	subject	NOUN
dad-10683	375	15	matter	matter	NOUN
dad-10683	375	16	being	be	AUX
dad-10683	375	17	discussed	discuss	VERB
dad-10683	375	18	.	.	PUNCT
dad-10683	376	1	it	it	PRON
dad-10683	376	2	is	be	AUX
dad-10683	376	3	clear	clear	ADJ
dad-10683	376	4	that	that	SCONJ
dad-10683	376	5	current	current	ADJ
dad-10683	376	6	end	end	NOUN
dad-10683	376	7	-	-	PUNCT
dad-10683	376	8	to	to	ADP
dad-10683	376	9	-	-	PUNCT
dad-10683	376	10	end	end	NOUN
dad-10683	376	11	dialogue	dialogue	NOUN
dad-10683	376	12	systems	system	NOUN
dad-10683	376	13	are	be	AUX
dad-10683	376	14	not	not	PART
dad-10683	376	15	able	able	ADJ
dad-10683	376	16	to	to	PART
dad-10683	376	17	adequately	adequately	ADV
dad-10683	376	18	address	address	VERB
dad-10683	376	19	all	all	DET
dad-10683	376	20	these	these	DET
dad-10683	376	21	problems	problem	NOUN
dad-10683	376	22	,	,	PUNCT
dad-10683	376	23	yet	yet	CCONJ
dad-10683	376	24	precisely	precisely	ADV
dad-10683	376	25	which	which	DET
dad-10683	376	26	aspects	aspect	NOUN
dad-10683	376	27	of	of	ADP
dad-10683	376	28	conversation	conversation	NOUN
dad-10683	376	29	are	be	AUX
dad-10683	376	30	the	the	DET
dad-10683	376	31	most	most	ADV
dad-10683	376	32	prevalent	prevalent	ADJ
dad-10683	376	33	sources	source	NOUN
dad-10683	376	34	of	of	ADP
dad-10683	376	35	errors	error	NOUN
dad-10683	376	36	remains	remain	VERB
dad-10683	376	37	relatively	relatively	ADV
dad-10683	376	38	unknown	unknown	ADJ
dad-10683	376	39	.	.	PUNCT
dad-10683	377	1	this	this	PRON
dad-10683	377	2	is	be	AUX
dad-10683	377	3	particularly	particularly	ADV
dad-10683	377	4	true	true	ADJ
dad-10683	377	5	for	for	ADP
dad-10683	377	6	neural	neural	ADJ
dad-10683	377	7	network	network	NOUN
dad-10683	377	8	models	model	NOUN
dad-10683	377	9	for	for	ADP
dad-10683	377	10	dialogue	dialogue	NOUN
dad-10683	377	11	,	,	PUNCT
dad-10683	377	12	which	which	PRON
dad-10683	377	13	have	have	AUX
dad-10683	377	14	only	only	ADV
dad-10683	377	15	recently	recently	ADV
dad-10683	377	16	come	come	VERB
dad-10683	377	17	into	into	ADP
dad-10683	377	18	prominence	prominence	NOUN
dad-10683	377	19	.	.	PUNCT
dad-10683	378	1	we	we	PRON
dad-10683	378	2	undertake	undertake	VERB
dad-10683	378	3	the	the	DET
dad-10683	378	4	task	task	NOUN
dad-10683	378	5	of	of	ADP
dad-10683	378	6	evaluating	evaluate	VERB
dad-10683	378	7	an	an	DET
dad-10683	378	8	end	end	NOUN
dad-10683	378	9	-	-	PUNCT
dad-10683	378	10	to	to	ADP
dad-10683	378	11	-	-	PUNCT
dad-10683	378	12	end	end	NOUN
dad-10683	378	13	dialogue	dialogue	NOUN
dad-10683	378	14	system	system	NOUN
dad-10683	378	15	,	,	PUNCT
dad-10683	378	16	the	the	DET
dad-10683	378	17	de	de	X
dad-10683	378	18	model	model	NOUN
dad-10683	378	19	with	with	ADP
dad-10683	378	20	lstm	lstm	ADJ
dad-10683	378	21	units	unit	NOUN
dad-10683	378	22	,	,	PUNCT
dad-10683	378	23	on	on	ADP
dad-10683	378	24	the	the	DET
dad-10683	378	25	ubuntu	ubuntu	ADJ
dad-10683	378	26	dialogue	dialogue	NOUN
dad-10683	378	27	corpus	corpus	NOUN
dad-10683	378	28	for	for	ADP
dad-10683	378	29	the	the	DET
dad-10683	378	30	nuc	nuc	PROPN
dad-10683	378	31	task	task	NOUN
dad-10683	378	32	.	.	PUNCT
dad-10683	379	1	we	we	PRON
dad-10683	379	2	hope	hope	VERB
dad-10683	379	3	that	that	SCONJ
dad-10683	379	4	an	an	DET
dad-10683	379	5	understanding	understanding	NOUN
dad-10683	379	6	of	of	ADP
dad-10683	379	7	the	the	DET
dad-10683	379	8	most	most	ADV
dad-10683	379	9	common	common	ADJ
dad-10683	379	10	errors	error	NOUN
dad-10683	379	11	made	make	VERB
dad-10683	379	12	by	by	ADP
dad-10683	379	13	this	this	DET
dad-10683	379	14	model	model	NOUN
dad-10683	379	15	can	can	AUX
dad-10683	379	16	help	help	VERB
dad-10683	379	17	inform	inform	VERB
dad-10683	379	18	future	future	ADJ
dad-10683	379	19	work	work	NOUN
dad-10683	379	20	on	on	ADP
dad-10683	379	21	neural	neural	ADJ
dad-10683	379	22	dialogue	dialogue	NOUN
dad-10683	379	23	systems	system	NOUN
dad-10683	379	24	,	,	PUNCT
dad-10683	379	25	particularly	particularly	ADV
dad-10683	379	26	on	on	ADP
dad-10683	379	27	the	the	DET
dad-10683	379	28	ubuntu	ubuntu	ADJ
dad-10683	379	29	dialogue	dialogue	NOUN
dad-10683	379	30	corpus	corpus	NOUN
dad-10683	379	31	.	.	PUNCT
dad-10683	380	1	we	we	PRON
dad-10683	380	2	conduct	conduct	VERB
dad-10683	380	3	an	an	DET
dad-10683	380	4	error	error	NOUN
dad-10683	380	5	analysis	analysis	NOUN
dad-10683	380	6	with	with	ADP
dad-10683	380	7	three	three	NUM
dad-10683	380	8	participants10	participants10	NOUN
dad-10683	380	9	evaluating	evaluate	VERB
dad-10683	380	10	a	a	DET
dad-10683	380	11	total	total	NOUN
dad-10683	380	12	of	of	ADP
dad-10683	380	13	100	100	NUM
dad-10683	380	14	randomly	randomly	ADV
dad-10683	380	15	chosen	choose	VERB
dad-10683	380	16	errors	error	NOUN
dad-10683	380	17	made	make	VERB
dad-10683	380	18	by	by	ADP
dad-10683	380	19	the	the	DET
dad-10683	380	20	dual	dual	ADJ
dad-10683	380	21	encoder	encoder	NOUN
dad-10683	380	22	.	.	PUNCT
dad-10683	381	1	for	for	ADP
dad-10683	381	2	each	each	DET
dad-10683	381	3	error	error	NOUN
dad-10683	381	4	made	make	VERB
dad-10683	381	5	by	by	ADP
dad-10683	381	6	the	the	DET
dad-10683	381	7	model11	model11	NOUN
dad-10683	381	8	,	,	PUNCT
dad-10683	381	9	we	we	PRON
dad-10683	381	10	consider	consider	VERB
dad-10683	381	11	what	what	PRON
dad-10683	381	12	abilities	ability	NOUN
dad-10683	381	13	the	the	DET
dad-10683	381	14	model	model	NOUN
dad-10683	381	15	would	would	AUX
dad-10683	381	16	need	need	VERB
dad-10683	381	17	to	to	PART
dad-10683	381	18	have	have	VERB
dad-10683	381	19	in	in	ADP
dad-10683	381	20	order	order	NOUN
dad-10683	381	21	to	to	PART
dad-10683	381	22	answer	answer	VERB
dad-10683	381	23	the	the	DET
dad-10683	381	24	question	question	NOUN
dad-10683	381	25	correctly	correctly	ADV
dad-10683	381	26	.	.	PUNCT
dad-10683	382	1	we	we	PRON
dad-10683	382	2	classify	classify	VERB
dad-10683	382	3	these	these	PRON
dad-10683	382	4	into	into	ADP
dad-10683	382	5	several	several	ADJ
dad-10683	382	6	10	10	NUM
dad-10683	382	7	.	.	PUNCT
dad-10683	383	1	participants	participant	NOUN
dad-10683	383	2	were	be	AUX
dad-10683	383	3	graduate	graduate	ADJ
dad-10683	383	4	students	student	NOUN
dad-10683	383	5	in	in	ADP
dad-10683	383	6	computer	computer	NOUN
dad-10683	383	7	science	science	NOUN
dad-10683	383	8	,	,	PUNCT
dad-10683	383	9	who	who	PRON
dad-10683	383	10	had	have	VERB
dad-10683	383	11	familiarity	familiarity	NOUN
dad-10683	383	12	with	with	ADP
dad-10683	383	13	both	both	DET
dad-10683	383	14	dialogue	dialogue	NOUN
dad-10683	383	15	systems	system	NOUN
dad-10683	383	16	and	and	CCONJ
dad-10683	383	17	the	the	DET
dad-10683	383	18	ubuntu	ubuntu	ADJ
dad-10683	383	19	domain	domain	NOUN
dad-10683	383	20	.	.	PUNCT
dad-10683	384	1	11	11	NUM
dad-10683	384	2	.	.	PUNCT
dad-10683	385	1	we	we	PRON
dad-10683	385	2	consider	consider	VERB
dad-10683	385	3	an	an	DET
dad-10683	385	4	error	error	NOUN
dad-10683	385	5	to	to	PART
dad-10683	385	6	be	be	AUX
dad-10683	385	7	any	any	DET
dad-10683	385	8	example	example	NOUN
dad-10683	385	9	where	where	SCONJ
dad-10683	385	10	the	the	DET
dad-10683	385	11	correct	correct	ADJ
dad-10683	385	12	response	response	NOUN
dad-10683	385	13	is	be	AUX
dad-10683	385	14	not	not	PART
dad-10683	385	15	the	the	DET
dad-10683	385	16	top	top	ADJ
dad-10683	385	17	1	1	NUM
dad-10683	385	18	response	response	NOUN
dad-10683	385	19	ranked	rank	VERB
dad-10683	385	20	by	by	ADP
dad-10683	385	21	the	the	DET
dad-10683	385	22	model	model	NOUN
dad-10683	385	23	.	.	PUNCT
dad-10683	386	1	47	47	NUM
dad-10683	386	2	lowe	lowe	PROPN
dad-10683	386	3	,	,	PUNCT
dad-10683	386	4	pow	pow	PROPN
dad-10683	386	5	,	,	PUNCT
dad-10683	386	6	serban	serban	PROPN
dad-10683	386	7	,	,	PUNCT
dad-10683	386	8	charlinn	charlinn	PROPN
dad-10683	386	9	,	,	PUNCT
dad-10683	386	10	liu	liu	PROPN
dad-10683	386	11	and	and	CCONJ
dad-10683	386	12	pineau	pineau	PROPN
dad-10683	386	13	categories	category	NOUN
dad-10683	386	14	:	:	PUNCT
dad-10683	386	15	using	use	VERB
dad-10683	386	16	knowledge	knowledge	NOUN
dad-10683	386	17	,	,	PUNCT
dad-10683	386	18	understanding	understand	VERB
dad-10683	386	19	tone	tone	NOUN
dad-10683	386	20	and	and	CCONJ
dad-10683	386	21	style	style	NOUN
dad-10683	386	22	of	of	ADP
dad-10683	386	23	the	the	DET
dad-10683	386	24	responses	response	NOUN
dad-10683	386	25	,	,	PUNCT
dad-10683	386	26	a	a	DET
dad-10683	386	27	better	well	ADJ
dad-10683	386	28	understanding	understanding	NOUN
dad-10683	386	29	of	of	ADP
dad-10683	386	30	the	the	DET
dad-10683	386	31	semantic	semantic	ADJ
dad-10683	386	32	similarity	similarity	NOUN
dad-10683	386	33	of	of	ADP
dad-10683	386	34	phrases	phrase	NOUN
dad-10683	386	35	,	,	PUNCT
dad-10683	386	36	and	and	CCONJ
dad-10683	386	37	explicitly	explicitly	ADV
dad-10683	386	38	considering	consider	VERB
dad-10683	386	39	the	the	DET
dad-10683	386	40	turn	turn	NOUN
dad-10683	386	41	-	-	PUNCT
dad-10683	386	42	taking	take	VERB
dad-10683	386	43	nature	nature	NOUN
dad-10683	386	44	of	of	ADP
dad-10683	386	45	dialogue	dialogue	NOUN
dad-10683	386	46	.	.	PUNCT
dad-10683	387	1	since	since	SCONJ
dad-10683	387	2	we	we	PRON
dad-10683	387	3	are	be	AUX
dad-10683	387	4	evaluating	evaluate	VERB
dad-10683	387	5	a	a	DET
dad-10683	387	6	classification	classification	NOUN
dad-10683	387	7	model	model	NOUN
dad-10683	387	8	,	,	PUNCT
dad-10683	387	9	we	we	PRON
dad-10683	387	10	do	do	AUX
dad-10683	387	11	not	not	PART
dad-10683	387	12	consider	consider	VERB
dad-10683	387	13	problems	problem	NOUN
dad-10683	387	14	associated	associate	VERB
dad-10683	387	15	with	with	ADP
dad-10683	387	16	natural	natural	ADJ
dad-10683	387	17	language	language	NOUN
dad-10683	387	18	generation	generation	NOUN
dad-10683	387	19	.	.	PUNCT
dad-10683	388	1	note	note	VERB
dad-10683	388	2	that	that	SCONJ
dad-10683	388	3	an	an	DET
dad-10683	388	4	error	error	NOUN
dad-10683	388	5	can	can	AUX
dad-10683	388	6	be	be	AUX
dad-10683	388	7	classified	classify	VERB
dad-10683	388	8	into	into	ADP
dad-10683	388	9	multiple	multiple	ADJ
dad-10683	388	10	categories	category	NOUN
dad-10683	388	11	,	,	PUNCT
dad-10683	388	12	if	if	SCONJ
dad-10683	388	13	they	they	PRON
dad-10683	388	14	are	be	AUX
dad-10683	388	15	each	each	PRON
dad-10683	388	16	necessary	necessary	ADJ
dad-10683	388	17	to	to	PART
dad-10683	388	18	answer	answer	VERB
dad-10683	388	19	the	the	DET
dad-10683	388	20	question	question	NOUN
dad-10683	388	21	correctly	correctly	ADV
dad-10683	388	22	.	.	PUNCT
dad-10683	389	1	in	in	ADP
dad-10683	389	2	addition	addition	NOUN
dad-10683	389	3	to	to	ADP
dad-10683	389	4	classifying	classify	VERB
dad-10683	389	5	the	the	DET
dad-10683	389	6	errors	error	NOUN
dad-10683	389	7	made	make	VERB
dad-10683	389	8	by	by	ADP
dad-10683	389	9	the	the	DET
dad-10683	389	10	model	model	NOUN
dad-10683	389	11	,	,	PUNCT
dad-10683	389	12	we	we	PRON
dad-10683	389	13	qualitatively	qualitatively	ADV
dad-10683	389	14	evaluate	evaluate	VERB
dad-10683	389	15	the	the	DET
dad-10683	389	16	difficulty	difficulty	NOUN
dad-10683	389	17	of	of	ADP
dad-10683	389	18	the	the	DET
dad-10683	389	19	questions	question	NOUN
dad-10683	389	20	on	on	ADP
dad-10683	389	21	a	a	DET
dad-10683	389	22	scale	scale	NOUN
dad-10683	389	23	from	from	ADP
dad-10683	389	24	1	1	NUM
dad-10683	389	25	-	-	SYM
dad-10683	389	26	5	5	NUM
dad-10683	389	27	.	.	PUNCT
dad-10683	390	1	a	a	DET
dad-10683	390	2	rating	rating	NOUN
dad-10683	390	3	of	of	ADP
dad-10683	390	4	1	1	NUM
dad-10683	390	5	on	on	ADP
dad-10683	390	6	the	the	DET
dad-10683	390	7	difficulty	difficulty	NOUN
dad-10683	390	8	scale	scale	NOUN
dad-10683	390	9	means	mean	VERB
dad-10683	390	10	that	that	SCONJ
dad-10683	390	11	the	the	DET
dad-10683	390	12	question	question	NOUN
dad-10683	390	13	is	be	AUX
dad-10683	390	14	easily	easily	ADV
dad-10683	390	15	answerable	answerable	ADJ
dad-10683	390	16	by	by	ADP
dad-10683	390	17	all	all	DET
dad-10683	390	18	humans	human	NOUN
dad-10683	390	19	.	.	PUNCT
dad-10683	391	1	a	a	DET
dad-10683	391	2	2	2	NUM
dad-10683	391	3	indicates	indicate	VERB
dad-10683	391	4	moderate	moderate	ADJ
dad-10683	391	5	difficulty	difficulty	NOUN
dad-10683	391	6	,	,	PUNCT
dad-10683	391	7	which	which	PRON
dad-10683	391	8	should	should	AUX
dad-10683	391	9	still	still	ADV
dad-10683	391	10	be	be	AUX
dad-10683	391	11	answerable	answerable	ADJ
dad-10683	391	12	by	by	ADP
dad-10683	391	13	all	all	DET
dad-10683	391	14	humans	human	NOUN
dad-10683	391	15	but	but	CCONJ
dad-10683	391	16	only	only	ADV
dad-10683	391	17	if	if	SCONJ
dad-10683	391	18	they	they	PRON
dad-10683	391	19	are	be	AUX
dad-10683	391	20	paying	pay	VERB
dad-10683	391	21	attention	attention	NOUN
dad-10683	391	22	.	.	PUNCT
dad-10683	392	1	a	a	DET
dad-10683	392	2	3	3	NUM
dad-10683	392	3	means	mean	VERB
dad-10683	392	4	that	that	SCONJ
dad-10683	392	5	the	the	DET
dad-10683	392	6	question	question	NOUN
dad-10683	392	7	is	be	AUX
dad-10683	392	8	fairly	fairly	ADV
dad-10683	392	9	challenging	challenging	ADJ
dad-10683	392	10	,	,	PUNCT
dad-10683	392	11	and	and	CCONJ
dad-10683	392	12	may	may	AUX
dad-10683	392	13	either	either	CCONJ
dad-10683	392	14	require	require	VERB
dad-10683	392	15	some	some	DET
dad-10683	392	16	familiarity	familiarity	NOUN
dad-10683	392	17	with	with	ADP
dad-10683	392	18	ubuntu	ubuntu	NOUN
dad-10683	392	19	or	or	CCONJ
dad-10683	392	20	the	the	DET
dad-10683	392	21	human	human	ADJ
dad-10683	392	22	respondent	respondent	NOUN
dad-10683	392	23	paying	pay	VERB
dad-10683	392	24	very	very	ADV
dad-10683	392	25	close	close	ADJ
dad-10683	392	26	attention	attention	NOUN
dad-10683	392	27	to	to	PART
dad-10683	392	28	answer	answer	VERB
dad-10683	392	29	correctly	correctly	ADV
dad-10683	392	30	.	.	PUNCT
dad-10683	393	1	a	a	DET
dad-10683	393	2	4	4	NUM
dad-10683	393	3	is	be	AUX
dad-10683	393	4	very	very	ADV
dad-10683	393	5	hard	hard	ADJ
dad-10683	393	6	,	,	PUNCT
dad-10683	393	7	usually	usually	ADV
dad-10683	393	8	meaning	mean	VERB
dad-10683	393	9	that	that	SCONJ
dad-10683	393	10	there	there	PRON
dad-10683	393	11	are	be	VERB
dad-10683	393	12	other	other	ADJ
dad-10683	393	13	responses	response	NOUN
dad-10683	393	14	that	that	PRON
dad-10683	393	15	are	be	AUX
dad-10683	393	16	nearly	nearly	ADV
dad-10683	393	17	as	as	ADV
dad-10683	393	18	good	good	ADJ
dad-10683	393	19	as	as	ADP
dad-10683	393	20	the	the	DET
dad-10683	393	21	true	true	ADJ
dad-10683	393	22	response	response	NOUN
dad-10683	393	23	;	;	PUNCT
dad-10683	393	24	many	many	ADJ
dad-10683	393	25	humans	human	NOUN
dad-10683	393	26	would	would	AUX
dad-10683	393	27	be	be	AUX
dad-10683	393	28	unable	unable	ADJ
dad-10683	393	29	to	to	PART
dad-10683	393	30	answer	answer	VERB
dad-10683	393	31	questions	question	NOUN
dad-10683	393	32	of	of	ADP
dad-10683	393	33	difficulty	difficulty	NOUN
dad-10683	393	34	4	4	NUM
dad-10683	393	35	correctly	correctly	ADV
dad-10683	393	36	.	.	PUNCT
dad-10683	394	1	a	a	DET
dad-10683	394	2	5	5	NUM
dad-10683	394	3	means	mean	VERB
dad-10683	394	4	that	that	SCONJ
dad-10683	394	5	the	the	DET
dad-10683	394	6	question	question	NOUN
dad-10683	394	7	is	be	AUX
dad-10683	394	8	effectively	effectively	ADV
dad-10683	394	9	impossible	impossible	ADJ
dad-10683	394	10	:	:	PUNCT
dad-10683	394	11	either	either	CCONJ
dad-10683	394	12	the	the	DET
dad-10683	394	13	true	true	ADJ
dad-10683	394	14	response	response	NOUN
dad-10683	394	15	is	be	AUX
dad-10683	394	16	completely	completely	ADV
dad-10683	394	17	unrelated	unrelated	ADJ
dad-10683	394	18	to	to	ADP
dad-10683	394	19	the	the	DET
dad-10683	394	20	context	context	NOUN
dad-10683	394	21	,	,	PUNCT
dad-10683	394	22	or	or	CCONJ
dad-10683	394	23	it	it	PRON
dad-10683	394	24	is	be	AUX
dad-10683	394	25	very	very	ADV
dad-10683	394	26	short	short	ADJ
dad-10683	394	27	and	and	CCONJ
dad-10683	394	28	generic	generic	ADJ
dad-10683	394	29	.	.	PUNCT
dad-10683	395	1	finally	finally	ADV
dad-10683	395	2	,	,	PUNCT
dad-10683	395	3	we	we	PRON
dad-10683	395	4	evaluate	evaluate	VERB
dad-10683	395	5	the	the	DET
dad-10683	395	6	appropriateness	appropriateness	NOUN
dad-10683	395	7	of	of	ADP
dad-10683	395	8	the	the	DET
dad-10683	395	9	response	response	NOUN
dad-10683	395	10	chosen	choose	VERB
dad-10683	395	11	by	by	ADP
dad-10683	395	12	the	the	DET
dad-10683	395	13	model	model	NOUN
dad-10683	395	14	for	for	ADP
dad-10683	395	15	each	each	DET
dad-10683	395	16	question	question	NOUN
dad-10683	395	17	on	on	ADP
dad-10683	395	18	a	a	DET
dad-10683	395	19	scale	scale	NOUN
dad-10683	395	20	from	from	ADP
dad-10683	395	21	1	1	NUM
dad-10683	395	22	-	-	SYM
dad-10683	395	23	3	3	NUM
dad-10683	395	24	.	.	PUNCT
dad-10683	396	1	a	a	DET
dad-10683	396	2	score	score	NOUN
dad-10683	396	3	of	of	ADP
dad-10683	396	4	1	1	NUM
dad-10683	396	5	indicates	indicate	VERB
dad-10683	396	6	that	that	SCONJ
dad-10683	396	7	the	the	DET
dad-10683	396	8	chosen	choose	VERB
dad-10683	396	9	response	response	NOUN
dad-10683	396	10	is	be	AUX
dad-10683	396	11	completely	completely	ADV
dad-10683	396	12	unreasonable	unreasonable	ADJ
dad-10683	396	13	given	give	VERB
dad-10683	396	14	the	the	DET
dad-10683	396	15	context	context	NOUN
dad-10683	396	16	.	.	PUNCT
dad-10683	397	1	a	a	DET
dad-10683	397	2	2	2	NUM
dad-10683	397	3	means	mean	VERB
dad-10683	397	4	that	that	SCONJ
dad-10683	397	5	the	the	DET
dad-10683	397	6	response	response	NOUN
dad-10683	397	7	chosen	choose	VERB
dad-10683	397	8	was	be	AUX
dad-10683	397	9	somewhat	somewhat	ADV
dad-10683	397	10	reasonable	reasonable	ADJ
dad-10683	397	11	,	,	PUNCT
dad-10683	397	12	and	and	CCONJ
dad-10683	397	13	that	that	SCONJ
dad-10683	397	14	it	it	PRON
dad-10683	397	15	’s	’	VERB
dad-10683	397	16	possible	possible	ADJ
dad-10683	397	17	for	for	SCONJ
dad-10683	397	18	a	a	DET
dad-10683	397	19	human	human	NOUN
dad-10683	397	20	to	to	PART
dad-10683	397	21	make	make	VERB
dad-10683	397	22	a	a	DET
dad-10683	397	23	similar	similar	ADJ
dad-10683	397	24	mistake	mistake	NOUN
dad-10683	397	25	.	.	PUNCT
dad-10683	398	1	a	a	DET
dad-10683	398	2	3	3	NUM
dad-10683	398	3	means	mean	VERB
dad-10683	398	4	that	that	SCONJ
dad-10683	398	5	the	the	DET
dad-10683	398	6	model	model	NOUN
dad-10683	398	7	’s	’s	PART
dad-10683	398	8	response	response	NOUN
dad-10683	398	9	was	be	AUX
dad-10683	398	10	more	more	ADV
dad-10683	398	11	suited	suited	ADJ
dad-10683	398	12	to	to	ADP
dad-10683	398	13	the	the	DET
dad-10683	398	14	context	context	NOUN
dad-10683	398	15	than	than	ADP
dad-10683	398	16	the	the	DET
dad-10683	398	17	actual	actual	ADJ
dad-10683	398	18	response	response	NOUN
dad-10683	398	19	.	.	PUNCT
dad-10683	399	1	we	we	PRON
dad-10683	399	2	note	note	VERB
dad-10683	399	3	that	that	SCONJ
dad-10683	399	4	such	such	DET
dad-10683	399	5	an	an	DET
dad-10683	399	6	analysis	analysis	NOUN
dad-10683	399	7	is	be	AUX
dad-10683	399	8	partially	partially	ADV
dad-10683	399	9	dependent	dependent	ADJ
dad-10683	399	10	on	on	ADP
dad-10683	399	11	the	the	DET
dad-10683	399	12	model	model	NOUN
dad-10683	399	13	and	and	CCONJ
dad-10683	399	14	the	the	DET
dad-10683	399	15	domain	domain	NOUN
dad-10683	399	16	.	.	PUNCT
dad-10683	400	1	the	the	DET
dad-10683	400	2	endto	endto	NOUN
dad-10683	400	3	-	-	PUNCT
dad-10683	400	4	end	end	NOUN
dad-10683	400	5	system	system	NOUN
dad-10683	400	6	from	from	ADP
dad-10683	400	7	wen	wen	PROPN
dad-10683	400	8	et	et	PROPN
dad-10683	400	9	al	al	PROPN
dad-10683	400	10	.	.	PROPN
dad-10683	401	1	(	(	PUNCT
dad-10683	401	2	2016	2016	NUM
dad-10683	401	3	)	)	PUNCT
dad-10683	401	4	achieves	achieve	VERB
dad-10683	401	5	very	very	ADV
dad-10683	401	6	strong	strong	ADJ
dad-10683	401	7	performance	performance	NOUN
dad-10683	401	8	,	,	PUNCT
dad-10683	401	9	and	and	CCONJ
dad-10683	401	10	thus	thus	ADV
dad-10683	401	11	would	would	AUX
dad-10683	401	12	not	not	PART
dad-10683	401	13	face	face	VERB
dad-10683	401	14	exactly	exactly	ADV
dad-10683	401	15	the	the	DET
dad-10683	401	16	same	same	ADJ
dad-10683	401	17	problems	problem	NOUN
dad-10683	401	18	as	as	ADP
dad-10683	401	19	the	the	DET
dad-10683	401	20	models	model	NOUN
dad-10683	401	21	we	we	PRON
dad-10683	401	22	present	present	VERB
dad-10683	401	23	here	here	ADV
dad-10683	401	24	.	.	PUNCT
dad-10683	402	1	however	however	ADV
dad-10683	402	2	,	,	PUNCT
dad-10683	402	3	this	this	PRON
dad-10683	402	4	is	be	AUX
dad-10683	402	5	because	because	SCONJ
dad-10683	402	6	the	the	DET
dad-10683	402	7	model	model	NOUN
dad-10683	402	8	was	be	AUX
dad-10683	402	9	trained	train	VERB
dad-10683	402	10	on	on	ADP
dad-10683	402	11	a	a	DET
dad-10683	402	12	very	very	ADV
dad-10683	402	13	narrow	narrow	ADJ
dad-10683	402	14	dataset	dataset	NOUN
dad-10683	402	15	of	of	ADP
dad-10683	402	16	restaurant	restaurant	NOUN
dad-10683	402	17	recommendations	recommendation	NOUN
dad-10683	402	18	,	,	PUNCT
dad-10683	402	19	and	and	CCONJ
dad-10683	402	20	thus	thus	ADV
dad-10683	402	21	the	the	DET
dad-10683	402	22	space	space	NOUN
dad-10683	402	23	of	of	ADP
dad-10683	402	24	generated	generate	VERB
dad-10683	402	25	responses	response	NOUN
dad-10683	402	26	is	be	AUX
dad-10683	402	27	comparatively	comparatively	ADV
dad-10683	402	28	small	small	ADJ
dad-10683	402	29	.	.	PUNCT
dad-10683	403	1	we	we	PRON
dad-10683	403	2	believe	believe	VERB
dad-10683	403	3	that	that	SCONJ
dad-10683	403	4	other	other	ADJ
dad-10683	403	5	conversational	conversational	ADJ
dad-10683	403	6	models	model	NOUN
dad-10683	403	7	trained	train	VERB
dad-10683	403	8	on	on	ADP
dad-10683	403	9	large	large	ADJ
dad-10683	403	10	,	,	PUNCT
dad-10683	403	11	complex	complex	ADJ
dad-10683	403	12	datasets	dataset	NOUN
dad-10683	403	13	are	be	AUX
dad-10683	403	14	likely	likely	ADJ
dad-10683	403	15	to	to	PART
dad-10683	403	16	encounter	encounter	VERB
dad-10683	403	17	the	the	DET
dad-10683	403	18	same	same	ADJ
dad-10683	403	19	problems	problem	NOUN
dad-10683	403	20	that	that	PRON
dad-10683	403	21	we	we	PRON
dad-10683	403	22	present	present	VERB
dad-10683	403	23	here	here	ADV
dad-10683	403	24	.	.	PUNCT
dad-10683	404	1	in	in	ADP
dad-10683	404	2	the	the	DET
dad-10683	404	3	ubuntu	ubuntu	ADJ
dad-10683	404	4	domain	domain	NOUN
dad-10683	404	5	,	,	PUNCT
dad-10683	404	6	questions	question	NOUN
dad-10683	404	7	where	where	SCONJ
dad-10683	404	8	using	use	VERB
dad-10683	404	9	external	external	ADJ
dad-10683	404	10	knowledge	knowledge	NOUN
dad-10683	404	11	would	would	AUX
dad-10683	404	12	be	be	AUX
dad-10683	404	13	helpful	helpful	ADJ
dad-10683	404	14	for	for	ADP
dad-10683	404	15	the	the	DET
dad-10683	404	16	model	model	NOUN
dad-10683	404	17	involve	involve	VERB
dad-10683	404	18	technical	technical	ADJ
dad-10683	404	19	terminology	terminology	NOUN
dad-10683	404	20	.	.	PUNCT
dad-10683	405	1	in	in	ADP
dad-10683	405	2	most	most	ADJ
dad-10683	405	3	cases	case	NOUN
dad-10683	405	4	,	,	PUNCT
dad-10683	405	5	the	the	DET
dad-10683	405	6	correct	correct	ADJ
dad-10683	405	7	response	response	NOUN
dad-10683	405	8	contains	contain	VERB
dad-10683	405	9	the	the	DET
dad-10683	405	10	name	name	NOUN
dad-10683	405	11	of	of	ADP
dad-10683	405	12	a	a	DET
dad-10683	405	13	command	command	NOUN
dad-10683	405	14	or	or	CCONJ
dad-10683	405	15	process	process	NOUN
dad-10683	405	16	that	that	PRON
dad-10683	405	17	is	be	AUX
dad-10683	405	18	related	relate	VERB
dad-10683	405	19	to	to	ADP
dad-10683	405	20	one	one	NUM
dad-10683	405	21	stated	state	VERB
dad-10683	405	22	in	in	ADP
dad-10683	405	23	the	the	DET
dad-10683	405	24	context	context	NOUN
dad-10683	405	25	;	;	PUNCT
dad-10683	405	26	however	however	ADV
dad-10683	405	27	,	,	PUNCT
dad-10683	405	28	the	the	DET
dad-10683	405	29	model	model	NOUN
dad-10683	405	30	is	be	AUX
dad-10683	405	31	unable	unable	ADJ
dad-10683	405	32	to	to	PART
dad-10683	405	33	link	link	VERB
dad-10683	405	34	the	the	DET
dad-10683	405	35	two	two	NUM
dad-10683	405	36	together	together	ADV
dad-10683	405	37	.	.	PUNCT
dad-10683	406	1	an	an	DET
dad-10683	406	2	example	example	NOUN
dad-10683	406	3	of	of	ADP
dad-10683	406	4	such	such	DET
dad-10683	406	5	a	a	DET
dad-10683	406	6	question	question	NOUN
dad-10683	406	7	is	be	AUX
dad-10683	406	8	shown	show	VERB
dad-10683	406	9	in	in	ADP
dad-10683	406	10	figure	figure	NOUN
dad-10683	406	11	8	8	NUM
dad-10683	406	12	.	.	PUNCT
dad-10683	407	1	in	in	ADP
dad-10683	407	2	this	this	DET
dad-10683	407	3	case	case	NOUN
dad-10683	407	4	,	,	PUNCT
dad-10683	407	5	the	the	DET
dad-10683	407	6	context	context	NOUN
dad-10683	407	7	of	of	ADP
dad-10683	407	8	the	the	DET
dad-10683	407	9	conversation	conversation	NOUN
dad-10683	407	10	is	be	AUX
dad-10683	407	11	about	about	ADP
dad-10683	407	12	file	file	NOUN
dad-10683	407	13	searching	search	VERB
dad-10683	407	14	in	in	ADP
dad-10683	407	15	ubuntu	ubuntu	NOUN
dad-10683	407	16	,	,	PUNCT
dad-10683	407	17	and	and	CCONJ
dad-10683	407	18	the	the	DET
dad-10683	407	19	correct	correct	ADJ
dad-10683	407	20	response	response	NOUN
dad-10683	407	21	(	(	PUNCT
dad-10683	407	22	in	in	ADP
dad-10683	407	23	italics	italic	NOUN
dad-10683	407	24	)	)	PUNCT
dad-10683	407	25	mentions	mention	VERB
dad-10683	407	26	the	the	DET
dad-10683	407	27	locate	locate	ADJ
dad-10683	407	28	command	command	NOUN
dad-10683	407	29	.	.	PUNCT
dad-10683	408	1	this	this	DET
dad-10683	408	2	response	response	NOUN
dad-10683	408	3	would	would	AUX
dad-10683	408	4	have	have	AUX
dad-10683	408	5	been	be	AUX
dad-10683	408	6	assigned	assign	VERB
dad-10683	408	7	a	a	DET
dad-10683	408	8	higher	high	ADJ
dad-10683	408	9	probability	probability	NOUN
dad-10683	408	10	if	if	SCONJ
dad-10683	408	11	it	it	PRON
dad-10683	408	12	was	be	AUX
dad-10683	408	13	able	able	ADJ
dad-10683	408	14	to	to	PART
dad-10683	408	15	determine	determine	VERB
dad-10683	408	16	the	the	DET
dad-10683	408	17	meaning	meaning	NOUN
dad-10683	408	18	of	of	ADP
dad-10683	408	19	the	the	DET
dad-10683	408	20	locate	locate	ADJ
dad-10683	408	21	command	command	NOUN
dad-10683	408	22	.	.	PUNCT
dad-10683	409	1	there	there	PRON
dad-10683	409	2	are	be	VERB
dad-10683	409	3	other	other	ADJ
dad-10683	409	4	examples	example	NOUN
dad-10683	409	5	where	where	SCONJ
dad-10683	409	6	the	the	DET
dad-10683	409	7	model	model	NOUN
dad-10683	409	8	may	may	AUX
dad-10683	409	9	be	be	AUX
dad-10683	409	10	incapable	incapable	ADJ
dad-10683	409	11	of	of	ADP
dad-10683	409	12	taking	take	VERB
dad-10683	409	13	into	into	ADP
dad-10683	409	14	account	account	NOUN
dad-10683	409	15	the	the	DET
dad-10683	409	16	specific	specific	ADJ
dad-10683	409	17	tone	tone	NOUN
dad-10683	409	18	or	or	CCONJ
dad-10683	409	19	style	style	NOUN
dad-10683	409	20	of	of	ADP
dad-10683	409	21	the	the	DET
dad-10683	409	22	users	user	NOUN
dad-10683	409	23	in	in	ADP
dad-10683	409	24	the	the	DET
dad-10683	409	25	conversation	conversation	NOUN
dad-10683	409	26	.	.	PUNCT
dad-10683	410	1	for	for	ADP
dad-10683	410	2	instance	instance	NOUN
dad-10683	410	3	,	,	PUNCT
dad-10683	410	4	a	a	DET
dad-10683	410	5	speaker	speaker	NOUN
dad-10683	410	6	may	may	AUX
dad-10683	410	7	use	use	VERB
dad-10683	410	8	many	many	ADJ
dad-10683	410	9	emoticons	emoticon	NOUN
dad-10683	410	10	,	,	PUNCT
dad-10683	410	11	have	have	VERB
dad-10683	410	12	poor	poor	ADJ
dad-10683	410	13	english	english	ADJ
dad-10683	410	14	grammar	grammar	NOUN
dad-10683	410	15	skills	skill	NOUN
dad-10683	410	16	and	and	CCONJ
dad-10683	410	17	be	be	AUX
dad-10683	410	18	prone	prone	ADJ
dad-10683	410	19	to	to	ADP
dad-10683	410	20	misspelling	misspelling	NOUN
dad-10683	410	21	words	word	NOUN
dad-10683	410	22	,	,	PUNCT
dad-10683	410	23	use	use	VERB
dad-10683	410	24	frequent	frequent	ADJ
dad-10683	410	25	abbreviations	abbreviation	NOUN
dad-10683	410	26	,	,	PUNCT
dad-10683	410	27	or	or	CCONJ
dad-10683	410	28	use	use	VERB
dad-10683	410	29	a	a	DET
dad-10683	410	30	particularly	particularly	ADV
dad-10683	410	31	formal	formal	ADJ
dad-10683	410	32	tone	tone	NOUN
dad-10683	410	33	.	.	PUNCT
dad-10683	411	1	being	be	AUX
dad-10683	411	2	able	able	ADJ
dad-10683	411	3	to	to	PART
dad-10683	411	4	spot	spot	VERB
dad-10683	411	5	these	these	DET
dad-10683	411	6	distinct	distinct	ADJ
dad-10683	411	7	language	language	NOUN
dad-10683	411	8	features	feature	NOUN
dad-10683	411	9	could	could	AUX
dad-10683	411	10	lead	lead	VERB
dad-10683	411	11	the	the	DET
dad-10683	411	12	model	model	NOUN
dad-10683	411	13	to	to	PART
dad-10683	411	14	improve	improve	VERB
dad-10683	411	15	its	its	PRON
dad-10683	411	16	performance	performance	NOUN
dad-10683	411	17	in	in	ADP
dad-10683	411	18	terms	term	NOUN
dad-10683	411	19	of	of	ADP
dad-10683	411	20	selecting	select	VERB
dad-10683	411	21	the	the	DET
dad-10683	411	22	actual	actual	ADJ
dad-10683	411	23	next	next	ADJ
dad-10683	411	24	response	response	NOUN
dad-10683	411	25	.	.	PUNCT
dad-10683	412	1	an	an	DET
dad-10683	412	2	example	example	NOUN
dad-10683	412	3	of	of	ADP
dad-10683	412	4	this	this	PRON
dad-10683	412	5	is	be	AUX
dad-10683	412	6	shown	show	VERB
dad-10683	412	7	in	in	ADP
dad-10683	412	8	figure	figure	NOUN
dad-10683	412	9	9	9	NUM
dad-10683	412	10	.	.	PUNCT
dad-10683	413	1	in	in	ADP
dad-10683	413	2	the	the	DET
dad-10683	413	3	context	context	NOUN
dad-10683	413	4	,	,	PUNCT
dad-10683	413	5	speaker	speaker	NOUN
dad-10683	413	6	a	a	DET
dad-10683	413	7	appends	append	NOUN
dad-10683	413	8	his	his	PRON
dad-10683	413	9	question	question	NOUN
dad-10683	413	10	with	with	ADP
dad-10683	413	11	an	an	DET
dad-10683	413	12	(	(	PUNCT
dad-10683	413	13	unnecessary	unnecessary	ADJ
dad-10683	413	14	)	)	PUNCT
dad-10683	413	15	smiley	smiley	ADJ
dad-10683	413	16	face	face	NOUN
dad-10683	413	17	.	.	PUNCT
dad-10683	414	1	thus	thus	ADV
dad-10683	414	2	,	,	PUNCT
dad-10683	414	3	it	it	PRON
dad-10683	414	4	is	be	AUX
dad-10683	414	5	more	more	ADV
dad-10683	414	6	likely	likely	ADJ
dad-10683	414	7	that	that	SCONJ
dad-10683	414	8	the	the	DET
dad-10683	414	9	candidate	candidate	NOUN
dad-10683	414	10	response	response	NOUN
dad-10683	414	11	with	with	ADP
dad-10683	414	12	multiple	multiple	ADJ
dad-10683	414	13	smiley	smiley	NOUN
dad-10683	414	14	faces	face	VERB
dad-10683	414	15	is	be	AUX
dad-10683	414	16	the	the	DET
dad-10683	414	17	correct	correct	ADJ
dad-10683	414	18	response	response	NOUN
dad-10683	414	19	.	.	PUNCT
dad-10683	415	1	one	one	NUM
dad-10683	415	2	of	of	ADP
dad-10683	415	3	the	the	DET
dad-10683	415	4	most	most	ADV
dad-10683	415	5	important	important	ADJ
dad-10683	415	6	challenges	challenge	NOUN
dad-10683	415	7	in	in	ADP
dad-10683	415	8	natural	natural	ADJ
dad-10683	415	9	language	language	NOUN
dad-10683	415	10	is	be	AUX
dad-10683	415	11	understanding	understand	VERB
dad-10683	415	12	the	the	DET
dad-10683	415	13	semantics	semantic	NOUN
dad-10683	415	14	of	of	ADP
dad-10683	415	15	phrases	phrase	NOUN
dad-10683	415	16	.	.	PUNCT
dad-10683	416	1	classification	classification	NOUN
dad-10683	416	2	dialogue	dialogue	NOUN
dad-10683	416	3	models	model	NOUN
dad-10683	416	4	can	can	AUX
dad-10683	416	5	make	make	VERB
dad-10683	416	6	errors	error	NOUN
dad-10683	416	7	due	due	ADJ
dad-10683	416	8	to	to	ADP
dad-10683	416	9	an	an	DET
dad-10683	416	10	inability	inability	NOUN
dad-10683	416	11	to	to	PART
dad-10683	416	12	detect	detect	VERB
dad-10683	416	13	semantic	semantic	ADJ
dad-10683	416	14	similarities	similarity	NOUN
dad-10683	416	15	between	between	ADP
dad-10683	416	16	sentences	sentence	NOUN
dad-10683	416	17	,	,	PUNCT
dad-10683	416	18	or	or	CCONJ
dad-10683	416	19	due	due	ADP
dad-10683	416	20	to	to	ADP
dad-10683	416	21	the	the	DET
dad-10683	416	22	detection	detection	NOUN
dad-10683	416	23	of	of	ADP
dad-10683	416	24	spurious	spurious	ADJ
dad-10683	416	25	similarities	similarity	NOUN
dad-10683	416	26	.	.	PUNCT
dad-10683	417	1	this	this	DET
dad-10683	417	2	category	category	NOUN
dad-10683	417	3	covers	cover	VERB
dad-10683	417	4	the	the	DET
dad-10683	417	5	general	general	ADJ
dad-10683	417	6	case	case	NOUN
dad-10683	417	7	where	where	SCONJ
dad-10683	417	8	the	the	DET
dad-10683	417	9	topic	topic	NOUN
dad-10683	417	10	of	of	ADP
dad-10683	417	11	the	the	DET
dad-10683	417	12	model	model	NOUN
dad-10683	417	13	’s	’s	PART
dad-10683	417	14	response	response	NOUN
dad-10683	417	15	is	be	AUX
dad-10683	417	16	clearly	clearly	ADV
dad-10683	417	17	different	different	ADJ
dad-10683	417	18	than	than	ADP
dad-10683	417	19	the	the	DET
dad-10683	417	20	topic	topic	NOUN
dad-10683	417	21	of	of	ADP
dad-10683	417	22	the	the	DET
dad-10683	417	23	context	context	NOUN
dad-10683	417	24	and	and	CCONJ
dad-10683	417	25	true	true	ADJ
dad-10683	417	26	response	response	NOUN
dad-10683	417	27	.	.	PUNCT
dad-10683	418	1	in	in	ADP
dad-10683	418	2	this	this	DET
dad-10683	418	3	category	category	NOUN
dad-10683	418	4	,	,	PUNCT
dad-10683	418	5	we	we	PRON
dad-10683	418	6	also	also	ADV
dad-10683	418	7	define	define	VERB
dad-10683	418	8	two	two	NUM
dad-10683	418	9	special	special	ADJ
dad-10683	418	10	cases	case	NOUN
dad-10683	418	11	:	:	PUNCT
dad-10683	418	12	one	one	NUM
dad-10683	418	13	where	where	SCONJ
dad-10683	418	14	there	there	PRON
dad-10683	418	15	is	be	VERB
dad-10683	418	16	a	a	DET
dad-10683	418	17	48	48	NUM
dad-10683	418	18	training	training	NOUN
dad-10683	418	19	end	end	NOUN
dad-10683	418	20	-	-	PUNCT
dad-10683	418	21	to	to	ADP
dad-10683	418	22	-	-	PUNCT
dad-10683	418	23	end	end	NOUN
dad-10683	418	24	dialogue	dialogue	NOUN
dad-10683	418	25	systems	system	NOUN
dad-10683	418	26	context	context	NOUN
dad-10683	418	27	:	:	PUNCT
dad-10683	418	28	speaker	speaker	NOUN
dad-10683	418	29	a	a	NOUN
dad-10683	418	30	:	:	PUNCT
dad-10683	418	31	is	be	AUX
dad-10683	418	32	there	there	PRON
dad-10683	418	33	anything	anything	PRON
dad-10683	418	34	i	i	PRON
dad-10683	418	35	can	can	AUX
dad-10683	418	36	do	do	VERB
dad-10683	418	37	to	to	PART
dad-10683	418	38	make	make	VERB
dad-10683	418	39	ubuntus	ubuntus	NOUN
dad-10683	418	40	filesearching	filesearche	VERB
dad-10683	418	41	faster	fast	ADV
dad-10683	418	42	?	?	PUNCT
dad-10683	419	1	i	i	PRON
dad-10683	419	2	am	be	AUX
dad-10683	419	3	running	run	VERB
dad-10683	419	4	from	from	ADP
dad-10683	419	5	an	an	DET
dad-10683	419	6	ssd	ssd	NOUN
dad-10683	419	7	and	and	CCONJ
dad-10683	419	8	it	it	PRON
dad-10683	419	9	’s	’	VERB
dad-10683	419	10	still	still	ADV
dad-10683	419	11	painfully	painfully	ADV
dad-10683	419	12	slow	slow	ADJ
dad-10683	419	13	eou	eou	PROPN
dad-10683	419	14	speaker	speaker	PROPN
dad-10683	419	15	b	b	PROPN
dad-10683	419	16	:	:	PUNCT
dad-10683	419	17	searching	search	VERB
dad-10683	419	18	how	how	SCONJ
dad-10683	419	19	?	?	PUNCT
dad-10683	420	1	eou	eou	PROPN
dad-10683	420	2	speaker	speaker	PROPN
dad-10683	420	3	a	a	NOUN
dad-10683	420	4	:	:	PUNCT
dad-10683	420	5	hitting	hit	VERB
dad-10683	420	6	the	the	DET
dad-10683	420	7	search	search	NOUN
dad-10683	420	8	button	button	NOUN
dad-10683	420	9	from	from	ADP
dad-10683	420	10	nautilus	nautilus	PROPN
dad-10683	420	11	eou	eou	PROPN
dad-10683	420	12	searching	search	VERB
dad-10683	420	13	systemwide	systemwide	ADJ
dad-10683	420	14	eou	eou	PROPN
dad-10683	420	15	binary	binary	PROPN
dad-10683	420	16	probability	probability	NOUN
dad-10683	420	17	candidate	candidate	NOUN
dad-10683	420	18	responses	response	VERB
dad-10683	420	19	0.10	0.10	NUM
dad-10683	420	20	i	i	PRON
dad-10683	420	21	tend	tend	VERB
dad-10683	420	22	to	to	PART
dad-10683	420	23	use	use	VERB
dad-10683	420	24	the	the	DET
dad-10683	420	25	locate	locate	ADJ
dad-10683	420	26	command	command	NOUN
dad-10683	420	27	.	.	PUNCT
dad-10683	421	1	eou	eou	PROPN
dad-10683	421	2	0.38	0.38	NUM
dad-10683	421	3	i‘m	i‘m	PROPN
dad-10683	421	4	not	not	PART
dad-10683	421	5	that	that	PRON
dad-10683	421	6	into	into	ADP
dad-10683	421	7	it	it	PRON
dad-10683	421	8	,	,	PUNCT
dad-10683	421	9	but	but	CCONJ
dad-10683	421	10	it	it	PRON
dad-10683	421	11	has	have	VERB
dad-10683	421	12	to	to	PART
dad-10683	421	13	be	be	AUX
dad-10683	421	14	session	session	NOUN
dad-10683	421	15	in	in	ADP
dad-10683	421	16	one	one	NUM
dad-10683	421	17	,	,	PUNCT
dad-10683	421	18	or	or	CCONJ
dad-10683	421	19	track	track	VERB
dad-10683	421	20	in	in	ADP
dad-10683	421	21	one	one	NUM
dad-10683	421	22	or	or	CCONJ
dad-10683	421	23	something	something	PRON
dad-10683	421	24	to	to	PART
dad-10683	421	25	have	have	VERB
dad-10683	421	26	the	the	DET
dad-10683	421	27	rw	rw	NOUN
dad-10683	421	28	funtion	funtion	NOUN
dad-10683	421	29	eou	eou	PROPN
dad-10683	421	30	0.06	0.06	NUM
dad-10683	421	31	np	np	PROPN
dad-10683	421	32	eou	eou	PROPN
dad-10683	421	33	0.56	0.56	NUM
dad-10683	422	1	probably	probably	ADV
dad-10683	422	2	just	just	ADV
dad-10683	422	3	junky	junky	PROPN
dad-10683	422	4	firefox	firefox	PROPN
dad-10683	422	5	eou	eou	PROPN
dad-10683	422	6	i	i	PRON
dad-10683	422	7	bet	bet	VERB
dad-10683	422	8	you	you	PRON
dad-10683	422	9	have	have	VERB
dad-10683	422	10	a	a	DET
dad-10683	422	11	tonne	tonne	NOUN
dad-10683	422	12	of	of	ADP
dad-10683	422	13	addons	addon	NOUN
dad-10683	422	14	eou	eou	VERB
dad-10683	422	15	that	that	SCONJ
dad-10683	422	16	all	all	PRON
dad-10683	422	17	takes	take	VERB
dad-10683	422	18	resources	resource	NOUN
dad-10683	422	19	eou	eou	PROPN
dad-10683	422	20	what	what	PRON
dad-10683	422	21	other	other	ADJ
dad-10683	422	22	apps	app	NOUN
dad-10683	422	23	are	be	AUX
dad-10683	422	24	you	you	PRON
dad-10683	422	25	running	run	VERB
dad-10683	422	26	?	?	PUNCT
dad-10683	423	1	eou	eou	PROPN
dad-10683	423	2	how	how	SCONJ
dad-10683	423	3	much	much	ADJ
dad-10683	423	4	ram	ram	NOUN
dad-10683	423	5	frees	free	VERB
dad-10683	423	6	after	after	SCONJ
dad-10683	423	7	you	you	PRON
dad-10683	423	8	close	close	VERB
dad-10683	423	9	skype	skype	NOUN
dad-10683	423	10	(	(	PUNCT
dad-10683	423	11	if	if	SCONJ
dad-10683	423	12	its	its	PRON
dad-10683	423	13	convenient	convenient	ADJ
dad-10683	423	14	)	)	PUNCT
dad-10683	423	15	0.48	0.48	NUM
dad-10683	423	16	except	except	SCONJ
dad-10683	423	17	i	i	PRON
dad-10683	423	18	get	get	VERB
dad-10683	423	19	an	an	DET
dad-10683	423	20	invite	invite	NOUN
dad-10683	423	21	eou	eou	PROPN
dad-10683	423	22	0.44	0.44	NUM
dad-10683	423	23	installing	instal	VERB
dad-10683	423	24	from	from	ADP
dad-10683	423	25	source	source	NOUN
dad-10683	423	26	on	on	ADP
dad-10683	423	27	ubuntu	ubuntu	NOUN
dad-10683	423	28	is	be	AUX
dad-10683	423	29	n’t	not	PART
dad-10683	423	30	a	a	DET
dad-10683	423	31	great	great	ADJ
dad-10683	423	32	idea	idea	NOUN
dad-10683	423	33	imo	imo	ADV
dad-10683	423	34	.	.	PUNCT
dad-10683	424	1	but	but	CCONJ
dad-10683	424	2	look	look	VERB
dad-10683	424	3	for	for	ADP
dad-10683	424	4	a	a	DET
dad-10683	424	5	make	make	NOUN
dad-10683	424	6	uninstall	uninstall	ADJ
dad-10683	424	7	option	option	NOUN
dad-10683	424	8	eou	eou	PROPN
dad-10683	424	9	0.16	0.16	NUM
dad-10683	425	1	oh	oh	INTJ
dad-10683	425	2	i	i	PRON
dad-10683	425	3	get	get	VERB
dad-10683	425	4	it	it	PRON
dad-10683	425	5	,	,	PUNCT
dad-10683	425	6	thanks	thank	NOUN
dad-10683	425	7	a	a	DET
dad-10683	425	8	lot	lot	NOUN
dad-10683	425	9	eou	eou	PROPN
dad-10683	425	10	0.21	0.21	NUM
dad-10683	425	11	the	the	DET
dad-10683	425	12	python	python	NOUN
dad-10683	425	13	one	one	NUM
dad-10683	425	14	eou	eou	NOUN
dad-10683	425	15	i	i	PRON
dad-10683	425	16	think	think	VERB
dad-10683	425	17	cron	cron	PROPN
dad-10683	425	18	may	may	AUX
dad-10683	425	19	be	be	AUX
dad-10683	425	20	able	able	ADJ
dad-10683	425	21	to	to	PART
dad-10683	425	22	do	do	VERB
dad-10683	425	23	that	that	PRON
dad-10683	425	24	..	..	PUNCT
dad-10683	425	25	to	to	PART
dad-10683	425	26	restart	restart	VERB
dad-10683	425	27	a	a	DET
dad-10683	425	28	task	task	NOUN
dad-10683	425	29	if	if	SCONJ
dad-10683	425	30	it	it	PRON
dad-10683	425	31	dies	die	VERB
dad-10683	425	32	out	out	ADP
dad-10683	425	33	prematurely	prematurely	ADV
dad-10683	425	34	eou	eou	PROPN
dad-10683	426	1	well	well	INTJ
dad-10683	426	2	you	you	PRON
dad-10683	426	3	can	can	AUX
dad-10683	426	4	show	show	VERB
dad-10683	426	5	your	your	PRON
dad-10683	426	6	#	#	SYM
dad-10683	426	7	python	python	NOUN
dad-10683	426	8	script	script	NOUN
dad-10683	426	9	and	and	CCONJ
dad-10683	426	10	people	people	NOUN
dad-10683	426	11	may	may	AUX
dad-10683	426	12	suggest	suggest	VERB
dad-10683	426	13	the	the	DET
dad-10683	426	14	best	good	ADJ
dad-10683	426	15	way	way	NOUN
dad-10683	426	16	to	to	PART
dad-10683	426	17	*	*	VERB
dad-10683	426	18	overcome	overcome	VERB
dad-10683	426	19	and	and	CCONJ
dad-10683	426	20	premature	premature	ADJ
dad-10683	426	21	*	*	PUNCT
dad-10683	426	22	unknown	unknown	ADJ
dad-10683	426	23	*	*	NOUN
dad-10683	426	24	*	*	PUNCT
dad-10683	426	25	..	..	PUNCT
dad-10683	426	26	eou	eou	PROPN
dad-10683	426	27	if	if	SCONJ
dad-10683	426	28	it	it	PRON
dad-10683	426	29	’s	’	VERB
dad-10683	426	30	a	a	DET
dad-10683	426	31	buggy	buggy	ADJ
dad-10683	426	32	script	script	NOUN
dad-10683	426	33	then	then	ADV
dad-10683	426	34	you	you	PRON
dad-10683	426	35	’d	’d	NOUN
dad-10683	426	36	expect	expect	VERB
dad-10683	426	37	it	it	PRON
dad-10683	426	38	to	to	PART
dad-10683	426	39	be	be	AUX
dad-10683	426	40	very	very	ADV
dad-10683	426	41	problematic	problematic	ADJ
dad-10683	426	42	with	with	ADP
dad-10683	426	43	anything	anything	PRON
dad-10683	426	44	starting	start	VERB
dad-10683	426	45	it	it	PRON
dad-10683	426	46	eou	eou	VERB
dad-10683	426	47	0.22	0.22	NUM
dad-10683	426	48	or	or	CCONJ
dad-10683	426	49	killall	killall	ADJ
dad-10683	426	50	ftl	ftl	PROPN
dad-10683	426	51	*	*	PUNCT
dad-10683	426	52	eou	eou	PROPN
dad-10683	426	53	0.54	0.54	NUM
dad-10683	426	54	any	any	DET
dad-10683	426	55	help	help	NOUN
dad-10683	426	56	with	with	ADP
dad-10683	426	57	custom	custom	NOUN
dad-10683	426	58	msg	msg	X
dad-10683	426	59	eou	eou	PROPN
dad-10683	426	60	figure	figure	NOUN
dad-10683	426	61	8	8	NUM
dad-10683	426	62	:	:	PUNCT
dad-10683	426	63	example	example	NOUN
dad-10683	426	64	where	where	SCONJ
dad-10683	426	65	the	the	DET
dad-10683	426	66	model	model	NOUN
dad-10683	426	67	would	would	AUX
dad-10683	426	68	benefit	benefit	VERB
dad-10683	426	69	from	from	ADP
dad-10683	426	70	using	use	VERB
dad-10683	426	71	an	an	DET
dad-10683	426	72	external	external	ADJ
dad-10683	426	73	knowledge	knowledge	NOUN
dad-10683	426	74	source	source	NOUN
dad-10683	426	75	.	.	PUNCT
dad-10683	427	1	correct	correct	ADJ
dad-10683	427	2	answer	answer	NOUN
dad-10683	427	3	is	be	AUX
dad-10683	427	4	in	in	ADP
dad-10683	427	5	italics	italic	NOUN
dad-10683	427	6	,	,	PUNCT
dad-10683	427	7	and	and	CCONJ
dad-10683	427	8	the	the	DET
dad-10683	427	9	model	model	NOUN
dad-10683	427	10	’s	’s	PART
dad-10683	427	11	selected	select	VERB
dad-10683	427	12	answer	answer	NOUN
dad-10683	427	13	is	be	AUX
dad-10683	427	14	in	in	ADP
dad-10683	427	15	bold	bold	ADJ
dad-10683	427	16	.	.	PUNCT
dad-10683	428	1	note	note	VERB
dad-10683	428	2	that	that	SCONJ
dad-10683	428	3	the	the	DET
dad-10683	428	4	probabilities	probability	NOUN
dad-10683	428	5	do	do	AUX
dad-10683	428	6	not	not	PART
dad-10683	428	7	sum	sum	VERB
dad-10683	428	8	up	up	ADP
dad-10683	428	9	to	to	PART
dad-10683	428	10	1	1	NUM
dad-10683	428	11	,	,	PUNCT
dad-10683	428	12	as	as	SCONJ
dad-10683	428	13	they	they	PRON
dad-10683	428	14	are	be	AUX
dad-10683	428	15	binary	binary	ADJ
dad-10683	428	16	probabilities	probability	NOUN
dad-10683	428	17	–	–	PUNCT
dad-10683	428	18	the	the	DET
dad-10683	428	19	model	model	NOUN
dad-10683	428	20	considers	consider	VERB
dad-10683	428	21	each	each	DET
dad-10683	428	22	candidate	candidate	NOUN
dad-10683	428	23	response	response	NOUN
dad-10683	428	24	independently	independently	ADV
dad-10683	428	25	.	.	PUNCT
dad-10683	429	1	context	context	NOUN
dad-10683	429	2	:	:	PUNCT
dad-10683	429	3	speaker	speaker	NOUN
dad-10683	429	4	a	a	NOUN
dad-10683	429	5	:	:	PUNCT
dad-10683	429	6	what	what	PRON
dad-10683	429	7	s	s	VERB
dad-10683	429	8	the	the	DET
dad-10683	429	9	best	good	ADJ
dad-10683	429	10	rdp	rdp	ADJ
dad-10683	429	11	software	software	NOUN
dad-10683	429	12	for	for	ADP
dad-10683	429	13	ubuntu	ubuntu	NOUN
dad-10683	429	14	?	?	PUNCT
dad-10683	430	1	i	i	PRON
dad-10683	430	2	want	want	VERB
dad-10683	430	3	to	to	PART
dad-10683	430	4	be	be	AUX
dad-10683	430	5	able	able	ADJ
dad-10683	430	6	to	to	PART
dad-10683	430	7	rdp	rdp	VERB
dad-10683	430	8	into	into	ADP
dad-10683	430	9	my	my	PRON
dad-10683	430	10	ubuntu	ubuntu	ADJ
dad-10683	430	11	desktop	desktop	NOUN
dad-10683	430	12	from	from	ADP
dad-10683	430	13	my	my	PRON
dad-10683	430	14	ubuntu	ubuntu	ADJ
dad-10683	430	15	laptop	laptop	NOUN
dad-10683	431	1	:)	:)	PUNCT
dad-10683	431	2	eou	eou	PROPN
dad-10683	431	3	speaker	speaker	PROPN
dad-10683	431	4	b	b	PROPN
dad-10683	431	5	:	:	PUNCT
dad-10683	431	6	then	then	ADV
dad-10683	431	7	just	just	ADV
dad-10683	431	8	use	use	VERB
dad-10683	431	9	vnc	vnc	NOUN
dad-10683	431	10	.	.	PUNCT
dad-10683	432	1	eou	eou	PROPN
dad-10683	432	2	binary	binary	PROPN
dad-10683	432	3	probability	probability	PROPN
dad-10683	432	4	candidate	candidate	NOUN
dad-10683	432	5	responses	response	VERB
dad-10683	432	6	0.15	0.15	NUM
dad-10683	432	7	what	what	DET
dad-10683	432	8	software	software	NOUN
dad-10683	432	9	,	,	PUNCT
dad-10683	432	10	do	do	AUX
dad-10683	432	11	you	you	PRON
dad-10683	432	12	have	have	VERB
dad-10683	432	13	any	any	DET
dad-10683	432	14	links	link	NOUN
dad-10683	432	15	to	to	PART
dad-10683	432	16	show	show	VERB
dad-10683	432	17	how	how	SCONJ
dad-10683	432	18	to	to	PART
dad-10683	432	19	to	to	ADP
dad-10683	432	20	it	it	PRON
dad-10683	432	21	?	?	PUNCT
dad-10683	433	1	:	:	PUNCT
dad-10683	433	2	)	)	PUNCT
dad-10683	434	1	eou	eou	PROPN
dad-10683	434	2	ur	ur	VERB
dad-10683	434	3	a	a	DET
dad-10683	434	4	beast	beast	NOUN
dad-10683	434	5	!	!	PUNCT
dad-10683	435	1	ty	ty	INTJ
dad-10683	435	2	again	again	ADV
dad-10683	435	3	:)	:)	INTJ
dad-10683	435	4	eou	eou	PROPN
dad-10683	435	5	0.04	0.04	NUM
dad-10683	435	6	wrong	wrong	ADJ
dad-10683	435	7	place	place	NOUN
dad-10683	435	8	stop	stop	VERB
dad-10683	435	9	it	it	PRON
dad-10683	435	10	eou	eou	NOUN
dad-10683	435	11	0.33	0.33	NUM
dad-10683	435	12	could	could	AUX
dad-10683	435	13	use	use	VERB
dad-10683	435	14	puppet	puppet	NOUN
dad-10683	436	1	:)	:)	INTJ
dad-10683	436	2	eou	eou	PROPN
dad-10683	436	3	0.29	0.29	NUM
dad-10683	436	4	lol	lol	ADJ
dad-10683	436	5	eou	eou	PROPN
dad-10683	436	6	0.17	0.17	NUM
dad-10683	436	7	i	i	PRON
dad-10683	436	8	’ve	’ve	AUX
dad-10683	436	9	installed	instal	VERB
dad-10683	436	10	mine	mine	PRON
dad-10683	436	11	from	from	ADP
dad-10683	436	12	“	"	PUNCT
dad-10683	436	13	additional	additional	ADJ
dad-10683	436	14	drivers	driver	NOUN
dad-10683	436	15	”	"	PUNCT
dad-10683	436	16	eou	eou	PROPN
dad-10683	436	17	0.36	0.36	NUM
dad-10683	436	18	xp	xp	INTJ
dad-10683	437	1	yeah	yeah	INTJ
dad-10683	437	2	that	that	SCONJ
dad-10683	437	3	its	its	PRON
dad-10683	437	4	been	be	AUX
dad-10683	437	5	a	a	DET
dad-10683	437	6	long	long	ADJ
dad-10683	437	7	time	time	NOUN
dad-10683	437	8	since	since	SCONJ
dad-10683	437	9	i	i	PRON
dad-10683	437	10	last	last	ADV
dad-10683	437	11	used	use	VERB
dad-10683	437	12	my	my	PRON
dad-10683	437	13	vpn	vpn	NOUN
dad-10683	437	14	server	server	NOUN
dad-10683	437	15	eou	eou	PROPN
dad-10683	437	16	0.00	0.00	NUM
dad-10683	437	17	yes	yes	INTJ
dad-10683	437	18	,	,	PUNCT
dad-10683	437	19	and	and	CCONJ
dad-10683	437	20	where	where	SCONJ
dad-10683	437	21	does	do	AUX
dad-10683	437	22	it	it	PRON
dad-10683	437	23	say	say	VERB
dad-10683	437	24	it	it	PRON
dad-10683	437	25	’s	’s	AUX
dad-10683	437	26	released	release	VERB
dad-10683	437	27	,	,	PUNCT
dad-10683	437	28	or	or	CCONJ
dad-10683	437	29	you	you	PRON
dad-10683	437	30	can	can	AUX
dad-10683	437	31	buy	buy	VERB
dad-10683	437	32	it	it	PRON
dad-10683	437	33	,	,	PUNCT
dad-10683	437	34	in	in	ADP
dad-10683	437	35	actually	actually	ADV
dad-10683	437	36	anything	anything	PRON
dad-10683	437	37	about	about	ADP
dad-10683	437	38	it	it	PRON
dad-10683	438	1	eou	eou	NOUN
dad-10683	438	2	*	*	PUNCT
dad-10683	438	3	*	*	PROPN
dad-10683	438	4	unknown	unknown	ADJ
dad-10683	438	5	*	*	PUNCT
dad-10683	438	6	*	*	AUX
dad-10683	438	7	’s	’s	AUX
dad-10683	438	8	not	not	PART
dad-10683	438	9	released	release	VERB
dad-10683	438	10	eou	eou	PROPN
dad-10683	439	1	it	it	PRON
dad-10683	439	2	’s	’	VERB
dad-10683	439	3	a	a	DET
dad-10683	439	4	concept	concept	NOUN
dad-10683	439	5	canonical	canonical	NOUN
dad-10683	439	6	are	be	AUX
dad-10683	439	7	working	work	VERB
dad-10683	439	8	on	on	ADP
dad-10683	439	9	/	/	PUNCT
dad-10683	439	10	trying	try	VERB
dad-10683	439	11	to	to	PART
dad-10683	439	12	create	create	VERB
dad-10683	439	13	eou	eou	PROPN
dad-10683	439	14	0.75	0.75	NUM
dad-10683	439	15	it	it	PRON
dad-10683	439	16	’s	’	VERB
dad-10683	439	17	that	that	SCONJ
dad-10683	439	18	“	"	PUNCT
dad-10683	439	19	persistence	persistence	NOUN
dad-10683	439	20	”	"	PUNCT
dad-10683	439	21	stuff	stuff	NOUN
dad-10683	439	22	?	?	PUNCT
dad-10683	440	1	what	what	PRON
dad-10683	440	2	do	do	AUX
dad-10683	440	3	you	you	PRON
dad-10683	440	4	mean	mean	VERB
dad-10683	440	5	?	?	PUNCT
dad-10683	441	1	eou	eou	PROPN
dad-10683	441	2	0.00	0.00	NUM
dad-10683	441	3	sudo	sudo	PROPN
dad-10683	441	4	chown	chown	PROPN
dad-10683	441	5	root	root	NOUN
dad-10683	441	6	:	:	PUNCT
dad-10683	441	7	root	root	NOUN
dad-10683	441	8	/tmp	/tmp	PROPN
dad-10683	441	9	&	&	CCONJ
dad-10683	441	10	&	&	CCONJ
dad-10683	441	11	sudo	sudo	PROPN
dad-10683	441	12	chmod	chmod	PROPN
dad-10683	441	13	1777	1777	NUM
dad-10683	441	14	/tmp	/tmp	PUNCT
dad-10683	442	1	eou	eou	PROPN
dad-10683	442	2	0.00	0.00	NUM
dad-10683	442	3	python	python	PROPN
dad-10683	442	4	eou	eou	PROPN
dad-10683	442	5	python	python	PROPN
dad-10683	442	6	eou	eou	PROPN
dad-10683	442	7	figure	figure	NOUN
dad-10683	442	8	9	9	NUM
dad-10683	442	9	:	:	PUNCT
dad-10683	442	10	example	example	NOUN
dad-10683	442	11	where	where	SCONJ
dad-10683	442	12	the	the	DET
dad-10683	442	13	model	model	NOUN
dad-10683	442	14	would	would	AUX
dad-10683	442	15	benefit	benefit	VERB
dad-10683	442	16	from	from	ADP
dad-10683	442	17	understanding	understand	VERB
dad-10683	442	18	the	the	DET
dad-10683	442	19	tone	tone	NOUN
dad-10683	442	20	and	and	CCONJ
dad-10683	442	21	style	style	NOUN
dad-10683	442	22	of	of	ADP
dad-10683	442	23	the	the	DET
dad-10683	442	24	speakers	speaker	NOUN
dad-10683	442	25	.	.	PUNCT
dad-10683	443	1	correct	correct	ADJ
dad-10683	443	2	answer	answer	NOUN
dad-10683	443	3	is	be	AUX
dad-10683	443	4	in	in	ADP
dad-10683	443	5	italics	italic	NOUN
dad-10683	443	6	,	,	PUNCT
dad-10683	443	7	and	and	CCONJ
dad-10683	443	8	the	the	DET
dad-10683	443	9	model	model	NOUN
dad-10683	443	10	’s	’s	PART
dad-10683	443	11	selected	select	VERB
dad-10683	443	12	answer	answer	NOUN
dad-10683	443	13	is	be	AUX
dad-10683	443	14	in	in	ADP
dad-10683	443	15	bold	bold	ADJ
dad-10683	443	16	.	.	PUNCT
dad-10683	444	1	49	49	NUM
dad-10683	444	2	lowe	lowe	PROPN
dad-10683	444	3	,	,	PUNCT
dad-10683	444	4	pow	pow	PROPN
dad-10683	444	5	,	,	PUNCT
dad-10683	444	6	serban	serban	PROPN
dad-10683	444	7	,	,	PUNCT
dad-10683	444	8	charlinn	charlinn	PROPN
dad-10683	444	9	,	,	PUNCT
dad-10683	444	10	liu	liu	PROPN
dad-10683	444	11	and	and	CCONJ
dad-10683	444	12	pineau	pineau	PROPN
dad-10683	444	13	context	context	PROPN
dad-10683	444	14	:	:	PUNCT
dad-10683	444	15	speaker	speaker	NOUN
dad-10683	444	16	a	a	NOUN
dad-10683	444	17	:	:	PUNCT
dad-10683	444	18	how	how	SCONJ
dad-10683	444	19	do	do	AUX
dad-10683	444	20	i	i	PRON
dad-10683	444	21	move	move	VERB
dad-10683	444	22	programs	program	NOUN
dad-10683	444	23	and	and	CCONJ
dad-10683	444	24	all	all	DET
dad-10683	444	25	their	their	PRON
dad-10683	444	26	dependencies	dependency	NOUN
dad-10683	444	27	automatically	automatically	ADV
dad-10683	444	28	?	?	PUNCT
dad-10683	445	1	eou	eou	PROPN
dad-10683	445	2	speaker	speaker	PROPN
dad-10683	445	3	b	b	PROPN
dad-10683	445	4	:	:	PUNCT
dad-10683	445	5	you	you	PRON
dad-10683	445	6	mean	mean	VERB
dad-10683	445	7	between	between	ADP
dad-10683	445	8	two	two	NUM
dad-10683	445	9	ubuntu	ubuntu	ADJ
dad-10683	445	10	installations	installation	NOUN
dad-10683	445	11	?	?	PUNCT
dad-10683	446	1	eou	eou	PROPN
dad-10683	446	2	binary	binary	PROPN
dad-10683	446	3	probability	probability	PROPN
dad-10683	446	4	candidate	candidate	NOUN
dad-10683	446	5	responses	response	VERB
dad-10683	446	6	0.04	0.04	NUM
dad-10683	447	1	i	i	PRON
dad-10683	447	2	have	have	VERB
dad-10683	447	3	a	a	DET
dad-10683	447	4	chroot	chroot	NOUN
dad-10683	447	5	partition	partition	NOUN
dad-10683	447	6	i	i	PRON
dad-10683	447	7	’ve	’ve	AUX
dad-10683	447	8	been	be	AUX
dad-10683	447	9	using	use	VERB
dad-10683	447	10	ldd	ldd	PROPN
dad-10683	447	11	and	and	CCONJ
dad-10683	447	12	doing	do	VERB
dad-10683	447	13	it	it	PRON
dad-10683	447	14	manually	manually	ADV
dad-10683	448	1	but	but	CCONJ
dad-10683	448	2	it	it	PRON
dad-10683	448	3	is	be	AUX
dad-10683	448	4	quite	quite	ADV
dad-10683	448	5	slow	slow	ADJ
dad-10683	448	6	eou	eou	NOUN
dad-10683	448	7	0.00	0.00	NUM
dad-10683	448	8	if	if	SCONJ
dad-10683	448	9	your	your	PRON
dad-10683	448	10	not	not	PART
dad-10683	448	11	a	a	DET
dad-10683	448	12	geek	geek	NOUN
dad-10683	448	13	then	then	ADV
dad-10683	448	14	you	you	PRON
dad-10683	448	15	wo	will	AUX
dad-10683	448	16	n’t	not	PART
dad-10683	448	17	understand	understand	VERB
dad-10683	448	18	eou	eou	PROPN
dad-10683	448	19	would	would	AUX
dad-10683	448	20	you	you	PRON
dad-10683	448	21	advise	advise	VERB
dad-10683	448	22	your	your	PRON
dad-10683	448	23	grandmother	grandmother	NOUN
dad-10683	448	24	to	to	PART
dad-10683	448	25	try	try	VERB
dad-10683	448	26	and	and	CCONJ
dad-10683	448	27	install	install	VERB
dad-10683	448	28	linux	linux	PROPN
dad-10683	448	29	?	?	PUNCT
dad-10683	449	1	eou	eou	PROPN
dad-10683	449	2	0.02	0.02	NUM
dad-10683	449	3	that	that	PRON
dad-10683	449	4	’s	’	VERB
dad-10683	449	5	where	where	SCONJ
dad-10683	449	6	i	i	PRON
dad-10683	449	7	m	m	VERB
dad-10683	449	8	stumped	stumped	ADJ
dad-10683	449	9	...	...	PUNCT
dad-10683	449	10	an	an	DET
dad-10683	449	11	older	old	ADJ
dad-10683	449	12	kernel	kernel	NOUN
dad-10683	449	13	made	make	VERB
dad-10683	449	14	no	no	DET
dad-10683	449	15	difference	difference	NOUN
dad-10683	449	16	,	,	PUNCT
dad-10683	449	17	whilst	whilst	SCONJ
dad-10683	449	18	an	an	DET
dad-10683	449	19	older	old	ADJ
dad-10683	449	20	release	release	NOUN
dad-10683	449	21	of	of	ADP
dad-10683	449	22	ubuntu	ubuntu	NOUN
dad-10683	449	23	did	do	AUX
dad-10683	449	24	eou	eou	PROPN
dad-10683	449	25	0.71	0.71	NUM
dad-10683	449	26	well	well	INTJ
dad-10683	449	27	,	,	PUNCT
dad-10683	449	28	everything	everything	PRON
dad-10683	449	29	with	with	ADP
dad-10683	449	30	indicators	indicator	NOUN
dad-10683	449	31	is	be	AUX
dad-10683	449	32	basically	basically	ADV
dad-10683	449	33	dbus	dbus	ADJ
dad-10683	449	34	eou	eou	PROPN
dad-10683	449	35	0.01	0.01	NUM
dad-10683	450	1	it	it	PRON
dad-10683	450	2	’s	’	VERB
dad-10683	450	3	cool	cool	ADJ
dad-10683	450	4	that	that	SCONJ
dad-10683	450	5	you	you	PRON
dad-10683	450	6	help	help	VERB
dad-10683	450	7	people	people	NOUN
dad-10683	450	8	who	who	PRON
dad-10683	450	9	run	run	VERB
dad-10683	450	10	free	free	ADJ
dad-10683	450	11	software	software	NOUN
dad-10683	450	12	;)	;)	PUNCT
dad-10683	451	1	bye	bye	INTJ
dad-10683	451	2	eou	eou	PROPN
dad-10683	451	3	0.00	0.00	NUM
dad-10683	451	4	not	not	PART
dad-10683	451	5	by	by	ADP
dad-10683	451	6	default	default	NOUN
dad-10683	451	7	,	,	PUNCT
dad-10683	451	8	but	but	CCONJ
dad-10683	451	9	it	it	PRON
dad-10683	451	10	can	can	AUX
dad-10683	451	11	.	.	PUNCT
dad-10683	451	12	/var	/var	PUNCT
dad-10683	452	1	holds	hold	VERB
dad-10683	452	2	a	a	DET
dad-10683	452	3	lot	lot	NOUN
dad-10683	452	4	of	of	ADP
dad-10683	452	5	temp	temp	NOUN
dad-10683	452	6	stuff	stuff	NOUN
dad-10683	452	7	like	like	ADP
dad-10683	452	8	logs	log	NOUN
dad-10683	452	9	and	and	CCONJ
dad-10683	452	10	debs	debs	NOUN
dad-10683	452	11	,	,	PUNCT
dad-10683	452	12	you	you	PRON
dad-10683	452	13	do	do	AUX
dad-10683	452	14	n’t	not	PART
dad-10683	452	15	need	need	VERB
dad-10683	452	16	those	those	PRON
dad-10683	452	17	cluttering	clutter	VERB
dad-10683	452	18	your	your	PRON
dad-10683	452	19	ssd	ssd	NOUN
dad-10683	452	20	and	and	CCONJ
dad-10683	452	21	using	use	VERB
dad-10683	452	22	write	write	NOUN
dad-10683	452	23	cycles	cycle	NOUN
dad-10683	452	24	eou	eou	PROPN
dad-10683	452	25	also	also	ADV
dad-10683	452	26	,	,	PUNCT
dad-10683	452	27	move	move	VERB
dad-10683	452	28	your	your	PRON
dad-10683	452	29	web	web	NOUN
dad-10683	452	30	cache	cache	NOUN
dad-10683	452	31	to	to	PART
dad-10683	452	32	ramdisk	ramdisk	VERB
dad-10683	452	33	to	to	PART
dad-10683	452	34	make	make	VERB
dad-10683	452	35	it	it	PRON
dad-10683	452	36	fast	fast	VERB
dad-10683	452	37	as	as	ADV
dad-10683	452	38	well	well	ADV
dad-10683	452	39	as	as	ADP
dad-10683	452	40	not	not	PART
dad-10683	452	41	use	use	VERB
dad-10683	452	42	your	your	PRON
dad-10683	452	43	hdd	hdd	NOUN
dad-10683	452	44	at	at	ADV
dad-10683	452	45	all	all	ADV
dad-10683	452	46	:)	:)	INTJ
dad-10683	452	47	eou	eou	VERB
dad-10683	452	48	its	its	PRON
dad-10683	452	49	a	a	DET
dad-10683	452	50	disk	disk	NOUN
dad-10683	452	51	space	space	NOUN
dad-10683	452	52	...	...	PUNCT
dad-10683	452	53	in	in	ADP
dad-10683	452	54	ram	ram	NOUN
dad-10683	452	55	eou	eou	PROPN
dad-10683	453	1	0.59	0.59	NUM
dad-10683	453	2	yes	yes	INTJ
dad-10683	453	3	,	,	PUNCT
dad-10683	454	1	but	but	CCONJ
dad-10683	454	2	they	they	PRON
dad-10683	454	3	speak	speak	VERB
dad-10683	454	4	http	http	ADV
dad-10683	455	1	so	so	SCONJ
dad-10683	455	2	i	i	PRON
dad-10683	455	3	could	could	AUX
dad-10683	455	4	use	use	VERB
dad-10683	455	5	the	the	DET
dad-10683	455	6	browser	browser	NOUN
dad-10683	455	7	as	as	ADP
dad-10683	455	8	a	a	DET
dad-10683	455	9	low	low	ADJ
dad-10683	455	10	-	-	PUNCT
dad-10683	455	11	level	level	NOUN
dad-10683	455	12	access	access	NOUN
dad-10683	455	13	tool	tool	NOUN
dad-10683	455	14	for	for	ADP
dad-10683	455	15	browing	browing	PROPN
dad-10683	455	16	repos	repos	PROPN
dad-10683	455	17	and	and	CCONJ
dad-10683	455	18	i	i	PRON
dad-10683	455	19	would	would	AUX
dad-10683	455	20	like	like	VERB
dad-10683	455	21	to	to	PART
dad-10683	455	22	do	do	VERB
dad-10683	455	23	that	that	PRON
dad-10683	455	24	,	,	PUNCT
dad-10683	455	25	but	but	CCONJ
dad-10683	455	26	that	that	PRON
dad-10683	455	27	does	do	AUX
dad-10683	455	28	n’t	not	PART
dad-10683	455	29	seem	seem	VERB
dad-10683	455	30	to	to	PART
dad-10683	455	31	work	work	VERB
dad-10683	455	32	.	.	PUNCT
dad-10683	456	1	eou	eou	PROPN
dad-10683	456	2	0.61	0.61	NUM
dad-10683	456	3	sure	sure	ADJ
dad-10683	456	4	eou	eou	PROPN
dad-10683	456	5	0.97	0.97	NUM
dad-10683	456	6	i	i	PRON
dad-10683	456	7	believe	believe	VERB
dad-10683	456	8	it	it	PRON
dad-10683	456	9	uses	use	VERB
dad-10683	456	10	gdm	gdm	NOUN
dad-10683	456	11	but	but	CCONJ
dad-10683	456	12	i	i	PRON
dad-10683	456	13	’m	’m	VERB
dad-10683	456	14	not	not	PART
dad-10683	456	15	sure	sure	ADJ
dad-10683	456	16	.	.	PUNCT
dad-10683	457	1	the	the	DET
dad-10683	457	2	login	login	NOUN
dad-10683	457	3	manager	manager	NOUN
dad-10683	457	4	thing	thing	NOUN
dad-10683	457	5	looks	look	VERB
dad-10683	457	6	the	the	DET
dad-10683	457	7	same	same	ADJ
dad-10683	457	8	as	as	SCONJ
dad-10683	457	9	the	the	DET
dad-10683	457	10	stock	stock	NOUN
dad-10683	457	11	ubuntu	ubuntu	NOUN
dad-10683	457	12	12.10	12.10	NUM
dad-10683	457	13	one	one	NUM
dad-10683	457	14	eou	eou	PROPN
dad-10683	457	15	0.14	0.14	NUM
dad-10683	457	16	ok	ok	INTJ
dad-10683	457	17	,	,	PUNCT
dad-10683	457	18	ty	ty	PRON
dad-10683	457	19	eou	eou	PROPN
dad-10683	457	20	figure	figure	NOUN
dad-10683	457	21	10	10	NUM
dad-10683	457	22	:	:	PUNCT
dad-10683	457	23	example	example	NOUN
dad-10683	457	24	where	where	SCONJ
dad-10683	457	25	the	the	DET
dad-10683	457	26	model	model	NOUN
dad-10683	457	27	would	would	AUX
dad-10683	457	28	benefit	benefit	VERB
dad-10683	457	29	from	from	ADP
dad-10683	457	30	the	the	DET
dad-10683	457	31	ability	ability	NOUN
dad-10683	457	32	to	to	PART
dad-10683	457	33	conduct	conduct	VERB
dad-10683	457	34	high	high	ADJ
dad-10683	457	35	-	-	PUNCT
dad-10683	457	36	level	level	NOUN
dad-10683	457	37	inference	inference	NOUN
dad-10683	457	38	to	to	PART
dad-10683	457	39	better	well	ADV
dad-10683	457	40	understand	understand	VERB
dad-10683	457	41	the	the	DET
dad-10683	457	42	semantic	semantic	ADJ
dad-10683	457	43	similarity	similarity	NOUN
dad-10683	457	44	between	between	ADP
dad-10683	457	45	context	context	NOUN
dad-10683	457	46	and	and	CCONJ
dad-10683	457	47	correct	correct	ADJ
dad-10683	457	48	response	response	NOUN
dad-10683	457	49	.	.	PUNCT
dad-10683	458	1	correct	correct	ADJ
dad-10683	458	2	answer	answer	NOUN
dad-10683	458	3	is	be	AUX
dad-10683	458	4	in	in	ADP
dad-10683	458	5	italics	italic	NOUN
dad-10683	458	6	,	,	PUNCT
dad-10683	458	7	and	and	CCONJ
dad-10683	458	8	the	the	DET
dad-10683	458	9	model	model	NOUN
dad-10683	458	10	’s	’s	PART
dad-10683	458	11	selected	select	VERB
dad-10683	458	12	answer	answer	NOUN
dad-10683	458	13	is	be	AUX
dad-10683	458	14	in	in	ADP
dad-10683	458	15	bold	bold	ADJ
dad-10683	458	16	.	.	PUNCT
dad-10683	459	1	direct	direct	ADJ
dad-10683	459	2	word	word	NOUN
dad-10683	459	3	copying	copying	NOUN
dad-10683	459	4	between	between	ADP
dad-10683	459	5	the	the	DET
dad-10683	459	6	context	context	NOUN
dad-10683	459	7	and	and	CCONJ
dad-10683	459	8	true	true	ADJ
dad-10683	459	9	response	response	NOUN
dad-10683	459	10	that	that	SCONJ
dad-10683	459	11	the	the	DET
dad-10683	459	12	model	model	NOUN
dad-10683	459	13	failed	fail	VERB
dad-10683	459	14	to	to	PART
dad-10683	459	15	detect	detect	VERB
dad-10683	459	16	,	,	PUNCT
dad-10683	459	17	and	and	CCONJ
dad-10683	459	18	one	one	NUM
dad-10683	459	19	where	where	SCONJ
dad-10683	459	20	some	some	DET
dad-10683	459	21	high	high	ADJ
dad-10683	459	22	-	-	PUNCT
dad-10683	459	23	level	level	NOUN
dad-10683	459	24	inference	inference	NOUN
dad-10683	459	25	is	be	AUX
dad-10683	459	26	required	require	VERB
dad-10683	459	27	to	to	PART
dad-10683	459	28	answer	answer	VERB
dad-10683	459	29	correctly	correctly	ADV
dad-10683	459	30	.	.	PUNCT
dad-10683	460	1	an	an	DET
dad-10683	460	2	example	example	NOUN
dad-10683	460	3	of	of	ADP
dad-10683	460	4	the	the	DET
dad-10683	460	5	latter	latter	ADJ
dad-10683	460	6	case	case	NOUN
dad-10683	460	7	is	be	AUX
dad-10683	460	8	shown	show	VERB
dad-10683	460	9	in	in	ADP
dad-10683	460	10	figure	figure	NOUN
dad-10683	460	11	10	10	NUM
dad-10683	460	12	;	;	PUNCT
dad-10683	460	13	speaker	speaker	NOUN
dad-10683	460	14	a	a	DET
dad-10683	460	15	asks	ask	VERB
dad-10683	460	16	how	how	SCONJ
dad-10683	460	17	to	to	PART
dad-10683	460	18	install	install	VERB
dad-10683	460	19	some	some	DET
dad-10683	460	20	programs	program	NOUN
dad-10683	460	21	automatically	automatically	ADV
dad-10683	460	22	,	,	PUNCT
dad-10683	460	23	and	and	CCONJ
dad-10683	460	24	the	the	DET
dad-10683	460	25	correct	correct	ADJ
dad-10683	460	26	response	response	NOUN
dad-10683	460	27	states	state	VERB
dad-10683	460	28	that	that	SCONJ
dad-10683	460	29	they	they	PRON
dad-10683	460	30	had	have	AUX
dad-10683	460	31	previously	previously	ADV
dad-10683	460	32	been	be	AUX
dad-10683	460	33	‘	'	PUNCT
dad-10683	460	34	doing	do	VERB
dad-10683	460	35	it	it	PRON
dad-10683	460	36	manually	manually	ADV
dad-10683	460	37	’	'	PUNCT
dad-10683	460	38	.	.	PUNCT
dad-10683	461	1	thus	thus	ADV
dad-10683	461	2	,	,	PUNCT
dad-10683	461	3	if	if	SCONJ
dad-10683	461	4	the	the	DET
dad-10683	461	5	model	model	NOUN
dad-10683	461	6	was	be	AUX
dad-10683	461	7	able	able	ADJ
dad-10683	461	8	to	to	PART
dad-10683	461	9	infer	infer	VERB
dad-10683	461	10	that	that	SCONJ
dad-10683	461	11	a	a	DET
dad-10683	461	12	person	person	NOUN
dad-10683	461	13	who	who	PRON
dad-10683	461	14	has	have	AUX
dad-10683	461	15	asked	ask	VERB
dad-10683	461	16	to	to	PART
dad-10683	461	17	perform	perform	VERB
dad-10683	461	18	an	an	DET
dad-10683	461	19	operation	operation	NOUN
dad-10683	461	20	automatically	automatically	ADV
dad-10683	461	21	could	could	AUX
dad-10683	461	22	previously	previously	ADV
dad-10683	461	23	have	have	AUX
dad-10683	461	24	been	be	AUX
dad-10683	461	25	doing	do	VERB
dad-10683	461	26	it	it	PRON
dad-10683	461	27	manually	manually	ADV
dad-10683	461	28	,	,	PUNCT
dad-10683	461	29	it	it	PRON
dad-10683	461	30	would	would	AUX
dad-10683	461	31	have	have	AUX
dad-10683	461	32	assigned	assign	VERB
dad-10683	461	33	a	a	DET
dad-10683	461	34	higher	high	ADJ
dad-10683	461	35	probability	probability	NOUN
dad-10683	461	36	to	to	ADP
dad-10683	461	37	the	the	DET
dad-10683	461	38	correct	correct	ADJ
dad-10683	461	39	response	response	NOUN
dad-10683	461	40	.	.	PUNCT
dad-10683	462	1	finally	finally	ADV
dad-10683	462	2	,	,	PUNCT
dad-10683	462	3	we	we	PRON
dad-10683	462	4	consider	consider	VERB
dad-10683	462	5	errors	error	NOUN
dad-10683	462	6	where	where	SCONJ
dad-10683	462	7	the	the	DET
dad-10683	462	8	model	model	NOUN
dad-10683	462	9	is	be	AUX
dad-10683	462	10	unable	unable	ADJ
dad-10683	462	11	to	to	PART
dad-10683	462	12	account	account	VERB
dad-10683	462	13	for	for	ADP
dad-10683	462	14	the	the	DET
dad-10683	462	15	turn	turn	NOUN
dad-10683	462	16	-	-	PUNCT
dad-10683	462	17	taking	take	VERB
dad-10683	462	18	structure	structure	NOUN
dad-10683	462	19	of	of	ADP
dad-10683	462	20	dialogue	dialogue	NOUN
dad-10683	462	21	.	.	PUNCT
dad-10683	463	1	for	for	ADP
dad-10683	463	2	the	the	DET
dad-10683	463	3	ubuntu	ubuntu	ADJ
dad-10683	463	4	dialogue	dialogue	NOUN
dad-10683	463	5	corpus	corpus	PROPN
dad-10683	463	6	,	,	PUNCT
dad-10683	463	7	interactions	interaction	NOUN
dad-10683	463	8	between	between	ADP
dad-10683	463	9	users	user	NOUN
dad-10683	463	10	usually	usually	ADV
dad-10683	463	11	take	take	VERB
dad-10683	463	12	a	a	DET
dad-10683	463	13	certain	certain	ADJ
dad-10683	463	14	form	form	NOUN
dad-10683	463	15	,	,	PUNCT
dad-10683	463	16	where	where	SCONJ
dad-10683	463	17	one	one	NUM
dad-10683	463	18	user	user	NOUN
dad-10683	463	19	is	be	AUX
dad-10683	463	20	asking	ask	VERB
dad-10683	463	21	for	for	ADP
dad-10683	463	22	help	help	NOUN
dad-10683	463	23	and	and	CCONJ
dad-10683	463	24	the	the	DET
dad-10683	463	25	other	other	ADJ
dad-10683	463	26	user	user	NOUN
dad-10683	463	27	is	be	AUX
dad-10683	463	28	providing	provide	VERB
dad-10683	463	29	answers	answer	NOUN
dad-10683	463	30	.	.	PUNCT
dad-10683	464	1	thus	thus	ADV
dad-10683	464	2	,	,	PUNCT
dad-10683	464	3	it	it	PRON
dad-10683	464	4	is	be	AUX
dad-10683	464	5	important	important	ADJ
dad-10683	464	6	to	to	PART
dad-10683	464	7	consider	consider	VERB
dad-10683	464	8	the	the	DET
dad-10683	464	9	role	role	NOUN
dad-10683	464	10	of	of	ADP
dad-10683	464	11	the	the	DET
dad-10683	464	12	current	current	ADJ
dad-10683	464	13	user	user	NOUN
dad-10683	464	14	when	when	SCONJ
dad-10683	464	15	selecting	select	VERB
dad-10683	464	16	the	the	DET
dad-10683	464	17	correct	correct	ADJ
dad-10683	464	18	response	response	NOUN
dad-10683	464	19	;	;	PUNCT
dad-10683	464	20	indeed	indeed	ADV
dad-10683	464	21	,	,	PUNCT
dad-10683	464	22	there	there	PRON
dad-10683	464	23	has	have	AUX
dad-10683	464	24	been	be	AUX
dad-10683	464	25	preliminary	preliminary	ADJ
dad-10683	464	26	work	work	NOUN
dad-10683	464	27	in	in	ADP
dad-10683	464	28	this	this	DET
dad-10683	464	29	direction	direction	NOUN
dad-10683	464	30	(	(	PUNCT
dad-10683	464	31	luan	luan	PROPN
dad-10683	464	32	et	et	PROPN
dad-10683	464	33	al	al	PROPN
dad-10683	464	34	.	.	PROPN
dad-10683	464	35	,	,	PUNCT
dad-10683	464	36	2016	2016	NUM
dad-10683	464	37	;	;	PUNCT
dad-10683	464	38	li	li	PROPN
dad-10683	464	39	et	et	PROPN
dad-10683	464	40	al	al	PROPN
dad-10683	464	41	.	.	PROPN
dad-10683	464	42	,	,	PUNCT
dad-10683	464	43	2016a	2016a	NUM
dad-10683	464	44	)	)	PUNCT
dad-10683	464	45	.	.	PUNCT
dad-10683	465	1	we	we	PRON
dad-10683	465	2	also	also	ADV
dad-10683	465	3	consider	consider	VERB
dad-10683	465	4	a	a	DET
dad-10683	465	5	special	special	ADJ
dad-10683	465	6	case	case	NOUN
dad-10683	465	7	of	of	ADP
dad-10683	465	8	this	this	DET
dad-10683	465	9	error	error	NOUN
dad-10683	465	10	,	,	PUNCT
dad-10683	465	11	when	when	SCONJ
dad-10683	465	12	the	the	DET
dad-10683	465	13	last	last	ADJ
dad-10683	465	14	utterance	utterance	NOUN
dad-10683	465	15	in	in	ADP
dad-10683	465	16	the	the	DET
dad-10683	465	17	context	context	NOUN
dad-10683	465	18	asks	ask	VERB
dad-10683	465	19	a	a	DET
dad-10683	465	20	question	question	NOUN
dad-10683	465	21	and	and	CCONJ
dad-10683	465	22	the	the	DET
dad-10683	465	23	response	response	NOUN
dad-10683	465	24	chosen	choose	VERB
dad-10683	465	25	by	by	ADP
dad-10683	465	26	the	the	DET
dad-10683	465	27	model	model	NOUN
dad-10683	465	28	is	be	AUX
dad-10683	465	29	not	not	PART
dad-10683	465	30	answering	answer	VERB
dad-10683	465	31	any	any	DET
dad-10683	465	32	question	question	NOUN
dad-10683	465	33	at	at	ADV
dad-10683	465	34	all	all	ADV
dad-10683	465	35	.	.	PUNCT
dad-10683	466	1	for	for	ADP
dad-10683	466	2	example	example	NOUN
dad-10683	466	3	,	,	PUNCT
dad-10683	466	4	in	in	ADP
dad-10683	466	5	figure	figure	NOUN
dad-10683	466	6	11	11	NUM
dad-10683	466	7	,	,	PUNCT
dad-10683	466	8	the	the	DET
dad-10683	466	9	final	final	ADJ
dad-10683	466	10	utterance	utterance	NOUN
dad-10683	466	11	asks	ask	VERB
dad-10683	466	12	the	the	DET
dad-10683	466	13	question	question	NOUN
dad-10683	466	14	‘	'	PUNCT
dad-10683	466	15	no	no	DET
dad-10683	466	16	mm	mm	NOUN
dad-10683	466	17	’s	’s	NOUN
dad-10683	466	18	?	?	PUNCT
dad-10683	466	19	’	'	PUNCT
dad-10683	466	20	.	.	PUNCT
dad-10683	467	1	the	the	DET
dad-10683	467	2	response	response	NOUN
dad-10683	467	3	selected	select	VERB
dad-10683	467	4	by	by	ADP
dad-10683	467	5	the	the	DET
dad-10683	467	6	model	model	NOUN
dad-10683	467	7	begins	begin	VERB
dad-10683	467	8	with	with	ADP
dad-10683	467	9	‘	'	PUNCT
dad-10683	467	10	thanks	thank	NOUN
dad-10683	467	11	’	'	PUNCT
dad-10683	467	12	,	,	PUNCT
dad-10683	467	13	which	which	PRON
dad-10683	467	14	is	be	AUX
dad-10683	467	15	clearly	clearly	ADV
dad-10683	467	16	not	not	PART
dad-10683	467	17	a	a	DET
dad-10683	467	18	reasonable	reasonable	ADJ
dad-10683	467	19	response	response	NOUN
dad-10683	467	20	to	to	ADP
dad-10683	467	21	a	a	DET
dad-10683	467	22	question	question	NOUN
dad-10683	467	23	.	.	PUNCT
dad-10683	468	1	an	an	DET
dad-10683	468	2	example	example	NOUN
dad-10683	468	3	of	of	ADP
dad-10683	468	4	the	the	DET
dad-10683	468	5	general	general	ADJ
dad-10683	468	6	turn	turn	NOUN
dad-10683	468	7	-	-	PUNCT
dad-10683	468	8	taking	take	VERB
dad-10683	468	9	error	error	NOUN
dad-10683	468	10	is	be	AUX
dad-10683	468	11	shown	show	VERB
dad-10683	468	12	in	in	ADP
dad-10683	468	13	figure	figure	NOUN
dad-10683	468	14	12	12	NUM
dad-10683	468	15	.	.	PUNCT
dad-10683	469	1	this	this	PRON
dad-10683	469	2	depicts	depict	VERB
dad-10683	469	3	a	a	DET
dad-10683	469	4	typical	typical	ADJ
dad-10683	469	5	dialogue	dialogue	NOUN
dad-10683	469	6	between	between	ADP
dad-10683	469	7	two	two	NUM
dad-10683	469	8	users	user	NOUN
dad-10683	469	9	in	in	ADP
dad-10683	469	10	the	the	DET
dad-10683	469	11	ubuntu	ubuntu	ADJ
dad-10683	469	12	dialogue	dialogue	NOUN
dad-10683	469	13	corpus	corpus	PROPN
dad-10683	469	14	:	:	PUNCT
dad-10683	469	15	speaker	speaker	NOUN
dad-10683	469	16	a	a	PRON
dad-10683	469	17	is	be	AUX
dad-10683	469	18	having	have	VERB
dad-10683	469	19	trouble	trouble	NOUN
dad-10683	469	20	with	with	ADP
dad-10683	469	21	their	their	PRON
dad-10683	469	22	brightness	brightness	NOUN
dad-10683	469	23	keys	key	NOUN
dad-10683	469	24	,	,	PUNCT
dad-10683	469	25	and	and	CCONJ
dad-10683	469	26	speaker	speaker	PROPN
dad-10683	469	27	b	b	PROPN
dad-10683	469	28	is	be	AUX
dad-10683	469	29	trying	try	VERB
dad-10683	469	30	to	to	PART
dad-10683	469	31	help	help	VERB
dad-10683	469	32	them	they	PRON
dad-10683	469	33	.	.	PUNCT
dad-10683	470	1	the	the	DET
dad-10683	470	2	model	model	NOUN
dad-10683	470	3	must	must	AUX
dad-10683	470	4	predict	predict	VERB
dad-10683	470	5	the	the	DET
dad-10683	470	6	next	next	ADJ
dad-10683	470	7	response	response	NOUN
dad-10683	470	8	of	of	ADP
dad-10683	470	9	speaker	speaker	NOUN
dad-10683	470	10	a.	a.	NOUN
dad-10683	470	11	in	in	ADP
dad-10683	470	12	the	the	DET
dad-10683	470	13	first	first	ADJ
dad-10683	470	14	response	response	NOUN
dad-10683	470	15	,	,	PUNCT
dad-10683	470	16	the	the	DET
dad-10683	470	17	user	user	NOUN
dad-10683	470	18	states	state	VERB
dad-10683	470	19	that	that	SCONJ
dad-10683	470	20	they	they	PRON
dad-10683	470	21	are	be	AUX
dad-10683	470	22	appreciative	appreciative	ADJ
dad-10683	470	23	of	of	ADP
dad-10683	470	24	the	the	DET
dad-10683	470	25	help	help	NOUN
dad-10683	470	26	being	be	AUX
dad-10683	470	27	given	give	VERB
dad-10683	470	28	,	,	PUNCT
dad-10683	470	29	which	which	PRON
dad-10683	470	30	fits	fit	VERB
dad-10683	470	31	with	with	ADP
dad-10683	470	32	speaker	speaker	NOUN
dad-10683	470	33	a	a	DET
dad-10683	470	34	’s	’s	ADJ
dad-10683	470	35	behaviour	behaviour	NOUN
dad-10683	470	36	in	in	ADP
dad-10683	470	37	the	the	DET
dad-10683	470	38	context	context	NOUN
dad-10683	470	39	;	;	PUNCT
dad-10683	470	40	thus	thus	ADV
dad-10683	470	41	,	,	PUNCT
dad-10683	470	42	it	it	PRON
dad-10683	470	43	is	be	AUX
dad-10683	470	44	more	more	ADV
dad-10683	470	45	likely	likely	ADJ
dad-10683	470	46	to	to	PART
dad-10683	470	47	be	be	AUX
dad-10683	470	48	the	the	DET
dad-10683	470	49	correct	correct	ADJ
dad-10683	470	50	response	response	NOUN
dad-10683	470	51	.	.	PUNCT
dad-10683	471	1	we	we	PRON
dad-10683	471	2	examined	examine	VERB
dad-10683	471	3	100	100	NUM
dad-10683	471	4	randomly	randomly	ADV
dad-10683	471	5	selected	select	VERB
dad-10683	471	6	errors	error	NOUN
dad-10683	471	7	of	of	ADP
dad-10683	471	8	the	the	DET
dad-10683	471	9	de	de	X
dad-10683	471	10	model	model	NOUN
dad-10683	471	11	on	on	ADP
dad-10683	471	12	the	the	DET
dad-10683	471	13	ubuntu	ubuntu	ADJ
dad-10683	471	14	corpus	corpus	PROPN
dad-10683	471	15	to	to	PART
dad-10683	471	16	compute	compute	VERB
dad-10683	471	17	the	the	DET
dad-10683	471	18	number	number	NOUN
dad-10683	471	19	of	of	ADP
dad-10683	471	20	errors	error	NOUN
dad-10683	471	21	in	in	ADP
dad-10683	471	22	each	each	DET
dad-10683	471	23	category	category	NOUN
dad-10683	471	24	;	;	PUNCT
dad-10683	471	25	the	the	DET
dad-10683	471	26	results	result	NOUN
dad-10683	471	27	are	be	AUX
dad-10683	471	28	shown	show	VERB
dad-10683	471	29	in	in	ADP
dad-10683	471	30	table	table	NOUN
dad-10683	471	31	13	13	NUM
dad-10683	471	32	.	.	PUNCT
dad-10683	472	1	we	we	PRON
dad-10683	472	2	can	can	AUX
dad-10683	472	3	first	first	ADV
dad-10683	472	4	note	note	VERB
dad-10683	472	5	that	that	SCONJ
dad-10683	472	6	there	there	PRON
dad-10683	472	7	is	be	VERB
dad-10683	472	8	significant	significant	ADJ
dad-10683	472	9	progress	progress	NOUN
dad-10683	472	10	to	to	PART
dad-10683	472	11	be	be	AUX
dad-10683	472	12	made	make	VERB
dad-10683	472	13	for	for	ADP
dad-10683	472	14	classification	classification	NOUN
dad-10683	472	15	models	model	NOUN
dad-10683	472	16	on	on	ADP
dad-10683	472	17	the	the	DET
dad-10683	472	18	ubuntu	ubuntu	ADJ
dad-10683	472	19	dialogue	dialogue	NOUN
dad-10683	472	20	corpus	corpu	NOUN
dad-10683	472	21	;	;	PUNCT
dad-10683	472	22	over	over	ADP
dad-10683	472	23	half	half	NOUN
dad-10683	472	24	(	(	PUNCT
dad-10683	472	25	60	60	NUM
dad-10683	472	26	%	%	NOUN
dad-10683	472	27	)	)	PUNCT
dad-10683	472	28	of	of	ADP
dad-10683	472	29	the	the	DET
dad-10683	472	30	errors	error	NOUN
dad-10683	472	31	made	make	VERB
dad-10683	472	32	by	by	ADP
dad-10683	472	33	the	the	DET
dad-10683	472	34	model	model	NOUN
dad-10683	472	35	can	can	AUX
dad-10683	472	36	be	be	AUX
dad-10683	472	37	considered	consider	VERB
dad-10683	472	38	feasible	feasible	ADJ
dad-10683	472	39	for	for	ADP
dad-10683	472	40	the	the	DET
dad-10683	472	41	majority	majority	NOUN
dad-10683	472	42	of	of	ADP
dad-10683	472	43	humans	human	NOUN
dad-10683	472	44	(	(	PUNCT
dad-10683	472	45	1	1	NUM
dad-10683	472	46	-	-	SYM
dad-10683	472	47	3	3	NUM
dad-10683	472	48	on	on	ADP
dad-10683	472	49	the	the	DET
dad-10683	472	50	difficulty	difficulty	NOUN
dad-10683	472	51	rating	rating	NOUN
dad-10683	472	52	)	)	PUNCT
dad-10683	472	53	.	.	PUNCT
dad-10683	473	1	however	however	ADV
dad-10683	473	2	,	,	PUNCT
dad-10683	473	3	the	the	DET
dad-10683	473	4	number	number	NOUN
dad-10683	473	5	of	of	ADP
dad-10683	473	6	questions	question	NOUN
dad-10683	473	7	that	that	PRON
dad-10683	473	8	every	every	DET
dad-10683	473	9	human	human	NOUN
dad-10683	473	10	could	could	AUX
dad-10683	473	11	50	50	NUM
dad-10683	473	12	training	training	NOUN
dad-10683	473	13	end	end	NOUN
dad-10683	473	14	-	-	PUNCT
dad-10683	473	15	to	to	ADP
dad-10683	473	16	-	-	PUNCT
dad-10683	473	17	end	end	NOUN
dad-10683	473	18	dialogue	dialogue	NOUN
dad-10683	473	19	systems	system	NOUN
dad-10683	473	20	context	context	NOUN
dad-10683	473	21	:	:	PUNCT
dad-10683	473	22	speaker	speaker	NOUN
dad-10683	473	23	a	a	NOUN
dad-10683	473	24	:	:	PUNCT
dad-10683	473	25	i	i	PRON
dad-10683	473	26	ca	can	AUX
dad-10683	473	27	n’t	not	PART
dad-10683	473	28	seem	seem	VERB
dad-10683	473	29	to	to	PART
dad-10683	473	30	get	get	VERB
dad-10683	473	31	audio	audio	NOUN
dad-10683	473	32	working	working	NOUN
dad-10683	473	33	as	as	ADP
dad-10683	473	34	a	a	DET
dad-10683	473	35	non	non	ADJ
dad-10683	473	36	-	-	ADJ
dad-10683	473	37	root	root	ADJ
dad-10683	473	38	user	user	NOUN
dad-10683	473	39	.	.	PUNCT
dad-10683	474	1	has	have	AUX
dad-10683	474	2	anyone	anyone	PRON
dad-10683	474	3	ever	ever	ADV
dad-10683	474	4	had	have	VERB
dad-10683	474	5	this	this	DET
dad-10683	474	6	problem	problem	NOUN
dad-10683	474	7	?	?	PUNCT
dad-10683	475	1	eou	eou	PROPN
dad-10683	475	2	speaker	speaker	PROPN
dad-10683	475	3	b	b	PROPN
dad-10683	475	4	:	:	PUNCT
dad-10683	475	5	alsamixer	alsamixer	NOUN
dad-10683	475	6	to	to	ADP
dad-10683	475	7	the	the	DET
dad-10683	475	8	rescue	rescue	NOUN
dad-10683	475	9	eou	eou	PROPN
dad-10683	475	10	speaker	speaker	NOUN
dad-10683	475	11	a	a	NOUN
dad-10683	475	12	:	:	PUNCT
dad-10683	475	13	alsamixer	alsamixer	PROPN
dad-10683	475	14	shows	show	VERB
dad-10683	475	15	everything	everything	PRON
dad-10683	475	16	turned	turn	VERB
dad-10683	475	17	on	on	ADP
dad-10683	475	18	,	,	PUNCT
dad-10683	475	19	and	and	CCONJ
dad-10683	475	20	looks	look	VERB
dad-10683	475	21	exactly	exactly	ADV
dad-10683	475	22	the	the	DET
dad-10683	475	23	same	same	ADJ
dad-10683	475	24	for	for	ADP
dad-10683	475	25	my	my	PRON
dad-10683	475	26	normal	normal	ADJ
dad-10683	475	27	user	user	NOUN
dad-10683	475	28	as	as	SCONJ
dad-10683	475	29	it	it	PRON
dad-10683	475	30	does	do	VERB
dad-10683	475	31	for	for	ADP
dad-10683	475	32	root	root	NOUN
dad-10683	475	33	eou	eou	PROPN
dad-10683	475	34	speaker	speaker	PROPN
dad-10683	475	35	b	b	PROPN
dad-10683	475	36	:	:	PUNCT
dad-10683	475	37	no	no	DET
dad-10683	475	38	mm	mm	NOUN
dad-10683	475	39	’s	’s	NOUN
dad-10683	475	40	?	?	PUNCT
dad-10683	476	1	eou	eou	PROPN
dad-10683	476	2	binary	binary	PROPN
dad-10683	476	3	probability	probability	PROPN
dad-10683	476	4	candidate	candidate	NOUN
dad-10683	476	5	responses	response	VERB
dad-10683	476	6	0.62	0.62	NUM
dad-10683	476	7	correct	correct	ADJ
dad-10683	476	8	eou	eou	PROPN
dad-10683	476	9	0.92	0.92	NUM
dad-10683	476	10	true	true	ADJ
dad-10683	476	11	but	but	CCONJ
dad-10683	476	12	how	how	SCONJ
dad-10683	476	13	will	will	AUX
dad-10683	476	14	he	he	PRON
dad-10683	476	15	find	find	VERB
dad-10683	476	16	my	my	PRON
dad-10683	476	17	new	new	ADJ
dad-10683	476	18	ip	ip	NOUN
dad-10683	476	19	so	so	ADV
dad-10683	476	20	easily	easily	ADV
dad-10683	476	21	if	if	SCONJ
dad-10683	476	22	i	i	PRON
dad-10683	476	23	get	get	VERB
dad-10683	476	24	it	it	PRON
dad-10683	476	25	changed	change	VERB
dad-10683	476	26	?	?	PUNCT
dad-10683	477	1	all	all	PRON
dad-10683	477	2	i	i	PRON
dad-10683	477	3	do	do	VERB
dad-10683	477	4	is	be	AUX
dad-10683	477	5	programming	program	VERB
dad-10683	477	6	c	c	PROPN
dad-10683	477	7	and	and	CCONJ
dad-10683	477	8	check	check	VERB
dad-10683	477	9	my	my	PRON
dad-10683	477	10	mail	mail	NOUN
dad-10683	477	11	usually	usually	ADV
dad-10683	477	12	eou	eou	PROPN
dad-10683	477	13	0.68	0.68	NUM
dad-10683	477	14	yes	yes	INTJ
dad-10683	477	15	strange	strange	ADJ
dad-10683	477	16	.	.	PUNCT
dad-10683	478	1	then	then	ADV
dad-10683	478	2	omit	omit	VERB
dad-10683	478	3	the	the	DET
dad-10683	478	4	dash	dash	NOUN
dad-10683	478	5	altogether	altogether	ADV
dad-10683	478	6	,	,	PUNCT
dad-10683	478	7	try	try	VERB
dad-10683	478	8	giving	give	VERB
dad-10683	478	9	set	set	VERB
dad-10683	478	10	default	default	NOUN
dad-10683	478	11	sink	sink	NOUN
dad-10683	478	12	:/	:/	PUNCT
dad-10683	478	13	eou	eou	PROPN
dad-10683	478	14	0.93	0.93	NUM
dad-10683	478	15	thanks	thank	NOUN
dad-10683	478	16	eou	eou	PROPN
dad-10683	478	17	where	where	SCONJ
dad-10683	478	18	is	be	AUX
dad-10683	478	19	the	the	DET
dad-10683	478	20	db	db	PROPN
dad-10683	478	21	app	app	NOUN
dad-10683	478	22	i	i	PRON
dad-10683	478	23	can	can	AUX
dad-10683	478	24	not	not	PART
dad-10683	478	25	locate	locate	VERB
dad-10683	478	26	it	it	PRON
dad-10683	478	27	(	(	PUNCT
dad-10683	478	28	sorry	sorry	ADJ
dad-10683	478	29	to	to	PART
dad-10683	478	30	be	be	AUX
dad-10683	478	31	such	such	DET
dad-10683	478	32	a	a	DET
dad-10683	478	33	noob	noob	NOUN
dad-10683	478	34	!	!	PUNCT
dad-10683	478	35	)	)	PUNCT
dad-10683	479	1	eou	eou	PROPN
dad-10683	479	2	0.03	0.03	NUM
dad-10683	479	3	i	i	NOUN
dad-10683	479	4	’m	’m	NOUN
dad-10683	479	5	switching	switch	VERB
dad-10683	479	6	the	the	DET
dad-10683	479	7	location	location	NOUN
dad-10683	479	8	to	to	ADP
dad-10683	479	9	my	my	PRON
dad-10683	479	10	on	on	ADP
dad-10683	479	11	board	board	PROPN
dad-10683	479	12	ssd	ssd	PROPN
dad-10683	479	13	drive	drive	NOUN
dad-10683	479	14	that	that	PRON
dad-10683	479	15	’s	’s	AUX
dad-10683	479	16	embedded	embed	VERB
dad-10683	479	17	to	to	ADP
dad-10683	479	18	the	the	DET
dad-10683	479	19	laptops	laptop	NOUN
dad-10683	479	20	board	board	NOUN
dad-10683	479	21	.	.	PUNCT
dad-10683	480	1	i	i	PRON
dad-10683	480	2	just	just	ADV
dad-10683	480	3	have	have	AUX
dad-10683	480	4	n’t	not	PART
dad-10683	480	5	been	be	AUX
dad-10683	480	6	using	use	VERB
dad-10683	480	7	the	the	DET
dad-10683	480	8	storage	storage	NOUN
dad-10683	480	9	so	so	SCONJ
dad-10683	480	10	i	i	PRON
dad-10683	480	11	figure	figure	VERB
dad-10683	480	12	i	i	PRON
dad-10683	480	13	could	could	AUX
dad-10683	480	14	try	try	VERB
dad-10683	480	15	and	and	CCONJ
dad-10683	480	16	utilize	utilize	VERB
dad-10683	480	17	the	the	DET
dad-10683	480	18	space	space	NOUN
dad-10683	480	19	while	while	SCONJ
dad-10683	480	20	the	the	DET
dad-10683	480	21	ram	ram	NOUN
dad-10683	480	22	being	be	AUX
dad-10683	480	23	8	8	NUM
dad-10683	480	24	gig	gig	NOUN
dad-10683	480	25	’s	’s	ADJ
dad-10683	480	26	itself	itself	PRON
dad-10683	480	27	i	i	PRON
dad-10683	480	28	see	see	VERB
dad-10683	480	29	no	no	DET
dad-10683	480	30	problem	problem	NOUN
dad-10683	480	31	with	with	ADP
dad-10683	480	32	the	the	DET
dad-10683	480	33	switch	switch	NOUN
dad-10683	480	34	.	.	PUNCT
dad-10683	481	1	do	do	AUX
dad-10683	481	2	you	you	PRON
dad-10683	481	3	understand	understand	VERB
dad-10683	481	4	what	what	PRON
dad-10683	481	5	i	i	PRON
dad-10683	481	6	’m	’m	AUX
dad-10683	481	7	doing	do	VERB
dad-10683	481	8	.	.	PUNCT
dad-10683	482	1	i	i	PRON
dad-10683	482	2	’m	’m	AUX
dad-10683	482	3	only	only	ADV
dad-10683	482	4	asking	ask	VERB
dad-10683	482	5	here	here	ADV
dad-10683	482	6	so	so	ADV
dad-10683	482	7	i	i	PRON
dad-10683	482	8	do	do	AUX
dad-10683	482	9	n’t	not	PART
dad-10683	482	10	go	go	VERB
dad-10683	482	11	screwing	screw	VERB
dad-10683	482	12	up	up	ADP
dad-10683	482	13	and	and	CCONJ
dad-10683	482	14	save	save	VERB
dad-10683	482	15	myself	myself	PRON
dad-10683	482	16	hours	hour	NOUN
dad-10683	482	17	of	of	ADP
dad-10683	482	18	headaches	headache	NOUN
dad-10683	482	19	eou	eou	PROPN
dad-10683	482	20	0.03	0.03	NUM
dad-10683	482	21	you	you	PRON
dad-10683	482	22	’ll	’ll	AUX
dad-10683	482	23	love	love	VERB
dad-10683	482	24	it	it	PRON
dad-10683	482	25	eou	eou	ADV
dad-10683	483	1	i	i	PRON
dad-10683	483	2	am	be	AUX
dad-10683	483	3	joking	joke	VERB
dad-10683	483	4	,	,	PUNCT
dad-10683	483	5	but	but	CCONJ
dad-10683	483	6	you	you	PRON
dad-10683	483	7	will	will	AUX
dad-10683	483	8	probably	probably	ADV
dad-10683	483	9	enjoy	enjoy	VERB
dad-10683	483	10	learning	learn	VERB
dad-10683	483	11	about	about	ADP
dad-10683	483	12	it	it	PRON
dad-10683	483	13	eou	eou	VERB
dad-10683	484	1	well	well	INTJ
dad-10683	484	2	it	it	PRON
dad-10683	484	3	’s	’	VERB
dad-10683	484	4	a	a	DET
dad-10683	484	5	step	step	NOUN
dad-10683	484	6	up	up	ADP
dad-10683	484	7	from	from	ADP
dad-10683	484	8	opening	open	VERB
dad-10683	484	9	your	your	PRON
dad-10683	484	10	hard	hard	ADJ
dad-10683	484	11	drive	drive	NOUN
dad-10683	484	12	up	up	ADP
dad-10683	484	13	and	and	CCONJ
dad-10683	484	14	using	use	VERB
dad-10683	484	15	a	a	DET
dad-10683	484	16	magnet	magnet	NOUN
dad-10683	484	17	eou	eou	PROPN
dad-10683	485	1	0.44	0.44	NUM
dad-10683	485	2	yeah	yeah	INTJ
dad-10683	485	3	,	,	PUNCT
dad-10683	485	4	512	512	NUM
dad-10683	485	5	is	be	AUX
dad-10683	485	6	plenty	plenty	ADJ
dad-10683	485	7	eou	eou	NOUN
dad-10683	485	8	0.20	0.20	NUM
dad-10683	485	9	i	i	PRON
dad-10683	485	10	use	use	VERB
dad-10683	485	11	it	it	PRON
dad-10683	485	12	on	on	ADP
dad-10683	485	13	a	a	DET
dad-10683	485	14	number	number	NOUN
dad-10683	485	15	of	of	ADP
dad-10683	485	16	machines	machine	NOUN
dad-10683	485	17	with	with	ADP
dad-10683	485	18	no	no	DET
dad-10683	485	19	problems	problem	NOUN
dad-10683	485	20	.	.	PUNCT
dad-10683	486	1	just	just	ADV
dad-10683	486	2	this	this	DET
dad-10683	486	3	one	one	NOUN
dad-10683	486	4	.	.	PUNCT
dad-10683	487	1	eou	eou	PROPN
dad-10683	487	2	modprobe	modprobe	PROPN
dad-10683	487	3	pulls	pull	VERB
dad-10683	487	4	up	up	ADP
dad-10683	487	5	a	a	DET
dad-10683	487	6	variety	variety	NOUN
dad-10683	487	7	of	of	ADP
dad-10683	487	8	mouse	mouse	NOUN
dad-10683	487	9	drivers	driver	NOUN
dad-10683	487	10	eou	eou	PROPN
dad-10683	487	11	0.62	0.62	NUM
dad-10683	487	12	so	so	SCONJ
dad-10683	487	13	the	the	DET
dad-10683	487	14	issue	issue	NOUN
dad-10683	487	15	is	be	AUX
dad-10683	487	16	*	*	ADJ
dad-10683	487	17	*	*	NOUN
dad-10683	487	18	unknown	unknown	ADJ
dad-10683	487	19	*	*	PUNCT
dad-10683	487	20	*	*	PUNCT
dad-10683	487	21	*	*	PUNCT
dad-10683	487	22	*	*	PUNCT
dad-10683	487	23	unknown	unknown	ADJ
dad-10683	487	24	*	*	NOUN
dad-10683	487	25	*	*	NOUN
dad-10683	487	26	.	.	PUNCT
dad-10683	488	1	gz	gz	PROPN
dad-10683	488	2	’	'	PUNCT
dad-10683	488	3	is	be	AUX
dad-10683	488	4	different	different	ADJ
dad-10683	488	5	from	from	ADP
dad-10683	488	6	the	the	DET
dad-10683	488	7	same	same	ADJ
dad-10683	488	8	file	file	NOUN
dad-10683	488	9	on	on	ADP
dad-10683	488	10	the	the	DET
dad-10683	488	11	system	system	NOUN
dad-10683	488	12	”	"	PUNCT
dad-10683	488	13	but	but	CCONJ
dad-10683	488	14	i	i	PRON
dad-10683	488	15	do	do	AUX
dad-10683	488	16	n’t	not	PART
dad-10683	488	17	have	have	VERB
dad-10683	488	18	any	any	DET
dad-10683	488	19	idea	idea	NOUN
dad-10683	488	20	why	why	SCONJ
dad-10683	488	21	/	/	ADP
dad-10683	488	22	what	what	PRON
dad-10683	488	23	that	that	PRON
dad-10683	488	24	means	mean	VERB
dad-10683	488	25	,	,	PUNCT
dad-10683	488	26	sorry	sorry	ADJ
dad-10683	488	27	.	.	PUNCT
dad-10683	489	1	best	good	ADJ
dad-10683	489	2	of	of	ADP
dad-10683	489	3	luck	luck	NOUN
dad-10683	489	4	.	.	PUNCT
dad-10683	490	1	eou	eou	PROPN
dad-10683	490	2	0.89	0.89	NUM
dad-10683	490	3	http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-680	http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-680	NOUN
dad-10683	490	4	m	m	VERB
dad-10683	490	5	comes	come	VERB
dad-10683	490	6	with	with	ADP
dad-10683	490	7	optimus	optimus	PROPN
dad-10683	490	8	technology	technology	NOUN
dad-10683	490	9	.	.	PUNCT
dad-10683	491	1	so	so	ADV
dad-10683	491	2	i	i	PRON
dad-10683	491	3	think	think	VERB
dad-10683	491	4	it	it	PRON
dad-10683	491	5	has	have	VERB
dad-10683	491	6	an	an	DET
dad-10683	491	7	onboard	onboard	ADJ
dad-10683	491	8	intel	intel	PROPN
dad-10683	491	9	card	card	NOUN
dad-10683	491	10	eou	eou	PROPN
dad-10683	491	11	figure	figure	NOUN
dad-10683	491	12	11	11	NUM
dad-10683	491	13	:	:	PUNCT
dad-10683	491	14	example	example	NOUN
dad-10683	491	15	where	where	SCONJ
dad-10683	491	16	the	the	DET
dad-10683	491	17	model	model	NOUN
dad-10683	491	18	selects	select	VERB
dad-10683	491	19	an	an	DET
dad-10683	491	20	inappropriate	inappropriate	ADJ
dad-10683	491	21	response	response	NOUN
dad-10683	491	22	to	to	ADP
dad-10683	491	23	a	a	DET
dad-10683	491	24	question	question	NOUN
dad-10683	491	25	.	.	PUNCT
dad-10683	492	1	correct	correct	ADJ
dad-10683	492	2	answer	answer	NOUN
dad-10683	492	3	is	be	AUX
dad-10683	492	4	in	in	ADP
dad-10683	492	5	italics	italic	NOUN
dad-10683	492	6	,	,	PUNCT
dad-10683	492	7	and	and	CCONJ
dad-10683	492	8	the	the	DET
dad-10683	492	9	model	model	NOUN
dad-10683	492	10	’s	’s	PART
dad-10683	492	11	selected	select	VERB
dad-10683	492	12	answer	answer	NOUN
dad-10683	492	13	is	be	AUX
dad-10683	492	14	in	in	ADP
dad-10683	492	15	bold	bold	ADJ
dad-10683	492	16	.	.	PUNCT
dad-10683	493	1	context	context	NOUN
dad-10683	493	2	:	:	PUNCT
dad-10683	493	3	speaker	speaker	NOUN
dad-10683	493	4	a	a	NOUN
dad-10683	493	5	:	:	PUNCT
dad-10683	493	6	hi	hi	INTJ
dad-10683	493	7	eou	eou	PROPN
dad-10683	494	1	i	i	PRON
dad-10683	494	2	have	have	VERB
dad-10683	494	3	a	a	DET
dad-10683	494	4	problem	problem	NOUN
dad-10683	494	5	with	with	ADP
dad-10683	494	6	fn	fn	NOUN
dad-10683	494	7	keys	key	NOUN
dad-10683	494	8	for	for	ADP
dad-10683	494	9	brightness	brightness	NOUN
dad-10683	494	10	with	with	ADP
dad-10683	494	11	my	my	PRON
dad-10683	494	12	laptop	laptop	NOUN
dad-10683	494	13	and	and	CCONJ
dad-10683	494	14	nvidia	nvidia	PROPN
dad-10683	494	15	propertiary	propertiary	PROPN
dad-10683	494	16	driver	driver	PROPN
dad-10683	494	17	eou	eou	PROPN
dad-10683	494	18	speaker	speaker	PROPN
dad-10683	494	19	b	b	PROPN
dad-10683	494	20	:	:	PUNCT
dad-10683	494	21	what	what	PRON
dad-10683	494	22	make	make	VERB
dad-10683	494	23	and	and	CCONJ
dad-10683	494	24	model	model	VERB
dad-10683	494	25	laptop	laptop	NOUN
dad-10683	494	26	?	?	PUNCT
dad-10683	495	1	¿	¿	PROPN
dad-10683	495	2	eou	eou	PROPN
dad-10683	495	3	speaker	speaker	NOUN
dad-10683	495	4	a	a	NOUN
dad-10683	495	5	:	:	PUNCT
dad-10683	495	6	sony	sony	PROPN
dad-10683	495	7	vaio	vaio	PROPN
dad-10683	495	8	vgn	vgn	PROPN
dad-10683	495	9	fz31z	fz31z	ADP
dad-10683	495	10	eou	eou	PROPN
dad-10683	495	11	and	and	CCONJ
dad-10683	495	12	i	i	PRON
dad-10683	495	13	m	m	AUX
dad-10683	495	14	using	use	VERB
dad-10683	495	15	nvidia	nvidia	PROPN
dad-10683	495	16	propertiary	propertiary	PROPN
dad-10683	495	17	driver	driver	PROPN
dad-10683	495	18	version	version	PROPN
dad-10683	495	19	current	current	NOUN
dad-10683	495	20	(	(	PUNCT
dad-10683	495	21	recommended	recommend	VERB
dad-10683	495	22	one	one	NUM
dad-10683	495	23	)	)	PUNCT
dad-10683	495	24	eou	eou	PROPN
dad-10683	495	25	speaker	speaker	PROPN
dad-10683	495	26	b	b	PROPN
dad-10683	495	27	:	:	PUNCT
dad-10683	495	28	try	try	VERB
dad-10683	495	29	the	the	DET
dad-10683	495	30	boot	boot	NOUN
dad-10683	495	31	option	option	NOUN
dad-10683	495	32	:	:	PUNCT
dad-10683	495	33	acpi	acpi	ADJ
dad-10683	495	34	backlight	backlight	ADJ
dad-10683	495	35	=	=	NOUN
dad-10683	495	36	vendor	vendor	NOUN
dad-10683	495	37	eou	eou	PROPN
dad-10683	495	38	speaker	speaker	PROPN
dad-10683	495	39	a	a	NOUN
dad-10683	495	40	:	:	PUNCT
dad-10683	495	41	i	i	PRON
dad-10683	495	42	have	have	AUX
dad-10683	495	43	added	add	VERB
dad-10683	495	44	acpi	acpi	ADJ
dad-10683	495	45	backlight	backlight	NOUN
dad-10683	495	46	for	for	ADP
dad-10683	495	47	vendor	vendor	NOUN
dad-10683	495	48	i	i	PRON
dad-10683	495	49	have	have	AUX
dad-10683	495	50	updated	update	VERB
dad-10683	495	51	grub	grub	NOUN
dad-10683	495	52	but	but	CCONJ
dad-10683	495	53	the	the	DET
dad-10683	495	54	keys	key	NOUN
dad-10683	495	55	are	be	AUX
dad-10683	495	56	not	not	PART
dad-10683	495	57	working	work	VERB
dad-10683	495	58	eou	eou	PROPN
dad-10683	495	59	my	my	PRON
dad-10683	495	60	grub	grub	NOUN
dad-10683	495	61	cmd	cmd	NOUN
dad-10683	495	62	line	line	PROPN
dad-10683	495	63	linxu	linxu	PROPN
dad-10683	495	64	default	default	NOUN
dad-10683	495	65	:	:	PUNCT
dad-10683	495	66	quiet	quiet	ADJ
dad-10683	495	67	splash	splash	NOUN
dad-10683	495	68	acpi	acpi	ADJ
dad-10683	495	69	backlight	backlight	ADJ
dad-10683	495	70	=	=	NOUN
dad-10683	495	71	vendor	vendor	NOUN
dad-10683	495	72	eou	eou	PROPN
dad-10683	495	73	speaker	speaker	PROPN
dad-10683	495	74	b	b	PROPN
dad-10683	495	75	:	:	PUNCT
dad-10683	495	76	try	try	VERB
dad-10683	495	77	the	the	DET
dad-10683	495	78	boot	boot	NOUN
dad-10683	495	79	option	option	NOUN
dad-10683	495	80	:	:	PUNCT
dad-10683	495	81	acpi	acpi	ADJ
dad-10683	495	82	osi	osi	PROPN
dad-10683	495	83	=	=	SYM
dad-10683	495	84	linux	linux	PROPN
dad-10683	495	85	eou	eou	PROPN
dad-10683	495	86	speaker	speaker	NOUN
dad-10683	495	87	a	a	NOUN
dad-10683	495	88	:	:	PUNCT
dad-10683	495	89	ok	ok	INTJ
dad-10683	495	90	i	i	PRON
dad-10683	495	91	must	must	AUX
dad-10683	495	92	remove	remove	VERB
dad-10683	495	93	the	the	DET
dad-10683	495	94	acpi	acpi	ADJ
dad-10683	495	95	backlihgt/	backlihgt/	NUM
dad-10683	495	96	eou	eou	PROPN
dad-10683	495	97	speaker	speaker	NOUN
dad-10683	495	98	b	b	PROPN
dad-10683	495	99	:	:	PUNCT
dad-10683	495	100	i	i	PROPN
dad-10683	495	101	’d	’d	NOUN
dad-10683	495	102	also	also	ADV
dad-10683	495	103	report	report	VERB
dad-10683	495	104	a	a	DET
dad-10683	495	105	bug	bug	NOUN
dad-10683	495	106	eou	eou	PROPN
dad-10683	495	107	could	could	AUX
dad-10683	495	108	try	try	VERB
dad-10683	495	109	quantal	quantal	ADJ
dad-10683	495	110	livecd	livecd	NOUN
dad-10683	495	111	to	to	PART
dad-10683	495	112	see	see	VERB
dad-10683	495	113	if	if	SCONJ
dad-10683	495	114	the	the	DET
dad-10683	495	115	newer	new	ADJ
dad-10683	495	116	kernel	kernel	NOUN
dad-10683	495	117	plays	play	VERB
dad-10683	495	118	nicer	nice	ADJ
dad-10683	495	119	eou	eou	PROPN
dad-10683	495	120	speaker	speaker	NOUN
dad-10683	495	121	a	a	NOUN
dad-10683	495	122	:	:	PUNCT
dad-10683	495	123	i	i	PRON
dad-10683	495	124	think	think	VERB
dad-10683	495	125	that	that	PRON
dad-10683	495	126	is	be	AUX
dad-10683	495	127	a	a	DET
dad-10683	495	128	nvidia	nvidia	ADJ
dad-10683	495	129	problem	problem	NOUN
dad-10683	495	130	with	with	ADP
dad-10683	495	131	the	the	DET
dad-10683	495	132	propertiary	propertiary	NOUN
dad-10683	495	133	eou	eou	PROPN
dad-10683	495	134	speaker	speaker	PROPN
dad-10683	495	135	b	b	PROPN
dad-10683	495	136	:	:	PUNCT
dad-10683	495	137	possibly	possibly	ADV
dad-10683	495	138	,	,	PUNCT
dad-10683	495	139	or	or	CCONJ
dad-10683	495	140	it	it	PRON
dad-10683	495	141	could	could	AUX
dad-10683	495	142	be	be	AUX
dad-10683	495	143	acpi	acpi	NOUN
dad-10683	495	144	based	base	VERB
dad-10683	495	145	eou	eou	PROPN
dad-10683	495	146	speaker	speaker	PROPN
dad-10683	495	147	a	a	NOUN
dad-10683	495	148	:	:	PUNCT
dad-10683	495	149	ok	ok	INTJ
dad-10683	495	150	thank	thank	VERB
dad-10683	495	151	you	you	PRON
dad-10683	495	152	i	i	PRON
dad-10683	495	153	must	must	AUX
dad-10683	495	154	remove	remove	VERB
dad-10683	495	155	the	the	DET
dad-10683	495	156	previous	previous	ADJ
dad-10683	495	157	about	about	ADP
dad-10683	495	158	the	the	DET
dad-10683	495	159	vendor	vendor	NOUN
dad-10683	495	160	ok	ok	INTJ
dad-10683	495	161	?	?	PUNCT
dad-10683	496	1	eou	eou	PROPN
dad-10683	496	2	speaker	speaker	PROPN
dad-10683	496	3	b	b	PROPN
dad-10683	496	4	:	:	PUNCT
dad-10683	496	5	could	could	AUX
dad-10683	496	6	try	try	VERB
dad-10683	496	7	both	both	PRON
dad-10683	496	8	and	and	CCONJ
dad-10683	496	9	then	then	ADV
dad-10683	496	10	just	just	ADV
dad-10683	496	11	one	one	NUM
dad-10683	496	12	eou	eou	PROPN
dad-10683	496	13	binary	binary	ADJ
dad-10683	496	14	probability	probability	NOUN
dad-10683	496	15	candidate	candidate	NOUN
dad-10683	496	16	responses	response	VERB
dad-10683	496	17	0.38	0.38	NUM
dad-10683	496	18	thanks	thank	NOUN
dad-10683	496	19	for	for	ADP
dad-10683	496	20	the	the	DET
dad-10683	496	21	help	help	NOUN
dad-10683	496	22	.	.	PUNCT
dad-10683	497	1	trying	try	VERB
dad-10683	497	2	now	now	ADV
dad-10683	497	3	.	.	PUNCT
dad-10683	498	1	is	be	AUX
dad-10683	498	2	there	there	PRON
dad-10683	498	3	any	any	DET
dad-10683	498	4	other	other	ADJ
dad-10683	498	5	same	same	ADJ
dad-10683	498	6	bug	bug	NOUN
dad-10683	498	7	report	report	NOUN
dad-10683	498	8	for	for	ADP
dad-10683	498	9	vaio/	vaio/	NUM
dad-10683	498	10	eou	eou	NOUN
dad-10683	498	11	0.20	0.20	NUM
dad-10683	498	12	it	it	PRON
dad-10683	498	13	’s	’	VERB
dad-10683	498	14	actually	actually	ADV
dad-10683	498	15	ubuntu	ubuntu	ADJ
dad-10683	498	16	support	support	NOUN
dad-10683	498	17	,	,	PUNCT
dad-10683	498	18	since	since	SCONJ
dad-10683	498	19	i	i	PRON
dad-10683	498	20	’m	’m	AUX
dad-10683	498	21	using	use	VERB
dad-10683	498	22	ubuntu	ubuntu	NOUN
dad-10683	498	23	,	,	PUNCT
dad-10683	498	24	is	be	AUX
dad-10683	498	25	n’t	not	PART
dad-10683	498	26	it	it	PRON
dad-10683	498	27	?	?	PUNCT
dad-10683	499	1	eou	eou	PROPN
dad-10683	500	1	0.49	0.49	NUM
dad-10683	500	2	yes	yes	NOUN
dad-10683	500	3	eou	eou	PROPN
dad-10683	500	4	the	the	DET
dad-10683	500	5	usb	usb	NOUN
dad-10683	500	6	disk	disk	NOUN
dad-10683	500	7	will	will	AUX
dad-10683	500	8	just	just	ADV
dad-10683	500	9	be	be	AUX
dad-10683	500	10	seen	see	VERB
dad-10683	500	11	as	as	ADP
dad-10683	500	12	a	a	DET
dad-10683	500	13	hard	hard	ADJ
dad-10683	500	14	disk	disk	NOUN
dad-10683	500	15	,	,	PUNCT
dad-10683	500	16	install	install	VERB
dad-10683	500	17	to	to	ADP
dad-10683	500	18	it	it	PRON
dad-10683	500	19	eou	eou	PROPN
dad-10683	500	20	0.50	0.50	NUM
dad-10683	500	21	if	if	SCONJ
dad-10683	500	22	you	you	PRON
dad-10683	500	23	do	do	VERB
dad-10683	500	24	unattended	unattended	ADJ
dad-10683	500	25	-	-	PUNCT
dad-10683	500	26	upgrades	upgrade	NOUN
dad-10683	500	27	-d	-d	PUNCT
dad-10683	500	28	,	,	PUNCT
dad-10683	500	29	that	that	PRON
dad-10683	500	30	might	might	AUX
dad-10683	500	31	tell	tell	VERB
dad-10683	500	32	you	you	PRON
dad-10683	500	33	a	a	DET
dad-10683	500	34	few	few	ADJ
dad-10683	500	35	things	thing	NOUN
dad-10683	500	36	?	?	PUNCT
dad-10683	501	1	eou	eou	PROPN
dad-10683	501	2	0.58	0.58	PROPN
dad-10683	501	3	does	do	AUX
dad-10683	501	4	this	this	PRON
dad-10683	501	5	have	have	VERB
dad-10683	501	6	’	'	PUNCT
dad-10683	501	7	open	open	ADJ
dad-10683	501	8	terminal	terminal	NOUN
dad-10683	501	9	here	here	ADV
dad-10683	501	10	’	'	PUNCT
dad-10683	501	11	and	and	CCONJ
dad-10683	501	12	’	'	PUNCT
dad-10683	501	13	2pane	2pane	NUM
dad-10683	501	14	mode	mode	NOUN
dad-10683	501	15	’	'	PUNCT
dad-10683	501	16	options	option	NOUN
dad-10683	501	17	?	?	PUNCT
dad-10683	502	1	eou	eou	PROPN
dad-10683	502	2	found	find	VERB
dad-10683	502	3	terminal	terminal	ADJ
dad-10683	502	4	option	option	NOUN
dad-10683	502	5	,	,	PUNCT
dad-10683	502	6	just	just	ADV
dad-10683	502	7	looking	look	VERB
dad-10683	502	8	for	for	ADP
dad-10683	502	9	2pane	2pane	PROPN
dad-10683	502	10	eou	eou	NOUN
dad-10683	502	11	0.71	0.71	NUM
dad-10683	502	12	it	it	PRON
dad-10683	502	13	’s	’	VERB
dad-10683	502	14	cool	cool	ADJ
dad-10683	502	15	eou	eou	PROPN
dad-10683	502	16	0.02	0.02	NUM
dad-10683	502	17	it	it	PRON
dad-10683	502	18	’s	’	VERB
dad-10683	502	19	like	like	ADP
dad-10683	502	20	hotel	hotel	NOUN
dad-10683	502	21	internet	internet	NOUN
dad-10683	502	22	eou	eou	PROPN
dad-10683	502	23	http://www.fdlinux.com/networksetuphowto.html	http://www.fdlinux.com/networksetuphowto.html	PROPN
dad-10683	502	24	eou	eou	PROPN
dad-10683	502	25	0.01	0.01	NUM
dad-10683	502	26	i	i	PRON
dad-10683	502	27	’ll	’ll	AUX
dad-10683	502	28	check	check	VERB
dad-10683	502	29	what	what	PRON
dad-10683	502	30	it	it	PRON
dad-10683	502	31	means	mean	VERB
dad-10683	502	32	in	in	ADP
dad-10683	502	33	google	google	PROPN
dad-10683	502	34	.	.	PUNCT
dad-10683	503	1	thank	thank	VERB
dad-10683	503	2	you	you	PRON
dad-10683	503	3	.	.	PUNCT
dad-10683	504	1	eou	eou	PROPN
dad-10683	504	2	0.36	0.36	NUM
dad-10683	504	3	i	i	PRON
dad-10683	504	4	never	never	ADV
dad-10683	504	5	liked	like	VERB
dad-10683	504	6	it	it	PRON
dad-10683	504	7	...	...	PUNCT
dad-10683	504	8	for	for	ADP
dad-10683	504	9	thin	thin	ADJ
dad-10683	504	10	versions	version	NOUN
dad-10683	504	11	,	,	PUNCT
dad-10683	504	12	i	i	PRON
dad-10683	504	13	use	use	VERB
dad-10683	504	14	fluxbox	fluxbox	NOUN
dad-10683	504	15	or	or	CCONJ
dad-10683	504	16	some	some	DET
dad-10683	504	17	other	other	ADJ
dad-10683	504	18	window	window	NOUN
dad-10683	504	19	manager	manager	NOUN
dad-10683	504	20	eou	eou	PROPN
dad-10683	504	21	0.49	0.49	NUM
dad-10683	504	22	so	so	ADV
dad-10683	504	23	do	do	AUX
dad-10683	504	24	i	i	PRON
dad-10683	504	25	just	just	ADV
dad-10683	504	26	paste	paste	VERB
dad-10683	504	27	that	that	DET
dad-10683	504	28	code	code	NOUN
dad-10683	504	29	in	in	ADP
dad-10683	504	30	to	to	ADP
dad-10683	504	31	the	the	DET
dad-10683	504	32	beginning	beginning	NOUN
dad-10683	504	33	of	of	ADP
dad-10683	504	34	the	the	DET
dad-10683	504	35	script	script	NOUN
dad-10683	504	36	...	...	PUNCT
dad-10683	504	37	?	?	PUNCT
dad-10683	505	1	eou	eou	PROPN
dad-10683	505	2	sorry	sorry	PROPN
dad-10683	505	3	experienced	experience	VERB
dad-10683	505	4	linux	linux	PROPN
dad-10683	505	5	user	user	NOUN
dad-10683	505	6	,	,	PUNCT
dad-10683	505	7	very	very	ADV
dad-10683	505	8	very	very	ADV
dad-10683	505	9	novice	novice	NOUN
dad-10683	505	10	coder	coder	NOUN
dad-10683	505	11	;	;	PUNCT
dad-10683	505	12	-p	-p	PUNCT
dad-10683	505	13	eou	eou	PROPN
dad-10683	505	14	figure	figure	NOUN
dad-10683	505	15	12	12	NUM
dad-10683	505	16	:	:	PUNCT
dad-10683	505	17	example	example	NOUN
dad-10683	505	18	where	where	SCONJ
dad-10683	505	19	the	the	DET
dad-10683	505	20	model	model	NOUN
dad-10683	505	21	does	do	AUX
dad-10683	505	22	not	not	PART
dad-10683	505	23	take	take	VERB
dad-10683	505	24	into	into	ADP
dad-10683	505	25	account	account	NOUN
dad-10683	505	26	the	the	DET
dad-10683	505	27	roles	role	NOUN
dad-10683	505	28	of	of	ADP
dad-10683	505	29	the	the	DET
dad-10683	505	30	participants	participant	NOUN
dad-10683	505	31	in	in	ADP
dad-10683	505	32	the	the	DET
dad-10683	505	33	dialogue	dialogue	NOUN
dad-10683	505	34	.	.	PUNCT
dad-10683	506	1	correct	correct	ADJ
dad-10683	506	2	answer	answer	NOUN
dad-10683	506	3	is	be	AUX
dad-10683	506	4	in	in	ADP
dad-10683	506	5	italics	italic	NOUN
dad-10683	506	6	,	,	PUNCT
dad-10683	506	7	and	and	CCONJ
dad-10683	506	8	the	the	DET
dad-10683	506	9	model	model	NOUN
dad-10683	506	10	’s	’s	PART
dad-10683	506	11	selected	select	VERB
dad-10683	506	12	answer	answer	NOUN
dad-10683	506	13	is	be	AUX
dad-10683	506	14	in	in	ADP
dad-10683	506	15	bold	bold	ADJ
dad-10683	506	16	.	.	PUNCT
dad-10683	507	1	51	51	NUM
dad-10683	507	2	lowe	lowe	PROPN
dad-10683	507	3	,	,	PUNCT
dad-10683	507	4	pow	pow	PROPN
dad-10683	507	5	,	,	PUNCT
dad-10683	507	6	serban	serban	PROPN
dad-10683	507	7	,	,	PUNCT
dad-10683	507	8	charlinn	charlinn	PROPN
dad-10683	507	9	,	,	PUNCT
dad-10683	507	10	liu	liu	PROPN
dad-10683	507	11	and	and	CCONJ
dad-10683	507	12	pineau	pineau	PROPN
dad-10683	507	13	difficulty	difficulty	NOUN
dad-10683	507	14	rating	rating	NOUN
dad-10683	507	15	(	(	PUNCT
dad-10683	507	16	1	1	NUM
dad-10683	507	17	-	-	SYM
dad-10683	507	18	5	5	NUM
dad-10683	507	19	)	)	PUNCT
dad-10683	507	20	number	number	NOUN
dad-10683	507	21	of	of	ADP
dad-10683	507	22	errors	error	NOUN
dad-10683	507	23	%	%	NOUN
dad-10683	507	24	of	of	ADP
dad-10683	507	25	errors	error	NOUN
dad-10683	507	26	impossible	impossible	ADJ
dad-10683	507	27	(	(	PUNCT
dad-10683	507	28	5	5	NUM
dad-10683	507	29	)	)	PUNCT
dad-10683	507	30	19	19	NUM
dad-10683	507	31	19	19	NUM
dad-10683	507	32	%	%	NOUN
dad-10683	507	33	very	very	ADV
dad-10683	507	34	difficult	difficult	ADJ
dad-10683	507	35	(	(	PUNCT
dad-10683	507	36	4	4	NUM
dad-10683	507	37	)	)	PUNCT
dad-10683	507	38	21	21	NUM
dad-10683	507	39	21	21	NUM
dad-10683	507	40	%	%	NOUN
dad-10683	507	41	difficult	difficult	ADJ
dad-10683	507	42	(	(	PUNCT
dad-10683	507	43	3	3	NUM
dad-10683	507	44	)	)	PUNCT
dad-10683	507	45	22	22	NUM
dad-10683	507	46	22	22	NUM
dad-10683	507	47	%	%	NOUN
dad-10683	507	48	moderate	moderate	ADJ
dad-10683	507	49	(	(	PUNCT
dad-10683	507	50	2	2	NUM
dad-10683	507	51	)	)	PUNCT
dad-10683	507	52	25	25	NUM
dad-10683	507	53	25	25	NUM
dad-10683	507	54	%	%	NOUN
dad-10683	507	55	easy	easy	ADJ
dad-10683	508	1	(	(	PUNCT
dad-10683	509	1	1	1	NUM
dad-10683	509	2	)	)	PUNCT
dad-10683	509	3	13	13	NUM
dad-10683	509	4	13	13	NUM
dad-10683	509	5	%	%	NOUN
dad-10683	509	6	model	model	NOUN
dad-10683	509	7	response	response	NOUN
dad-10683	509	8	rating	rating	NOUN
dad-10683	509	9	(	(	PUNCT
dad-10683	509	10	1	1	NUM
dad-10683	509	11	-	-	SYM
dad-10683	509	12	3	3	NUM
dad-10683	509	13	)	)	PUNCT
dad-10683	509	14	very	very	ADV
dad-10683	509	15	reasonable	reasonable	ADJ
dad-10683	509	16	(	(	PUNCT
dad-10683	509	17	3	3	NUM
dad-10683	509	18	)	)	PUNCT
dad-10683	509	19	14	14	NUM
dad-10683	509	20	14	14	NUM
dad-10683	509	21	%	%	NOUN
dad-10683	509	22	somewhat	somewhat	ADV
dad-10683	509	23	reasonable	reasonable	ADJ
dad-10683	509	24	(	(	PUNCT
dad-10683	509	25	2	2	NUM
dad-10683	509	26	)	)	PUNCT
dad-10683	509	27	37	37	NUM
dad-10683	509	28	37	37	NUM
dad-10683	509	29	%	%	NOUN
dad-10683	509	30	unreasonable	unreasonable	ADJ
dad-10683	509	31	(	(	PUNCT
dad-10683	509	32	1	1	NUM
dad-10683	509	33	)	)	PUNCT
dad-10683	509	34	49	49	NUM
dad-10683	509	35	49	49	NUM
dad-10683	509	36	%	%	NOUN
dad-10683	509	37	error	error	NOUN
dad-10683	509	38	category	category	NOUN
dad-10683	509	39	tone	tone	NOUN
dad-10683	509	40	and	and	CCONJ
dad-10683	509	41	style	style	NOUN
dad-10683	509	42	8	8	NUM
dad-10683	509	43	9	9	NUM
dad-10683	509	44	%	%	NOUN
dad-10683	509	45	knowledge	knowledge	NOUN
dad-10683	509	46	18	18	NUM
dad-10683	509	47	20	20	NUM
dad-10683	509	48	%	%	NOUN
dad-10683	509	49	semantic	semantic	ADJ
dad-10683	509	50	similarity	similarity	NOUN
dad-10683	509	51	45	45	NUM
dad-10683	509	52	49	49	NUM
dad-10683	509	53	%	%	NOUN
dad-10683	509	54	word	word	NOUN
dad-10683	509	55	copying	copy	VERB
dad-10683	509	56	11	11	NUM
dad-10683	509	57	12	12	NUM
dad-10683	509	58	%	%	NOUN
dad-10683	509	59	high	high	ADJ
dad-10683	509	60	-	-	PUNCT
dad-10683	509	61	level	level	NOUN
dad-10683	509	62	inference	inference	NOUN
dad-10683	509	63	16	16	NUM
dad-10683	509	64	18	18	NUM
dad-10683	509	65	%	%	NOUN
dad-10683	509	66	turn	turn	NOUN
dad-10683	509	67	-	-	PUNCT
dad-10683	509	68	taking	take	VERB
dad-10683	509	69	structure	structure	NOUN
dad-10683	509	70	20	20	NUM
dad-10683	509	71	22	22	NUM
dad-10683	509	72	%	%	NOUN
dad-10683	509	73	answering	answer	VERB
dad-10683	509	74	questions	question	NOUN
dad-10683	509	75	6	6	NUM
dad-10683	509	76	7	7	NUM
dad-10683	509	77	%	%	NOUN
dad-10683	509	78	figure	figure	NOUN
dad-10683	509	79	13	13	NUM
dad-10683	509	80	:	:	PUNCT
dad-10683	509	81	qualitative	qualitative	ADJ
dad-10683	509	82	evaluation	evaluation	NOUN
dad-10683	509	83	of	of	ADP
dad-10683	509	84	the	the	DET
dad-10683	509	85	errors	error	NOUN
dad-10683	509	86	from	from	ADP
dad-10683	509	87	the	the	DET
dad-10683	509	88	de	de	PROPN
dad-10683	509	89	model	model	NOUN
dad-10683	509	90	.	.	PUNCT
dad-10683	510	1	note	note	VERB
dad-10683	510	2	that	that	SCONJ
dad-10683	510	3	counts	count	VERB
dad-10683	510	4	for	for	ADP
dad-10683	510	5	parent	parent	NOUN
dad-10683	510	6	categories	category	NOUN
dad-10683	510	7	(	(	PUNCT
dad-10683	510	8	semantic	semantic	ADJ
dad-10683	510	9	similarity	similarity	NOUN
dad-10683	510	10	and	and	CCONJ
dad-10683	510	11	turn	turn	NOUN
dad-10683	510	12	-	-	PUNCT
dad-10683	510	13	taking	take	VERB
dad-10683	510	14	structure	structure	NOUN
dad-10683	510	15	)	)	PUNCT
dad-10683	510	16	include	include	VERB
dad-10683	510	17	the	the	DET
dad-10683	510	18	counts	count	NOUN
dad-10683	510	19	for	for	ADP
dad-10683	510	20	the	the	DET
dad-10683	510	21	child	child	NOUN
dad-10683	510	22	categories	category	NOUN
dad-10683	510	23	.	.	PUNCT
dad-10683	511	1	error	error	NOUN
dad-10683	511	2	categories	category	NOUN
dad-10683	511	3	are	be	AUX
dad-10683	511	4	not	not	PART
dad-10683	511	5	classified	classify	VERB
dad-10683	511	6	for	for	ADP
dad-10683	511	7	impossible	impossible	ADJ
dad-10683	511	8	questions	question	NOUN
dad-10683	511	9	and	and	CCONJ
dad-10683	511	10	are	be	AUX
dad-10683	511	11	not	not	PART
dad-10683	511	12	mutually	mutually	ADV
dad-10683	511	13	exclusive	exclusive	ADJ
dad-10683	511	14	,	,	PUNCT
dad-10683	511	15	thus	thus	ADV
dad-10683	511	16	totals	total	NOUN
dad-10683	511	17	may	may	AUX
dad-10683	511	18	not	not	PART
dad-10683	511	19	add	add	VERB
dad-10683	511	20	up	up	ADP
dad-10683	511	21	to	to	PART
dad-10683	511	22	100	100	NUM
dad-10683	511	23	.	.	PUNCT
dad-10683	512	1	answer	answer	NOUN
dad-10683	512	2	unconditionally	unconditionally	ADV
dad-10683	512	3	is	be	AUX
dad-10683	512	4	small	small	ADJ
dad-10683	512	5	,	,	PUNCT
dad-10683	512	6	as	as	SCONJ
dad-10683	512	7	technical	technical	ADJ
dad-10683	512	8	language	language	NOUN
dad-10683	512	9	can	can	AUX
dad-10683	512	10	often	often	ADV
dad-10683	512	11	be	be	AUX
dad-10683	512	12	confusing	confusing	ADJ
dad-10683	512	13	for	for	ADP
dad-10683	512	14	people	people	NOUN
dad-10683	512	15	who	who	PRON
dad-10683	512	16	are	be	AUX
dad-10683	512	17	unaccustomed	unaccustomed	ADJ
dad-10683	512	18	to	to	ADP
dad-10683	512	19	it	it	PRON
dad-10683	512	20	.	.	PUNCT
dad-10683	513	1	the	the	DET
dad-10683	513	2	other	other	ADJ
dad-10683	513	3	questions	question	NOUN
dad-10683	513	4	are	be	AUX
dad-10683	513	5	roughly	roughly	ADV
dad-10683	513	6	uniformly	uniformly	ADV
dad-10683	513	7	distributed	distribute	VERB
dad-10683	513	8	over	over	ADP
dad-10683	513	9	the	the	DET
dad-10683	513	10	remaining	remain	VERB
dad-10683	513	11	levels	level	NOUN
dad-10683	513	12	of	of	ADP
dad-10683	513	13	difficulty	difficulty	NOUN
dad-10683	513	14	,	,	PUNCT
dad-10683	513	15	from	from	ADP
dad-10683	513	16	moderate	moderate	ADJ
dad-10683	513	17	to	to	ADP
dad-10683	513	18	impossible	impossible	ADJ
dad-10683	513	19	.	.	PUNCT
dad-10683	514	1	we	we	PRON
dad-10683	514	2	also	also	ADV
dad-10683	514	3	note	note	VERB
dad-10683	514	4	that	that	SCONJ
dad-10683	514	5	there	there	PRON
dad-10683	514	6	are	be	VERB
dad-10683	514	7	a	a	DET
dad-10683	514	8	large	large	ADJ
dad-10683	514	9	number	number	NOUN
dad-10683	514	10	of	of	ADP
dad-10683	514	11	cases	case	NOUN
dad-10683	514	12	(	(	PUNCT
dad-10683	514	13	49	49	NUM
dad-10683	514	14	%	%	NOUN
dad-10683	514	15	)	)	PUNCT
dad-10683	514	16	where	where	SCONJ
dad-10683	514	17	the	the	DET
dad-10683	514	18	response	response	NOUN
dad-10683	514	19	retrieved	retrieve	VERB
dad-10683	514	20	by	by	ADP
dad-10683	514	21	the	the	DET
dad-10683	514	22	model	model	NOUN
dad-10683	514	23	was	be	AUX
dad-10683	514	24	completely	completely	ADV
dad-10683	514	25	unreasonable	unreasonable	ADJ
dad-10683	514	26	given	give	VERB
dad-10683	514	27	the	the	DET
dad-10683	514	28	context	context	NOUN
dad-10683	514	29	,	,	PUNCT
dad-10683	514	30	which	which	PRON
dad-10683	514	31	further	far	ADV
dad-10683	514	32	indicates	indicate	VERB
dad-10683	514	33	that	that	SCONJ
dad-10683	514	34	there	there	PRON
dad-10683	514	35	is	be	VERB
dad-10683	514	36	room	room	NOUN
dad-10683	514	37	for	for	ADP
dad-10683	514	38	improvement	improvement	NOUN
dad-10683	514	39	in	in	ADP
dad-10683	514	40	these	these	DET
dad-10683	514	41	models	model	NOUN
dad-10683	514	42	.	.	PUNCT
dad-10683	515	1	it	it	PRON
dad-10683	515	2	is	be	AUX
dad-10683	515	3	also	also	ADV
dad-10683	515	4	interesting	interesting	ADJ
dad-10683	515	5	to	to	PART
dad-10683	515	6	examine	examine	VERB
dad-10683	515	7	the	the	DET
dad-10683	515	8	distribution	distribution	NOUN
dad-10683	515	9	of	of	ADP
dad-10683	515	10	errors	error	NOUN
dad-10683	515	11	across	across	ADP
dad-10683	515	12	the	the	DET
dad-10683	515	13	examples	example	NOUN
dad-10683	515	14	.	.	PUNCT
dad-10683	516	1	as	as	SCONJ
dad-10683	516	2	can	can	AUX
dad-10683	516	3	perhaps	perhaps	ADV
dad-10683	516	4	be	be	AUX
dad-10683	516	5	expected	expect	VERB
dad-10683	516	6	,	,	PUNCT
dad-10683	516	7	the	the	DET
dad-10683	516	8	most	most	ADV
dad-10683	516	9	common	common	ADJ
dad-10683	516	10	form	form	NOUN
dad-10683	516	11	of	of	ADP
dad-10683	516	12	error	error	NOUN
dad-10683	516	13	was	be	AUX
dad-10683	516	14	a	a	DET
dad-10683	516	15	lack	lack	NOUN
dad-10683	516	16	of	of	ADP
dad-10683	516	17	understanding	understanding	NOUN
dad-10683	516	18	of	of	ADP
dad-10683	516	19	the	the	DET
dad-10683	516	20	semantics	semantic	NOUN
dad-10683	516	21	of	of	ADP
dad-10683	516	22	the	the	DET
dad-10683	516	23	responses	response	NOUN
dad-10683	516	24	.	.	PUNCT
dad-10683	517	1	what	what	PRON
dad-10683	517	2	is	be	AUX
dad-10683	517	3	more	more	ADV
dad-10683	517	4	surprising	surprising	ADJ
dad-10683	517	5	is	be	AUX
dad-10683	517	6	that	that	SCONJ
dad-10683	517	7	there	there	PRON
dad-10683	517	8	is	be	VERB
dad-10683	517	9	a	a	DET
dad-10683	517	10	significant	significant	ADJ
dad-10683	517	11	number	number	NOUN
dad-10683	517	12	of	of	ADP
dad-10683	517	13	examples	example	NOUN
dad-10683	517	14	where	where	SCONJ
dad-10683	517	15	the	the	DET
dad-10683	517	16	model	model	NOUN
dad-10683	517	17	failed	fail	VERB
dad-10683	517	18	to	to	PART
dad-10683	517	19	observe	observe	VERB
dad-10683	517	20	that	that	SCONJ
dad-10683	517	21	there	there	PRON
dad-10683	517	22	was	be	VERB
dad-10683	517	23	a	a	DET
dad-10683	517	24	key	key	ADJ
dad-10683	517	25	word	word	NOUN
dad-10683	517	26	shared	share	VERB
dad-10683	517	27	between	between	ADP
dad-10683	517	28	the	the	DET
dad-10683	517	29	context	context	NOUN
dad-10683	517	30	and	and	CCONJ
dad-10683	517	31	the	the	DET
dad-10683	517	32	correct	correct	ADJ
dad-10683	517	33	response	response	NOUN
dad-10683	517	34	;	;	PUNCT
dad-10683	517	35	this	this	PRON
dad-10683	517	36	could	could	AUX
dad-10683	517	37	be	be	AUX
dad-10683	517	38	because	because	SCONJ
dad-10683	517	39	there	there	PRON
dad-10683	517	40	are	be	VERB
dad-10683	517	41	often	often	ADV
dad-10683	517	42	common	common	ADJ
dad-10683	517	43	words	word	NOUN
dad-10683	517	44	between	between	ADP
dad-10683	517	45	the	the	DET
dad-10683	517	46	context	context	NOUN
dad-10683	517	47	and	and	CCONJ
dad-10683	517	48	false	false	ADJ
dad-10683	517	49	responses	response	NOUN
dad-10683	517	50	in	in	ADP
dad-10683	517	51	the	the	DET
dad-10683	517	52	training	training	NOUN
dad-10683	517	53	set	set	NOUN
dad-10683	517	54	,	,	PUNCT
dad-10683	517	55	and	and	CCONJ
dad-10683	517	56	the	the	DET
dad-10683	517	57	model	model	NOUN
dad-10683	517	58	is	be	AUX
dad-10683	517	59	unable	unable	ADJ
dad-10683	517	60	to	to	PART
dad-10683	517	61	distinguish	distinguish	VERB
dad-10683	517	62	between	between	ADP
dad-10683	517	63	words	word	NOUN
dad-10683	517	64	that	that	PRON
dad-10683	517	65	are	be	AUX
dad-10683	517	66	relatively	relatively	ADV
dad-10683	517	67	unimportant	unimportant	ADJ
dad-10683	517	68	and	and	CCONJ
dad-10683	517	69	those	those	PRON
dad-10683	517	70	that	that	PRON
dad-10683	517	71	carry	carry	VERB
dad-10683	517	72	significant	significant	ADJ
dad-10683	517	73	semantic	semantic	ADJ
dad-10683	517	74	meaning	meaning	NOUN
dad-10683	517	75	.	.	PUNCT
dad-10683	518	1	thus	thus	ADV
dad-10683	518	2	,	,	PUNCT
dad-10683	518	3	there	there	PRON
dad-10683	518	4	is	be	VERB
dad-10683	518	5	much	much	ADJ
dad-10683	518	6	progress	progress	NOUN
dad-10683	518	7	to	to	PART
dad-10683	518	8	be	be	AUX
dad-10683	518	9	made	make	VERB
dad-10683	518	10	in	in	ADP
dad-10683	518	11	dialogue	dialogue	NOUN
dad-10683	518	12	systems	system	NOUN
dad-10683	518	13	by	by	ADP
dad-10683	518	14	working	work	VERB
dad-10683	518	15	on	on	ADP
dad-10683	518	16	the	the	DET
dad-10683	518	17	general	general	ADJ
dad-10683	518	18	problem	problem	NOUN
dad-10683	518	19	of	of	ADP
dad-10683	518	20	natural	natural	ADJ
dad-10683	518	21	language	language	NOUN
dad-10683	518	22	understanding	understanding	NOUN
dad-10683	518	23	.	.	PUNCT
dad-10683	519	1	there	there	PRON
dad-10683	519	2	are	be	VERB
dad-10683	519	3	many	many	ADJ
dad-10683	519	4	examples	example	NOUN
dad-10683	519	5	where	where	SCONJ
dad-10683	519	6	the	the	DET
dad-10683	519	7	model	model	NOUN
dad-10683	519	8	could	could	AUX
dad-10683	519	9	be	be	AUX
dad-10683	519	10	improved	improve	VERB
dad-10683	519	11	by	by	ADP
dad-10683	519	12	explicitly	explicitly	ADV
dad-10683	519	13	accounting	account	VERB
dad-10683	519	14	for	for	ADP
dad-10683	519	15	the	the	DET
dad-10683	519	16	turn	turn	NOUN
dad-10683	519	17	-	-	PUNCT
dad-10683	519	18	taking	take	VERB
dad-10683	519	19	structure	structure	NOUN
dad-10683	519	20	of	of	ADP
dad-10683	519	21	dialogue	dialogue	NOUN
dad-10683	519	22	,	,	PUNCT
dad-10683	519	23	as	as	SCONJ
dad-10683	519	24	there	there	PRON
dad-10683	519	25	were	be	VERB
dad-10683	519	26	often	often	ADV
dad-10683	519	27	instances	instance	NOUN
dad-10683	519	28	where	where	SCONJ
dad-10683	519	29	the	the	DET
dad-10683	519	30	model	model	NOUN
dad-10683	519	31	selected	select	VERB
dad-10683	519	32	a	a	DET
dad-10683	519	33	response	response	NOUN
dad-10683	519	34	that	that	PRON
dad-10683	519	35	was	be	AUX
dad-10683	519	36	not	not	PART
dad-10683	519	37	suited	suit	VERB
dad-10683	519	38	to	to	ADP
dad-10683	519	39	the	the	DET
dad-10683	519	40	current	current	ADJ
dad-10683	519	41	speaker	speaker	NOUN
dad-10683	519	42	.	.	PUNCT
dad-10683	520	1	in	in	ADP
dad-10683	520	2	several	several	ADJ
dad-10683	520	3	cases	case	NOUN
dad-10683	520	4	,	,	PUNCT
dad-10683	520	5	the	the	DET
dad-10683	520	6	model	model	NOUN
dad-10683	520	7	also	also	ADV
dad-10683	520	8	needed	need	VERB
dad-10683	520	9	some	some	DET
dad-10683	520	10	form	form	NOUN
dad-10683	520	11	of	of	ADP
dad-10683	520	12	external	external	ADJ
dad-10683	520	13	knowledge	knowledge	NOUN
dad-10683	520	14	base	base	NOUN
dad-10683	520	15	in	in	ADP
dad-10683	520	16	order	order	NOUN
dad-10683	520	17	to	to	PART
dad-10683	520	18	answer	answer	VERB
dad-10683	520	19	the	the	DET
dad-10683	520	20	question	question	NOUN
dad-10683	520	21	correctly	correctly	ADV
dad-10683	520	22	.	.	PUNCT
dad-10683	521	1	note	note	VERB
dad-10683	521	2	that	that	SCONJ
dad-10683	521	3	the	the	DET
dad-10683	521	4	number	number	NOUN
dad-10683	521	5	of	of	ADP
dad-10683	521	6	such	such	ADJ
dad-10683	521	7	examples	example	NOUN
dad-10683	521	8	in	in	ADP
dad-10683	521	9	table	table	NOUN
dad-10683	521	10	13	13	NUM
dad-10683	521	11	refers	refer	VERB
dad-10683	521	12	to	to	ADP
dad-10683	521	13	instances	instance	NOUN
dad-10683	521	14	where	where	SCONJ
dad-10683	521	15	the	the	DET
dad-10683	521	16	correct	correct	ADJ
dad-10683	521	17	response	response	NOUN
dad-10683	521	18	mentions	mention	VERB
dad-10683	521	19	a	a	DET
dad-10683	521	20	ubuntu	ubuntu	ADJ
dad-10683	521	21	term	term	NOUN
dad-10683	521	22	that	that	PRON
dad-10683	521	23	is	be	AUX
dad-10683	521	24	related	relate	VERB
dad-10683	521	25	but	but	CCONJ
dad-10683	521	26	not	not	PART
dad-10683	521	27	identical	identical	ADJ
dad-10683	521	28	to	to	ADP
dad-10683	521	29	the	the	DET
dad-10683	521	30	terminology	terminology	NOUN
dad-10683	521	31	in	in	ADP
dad-10683	521	32	the	the	DET
dad-10683	521	33	context	context	NOUN
dad-10683	521	34	;	;	PUNCT
dad-10683	521	35	if	if	SCONJ
dad-10683	521	36	this	this	PRON
dad-10683	521	37	were	be	AUX
dad-10683	521	38	to	to	PART
dad-10683	521	39	be	be	AUX
dad-10683	521	40	extended	extend	VERB
dad-10683	521	41	to	to	ADP
dad-10683	521	42	all	all	DET
dad-10683	521	43	questions	question	NOUN
dad-10683	521	44	where	where	SCONJ
dad-10683	521	45	technical	technical	ADJ
dad-10683	521	46	vocabulary	vocabulary	NOUN
dad-10683	521	47	is	be	AUX
dad-10683	521	48	mentioned	mention	VERB
dad-10683	521	49	,	,	PUNCT
dad-10683	521	50	the	the	DET
dad-10683	521	51	number	number	NOUN
dad-10683	521	52	would	would	AUX
dad-10683	521	53	be	be	AUX
dad-10683	521	54	significantly	significantly	ADV
dad-10683	521	55	higher	high	ADJ
dad-10683	521	56	.	.	PUNCT
dad-10683	522	1	finally	finally	ADV
dad-10683	522	2	,	,	PUNCT
dad-10683	522	3	there	there	PRON
dad-10683	522	4	is	be	VERB
dad-10683	522	5	a	a	DET
dad-10683	522	6	small	small	ADJ
dad-10683	522	7	number	number	NOUN
dad-10683	522	8	of	of	ADP
dad-10683	522	9	cases	case	NOUN
dad-10683	522	10	where	where	SCONJ
dad-10683	522	11	a	a	DET
dad-10683	522	12	better	well	ADJ
dad-10683	522	13	understanding	understanding	NOUN
dad-10683	522	14	of	of	ADP
dad-10683	522	15	the	the	DET
dad-10683	522	16	tone	tone	NOUN
dad-10683	522	17	of	of	ADP
dad-10683	522	18	the	the	DET
dad-10683	522	19	dialogue	dialogue	NOUN
dad-10683	522	20	would	would	AUX
dad-10683	522	21	help	help	VERB
dad-10683	522	22	the	the	DET
dad-10683	522	23	model	model	NOUN
dad-10683	522	24	,	,	PUNCT
dad-10683	522	25	however	however	ADV
dad-10683	522	26	this	this	PRON
dad-10683	522	27	does	do	AUX
dad-10683	522	28	not	not	PART
dad-10683	522	29	seem	seem	VERB
dad-10683	522	30	to	to	PART
dad-10683	522	31	be	be	AUX
dad-10683	522	32	the	the	DET
dad-10683	522	33	best	good	ADJ
dad-10683	522	34	direction	direction	NOUN
dad-10683	522	35	for	for	ADP
dad-10683	522	36	future	future	ADJ
dad-10683	522	37	research	research	NOUN
dad-10683	522	38	.	.	PUNCT
dad-10683	523	1	52	52	NUM
dad-10683	523	2	training	training	NOUN
dad-10683	523	3	end	end	NOUN
dad-10683	523	4	-	-	PUNCT
dad-10683	523	5	to	to	ADP
dad-10683	523	6	-	-	PUNCT
dad-10683	523	7	end	end	NOUN
dad-10683	523	8	dialogue	dialogue	NOUN
dad-10683	523	9	systems	system	NOUN
dad-10683	523	10	figure	figure	VERB
dad-10683	523	11	14	14	NUM
dad-10683	523	12	:	:	PUNCT
dad-10683	523	13	diagram	diagram	NOUN
dad-10683	523	14	of	of	ADP
dad-10683	523	15	the	the	DET
dad-10683	523	16	rnn	rnn	NOUN
dad-10683	523	17	architecture	architecture	NOUN
dad-10683	523	18	for	for	ADP
dad-10683	523	19	dialogue	dialogue	NOUN
dad-10683	523	20	modeling	modeling	NOUN
dad-10683	523	21	.	.	PUNCT
dad-10683	524	1	note	note	VERB
dad-10683	524	2	that	that	SCONJ
dad-10683	524	3	utterances	utterance	NOUN
dad-10683	524	4	in	in	ADP
dad-10683	524	5	the	the	DET
dad-10683	524	6	context	context	NOUN
dad-10683	524	7	are	be	AUX
dad-10683	524	8	concatenated	concatenate	VERB
dad-10683	524	9	together	together	ADV
dad-10683	524	10	before	before	ADP
dad-10683	524	11	being	be	AUX
dad-10683	524	12	fed	feed	VERB
dad-10683	524	13	into	into	ADP
dad-10683	524	14	the	the	DET
dad-10683	524	15	rnn	rnn	NOUN
dad-10683	524	16	.	.	PROPN
dad-10683	525	1	5	5	NUM
dad-10683	525	2	.	.	X
dad-10683	525	3	generative	generative	ADJ
dad-10683	525	4	response	response	NOUN
dad-10683	525	5	architectures	architecture	NOUN
dad-10683	525	6	in	in	ADP
dad-10683	525	7	order	order	NOUN
dad-10683	525	8	to	to	PART
dad-10683	525	9	aid	aid	VERB
dad-10683	525	10	progress	progress	NOUN
dad-10683	525	11	towards	towards	ADP
dad-10683	525	12	the	the	DET
dad-10683	525	13	goal	goal	NOUN
dad-10683	525	14	of	of	ADP
dad-10683	525	15	building	build	VERB
dad-10683	525	16	fully	fully	ADV
dad-10683	525	17	generative	generative	ADJ
dad-10683	525	18	conversational	conversational	ADJ
dad-10683	525	19	models	model	NOUN
dad-10683	525	20	,	,	PUNCT
dad-10683	525	21	we	we	PRON
dad-10683	525	22	present	present	VERB
dad-10683	525	23	baseline	baseline	NOUN
dad-10683	525	24	models	model	NOUN
dad-10683	525	25	for	for	ADP
dad-10683	525	26	generating	generate	VERB
dad-10683	525	27	responses	response	NOUN
dad-10683	525	28	conditioned	condition	VERB
dad-10683	525	29	on	on	ADP
dad-10683	525	30	the	the	DET
dad-10683	525	31	context	context	NOUN
dad-10683	525	32	of	of	ADP
dad-10683	525	33	the	the	DET
dad-10683	525	34	conversation	conversation	NOUN
dad-10683	525	35	for	for	ADP
dad-10683	525	36	the	the	DET
dad-10683	525	37	ubuntu	ubuntu	ADJ
dad-10683	525	38	dialogue	dialogue	NOUN
dad-10683	525	39	corpus	corpus	NOUN
dad-10683	525	40	.	.	PUNCT
dad-10683	526	1	it	it	PRON
dad-10683	526	2	should	should	AUX
dad-10683	526	3	be	be	AUX
dad-10683	526	4	noted	note	VERB
dad-10683	526	5	that	that	SCONJ
dad-10683	526	6	the	the	DET
dad-10683	526	7	format	format	NOUN
dad-10683	526	8	of	of	ADP
dad-10683	526	9	the	the	DET
dad-10683	526	10	dataset	dataset	NOUN
dad-10683	526	11	can	can	AUX
dad-10683	526	12	easily	easily	ADV
dad-10683	526	13	be	be	AUX
dad-10683	526	14	altered	alter	VERB
dad-10683	526	15	to	to	PART
dad-10683	526	16	support	support	VERB
dad-10683	526	17	training	training	NOUN
dad-10683	526	18	in	in	ADP
dad-10683	526	19	this	this	DET
dad-10683	526	20	manner	manner	NOUN
dad-10683	526	21	:	:	PUNCT
dad-10683	526	22	one	one	PRON
dad-10683	526	23	can	can	AUX
dad-10683	526	24	simply	simply	ADV
dad-10683	526	25	remove	remove	VERB
dad-10683	526	26	all	all	DET
dad-10683	526	27	(	(	PUNCT
dad-10683	526	28	context	context	ADJ
dad-10683	526	29	,	,	PUNCT
dad-10683	526	30	response	response	NOUN
dad-10683	526	31	,	,	PUNCT
dad-10683	526	32	flag	flag	NOUN
dad-10683	526	33	)	)	PUNCT
dad-10683	526	34	triples	triple	NOUN
dad-10683	526	35	with	with	ADP
dad-10683	526	36	flag	flag	NOUN
dad-10683	526	37	=	=	SYM
dad-10683	526	38	0	0	NUM
dad-10683	526	39	,	,	PUNCT
dad-10683	526	40	and	and	CCONJ
dad-10683	526	41	be	be	AUX
dad-10683	526	42	left	leave	VERB
dad-10683	526	43	with	with	ADP
dad-10683	526	44	only	only	ADV
dad-10683	526	45	the	the	DET
dad-10683	526	46	valid	valid	ADJ
dad-10683	526	47	(	(	PUNCT
dad-10683	526	48	context	context	ADJ
dad-10683	526	49	,	,	PUNCT
dad-10683	526	50	response	response	NOUN
dad-10683	526	51	)	)	PUNCT
dad-10683	526	52	pairs	pair	NOUN
dad-10683	526	53	.	.	PUNCT
dad-10683	527	1	5.1	5.1	NUM
dad-10683	527	2	generative	generative	ADJ
dad-10683	527	3	recurrent	recurrent	ADJ
dad-10683	527	4	neural	neural	ADJ
dad-10683	527	5	language	language	NOUN
dad-10683	527	6	model	model	NOUN
dad-10683	527	7	we	we	PRON
dad-10683	527	8	first	first	ADV
dad-10683	527	9	describe	describe	VERB
dad-10683	527	10	the	the	DET
dad-10683	527	11	standard	standard	ADJ
dad-10683	527	12	recurrent	recurrent	ADJ
dad-10683	527	13	neural	neural	ADJ
dad-10683	527	14	network	network	NOUN
dad-10683	527	15	language	language	NOUN
dad-10683	527	16	model	model	NOUN
dad-10683	527	17	(	(	PUNCT
dad-10683	527	18	rnn	rnn	PROPN
dad-10683	527	19	-	-	PUNCT
dad-10683	527	20	lm	lm	NOUN
dad-10683	527	21	)	)	PUNCT
dad-10683	527	22	,	,	PUNCT
dad-10683	527	23	as	as	SCONJ
dad-10683	527	24	shown	show	VERB
dad-10683	527	25	in	in	ADP
dad-10683	527	26	figure	figure	NOUN
dad-10683	527	27	4.6	4.6	NUM
dad-10683	527	28	,	,	PUNCT
dad-10683	527	29	a	a	DET
dad-10683	527	30	neural	neural	ADJ
dad-10683	527	31	network	network	NOUN
dad-10683	527	32	model	model	NOUN
dad-10683	527	33	which	which	PRON
dad-10683	527	34	is	be	AUX
dad-10683	527	35	used	use	VERB
dad-10683	527	36	to	to	PART
dad-10683	527	37	predict	predict	VERB
dad-10683	527	38	the	the	DET
dad-10683	527	39	next	next	ADJ
dad-10683	527	40	word	word	NOUN
dad-10683	527	41	in	in	ADP
dad-10683	527	42	a	a	DET
dad-10683	527	43	sequence	sequence	NOUN
dad-10683	527	44	of	of	ADP
dad-10683	527	45	words	word	NOUN
dad-10683	527	46	(	(	PUNCT
dad-10683	527	47	mikolov	mikolov	NOUN
dad-10683	527	48	et	et	PROPN
dad-10683	527	49	al	al	PROPN
dad-10683	527	50	.	.	PROPN
dad-10683	527	51	,	,	PUNCT
dad-10683	527	52	2010	2010	NUM
dad-10683	527	53	)	)	PUNCT
dad-10683	527	54	.	.	PUNCT
dad-10683	528	1	the	the	DET
dad-10683	528	2	model	model	NOUN
dad-10683	528	3	observes	observe	VERB
dad-10683	528	4	the	the	DET
dad-10683	528	5	dialogue	dialogue	NOUN
dad-10683	528	6	word	word	NOUN
dad-10683	528	7	-	-	PUNCT
dad-10683	528	8	by	by	ADP
dad-10683	528	9	-	-	PUNCT
dad-10683	528	10	word	word	NOUN
dad-10683	528	11	and	and	CCONJ
dad-10683	528	12	updates	update	VERB
dad-10683	528	13	its	its	PRON
dad-10683	528	14	hidden	hidden	ADJ
dad-10683	528	15	state	state	NOUN
dad-10683	528	16	ht	ht	PROPN
dad-10683	528	17	at	at	ADP
dad-10683	528	18	time	time	NOUN
dad-10683	528	19	step	step	NOUN
dad-10683	528	20	(	(	PUNCT
dad-10683	528	21	word	word	NOUN
dad-10683	528	22	index	index	NOUN
dad-10683	528	23	)	)	PUNCT
dad-10683	529	1	t.	t.	NOUN
dad-10683	529	2	given	give	VERB
dad-10683	529	3	a	a	DET
dad-10683	529	4	hidden	hidden	ADJ
dad-10683	529	5	state	state	NOUN
dad-10683	529	6	ht	ht	PROPN
dad-10683	529	7	the	the	DET
dad-10683	529	8	model	model	NOUN
dad-10683	529	9	outputs	output	VERB
dad-10683	529	10	a	a	DET
dad-10683	529	11	probability	probability	NOUN
dad-10683	529	12	distribution	distribution	NOUN
dad-10683	529	13	over	over	ADP
dad-10683	529	14	all	all	DET
dad-10683	529	15	words	word	NOUN
dad-10683	529	16	in	in	ADP
dad-10683	529	17	the	the	DET
dad-10683	529	18	vocabulary	vocabulary	NOUN
dad-10683	529	19	.	.	PUNCT
dad-10683	530	1	formally	formally	ADV
dad-10683	530	2	,	,	PUNCT
dad-10683	530	3	the	the	DET
dad-10683	530	4	model	model	NOUN
dad-10683	530	5	computes	compute	VERB
dad-10683	530	6	the	the	DET
dad-10683	530	7	hidden	hidden	ADJ
dad-10683	530	8	state	state	NOUN
dad-10683	530	9	as	as	SCONJ
dad-10683	530	10	described	describe	VERB
dad-10683	530	11	in	in	ADP
dad-10683	530	12	equation	equation	NOUN
dad-10683	530	13	(	(	PUNCT
dad-10683	530	14	2	2	NUM
dad-10683	530	15	):	):	PUNCT
dad-10683	530	16	ht	ht	PROPN
dad-10683	530	17	=	=	PUNCT
dad-10683	530	18	f(w	f(w	PROPN
dad-10683	530	19	hht−1	hht−1	PROPN
dad-10683	531	1	+	+	CCONJ
dad-10683	531	2	w	w	PROPN
dad-10683	531	3	xxt	xxt	PROPN
dad-10683	531	4	)	)	PUNCT
dad-10683	531	5	,	,	PUNCT
dad-10683	531	6	where	where	SCONJ
dad-10683	531	7	,	,	PUNCT
dad-10683	531	8	ht−1	ht−1	PROPN
dad-10683	531	9	is	be	AUX
dad-10683	531	10	the	the	DET
dad-10683	531	11	previous	previous	ADJ
dad-10683	531	12	hidden	hidden	ADJ
dad-10683	531	13	state	state	NOUN
dad-10683	531	14	,	,	PUNCT
dad-10683	531	15	xt	xt	PROPN
dad-10683	531	16	is	be	AUX
dad-10683	531	17	the	the	DET
dad-10683	531	18	current	current	ADJ
dad-10683	531	19	input	input	NOUN
dad-10683	531	20	word	word	NOUN
dad-10683	531	21	,	,	PUNCT
dad-10683	531	22	f	f	X
dad-10683	531	23	(	(	PUNCT
dad-10683	531	24	·	·	PUNCT
dad-10683	531	25	)	)	PUNCT
dad-10683	531	26	is	be	AUX
dad-10683	531	27	a	a	DET
dad-10683	531	28	non	non	ADJ
dad-10683	531	29	-	-	ADJ
dad-10683	531	30	linear	linear	ADJ
dad-10683	531	31	activation	activation	NOUN
dad-10683	531	32	function	function	NOUN
dad-10683	531	33	such	such	ADJ
dad-10683	531	34	as	as	ADP
dad-10683	531	35	tanh	tanh	NOUN
dad-10683	531	36	,	,	PUNCT
dad-10683	531	37	and	and	CCONJ
dad-10683	531	38	w	w	NOUN
dad-10683	531	39	x	x	X
dad-10683	531	40	and	and	CCONJ
dad-10683	531	41	w	w	PROPN
dad-10683	531	42	h	h	NOUN
dad-10683	531	43	are	be	AUX
dad-10683	531	44	model	model	NOUN
dad-10683	531	45	parameters	parameter	NOUN
dad-10683	531	46	.	.	PUNCT
dad-10683	532	1	we	we	PRON
dad-10683	532	2	will	will	AUX
dad-10683	532	3	take	take	VERB
dad-10683	532	4	ŷn	ŷn	NOUN
dad-10683	532	5	=	=	PUNCT
dad-10683	532	6	{	{	PUNCT
dad-10683	532	7	w1	w1	NOUN
dad-10683	532	8	,	,	PUNCT
dad-10683	532	9	...	...	PUNCT
dad-10683	532	10	,	,	PUNCT
dad-10683	532	11	wt	wt	INTJ
dad-10683	532	12	}	}	PUNCT
dad-10683	532	13	as	as	ADP
dad-10683	532	14	the	the	DET
dad-10683	532	15	target	target	NOUN
dad-10683	532	16	output	output	NOUN
dad-10683	532	17	sequence	sequence	NOUN
dad-10683	532	18	for	for	ADP
dad-10683	532	19	the	the	DET
dad-10683	532	20	nth	nth	NOUN
dad-10683	532	21	training	training	NOUN
dad-10683	532	22	example	example	NOUN
dad-10683	532	23	,	,	PUNCT
dad-10683	532	24	given	give	VERB
dad-10683	532	25	the	the	DET
dad-10683	532	26	input	input	NOUN
dad-10683	532	27	sequence	sequence	NOUN
dad-10683	532	28	x̂n	x̂n	X
dad-10683	532	29	=	=	SYM
dad-10683	532	30	{	{	PUNCT
dad-10683	532	31	x1	x1	PROPN
dad-10683	532	32	,	,	PUNCT
dad-10683	532	33	...	...	PUNCT
dad-10683	532	34	,	,	PUNCT
dad-10683	532	35	xt	xt	ADP
dad-10683	532	36	}	}	PUNCT
dad-10683	532	37	.	.	PUNCT
dad-10683	533	1	the	the	DET
dad-10683	533	2	conditional	conditional	ADJ
dad-10683	533	3	distribution	distribution	NOUN
dad-10683	533	4	over	over	ADP
dad-10683	533	5	each	each	DET
dad-10683	533	6	output	output	NOUN
dad-10683	533	7	symbol	symbol	NOUN
dad-10683	533	8	is	be	AUX
dad-10683	533	9	computed	compute	VERB
dad-10683	533	10	in	in	ADP
dad-10683	533	11	a	a	DET
dad-10683	533	12	similar	similar	ADJ
dad-10683	533	13	manner	manner	NOUN
dad-10683	533	14	,	,	PUNCT
dad-10683	533	15	and	and	CCONJ
dad-10683	533	16	depends	depend	VERB
dad-10683	533	17	on	on	ADP
dad-10683	533	18	the	the	DET
dad-10683	533	19	current	current	ADJ
dad-10683	533	20	hidden	hide	VERB
dad-10683	533	21	state	state	NOUN
dad-10683	533	22	ht	ht	PROPN
dad-10683	533	23	:	:	PUNCT
dad-10683	533	24	p(wt|wt−1	p(wt|wt−1	PROPN
dad-10683	533	25	,	,	PUNCT
dad-10683	533	26	...	...	PUNCT
dad-10683	533	27	,	,	PUNCT
dad-10683	533	28	w1	w1	NOUN
dad-10683	533	29	)	)	PUNCT
dad-10683	533	30	=	=	SYM
dad-10683	534	1	exp(w	exp(w	X
dad-10683	535	1	o	o	NOUN
dad-10683	535	2	wt	wt	PROPN
dad-10683	535	3	·	·	PUNCT
dad-10683	535	4	ht)∑	ht)∑	X
dad-10683	535	5	w	w	PROPN
dad-10683	535	6	exp([w	exp([w	NOUN
dad-10683	535	7	o]w	o]w	X
dad-10683	535	8	·	·	PUNCT
dad-10683	535	9	ht	ht	X
dad-10683	535	10	)	)	PUNCT
dad-10683	535	11	,	,	PUNCT
dad-10683	535	12	(	(	PUNCT
dad-10683	535	13	5	5	X
dad-10683	535	14	)	)	PUNCT
dad-10683	535	15	where	where	SCONJ
dad-10683	535	16	w	w	NOUN
dad-10683	535	17	o	o	PROPN
dad-10683	535	18	is	be	AUX
dad-10683	535	19	the	the	DET
dad-10683	535	20	output	output	NOUN
dad-10683	535	21	matrix	matrix	NOUN
dad-10683	535	22	,	,	PUNCT
dad-10683	535	23	and	and	CCONJ
dad-10683	535	24	w	w	ADP
dad-10683	535	25	o	o	PROPN
dad-10683	535	26	w	w	NOUN
dad-10683	535	27	indicates	indicate	VERB
dad-10683	535	28	the	the	DET
dad-10683	535	29	row	row	NOUN
dad-10683	535	30	of	of	ADP
dad-10683	535	31	the	the	DET
dad-10683	535	32	w	w	PROPN
dad-10683	535	33	o	o	NOUN
dad-10683	535	34	matrix	matrix	NOUN
dad-10683	535	35	corresponding	correspond	VERB
dad-10683	535	36	to	to	ADP
dad-10683	535	37	the	the	DET
dad-10683	535	38	output	output	NOUN
dad-10683	535	39	index	index	NOUN
dad-10683	535	40	for	for	ADP
dad-10683	535	41	word	word	NOUN
dad-10683	535	42	w.	w.	PROPN
dad-10683	535	43	the	the	DET
dad-10683	535	44	model	model	NOUN
dad-10683	535	45	is	be	AUX
dad-10683	535	46	trained	train	VERB
dad-10683	535	47	with	with	ADP
dad-10683	535	48	teacher	teacher	NOUN
dad-10683	535	49	forcing	force	VERB
dad-10683	535	50	,	,	PUNCT
dad-10683	535	51	meaning	mean	VERB
dad-10683	535	52	the	the	DET
dad-10683	535	53	input	input	NOUN
dad-10683	535	54	xt	xt	PROPN
dad-10683	535	55	to	to	ADP
dad-10683	535	56	the	the	DET
dad-10683	535	57	network	network	NOUN
dad-10683	535	58	is	be	AUX
dad-10683	535	59	the	the	DET
dad-10683	535	60	previous	previous	ADJ
dad-10683	535	61	ground	ground	NOUN
dad-10683	535	62	-	-	PUNCT
dad-10683	535	63	truth	truth	NOUN
dad-10683	535	64	output	output	NOUN
dad-10683	535	65	word	word	NOUN
dad-10683	535	66	wt−1	wt−1	PROPN
dad-10683	535	67	.	.	PUNCT
dad-10683	536	1	the	the	DET
dad-10683	536	2	model	model	NOUN
dad-10683	536	3	is	be	AUX
dad-10683	536	4	trained	train	VERB
dad-10683	536	5	in	in	ADP
dad-10683	536	6	an	an	DET
dad-10683	536	7	end	end	NOUN
dad-10683	536	8	-	-	PUNCT
dad-10683	536	9	to	to	ADP
dad-10683	536	10	-	-	PUNCT
dad-10683	536	11	end	end	NOUN
dad-10683	536	12	fashion	fashion	NOUN
dad-10683	536	13	by	by	ADP
dad-10683	536	14	gradient	gradient	ADJ
dad-10683	536	15	descent	descent	NOUN
dad-10683	536	16	to	to	PART
dad-10683	536	17	maximize	maximize	VERB
dad-10683	536	18	the	the	DET
dad-10683	536	19	conditional	conditional	ADJ
dad-10683	536	20	log	log	NOUN
dad-10683	536	21	-	-	PUNCT
dad-10683	536	22	likelihood	likelihood	NOUN
dad-10683	536	23	of	of	ADP
dad-10683	536	24	input	input	NOUN
dad-10683	536	25	-	-	PUNCT
dad-10683	536	26	output	output	NOUN
dad-10683	536	27	pairs	pair	NOUN
dad-10683	536	28	from	from	ADP
dad-10683	536	29	the	the	DET
dad-10683	536	30	training	training	NOUN
dad-10683	536	31	set	set	NOUN
dad-10683	536	32	,	,	PUNCT
dad-10683	536	33	{	{	PUNCT
dad-10683	536	34	x̂n	x̂n	X
dad-10683	536	35	,	,	PUNCT
dad-10683	536	36	ŷn	ŷn	PROPN
dad-10683	536	37	}	}	PUNCT
dad-10683	536	38	:	:	PUNCT
dad-10683	536	39	max	max	PROPN
dad-10683	536	40	θ	θ	PROPN
dad-10683	536	41	1	1	NUM
dad-10683	536	42	n	n	NOUN
dad-10683	536	43	n∑	n∑	NOUN
dad-10683	536	44	n=1	n=1	PROPN
dad-10683	536	45	log	log	VERB
dad-10683	536	46	pθ(ŷn|x̂n	pθ(ŷn|x̂n	PROPN
dad-10683	536	47	)	)	PUNCT
dad-10683	536	48	,	,	PUNCT
dad-10683	536	49	(	(	PUNCT
dad-10683	536	50	6	6	X
dad-10683	536	51	)	)	PUNCT
dad-10683	536	52	53	53	NUM
dad-10683	536	53	lowe	lowe	NOUN
dad-10683	536	54	,	,	PUNCT
dad-10683	536	55	pow	pow	PROPN
dad-10683	536	56	,	,	PUNCT
dad-10683	536	57	serban	serban	PROPN
dad-10683	536	58	,	,	PUNCT
dad-10683	536	59	charlinn	charlinn	PROPN
dad-10683	536	60	,	,	PUNCT
dad-10683	536	61	liu	liu	PROPN
dad-10683	536	62	and	and	CCONJ
dad-10683	536	63	pineau	pineau	PROPN
dad-10683	536	64	figure	figure	NOUN
dad-10683	536	65	15	15	NUM
dad-10683	536	66	:	:	PUNCT
dad-10683	536	67	diagram	diagram	NOUN
dad-10683	536	68	of	of	ADP
dad-10683	536	69	the	the	DET
dad-10683	536	70	hred	hre	VERB
dad-10683	536	71	model	model	NOUN
dad-10683	536	72	.	.	PUNCT
dad-10683	537	1	note	note	VERB
dad-10683	537	2	that	that	SCONJ
dad-10683	537	3	each	each	DET
dad-10683	537	4	utterance	utterance	NOUN
dad-10683	537	5	in	in	ADP
dad-10683	537	6	the	the	DET
dad-10683	537	7	context	context	NOUN
dad-10683	537	8	is	be	AUX
dad-10683	537	9	encoded	encode	VERB
dad-10683	537	10	with	with	ADP
dad-10683	537	11	a	a	DET
dad-10683	537	12	separate	separate	ADJ
dad-10683	537	13	‘	'	PUNCT
dad-10683	537	14	utterance	utterance	ADJ
dad-10683	537	15	-	-	PUNCT
dad-10683	537	16	level	level	NOUN
dad-10683	537	17	’	'	PUNCT
dad-10683	537	18	encoder	encoder	NOUN
dad-10683	537	19	,	,	PUNCT
dad-10683	537	20	which	which	PRON
dad-10683	537	21	is	be	AUX
dad-10683	537	22	then	then	ADV
dad-10683	537	23	fed	feed	VERB
dad-10683	537	24	into	into	ADP
dad-10683	537	25	a	a	DET
dad-10683	537	26	‘	'	PUNCT
dad-10683	537	27	context	context	NOUN
dad-10683	537	28	-	-	PUNCT
dad-10683	537	29	level	level	NOUN
dad-10683	537	30	’	'	PUNCT
dad-10683	537	31	encoder	encoder	NOUN
dad-10683	537	32	.	.	PUNCT
dad-10683	538	1	where	where	SCONJ
dad-10683	538	2	θ	θ	PROPN
dad-10683	538	3	are	be	AUX
dad-10683	538	4	the	the	DET
dad-10683	538	5	parameters	parameter	NOUN
dad-10683	538	6	of	of	ADP
dad-10683	538	7	the	the	DET
dad-10683	538	8	model	model	NOUN
dad-10683	538	9	,	,	PUNCT
dad-10683	538	10	including	include	VERB
dad-10683	538	11	w	w	PROPN
dad-10683	538	12	x	x	PROPN
dad-10683	538	13	,	,	PUNCT
dad-10683	538	14	w	w	PROPN
dad-10683	538	15	h	h	PROPN
dad-10683	538	16	,	,	PUNCT
dad-10683	538	17	w	w	PROPN
dad-10683	538	18	o	o	NOUN
dad-10683	538	19	,	,	PUNCT
dad-10683	538	20	and	and	CCONJ
dad-10683	538	21	the	the	DET
dad-10683	538	22	corresponding	correspond	VERB
dad-10683	538	23	biases	bias	NOUN
dad-10683	538	24	.	.	PUNCT
dad-10683	539	1	thus	thus	ADV
dad-10683	539	2	,	,	PUNCT
dad-10683	539	3	the	the	DET
dad-10683	539	4	model	model	NOUN
dad-10683	539	5	learns	learn	VERB
dad-10683	539	6	a	a	DET
dad-10683	539	7	probability	probability	NOUN
dad-10683	539	8	distribution	distribution	NOUN
dad-10683	539	9	over	over	ADP
dad-10683	539	10	all	all	DET
dad-10683	539	11	output	output	NOUN
dad-10683	539	12	sequences	sequence	NOUN
dad-10683	539	13	,	,	PUNCT
dad-10683	539	14	p(ŷ1	p(ŷ1	PROPN
dad-10683	539	15	,	,	PUNCT
dad-10683	539	16	...	...	PUNCT
dad-10683	539	17	,	,	PUNCT
dad-10683	539	18	ŷt	ŷt	VERB
dad-10683	539	19	)	)	PUNCT
dad-10683	539	20	.	.	PUNCT
dad-10683	540	1	for	for	ADP
dad-10683	540	2	dialogue	dialogue	NOUN
dad-10683	540	3	response	response	NOUN
dad-10683	540	4	generation	generation	NOUN
dad-10683	540	5	,	,	PUNCT
dad-10683	540	6	the	the	DET
dad-10683	540	7	model	model	NOUN
dad-10683	540	8	is	be	AUX
dad-10683	540	9	conditioned	condition	VERB
dad-10683	540	10	on	on	ADP
dad-10683	540	11	the	the	DET
dad-10683	540	12	previous	previous	ADJ
dad-10683	540	13	dialogue	dialogue	NOUN
dad-10683	540	14	context	context	NOUN
dad-10683	540	15	and	and	CCONJ
dad-10683	540	16	used	use	VERB
dad-10683	540	17	to	to	PART
dad-10683	540	18	generate	generate	VERB
dad-10683	540	19	a	a	DET
dad-10683	540	20	response	response	NOUN
dad-10683	540	21	,	,	PUNCT
dad-10683	540	22	i.e.	i.e.	X
dad-10683	540	23	the	the	DET
dad-10683	540	24	next	next	ADJ
dad-10683	540	25	utterance	utterance	NOUN
dad-10683	540	26	in	in	ADP
dad-10683	540	27	the	the	DET
dad-10683	540	28	dialogue	dialogue	NOUN
dad-10683	540	29	.	.	PUNCT
dad-10683	541	1	such	such	DET
dad-10683	541	2	a	a	DET
dad-10683	541	3	model	model	NOUN
dad-10683	541	4	could	could	AUX
dad-10683	541	5	be	be	AUX
dad-10683	541	6	used	use	VERB
dad-10683	541	7	as	as	ADP
dad-10683	541	8	a	a	DET
dad-10683	541	9	full	full	ADJ
dad-10683	541	10	dialogue	dialogue	NOUN
dad-10683	541	11	system	system	NOUN
dad-10683	541	12	,	,	PUNCT
dad-10683	541	13	as	as	SCONJ
dad-10683	541	14	defined	define	VERB
dad-10683	541	15	in	in	ADP
dad-10683	541	16	section	section	NOUN
dad-10683	541	17	1.1	1.1	NUM
dad-10683	541	18	,	,	PUNCT
dad-10683	541	19	to	to	PART
dad-10683	541	20	carry	carry	VERB
dad-10683	541	21	out	out	ADP
dad-10683	541	22	a	a	DET
dad-10683	541	23	conversation	conversation	NOUN
dad-10683	541	24	with	with	ADP
dad-10683	541	25	a	a	DET
dad-10683	541	26	user	user	NOUN
dad-10683	541	27	.	.	PUNCT
dad-10683	542	1	5.2	5.2	NUM
dad-10683	542	2	hierarchical	hierarchical	ADJ
dad-10683	542	3	recurrent	recurrent	ADJ
dad-10683	542	4	encoder	encoder	NOUN
dad-10683	542	5	-	-	PUNCT
dad-10683	542	6	decoder	decoder	NOUN
dad-10683	542	7	one	one	NUM
dad-10683	542	8	problem	problem	NOUN
dad-10683	542	9	with	with	ADP
dad-10683	542	10	directly	directly	ADV
dad-10683	542	11	applying	apply	VERB
dad-10683	542	12	a	a	DET
dad-10683	542	13	standard	standard	ADJ
dad-10683	542	14	rnn	rnn	NOUN
dad-10683	542	15	language	language	NOUN
dad-10683	542	16	model	model	NOUN
dad-10683	542	17	to	to	ADP
dad-10683	542	18	modeling	model	VERB
dad-10683	542	19	dialogues	dialogue	NOUN
dad-10683	542	20	is	be	AUX
dad-10683	542	21	that	that	SCONJ
dad-10683	542	22	it	it	PRON
dad-10683	542	23	does	do	AUX
dad-10683	542	24	not	not	PART
dad-10683	542	25	take	take	VERB
dad-10683	542	26	into	into	ADP
dad-10683	542	27	account	account	NOUN
dad-10683	542	28	the	the	DET
dad-10683	542	29	turn	turn	NOUN
dad-10683	542	30	-	-	PUNCT
dad-10683	542	31	taking	take	VERB
dad-10683	542	32	nature	nature	NOUN
dad-10683	542	33	of	of	ADP
dad-10683	542	34	conversations	conversation	NOUN
dad-10683	542	35	.	.	PUNCT
dad-10683	543	1	it	it	PRON
dad-10683	543	2	is	be	AUX
dad-10683	543	3	well	well	ADV
dad-10683	543	4	known	know	VERB
dad-10683	543	5	that	that	SCONJ
dad-10683	543	6	recurrent	recurrent	ADJ
dad-10683	543	7	neural	neural	ADJ
dad-10683	543	8	networks	network	NOUN
dad-10683	543	9	have	have	VERB
dad-10683	543	10	trouble	trouble	NOUN
dad-10683	543	11	learning	learn	VERB
dad-10683	543	12	long	long	ADJ
dad-10683	543	13	-	-	PUNCT
dad-10683	543	14	term	term	NOUN
dad-10683	543	15	dependencies	dependency	NOUN
dad-10683	543	16	(	(	PUNCT
dad-10683	543	17	bengio	bengio	NOUN
dad-10683	543	18	et	et	PROPN
dad-10683	543	19	al	al	PROPN
dad-10683	543	20	.	.	PROPN
dad-10683	543	21	,	,	PUNCT
dad-10683	543	22	1994	1994	NUM
dad-10683	543	23	)	)	PUNCT
dad-10683	543	24	,	,	PUNCT
dad-10683	543	25	a	a	DET
dad-10683	543	26	problem	problem	NOUN
dad-10683	543	27	only	only	ADV
dad-10683	543	28	partially	partially	ADV
dad-10683	543	29	alleviated	alleviate	VERB
dad-10683	543	30	with	with	ADP
dad-10683	543	31	lstms	lstms	NOUN
dad-10683	543	32	.	.	PUNCT
dad-10683	544	1	thus	thus	ADV
dad-10683	544	2	,	,	PUNCT
dad-10683	544	3	if	if	SCONJ
dad-10683	544	4	a	a	DET
dad-10683	544	5	long	long	ADJ
dad-10683	544	6	context	context	NOUN
dad-10683	544	7	is	be	AUX
dad-10683	544	8	fed	feed	VERB
dad-10683	544	9	into	into	ADP
dad-10683	544	10	the	the	DET
dad-10683	544	11	encoder	encoder	NOUN
dad-10683	544	12	,	,	PUNCT
dad-10683	544	13	it	it	PRON
dad-10683	544	14	is	be	AUX
dad-10683	544	15	possible	possible	ADJ
dad-10683	544	16	for	for	SCONJ
dad-10683	544	17	the	the	DET
dad-10683	544	18	model	model	NOUN
dad-10683	544	19	to	to	PART
dad-10683	544	20	put	put	VERB
dad-10683	544	21	a	a	DET
dad-10683	544	22	large	large	ADJ
dad-10683	544	23	weight	weight	NOUN
dad-10683	544	24	on	on	ADP
dad-10683	544	25	only	only	ADV
dad-10683	544	26	the	the	DET
dad-10683	544	27	most	most	ADV
dad-10683	544	28	recent	recent	ADJ
dad-10683	544	29	utterance	utterance	NOUN
dad-10683	544	30	.	.	PUNCT
dad-10683	545	1	in	in	ADP
dad-10683	545	2	order	order	NOUN
dad-10683	545	3	to	to	PART
dad-10683	545	4	investigate	investigate	VERB
dad-10683	545	5	models	model	NOUN
dad-10683	545	6	that	that	PRON
dad-10683	545	7	are	be	AUX
dad-10683	545	8	able	able	ADJ
dad-10683	545	9	to	to	PART
dad-10683	545	10	retain	retain	VERB
dad-10683	545	11	state	state	NOUN
dad-10683	545	12	over	over	ADP
dad-10683	545	13	long	long	ADJ
dad-10683	545	14	conversations	conversation	NOUN
dad-10683	545	15	,	,	PUNCT
dad-10683	545	16	we	we	PRON
dad-10683	545	17	implement	implement	VERB
dad-10683	545	18	the	the	DET
dad-10683	545	19	recently	recently	ADV
dad-10683	545	20	proposed	propose	VERB
dad-10683	545	21	hierarchical	hierarchical	ADJ
dad-10683	545	22	recurrent	recurrent	ADJ
dad-10683	545	23	encoder	encoder	NOUN
dad-10683	545	24	-	-	PUNCT
dad-10683	545	25	decoder	decoder	NOUN
dad-10683	545	26	(	(	PUNCT
dad-10683	545	27	hred	hre	VERB
dad-10683	545	28	)	)	PUNCT
dad-10683	545	29	sordoni	sordoni	INTJ
dad-10683	545	30	et	et	PROPN
dad-10683	545	31	al	al	PROPN
dad-10683	545	32	.	.	PROPN
dad-10683	546	1	(	(	PUNCT
dad-10683	546	2	2015a	2015a	NUM
dad-10683	546	3	)	)	PUNCT
dad-10683	546	4	.	.	PUNCT
dad-10683	547	1	while	while	SCONJ
dad-10683	547	2	this	this	DET
dad-10683	547	3	model	model	NOUN
dad-10683	547	4	was	be	AUX
dad-10683	547	5	initially	initially	ADV
dad-10683	547	6	proposed	propose	VERB
dad-10683	547	7	for	for	ADP
dad-10683	547	8	context	context	NOUN
dad-10683	547	9	-	-	PUNCT
dad-10683	547	10	sensitive	sensitive	ADJ
dad-10683	547	11	query	query	NOUN
dad-10683	547	12	suggestion	suggestion	NOUN
dad-10683	547	13	,	,	PUNCT
dad-10683	547	14	it	it	PRON
dad-10683	547	15	has	have	AUX
dad-10683	547	16	been	be	AUX
dad-10683	547	17	adapted	adapt	VERB
dad-10683	547	18	for	for	ADP
dad-10683	547	19	dialogue	dialogue	NOUN
dad-10683	547	20	response	response	NOUN
dad-10683	547	21	generation	generation	NOUN
dad-10683	547	22	on	on	ADP
dad-10683	547	23	a	a	DET
dad-10683	547	24	dataset	dataset	NOUN
dad-10683	547	25	of	of	ADP
dad-10683	547	26	movie	movie	NOUN
dad-10683	547	27	subtitles	subtitle	NOUN
dad-10683	547	28	(	(	PUNCT
dad-10683	547	29	serban	serban	PROPN
dad-10683	547	30	et	et	PROPN
dad-10683	547	31	al	al	PROPN
dad-10683	547	32	.	.	PROPN
dad-10683	547	33	,	,	PUNCT
dad-10683	547	34	2016	2016	NUM
dad-10683	547	35	)	)	PUNCT
dad-10683	547	36	.	.	PUNCT
dad-10683	548	1	the	the	DET
dad-10683	548	2	hred	hre	VERB
dad-10683	548	3	model	model	NOUN
dad-10683	548	4	builds	build	VERB
dad-10683	548	5	on	on	ADP
dad-10683	548	6	the	the	DET
dad-10683	548	7	traditional	traditional	ADJ
dad-10683	548	8	encoder	encoder	NOUN
dad-10683	548	9	-	-	PUNCT
dad-10683	548	10	decoder	decoder	NOUN
dad-10683	548	11	model	model	NOUN
dad-10683	548	12	(	(	PUNCT
dad-10683	548	13	cho	cho	PROPN
dad-10683	548	14	et	et	PROPN
dad-10683	548	15	al	al	PROPN
dad-10683	548	16	.	.	PROPN
dad-10683	548	17	,	,	PUNCT
dad-10683	548	18	2014	2014	NUM
dad-10683	548	19	)	)	PUNCT
dad-10683	548	20	.	.	PUNCT
dad-10683	549	1	the	the	DET
dad-10683	549	2	main	main	ADJ
dad-10683	549	3	addition	addition	NOUN
dad-10683	549	4	is	be	AUX
dad-10683	549	5	a	a	DET
dad-10683	549	6	second	second	ADJ
dad-10683	549	7	encoder	encoder	NOUN
dad-10683	549	8	,	,	PUNCT
dad-10683	549	9	the	the	DET
dad-10683	549	10	utterance	utterance	ADJ
dad-10683	549	11	-	-	PUNCT
dad-10683	549	12	level	level	NOUN
dad-10683	549	13	encoder	encoder	NOUN
dad-10683	549	14	,	,	PUNCT
dad-10683	549	15	that	that	PRON
dad-10683	549	16	takes	take	VERB
dad-10683	549	17	as	as	ADP
dad-10683	549	18	input	input	NOUN
dad-10683	549	19	the	the	DET
dad-10683	549	20	fixed	fix	VERB
dad-10683	549	21	-	-	PUNCT
dad-10683	549	22	length	length	NOUN
dad-10683	549	23	vectors	vector	NOUN
dad-10683	549	24	produced	produce	VERB
dad-10683	549	25	by	by	ADP
dad-10683	549	26	the	the	DET
dad-10683	549	27	lower	low	ADJ
dad-10683	549	28	-	-	PUNCT
dad-10683	549	29	level	level	NOUN
dad-10683	549	30	encoder	encoder	NOUN
dad-10683	549	31	,	,	PUNCT
dad-10683	549	32	which	which	PRON
dad-10683	549	33	we	we	PRON
dad-10683	549	34	refer	refer	VERB
dad-10683	549	35	to	to	ADP
dad-10683	549	36	as	as	ADP
dad-10683	549	37	the	the	DET
dad-10683	549	38	word	word	NOUN
dad-10683	549	39	-	-	PUNCT
dad-10683	549	40	level	level	NOUN
dad-10683	549	41	encoder	encoder	NOUN
dad-10683	549	42	.	.	PUNCT
dad-10683	550	1	instead	instead	ADV
dad-10683	550	2	of	of	ADP
dad-10683	550	3	letting	let	VERB
dad-10683	550	4	the	the	DET
dad-10683	550	5	decoder	decoder	NOUN
dad-10683	550	6	take	take	VERB
dad-10683	550	7	as	as	SCONJ
dad-10683	550	8	input	input	NOUN
dad-10683	550	9	the	the	DET
dad-10683	550	10	fixed	fix	VERB
dad-10683	550	11	-	-	PUNCT
dad-10683	550	12	length	length	NOUN
dad-10683	550	13	vectors	vector	NOUN
dad-10683	550	14	from	from	ADP
dad-10683	550	15	the	the	DET
dad-10683	550	16	word	word	NOUN
dad-10683	550	17	-	-	PUNCT
dad-10683	550	18	level	level	NOUN
dad-10683	550	19	encoder	encoder	NOUN
dad-10683	550	20	,	,	PUNCT
dad-10683	550	21	the	the	DET
dad-10683	550	22	decoder	decoder	NOUN
dad-10683	550	23	takes	take	VERB
dad-10683	550	24	as	as	SCONJ
dad-10683	550	25	input	input	NOUN
dad-10683	550	26	the	the	DET
dad-10683	550	27	output	output	NOUN
dad-10683	550	28	of	of	ADP
dad-10683	550	29	the	the	DET
dad-10683	550	30	utterance	utterance	ADJ
dad-10683	550	31	-	-	PUNCT
dad-10683	550	32	level	level	NOUN
dad-10683	550	33	encoder	encoder	NOUN
dad-10683	550	34	.	.	PUNCT
dad-10683	551	1	intuitively	intuitively	ADV
dad-10683	551	2	,	,	PUNCT
dad-10683	551	3	the	the	DET
dad-10683	551	4	utterance	utterance	ADJ
dad-10683	551	5	-	-	PUNCT
dad-10683	551	6	level	level	NOUN
dad-10683	551	7	encoder	encoder	NOUN
dad-10683	551	8	summarizes	summarize	VERB
dad-10683	551	9	the	the	DET
dad-10683	551	10	history	history	NOUN
dad-10683	551	11	of	of	ADP
dad-10683	551	12	the	the	DET
dad-10683	551	13	conversations	conversation	NOUN
dad-10683	551	14	into	into	ADP
dad-10683	551	15	a	a	DET
dad-10683	551	16	single	single	ADJ
dad-10683	551	17	vector	vector	NOUN
dad-10683	551	18	,	,	PUNCT
dad-10683	551	19	which	which	PRON
dad-10683	551	20	is	be	AUX
dad-10683	551	21	more	more	ADV
dad-10683	551	22	sensitive	sensitive	ADJ
dad-10683	551	23	to	to	ADP
dad-10683	551	24	previous	previous	ADJ
dad-10683	551	25	utterances	utterance	NOUN
dad-10683	551	26	in	in	ADP
dad-10683	551	27	the	the	DET
dad-10683	551	28	conversation	conversation	NOUN
dad-10683	551	29	.	.	PUNCT
dad-10683	552	1	this	this	PRON
dad-10683	552	2	provides	provide	VERB
dad-10683	552	3	a	a	DET
dad-10683	552	4	more	more	ADV
dad-10683	552	5	powerful	powerful	ADJ
dad-10683	552	6	architecture	architecture	NOUN
dad-10683	552	7	as	as	SCONJ
dad-10683	552	8	it	it	PRON
dad-10683	552	9	is	be	AUX
dad-10683	552	10	now	now	ADV
dad-10683	552	11	possible	possible	ADJ
dad-10683	552	12	for	for	SCONJ
dad-10683	552	13	the	the	DET
dad-10683	552	14	model	model	NOUN
dad-10683	552	15	to	to	PART
dad-10683	552	16	encode	encode	VERB
dad-10683	552	17	order	order	NOUN
dad-10683	552	18	-	-	PUNCT
dad-10683	552	19	dependent	dependent	ADJ
dad-10683	552	20	patterns	pattern	NOUN
dad-10683	552	21	inherent	inherent	ADJ
dad-10683	552	22	in	in	ADP
dad-10683	552	23	the	the	DET
dad-10683	552	24	turn	turn	NOUN
dad-10683	552	25	-	-	PUNCT
dad-10683	552	26	taking	take	VERB
dad-10683	552	27	nature	nature	NOUN
dad-10683	552	28	of	of	ADP
dad-10683	552	29	dialogue	dialogue	NOUN
dad-10683	552	30	.	.	PUNCT
dad-10683	553	1	as	as	ADP
dad-10683	553	2	before	before	ADV
dad-10683	553	3	,	,	PUNCT
dad-10683	553	4	the	the	DET
dad-10683	553	5	54	54	NUM
dad-10683	553	6	training	training	NOUN
dad-10683	553	7	end	end	NOUN
dad-10683	553	8	-	-	PUNCT
dad-10683	553	9	to	to	ADP
dad-10683	553	10	-	-	PUNCT
dad-10683	553	11	end	end	NOUN
dad-10683	553	12	dialogue	dialogue	NOUN
dad-10683	553	13	systems	system	NOUN
dad-10683	553	14	model	model	NOUN
dad-10683	553	15	is	be	AUX
dad-10683	553	16	trained	train	VERB
dad-10683	553	17	end	end	NOUN
dad-10683	553	18	-	-	PUNCT
dad-10683	553	19	to	to	ADP
dad-10683	553	20	-	-	PUNCT
dad-10683	553	21	end	end	NOUN
dad-10683	553	22	to	to	PART
dad-10683	553	23	maximize	maximize	VERB
dad-10683	553	24	the	the	DET
dad-10683	553	25	log	log	NOUN
dad-10683	553	26	-	-	PUNCT
dad-10683	553	27	likelihood	likelihood	NOUN
dad-10683	553	28	of	of	ADP
dad-10683	553	29	the	the	DET
dad-10683	553	30	generated	generate	VERB
dad-10683	553	31	utterance	utterance	NOUN
dad-10683	553	32	.	.	PUNCT
dad-10683	554	1	a	a	DET
dad-10683	554	2	diagram	diagram	NOUN
dad-10683	554	3	of	of	ADP
dad-10683	554	4	the	the	DET
dad-10683	554	5	hred	hre	VERB
dad-10683	554	6	model	model	NOUN
dad-10683	554	7	can	can	AUX
dad-10683	554	8	be	be	AUX
dad-10683	554	9	seen	see	VERB
dad-10683	554	10	in	in	ADP
dad-10683	554	11	figure	figure	NOUN
dad-10683	554	12	5.1	5.1	NUM
dad-10683	554	13	.	.	PUNCT
dad-10683	555	1	to	to	PART
dad-10683	555	2	summarize	summarize	VERB
dad-10683	555	3	:	:	PUNCT
dad-10683	555	4	the	the	DET
dad-10683	555	5	rnn	rnn	PROPN
dad-10683	555	6	-	-	PUNCT
dad-10683	555	7	lm	lm	PROPN
dad-10683	555	8	(	(	PUNCT
dad-10683	555	9	figure	figure	NOUN
dad-10683	555	10	4.6	4.6	NUM
dad-10683	555	11	)	)	PUNCT
dad-10683	555	12	uses	use	VERB
dad-10683	555	13	a	a	DET
dad-10683	555	14	single	single	ADJ
dad-10683	555	15	rnn	rnn	NOUN
dad-10683	555	16	(	(	PUNCT
dad-10683	555	17	or	or	CCONJ
dad-10683	555	18	lstm	lstm	ADJ
dad-10683	555	19	)	)	PUNCT
dad-10683	555	20	to	to	PART
dad-10683	555	21	encode	encode	VERB
dad-10683	555	22	the	the	DET
dad-10683	555	23	entire	entire	ADJ
dad-10683	555	24	history	history	NOUN
dad-10683	555	25	of	of	ADP
dad-10683	555	26	the	the	DET
dad-10683	555	27	dialogue	dialogue	NOUN
dad-10683	555	28	,	,	PUNCT
dad-10683	555	29	which	which	PRON
dad-10683	555	30	consists	consist	VERB
dad-10683	555	31	of	of	ADP
dad-10683	555	32	all	all	DET
dad-10683	555	33	utterances	utterance	NOUN
dad-10683	555	34	in	in	ADP
dad-10683	555	35	the	the	DET
dad-10683	555	36	context	context	NOUN
dad-10683	555	37	concatenated	concatenate	VERB
dad-10683	555	38	together	together	ADV
dad-10683	555	39	.	.	PUNCT
dad-10683	556	1	it	it	PRON
dad-10683	556	2	then	then	ADV
dad-10683	556	3	uses	use	VERB
dad-10683	556	4	the	the	DET
dad-10683	556	5	same	same	ADJ
dad-10683	556	6	rnn	rnn	NOUN
dad-10683	556	7	(	(	PUNCT
dad-10683	556	8	i.e.	i.e.	X
dad-10683	556	9	an	an	DET
dad-10683	556	10	rnn	rnn	NOUN
dad-10683	556	11	with	with	ADP
dad-10683	556	12	the	the	DET
dad-10683	556	13	same	same	ADJ
dad-10683	556	14	parameters	parameter	NOUN
dad-10683	556	15	)	)	PUNCT
dad-10683	556	16	to	to	PART
dad-10683	556	17	decode	decode	VERB
dad-10683	556	18	the	the	DET
dad-10683	556	19	prediction	prediction	NOUN
dad-10683	556	20	utterance	utterance	ADV
dad-10683	556	21	.	.	PUNCT
dad-10683	557	1	the	the	DET
dad-10683	557	2	encoder	encoder	NOUN
dad-10683	557	3	-	-	PUNCT
dad-10683	557	4	decoder	decoder	NOUN
dad-10683	557	5	model	model	NOUN
dad-10683	557	6	(	(	PUNCT
dad-10683	557	7	not	not	PART
dad-10683	557	8	shown	show	VERB
dad-10683	557	9	here	here	ADV
dad-10683	557	10	)	)	PUNCT
dad-10683	557	11	augments	augment	VERB
dad-10683	557	12	this	this	PRON
dad-10683	557	13	with	with	ADP
dad-10683	557	14	a	a	DET
dad-10683	557	15	second	second	ADJ
dad-10683	557	16	rnn	rnn	NOUN
dad-10683	557	17	with	with	ADP
dad-10683	557	18	different	different	ADJ
dad-10683	557	19	parameters	parameter	NOUN
dad-10683	557	20	for	for	ADP
dad-10683	557	21	the	the	DET
dad-10683	557	22	decoder	decoder	NOUN
dad-10683	557	23	.	.	PUNCT
dad-10683	558	1	all	all	DET
dad-10683	558	2	context	context	NOUN
dad-10683	558	3	utterances	utterance	NOUN
dad-10683	558	4	are	be	AUX
dad-10683	558	5	still	still	ADV
dad-10683	558	6	concatenated	concatenate	VERB
dad-10683	558	7	in	in	ADP
dad-10683	558	8	the	the	DET
dad-10683	558	9	encoder	encoder	NOUN
dad-10683	558	10	,	,	PUNCT
dad-10683	558	11	and	and	CCONJ
dad-10683	558	12	thus	thus	ADV
dad-10683	558	13	it	it	PRON
dad-10683	558	14	is	be	AUX
dad-10683	558	15	difficult	difficult	ADJ
dad-10683	558	16	to	to	PART
dad-10683	558	17	model	model	VERB
dad-10683	558	18	long	long	ADJ
dad-10683	558	19	-	-	PUNCT
dad-10683	558	20	term	term	NOUN
dad-10683	558	21	dependencies	dependency	NOUN
dad-10683	558	22	for	for	ADP
dad-10683	558	23	utterances	utterance	NOUN
dad-10683	558	24	that	that	PRON
dad-10683	558	25	occur	occur	VERB
dad-10683	558	26	earlier	early	ADV
dad-10683	558	27	in	in	ADP
dad-10683	558	28	the	the	DET
dad-10683	558	29	dialogue	dialogue	NOUN
dad-10683	558	30	.	.	PUNCT
dad-10683	559	1	this	this	DET
dad-10683	559	2	problem	problem	NOUN
dad-10683	559	3	is	be	AUX
dad-10683	559	4	alleviated	alleviate	VERB
dad-10683	559	5	with	with	ADP
dad-10683	559	6	the	the	DET
dad-10683	559	7	hred	hre	VERB
dad-10683	559	8	model	model	NOUN
dad-10683	559	9	(	(	PUNCT
dad-10683	559	10	figure	figure	NOUN
dad-10683	559	11	5.1	5.1	NUM
dad-10683	559	12	)	)	PUNCT
dad-10683	559	13	,	,	PUNCT
dad-10683	559	14	which	which	PRON
dad-10683	559	15	does	do	AUX
dad-10683	559	16	not	not	PART
dad-10683	559	17	concatenate	concatenate	VERB
dad-10683	559	18	the	the	DET
dad-10683	559	19	context	context	NOUN
dad-10683	559	20	utterances	utterance	NOUN
dad-10683	559	21	:	:	PUNCT
dad-10683	559	22	each	each	PRON
dad-10683	559	23	is	be	AUX
dad-10683	559	24	encoded	encode	VERB
dad-10683	559	25	with	with	ADP
dad-10683	559	26	a	a	DET
dad-10683	559	27	separate	separate	ADJ
dad-10683	559	28	utterance	utterance	ADJ
dad-10683	559	29	-	-	PUNCT
dad-10683	559	30	level	level	NOUN
dad-10683	559	31	encoder	encoder	NOUN
dad-10683	559	32	,	,	PUNCT
dad-10683	559	33	whose	whose	DET
dad-10683	559	34	output	output	NOUN
dad-10683	559	35	is	be	AUX
dad-10683	559	36	fed	feed	VERB
dad-10683	559	37	into	into	ADP
dad-10683	559	38	an	an	DET
dad-10683	559	39	additional	additional	ADJ
dad-10683	559	40	context	context	NOUN
dad-10683	559	41	-	-	PUNCT
dad-10683	559	42	level	level	NOUN
dad-10683	559	43	encoder	encoder	NOUN
dad-10683	559	44	.	.	PUNCT
dad-10683	560	1	the	the	DET
dad-10683	560	2	output	output	NOUN
dad-10683	560	3	of	of	ADP
dad-10683	560	4	the	the	DET
dad-10683	560	5	context	context	NOUN
dad-10683	560	6	-	-	PUNCT
dad-10683	560	7	level	level	NOUN
dad-10683	560	8	encoder	encoder	NOUN
dad-10683	560	9	depends	depend	VERB
dad-10683	560	10	on	on	ADP
dad-10683	560	11	all	all	DET
dad-10683	560	12	the	the	DET
dad-10683	560	13	utterances	utterance	NOUN
dad-10683	560	14	in	in	ADP
dad-10683	560	15	the	the	DET
dad-10683	560	16	context	context	NOUN
dad-10683	560	17	,	,	PUNCT
dad-10683	560	18	and	and	CCONJ
dad-10683	560	19	is	be	AUX
dad-10683	560	20	fed	feed	VERB
dad-10683	560	21	into	into	ADP
dad-10683	560	22	the	the	DET
dad-10683	560	23	decoder	decoder	NOUN
dad-10683	560	24	.	.	PUNCT
dad-10683	561	1	each	each	DET
dad-10683	561	2	rnn	rnn	NOUN
dad-10683	561	3	has	have	VERB
dad-10683	561	4	separate	separate	ADJ
dad-10683	561	5	parameters	parameter	NOUN
dad-10683	561	6	.	.	PUNCT
dad-10683	562	1	generative	generative	ADJ
dad-10683	562	2	metrics	metric	NOUN
dad-10683	562	3	embedding	embed	VERB
dad-10683	562	4	average	average	ADJ
dad-10683	562	5	greedy	greedy	ADJ
dad-10683	562	6	matching	matching	NOUN
dad-10683	562	7	vector	vector	NOUN
dad-10683	562	8	extrema	extrema	NOUN
dad-10683	562	9	lstm	lstm	NOUN
dad-10683	562	10	-	-	PUNCT
dad-10683	562	11	lm	lm	PROPN
dad-10683	562	12	0.561	0.561	NUM
dad-10683	562	13	0.425	0.425	NUM
dad-10683	562	14	0.380	0.380	NUM
dad-10683	562	15	hred	hre	VERB
dad-10683	562	16	0.617	0.617	NUM
dad-10683	562	17	0.452	0.452	NUM
dad-10683	562	18	0.408	0.408	NUM
dad-10683	562	19	tf	tf	PROPN
dad-10683	562	20	-	-	PUNCT
dad-10683	562	21	idf	idf	PROPN
dad-10683	562	22	0.536	0.536	NUM
dad-10683	562	23	0.370	0.370	NUM
dad-10683	562	24	0.342	0.342	NUM
dad-10683	562	25	dual	dual	ADJ
dad-10683	562	26	encoder	encoder	NOUN
dad-10683	562	27	w/	w/	NOUN
dad-10683	562	28	lstm	lstm	PROPN
dad-10683	562	29	units	unit	NOUN
dad-10683	562	30	0.650	0.650	NUM
dad-10683	562	31	0.413	0.413	NUM
dad-10683	562	32	0.376	0.376	NUM
dad-10683	562	33	table	table	NOUN
dad-10683	562	34	5	5	NUM
dad-10683	562	35	:	:	PUNCT
dad-10683	562	36	results	result	NOUN
dad-10683	562	37	for	for	ADP
dad-10683	562	38	both	both	CCONJ
dad-10683	562	39	the	the	DET
dad-10683	562	40	generative	generative	ADJ
dad-10683	562	41	and	and	CCONJ
dad-10683	562	42	retrieval	retrieval	NOUN
dad-10683	562	43	models	model	NOUN
dad-10683	562	44	on	on	ADP
dad-10683	562	45	the	the	DET
dad-10683	562	46	embedding	embed	VERB
dad-10683	562	47	average	average	ADJ
dad-10683	562	48	,	,	PUNCT
dad-10683	562	49	greedy	greedy	ADJ
dad-10683	562	50	matching	matching	NOUN
dad-10683	562	51	,	,	PUNCT
dad-10683	562	52	and	and	CCONJ
dad-10683	562	53	vector	vector	NOUN
dad-10683	562	54	extrema	extrema	NOUN
dad-10683	562	55	scores	score	NOUN
dad-10683	562	56	.	.	PUNCT
dad-10683	563	1	these	these	DET
dad-10683	563	2	scores	score	NOUN
dad-10683	563	3	provide	provide	VERB
dad-10683	563	4	an	an	DET
dad-10683	563	5	estimate	estimate	NOUN
dad-10683	563	6	of	of	ADP
dad-10683	563	7	the	the	DET
dad-10683	563	8	topic	topic	NOUN
dad-10683	563	9	consistency	consistency	NOUN
dad-10683	563	10	of	of	ADP
dad-10683	563	11	the	the	DET
dad-10683	563	12	generated	generate	VERB
dad-10683	563	13	responses	response	NOUN
dad-10683	563	14	.	.	PUNCT
dad-10683	564	1	retrieval	retrieval	NOUN
dad-10683	564	2	metrics	metric	NOUN
dad-10683	564	3	1	1	NUM
dad-10683	564	4	in	in	ADP
dad-10683	564	5	2	2	NUM
dad-10683	564	6	r@1	r@1	PROPN
dad-10683	564	7	1	1	NUM
dad-10683	564	8	in	in	ADP
dad-10683	564	9	10	10	NUM
dad-10683	564	10	r@1	r@1	PROPN
dad-10683	564	11	1	1	NUM
dad-10683	564	12	in	in	ADP
dad-10683	564	13	10	10	NUM
dad-10683	564	14	r@2	r@2	VERB
dad-10683	564	15	1	1	NUM
dad-10683	564	16	in	in	ADP
dad-10683	564	17	10	10	NUM
dad-10683	564	18	r@5	r@5	NOUN
dad-10683	564	19	lstm	lstm	PROPN
dad-10683	564	20	-	-	PUNCT
dad-10683	564	21	lm	lm	PROPN
dad-10683	564	22	58.9	58.9	NUM
dad-10683	564	23	%	%	NOUN
dad-10683	564	24	19.6	19.6	NUM
dad-10683	564	25	%	%	NOUN
dad-10683	564	26	33.1	33.1	NUM
dad-10683	564	27	%	%	NOUN
dad-10683	564	28	61.4	61.4	NUM
dad-10683	564	29	%	%	NOUN
dad-10683	564	30	hred	hre	VERB
dad-10683	564	31	61.8	61.8	NUM
dad-10683	564	32	%	%	NOUN
dad-10683	564	33	21.5	21.5	NUM
dad-10683	564	34	%	%	NOUN
dad-10683	564	35	35.8	35.8	NUM
dad-10683	564	36	%	%	NOUN
dad-10683	564	37	64.5	64.5	NUM
dad-10683	564	38	%	%	NOUN
dad-10683	565	1	tf	tf	PROPN
dad-10683	565	2	-	-	PUNCT
dad-10683	565	3	idf	idf	NOUN
dad-10683	565	4	74.9	74.9	NUM
dad-10683	565	5	%	%	NOUN
dad-10683	565	6	48.8	48.8	NUM
dad-10683	565	7	%	%	NOUN
dad-10683	565	8	58.7	58.7	NUM
dad-10683	565	9	%	%	NOUN
dad-10683	565	10	76.3	76.3	NUM
dad-10683	565	11	%	%	NOUN
dad-10683	565	12	dual	dual	ADJ
dad-10683	565	13	encoder	encoder	NOUN
dad-10683	565	14	w	w	PROPN
dad-10683	565	15	/	/	SYM
dad-10683	565	16	rnn	rnn	NOUN
dad-10683	565	17	units	unit	NOUN
dad-10683	565	18	77.7	77.7	NUM
dad-10683	565	19	%	%	NOUN
dad-10683	565	20	37.9	37.9	NUM
dad-10683	565	21	%	%	NOUN
dad-10683	565	22	56.1	56.1	NUM
dad-10683	565	23	%	%	NOUN
dad-10683	565	24	83.6	83.6	NUM
dad-10683	565	25	%	%	NOUN
dad-10683	565	26	dual	dual	ADJ
dad-10683	565	27	encoder	encoder	NOUN
dad-10683	565	28	w	w	PROPN
dad-10683	565	29	/	/	SYM
dad-10683	565	30	lstm	lstm	ADJ
dad-10683	565	31	units	unit	NOUN
dad-10683	565	32	86.9	86.9	NUM
dad-10683	565	33	%	%	NOUN
dad-10683	565	34	55.2	55.2	NUM
dad-10683	565	35	%	%	NOUN
dad-10683	565	36	72.1	72.1	NUM
dad-10683	565	37	%	%	NOUN
dad-10683	565	38	92.4	92.4	NUM
dad-10683	565	39	%	%	NOUN
dad-10683	565	40	memn2n	memn2n	PROPN
dad-10683	565	41	(	(	PUNCT
dad-10683	565	42	dodge	dodge	VERB
dad-10683	565	43	et	et	PROPN
dad-10683	565	44	al	al	PROPN
dad-10683	565	45	.	.	PROPN
dad-10683	565	46	,	,	PUNCT
dad-10683	565	47	2015	2015	NUM
dad-10683	565	48	)	)	PUNCT
dad-10683	565	49	—	—	PUNCT
dad-10683	566	1	63.72	63.72	NUM
dad-10683	566	2	%	%	NOUN
dad-10683	566	3	—	—	PUNCT
dad-10683	566	4	—	—	PUNCT
dad-10683	566	5	rnn	rnn	PROPN
dad-10683	566	6	-	-	PUNCT
dad-10683	566	7	cnn	cnn	PROPN
dad-10683	566	8	(	(	PUNCT
dad-10683	566	9	baudiš	baudiš	NOUN
dad-10683	566	10	and	and	CCONJ
dad-10683	566	11	šedivỳ	šedivỳ	PROPN
dad-10683	566	12	,	,	PUNCT
dad-10683	566	13	2016	2016	NUM
dad-10683	566	14	)	)	PUNCT
dad-10683	566	15	91.1	91.1	NUM
dad-10683	566	16	%	%	NOUN
dad-10683	566	17	67.2	67.2	NUM
dad-10683	566	18	%	%	NOUN
dad-10683	566	19	80.9	80.9	NUM
dad-10683	566	20	%	%	NOUN
dad-10683	566	21	95.6	95.6	NUM
dad-10683	566	22	%	%	NOUN
dad-10683	566	23	ensemble	ensemble	ADJ
dad-10683	566	24	(	(	PUNCT
dad-10683	566	25	kadlec	kadlec	PROPN
dad-10683	566	26	et	et	PROPN
dad-10683	566	27	al	al	PROPN
dad-10683	566	28	.	.	PROPN
dad-10683	566	29	,	,	PUNCT
dad-10683	566	30	2015	2015	NUM
dad-10683	566	31	)	)	PUNCT
dad-10683	566	32	91.5	91.5	NUM
dad-10683	566	33	%	%	NOUN
dad-10683	566	34	68.3	68.3	NUM
dad-10683	566	35	%	%	NOUN
dad-10683	566	36	81.8	81.8	NUM
dad-10683	566	37	%	%	NOUN
dad-10683	566	38	95.7	95.7	NUM
dad-10683	566	39	%	%	NOUN
dad-10683	566	40	r	r	NOUN
dad-10683	566	41	-	-	PUNCT
dad-10683	566	42	lstm	lstm	NOUN
dad-10683	566	43	(	(	PUNCT
dad-10683	566	44	xu	xu	INTJ
dad-10683	566	45	et	et	PROPN
dad-10683	566	46	al	al	PROPN
dad-10683	566	47	.	.	PROPN
dad-10683	566	48	,	,	PUNCT
dad-10683	566	49	2016	2016	NUM
dad-10683	566	50	)	)	PUNCT
dad-10683	566	51	88.9	88.9	NUM
dad-10683	566	52	%	%	NOUN
dad-10683	566	53	64.9	64.9	NUM
dad-10683	566	54	%	%	NOUN
dad-10683	566	55	78.5	78.5	NUM
dad-10683	566	56	%	%	NOUN
dad-10683	566	57	93.2	93.2	NUM
dad-10683	566	58	%	%	NOUN
dad-10683	566	59	table	table	NOUN
dad-10683	566	60	6	6	NUM
dad-10683	566	61	:	:	PUNCT
dad-10683	566	62	results	result	NOUN
dad-10683	566	63	for	for	ADP
dad-10683	566	64	both	both	CCONJ
dad-10683	566	65	the	the	DET
dad-10683	566	66	generative	generative	ADJ
dad-10683	566	67	and	and	CCONJ
dad-10683	566	68	retrieval	retrieval	NOUN
dad-10683	566	69	models	model	NOUN
dad-10683	566	70	using	use	VERB
dad-10683	566	71	various	various	ADJ
dad-10683	566	72	recall	recall	NOUN
dad-10683	566	73	measures	measure	NOUN
dad-10683	566	74	for	for	ADP
dad-10683	566	75	binary	binary	NOUN
dad-10683	566	76	(	(	PUNCT
dad-10683	566	77	1	1	NUM
dad-10683	566	78	in	in	ADP
dad-10683	566	79	2	2	NUM
dad-10683	566	80	)	)	PUNCT
dad-10683	566	81	and	and	CCONJ
dad-10683	566	82	1	1	NUM
dad-10683	566	83	in	in	ADP
dad-10683	566	84	10	10	NUM
dad-10683	566	85	(	(	PUNCT
dad-10683	566	86	1	1	NUM
dad-10683	566	87	in	in	ADP
dad-10683	566	88	10	10	NUM
dad-10683	566	89	)	)	PUNCT
dad-10683	566	90	next	next	ADJ
dad-10683	566	91	utterance	utterance	NOUN
dad-10683	566	92	classification	classification	NOUN
dad-10683	566	93	%	%	NOUN
dad-10683	566	94	.	.	PUNCT
dad-10683	567	1	we	we	PRON
dad-10683	567	2	include	include	VERB
dad-10683	567	3	state	state	NOUN
dad-10683	567	4	-	-	PUNCT
dad-10683	567	5	of	of	ADP
dad-10683	567	6	-	-	PUNCT
dad-10683	567	7	theart	theart	NOUN
dad-10683	567	8	results	result	NOUN
dad-10683	567	9	from	from	ADP
dad-10683	567	10	more	more	ADJ
dad-10683	567	11	recent	recent	ADJ
dad-10683	567	12	papers	paper	NOUN
dad-10683	567	13	.	.	PUNCT
dad-10683	568	1	55	55	NUM
dad-10683	568	2	lowe	lowe	NOUN
dad-10683	568	3	,	,	PUNCT
dad-10683	568	4	pow	pow	PROPN
dad-10683	568	5	,	,	PUNCT
dad-10683	568	6	serban	serban	PROPN
dad-10683	568	7	,	,	PUNCT
dad-10683	568	8	charlinn	charlinn	PROPN
dad-10683	568	9	,	,	PUNCT
dad-10683	568	10	liu	liu	PROPN
dad-10683	568	11	and	and	CCONJ
dad-10683	568	12	pineau	pineau	PROPN
dad-10683	568	13	5.3	5.3	NUM
dad-10683	568	14	experimental	experimental	ADJ
dad-10683	568	15	results	result	NOUN
dad-10683	568	16	we	we	PRON
dad-10683	568	17	now	now	ADV
dad-10683	568	18	examine	examine	VERB
dad-10683	568	19	the	the	DET
dad-10683	568	20	performance	performance	NOUN
dad-10683	568	21	of	of	ADP
dad-10683	568	22	the	the	DET
dad-10683	568	23	generative	generative	ADJ
dad-10683	568	24	models	model	NOUN
dad-10683	568	25	using	use	VERB
dad-10683	568	26	the	the	DET
dad-10683	568	27	vector	vector	NOUN
dad-10683	568	28	-	-	PUNCT
dad-10683	568	29	based	base	VERB
dad-10683	568	30	metrics	metric	NOUN
dad-10683	568	31	defined	define	VERB
dad-10683	568	32	in	in	ADP
dad-10683	568	33	section	section	NOUN
dad-10683	568	34	4.4	4.4	NUM
dad-10683	568	35	.	.	PUNCT
dad-10683	569	1	the	the	DET
dad-10683	569	2	results	result	NOUN
dad-10683	569	3	for	for	ADP
dad-10683	569	4	the	the	DET
dad-10683	569	5	rnn	rnn	PROPN
dad-10683	569	6	language	language	NOUN
dad-10683	569	7	model	model	NOUN
dad-10683	569	8	with	with	ADP
dad-10683	569	9	lstm	lstm	ADJ
dad-10683	569	10	units	unit	NOUN
dad-10683	569	11	(	(	PUNCT
dad-10683	569	12	lstm	lstm	NOUN
dad-10683	569	13	-	-	PUNCT
dad-10683	569	14	lm	lm	NOUN
dad-10683	569	15	)	)	PUNCT
dad-10683	569	16	and	and	CCONJ
dad-10683	569	17	the	the	DET
dad-10683	569	18	hred	hre	VERB
dad-10683	569	19	model	model	NOUN
dad-10683	569	20	can	can	AUX
dad-10683	569	21	be	be	AUX
dad-10683	569	22	seen	see	VERB
dad-10683	569	23	in	in	ADP
dad-10683	569	24	figure	figure	NOUN
dad-10683	569	25	5	5	NUM
dad-10683	569	26	.	.	PUNCT
dad-10683	569	27	as	as	SCONJ
dad-10683	569	28	expected	expect	VERB
dad-10683	569	29	,	,	PUNCT
dad-10683	569	30	the	the	DET
dad-10683	569	31	hred	hre	VERB
dad-10683	569	32	model	model	NOUN
dad-10683	569	33	outperforms	outperform	VERB
dad-10683	569	34	the	the	DET
dad-10683	569	35	baseline	baseline	ADJ
dad-10683	569	36	lstm	lstm	PROPN
dad-10683	569	37	model	model	NOUN
dad-10683	569	38	across	across	ADP
dad-10683	569	39	all	all	PRON
dad-10683	569	40	of	of	ADP
dad-10683	569	41	the	the	DET
dad-10683	569	42	metrics	metric	NOUN
dad-10683	569	43	.	.	PUNCT
dad-10683	570	1	however	however	ADV
dad-10683	570	2	,	,	PUNCT
dad-10683	570	3	it	it	PRON
dad-10683	570	4	is	be	AUX
dad-10683	570	5	interesting	interesting	ADJ
dad-10683	570	6	to	to	PART
dad-10683	570	7	note	note	VERB
dad-10683	570	8	that	that	SCONJ
dad-10683	570	9	in	in	ADP
dad-10683	570	10	a	a	DET
dad-10683	570	11	direct	direct	ADJ
dad-10683	570	12	comparison	comparison	NOUN
dad-10683	570	13	with	with	ADP
dad-10683	570	14	the	the	DET
dad-10683	570	15	dual	dual	ADJ
dad-10683	570	16	encoder	encoder	NOUN
dad-10683	570	17	model	model	NOUN
dad-10683	570	18	,	,	PUNCT
dad-10683	570	19	the	the	DET
dad-10683	570	20	hred	hre	VERB
dad-10683	570	21	model	model	NOUN
dad-10683	570	22	has	have	VERB
dad-10683	570	23	a	a	DET
dad-10683	570	24	higher	high	ADJ
dad-10683	570	25	score	score	NOUN
dad-10683	570	26	in	in	ADP
dad-10683	570	27	2	2	NUM
dad-10683	570	28	out	out	ADP
dad-10683	570	29	of	of	ADP
dad-10683	570	30	the	the	DET
dad-10683	570	31	3	3	NUM
dad-10683	570	32	metrics	metric	NOUN
dad-10683	570	33	considered	consider	VERB
dad-10683	570	34	.	.	PUNCT
dad-10683	571	1	thus	thus	ADV
dad-10683	571	2	,	,	PUNCT
dad-10683	571	3	it	it	PRON
dad-10683	571	4	is	be	AUX
dad-10683	571	5	likely	likely	ADJ
dad-10683	571	6	that	that	SCONJ
dad-10683	571	7	the	the	DET
dad-10683	571	8	hred	hre	VERB
dad-10683	571	9	model	model	NOUN
dad-10683	571	10	is	be	AUX
dad-10683	571	11	generating	generate	VERB
dad-10683	571	12	responses	response	NOUN
dad-10683	571	13	that	that	PRON
dad-10683	571	14	are	be	AUX
dad-10683	571	15	more	more	ADV
dad-10683	571	16	semantically	semantically	ADV
dad-10683	571	17	similar	similar	ADJ
dad-10683	571	18	to	to	ADP
dad-10683	571	19	the	the	DET
dad-10683	571	20	ground	ground	NOUN
dad-10683	571	21	-	-	PUNCT
dad-10683	571	22	truth	truth	NOUN
dad-10683	571	23	response	response	NOUN
dad-10683	571	24	,	,	PUNCT
dad-10683	571	25	and	and	CCONJ
dad-10683	571	26	is	be	AUX
dad-10683	571	27	better	well	ADJ
dad-10683	571	28	at	at	ADP
dad-10683	571	29	staying	stay	VERB
dad-10683	571	30	on	on	ADP
dad-10683	571	31	topic	topic	NOUN
dad-10683	571	32	.	.	PUNCT
dad-10683	572	1	we	we	PRON
dad-10683	572	2	also	also	ADV
dad-10683	572	3	present	present	VERB
dad-10683	572	4	the	the	DET
dad-10683	572	5	results	result	NOUN
dad-10683	572	6	for	for	ADP
dad-10683	572	7	these	these	DET
dad-10683	572	8	models	model	NOUN
dad-10683	572	9	on	on	ADP
dad-10683	572	10	the	the	DET
dad-10683	572	11	nuc	nuc	PROPN
dad-10683	572	12	task	task	NOUN
dad-10683	572	13	.	.	PUNCT
dad-10683	573	1	it	it	PRON
dad-10683	573	2	is	be	AUX
dad-10683	573	3	possible	possible	ADJ
dad-10683	573	4	to	to	PART
dad-10683	573	5	apply	apply	VERB
dad-10683	573	6	the	the	DET
dad-10683	573	7	generative	generative	ADJ
dad-10683	573	8	models	model	NOUN
dad-10683	573	9	to	to	ADP
dad-10683	573	10	the	the	DET
dad-10683	573	11	nuc	nuc	PROPN
dad-10683	573	12	task	task	NOUN
dad-10683	573	13	,	,	PUNCT
dad-10683	573	14	as	as	SCONJ
dad-10683	573	15	all	all	PRON
dad-10683	573	16	that	that	PRON
dad-10683	573	17	is	be	AUX
dad-10683	573	18	required	require	VERB
dad-10683	573	19	is	be	AUX
dad-10683	573	20	the	the	DET
dad-10683	573	21	ability	ability	NOUN
dad-10683	573	22	to	to	PART
dad-10683	573	23	assign	assign	VERB
dad-10683	573	24	probabilities	probability	NOUN
dad-10683	573	25	to	to	ADP
dad-10683	573	26	sequences	sequence	NOUN
dad-10683	573	27	of	of	ADP
dad-10683	573	28	utterances	utterance	NOUN
dad-10683	573	29	.	.	PUNCT
dad-10683	574	1	again	again	ADV
dad-10683	574	2	,	,	PUNCT
dad-10683	574	3	the	the	DET
dad-10683	574	4	hred	hre	VERB
dad-10683	574	5	model	model	NOUN
dad-10683	574	6	predictably	predictably	ADV
dad-10683	574	7	outperforms	outperform	VERB
dad-10683	574	8	the	the	DET
dad-10683	574	9	lstm	lstm	PROPN
dad-10683	574	10	-	-	PUNCT
dad-10683	574	11	lm	lm	ADJ
dad-10683	574	12	model	model	NOUN
dad-10683	574	13	on	on	ADP
dad-10683	574	14	all	all	DET
dad-10683	574	15	metrics	metric	NOUN
dad-10683	574	16	.	.	PUNCT
dad-10683	575	1	this	this	PRON
dad-10683	575	2	coincides	coincide	VERB
dad-10683	575	3	with	with	ADP
dad-10683	575	4	the	the	DET
dad-10683	575	5	results	result	NOUN
dad-10683	575	6	of	of	ADP
dad-10683	575	7	(	(	PUNCT
dad-10683	575	8	serban	serban	PROPN
dad-10683	575	9	et	et	PROPN
dad-10683	575	10	al	al	PROPN
dad-10683	575	11	.	.	PROPN
dad-10683	575	12	,	,	PUNCT
dad-10683	575	13	2016	2016	NUM
dad-10683	575	14	)	)	PUNCT
dad-10683	575	15	on	on	ADP
dad-10683	575	16	a	a	DET
dad-10683	575	17	different	different	ADJ
dad-10683	575	18	dataset	dataset	NOUN
dad-10683	575	19	,	,	PUNCT
dad-10683	575	20	using	use	VERB
dad-10683	575	21	different	different	ADJ
dad-10683	575	22	metrics	metric	NOUN
dad-10683	575	23	.	.	PUNCT
dad-10683	576	1	we	we	PRON
dad-10683	576	2	can	can	AUX
dad-10683	576	3	also	also	ADV
dad-10683	576	4	see	see	VERB
dad-10683	576	5	that	that	SCONJ
dad-10683	576	6	the	the	DET
dad-10683	576	7	generative	generative	ADJ
dad-10683	576	8	models	model	NOUN
dad-10683	576	9	perform	perform	VERB
dad-10683	576	10	much	much	ADV
dad-10683	576	11	worse	bad	ADJ
dad-10683	576	12	than	than	ADP
dad-10683	576	13	the	the	DET
dad-10683	576	14	models	model	NOUN
dad-10683	576	15	explicitly	explicitly	ADV
dad-10683	576	16	trained	train	VERB
dad-10683	576	17	to	to	PART
dad-10683	576	18	retrieve	retrieve	VERB
dad-10683	576	19	utterances	utterance	NOUN
dad-10683	576	20	from	from	ADP
dad-10683	576	21	a	a	DET
dad-10683	576	22	list	list	NOUN
dad-10683	576	23	.	.	PUNCT
dad-10683	577	1	this	this	PRON
dad-10683	577	2	is	be	AUX
dad-10683	577	3	to	to	PART
dad-10683	577	4	be	be	AUX
dad-10683	577	5	expected	expect	VERB
dad-10683	577	6	,	,	PUNCT
dad-10683	577	7	as	as	SCONJ
dad-10683	577	8	the	the	DET
dad-10683	577	9	retrieval	retrieval	NOUN
dad-10683	577	10	models	model	NOUN
dad-10683	577	11	were	be	AUX
dad-10683	577	12	trained	train	VERB
dad-10683	577	13	explicitly	explicitly	ADV
dad-10683	577	14	on	on	ADP
dad-10683	577	15	the	the	DET
dad-10683	577	16	nuc	nuc	PROPN
dad-10683	577	17	task	task	NOUN
dad-10683	577	18	,	,	PUNCT
dad-10683	577	19	while	while	SCONJ
dad-10683	577	20	the	the	DET
dad-10683	577	21	generative	generative	ADJ
dad-10683	577	22	models	model	NOUN
dad-10683	577	23	were	be	AUX
dad-10683	577	24	not	not	PART
dad-10683	577	25	.	.	PUNCT
dad-10683	578	1	because	because	SCONJ
dad-10683	578	2	of	of	ADP
dad-10683	578	3	the	the	DET
dad-10683	578	4	discrepancy	discrepancy	NOUN
dad-10683	578	5	in	in	ADP
dad-10683	578	6	training	training	NOUN
dad-10683	578	7	objectives	objective	NOUN
dad-10683	578	8	,	,	PUNCT
dad-10683	578	9	we	we	PRON
dad-10683	578	10	do	do	AUX
dad-10683	578	11	not	not	PART
dad-10683	578	12	recommend	recommend	VERB
dad-10683	578	13	the	the	DET
dad-10683	578	14	use	use	NOUN
dad-10683	578	15	of	of	ADP
dad-10683	578	16	nuc	nuc	NOUN
dad-10683	578	17	for	for	ADP
dad-10683	578	18	comparing	compare	VERB
dad-10683	578	19	generative	generative	ADJ
dad-10683	578	20	models	model	NOUN
dad-10683	578	21	with	with	ADP
dad-10683	578	22	retrieval	retrieval	NOUN
dad-10683	578	23	models	model	NOUN
dad-10683	578	24	.	.	PUNCT
dad-10683	579	1	however	however	ADV
dad-10683	579	2	,	,	PUNCT
dad-10683	579	3	we	we	PRON
dad-10683	579	4	believe	believe	VERB
dad-10683	579	5	that	that	SCONJ
dad-10683	579	6	nuc	nuc	PROPN
dad-10683	579	7	is	be	AUX
dad-10683	579	8	very	very	ADV
dad-10683	579	9	useful	useful	ADJ
dad-10683	579	10	for	for	ADP
dad-10683	579	11	comparing	compare	VERB
dad-10683	579	12	generative	generative	ADJ
dad-10683	579	13	models	model	NOUN
dad-10683	579	14	with	with	ADP
dad-10683	579	15	other	other	ADJ
dad-10683	579	16	generative	generative	ADJ
dad-10683	579	17	models	model	NOUN
dad-10683	579	18	,	,	PUNCT
dad-10683	579	19	and	and	CCONJ
dad-10683	579	20	retrieval	retrieval	NOUN
dad-10683	579	21	models	model	NOUN
dad-10683	579	22	with	with	ADP
dad-10683	579	23	retrieval	retrieval	NOUN
dad-10683	579	24	models	model	NOUN
dad-10683	579	25	;	;	PUNCT
dad-10683	579	26	we	we	PRON
dad-10683	579	27	justify	justify	VERB
dad-10683	579	28	this	this	PRON
dad-10683	579	29	further	far	ADV
dad-10683	579	30	in	in	ADP
dad-10683	579	31	section	section	NOUN
dad-10683	579	32	6.3	6.3	NUM
dad-10683	579	33	.	.	PUNCT
dad-10683	580	1	5.4	5.4	NUM
dad-10683	580	2	examples	example	NOUN
dad-10683	580	3	of	of	ADP
dad-10683	580	4	generated	generate	VERB
dad-10683	580	5	responses	response	NOUN
dad-10683	580	6	in	in	ADP
dad-10683	580	7	order	order	NOUN
dad-10683	580	8	to	to	PART
dad-10683	580	9	obtain	obtain	VERB
dad-10683	580	10	a	a	DET
dad-10683	580	11	better	well	ADJ
dad-10683	580	12	understanding	understanding	NOUN
dad-10683	580	13	of	of	ADP
dad-10683	580	14	the	the	DET
dad-10683	580	15	quality	quality	NOUN
dad-10683	580	16	of	of	ADP
dad-10683	580	17	responses	response	NOUN
dad-10683	580	18	from	from	ADP
dad-10683	580	19	the	the	DET
dad-10683	580	20	generative	generative	ADJ
dad-10683	580	21	model	model	NOUN
dad-10683	580	22	,	,	PUNCT
dad-10683	580	23	we	we	PRON
dad-10683	580	24	provide	provide	VERB
dad-10683	580	25	a	a	DET
dad-10683	580	26	table	table	NOUN
dad-10683	580	27	of	of	ADP
dad-10683	580	28	examples	example	NOUN
dad-10683	580	29	from	from	ADP
dad-10683	580	30	the	the	DET
dad-10683	580	31	lstm	lstm	PROPN
dad-10683	580	32	-	-	PUNCT
dad-10683	580	33	lm	lm	ADJ
dad-10683	580	34	and	and	CCONJ
dad-10683	580	35	hred	hre	VERB
dad-10683	580	36	models	model	NOUN
dad-10683	580	37	in	in	ADP
dad-10683	580	38	table	table	NOUN
dad-10683	580	39	7	7	NUM
dad-10683	580	40	.	.	PUNCT
dad-10683	581	1	we	we	PRON
dad-10683	581	2	chose	choose	VERB
dad-10683	581	3	several	several	ADJ
dad-10683	581	4	representative	representative	ADJ
dad-10683	581	5	samples	sample	NOUN
dad-10683	581	6	that	that	PRON
dad-10683	581	7	provide	provide	VERB
dad-10683	581	8	insight	insight	NOUN
dad-10683	581	9	into	into	ADP
dad-10683	581	10	the	the	DET
dad-10683	581	11	behaviour	behaviour	NOUN
dad-10683	581	12	,	,	PUNCT
dad-10683	581	13	and	and	CCONJ
dad-10683	581	14	the	the	DET
dad-10683	581	15	limitations	limitation	NOUN
dad-10683	581	16	,	,	PUNCT
dad-10683	581	17	of	of	ADP
dad-10683	581	18	the	the	DET
dad-10683	581	19	generative	generative	ADJ
dad-10683	581	20	models.12	models.12	PROPN
dad-10683	581	21	first	first	ADV
dad-10683	581	22	,	,	PUNCT
dad-10683	581	23	we	we	PRON
dad-10683	581	24	note	note	VERB
dad-10683	581	25	that	that	SCONJ
dad-10683	581	26	there	there	PRON
dad-10683	581	27	are	be	VERB
dad-10683	581	28	several	several	ADJ
dad-10683	581	29	situations	situation	NOUN
dad-10683	581	30	where	where	SCONJ
dad-10683	581	31	the	the	DET
dad-10683	581	32	generative	generative	ADJ
dad-10683	581	33	models	model	NOUN
dad-10683	581	34	,	,	PUNCT
dad-10683	581	35	particularly	particularly	ADV
dad-10683	581	36	the	the	DET
dad-10683	581	37	hred	hre	VERB
dad-10683	581	38	model	model	NOUN
dad-10683	581	39	,	,	PUNCT
dad-10683	581	40	produce	produce	VERB
dad-10683	581	41	reasonable	reasonable	ADJ
dad-10683	581	42	responses	response	NOUN
dad-10683	581	43	.	.	PUNCT
dad-10683	582	1	for	for	ADP
dad-10683	582	2	instance	instance	NOUN
dad-10683	582	3	,	,	PUNCT
dad-10683	582	4	in	in	ADP
dad-10683	582	5	example	example	NOUN
dad-10683	582	6	8)	8)	NUM
dad-10683	582	7	,	,	PUNCT
dad-10683	582	8	the	the	DET
dad-10683	582	9	models	model	NOUN
dad-10683	582	10	are	be	AUX
dad-10683	582	11	able	able	ADJ
dad-10683	582	12	to	to	PART
dad-10683	582	13	generate	generate	VERB
dad-10683	582	14	a	a	DET
dad-10683	582	15	coherent	coherent	ADJ
dad-10683	582	16	and	and	CCONJ
dad-10683	582	17	useful	useful	ADJ
dad-10683	582	18	response	response	NOUN
dad-10683	582	19	concerning	concern	VERB
dad-10683	582	20	the	the	DET
dad-10683	582	21	installation	installation	NOUN
dad-10683	582	22	of	of	ADP
dad-10683	582	23	grub	grub	NOUN
dad-10683	582	24	.	.	PUNCT
dad-10683	583	1	the	the	DET
dad-10683	583	2	models	model	NOUN
dad-10683	583	3	achieve	achieve	VERB
dad-10683	583	4	this	this	PRON
dad-10683	583	5	despite	despite	SCONJ
dad-10683	583	6	having	have	VERB
dad-10683	583	7	no	no	DET
dad-10683	583	8	knowledge	knowledge	NOUN
dad-10683	583	9	of	of	ADP
dad-10683	583	10	ubuntu	ubuntu	NOUN
dad-10683	583	11	other	other	ADJ
dad-10683	583	12	than	than	ADP
dad-10683	583	13	what	what	PRON
dad-10683	583	14	it	it	PRON
dad-10683	583	15	has	have	AUX
dad-10683	583	16	observed	observe	VERB
dad-10683	583	17	in	in	ADP
dad-10683	583	18	the	the	DET
dad-10683	583	19	training	training	NOUN
dad-10683	583	20	corpus	corpus	NOUN
dad-10683	583	21	.	.	PUNCT
dad-10683	584	1	the	the	DET
dad-10683	584	2	hred	hre	VERB
dad-10683	584	3	model	model	NOUN
dad-10683	584	4	is	be	AUX
dad-10683	584	5	also	also	ADV
dad-10683	584	6	able	able	ADJ
dad-10683	584	7	to	to	PART
dad-10683	584	8	generate	generate	VERB
dad-10683	584	9	reasonable	reasonable	ADJ
dad-10683	584	10	responses	response	NOUN
dad-10683	584	11	when	when	SCONJ
dad-10683	584	12	it	it	PRON
dad-10683	584	13	is	be	AUX
dad-10683	584	14	playing	play	VERB
dad-10683	584	15	the	the	DET
dad-10683	584	16	role	role	NOUN
dad-10683	584	17	of	of	ADP
dad-10683	584	18	the	the	DET
dad-10683	584	19	non	non	NOUN
dad-10683	584	20	-	-	NOUN
dad-10683	584	21	expert	expert	ADJ
dad-10683	584	22	who	who	PRON
dad-10683	584	23	is	be	AUX
dad-10683	584	24	seeking	seek	VERB
dad-10683	584	25	help	help	NOUN
dad-10683	584	26	.	.	PUNCT
dad-10683	585	1	this	this	PRON
dad-10683	585	2	is	be	AUX
dad-10683	585	3	the	the	DET
dad-10683	585	4	case	case	NOUN
dad-10683	585	5	in	in	ADP
dad-10683	585	6	examples	example	NOUN
dad-10683	585	7	1	1	NUM
dad-10683	585	8	)	)	PUNCT
dad-10683	585	9	,	,	PUNCT
dad-10683	585	10	3	3	NUM
dad-10683	585	11	)	)	PUNCT
dad-10683	585	12	,	,	PUNCT
dad-10683	585	13	and	and	CCONJ
dad-10683	585	14	6	6	NUM
dad-10683	585	15	)	)	PUNCT
dad-10683	585	16	.	.	PUNCT
dad-10683	586	1	however	however	ADV
dad-10683	586	2	,	,	PUNCT
dad-10683	586	3	the	the	DET
dad-10683	586	4	responses	response	NOUN
dad-10683	586	5	from	from	ADP
dad-10683	586	6	the	the	DET
dad-10683	586	7	model	model	NOUN
dad-10683	586	8	are	be	AUX
dad-10683	586	9	fairly	fairly	ADV
dad-10683	586	10	homogeneous	homogeneous	ADJ
dad-10683	586	11	;	;	PUNCT
dad-10683	586	12	it	it	PRON
dad-10683	586	13	is	be	AUX
dad-10683	586	14	simply	simply	ADV
dad-10683	586	15	stating	state	VERB
dad-10683	586	16	that	that	SCONJ
dad-10683	586	17	it	it	PRON
dad-10683	586	18	does	do	AUX
dad-10683	586	19	not	not	PART
dad-10683	586	20	know	know	VERB
dad-10683	586	21	how	how	SCONJ
dad-10683	586	22	to	to	PART
dad-10683	586	23	solve	solve	VERB
dad-10683	586	24	the	the	DET
dad-10683	586	25	problem	problem	NOUN
dad-10683	586	26	.	.	PUNCT
dad-10683	587	1	this	this	PRON
dad-10683	587	2	is	be	AUX
dad-10683	587	3	a	a	DET
dad-10683	587	4	commonly	commonly	ADV
dad-10683	587	5	observed	observe	VERB
dad-10683	587	6	behaviour	behaviour	NOUN
dad-10683	587	7	in	in	ADP
dad-10683	587	8	the	the	DET
dad-10683	587	9	generated	generate	VERB
dad-10683	587	10	samples	sample	NOUN
dad-10683	587	11	of	of	ADP
dad-10683	587	12	the	the	DET
dad-10683	587	13	model	model	NOUN
dad-10683	587	14	.	.	PUNCT
dad-10683	588	1	given	give	VERB
dad-10683	588	2	a	a	DET
dad-10683	588	3	context	context	NOUN
dad-10683	588	4	,	,	PUNCT
dad-10683	588	5	it	it	PRON
dad-10683	588	6	is	be	AUX
dad-10683	588	7	often	often	ADV
dad-10683	588	8	reasonable	reasonable	ADJ
dad-10683	588	9	to	to	ADP
dad-10683	588	10	simply	simply	ADV
dad-10683	588	11	state	state	VERB
dad-10683	588	12	‘	'	PUNCT
dad-10683	588	13	i	i	PRON
dad-10683	588	14	do	do	AUX
dad-10683	588	15	n’t	not	PART
dad-10683	588	16	know	know	VERB
dad-10683	588	17	’	'	PUNCT
dad-10683	588	18	,	,	PUNCT
dad-10683	588	19	or	or	CCONJ
dad-10683	588	20	‘	'	PUNCT
dad-10683	588	21	i	i	PRON
dad-10683	588	22	’m	’m	VERB
dad-10683	588	23	not	not	PART
dad-10683	588	24	sure	sure	ADJ
dad-10683	588	25	’	'	PUNCT
dad-10683	588	26	,	,	PUNCT
dad-10683	588	27	and	and	CCONJ
dad-10683	588	28	indeed	indeed	ADV
dad-10683	588	29	these	these	PRON
dad-10683	588	30	are	be	AUX
dad-10683	588	31	likely	likely	ADJ
dad-10683	588	32	responses	response	NOUN
dad-10683	588	33	in	in	ADP
dad-10683	588	34	the	the	DET
dad-10683	588	35	training	training	NOUN
dad-10683	588	36	set	set	NOUN
dad-10683	588	37	given	give	VERB
dad-10683	588	38	many	many	ADJ
dad-10683	588	39	different	different	ADJ
dad-10683	588	40	contexts	contexts	NOUN
dad-10683	588	41	.	.	PUNCT
dad-10683	589	1	this	this	PRON
dad-10683	589	2	is	be	AUX
dad-10683	589	3	the	the	DET
dad-10683	589	4	problem	problem	NOUN
dad-10683	589	5	of	of	ADP
dad-10683	589	6	generic	generic	ADJ
dad-10683	589	7	responses	response	NOUN
dad-10683	589	8	,	,	PUNCT
dad-10683	589	9	and	and	CCONJ
dad-10683	589	10	has	have	AUX
dad-10683	589	11	previously	previously	ADV
dad-10683	589	12	been	be	AUX
dad-10683	589	13	observed	observe	VERB
dad-10683	589	14	in	in	ADP
dad-10683	589	15	end	end	NOUN
dad-10683	589	16	-	-	PUNCT
dad-10683	589	17	to	to	ADP
dad-10683	589	18	-	-	PUNCT
dad-10683	589	19	end	end	NOUN
dad-10683	589	20	generative	generative	ADJ
dad-10683	589	21	dialogue	dialogue	NOUN
dad-10683	589	22	models	model	NOUN
dad-10683	589	23	(	(	PUNCT
dad-10683	589	24	vinyals	vinyal	NOUN
dad-10683	589	25	and	and	CCONJ
dad-10683	589	26	le	le	NOUN
dad-10683	589	27	,	,	PUNCT
dad-10683	589	28	2015	2015	NUM
dad-10683	589	29	)	)	PUNCT
dad-10683	589	30	.	.	PUNCT
dad-10683	590	1	there	there	PRON
dad-10683	590	2	are	be	VERB
dad-10683	590	3	other	other	ADJ
dad-10683	590	4	instances	instance	NOUN
dad-10683	590	5	where	where	SCONJ
dad-10683	590	6	the	the	DET
dad-10683	590	7	lack	lack	NOUN
dad-10683	590	8	of	of	ADP
dad-10683	590	9	semantic	semantic	ADJ
dad-10683	590	10	understanding	understanding	NOUN
dad-10683	590	11	is	be	AUX
dad-10683	590	12	clear	clear	ADJ
dad-10683	590	13	.	.	PUNCT
dad-10683	591	1	in	in	ADP
dad-10683	591	2	example	example	NOUN
dad-10683	591	3	2	2	NUM
dad-10683	591	4	)	)	PUNCT
dad-10683	591	5	,	,	PUNCT
dad-10683	591	6	the	the	DET
dad-10683	591	7	hred	hre	VERB
dad-10683	591	8	makes	make	VERB
dad-10683	591	9	a	a	DET
dad-10683	591	10	relevant	relevant	ADJ
dad-10683	591	11	statement	statement	NOUN
dad-10683	591	12	about	about	ADP
dad-10683	591	13	the	the	DET
dad-10683	591	14	command	command	NOUN
dad-10683	591	15	‘	'	PUNCT
dad-10683	591	16	sudo	sudo	ADJ
dad-10683	591	17	apt	apt	ADJ
dad-10683	591	18	-	-	PUNCT
dad-10683	591	19	get	get	NOUN
dad-10683	591	20	’	'	PUNCT
dad-10683	591	21	,	,	PUNCT
dad-10683	591	22	stating	state	VERB
dad-10683	591	23	that	that	SCONJ
dad-10683	591	24	the	the	DET
dad-10683	591	25	user	user	NOUN
dad-10683	591	26	should	should	AUX
dad-10683	591	27	try	try	VERB
dad-10683	591	28	to	to	PART
dad-10683	591	29	run	run	VERB
dad-10683	591	30	this	this	DET
dad-10683	591	31	command	command	NOUN
dad-10683	591	32	.	.	PUNCT
dad-10683	592	1	however	however	ADV
dad-10683	592	2	,	,	PUNCT
dad-10683	592	3	it	it	PRON
dad-10683	592	4	fails	fail	VERB
dad-10683	592	5	to	to	PART
dad-10683	592	6	understand	understand	VERB
dad-10683	592	7	that	that	SCONJ
dad-10683	592	8	the	the	DET
dad-10683	592	9	last	last	ADJ
dad-10683	592	10	user	user	NOUN
dad-10683	592	11	utterance	utterance	NOUN
dad-10683	592	12	(	(	PUNCT
dad-10683	592	13	‘	'	PUNCT
dad-10683	592	14	i	i	PRON
dad-10683	592	15	tried	try	VERB
dad-10683	592	16	that	that	PRON
dad-10683	592	17	,	,	PUNCT
dad-10683	592	18	it	it	PRON
dad-10683	592	19	comes	come	VERB
dad-10683	592	20	the	the	DET
dad-10683	592	21	same	same	ADJ
dad-10683	592	22	error	error	NOUN
dad-10683	592	23	’	'	PUNCT
dad-10683	592	24	)	)	PUNCT
dad-10683	592	25	,	,	PUNCT
dad-10683	592	26	refers	refer	VERB
dad-10683	592	27	to	to	ADP
dad-10683	592	28	the	the	DET
dad-10683	592	29	user	user	NOUN
dad-10683	592	30	having	having	AUX
dad-10683	592	31	already	already	ADV
dad-10683	592	32	tried	try	VERB
dad-10683	592	33	this	this	DET
dad-10683	592	34	exact	exact	ADJ
dad-10683	592	35	command	command	NOUN
dad-10683	592	36	.	.	PUNCT
dad-10683	593	1	it	it	PRON
dad-10683	593	2	is	be	AUX
dad-10683	593	3	difficult	difficult	ADJ
dad-10683	593	4	for	for	SCONJ
dad-10683	593	5	these	these	DET
dad-10683	593	6	models	model	NOUN
dad-10683	593	7	to	to	PART
dad-10683	593	8	integrate	integrate	VERB
dad-10683	593	9	information	information	NOUN
dad-10683	593	10	from	from	ADP
dad-10683	593	11	multiple	multiple	ADJ
dad-10683	593	12	utterances	utterance	NOUN
dad-10683	593	13	(	(	PUNCT
dad-10683	593	14	li	li	PROPN
dad-10683	593	15	et	et	PROPN
dad-10683	593	16	al	al	PROPN
dad-10683	593	17	.	.	PROPN
dad-10683	593	18	,	,	PUNCT
dad-10683	593	19	2016b	2016b	NUM
dad-10683	593	20	)	)	PUNCT
dad-10683	593	21	,	,	PUNCT
dad-10683	593	22	and	and	CCONJ
dad-10683	593	23	to	to	PART
dad-10683	593	24	understand	understand	VERB
dad-10683	593	25	the	the	DET
dad-10683	593	26	concept	concept	NOUN
dad-10683	593	27	of	of	ADP
dad-10683	593	28	coreference	coreference	NOUN
dad-10683	593	29	.	.	PUNCT
dad-10683	594	1	12	12	NUM
dad-10683	594	2	.	.	PUNCT
dad-10683	595	1	the	the	DET
dad-10683	595	2	full	full	ADJ
dad-10683	595	3	set	set	NOUN
dad-10683	595	4	of	of	ADP
dad-10683	595	5	model	model	NOUN
dad-10683	595	6	samples	sample	NOUN
dad-10683	595	7	are	be	AUX
dad-10683	595	8	available	available	ADJ
dad-10683	595	9	at	at	ADP
dad-10683	595	10	:	:	PUNCT
dad-10683	595	11	www.iulianserban.com/files/ubuntudialoguecorpus.zip	www.iulianserban.com/files/ubuntudialoguecorpus.zip	VERB
dad-10683	595	12	56	56	NUM
dad-10683	595	13	training	training	NOUN
dad-10683	595	14	end	end	NOUN
dad-10683	595	15	-	-	PUNCT
dad-10683	595	16	to	to	ADP
dad-10683	595	17	-	-	PUNCT
dad-10683	595	18	end	end	NOUN
dad-10683	595	19	dialogue	dialogue	NOUN
dad-10683	595	20	systems	system	NOUN
dad-10683	595	21	context	context	PROPN
dad-10683	595	22	response	response	NOUN
dad-10683	595	23	1	1	NUM
dad-10683	595	24	)	)	PUNCT
dad-10683	595	25	i	i	PRON
dad-10683	595	26	set	set	VERB
dad-10683	595	27	up	up	ADP
dad-10683	595	28	my	my	PRON
dad-10683	595	29	hd	hd	NOUN
dad-10683	595	30	such	such	ADJ
dad-10683	595	31	that	that	SCONJ
dad-10683	595	32	i	i	PRON
dad-10683	595	33	have	have	VERB
dad-10683	595	34	to	to	PART
dad-10683	595	35	type	type	VERB
dad-10683	595	36	a	a	DET
dad-10683	595	37	passphrase	passphrase	NOUN
dad-10683	595	38	to	to	PART
dad-10683	595	39	access	access	VERB
dad-10683	595	40	it	it	PRON
dad-10683	595	41	at	at	ADP
dad-10683	595	42	boot	boot	NOUN
dad-10683	595	43	.	.	PUNCT
dad-10683	596	1	how	how	SCONJ
dad-10683	596	2	can	can	AUX
dad-10683	596	3	i	i	PRON
dad-10683	596	4	remove	remove	VERB
dad-10683	596	5	that	that	DET
dad-10683	596	6	passwrd	passwrd	NOUN
dad-10683	596	7	,	,	PUNCT
dad-10683	596	8	and	and	CCONJ
dad-10683	596	9	just	just	ADV
dad-10683	596	10	boot	boot	VERB
dad-10683	596	11	up	up	ADP
dad-10683	596	12	normal	normal	ADJ
dad-10683	596	13	.	.	PUNCT
dad-10683	597	1	i	i	PRON
dad-10683	597	2	did	do	VERB
dad-10683	597	3	this	this	PRON
dad-10683	597	4	at	at	ADP
dad-10683	597	5	install	install	NOUN
dad-10683	597	6	,	,	PUNCT
dad-10683	597	7	it	it	PRON
dad-10683	597	8	works	work	VERB
dad-10683	597	9	fine	fine	ADV
dad-10683	597	10	,	,	PUNCT
dad-10683	597	11	just	just	ADV
dad-10683	597	12	tired	tired	ADJ
dad-10683	597	13	of	of	ADP
dad-10683	597	14	having	have	VERB
dad-10683	597	15	reboots	reboot	NOUN
dad-10683	597	16	where	where	SCONJ
dad-10683	597	17	i	i	PRON
dad-10683	597	18	need	need	VERB
dad-10683	597	19	to	to	PART
dad-10683	597	20	be	be	AUX
dad-10683	597	21	at	at	ADP
dad-10683	597	22	terminal	terminal	ADJ
dad-10683	597	23	to	to	PART
dad-10683	597	24	type	type	VERB
dad-10683	597	25	passwd	passwd	NOUN
dad-10683	597	26	in	in	ADP
dad-10683	597	27	.	.	PUNCT
dad-10683	598	1	help	help	VERB
dad-10683	598	2	?	?	PUNCT
dad-10683	599	1	→	→	SYM
dad-10683	599	2	backup	backup	ADJ
dad-10683	599	3	your	your	PRON
dad-10683	599	4	data	datum	NOUN
dad-10683	599	5	,	,	PUNCT
dad-10683	599	6	and	and	CCONJ
dad-10683	599	7	re	re	NOUN
dad-10683	599	8	-	-	NOUN
dad-10683	599	9	install	install	NOUN
dad-10683	599	10	without	without	ADP
dad-10683	599	11	encryption	encryption	NOUN
dad-10683	599	12	“	"	PUNCT
dad-10683	599	13	might	might	AUX
dad-10683	599	14	”	"	PUNCT
dad-10683	599	15	be	be	AUX
dad-10683	599	16	the	the	DET
dad-10683	599	17	easiest	easy	ADJ
dad-10683	599	18	method	method	NOUN
dad-10683	599	19	hred	hre	VERB
dad-10683	599	20	:	:	PUNCT
dad-10683	599	21	how	how	SCONJ
dad-10683	599	22	would	would	AUX
dad-10683	599	23	i	i	PRON
dad-10683	599	24	do	do	VERB
dad-10683	599	25	that	that	PRON
dad-10683	599	26	?	?	PUNCT
dad-10683	600	1	lstm	lstm	NOUN
dad-10683	600	2	:	:	PUNCT
dad-10683	600	3	how	how	SCONJ
dad-10683	600	4	do	do	AUX
dad-10683	600	5	i	i	PRON
dad-10683	600	6	do	do	VERB
dad-10683	600	7	that	that	PRON
dad-10683	600	8	?	?	PUNCT
dad-10683	601	1	2	2	X
dad-10683	601	2	)	)	PUNCT
dad-10683	601	3	hi	hi	INTJ
dad-10683	601	4	,	,	PUNCT
dad-10683	601	5	everyone	everyone	PRON
dad-10683	601	6	.	.	PUNCT
dad-10683	602	1	i	i	PRON
dad-10683	602	2	download	download	VERB
dad-10683	602	3	the	the	DET
dad-10683	602	4	googleearth	googleearth	PROPN
dad-10683	602	5	.	.	PUNCT
dad-10683	603	1	deb	deb	PROPN
dad-10683	603	2	,	,	PUNCT
dad-10683	603	3	and	and	CCONJ
dad-10683	603	4	want	want	VERB
dad-10683	603	5	to	to	PART
dad-10683	603	6	install	install	VERB
dad-10683	603	7	it	it	PRON
dad-10683	603	8	with	with	ADP
dad-10683	603	9	:	:	PUNCT
dad-10683	603	10	sudo	sudo	VERB
dad-10683	603	11	dpkg	dpkg	NOUN
dad-10683	603	12	-i	-i	PUNCT
dad-10683	603	13	something	something	PRON
dad-10683	603	14	.	.	PUNCT
dad-10683	604	1	deb	deb	PROPN
dad-10683	604	2	,	,	PUNCT
dad-10683	604	3	it	it	PRON
dad-10683	604	4	tells	tell	VERB
dad-10683	604	5	me	i	PRON
dad-10683	604	6	:	:	PUNCT
dad-10683	604	7	google	google	NOUN
dad-10683	604	8	-	-	PUNCT
dad-10683	604	9	earth	earth	NOUN
dad-10683	604	10	-	-	PUNCT
dad-10683	604	11	stable	stable	NOUN
dad-10683	604	12	depends	depend	VERB
dad-10683	604	13	on	on	ADP
dad-10683	604	14	lsb	lsb	PROPN
dad-10683	604	15	-	-	PUNCT
dad-10683	604	16	core	core	NOUN
dad-10683	604	17	(	(	PUNCT
dad-10683	604	18	>	>	PUNCT
dad-10683	604	19	=	=	PROPN
dad-10683	604	20	3.2	3.2	NUM
dad-10683	604	21	)	)	PUNCT
dad-10683	604	22	;	;	PUNCT
dad-10683	604	23	however	however	ADV
dad-10683	604	24	:	:	PUNCT
dad-10683	604	25	package	package	NOUN
dad-10683	604	26	lsb	lsb	PROPN
dad-10683	604	27	-	-	PUNCT
dad-10683	604	28	core	core	NOUN
dad-10683	604	29	is	be	AUX
dad-10683	604	30	not	not	PART
dad-10683	604	31	installed	instal	VERB
dad-10683	604	32	.	.	PUNCT
dad-10683	605	1	so	so	ADV
dad-10683	605	2	i	i	PRON
dad-10683	605	3	just	just	ADV
dad-10683	605	4	install	install	VERB
dad-10683	605	5	the	the	DET
dad-10683	605	6	lsb	lsb	PROPN
dad-10683	605	7	-	-	PUNCT
dad-10683	605	8	core	core	NOUN
dad-10683	605	9	,	,	PUNCT
dad-10683	605	10	but	but	CCONJ
dad-10683	605	11	it	it	PRON
dad-10683	605	12	comes	come	VERB
dad-10683	605	13	error	error	NOUN
dad-10683	605	14	:	:	PUNCT
dad-10683	605	15	<	<	X
dad-10683	605	16	url	url	X
dad-10683	605	17	>	>	X
dad-10683	605	18	what	what	PRON
dad-10683	605	19	should	should	AUX
dad-10683	605	20	i	i	PRON
dad-10683	605	21	do	do	VERB
dad-10683	605	22	?	?	PUNCT
dad-10683	606	1	→	→	PUNCT
dad-10683	606	2	the	the	DET
dad-10683	606	3	error	error	NOUN
dad-10683	606	4	suggests	suggest	VERB
dad-10683	606	5	running	run	VERB
dad-10683	606	6	“	"	PUNCT
dad-10683	606	7	sudo	sudo	X
dad-10683	606	8	apt	apt	ADJ
dad-10683	606	9	-	-	PUNCT
dad-10683	606	10	get	get	VERB
dad-10683	606	11	-f	-f	NUM
dad-10683	606	12	install	install	NOUN
dad-10683	606	13	”	"	PUNCT
dad-10683	606	14	→	→	PUNCT
dad-10683	606	15	i	i	PRON
dad-10683	606	16	tried	try	VERB
dad-10683	606	17	that	that	SCONJ
dad-10683	606	18	,	,	PUNCT
dad-10683	606	19	it	it	PRON
dad-10683	606	20	comes	come	VERB
dad-10683	606	21	the	the	DET
dad-10683	606	22	same	same	ADJ
dad-10683	606	23	error	error	NOUN
dad-10683	606	24	hred	hre	VERB
dad-10683	606	25	:	:	PUNCT
dad-10683	606	26	try	try	VERB
dad-10683	606	27	“	"	PUNCT
dad-10683	606	28	sudo	sudo	VERB
dad-10683	606	29	aptget	aptget	NOUN
dad-10683	606	30	-f	-f	NUM
dad-10683	606	31	install	install	VERB
dad-10683	606	32	”	"	PUNCT
dad-10683	606	33	lstm	lstm	NOUN
dad-10683	606	34	:	:	PUNCT
dad-10683	606	35	<	<	X
dad-10683	606	36	unk	unk	X
dad-10683	606	37	>	>	X
dad-10683	606	38	3	3	NUM
dad-10683	606	39	)	)	PUNCT
dad-10683	606	40	i	i	PRON
dad-10683	606	41	’ve	’ve	VERB
dad-10683	606	42	a	a	DET
dad-10683	606	43	2.0	2.0	NUM
dad-10683	606	44	tb	tb	ADP
dad-10683	606	45	hard	hard	ADJ
dad-10683	606	46	drive	drive	NOUN
dad-10683	606	47	.	.	PUNCT
dad-10683	607	1	it	it	PRON
dad-10683	607	2	no	no	ADV
dad-10683	607	3	longer	long	ADV
dad-10683	607	4	boots	boot	NOUN
dad-10683	607	5	from	from	ADP
dad-10683	607	6	the	the	DET
dad-10683	607	7	hard	hard	ADJ
dad-10683	607	8	drive	drive	NOUN
dad-10683	607	9	,	,	PUNCT
dad-10683	607	10	but	but	CCONJ
dad-10683	607	11	i	i	PRON
dad-10683	607	12	can	can	AUX
dad-10683	607	13	see	see	VERB
dad-10683	607	14	it	it	PRON
dad-10683	607	15	from	from	ADP
dad-10683	607	16	gpart	gpart	NOUN
dad-10683	607	17	and	and	CCONJ
dad-10683	607	18	it	it	PRON
dad-10683	607	19	has	have	VERB
dad-10683	607	20	a	a	DET
dad-10683	607	21	“	"	PUNCT
dad-10683	607	22	boot	boot	NOUN
dad-10683	607	23	”	"	PUNCT
dad-10683	607	24	flag	flag	NOUN
dad-10683	607	25	.	.	PUNCT
dad-10683	608	1	how	how	SCONJ
dad-10683	608	2	do	do	AUX
dad-10683	608	3	i	i	PRON
dad-10683	608	4	recover	recover	VERB
dad-10683	608	5	it	it	PRON
dad-10683	608	6	?	?	PUNCT
dad-10683	609	1	→	→	PUNCT
dad-10683	609	2	the	the	DET
dad-10683	609	3	boot	boot	NOUN
dad-10683	609	4	flag	flag	NOUN
dad-10683	609	5	does	do	AUX
dad-10683	609	6	n’t	not	PART
dad-10683	609	7	mean	mean	VERB
dad-10683	609	8	anything	anything	PRON
dad-10683	609	9	in	in	ADP
dad-10683	609	10	linux	linux	PROPN
dad-10683	609	11	why	why	SCONJ
dad-10683	609	12	do	do	AUX
dad-10683	609	13	you	you	PRON
dad-10683	609	14	not	not	PART
dad-10683	609	15	have	have	VERB
dad-10683	609	16	a	a	DET
dad-10683	609	17	backup	backup	NOUN
dad-10683	609	18	?	?	PUNCT
dad-10683	610	1	you	you	PRON
dad-10683	610	2	can	can	AUX
dad-10683	610	3	mount	mount	VERB
dad-10683	610	4	the	the	DET
dad-10683	610	5	partition	partition	NOUN
dad-10683	610	6	in	in	ADP
dad-10683	610	7	livecd	livecd	NOUN
dad-10683	610	8	and	and	CCONJ
dad-10683	610	9	access	access	VERB
dad-10683	610	10	the	the	DET
dad-10683	610	11	data	datum	NOUN
dad-10683	610	12	easily	easily	ADV
dad-10683	610	13	(	(	PUNCT
dad-10683	610	14	assuming	assume	VERB
dad-10683	610	15	the	the	DET
dad-10683	610	16	partition	partition	NOUN
dad-10683	610	17	is	be	AUX
dad-10683	610	18	healthy	healthy	ADJ
dad-10683	610	19	)	)	PUNCT
dad-10683	611	1	→	→	PUNCT
dad-10683	611	2	i	i	PRON
dad-10683	611	3	have	have	VERB
dad-10683	611	4	one	one	NUM
dad-10683	611	5	hard	hard	ADJ
dad-10683	611	6	drive	drive	NOUN
dad-10683	611	7	for	for	ADP
dad-10683	611	8	the	the	DET
dad-10683	611	9	moment	moment	NOUN
dad-10683	611	10	(	(	PUNCT
dad-10683	611	11	well	well	INTJ
dad-10683	611	12	,	,	PUNCT
dad-10683	611	13	i	i	PRON
dad-10683	611	14	do	do	AUX
dad-10683	611	15	n’t	not	PART
dad-10683	611	16	count	count	VERB
dad-10683	611	17	the	the	DET
dad-10683	611	18	250	250	NUM
dad-10683	611	19	gb	gb	NOUN
dad-10683	611	20	drive	drive	NOUN
dad-10683	611	21	)	)	PUNCT
dad-10683	611	22	besides	besides	SCONJ
dad-10683	611	23	,	,	PUNCT
dad-10683	611	24	i	i	PRON
dad-10683	611	25	do	do	AUX
dad-10683	611	26	n’t	not	PART
dad-10683	611	27	have	have	VERB
dad-10683	611	28	two	two	NUM
dad-10683	611	29	sata	sata	NOUN
dad-10683	611	30	cables	cable	NOUN
dad-10683	611	31	i	i	PRON
dad-10683	611	32	will	will	AUX
dad-10683	611	33	make	make	VERB
dad-10683	611	34	a	a	DET
dad-10683	611	35	backup	backup	NOUN
dad-10683	611	36	as	as	ADV
dad-10683	611	37	soon	soon	ADV
dad-10683	611	38	as	as	SCONJ
dad-10683	611	39	i	i	PRON
dad-10683	611	40	get	get	VERB
dad-10683	611	41	another	another	DET
dad-10683	611	42	nice	nice	ADJ
dad-10683	611	43	sized	sized	ADJ
dad-10683	611	44	hard	hard	ADJ
dad-10683	611	45	drive	drive	NOUN
dad-10683	611	46	and	and	CCONJ
dad-10683	611	47	a	a	DET
dad-10683	611	48	sata	sata	NOUN
dad-10683	611	49	cable	cable	NOUN
dad-10683	611	50	and	and	CCONJ
dad-10683	611	51	how	how	SCONJ
dad-10683	611	52	do	do	AUX
dad-10683	611	53	i	i	PRON
dad-10683	611	54	find	find	VERB
dad-10683	611	55	out	out	ADP
dad-10683	611	56	if	if	SCONJ
dad-10683	611	57	the	the	DET
dad-10683	611	58	partition	partition	NOUN
dad-10683	611	59	is	be	AUX
dad-10683	611	60	healthy	healthy	ADJ
dad-10683	611	61	?	?	PUNCT
dad-10683	612	1	→	→	PUNCT
dad-10683	612	2	boot	boot	NOUN
dad-10683	612	3	to	to	PART
dad-10683	612	4	ubuntu	ubuntu	VERB
dad-10683	612	5	livecd	livecd	NOUN
dad-10683	612	6	and	and	CCONJ
dad-10683	612	7	try	try	VERB
dad-10683	612	8	and	and	CCONJ
dad-10683	612	9	mount	mount	VERB
dad-10683	612	10	the	the	DET
dad-10683	612	11	partition	partition	NOUN
dad-10683	612	12	hred	hre	VERB
dad-10683	612	13	:	:	PUNCT
dad-10683	612	14	how	how	SCONJ
dad-10683	612	15	do	do	AUX
dad-10683	612	16	i	i	PRON
dad-10683	612	17	do	do	VERB
dad-10683	612	18	that	that	PRON
dad-10683	612	19	?	?	PUNCT
dad-10683	613	1	lstm	lstm	NOUN
dad-10683	613	2	:	:	PUNCT
dad-10683	613	3	on	on	ADP
dad-10683	613	4	the	the	DET
dad-10683	613	5	live	live	ADJ
dad-10683	613	6	cd	cd	NOUN
dad-10683	613	7	?	?	PROPN
dad-10683	614	1	4	4	NUM
dad-10683	614	2	)	)	PUNCT
dad-10683	614	3	bottom	bottom	ADJ
dad-10683	614	4	line	line	NOUN
dad-10683	614	5	is	be	AUX
dad-10683	614	6	usually	usually	ADV
dad-10683	614	7	if	if	SCONJ
dad-10683	614	8	you	you	PRON
dad-10683	614	9	got	get	VERB
dad-10683	614	10	64bit	64bit	NOUN
dad-10683	614	11	why	why	SCONJ
dad-10683	614	12	not	not	PART
dad-10683	614	13	use	use	VERB
dad-10683	614	14	it	it	PRON
dad-10683	614	15	?	?	PUNCT
dad-10683	615	1	(	(	PUNCT
dad-10683	615	2	unless	unless	SCONJ
dad-10683	615	3	there	there	PRON
dad-10683	615	4	is	be	VERB
dad-10683	615	5	something	something	PRON
dad-10683	615	6	that	that	PRON
dad-10683	615	7	does	do	AUX
dad-10683	615	8	n’t	not	PART
dad-10683	615	9	work	work	VERB
dad-10683	615	10	on	on	ADP
dad-10683	615	11	64bit	64bit	NOUN
dad-10683	615	12	,	,	PUNCT
dad-10683	615	13	which	which	PRON
dad-10683	615	14	is	be	AUX
dad-10683	615	15	amitedly	amitedly	ADV
dad-10683	615	16	not	not	PART
dad-10683	615	17	much	much	ADJ
dad-10683	615	18	)	)	PUNCT
dad-10683	616	1	→	→	SYM
dad-10683	616	2	well	well	INTJ
dad-10683	616	3	,	,	PUNCT
dad-10683	616	4	now	now	ADV
dad-10683	616	5	clip2net	clip2net	NOUN
dad-10683	616	6	is	be	AUX
dad-10683	616	7	not	not	PART
dad-10683	616	8	working	work	VERB
dad-10683	616	9	...	...	PUNCT
dad-10683	616	10	just	just	ADV
dad-10683	616	11	seems	seem	VERB
dad-10683	616	12	ubuntu	ubuntu	ADJ
dad-10683	616	13	runs	run	VERB
dad-10683	616	14	more	more	ADV
dad-10683	616	15	seamlessly	seamlessly	ADV
dad-10683	616	16	on	on	ADP
dad-10683	616	17	32	32	NUM
dad-10683	616	18	-	-	PUNCT
dad-10683	616	19	bit	bit	NOUN
dad-10683	616	20	and	and	CCONJ
dad-10683	616	21	after	after	ADP
dad-10683	616	22	reading	read	VERB
dad-10683	616	23	this	this	DET
dad-10683	616	24	article	article	NOUN
dad-10683	616	25	,	,	PUNCT
dad-10683	616	26	i	i	PRON
dad-10683	616	27	m	m	AUX
dad-10683	616	28	wondering	wonder	VERB
dad-10683	616	29	if	if	SCONJ
dad-10683	616	30	64	64	NUM
dad-10683	616	31	is	be	AUX
dad-10683	616	32	a	a	DET
dad-10683	616	33	bit	bit	NOUN
dad-10683	616	34	hyped	hype	VERB
dad-10683	616	35	?	?	PUNCT
dad-10683	617	1	<	<	AUX
dad-10683	617	2	url	url	X
dad-10683	617	3	>	>	X
dad-10683	617	4	hred	hre	VERB
dad-10683	617	5	:	:	PUNCT
dad-10683	617	6	i	i	PRON
dad-10683	617	7	have	have	VERB
dad-10683	617	8	no	no	DET
dad-10683	617	9	experience	experience	NOUN
dad-10683	617	10	with	with	ADP
dad-10683	617	11	64bit	64bit	NOUN
dad-10683	617	12	,	,	PUNCT
dad-10683	617	13	so	so	CCONJ
dad-10683	617	14	i	i	PRON
dad-10683	617	15	do	do	AUX
dad-10683	617	16	n’t	not	PART
dad-10683	617	17	know	know	VERB
dad-10683	617	18	what	what	PRON
dad-10683	617	19	you	you	PRON
dad-10683	617	20	mean	mean	VERB
dad-10683	617	21	.	.	PUNCT
dad-10683	618	1	lstm	lstm	NOUN
dad-10683	618	2	:	:	PUNCT
dad-10683	619	1	<	<	X
dad-10683	619	2	unk	unk	INTJ
dad-10683	619	3	>	>	X
dad-10683	619	4	,	,	PUNCT
dad-10683	619	5	<	<	X
dad-10683	619	6	unk	unk	INTJ
dad-10683	619	7	>	>	X
dad-10683	619	8	,	,	PUNCT
dad-10683	619	9	<	<	X
dad-10683	619	10	unk	unk	INTJ
dad-10683	619	11	>	>	X
dad-10683	619	12	,	,	PUNCT
dad-10683	619	13	<	<	X
dad-10683	619	14	unk	unk	INTJ
dad-10683	619	15	>	>	X
dad-10683	619	16	,	,	PUNCT
dad-10683	619	17	<	<	X
dad-10683	619	18	unk	unk	X
dad-10683	619	19	>	>	X
dad-10683	619	20	5	5	NUM
dad-10683	619	21	)	)	PUNCT
dad-10683	619	22	how	how	SCONJ
dad-10683	619	23	would	would	AUX
dad-10683	619	24	i	i	PRON
dad-10683	619	25	enable	enable	VERB
dad-10683	619	26	vim	vim	NOUN
dad-10683	619	27	syntax	syntax	NOUN
dad-10683	619	28	highlighting	highlighting	NOUN
dad-10683	619	29	on	on	ADP
dad-10683	619	30	a	a	DET
dad-10683	619	31	default	default	NOUN
dad-10683	619	32	install	install	NOUN
dad-10683	619	33	of	of	ADP
dad-10683	619	34	ubuntu	ubuntu	NOUN
dad-10683	619	35	12.04	12.04	NUM
dad-10683	619	36	?	?	PUNCT
dad-10683	620	1	→	→	PUNCT
dad-10683	620	2	have	have	AUX
dad-10683	620	3	you	you	PRON
dad-10683	620	4	insalled	insalle	VERB
dad-10683	620	5	vim	vim	NOUN
dad-10683	620	6	instead	instead	ADV
dad-10683	620	7	of	of	ADP
dad-10683	620	8	the	the	DET
dad-10683	620	9	default	default	NOUN
dad-10683	620	10	vim	vim	NOUN
dad-10683	620	11	-	-	PUNCT
dad-10683	620	12	tiny	tiny	ADJ
dad-10683	620	13	?	?	PUNCT
dad-10683	621	1	→	→	PUNCT
dad-10683	621	2	i	i	PRON
dad-10683	621	3	installed	instal	VERB
dad-10683	621	4	vim	vim	NOUN
dad-10683	621	5	-	-	PUNCT
dad-10683	621	6	runtime	runtime	NOUN
dad-10683	621	7	but	but	CCONJ
dad-10683	621	8	still	still	ADV
dad-10683	621	9	no	no	DET
dad-10683	621	10	highlighting	highlighting	NOUN
dad-10683	621	11	hred	hre	VERB
dad-10683	621	12	:	:	PUNCT
dad-10683	621	13	yes	yes	INTJ
dad-10683	621	14	lstm	lstm	ADJ
dad-10683	621	15	:	:	PUNCT
dad-10683	622	1	yes	yes	INTJ
dad-10683	622	2	6	6	X
dad-10683	622	3	)	)	PUNCT
dad-10683	622	4	its	its	PRON
dad-10683	622	5	comming	comming	NOUN
dad-10683	622	6	from	from	ADP
dad-10683	622	7	my	my	PRON
dad-10683	622	8	computer	computer	NOUN
dad-10683	622	9	15	15	NUM
dad-10683	622	10	people	people	NOUN
dad-10683	622	11	using	use	VERB
dad-10683	622	12	same	same	ADJ
dad-10683	622	13	router	router	NOUN
dad-10683	622	14	and	and	CCONJ
dad-10683	622	15	they	they	PRON
dad-10683	622	16	have	have	VERB
dad-10683	622	17	no	no	DET
dad-10683	622	18	issues	issue	NOUN
dad-10683	622	19	,	,	PUNCT
dad-10683	622	20	also	also	ADV
dad-10683	622	21	it	it	PRON
dad-10683	622	22	happened	happen	VERB
dad-10683	622	23	recently	recently	ADV
dad-10683	622	24	,	,	PUNCT
dad-10683	622	25	before	before	SCONJ
dad-10683	622	26	it	it	PRON
dad-10683	622	27	was	be	AUX
dad-10683	622	28	ok	ok	ADJ
dad-10683	622	29	→	→	PUNCT
dad-10683	623	1	yes	yes	INTJ
dad-10683	624	1	but	but	CCONJ
dad-10683	624	2	your	your	PRON
dad-10683	624	3	router	router	NOUN
dad-10683	624	4	may	may	AUX
dad-10683	624	5	have	have	VERB
dad-10683	624	6	optional	optional	ADJ
dad-10683	624	7	settings	setting	NOUN
dad-10683	624	8	for	for	ADP
dad-10683	624	9	dns	dns	NOUN
dad-10683	624	10	which	which	PRON
dad-10683	624	11	if	if	SCONJ
dad-10683	624	12	not	not	PART
dad-10683	624	13	set	set	VERB
dad-10683	624	14	to	to	ADP
dad-10683	624	15	the	the	DET
dad-10683	624	16	same	same	ADJ
dad-10683	624	17	dns	dns	NOUN
dad-10683	624	18	will	will	AUX
dad-10683	624	19	override	override	VERB
dad-10683	624	20	your	your	PRON
dad-10683	624	21	network	network	NOUN
dad-10683	624	22	interface	interface	NOUN
dad-10683	624	23	...	...	PUNCT
dad-10683	625	1	it	it	PRON
dad-10683	625	2	’s	’	VERB
dad-10683	625	3	an	an	DET
dad-10683	625	4	option	option	NOUN
dad-10683	625	5	to	to	PART
dad-10683	625	6	consider	consider	VERB
dad-10683	625	7	hred	hre	VERB
dad-10683	625	8	:	:	PUNCT
dad-10683	625	9	how	how	SCONJ
dad-10683	625	10	do	do	AUX
dad-10683	625	11	i	i	PRON
dad-10683	625	12	do	do	VERB
dad-10683	625	13	that	that	PRON
dad-10683	625	14	?	?	PUNCT
dad-10683	626	1	lstm	lstm	NOUN
dad-10683	626	2	:	:	PUNCT
dad-10683	626	3	that	that	PRON
dad-10683	626	4	is	be	AUX
dad-10683	626	5	what	what	PRON
dad-10683	626	6	i	i	PRON
dad-10683	626	7	was	be	AUX
dad-10683	626	8	looking	look	VERB
dad-10683	626	9	for	for	ADP
dad-10683	626	10	,	,	PUNCT
dad-10683	626	11	thanks	thank	NOUN
dad-10683	626	12	7	7	NUM
dad-10683	626	13	)	)	PUNCT
dad-10683	626	14	it	it	PRON
dad-10683	626	15	looks	look	VERB
dad-10683	626	16	as	as	SCONJ
dad-10683	626	17	if	if	SCONJ
dad-10683	626	18	the	the	DET
dad-10683	626	19	network	network	NOUN
dad-10683	626	20	’s	’s	PART
dad-10683	626	21	dhcp	dhcp	PROPN
dad-10683	626	22	server	server	NOUN
dad-10683	626	23	is	be	AUX
dad-10683	626	24	dead	dead	ADJ
dad-10683	626	25	:	:	PUNCT
dad-10683	626	26	“	"	PUNCT
dad-10683	626	27	<	<	X
dad-10683	626	28	warn	warn	VERB
dad-10683	626	29	>	>	X
dad-10683	626	30	(	(	PUNCT
dad-10683	626	31	eth0	eth0	PROPN
dad-10683	626	32	)	)	PUNCT
dad-10683	626	33	:	:	PUNCT
dad-10683	626	34	dhcpv4	dhcpv4	PROPN
dad-10683	626	35	request	request	NOUN
dad-10683	626	36	timed	time	VERB
dad-10683	626	37	out	out	ADP
dad-10683	626	38	”	"	PUNCT
dad-10683	626	39	→	→	PUNCT
dad-10683	626	40	but	but	CCONJ
dad-10683	626	41	it	it	PRON
dad-10683	626	42	is	be	AUX
dad-10683	626	43	n’t	n’t	ADJ
dad-10683	626	44	.	.	PUNCT
dad-10683	627	1	windows	window	NOUN
dad-10683	627	2	works	work	NOUN
dad-10683	627	3	hred	hre	VERB
dad-10683	627	4	:	:	PUNCT
dad-10683	627	5	that	that	PRON
dad-10683	627	6	’s	’	VERB
dad-10683	627	7	what	what	PRON
dad-10683	627	8	i	i	PRON
dad-10683	627	9	thought	think	VERB
dad-10683	627	10	,	,	PUNCT
dad-10683	627	11	but	but	CCONJ
dad-10683	627	12	i	i	PRON
dad-10683	627	13	do	do	AUX
dad-10683	627	14	n’t	not	PART
dad-10683	627	15	know	know	VERB
dad-10683	627	16	how	how	SCONJ
dad-10683	627	17	to	to	PART
dad-10683	627	18	fix	fix	VERB
dad-10683	627	19	it	it	PRON
dad-10683	627	20	lstm	lstm	ADJ
dad-10683	627	21	:	:	PUNCT
dad-10683	627	22	?	?	PUNCT
dad-10683	628	1	8)	8)	NUM
dad-10683	628	2	i	i	PRON
dad-10683	628	3	’ve	’ve	AUX
dad-10683	628	4	been	be	AUX
dad-10683	628	5	having	have	VERB
dad-10683	628	6	trouble	trouble	NOUN
dad-10683	628	7	with	with	ADP
dad-10683	628	8	a	a	DET
dad-10683	628	9	new	new	ADJ
dad-10683	628	10	3	3	NUM
dad-10683	628	11	tb	tb	NOUN
dad-10683	628	12	harddrive	harddrive	ADJ
dad-10683	628	13	and	and	CCONJ
dad-10683	628	14	a	a	DET
dad-10683	628	15	fresh	fresh	ADJ
dad-10683	628	16	ubuntu	ubuntu	ADJ
dad-10683	628	17	install	install	NOUN
dad-10683	628	18	.	.	PUNCT
dad-10683	629	1	i	i	PRON
dad-10683	629	2	install	install	VERB
dad-10683	629	3	(	(	PUNCT
dad-10683	629	4	as	as	ADP
dad-10683	629	5	gpt	gpt	NOUN
dad-10683	629	6	)	)	PUNCT
dad-10683	629	7	and	and	CCONJ
dad-10683	629	8	the	the	DET
dad-10683	629	9	install	install	NOUN
dad-10683	629	10	looks	look	VERB
dad-10683	629	11	ok	ok	ADJ
dad-10683	629	12	.	.	PUNCT
dad-10683	630	1	but	but	CCONJ
dad-10683	630	2	then	then	ADV
dad-10683	630	3	get	get	VERB
dad-10683	630	4	grub	grub	NOUN
dad-10683	630	5	“	"	PUNCT
dad-10683	630	6	error	error	NOUN
dad-10683	630	7	:	:	PUNCT
dad-10683	630	8	out	out	ADP
dad-10683	630	9	of	of	ADP
dad-10683	630	10	disk	disk	NOUN
dad-10683	630	11	”	"	PUNCT
dad-10683	630	12	after	after	ADP
dad-10683	630	13	reboot	reboot	NOUN
dad-10683	630	14	.	.	PUNCT
dad-10683	631	1	i	i	PRON
dad-10683	631	2	have	have	VERB
dad-10683	631	3	an	an	DET
dad-10683	631	4	october	october	PROPN
dad-10683	631	5	2011	2011	NUM
dad-10683	631	6	pc	pc	NOUN
dad-10683	631	7	with	with	ADP
dad-10683	631	8	updated	update	VERB
dad-10683	631	9	bios	bios	NOUN
dad-10683	631	10	and	and	CCONJ
dad-10683	631	11	i	i	PRON
dad-10683	631	12	do	do	AUX
dad-10683	631	13	see	see	VERB
dad-10683	631	14	the	the	DET
dad-10683	631	15	3	3	NUM
dad-10683	631	16	tb	tb	NOUN
dad-10683	631	17	.	.	PUNCT
dad-10683	632	1	google	google	PROPN
dad-10683	632	2	searches	search	NOUN
dad-10683	632	3	have	have	AUX
dad-10683	632	4	n’t	not	PART
dad-10683	632	5	been	be	AUX
dad-10683	632	6	to	to	ADP
dad-10683	632	7	helpful	helpful	ADJ
dad-10683	632	8	so	so	ADV
dad-10683	632	9	far	far	ADV
dad-10683	632	10	.	.	PUNCT
dad-10683	633	1	has	have	AUX
dad-10683	633	2	anyone	anyone	PRON
dad-10683	633	3	seen	see	VERB
dad-10683	633	4	grub	grub	NOUN
dad-10683	633	5	trouble	trouble	NOUN
dad-10683	633	6	when	when	SCONJ
dad-10683	633	7	installing	instal	VERB
dad-10683	633	8	on	on	ADP
dad-10683	633	9	3	3	NUM
dad-10683	633	10	tb	tb	ADP
dad-10683	633	11	harddrives	harddrive	NOUN
dad-10683	633	12	?	?	PUNCT
dad-10683	634	1	→	→	PUNCT
dad-10683	634	2	have	have	AUX
dad-10683	634	3	you	you	PRON
dad-10683	634	4	tried	try	VERB
dad-10683	634	5	installing	instal	VERB
dad-10683	634	6	grub	grub	NOUN
dad-10683	634	7	again	again	ADV
dad-10683	634	8	and	and	CCONJ
dad-10683	634	9	updating	update	VERB
dad-10683	634	10	it	it	PRON
dad-10683	634	11	?	?	PUNCT
dad-10683	635	1	sudo	sudo	VERB
dad-10683	635	2	grub	grub	NOUN
dad-10683	635	3	-	-	PUNCT
dad-10683	635	4	install	install	NOUN
dad-10683	635	5	sudo	sudo	NOUN
dad-10683	635	6	update	update	NOUN
dad-10683	635	7	-	-	PUNCT
dad-10683	635	8	grub	grub	NOUN
dad-10683	635	9	?	?	PUNCT
dad-10683	636	1	→	→	PUNCT
dad-10683	636	2	i	i	PRON
dad-10683	636	3	guess	guess	VERB
dad-10683	636	4	i	i	PRON
dad-10683	636	5	could	could	AUX
dad-10683	636	6	try	try	VERB
dad-10683	636	7	that	that	PRON
dad-10683	636	8	.	.	PUNCT
dad-10683	637	1	so	so	ADV
dad-10683	637	2	i	i	PRON
dad-10683	637	3	would	would	AUX
dad-10683	637	4	have	have	VERB
dad-10683	637	5	to	to	PART
dad-10683	637	6	install	install	VERB
dad-10683	637	7	grub	grub	NOUN
dad-10683	637	8	from	from	ADP
dad-10683	637	9	an	an	DET
dad-10683	637	10	older	old	ADJ
dad-10683	637	11	ubuntu	ubuntu	ADJ
dad-10683	637	12	install	install	NOUN
dad-10683	637	13	?	?	PUNCT
dad-10683	638	1	hred	hre	VERB
dad-10683	638	2	:	:	PUNCT
dad-10683	639	1	yes	yes	INTJ
dad-10683	639	2	,	,	PUNCT
dad-10683	639	3	you	you	PRON
dad-10683	639	4	can	can	AUX
dad-10683	639	5	install	install	VERB
dad-10683	639	6	grub	grub	NOUN
dad-10683	639	7	to	to	ADP
dad-10683	639	8	the	the	DET
dad-10683	639	9	mbr	mbr	PROPN
dad-10683	639	10	of	of	ADP
dad-10683	639	11	your	your	PRON
dad-10683	639	12	hard	hard	ADJ
dad-10683	639	13	drive	drive	NOUN
dad-10683	639	14	lstm	lstm	PROPN
dad-10683	639	15	:	:	PUNCT
dad-10683	639	16	grub	grub	NOUN
dad-10683	639	17	is	be	AUX
dad-10683	639	18	installed	instal	VERB
dad-10683	639	19	on	on	ADP
dad-10683	639	20	the	the	DET
dad-10683	639	21	mbr	mbr	NOUN
dad-10683	639	22	of	of	ADP
dad-10683	639	23	the	the	DET
dad-10683	639	24	hard	hard	ADJ
dad-10683	639	25	drive	drive	NOUN
dad-10683	639	26	table	table	NOUN
dad-10683	639	27	7	7	NUM
dad-10683	639	28	:	:	PUNCT
dad-10683	639	29	ubuntu	ubuntu	ADJ
dad-10683	639	30	model	model	NOUN
dad-10683	639	31	examples	example	NOUN
dad-10683	639	32	.	.	PUNCT
dad-10683	640	1	the	the	DET
dad-10683	640	2	→	→	PUNCT
dad-10683	640	3	token	token	PROPN
dad-10683	640	4	indicates	indicate	VERB
dad-10683	640	5	a	a	DET
dad-10683	640	6	change	change	NOUN
dad-10683	640	7	of	of	ADP
dad-10683	640	8	turn	turn	NOUN
dad-10683	640	9	,	,	PUNCT
dad-10683	640	10	and	and	CCONJ
dad-10683	640	11	<	<	X
dad-10683	640	12	unk	unk	X
dad-10683	640	13	>	>	X
dad-10683	640	14	is	be	AUX
dad-10683	640	15	the	the	DET
dad-10683	640	16	unknown	unknown	ADJ
dad-10683	640	17	token	token	NOUN
dad-10683	640	18	.	.	PUNCT
dad-10683	641	1	finally	finally	ADV
dad-10683	641	2	,	,	PUNCT
dad-10683	641	3	we	we	PRON
dad-10683	641	4	also	also	ADV
dad-10683	641	5	observed	observe	VERB
dad-10683	641	6	that	that	SCONJ
dad-10683	641	7	the	the	DET
dad-10683	641	8	lstm	lstm	PROPN
dad-10683	641	9	language	language	NOUN
dad-10683	641	10	model	model	NOUN
dad-10683	641	11	in	in	ADP
dad-10683	641	12	general	general	ADJ
dad-10683	641	13	produced	produce	VERB
dad-10683	641	14	many	many	ADJ
dad-10683	641	15	poor	poor	ADJ
dad-10683	641	16	responses	response	NOUN
dad-10683	641	17	,	,	PUNCT
dad-10683	641	18	consisting	consist	VERB
dad-10683	641	19	either	either	PRON
dad-10683	641	20	of	of	ADP
dad-10683	641	21	a	a	DET
dad-10683	641	22	single	single	ADJ
dad-10683	641	23	punctuation	punctuation	NOUN
dad-10683	641	24	mark	mark	NOUN
dad-10683	641	25	or	or	CCONJ
dad-10683	641	26	a	a	DET
dad-10683	641	27	series	series	NOUN
dad-10683	641	28	of	of	ADP
dad-10683	641	29	unknown	unknown	ADJ
dad-10683	641	30	tokens	token	NOUN
dad-10683	641	31	.	.	PUNCT
dad-10683	642	1	this	this	PRON
dad-10683	642	2	is	be	AUX
dad-10683	642	3	seen	see	VERB
dad-10683	642	4	in	in	ADP
dad-10683	642	5	examples	example	NOUN
dad-10683	642	6	2	2	NUM
dad-10683	642	7	)	)	PUNCT
dad-10683	642	8	,	,	PUNCT
dad-10683	642	9	4	4	NUM
dad-10683	642	10	)	)	PUNCT
dad-10683	642	11	,	,	PUNCT
dad-10683	642	12	and	and	CCONJ
dad-10683	642	13	7	7	NUM
dad-10683	642	14	)	)	PUNCT
dad-10683	642	15	.	.	PUNCT
dad-10683	643	1	on	on	ADP
dad-10683	643	2	the	the	DET
dad-10683	643	3	contrary	contrary	NOUN
dad-10683	643	4	,	,	PUNCT
dad-10683	643	5	the	the	DET
dad-10683	643	6	hred	hre	VERB
dad-10683	643	7	model	model	NOUN
dad-10683	643	8	rarely	rarely	ADV
dad-10683	643	9	produced	produce	VERB
dad-10683	643	10	such	such	ADJ
dad-10683	643	11	responses	response	NOUN
dad-10683	643	12	;	;	PUNCT
dad-10683	643	13	thus	thus	ADV
dad-10683	643	14	,	,	PUNCT
dad-10683	643	15	there	there	PRON
dad-10683	643	16	is	be	VERB
dad-10683	643	17	some	some	DET
dad-10683	643	18	indication	indication	NOUN
dad-10683	643	19	that	that	SCONJ
dad-10683	643	20	by	by	ADP
dad-10683	643	21	improving	improve	VERB
dad-10683	643	22	the	the	DET
dad-10683	643	23	model	model	NOUN
dad-10683	643	24	architecture	architecture	NOUN
dad-10683	643	25	,	,	PUNCT
dad-10683	643	26	we	we	PRON
dad-10683	643	27	will	will	AUX
dad-10683	643	28	produce	produce	VERB
dad-10683	643	29	models	model	NOUN
dad-10683	643	30	that	that	PRON
dad-10683	643	31	generate	generate	VERB
dad-10683	643	32	longer	long	ADV
dad-10683	643	33	and	and	CCONJ
dad-10683	643	34	more	more	ADV
dad-10683	643	35	interesting	interesting	ADJ
dad-10683	643	36	responses	response	NOUN
dad-10683	643	37	.	.	PUNCT
dad-10683	644	1	57	57	NUM
dad-10683	644	2	lowe	lowe	NOUN
dad-10683	644	3	,	,	PUNCT
dad-10683	644	4	pow	pow	PROPN
dad-10683	644	5	,	,	PUNCT
dad-10683	644	6	serban	serban	PROPN
dad-10683	644	7	,	,	PUNCT
dad-10683	644	8	charlinn	charlinn	PROPN
dad-10683	644	9	,	,	PUNCT
dad-10683	644	10	liu	liu	PROPN
dad-10683	644	11	and	and	CCONJ
dad-10683	644	12	pineau	pineau	PROPN
dad-10683	644	13	6	6	NUM
dad-10683	644	14	.	.	PUNCT
dad-10683	644	15	discussion	discussion	NOUN
dad-10683	644	16	this	this	DET
dad-10683	644	17	paper	paper	NOUN
dad-10683	644	18	presents	present	VERB
dad-10683	644	19	the	the	DET
dad-10683	644	20	ubuntu	ubuntu	ADJ
dad-10683	644	21	dialogue	dialogue	NOUN
dad-10683	644	22	corpus	corpus	PROPN
dad-10683	644	23	v2	v2	PROPN
dad-10683	644	24	,	,	PUNCT
dad-10683	644	25	a	a	DET
dad-10683	644	26	large	large	ADJ
dad-10683	644	27	dataset	dataset	NOUN
dad-10683	644	28	for	for	ADP
dad-10683	644	29	research	research	NOUN
dad-10683	644	30	in	in	ADP
dad-10683	644	31	unstructured	unstructured	ADJ
dad-10683	644	32	multi	multi	ADJ
dad-10683	644	33	-	-	ADJ
dad-10683	644	34	turn	turn	ADJ
dad-10683	644	35	dialogue	dialogue	NOUN
dad-10683	644	36	systems	system	NOUN
dad-10683	644	37	.	.	PUNCT
dad-10683	645	1	we	we	PRON
dad-10683	645	2	describe	describe	VERB
dad-10683	645	3	the	the	DET
dad-10683	645	4	construction	construction	NOUN
dad-10683	645	5	of	of	ADP
dad-10683	645	6	the	the	DET
dad-10683	645	7	dataset	dataset	NOUN
dad-10683	645	8	and	and	CCONJ
dad-10683	645	9	its	its	PRON
dad-10683	645	10	properties	property	NOUN
dad-10683	645	11	.	.	PUNCT
dad-10683	646	1	the	the	DET
dad-10683	646	2	availability	availability	NOUN
dad-10683	646	3	of	of	ADP
dad-10683	646	4	a	a	DET
dad-10683	646	5	dataset	dataset	NOUN
dad-10683	646	6	of	of	ADP
dad-10683	646	7	this	this	DET
dad-10683	646	8	size	size	NOUN
dad-10683	646	9	opens	open	VERB
dad-10683	646	10	up	up	ADP
dad-10683	646	11	several	several	ADJ
dad-10683	646	12	interesting	interesting	ADJ
dad-10683	646	13	possibilities	possibility	NOUN
dad-10683	646	14	for	for	ADP
dad-10683	646	15	research	research	NOUN
dad-10683	646	16	into	into	ADP
dad-10683	646	17	dialogue	dialogue	NOUN
dad-10683	646	18	systems	system	NOUN
dad-10683	646	19	based	base	VERB
dad-10683	646	20	on	on	ADP
dad-10683	646	21	rich	rich	ADJ
dad-10683	646	22	neural	neural	ADJ
dad-10683	646	23	-	-	PUNCT
dad-10683	646	24	network	network	NOUN
dad-10683	646	25	architectures	architecture	NOUN
dad-10683	646	26	.	.	PUNCT
dad-10683	647	1	we	we	PRON
dad-10683	647	2	present	present	VERB
dad-10683	647	3	results	result	NOUN
dad-10683	647	4	demonstrating	demonstrate	VERB
dad-10683	647	5	use	use	NOUN
dad-10683	647	6	of	of	ADP
dad-10683	647	7	this	this	DET
dad-10683	647	8	dataset	dataset	NOUN
dad-10683	647	9	to	to	PART
dad-10683	647	10	train	train	VERB
dad-10683	647	11	end	end	VERB
dad-10683	647	12	-	-	PUNCT
dad-10683	647	13	to	to	ADP
dad-10683	647	14	-	-	PUNCT
dad-10683	647	15	end	end	NOUN
dad-10683	647	16	rnn	rnn	NOUN
dad-10683	647	17	-	-	PUNCT
dad-10683	647	18	based	base	VERB
dad-10683	647	19	models	model	NOUN
dad-10683	647	20	,	,	PUNCT
dad-10683	647	21	and	and	CCONJ
dad-10683	647	22	critically	critically	ADV
dad-10683	647	23	evaluate	evaluate	VERB
dad-10683	647	24	the	the	DET
dad-10683	647	25	errors	error	NOUN
dad-10683	647	26	they	they	PRON
dad-10683	647	27	make	make	VERB
dad-10683	647	28	.	.	PUNCT
dad-10683	648	1	we	we	PRON
dad-10683	648	2	find	find	VERB
dad-10683	648	3	that	that	SCONJ
dad-10683	648	4	,	,	PUNCT
dad-10683	648	5	while	while	SCONJ
dad-10683	648	6	these	these	DET
dad-10683	648	7	models	model	NOUN
dad-10683	648	8	hold	hold	VERB
dad-10683	648	9	promise	promise	NOUN
dad-10683	648	10	for	for	ADP
dad-10683	648	11	building	build	VERB
dad-10683	648	12	non	non	ADJ
dad-10683	648	13	-	-	ADJ
dad-10683	648	14	task	task	ADJ
dad-10683	648	15	oriented	orient	VERB
dad-10683	648	16	dialogue	dialogue	NOUN
dad-10683	648	17	systems	system	NOUN
dad-10683	648	18	,	,	PUNCT
dad-10683	648	19	they	they	PRON
dad-10683	648	20	still	still	ADV
dad-10683	648	21	make	make	VERB
dad-10683	648	22	many	many	ADJ
dad-10683	648	23	obvious	obvious	ADJ
dad-10683	648	24	errors	error	NOUN
dad-10683	648	25	,	,	PUNCT
dad-10683	648	26	and	and	CCONJ
dad-10683	648	27	there	there	PRON
dad-10683	648	28	is	be	VERB
dad-10683	648	29	significant	significant	ADJ
dad-10683	648	30	room	room	NOUN
dad-10683	648	31	for	for	ADP
dad-10683	648	32	improvement	improvement	NOUN
dad-10683	648	33	.	.	PUNCT
dad-10683	649	1	next	next	ADV
dad-10683	649	2	,	,	PUNCT
dad-10683	649	3	we	we	PRON
dad-10683	649	4	outline	outline	VERB
dad-10683	649	5	several	several	ADJ
dad-10683	649	6	interesting	interesting	ADJ
dad-10683	649	7	directions	direction	NOUN
dad-10683	649	8	for	for	ADP
dad-10683	649	9	future	future	ADJ
dad-10683	649	10	work	work	NOUN
dad-10683	649	11	.	.	PUNCT
dad-10683	650	1	6.1	6.1	NUM
dad-10683	650	2	conversation	conversation	NOUN
dad-10683	650	3	disentanglement	disentanglement	NOUN
dad-10683	650	4	our	our	PRON
dad-10683	650	5	approach	approach	NOUN
dad-10683	650	6	to	to	ADP
dad-10683	650	7	conversation	conversation	NOUN
dad-10683	650	8	disentanglement	disentanglement	NOUN
dad-10683	650	9	consists	consist	VERB
dad-10683	650	10	of	of	ADP
dad-10683	650	11	a	a	DET
dad-10683	650	12	small	small	ADJ
dad-10683	650	13	set	set	NOUN
dad-10683	650	14	of	of	ADP
dad-10683	650	15	rules	rule	NOUN
dad-10683	650	16	.	.	PUNCT
dad-10683	651	1	more	more	ADV
dad-10683	651	2	sophisticated	sophisticated	ADJ
dad-10683	651	3	techniques	technique	NOUN
dad-10683	651	4	have	have	AUX
dad-10683	651	5	been	be	AUX
dad-10683	651	6	proposed	propose	VERB
dad-10683	651	7	,	,	PUNCT
dad-10683	651	8	such	such	ADJ
dad-10683	651	9	as	as	ADP
dad-10683	651	10	training	train	VERB
dad-10683	651	11	a	a	DET
dad-10683	651	12	maximum	maximum	ADJ
dad-10683	651	13	-	-	PUNCT
dad-10683	651	14	entropy	entropy	NOUN
dad-10683	651	15	classifier	classifier	NOUN
dad-10683	651	16	to	to	PART
dad-10683	651	17	cluster	cluster	VERB
dad-10683	651	18	utterances	utterance	NOUN
dad-10683	651	19	into	into	ADP
dad-10683	651	20	separate	separate	ADJ
dad-10683	651	21	dialogues	dialogue	NOUN
dad-10683	651	22	(	(	PUNCT
dad-10683	651	23	elsner	elsner	NOUN
dad-10683	651	24	and	and	CCONJ
dad-10683	651	25	charniak	charniak	NOUN
dad-10683	651	26	,	,	PUNCT
dad-10683	651	27	2008	2008	NUM
dad-10683	651	28	)	)	PUNCT
dad-10683	651	29	.	.	PUNCT
dad-10683	652	1	however	however	ADV
dad-10683	652	2	,	,	PUNCT
dad-10683	652	3	since	since	SCONJ
dad-10683	652	4	we	we	PRON
dad-10683	652	5	are	be	AUX
dad-10683	652	6	not	not	PART
dad-10683	652	7	trying	try	VERB
dad-10683	652	8	to	to	PART
dad-10683	652	9	replicate	replicate	VERB
dad-10683	652	10	the	the	DET
dad-10683	652	11	exact	exact	ADJ
dad-10683	652	12	conversation	conversation	NOUN
dad-10683	652	13	between	between	ADP
dad-10683	652	14	two	two	NUM
dad-10683	652	15	users	user	NOUN
dad-10683	652	16	,	,	PUNCT
dad-10683	652	17	but	but	CCONJ
dad-10683	652	18	only	only	ADV
dad-10683	652	19	to	to	PART
dad-10683	652	20	retrieve	retrieve	VERB
dad-10683	652	21	plausible	plausible	ADJ
dad-10683	652	22	natural	natural	ADJ
dad-10683	652	23	dialogues	dialogue	NOUN
dad-10683	652	24	,	,	PUNCT
dad-10683	652	25	the	the	DET
dad-10683	652	26	heuristic	heuristic	ADJ
dad-10683	652	27	method	method	NOUN
dad-10683	652	28	presented	present	VERB
dad-10683	652	29	in	in	ADP
dad-10683	652	30	this	this	DET
dad-10683	652	31	paper	paper	NOUN
dad-10683	652	32	may	may	AUX
dad-10683	652	33	be	be	AUX
dad-10683	652	34	sufficient	sufficient	ADJ
dad-10683	652	35	.	.	PUNCT
dad-10683	653	1	this	this	PRON
dad-10683	653	2	seems	seem	VERB
dad-10683	653	3	supported	support	VERB
dad-10683	653	4	through	through	ADP
dad-10683	653	5	qualitative	qualitative	ADJ
dad-10683	653	6	examination	examination	NOUN
dad-10683	653	7	of	of	ADP
dad-10683	653	8	the	the	DET
dad-10683	653	9	data	datum	NOUN
dad-10683	653	10	,	,	PUNCT
dad-10683	653	11	but	but	CCONJ
dad-10683	653	12	could	could	AUX
dad-10683	653	13	be	be	AUX
dad-10683	653	14	investigated	investigate	VERB
dad-10683	653	15	with	with	ADP
dad-10683	653	16	a	a	DET
dad-10683	653	17	more	more	ADV
dad-10683	653	18	formal	formal	ADJ
dad-10683	653	19	evaluation	evaluation	NOUN
dad-10683	653	20	.	.	PUNCT
dad-10683	654	1	6.2	6.2	NUM
dad-10683	654	2	non	non	ADJ
dad-10683	654	3	-	-	ADJ
dad-10683	654	4	task	task	ADJ
dad-10683	654	5	oriented	orient	VERB
dad-10683	654	6	model	model	NOUN
dad-10683	654	7	evaluation	evaluation	NOUN
dad-10683	654	8	it	it	PRON
dad-10683	654	9	may	may	AUX
dad-10683	654	10	seem	seem	VERB
dad-10683	654	11	unconventional	unconventional	ADJ
dad-10683	654	12	that	that	SCONJ
dad-10683	654	13	,	,	PUNCT
dad-10683	654	14	given	give	VERB
dad-10683	654	15	the	the	DET
dad-10683	654	16	technical	technical	ADJ
dad-10683	654	17	nature	nature	NOUN
dad-10683	654	18	of	of	ADP
dad-10683	654	19	the	the	DET
dad-10683	654	20	ubuntu	ubuntu	ADJ
dad-10683	654	21	dialogue	dialogue	NOUN
dad-10683	654	22	corpus	corpus	NOUN
dad-10683	654	23	and	and	CCONJ
dad-10683	654	24	the	the	DET
dad-10683	654	25	fact	fact	NOUN
dad-10683	654	26	that	that	SCONJ
dad-10683	654	27	it	it	PRON
dad-10683	654	28	involves	involve	VERB
dad-10683	654	29	interactions	interaction	NOUN
dad-10683	654	30	where	where	SCONJ
dad-10683	654	31	the	the	DET
dad-10683	654	32	end	end	NOUN
dad-10683	654	33	goal	goal	NOUN
dad-10683	654	34	is	be	AUX
dad-10683	654	35	solving	solve	VERB
dad-10683	654	36	a	a	DET
dad-10683	654	37	user	user	NOUN
dad-10683	654	38	’s	’s	PART
dad-10683	654	39	problem	problem	NOUN
dad-10683	654	40	,	,	PUNCT
dad-10683	654	41	we	we	PRON
dad-10683	654	42	are	be	AUX
dad-10683	654	43	treating	treat	VERB
dad-10683	654	44	our	our	PRON
dad-10683	654	45	models	model	NOUN
dad-10683	654	46	as	as	ADP
dad-10683	654	47	non	non	ADJ
dad-10683	654	48	-	-	ADJ
dad-10683	654	49	task	task	NOUN
dad-10683	654	50	oriented	orient	VERB
dad-10683	654	51	,	,	PUNCT
dad-10683	654	52	meaning	mean	VERB
dad-10683	654	53	that	that	SCONJ
dad-10683	654	54	we	we	PRON
dad-10683	654	55	do	do	AUX
dad-10683	654	56	not	not	PART
dad-10683	654	57	incorporate	incorporate	VERB
dad-10683	654	58	a	a	DET
dad-10683	654	59	supervised	supervised	ADJ
dad-10683	654	60	task	task	NOUN
dad-10683	654	61	completion	completion	NOUN
dad-10683	654	62	or	or	CCONJ
dad-10683	654	63	user	user	NOUN
dad-10683	654	64	satisfaction	satisfaction	NOUN
dad-10683	654	65	signal	signal	NOUN
dad-10683	654	66	during	during	ADP
dad-10683	654	67	training	training	NOUN
dad-10683	654	68	or	or	CCONJ
dad-10683	654	69	evaluation	evaluation	NOUN
dad-10683	654	70	.	.	PUNCT
dad-10683	655	1	the	the	DET
dad-10683	655	2	reasons	reason	NOUN
dad-10683	655	3	for	for	ADP
dad-10683	655	4	this	this	PRON
dad-10683	655	5	are	be	AUX
dad-10683	655	6	purely	purely	ADV
dad-10683	655	7	practical	practical	ADJ
dad-10683	655	8	;	;	PUNCT
dad-10683	655	9	in	in	ADP
dad-10683	655	10	general	general	ADJ
dad-10683	655	11	,	,	PUNCT
dad-10683	655	12	training	train	VERB
dad-10683	655	13	large	large	ADJ
dad-10683	655	14	,	,	PUNCT
dad-10683	655	15	end	end	NOUN
dad-10683	655	16	-	-	PUNCT
dad-10683	655	17	to	to	ADP
dad-10683	655	18	-	-	PUNCT
dad-10683	655	19	end	end	NOUN
dad-10683	655	20	goal	goal	NOUN
dad-10683	655	21	-	-	PUNCT
dad-10683	655	22	driven	drive	VERB
dad-10683	655	23	models	model	NOUN
dad-10683	655	24	is	be	AUX
dad-10683	655	25	very	very	ADV
dad-10683	655	26	difficult	difficult	ADJ
dad-10683	655	27	as	as	SCONJ
dad-10683	655	28	it	it	PRON
dad-10683	655	29	requires	require	VERB
dad-10683	655	30	the	the	DET
dad-10683	655	31	collection	collection	NOUN
dad-10683	655	32	of	of	ADP
dad-10683	655	33	a	a	DET
dad-10683	655	34	large	large	ADJ
dad-10683	655	35	amount	amount	NOUN
dad-10683	655	36	of	of	ADP
dad-10683	655	37	task	task	NOUN
dad-10683	655	38	completion	completion	NOUN
dad-10683	655	39	data	datum	NOUN
dad-10683	655	40	.	.	PUNCT
dad-10683	656	1	annotating	annotate	VERB
dad-10683	656	2	data	datum	NOUN
dad-10683	656	3	in	in	ADP
dad-10683	656	4	this	this	DET
dad-10683	656	5	way	way	NOUN
dad-10683	656	6	on	on	ADP
dad-10683	656	7	a	a	DET
dad-10683	656	8	large	large	ADJ
dad-10683	656	9	scale	scale	NOUN
dad-10683	656	10	is	be	AUX
dad-10683	656	11	extremely	extremely	ADV
dad-10683	656	12	expensive	expensive	ADJ
dad-10683	656	13	,	,	PUNCT
dad-10683	656	14	and	and	CCONJ
dad-10683	656	15	is	be	AUX
dad-10683	656	16	usually	usually	ADV
dad-10683	656	17	only	only	ADV
dad-10683	656	18	feasible	feasible	ADJ
dad-10683	656	19	for	for	ADP
dad-10683	656	20	technical	technical	ADJ
dad-10683	656	21	support	support	NOUN
dad-10683	656	22	channels	channel	NOUN
dad-10683	656	23	at	at	ADP
dad-10683	656	24	large	large	ADJ
dad-10683	656	25	corporations	corporation	NOUN
dad-10683	656	26	,	,	PUNCT
dad-10683	656	27	which	which	PRON
dad-10683	656	28	are	be	AUX
dad-10683	656	29	rarely	rarely	ADV
dad-10683	656	30	released	release	VERB
dad-10683	656	31	publicly	publicly	ADV
dad-10683	656	32	.	.	PUNCT
dad-10683	657	1	indeed	indeed	ADV
dad-10683	657	2	,	,	PUNCT
dad-10683	657	3	the	the	DET
dad-10683	657	4	ubuntu	ubuntu	ADJ
dad-10683	657	5	dialogue	dialogue	NOUN
dad-10683	657	6	corpus	corpus	NOUN
dad-10683	657	7	has	have	VERB
dad-10683	657	8	no	no	DET
dad-10683	657	9	such	such	ADJ
dad-10683	657	10	labelled	label	VERB
dad-10683	657	11	task	task	NOUN
dad-10683	657	12	completion	completion	NOUN
dad-10683	657	13	data	datum	NOUN
dad-10683	657	14	,	,	PUNCT
dad-10683	657	15	and	and	CCONJ
dad-10683	657	16	thus	thus	ADV
dad-10683	657	17	can	can	AUX
dad-10683	657	18	not	not	PART
dad-10683	657	19	be	be	AUX
dad-10683	657	20	analyzed	analyze	VERB
dad-10683	657	21	in	in	ADP
dad-10683	657	22	the	the	DET
dad-10683	657	23	task	task	NOUN
dad-10683	657	24	-	-	PUNCT
dad-10683	657	25	oriented	orient	VERB
dad-10683	657	26	setting	setting	NOUN
dad-10683	657	27	for	for	ADP
dad-10683	657	28	the	the	DET
dad-10683	657	29	time	time	NOUN
dad-10683	657	30	being	be	AUX
dad-10683	657	31	.	.	PUNCT
dad-10683	658	1	obtaining	obtain	VERB
dad-10683	658	2	such	such	ADJ
dad-10683	658	3	signals	signal	NOUN
dad-10683	658	4	automatically	automatically	ADV
dad-10683	658	5	remains	remain	VERB
dad-10683	658	6	an	an	DET
dad-10683	658	7	open	open	ADJ
dad-10683	658	8	problem	problem	NOUN
dad-10683	658	9	.	.	PUNCT
dad-10683	659	1	on	on	ADP
dad-10683	659	2	the	the	DET
dad-10683	659	3	other	other	ADJ
dad-10683	659	4	hand	hand	NOUN
dad-10683	659	5	,	,	PUNCT
dad-10683	659	6	training	train	VERB
dad-10683	659	7	non	non	ADJ
dad-10683	659	8	-	-	ADJ
dad-10683	659	9	task	task	ADJ
dad-10683	659	10	oriented	orient	VERB
dad-10683	659	11	dialogue	dialogue	NOUN
dad-10683	659	12	systems	system	NOUN
dad-10683	659	13	such	such	ADJ
dad-10683	659	14	as	as	ADP
dad-10683	659	15	chatbots	chatbot	NOUN
dad-10683	659	16	only	only	ADV
dad-10683	659	17	requires	require	VERB
dad-10683	659	18	conversational	conversational	ADJ
dad-10683	659	19	data	datum	NOUN
dad-10683	659	20	,	,	PUNCT
dad-10683	659	21	which	which	PRON
dad-10683	659	22	can	can	AUX
dad-10683	659	23	be	be	AUX
dad-10683	659	24	obtained	obtain	VERB
dad-10683	659	25	and	and	CCONJ
dad-10683	659	26	shared	share	VERB
dad-10683	659	27	publicly	publicly	ADV
dad-10683	659	28	on	on	ADP
dad-10683	659	29	a	a	DET
dad-10683	659	30	large	large	ADJ
dad-10683	659	31	scale	scale	NOUN
dad-10683	659	32	.	.	PUNCT
dad-10683	660	1	we	we	PRON
dad-10683	660	2	believe	believe	VERB
dad-10683	660	3	that	that	SCONJ
dad-10683	660	4	significant	significant	ADJ
dad-10683	660	5	progress	progress	NOUN
dad-10683	660	6	in	in	ADP
dad-10683	660	7	dialogue	dialogue	NOUN
dad-10683	660	8	systems	system	NOUN
dad-10683	660	9	can	can	AUX
dad-10683	660	10	be	be	AUX
dad-10683	660	11	made	make	VERB
dad-10683	660	12	in	in	ADP
dad-10683	660	13	this	this	DET
dad-10683	660	14	manner	manner	NOUN
dad-10683	660	15	,	,	PUNCT
dad-10683	660	16	as	as	SCONJ
dad-10683	660	17	there	there	PRON
dad-10683	660	18	remains	remain	VERB
dad-10683	660	19	many	many	ADJ
dad-10683	660	20	unsolved	unsolved	ADJ
dad-10683	660	21	problems	problem	NOUN
dad-10683	660	22	as	as	SCONJ
dad-10683	660	23	illustrated	illustrate	VERB
dad-10683	660	24	in	in	ADP
dad-10683	660	25	section	section	NOUN
dad-10683	660	26	4.6	4.6	NUM
dad-10683	660	27	.	.	PUNCT
dad-10683	661	1	6.3	6.3	NUM
dad-10683	661	2	automatic	automatic	ADJ
dad-10683	661	3	evaluation	evaluation	NOUN
dad-10683	661	4	of	of	ADP
dad-10683	661	5	dialogue	dialogue	NOUN
dad-10683	661	6	systems	system	NOUN
dad-10683	661	7	a	a	DET
dad-10683	661	8	crucial	crucial	ADJ
dad-10683	661	9	part	part	NOUN
dad-10683	661	10	of	of	ADP
dad-10683	661	11	research	research	NOUN
dad-10683	661	12	in	in	ADP
dad-10683	661	13	building	build	VERB
dad-10683	661	14	dialogue	dialogue	NOUN
dad-10683	661	15	systems	system	NOUN
dad-10683	661	16	concerns	concern	VERB
dad-10683	661	17	the	the	DET
dad-10683	661	18	problem	problem	NOUN
dad-10683	661	19	of	of	ADP
dad-10683	661	20	evaluation	evaluation	NOUN
dad-10683	661	21	.	.	PUNCT
dad-10683	662	1	in	in	ADP
dad-10683	662	2	the	the	DET
dad-10683	662	3	goal	goal	NOUN
dad-10683	662	4	-	-	PUNCT
dad-10683	662	5	oriented	orient	VERB
dad-10683	662	6	setting	setting	NOUN
dad-10683	662	7	,	,	PUNCT
dad-10683	662	8	when	when	SCONJ
dad-10683	662	9	there	there	PRON
dad-10683	662	10	is	be	VERB
dad-10683	662	11	a	a	DET
dad-10683	662	12	supervised	supervised	ADJ
dad-10683	662	13	task	task	NOUN
dad-10683	662	14	completion	completion	NOUN
dad-10683	662	15	signal	signal	NOUN
dad-10683	662	16	available	available	ADJ
dad-10683	662	17	with	with	ADP
dad-10683	662	18	the	the	DET
dad-10683	662	19	data	datum	NOUN
dad-10683	662	20	,	,	PUNCT
dad-10683	662	21	methods	method	NOUN
dad-10683	662	22	for	for	ADP
dad-10683	662	23	automatic	automatic	ADJ
dad-10683	662	24	evaluation	evaluation	NOUN
dad-10683	662	25	are	be	AUX
dad-10683	662	26	well	well	ADV
dad-10683	662	27	-	-	PUNCT
dad-10683	662	28	established	establish	VERB
dad-10683	662	29	,	,	PUNCT
dad-10683	662	30	such	such	ADJ
dad-10683	662	31	as	as	ADP
dad-10683	662	32	paradise	paradise	PROPN
dad-10683	662	33	(	(	PUNCT
dad-10683	662	34	walker	walker	PROPN
dad-10683	662	35	et	et	PROPN
dad-10683	662	36	al	al	PROPN
dad-10683	662	37	.	.	PROPN
dad-10683	662	38	,	,	PUNCT
dad-10683	662	39	1997	1997	NUM
dad-10683	662	40	)	)	PUNCT
dad-10683	662	41	and	and	CCONJ
dad-10683	662	42	memo	memo	NOUN
dad-10683	662	43	(	(	PUNCT
dad-10683	662	44	möller	möller	NOUN
dad-10683	662	45	et	et	PROPN
dad-10683	662	46	al	al	PROPN
dad-10683	662	47	.	.	PROPN
dad-10683	662	48	,	,	PUNCT
dad-10683	662	49	2006	2006	NUM
dad-10683	662	50	)	)	PUNCT
dad-10683	662	51	.	.	PUNCT
dad-10683	663	1	an	an	DET
dad-10683	663	2	overview	overview	NOUN
dad-10683	663	3	of	of	ADP
dad-10683	663	4	such	such	ADJ
dad-10683	663	5	methods	method	NOUN
dad-10683	663	6	can	can	AUX
dad-10683	663	7	be	be	AUX
dad-10683	663	8	found	find	VERB
dad-10683	663	9	in	in	ADP
dad-10683	663	10	jokinen	jokinen	PROPN
dad-10683	663	11	and	and	CCONJ
dad-10683	663	12	mctear	mctear	PROPN
dad-10683	663	13	(	(	PUNCT
dad-10683	663	14	2009	2009	NUM
dad-10683	663	15	)	)	PUNCT
dad-10683	663	16	and	and	CCONJ
dad-10683	663	17	hastie	hastie	PROPN
dad-10683	663	18	(	(	PUNCT
dad-10683	663	19	2012	2012	NUM
dad-10683	663	20	)	)	PUNCT
dad-10683	663	21	.	.	PUNCT
dad-10683	664	1	however	however	ADV
dad-10683	664	2	,	,	PUNCT
dad-10683	664	3	in	in	ADP
dad-10683	664	4	the	the	DET
dad-10683	664	5	non	non	ADJ
dad-10683	664	6	-	-	ADJ
dad-10683	664	7	goal	goal	ADJ
dad-10683	664	8	oriented	orient	VERB
dad-10683	664	9	setting	setting	NOUN
dad-10683	664	10	we	we	PRON
dad-10683	664	11	consider	consider	VERB
dad-10683	664	12	here	here	ADV
dad-10683	664	13	,	,	PUNCT
dad-10683	664	14	evaluation	evaluation	NOUN
dad-10683	664	15	is	be	AUX
dad-10683	664	16	more	more	ADV
dad-10683	664	17	difficult	difficult	ADJ
dad-10683	664	18	.	.	PUNCT
dad-10683	665	1	this	this	PRON
dad-10683	665	2	is	be	AUX
dad-10683	665	3	particularly	particularly	ADV
dad-10683	665	4	true	true	ADJ
dad-10683	665	5	for	for	ADP
dad-10683	665	6	end	end	NOUN
dad-10683	665	7	-	-	PUNCT
dad-10683	665	8	to	to	ADP
dad-10683	665	9	-	-	PUNCT
dad-10683	665	10	end	end	NOUN
dad-10683	665	11	systems	system	NOUN
dad-10683	665	12	,	,	PUNCT
dad-10683	665	13	as	as	SCONJ
dad-10683	665	14	there	there	PRON
dad-10683	665	15	is	be	VERB
dad-10683	665	16	no	no	DET
dad-10683	665	17	way	way	NOUN
dad-10683	665	18	to	to	PART
dad-10683	665	19	measure	measure	VERB
dad-10683	665	20	the	the	DET
dad-10683	665	21	accuracy	accuracy	NOUN
dad-10683	665	22	of	of	ADP
dad-10683	665	23	the	the	DET
dad-10683	665	24	state	state	NOUN
dad-10683	665	25	tracking	tracking	NOUN
dad-10683	665	26	module	module	NOUN
dad-10683	665	27	using	use	VERB
dad-10683	665	28	tasks	task	NOUN
dad-10683	665	29	such	such	ADJ
dad-10683	665	30	as	as	ADP
dad-10683	665	31	slot	slot	NOUN
dad-10683	665	32	filling	filling	NOUN
dad-10683	665	33	,	,	PUNCT
dad-10683	665	34	since	since	SCONJ
dad-10683	665	35	they	they	PRON
dad-10683	665	36	are	be	AUX
dad-10683	665	37	not	not	PART
dad-10683	665	38	modular	modular	ADJ
dad-10683	665	39	systems	system	NOUN
dad-10683	665	40	.	.	PUNCT
dad-10683	666	1	indeed	indeed	ADV
dad-10683	666	2	,	,	PUNCT
dad-10683	666	3	there	there	PRON
dad-10683	666	4	58	58	NUM
dad-10683	666	5	training	training	NOUN
dad-10683	666	6	end	end	NOUN
dad-10683	666	7	-	-	PUNCT
dad-10683	666	8	to	to	ADP
dad-10683	666	9	-	-	PUNCT
dad-10683	666	10	end	end	NOUN
dad-10683	666	11	dialogue	dialogue	NOUN
dad-10683	666	12	systems	system	NOUN
dad-10683	666	13	are	be	AUX
dad-10683	666	14	several	several	ADJ
dad-10683	666	15	reasons	reason	NOUN
dad-10683	666	16	for	for	ADP
dad-10683	666	17	wanting	want	VERB
dad-10683	666	18	to	to	PART
dad-10683	666	19	move	move	VERB
dad-10683	666	20	away	away	ADV
dad-10683	666	21	from	from	ADP
dad-10683	666	22	the	the	DET
dad-10683	666	23	slot	slot	NOUN
dad-10683	666	24	filling	fill	VERB
dad-10683	666	25	metrics	metric	NOUN
dad-10683	666	26	that	that	PRON
dad-10683	666	27	have	have	AUX
dad-10683	666	28	become	become	VERB
dad-10683	666	29	common	common	ADJ
dad-10683	666	30	for	for	ADP
dad-10683	666	31	modular	modular	ADJ
dad-10683	666	32	systems	system	NOUN
dad-10683	666	33	.	.	PUNCT
dad-10683	667	1	in	in	ADP
dad-10683	667	2	slot	slot	NOUN
dad-10683	667	3	filling	filling	NOUN
dad-10683	667	4	,	,	PUNCT
dad-10683	667	5	the	the	DET
dad-10683	667	6	set	set	NOUN
dad-10683	667	7	of	of	ADP
dad-10683	667	8	candidate	candidate	NOUN
dad-10683	667	9	outputs	output	NOUN
dad-10683	667	10	(	(	PUNCT
dad-10683	667	11	states	state	NOUN
dad-10683	667	12	)	)	PUNCT
dad-10683	667	13	is	be	AUX
dad-10683	667	14	identified	identify	VERB
dad-10683	667	15	a	a	DET
dad-10683	667	16	priori	priori	NOUN
dad-10683	667	17	through	through	ADP
dad-10683	667	18	knowledge	knowledge	NOUN
dad-10683	667	19	engineering	engineering	NOUN
dad-10683	667	20	,	,	PUNCT
dad-10683	667	21	and	and	CCONJ
dad-10683	667	22	is	be	AUX
dad-10683	667	23	typically	typically	ADV
dad-10683	667	24	rather	rather	ADV
dad-10683	667	25	small	small	ADJ
dad-10683	667	26	in	in	ADP
dad-10683	667	27	comparison	comparison	NOUN
dad-10683	667	28	to	to	ADP
dad-10683	667	29	the	the	DET
dad-10683	667	30	set	set	NOUN
dad-10683	667	31	of	of	ADP
dad-10683	667	32	responses	response	NOUN
dad-10683	667	33	considered	consider	VERB
dad-10683	667	34	by	by	ADP
dad-10683	667	35	nuc	nuc	PROPN
dad-10683	667	36	.	.	PUNCT
dad-10683	667	37	further	far	ADV
dad-10683	667	38	,	,	PUNCT
dad-10683	667	39	it	it	PRON
dad-10683	667	40	has	have	AUX
dad-10683	667	41	been	be	AUX
dad-10683	667	42	speculated	speculate	VERB
dad-10683	667	43	that	that	SCONJ
dad-10683	667	44	state	state	NOUN
dad-10683	667	45	-	-	PUNCT
dad-10683	667	46	of	of	ADP
dad-10683	667	47	-	-	PUNCT
dad-10683	667	48	the	the	DET
dad-10683	667	49	-	-	PUNCT
dad-10683	667	50	art	art	NOUN
dad-10683	667	51	state	state	NOUN
dad-10683	667	52	-	-	PUNCT
dad-10683	667	53	tracking	track	VERB
dad-10683	667	54	models	model	NOUN
dad-10683	667	55	(	(	PUNCT
dad-10683	667	56	henderson	henderson	PROPN
dad-10683	667	57	et	et	PROPN
dad-10683	667	58	al	al	PROPN
dad-10683	667	59	.	.	PROPN
dad-10683	667	60	,	,	PUNCT
dad-10683	667	61	2014b	2014b	NUM
dad-10683	667	62	;	;	PUNCT
dad-10683	667	63	williams	williams	PROPN
dad-10683	667	64	,	,	PUNCT
dad-10683	667	65	2014	2014	NUM
dad-10683	667	66	)	)	PUNCT
dad-10683	667	67	are	be	AUX
dad-10683	667	68	achieving	achieve	VERB
dad-10683	667	69	close	close	ADV
dad-10683	667	70	to	to	ADP
dad-10683	667	71	human	human	ADJ
dad-10683	667	72	-	-	PUNCT
dad-10683	667	73	level	level	NOUN
dad-10683	667	74	performance	performance	NOUN
dad-10683	667	75	.	.	PUNCT
dad-10683	668	1	thus	thus	ADV
dad-10683	668	2	it	it	PRON
dad-10683	668	3	is	be	AUX
dad-10683	668	4	desirable	desirable	ADJ
dad-10683	668	5	to	to	PART
dad-10683	668	6	move	move	VERB
dad-10683	668	7	beyond	beyond	ADP
dad-10683	668	8	this	this	DET
dad-10683	668	9	domain	domain	NOUN
dad-10683	668	10	into	into	ADP
dad-10683	668	11	more	more	ADV
dad-10683	668	12	difficult	difficult	ADJ
dad-10683	668	13	problems	problem	NOUN
dad-10683	668	14	.	.	PUNCT
dad-10683	669	1	to	to	PART
dad-10683	669	2	do	do	VERB
dad-10683	669	3	this	this	PRON
dad-10683	669	4	,	,	PUNCT
dad-10683	669	5	it	it	PRON
dad-10683	669	6	is	be	AUX
dad-10683	669	7	crucial	crucial	ADJ
dad-10683	669	8	to	to	PART
dad-10683	669	9	investigate	investigate	VERB
dad-10683	669	10	ways	way	NOUN
dad-10683	669	11	to	to	PART
dad-10683	669	12	evaluate	evaluate	VERB
dad-10683	669	13	models	model	NOUN
dad-10683	669	14	in	in	ADP
dad-10683	669	15	the	the	DET
dad-10683	669	16	non	non	ADJ
dad-10683	669	17	-	-	ADJ
dad-10683	669	18	goal	goal	ADJ
dad-10683	669	19	oriented	orient	VERB
dad-10683	669	20	setting	setting	NOUN
dad-10683	669	21	that	that	PRON
dad-10683	669	22	do	do	AUX
dad-10683	669	23	not	not	PART
dad-10683	669	24	require	require	VERB
dad-10683	669	25	supervised	supervised	ADJ
dad-10683	669	26	test	test	NOUN
dad-10683	669	27	data	datum	NOUN
dad-10683	669	28	for	for	ADP
dad-10683	669	29	the	the	DET
dad-10683	669	30	internal	internal	ADJ
dad-10683	669	31	modules	module	NOUN
dad-10683	669	32	of	of	ADP
dad-10683	669	33	a	a	DET
dad-10683	669	34	system	system	NOUN
dad-10683	669	35	.	.	PUNCT
dad-10683	670	1	researchers	researcher	NOUN
dad-10683	670	2	have	have	AUX
dad-10683	670	3	previously	previously	ADV
dad-10683	670	4	proposed	propose	VERB
dad-10683	670	5	measuring	measure	VERB
dad-10683	670	6	word	word	NOUN
dad-10683	670	7	perplexity	perplexity	NOUN
dad-10683	670	8	and	and	CCONJ
dad-10683	670	9	word	word	NOUN
dad-10683	670	10	classification	classification	NOUN
dad-10683	670	11	error	error	NOUN
dad-10683	670	12	rate	rate	NOUN
dad-10683	670	13	,	,	PUNCT
dad-10683	670	14	as	as	SCONJ
dad-10683	670	15	these	these	PRON
dad-10683	670	16	are	be	AUX
dad-10683	670	17	widely	widely	ADV
dad-10683	670	18	applied	apply	VERB
dad-10683	670	19	in	in	ADP
dad-10683	670	20	the	the	DET
dad-10683	670	21	language	language	NOUN
dad-10683	670	22	modeling	modeling	NOUN
dad-10683	670	23	and	and	CCONJ
dad-10683	670	24	automatic	automatic	ADJ
dad-10683	670	25	speech	speech	NOUN
dad-10683	670	26	recognition	recognition	NOUN
dad-10683	670	27	community	community	NOUN
dad-10683	670	28	(	(	PUNCT
dad-10683	670	29	serban	serban	PROPN
dad-10683	670	30	et	et	PROPN
dad-10683	670	31	al	al	PROPN
dad-10683	670	32	.	.	PROPN
dad-10683	670	33	,	,	PUNCT
dad-10683	670	34	2016	2016	NUM
dad-10683	670	35	;	;	PUNCT
dad-10683	670	36	vinyals	vinyal	NOUN
dad-10683	670	37	and	and	CCONJ
dad-10683	670	38	le	le	NOUN
dad-10683	670	39	,	,	PUNCT
dad-10683	670	40	2015	2015	NUM
dad-10683	670	41	;	;	PUNCT
dad-10683	670	42	pietquin	pietquin	NOUN
dad-10683	670	43	and	and	CCONJ
dad-10683	670	44	hastie	hastie	PROPN
dad-10683	670	45	,	,	PUNCT
dad-10683	670	46	2013	2013	NUM
dad-10683	670	47	)	)	PUNCT
dad-10683	670	48	.	.	PUNCT
dad-10683	671	1	however	however	ADV
dad-10683	671	2	,	,	PUNCT
dad-10683	671	3	these	these	DET
dad-10683	671	4	metrics	metric	NOUN
dad-10683	671	5	can	can	AUX
dad-10683	671	6	not	not	PART
dad-10683	671	7	be	be	AUX
dad-10683	671	8	computed	compute	VERB
dad-10683	671	9	for	for	ADP
dad-10683	671	10	retrieval	retrieval	NOUN
dad-10683	671	11	models	model	NOUN
dad-10683	671	12	.	.	PUNCT
dad-10683	672	1	researchers	researcher	NOUN
dad-10683	672	2	have	have	AUX
dad-10683	672	3	also	also	ADV
dad-10683	672	4	proposed	propose	VERB
dad-10683	672	5	to	to	PART
dad-10683	672	6	use	use	VERB
dad-10683	672	7	word	word	NOUN
dad-10683	672	8	overlap	overlap	NOUN
dad-10683	672	9	metrics	metric	NOUN
dad-10683	672	10	from	from	ADP
dad-10683	672	11	machine	machine	NOUN
dad-10683	672	12	translation	translation	NOUN
dad-10683	672	13	(	(	PUNCT
dad-10683	672	14	galley	galley	NOUN
dad-10683	672	15	et	et	PROPN
dad-10683	672	16	al	al	PROPN
dad-10683	672	17	.	.	PROPN
dad-10683	672	18	,	,	PUNCT
dad-10683	672	19	2015	2015	NUM
dad-10683	672	20	;	;	PUNCT
dad-10683	672	21	sordoni	sordoni	VERB
dad-10683	672	22	et	et	PROPN
dad-10683	672	23	al	al	PROPN
dad-10683	672	24	.	.	PROPN
dad-10683	672	25	,	,	PUNCT
dad-10683	672	26	2015b	2015b	NUM
dad-10683	672	27	)	)	PUNCT
dad-10683	672	28	.	.	PUNCT
dad-10683	673	1	however	however	ADV
dad-10683	673	2	,	,	PUNCT
dad-10683	673	3	such	such	ADJ
dad-10683	673	4	metrics	metric	NOUN
dad-10683	673	5	based	base	VERB
dad-10683	673	6	on	on	ADP
dad-10683	673	7	word	word	NOUN
dad-10683	673	8	overlaps	overlap	NOUN
dad-10683	673	9	suffer	suffer	VERB
dad-10683	673	10	from	from	ADP
dad-10683	673	11	severe	severe	ADJ
dad-10683	673	12	sparsity	sparsity	NOUN
dad-10683	673	13	issues	issue	NOUN
dad-10683	673	14	,	,	PUNCT
dad-10683	673	15	since	since	SCONJ
dad-10683	673	16	it	it	PRON
dad-10683	673	17	is	be	AUX
dad-10683	673	18	unlikely	unlikely	ADJ
dad-10683	673	19	that	that	SCONJ
dad-10683	673	20	any	any	DET
dad-10683	673	21	sequence	sequence	NOUN
dad-10683	673	22	of	of	ADP
dad-10683	673	23	words	word	NOUN
dad-10683	673	24	will	will	AUX
dad-10683	673	25	be	be	AUX
dad-10683	673	26	identical	identical	ADJ
dad-10683	673	27	in	in	ADP
dad-10683	673	28	both	both	CCONJ
dad-10683	673	29	the	the	DET
dad-10683	673	30	generated	generate	VERB
dad-10683	673	31	and	and	CCONJ
dad-10683	673	32	reference	reference	NOUN
dad-10683	673	33	responses	response	NOUN
dad-10683	673	34	.	.	PUNCT
dad-10683	674	1	these	these	PRON
dad-10683	674	2	have	have	AUX
dad-10683	674	3	shown	show	VERB
dad-10683	674	4	to	to	PART
dad-10683	674	5	correlate	correlate	VERB
dad-10683	674	6	poorly	poorly	ADV
dad-10683	674	7	with	with	ADP
dad-10683	674	8	human	human	ADJ
dad-10683	674	9	judgements	judgement	NOUN
dad-10683	674	10	when	when	SCONJ
dad-10683	674	11	only	only	ADV
dad-10683	674	12	a	a	DET
dad-10683	674	13	single	single	ADJ
dad-10683	674	14	ground	ground	NOUN
dad-10683	674	15	-	-	PUNCT
dad-10683	674	16	truth	truth	NOUN
dad-10683	674	17	response	response	NOUN
dad-10683	674	18	is	be	AUX
dad-10683	674	19	available	available	ADJ
dad-10683	674	20	(	(	PUNCT
dad-10683	674	21	liu	liu	PROPN
dad-10683	674	22	et	et	PROPN
dad-10683	674	23	al	al	PROPN
dad-10683	674	24	.	.	PROPN
dad-10683	674	25	,	,	PUNCT
dad-10683	674	26	2016	2016	NUM
dad-10683	674	27	)	)	PUNCT
dad-10683	674	28	,	,	PUNCT
dad-10683	674	29	and	and	CCONJ
dad-10683	674	30	they	they	PRON
dad-10683	674	31	have	have	VERB
dad-10683	674	32	at	at	ADP
dad-10683	674	33	best	good	ADJ
dad-10683	674	34	a	a	DET
dad-10683	674	35	mediocre	mediocre	ADJ
dad-10683	674	36	correlation	correlation	NOUN
dad-10683	674	37	when	when	SCONJ
dad-10683	674	38	multiple	multiple	ADJ
dad-10683	674	39	ground	ground	NOUN
dad-10683	674	40	-	-	PUNCT
dad-10683	674	41	truth	truth	NOUN
dad-10683	674	42	responses	response	NOUN
dad-10683	674	43	are	be	AUX
dad-10683	674	44	available	available	ADJ
dad-10683	674	45	(	(	PUNCT
dad-10683	674	46	galley	galley	NOUN
dad-10683	674	47	et	et	PROPN
dad-10683	674	48	al	al	PROPN
dad-10683	674	49	.	.	PROPN
dad-10683	674	50	,	,	PUNCT
dad-10683	674	51	2015	2015	NUM
dad-10683	674	52	)	)	PUNCT
dad-10683	674	53	.	.	PUNCT
dad-10683	675	1	furthermore	furthermore	ADV
dad-10683	675	2	,	,	PUNCT
dad-10683	675	3	it	it	PRON
dad-10683	675	4	has	have	AUX
dad-10683	675	5	been	be	AUX
dad-10683	675	6	argued	argue	VERB
dad-10683	675	7	that	that	SCONJ
dad-10683	675	8	such	such	ADJ
dad-10683	675	9	metrics	metric	NOUN
dad-10683	675	10	mainly	mainly	ADV
dad-10683	675	11	focus	focus	VERB
dad-10683	675	12	on	on	ADP
dad-10683	675	13	pronouns	pronoun	NOUN
dad-10683	675	14	and	and	CCONJ
dad-10683	675	15	punctuation	punctuation	NOUN
dad-10683	675	16	marks	mark	NOUN
dad-10683	675	17	when	when	SCONJ
dad-10683	675	18	applied	apply	VERB
dad-10683	675	19	to	to	ADP
dad-10683	675	20	non	non	ADJ
dad-10683	675	21	-	-	ADJ
dad-10683	675	22	task	task	ADJ
dad-10683	675	23	oriented	orient	VERB
dad-10683	675	24	dialogue	dialogue	NOUN
dad-10683	675	25	datasets	dataset	NOUN
dad-10683	675	26	(	(	PUNCT
dad-10683	675	27	serban	serban	PROPN
dad-10683	675	28	et	et	PROPN
dad-10683	675	29	al	al	PROPN
dad-10683	675	30	.	.	PROPN
dad-10683	675	31	,	,	PUNCT
dad-10683	675	32	2016	2016	NUM
dad-10683	675	33	)	)	PUNCT
dad-10683	675	34	.	.	PUNCT
dad-10683	676	1	while	while	SCONJ
dad-10683	676	2	the	the	DET
dad-10683	676	3	word	word	NOUN
dad-10683	676	4	embedding	embed	VERB
dad-10683	676	5	metrics	metric	NOUN
dad-10683	676	6	used	use	VERB
dad-10683	676	7	here	here	ADV
dad-10683	676	8	do	do	AUX
dad-10683	676	9	not	not	PART
dad-10683	676	10	have	have	VERB
dad-10683	676	11	a	a	DET
dad-10683	676	12	strong	strong	ADJ
dad-10683	676	13	correlation	correlation	NOUN
dad-10683	676	14	with	with	ADP
dad-10683	676	15	human	human	ADJ
dad-10683	676	16	judgement	judgement	NOUN
dad-10683	676	17	,	,	PUNCT
dad-10683	676	18	they	they	PRON
dad-10683	676	19	have	have	VERB
dad-10683	676	20	an	an	DET
dad-10683	676	21	additional	additional	ADJ
dad-10683	676	22	interpretation	interpretation	NOUN
dad-10683	676	23	of	of	ADP
dad-10683	676	24	measuring	measure	VERB
dad-10683	676	25	the	the	DET
dad-10683	676	26	semantic	semantic	ADJ
dad-10683	676	27	similarity	similarity	NOUN
dad-10683	676	28	between	between	ADP
dad-10683	676	29	the	the	DET
dad-10683	676	30	generated	generate	VERB
dad-10683	676	31	and	and	CCONJ
dad-10683	676	32	reference	reference	NOUN
dad-10683	676	33	responses	response	NOUN
dad-10683	676	34	(	(	PUNCT
dad-10683	676	35	as	as	SCONJ
dad-10683	676	36	argued	argue	VERB
dad-10683	676	37	in	in	ADP
dad-10683	676	38	section	section	NOUN
dad-10683	676	39	4.4	4.4	NUM
dad-10683	676	40	)	)	PUNCT
dad-10683	676	41	,	,	PUNCT
dad-10683	676	42	which	which	PRON
dad-10683	676	43	is	be	AUX
dad-10683	676	44	why	why	SCONJ
dad-10683	676	45	we	we	PRON
dad-10683	676	46	favour	favour	VERB
dad-10683	676	47	them	they	PRON
dad-10683	676	48	over	over	ADP
dad-10683	676	49	word	word	NOUN
dad-10683	676	50	-	-	PUNCT
dad-10683	676	51	overlap	overlap	NOUN
dad-10683	676	52	scores	score	NOUN
dad-10683	676	53	such	such	ADJ
dad-10683	676	54	as	as	ADP
dad-10683	676	55	bleu	bleu	NOUN
dad-10683	676	56	.	.	PUNCT
dad-10683	677	1	however	however	ADV
dad-10683	677	2	,	,	PUNCT
dad-10683	677	3	we	we	PRON
dad-10683	677	4	reiterate	reiterate	VERB
dad-10683	677	5	that	that	SCONJ
dad-10683	677	6	none	none	NOUN
dad-10683	677	7	of	of	ADP
dad-10683	677	8	these	these	DET
dad-10683	677	9	metrics	metric	NOUN
dad-10683	677	10	measures	measure	VERB
dad-10683	677	11	the	the	DET
dad-10683	677	12	coherence	coherence	NOUN
dad-10683	677	13	of	of	ADP
dad-10683	677	14	the	the	DET
dad-10683	677	15	generated	generate	VERB
dad-10683	677	16	responses	response	NOUN
dad-10683	677	17	,	,	PUNCT
dad-10683	677	18	and	and	CCONJ
dad-10683	677	19	this	this	PRON
dad-10683	677	20	remains	remain	VERB
dad-10683	677	21	an	an	DET
dad-10683	677	22	important	important	ADJ
dad-10683	677	23	direction	direction	NOUN
dad-10683	677	24	for	for	ADP
dad-10683	677	25	future	future	ADJ
dad-10683	677	26	work	work	NOUN
dad-10683	677	27	.	.	PUNCT
dad-10683	678	1	another	another	DET
dad-10683	678	2	option	option	NOUN
dad-10683	678	3	for	for	ADP
dad-10683	678	4	evaluating	evaluate	VERB
dad-10683	678	5	dialogue	dialogue	NOUN
dad-10683	678	6	systems	system	NOUN
dad-10683	678	7	trained	train	VERB
dad-10683	678	8	in	in	ADP
dad-10683	678	9	an	an	DET
dad-10683	678	10	end	end	NOUN
dad-10683	678	11	-	-	PUNCT
dad-10683	678	12	to	to	ADP
dad-10683	678	13	-	-	PUNCT
dad-10683	678	14	end	end	NOUN
dad-10683	678	15	manner	manner	NOUN
dad-10683	678	16	is	be	AUX
dad-10683	678	17	using	use	VERB
dad-10683	678	18	an	an	DET
dad-10683	678	19	alternative	alternative	ADJ
dad-10683	678	20	task	task	NOUN
dad-10683	678	21	such	such	ADJ
dad-10683	678	22	as	as	ADP
dad-10683	678	23	next	next	ADJ
dad-10683	678	24	utterance	utterance	X
dad-10683	678	25	classification	classification	NOUN
dad-10683	678	26	.	.	PUNCT
dad-10683	679	1	while	while	SCONJ
dad-10683	679	2	this	this	PRON
dad-10683	679	3	does	do	AUX
dad-10683	679	4	not	not	PART
dad-10683	679	5	directly	directly	ADV
dad-10683	679	6	compare	compare	VERB
dad-10683	679	7	the	the	DET
dad-10683	679	8	generated	generate	VERB
dad-10683	679	9	response	response	NOUN
dad-10683	679	10	of	of	ADP
dad-10683	679	11	the	the	DET
dad-10683	679	12	system	system	NOUN
dad-10683	679	13	to	to	ADP
dad-10683	679	14	the	the	DET
dad-10683	679	15	ground	ground	NOUN
dad-10683	679	16	-	-	PUNCT
dad-10683	679	17	truth	truth	NOUN
dad-10683	679	18	response	response	NOUN
dad-10683	679	19	,	,	PUNCT
dad-10683	679	20	there	there	PRON
dad-10683	679	21	are	be	VERB
dad-10683	679	22	several	several	ADJ
dad-10683	679	23	reasons	reason	NOUN
dad-10683	679	24	for	for	ADP
dad-10683	679	25	preferring	prefer	VERB
dad-10683	679	26	the	the	DET
dad-10683	679	27	recall	recall	NOUN
dad-10683	679	28	metric	metric	NOUN
dad-10683	679	29	:	:	PUNCT
dad-10683	680	1	1	1	X
dad-10683	680	2	.	.	X
dad-10683	680	3	it	it	PRON
dad-10683	680	4	is	be	AUX
dad-10683	680	5	a	a	DET
dad-10683	680	6	more	more	ADV
dad-10683	680	7	difficult	difficult	ADJ
dad-10683	680	8	task	task	NOUN
dad-10683	680	9	than	than	ADP
dad-10683	680	10	slot	slot	NOUN
dad-10683	680	11	filling	filling	NOUN
dad-10683	680	12	,	,	PUNCT
dad-10683	680	13	and	and	CCONJ
dad-10683	680	14	thus	thus	ADV
dad-10683	680	15	will	will	AUX
dad-10683	680	16	require	require	VERB
dad-10683	680	17	further	further	ADJ
dad-10683	680	18	development	development	NOUN
dad-10683	680	19	of	of	ADP
dad-10683	680	20	more	more	ADV
dad-10683	680	21	sophisticated	sophisticated	ADJ
dad-10683	680	22	dialogue	dialogue	NOUN
dad-10683	680	23	systems	system	NOUN
dad-10683	680	24	in	in	ADP
dad-10683	680	25	order	order	NOUN
dad-10683	680	26	to	to	PART
dad-10683	680	27	solve	solve	VERB
dad-10683	680	28	the	the	DET
dad-10683	680	29	task	task	NOUN
dad-10683	680	30	.	.	PUNCT
dad-10683	681	1	2	2	X
dad-10683	681	2	.	.	X
dad-10683	681	3	it	it	PRON
dad-10683	681	4	does	do	AUX
dad-10683	681	5	not	not	PART
dad-10683	681	6	suffer	suffer	VERB
dad-10683	681	7	from	from	ADP
dad-10683	681	8	the	the	DET
dad-10683	681	9	same	same	ADJ
dad-10683	681	10	problems	problem	NOUN
dad-10683	681	11	as	as	ADP
dad-10683	681	12	the	the	DET
dad-10683	681	13	word	word	NOUN
dad-10683	681	14	overlap	overlap	NOUN
dad-10683	681	15	metrics	metric	NOUN
dad-10683	681	16	,	,	PUNCT
dad-10683	681	17	as	as	SCONJ
dad-10683	681	18	it	it	PRON
dad-10683	681	19	does	do	AUX
dad-10683	681	20	not	not	PART
dad-10683	681	21	have	have	VERB
dad-10683	681	22	to	to	PART
dad-10683	681	23	directly	directly	ADV
dad-10683	681	24	compare	compare	VERB
dad-10683	681	25	the	the	DET
dad-10683	681	26	quality	quality	NOUN
dad-10683	681	27	of	of	ADP
dad-10683	681	28	a	a	DET
dad-10683	681	29	generated	generate	VERB
dad-10683	681	30	response	response	NOUN
dad-10683	681	31	to	to	ADP
dad-10683	681	32	the	the	DET
dad-10683	681	33	ground	ground	NOUN
dad-10683	681	34	-	-	PUNCT
dad-10683	681	35	truth	truth	NOUN
dad-10683	681	36	response	response	NOUN
dad-10683	681	37	,	,	PUNCT
dad-10683	681	38	an	an	DET
dad-10683	681	39	inherently	inherently	ADV
dad-10683	681	40	noisy	noisy	ADJ
dad-10683	681	41	process	process	NOUN
dad-10683	681	42	.	.	PUNCT
dad-10683	682	1	instead	instead	ADV
dad-10683	682	2	,	,	PUNCT
dad-10683	682	3	it	it	PRON
dad-10683	682	4	measures	measure	VERB
dad-10683	682	5	the	the	DET
dad-10683	682	6	model	model	NOUN
dad-10683	682	7	’s	’s	PART
dad-10683	682	8	capacity	capacity	NOUN
dad-10683	682	9	to	to	PART
dad-10683	682	10	pick	pick	VERB
dad-10683	682	11	out	out	ADP
dad-10683	682	12	the	the	DET
dad-10683	682	13	correct	correct	ADJ
dad-10683	682	14	response	response	NOUN
dad-10683	682	15	from	from	ADP
dad-10683	682	16	a	a	DET
dad-10683	682	17	list	list	NOUN
dad-10683	682	18	of	of	ADP
dad-10683	682	19	responses	response	NOUN
dad-10683	682	20	.	.	PUNCT
dad-10683	683	1	3	3	X
dad-10683	683	2	.	.	X
dad-10683	683	3	performance	performance	NOUN
dad-10683	683	4	using	use	VERB
dad-10683	683	5	the	the	DET
dad-10683	683	6	recall	recall	NOUN
dad-10683	683	7	metrics	metric	NOUN
dad-10683	683	8	is	be	AUX
dad-10683	683	9	easily	easily	ADV
dad-10683	683	10	interpretable	interpretable	ADJ
dad-10683	683	11	,	,	PUNCT
dad-10683	683	12	and	and	CCONJ
dad-10683	683	13	can	can	AUX
dad-10683	683	14	easily	easily	ADV
dad-10683	683	15	be	be	AUX
dad-10683	683	16	compared	compare	VERB
dad-10683	683	17	to	to	ADP
dad-10683	683	18	human	human	ADJ
dad-10683	683	19	performance	performance	NOUN
dad-10683	683	20	.	.	PUNCT
dad-10683	684	1	indeed	indeed	ADV
dad-10683	684	2	,	,	PUNCT
dad-10683	684	3	this	this	PRON
dad-10683	684	4	has	have	AUX
dad-10683	684	5	been	be	AUX
dad-10683	684	6	recently	recently	ADV
dad-10683	684	7	done	do	VERB
dad-10683	684	8	by	by	ADP
dad-10683	684	9	lowe	lowe	PROPN
dad-10683	684	10	et	et	PROPN
dad-10683	684	11	al	al	PROPN
dad-10683	684	12	.	.	PUNCT
dad-10683	685	1	(	(	PUNCT
dad-10683	685	2	2016	2016	NUM
dad-10683	685	3	)	)	PUNCT
dad-10683	685	4	,	,	PUNCT
dad-10683	685	5	who	who	PRON
dad-10683	685	6	show	show	VERB
dad-10683	685	7	that	that	SCONJ
dad-10683	685	8	human	human	ADJ
dad-10683	685	9	performance	performance	NOUN
dad-10683	685	10	on	on	ADP
dad-10683	685	11	this	this	DET
dad-10683	685	12	task	task	NOUN
dad-10683	685	13	is	be	AUX
dad-10683	685	14	above	above	ADP
dad-10683	685	15	the	the	DET
dad-10683	685	16	performance	performance	NOUN
dad-10683	685	17	for	for	ADP
dad-10683	685	18	the	the	DET
dad-10683	685	19	dual	dual	ADJ
dad-10683	685	20	encoder	encoder	NOUN
dad-10683	685	21	model	model	NOUN
dad-10683	685	22	on	on	ADP
dad-10683	685	23	the	the	DET
dad-10683	685	24	ubuntu	ubuntu	ADJ
dad-10683	685	25	dialogue	dialogue	NOUN
dad-10683	685	26	corpus	corpus	NOUN
dad-10683	685	27	,	,	PUNCT
dad-10683	685	28	as	as	ADV
dad-10683	685	29	well	well	ADV
dad-10683	685	30	as	as	ADP
dad-10683	685	31	on	on	ADP
dad-10683	685	32	movie	movie	NOUN
dad-10683	685	33	and	and	CCONJ
dad-10683	685	34	twitter	twitter	NOUN
dad-10683	685	35	corpora	corpora	PROPN
dad-10683	685	36	.	.	PUNCT
dad-10683	686	1	thus	thus	ADV
dad-10683	686	2	,	,	PUNCT
dad-10683	686	3	there	there	PRON
dad-10683	686	4	is	be	VERB
dad-10683	686	5	room	room	NOUN
dad-10683	686	6	for	for	ADP
dad-10683	686	7	improvement	improvement	NOUN
dad-10683	686	8	for	for	ADP
dad-10683	686	9	models	model	NOUN
dad-10683	686	10	on	on	ADP
dad-10683	686	11	this	this	DET
dad-10683	686	12	task	task	NOUN
dad-10683	686	13	.	.	PUNCT
dad-10683	687	1	4	4	X
dad-10683	687	2	.	.	X
dad-10683	687	3	the	the	DET
dad-10683	687	4	task	task	NOUN
dad-10683	687	5	is	be	AUX
dad-10683	687	6	consistent	consistent	ADJ
dad-10683	687	7	with	with	ADP
dad-10683	687	8	the	the	DET
dad-10683	687	9	end	end	NOUN
dad-10683	687	10	goal	goal	NOUN
dad-10683	687	11	of	of	ADP
dad-10683	687	12	building	build	VERB
dad-10683	687	13	dialogue	dialogue	NOUN
dad-10683	687	14	systems	system	NOUN
dad-10683	687	15	that	that	PRON
dad-10683	687	16	can	can	AUX
dad-10683	687	17	converse	converse	VERB
dad-10683	687	18	naturally	naturally	ADV
dad-10683	687	19	with	with	ADP
dad-10683	687	20	humans	human	NOUN
dad-10683	687	21	.	.	PUNCT
dad-10683	688	1	more	more	ADV
dad-10683	688	2	precisely	precisely	ADV
dad-10683	688	3	,	,	PUNCT
dad-10683	688	4	models	model	NOUN
dad-10683	688	5	that	that	PRON
dad-10683	688	6	are	be	AUX
dad-10683	688	7	able	able	ADJ
dad-10683	688	8	to	to	PART
dad-10683	688	9	generate	generate	VERB
dad-10683	688	10	good	good	ADJ
dad-10683	688	11	responses	response	NOUN
dad-10683	688	12	should	should	AUX
dad-10683	688	13	also	also	ADV
dad-10683	688	14	be	be	AUX
dad-10683	688	15	able	able	ADJ
dad-10683	688	16	to	to	PART
dad-10683	688	17	pick	pick	VERB
dad-10683	688	18	good	good	ADJ
dad-10683	688	19	responses	response	NOUN
dad-10683	688	20	from	from	ADP
dad-10683	688	21	a	a	DET
dad-10683	688	22	list	list	NOUN
dad-10683	688	23	of	of	ADP
dad-10683	688	24	candidates	candidate	NOUN
dad-10683	688	25	,	,	PUNCT
dad-10683	688	26	as	as	ADP
dad-10683	688	27	in	in	ADP
dad-10683	688	28	nuc	nuc	PROPN
dad-10683	688	29	.	.	PROPN
dad-10683	688	30	59	59	NUM
dad-10683	688	31	lowe	lowe	PROPN
dad-10683	688	32	,	,	PUNCT
dad-10683	688	33	pow	pow	PROPN
dad-10683	688	34	,	,	PUNCT
dad-10683	688	35	serban	serban	PROPN
dad-10683	688	36	,	,	PUNCT
dad-10683	688	37	charlinn	charlinn	PROPN
dad-10683	688	38	,	,	PUNCT
dad-10683	688	39	liu	liu	PROPN
dad-10683	688	40	and	and	CCONJ
dad-10683	688	41	pineau	pineau	PROPN
dad-10683	688	42	5	5	NUM
dad-10683	688	43	.	.	PUNCT
dad-10683	689	1	it	it	PRON
dad-10683	689	2	is	be	AUX
dad-10683	689	3	easy	easy	ADJ
dad-10683	689	4	to	to	PART
dad-10683	689	5	alter	alter	VERB
dad-10683	689	6	the	the	DET
dad-10683	689	7	task	task	NOUN
dad-10683	689	8	difficulty	difficulty	NOUN
dad-10683	689	9	in	in	ADP
dad-10683	689	10	a	a	DET
dad-10683	689	11	controlled	control	VERB
dad-10683	689	12	manner	manner	NOUN
dad-10683	689	13	.	.	PUNCT
dad-10683	690	1	we	we	PRON
dad-10683	690	2	demonstrated	demonstrate	VERB
dad-10683	690	3	this	this	PRON
dad-10683	690	4	by	by	ADP
dad-10683	690	5	moving	move	VERB
dad-10683	690	6	from	from	ADP
dad-10683	690	7	1	1	NUM
dad-10683	690	8	to	to	ADP
dad-10683	690	9	9	9	NUM
dad-10683	690	10	false	false	ADJ
dad-10683	690	11	responses	response	NOUN
dad-10683	690	12	,	,	PUNCT
dad-10683	690	13	and	and	CCONJ
dad-10683	690	14	by	by	ADP
dad-10683	690	15	varying	vary	VERB
dad-10683	690	16	the	the	DET
dad-10683	690	17	recall@k	recall@k	NOUN
dad-10683	690	18	parameter	parameter	NOUN
dad-10683	690	19	.	.	PUNCT
dad-10683	691	1	in	in	ADP
dad-10683	691	2	the	the	DET
dad-10683	691	3	future	future	NOUN
dad-10683	691	4	,	,	PUNCT
dad-10683	691	5	instead	instead	ADV
dad-10683	691	6	of	of	ADP
dad-10683	691	7	choosing	choose	VERB
dad-10683	691	8	false	false	ADJ
dad-10683	691	9	responses	response	NOUN
dad-10683	691	10	randomly	randomly	ADV
dad-10683	691	11	,	,	PUNCT
dad-10683	691	12	one	one	PRON
dad-10683	691	13	could	could	AUX
dad-10683	691	14	consider	consider	VERB
dad-10683	691	15	selecting	select	VERB
dad-10683	691	16	false	false	ADJ
dad-10683	691	17	responses	response	NOUN
dad-10683	691	18	that	that	PRON
dad-10683	691	19	are	be	AUX
dad-10683	691	20	similar	similar	ADJ
dad-10683	691	21	to	to	ADP
dad-10683	691	22	the	the	DET
dad-10683	691	23	actual	actual	ADJ
dad-10683	691	24	response	response	NOUN
dad-10683	691	25	(	(	PUNCT
dad-10683	691	26	e.g.	e.g.	ADV
dad-10683	691	27	as	as	SCONJ
dad-10683	691	28	measured	measure	VERB
dad-10683	691	29	by	by	ADP
dad-10683	691	30	tf	tf	PROPN
dad-10683	691	31	-	-	PUNCT
dad-10683	691	32	idf	idf	PROPN
dad-10683	691	33	cosine	cosine	NOUN
dad-10683	691	34	similarity	similarity	PROPN
dad-10683	691	35	)	)	PUNCT
dad-10683	691	36	.	.	PUNCT
dad-10683	692	1	a	a	DET
dad-10683	692	2	dialogue	dialogue	NOUN
dad-10683	692	3	model	model	NOUN
dad-10683	692	4	that	that	PRON
dad-10683	692	5	performs	perform	VERB
dad-10683	692	6	well	well	ADV
dad-10683	692	7	on	on	ADP
dad-10683	692	8	this	this	DET
dad-10683	692	9	more	more	ADV
dad-10683	692	10	difficult	difficult	ADJ
dad-10683	692	11	task	task	NOUN
dad-10683	692	12	should	should	AUX
dad-10683	692	13	also	also	ADV
dad-10683	692	14	manage	manage	VERB
dad-10683	692	15	to	to	PART
dad-10683	692	16	capture	capture	VERB
dad-10683	692	17	a	a	DET
dad-10683	692	18	more	more	ADV
dad-10683	692	19	fine	fine	ADV
dad-10683	692	20	-	-	PUNCT
dad-10683	692	21	grained	grain	VERB
dad-10683	692	22	semantic	semantic	ADJ
dad-10683	692	23	meaning	meaning	NOUN
dad-10683	692	24	of	of	ADP
dad-10683	692	25	sentences	sentence	NOUN
dad-10683	692	26	,	,	PUNCT
dad-10683	692	27	as	as	SCONJ
dad-10683	692	28	compared	compare	VERB
dad-10683	692	29	to	to	ADP
dad-10683	692	30	a	a	DET
dad-10683	692	31	model	model	NOUN
dad-10683	692	32	that	that	PRON
dad-10683	692	33	naively	naively	ADV
dad-10683	692	34	picks	pick	VERB
dad-10683	692	35	replies	reply	NOUN
dad-10683	692	36	with	with	ADP
dad-10683	692	37	the	the	DET
dad-10683	692	38	most	most	ADJ
dad-10683	692	39	words	word	NOUN
dad-10683	692	40	in	in	ADP
dad-10683	692	41	common	common	ADJ
dad-10683	692	42	with	with	ADP
dad-10683	692	43	the	the	DET
dad-10683	692	44	context	context	NOUN
dad-10683	692	45	such	such	ADJ
dad-10683	692	46	as	as	ADP
dad-10683	692	47	tf	tf	PROPN
dad-10683	692	48	-	-	PUNCT
dad-10683	692	49	idf	idf	PROPN
dad-10683	692	50	.	.	PUNCT
dad-10683	693	1	in	in	ADP
dad-10683	693	2	fact	fact	NOUN
dad-10683	693	3	,	,	PUNCT
dad-10683	693	4	when	when	SCONJ
dad-10683	693	5	the	the	DET
dad-10683	693	6	set	set	NOUN
dad-10683	693	7	of	of	ADP
dad-10683	693	8	candidate	candidate	NOUN
dad-10683	693	9	responses	response	NOUN
dad-10683	693	10	for	for	ADP
dad-10683	693	11	the	the	DET
dad-10683	693	12	model	model	NOUN
dad-10683	693	13	to	to	PART
dad-10683	693	14	choose	choose	VERB
dad-10683	693	15	from	from	ADP
dad-10683	693	16	is	be	AUX
dad-10683	693	17	close	close	ADJ
dad-10683	693	18	to	to	ADP
dad-10683	693	19	the	the	DET
dad-10683	693	20	size	size	NOUN
dad-10683	693	21	of	of	ADP
dad-10683	693	22	the	the	DET
dad-10683	693	23	dataset	dataset	NOUN
dad-10683	693	24	(	(	PUNCT
dad-10683	693	25	e.g.	e.g.	ADV
dad-10683	693	26	all	all	DET
dad-10683	693	27	utterances	utterance	NOUN
dad-10683	693	28	ever	ever	ADV
dad-10683	693	29	recorded	record	VERB
dad-10683	693	30	)	)	PUNCT
dad-10683	693	31	,	,	PUNCT
dad-10683	693	32	then	then	ADV
dad-10683	693	33	nuc	nuc	PROPN
dad-10683	693	34	becomes	become	VERB
dad-10683	693	35	close	close	ADJ
dad-10683	693	36	to	to	ADP
dad-10683	693	37	the	the	DET
dad-10683	693	38	response	response	NOUN
dad-10683	693	39	generation	generation	NOUN
dad-10683	693	40	case	case	NOUN
dad-10683	693	41	.	.	PUNCT
dad-10683	694	1	given	give	VERB
dad-10683	694	2	the	the	DET
dad-10683	694	3	above	above	ADJ
dad-10683	694	4	points	point	NOUN
dad-10683	694	5	,	,	PUNCT
dad-10683	694	6	we	we	PRON
dad-10683	694	7	believe	believe	VERB
dad-10683	694	8	that	that	SCONJ
dad-10683	694	9	evaluating	evaluate	VERB
dad-10683	694	10	models	model	NOUN
dad-10683	694	11	with	with	ADP
dad-10683	694	12	the	the	DET
dad-10683	694	13	nuc	nuc	PROPN
dad-10683	694	14	task	task	NOUN
dad-10683	694	15	is	be	AUX
dad-10683	694	16	very	very	ADV
dad-10683	694	17	useful	useful	ADJ
dad-10683	694	18	for	for	ADP
dad-10683	694	19	the	the	DET
dad-10683	694	20	time	time	NOUN
dad-10683	694	21	being	be	AUX
dad-10683	694	22	.	.	PUNCT
dad-10683	695	1	however	however	ADV
dad-10683	695	2	,	,	PUNCT
dad-10683	695	3	we	we	PRON
dad-10683	695	4	believe	believe	VERB
dad-10683	695	5	that	that	SCONJ
dad-10683	695	6	caution	caution	NOUN
dad-10683	695	7	should	should	AUX
dad-10683	695	8	be	be	AUX
dad-10683	695	9	used	use	VERB
dad-10683	695	10	when	when	SCONJ
dad-10683	695	11	comparing	compare	VERB
dad-10683	695	12	retrieval	retrieval	NOUN
dad-10683	695	13	models	model	NOUN
dad-10683	695	14	to	to	ADP
dad-10683	695	15	generative	generative	ADJ
dad-10683	695	16	models	model	NOUN
dad-10683	695	17	using	use	VERB
dad-10683	695	18	nuc	nuc	NOUN
dad-10683	695	19	,	,	PUNCT
dad-10683	695	20	as	as	SCONJ
dad-10683	695	21	the	the	DET
dad-10683	695	22	retrieval	retrieval	NOUN
dad-10683	695	23	models	model	NOUN
dad-10683	695	24	are	be	AUX
dad-10683	695	25	directly	directly	ADV
dad-10683	695	26	trained	train	VERB
dad-10683	695	27	on	on	ADP
dad-10683	695	28	the	the	DET
dad-10683	695	29	task	task	NOUN
dad-10683	695	30	of	of	ADP
dad-10683	695	31	nuc	nuc	PROPN
dad-10683	695	32	,	,	PUNCT
dad-10683	695	33	rendering	render	VERB
dad-10683	695	34	it	it	PRON
dad-10683	695	35	an	an	DET
dad-10683	695	36	unfair	unfair	ADJ
dad-10683	695	37	comparison	comparison	NOUN
dad-10683	695	38	.	.	PUNCT
dad-10683	696	1	6.4	6.4	NUM
dad-10683	696	2	future	future	ADJ
dad-10683	696	3	research	research	NOUN
dad-10683	696	4	directions	direction	NOUN
dad-10683	696	5	for	for	ADP
dad-10683	696	6	end	end	NOUN
dad-10683	696	7	-	-	PUNCT
dad-10683	696	8	to	to	ADP
dad-10683	696	9	-	-	PUNCT
dad-10683	696	10	end	end	NOUN
dad-10683	696	11	systems	system	NOUN
dad-10683	696	12	given	give	VERB
dad-10683	696	13	the	the	DET
dad-10683	696	14	analysis	analysis	NOUN
dad-10683	696	15	performed	perform	VERB
dad-10683	696	16	in	in	ADP
dad-10683	696	17	section	section	NOUN
dad-10683	696	18	4.6	4.6	NUM
dad-10683	696	19	,	,	PUNCT
dad-10683	696	20	we	we	PRON
dad-10683	696	21	postulate	postulate	VERB
dad-10683	696	22	several	several	ADJ
dad-10683	696	23	interesting	interesting	ADJ
dad-10683	696	24	directions	direction	NOUN
dad-10683	696	25	for	for	ADP
dad-10683	696	26	future	future	ADJ
dad-10683	696	27	research	research	NOUN
dad-10683	696	28	on	on	ADP
dad-10683	696	29	end	end	NOUN
dad-10683	696	30	-	-	PUNCT
dad-10683	696	31	to	to	ADP
dad-10683	696	32	-	-	PUNCT
dad-10683	696	33	end	end	NOUN
dad-10683	696	34	dialogue	dialogue	NOUN
dad-10683	696	35	systems	system	NOUN
dad-10683	696	36	,	,	PUNCT
dad-10683	696	37	particularly	particularly	ADV
dad-10683	696	38	on	on	ADP
dad-10683	696	39	the	the	DET
dad-10683	696	40	ubuntu	ubuntu	ADJ
dad-10683	696	41	dialogue	dialogue	NOUN
dad-10683	696	42	corpus	corpus	PROPN
dad-10683	696	43	.	.	PUNCT
dad-10683	697	1	an	an	DET
dad-10683	697	2	important	important	ADJ
dad-10683	697	3	challenge	challenge	NOUN
dad-10683	697	4	in	in	ADP
dad-10683	697	5	dialogue	dialogue	NOUN
dad-10683	697	6	systems	system	NOUN
dad-10683	697	7	is	be	AUX
dad-10683	697	8	the	the	DET
dad-10683	697	9	ability	ability	NOUN
dad-10683	697	10	to	to	PART
dad-10683	697	11	understand	understand	VERB
dad-10683	697	12	the	the	DET
dad-10683	697	13	turn	turn	NOUN
dad-10683	697	14	-	-	PUNCT
dad-10683	697	15	taking	take	VERB
dad-10683	697	16	structure	structure	NOUN
dad-10683	697	17	of	of	ADP
dad-10683	697	18	dialogue	dialogue	NOUN
dad-10683	697	19	.	.	PUNCT
dad-10683	698	1	this	this	PRON
dad-10683	698	2	is	be	AUX
dad-10683	698	3	a	a	DET
dad-10683	698	4	significant	significant	ADJ
dad-10683	698	5	source	source	NOUN
dad-10683	698	6	of	of	ADP
dad-10683	698	7	errors	error	NOUN
dad-10683	698	8	for	for	ADP
dad-10683	698	9	the	the	DET
dad-10683	698	10	dual	dual	ADJ
dad-10683	698	11	encoder	encoder	NOUN
dad-10683	698	12	model	model	NOUN
dad-10683	698	13	.	.	PUNCT
dad-10683	699	1	some	some	DET
dad-10683	699	2	progress	progress	NOUN
dad-10683	699	3	in	in	ADP
dad-10683	699	4	this	this	DET
dad-10683	699	5	direction	direction	NOUN
dad-10683	699	6	has	have	AUX
dad-10683	699	7	been	be	AUX
dad-10683	699	8	made	make	VERB
dad-10683	699	9	for	for	ADP
dad-10683	699	10	end	end	NOUN
dad-10683	699	11	-	-	PUNCT
dad-10683	699	12	to	to	ADP
dad-10683	699	13	-	-	PUNCT
dad-10683	699	14	end	end	NOUN
dad-10683	699	15	dialogue	dialogue	NOUN
dad-10683	699	16	systems	system	NOUN
dad-10683	699	17	(	(	PUNCT
dad-10683	699	18	luan	luan	PROPN
dad-10683	699	19	et	et	PROPN
dad-10683	699	20	al	al	PROPN
dad-10683	699	21	.	.	PROPN
dad-10683	699	22	,	,	PUNCT
dad-10683	699	23	2016	2016	NUM
dad-10683	699	24	;	;	PUNCT
dad-10683	699	25	li	li	PROPN
dad-10683	699	26	et	et	PROPN
dad-10683	699	27	al	al	PROPN
dad-10683	699	28	.	.	PROPN
dad-10683	699	29	,	,	PUNCT
dad-10683	699	30	2016a	2016a	NUM
dad-10683	699	31	)	)	PUNCT
dad-10683	699	32	,	,	PUNCT
dad-10683	699	33	using	use	VERB
dad-10683	699	34	approaches	approach	NOUN
dad-10683	699	35	derived	derive	VERB
dad-10683	699	36	from	from	ADP
dad-10683	699	37	topic	topic	NOUN
dad-10683	699	38	modelling	modelling	NOUN
dad-10683	699	39	or	or	CCONJ
dad-10683	699	40	by	by	ADP
dad-10683	699	41	explicitly	explicitly	ADV
dad-10683	699	42	modelling	model	VERB
dad-10683	699	43	each	each	DET
dad-10683	699	44	user	user	NOUN
dad-10683	699	45	with	with	ADP
dad-10683	699	46	a	a	DET
dad-10683	699	47	continuousvalued	continuousvalue	VERB
dad-10683	699	48	vector	vector	NOUN
dad-10683	699	49	.	.	PUNCT
dad-10683	700	1	however	however	ADV
dad-10683	700	2	,	,	PUNCT
dad-10683	700	3	this	this	PRON
dad-10683	700	4	is	be	AUX
dad-10683	700	5	still	still	ADV
dad-10683	700	6	an	an	DET
dad-10683	700	7	open	open	ADJ
dad-10683	700	8	problem	problem	NOUN
dad-10683	700	9	.	.	PUNCT
dad-10683	701	1	this	this	PRON
dad-10683	701	2	is	be	AUX
dad-10683	701	3	related	relate	VERB
dad-10683	701	4	to	to	ADP
dad-10683	701	5	the	the	DET
dad-10683	701	6	issue	issue	NOUN
dad-10683	701	7	of	of	ADP
dad-10683	701	8	end	end	NOUN
dad-10683	701	9	-	-	PUNCT
dad-10683	701	10	to	to	ADP
dad-10683	701	11	-	-	PUNCT
dad-10683	701	12	end	end	NOUN
dad-10683	701	13	dialogue	dialogue	NOUN
dad-10683	701	14	personalization	personalization	NOUN
dad-10683	701	15	,	,	PUNCT
dad-10683	701	16	which	which	PRON
dad-10683	701	17	involves	involve	VERB
dad-10683	701	18	building	build	VERB
dad-10683	701	19	end	end	NOUN
dad-10683	701	20	-	-	PUNCT
dad-10683	701	21	to	to	ADP
dad-10683	701	22	-	-	PUNCT
dad-10683	701	23	end	end	NOUN
dad-10683	701	24	dialogue	dialogue	NOUN
dad-10683	701	25	systems	system	NOUN
dad-10683	701	26	that	that	PRON
dad-10683	701	27	are	be	AUX
dad-10683	701	28	tailored	tailor	VERB
dad-10683	701	29	to	to	ADP
dad-10683	701	30	a	a	DET
dad-10683	701	31	particular	particular	ADJ
dad-10683	701	32	user	user	NOUN
dad-10683	701	33	and	and	CCONJ
dad-10683	701	34	that	that	PRON
dad-10683	701	35	evolve	evolve	VERB
dad-10683	701	36	over	over	ADP
dad-10683	701	37	time	time	NOUN
dad-10683	701	38	as	as	SCONJ
dad-10683	701	39	the	the	DET
dad-10683	701	40	user	user	NOUN
dad-10683	701	41	’s	’s	PART
dad-10683	701	42	preferences	preference	NOUN
dad-10683	701	43	change	change	VERB
dad-10683	701	44	.	.	PUNCT
dad-10683	702	1	the	the	DET
dad-10683	702	2	largest	large	ADJ
dad-10683	702	3	source	source	NOUN
dad-10683	702	4	of	of	ADP
dad-10683	702	5	errors	error	NOUN
dad-10683	702	6	from	from	ADP
dad-10683	702	7	the	the	DET
dad-10683	702	8	analysis	analysis	NOUN
dad-10683	702	9	in	in	ADP
dad-10683	702	10	section	section	NOUN
dad-10683	702	11	4.6	4.6	NUM
dad-10683	702	12	was	be	AUX
dad-10683	702	13	in	in	ADP
dad-10683	702	14	the	the	DET
dad-10683	702	15	failure	failure	NOUN
dad-10683	702	16	to	to	PART
dad-10683	702	17	understand	understand	VERB
dad-10683	702	18	the	the	DET
dad-10683	702	19	semantic	semantic	ADJ
dad-10683	702	20	similarity	similarity	NOUN
dad-10683	702	21	between	between	ADP
dad-10683	702	22	the	the	DET
dad-10683	702	23	context	context	NOUN
dad-10683	702	24	and	and	CCONJ
dad-10683	702	25	response	response	NOUN
dad-10683	702	26	.	.	PUNCT
dad-10683	703	1	this	this	PRON
dad-10683	703	2	falls	fall	VERB
dad-10683	703	3	under	under	ADP
dad-10683	703	4	the	the	DET
dad-10683	703	5	more	more	ADV
dad-10683	703	6	general	general	ADJ
dad-10683	703	7	problem	problem	NOUN
dad-10683	703	8	of	of	ADP
dad-10683	703	9	natural	natural	ADJ
dad-10683	703	10	language	language	NOUN
dad-10683	703	11	understanding	understanding	NOUN
dad-10683	703	12	,	,	PUNCT
dad-10683	703	13	which	which	PRON
dad-10683	703	14	arises	arise	VERB
dad-10683	703	15	in	in	ADP
dad-10683	703	16	many	many	ADJ
dad-10683	703	17	nlp	nlp	NOUN
dad-10683	703	18	tasks	task	NOUN
dad-10683	703	19	.	.	PUNCT
dad-10683	704	1	this	this	PRON
dad-10683	704	2	will	will	AUX
dad-10683	704	3	require	require	VERB
dad-10683	704	4	adjustments	adjustment	NOUN
dad-10683	704	5	in	in	ADP
dad-10683	704	6	the	the	DET
dad-10683	704	7	architecture	architecture	NOUN
dad-10683	704	8	of	of	ADP
dad-10683	704	9	end	end	NOUN
dad-10683	704	10	-	-	PUNCT
dad-10683	704	11	to	to	ADP
dad-10683	704	12	-	-	PUNCT
dad-10683	704	13	end	end	NOUN
dad-10683	704	14	models	model	NOUN
dad-10683	704	15	to	to	PART
dad-10683	704	16	render	render	VERB
dad-10683	704	17	them	they	PRON
dad-10683	704	18	more	more	ADV
dad-10683	704	19	suited	suited	ADJ
dad-10683	704	20	to	to	ADP
dad-10683	704	21	processing	processing	NOUN
dad-10683	704	22	language	language	NOUN
dad-10683	704	23	.	.	PUNCT
dad-10683	705	1	it	it	PRON
dad-10683	705	2	is	be	AUX
dad-10683	705	3	possible	possible	ADJ
dad-10683	705	4	that	that	SCONJ
dad-10683	705	5	insights	insight	NOUN
dad-10683	705	6	can	can	AUX
dad-10683	705	7	be	be	AUX
dad-10683	705	8	derived	derive	VERB
dad-10683	705	9	from	from	ADP
dad-10683	705	10	architectures	architecture	NOUN
dad-10683	705	11	developed	develop	VERB
dad-10683	705	12	on	on	ADP
dad-10683	705	13	more	more	ADJ
dad-10683	705	14	targeted	target	VERB
dad-10683	705	15	language	language	NOUN
dad-10683	705	16	understanding	understanding	NOUN
dad-10683	705	17	tasks	task	NOUN
dad-10683	705	18	,	,	PUNCT
dad-10683	705	19	such	such	ADJ
dad-10683	705	20	as	as	ADP
dad-10683	705	21	the	the	DET
dad-10683	705	22	cnn/	cnn/	NOUN
dad-10683	705	23	daily	daily	ADJ
dad-10683	705	24	mail	mail	NOUN
dad-10683	705	25	reading	reading	NOUN
dad-10683	705	26	comprehension	comprehension	NOUN
dad-10683	705	27	dataset	dataset	NOUN
dad-10683	705	28	(	(	PUNCT
dad-10683	705	29	hermann	hermann	PROPN
dad-10683	705	30	et	et	PROPN
dad-10683	705	31	al	al	PROPN
dad-10683	705	32	.	.	PROPN
dad-10683	705	33	,	,	PUNCT
dad-10683	705	34	2015	2015	NUM
dad-10683	705	35	)	)	PUNCT
dad-10683	705	36	,	,	PUNCT
dad-10683	705	37	where	where	SCONJ
dad-10683	705	38	attention	attention	NOUN
dad-10683	705	39	-	-	PUNCT
dad-10683	705	40	based	base	VERB
dad-10683	705	41	models	model	NOUN
dad-10683	705	42	have	have	AUX
dad-10683	705	43	achieved	achieve	VERB
dad-10683	705	44	strong	strong	ADJ
dad-10683	705	45	performance	performance	NOUN
dad-10683	705	46	.	.	PUNCT
dad-10683	706	1	in	in	ADP
dad-10683	706	2	order	order	NOUN
dad-10683	706	3	to	to	PART
dad-10683	706	4	be	be	AUX
dad-10683	706	5	able	able	ADJ
dad-10683	706	6	to	to	PART
dad-10683	706	7	correctly	correctly	ADV
dad-10683	706	8	answer	answer	VERB
dad-10683	706	9	questions	question	NOUN
dad-10683	706	10	regarding	regard	VERB
dad-10683	706	11	ubuntu	ubuntu	NOUN
dad-10683	706	12	and	and	CCONJ
dad-10683	706	13	solve	solve	VERB
dad-10683	706	14	the	the	DET
dad-10683	706	15	user	user	NOUN
dad-10683	706	16	’s	’s	PART
dad-10683	706	17	problem	problem	NOUN
dad-10683	706	18	,	,	PUNCT
dad-10683	706	19	dialogue	dialogue	NOUN
dad-10683	706	20	models	model	NOUN
dad-10683	706	21	will	will	AUX
dad-10683	706	22	inevitably	inevitably	ADV
dad-10683	706	23	require	require	VERB
dad-10683	706	24	some	some	DET
dad-10683	706	25	knowledge	knowledge	NOUN
dad-10683	706	26	of	of	ADP
dad-10683	706	27	the	the	DET
dad-10683	706	28	ubuntu	ubuntu	ADJ
dad-10683	706	29	domain	domain	NOUN
dad-10683	706	30	.	.	PUNCT
dad-10683	707	1	this	this	PRON
dad-10683	707	2	will	will	AUX
dad-10683	707	3	most	most	ADV
dad-10683	707	4	likely	likely	ADV
dad-10683	707	5	be	be	AUX
dad-10683	707	6	achieved	achieve	VERB
dad-10683	707	7	by	by	ADP
dad-10683	707	8	using	use	VERB
dad-10683	707	9	some	some	DET
dad-10683	707	10	source	source	NOUN
dad-10683	707	11	of	of	ADP
dad-10683	707	12	external	external	ADJ
dad-10683	707	13	knowledge	knowledge	NOUN
dad-10683	707	14	,	,	PUNCT
dad-10683	707	15	in	in	ADP
dad-10683	707	16	addition	addition	NOUN
dad-10683	707	17	to	to	ADP
dad-10683	707	18	the	the	DET
dad-10683	707	19	knowledge	knowledge	NOUN
dad-10683	707	20	that	that	PRON
dad-10683	707	21	is	be	AUX
dad-10683	707	22	present	present	ADJ
dad-10683	707	23	in	in	ADP
dad-10683	707	24	the	the	DET
dad-10683	707	25	dialogue	dialogue	NOUN
dad-10683	707	26	of	of	ADP
dad-10683	707	27	the	the	DET
dad-10683	707	28	ubuntu	ubuntu	ADJ
dad-10683	707	29	dialogue	dialogue	NOUN
dad-10683	707	30	corpus	corpus	NOUN
dad-10683	707	31	.	.	PUNCT
dad-10683	708	1	thus	thus	ADV
dad-10683	708	2	,	,	PUNCT
dad-10683	708	3	an	an	DET
dad-10683	708	4	important	important	ADJ
dad-10683	708	5	direction	direction	NOUN
dad-10683	708	6	for	for	ADP
dad-10683	708	7	research	research	NOUN
dad-10683	708	8	is	be	AUX
dad-10683	708	9	the	the	DET
dad-10683	708	10	investigation	investigation	NOUN
dad-10683	708	11	of	of	ADP
dad-10683	708	12	methods	method	NOUN
dad-10683	708	13	that	that	PRON
dad-10683	708	14	incorporate	incorporate	VERB
dad-10683	708	15	external	external	ADJ
dad-10683	708	16	knowledge	knowledge	NOUN
dad-10683	708	17	sources	source	NOUN
dad-10683	708	18	with	with	ADP
dad-10683	708	19	end	end	NOUN
dad-10683	708	20	-	-	PUNCT
dad-10683	708	21	to	to	ADP
dad-10683	708	22	-	-	PUNCT
dad-10683	708	23	end	end	NOUN
dad-10683	708	24	dialogue	dialogue	NOUN
dad-10683	708	25	systems	system	NOUN
dad-10683	708	26	.	.	PUNCT
dad-10683	709	1	this	this	PRON
dad-10683	709	2	applies	apply	VERB
dad-10683	709	3	more	more	ADV
dad-10683	709	4	generally	generally	ADV
dad-10683	709	5	to	to	ADP
dad-10683	709	6	any	any	DET
dad-10683	709	7	end	end	NOUN
dad-10683	709	8	-	-	PUNCT
dad-10683	709	9	to	to	ADP
dad-10683	709	10	-	-	PUNCT
dad-10683	709	11	end	end	NOUN
dad-10683	709	12	system	system	NOUN
dad-10683	709	13	that	that	PRON
dad-10683	709	14	is	be	AUX
dad-10683	709	15	developed	develop	VERB
dad-10683	709	16	for	for	ADP
dad-10683	709	17	the	the	DET
dad-10683	709	18	goaloriented	goaloriente	VERB
dad-10683	709	19	setting	setting	NOUN
dad-10683	709	20	,	,	PUNCT
dad-10683	709	21	and	and	CCONJ
dad-10683	709	22	may	may	AUX
dad-10683	709	23	require	require	VERB
dad-10683	709	24	imposing	impose	VERB
dad-10683	709	25	additional	additional	ADJ
dad-10683	709	26	structure	structure	NOUN
dad-10683	709	27	on	on	ADP
dad-10683	709	28	the	the	DET
dad-10683	709	29	output	output	NOUN
dad-10683	709	30	space	space	NOUN
dad-10683	709	31	of	of	ADP
dad-10683	709	32	the	the	DET
dad-10683	709	33	model	model	NOUN
dad-10683	709	34	.	.	PUNCT
dad-10683	710	1	there	there	PRON
dad-10683	710	2	is	be	VERB
dad-10683	710	3	promising	promise	VERB
dad-10683	710	4	work	work	NOUN
dad-10683	710	5	in	in	ADP
dad-10683	710	6	this	this	DET
dad-10683	710	7	direction	direction	NOUN
dad-10683	710	8	from	from	ADP
dad-10683	710	9	wen	wen	PROPN
dad-10683	710	10	et	et	PROPN
dad-10683	710	11	al	al	PROPN
dad-10683	710	12	.	.	PROPN
dad-10683	711	1	(	(	PUNCT
dad-10683	711	2	2016	2016	NUM
dad-10683	711	3	)	)	PUNCT
dad-10683	711	4	,	,	PUNCT
dad-10683	711	5	however	however	SCONJ
dad-10683	711	6	methods	method	NOUN
dad-10683	711	7	must	must	AUX
dad-10683	711	8	be	be	AUX
dad-10683	711	9	derived	derive	VERB
dad-10683	711	10	that	that	PRON
dad-10683	711	11	are	be	AUX
dad-10683	711	12	effective	effective	ADJ
dad-10683	711	13	in	in	ADP
dad-10683	711	14	a	a	DET
dad-10683	711	15	larger	large	ADJ
dad-10683	711	16	and	and	CCONJ
dad-10683	711	17	more	more	ADV
dad-10683	711	18	general	general	ADJ
dad-10683	711	19	setting	setting	NOUN
dad-10683	711	20	than	than	ADP
dad-10683	711	21	restaurant	restaurant	NOUN
dad-10683	711	22	recommendation	recommendation	NOUN
dad-10683	711	23	.	.	PUNCT
dad-10683	712	1	a	a	DET
dad-10683	712	2	common	common	ADJ
dad-10683	712	3	problem	problem	NOUN
dad-10683	712	4	that	that	PRON
dad-10683	712	5	has	have	AUX
dad-10683	712	6	been	be	AUX
dad-10683	712	7	observed	observe	VERB
dad-10683	712	8	when	when	SCONJ
dad-10683	712	9	training	training	NOUN
dad-10683	712	10	generative	generative	ADJ
dad-10683	712	11	end	end	NOUN
dad-10683	712	12	-	-	PUNCT
dad-10683	712	13	to	to	ADP
dad-10683	712	14	-	-	PUNCT
dad-10683	712	15	end	end	NOUN
dad-10683	712	16	models	model	NOUN
dad-10683	712	17	that	that	PRON
dad-10683	712	18	maximize	maximize	VERB
dad-10683	712	19	the	the	DET
dad-10683	712	20	log	log	NOUN
dad-10683	712	21	-	-	PUNCT
dad-10683	712	22	likelihood	likelihood	NOUN
dad-10683	712	23	of	of	ADP
dad-10683	712	24	the	the	DET
dad-10683	712	25	conversational	conversational	ADJ
dad-10683	712	26	response	response	NOUN
dad-10683	712	27	is	be	AUX
dad-10683	712	28	that	that	SCONJ
dad-10683	712	29	these	these	DET
dad-10683	712	30	models	model	NOUN
dad-10683	712	31	tend	tend	VERB
dad-10683	712	32	to	to	PART
dad-10683	712	33	produce	produce	VERB
dad-10683	712	34	generic	generic	ADJ
dad-10683	712	35	responses	response	NOUN
dad-10683	712	36	at	at	ADP
dad-10683	712	37	test	test	NOUN
dad-10683	712	38	time	time	NOUN
dad-10683	712	39	.	.	PUNCT
dad-10683	713	1	this	this	PRON
dad-10683	713	2	has	have	AUX
dad-10683	713	3	been	be	AUX
dad-10683	713	4	observed	observe	VERB
dad-10683	713	5	empirically	empirically	ADV
dad-10683	713	6	(	(	PUNCT
dad-10683	713	7	vinyals	vinyal	NOUN
dad-10683	713	8	and	and	CCONJ
dad-10683	713	9	le	le	NOUN
dad-10683	713	10	,	,	PUNCT
dad-10683	713	11	2015	2015	NUM
dad-10683	713	12	;	;	PUNCT
dad-10683	713	13	serban	serban	PROPN
dad-10683	713	14	et	et	PROPN
dad-10683	713	15	al	al	PROPN
dad-10683	713	16	.	.	PROPN
dad-10683	713	17	,	,	PUNCT
dad-10683	713	18	2016	2016	NUM
dad-10683	713	19	)	)	PUNCT
dad-10683	713	20	,	,	PUNCT
dad-10683	713	21	and	and	CCONJ
dad-10683	713	22	was	be	AUX
dad-10683	713	23	also	also	ADV
dad-10683	713	24	seen	see	VERB
dad-10683	713	25	in	in	ADP
dad-10683	713	26	some	some	PRON
dad-10683	713	27	of	of	ADP
dad-10683	713	28	the	the	DET
dad-10683	713	29	lstm	lstm	NOUN
dad-10683	713	30	and	and	CCONJ
dad-10683	713	31	hred	hre	VERB
dad-10683	713	32	examples	example	NOUN
dad-10683	713	33	presented	present	VERB
dad-10683	713	34	in	in	ADP
dad-10683	713	35	section	section	NOUN
dad-10683	713	36	60	60	NUM
dad-10683	713	37	training	training	NOUN
dad-10683	713	38	end	end	NOUN
dad-10683	713	39	-	-	PUNCT
dad-10683	713	40	to	to	ADP
dad-10683	713	41	-	-	PUNCT
dad-10683	713	42	end	end	NOUN
dad-10683	713	43	dialogue	dialogue	NOUN
dad-10683	713	44	systems	system	NOUN
dad-10683	713	45	5.4	5.4	NUM
dad-10683	713	46	.	.	PUNCT
dad-10683	714	1	this	this	PRON
dad-10683	714	2	has	have	AUX
dad-10683	714	3	been	be	AUX
dad-10683	714	4	investigated	investigate	VERB
dad-10683	714	5	in	in	ADP
dad-10683	714	6	(	(	PUNCT
dad-10683	714	7	li	li	PROPN
dad-10683	714	8	et	et	PROPN
dad-10683	714	9	al	al	PROPN
dad-10683	714	10	.	.	PROPN
dad-10683	714	11	,	,	PUNCT
dad-10683	714	12	2015	2015	NUM
dad-10683	714	13	)	)	PUNCT
dad-10683	714	14	,	,	PUNCT
dad-10683	714	15	where	where	SCONJ
dad-10683	714	16	the	the	DET
dad-10683	714	17	authors	author	NOUN
dad-10683	714	18	construct	construct	VERB
dad-10683	714	19	an	an	DET
dad-10683	714	20	objective	objective	ADJ
dad-10683	714	21	function	function	NOUN
dad-10683	714	22	based	base	VERB
dad-10683	714	23	on	on	ADP
dad-10683	714	24	mutual	mutual	ADJ
dad-10683	714	25	information	information	NOUN
dad-10683	714	26	that	that	PRON
dad-10683	714	27	promotes	promote	VERB
dad-10683	714	28	diversity	diversity	NOUN
dad-10683	714	29	,	,	PUNCT
dad-10683	714	30	however	however	ADV
dad-10683	714	31	they	they	PRON
dad-10683	714	32	achieve	achieve	VERB
dad-10683	714	33	only	only	ADV
dad-10683	714	34	modest	modest	ADJ
dad-10683	714	35	improvements	improvement	NOUN
dad-10683	714	36	.	.	PUNCT
dad-10683	715	1	this	this	PRON
dad-10683	715	2	is	be	AUX
dad-10683	715	3	a	a	DET
dad-10683	715	4	large	large	ADJ
dad-10683	715	5	impediment	impediment	NOUN
dad-10683	715	6	for	for	ADP
dad-10683	715	7	building	build	VERB
dad-10683	715	8	end	end	NOUN
dad-10683	715	9	-	-	PUNCT
dad-10683	715	10	to	to	ADP
dad-10683	715	11	-	-	PUNCT
dad-10683	715	12	end	end	NOUN
dad-10683	715	13	systems	system	NOUN
dad-10683	715	14	that	that	PRON
dad-10683	715	15	can	can	AUX
dad-10683	715	16	have	have	AUX
dad-10683	715	17	interesting	interesting	ADJ
dad-10683	715	18	and	and	CCONJ
dad-10683	715	19	engaging	engage	VERB
dad-10683	715	20	interactions	interaction	NOUN
dad-10683	715	21	with	with	ADP
dad-10683	715	22	users	user	NOUN
dad-10683	715	23	.	.	PUNCT
dad-10683	716	1	finally	finally	ADV
dad-10683	716	2	,	,	PUNCT
dad-10683	716	3	an	an	DET
dad-10683	716	4	important	important	ADJ
dad-10683	716	5	direction	direction	NOUN
dad-10683	716	6	for	for	ADP
dad-10683	716	7	future	future	ADJ
dad-10683	716	8	research	research	NOUN
dad-10683	716	9	is	be	AUX
dad-10683	716	10	building	build	VERB
dad-10683	716	11	large	large	ADJ
dad-10683	716	12	-	-	PUNCT
dad-10683	716	13	scale	scale	NOUN
dad-10683	716	14	datasets	dataset	NOUN
dad-10683	716	15	that	that	PRON
dad-10683	716	16	allow	allow	VERB
dad-10683	716	17	the	the	DET
dad-10683	716	18	training	training	NOUN
dad-10683	716	19	of	of	ADP
dad-10683	716	20	goal	goal	NOUN
dad-10683	716	21	-	-	PUNCT
dad-10683	716	22	oriented	orient	VERB
dad-10683	716	23	systems	system	NOUN
dad-10683	716	24	.	.	PUNCT
dad-10683	717	1	the	the	DET
dad-10683	717	2	ubuntu	ubuntu	ADJ
dad-10683	717	3	domain	domain	NOUN
dad-10683	717	4	is	be	AUX
dad-10683	717	5	particularly	particularly	ADV
dad-10683	717	6	suited	suit	VERB
dad-10683	717	7	for	for	ADP
dad-10683	717	8	training	training	NOUN
dad-10683	717	9	goaloriented	goaloriente	VERB
dad-10683	717	10	systems	system	NOUN
dad-10683	717	11	,	,	PUNCT
dad-10683	717	12	however	however	ADV
dad-10683	717	13	this	this	PRON
dad-10683	717	14	is	be	AUX
dad-10683	717	15	not	not	PART
dad-10683	717	16	yet	yet	ADV
dad-10683	717	17	possible	possible	ADJ
dad-10683	717	18	on	on	ADP
dad-10683	717	19	the	the	DET
dad-10683	717	20	ubuntu	ubuntu	ADJ
dad-10683	717	21	dialogue	dialogue	NOUN
dad-10683	717	22	corpus	corpus	NOUN
dad-10683	717	23	as	as	SCONJ
dad-10683	717	24	there	there	PRON
dad-10683	717	25	are	be	VERB
dad-10683	717	26	no	no	DET
dad-10683	717	27	supervised	supervised	ADJ
dad-10683	717	28	task	task	NOUN
dad-10683	717	29	completion	completion	NOUN
dad-10683	717	30	signals	signal	NOUN
dad-10683	717	31	,	,	PUNCT
dad-10683	717	32	as	as	SCONJ
dad-10683	717	33	mentioned	mention	VERB
dad-10683	717	34	in	in	ADP
dad-10683	717	35	section	section	NOUN
dad-10683	717	36	6.2	6.2	NUM
dad-10683	717	37	.	.	PUNCT
dad-10683	718	1	building	building	NOUN
dad-10683	718	2	models	model	NOUN
dad-10683	718	3	that	that	PRON
dad-10683	718	4	can	can	AUX
dad-10683	718	5	approximate	approximate	VERB
dad-10683	718	6	such	such	ADJ
dad-10683	718	7	signals	signal	NOUN
dad-10683	718	8	is	be	AUX
dad-10683	718	9	challenging	challenging	ADJ
dad-10683	718	10	,	,	PUNCT
dad-10683	718	11	yet	yet	CCONJ
dad-10683	718	12	it	it	PRON
dad-10683	718	13	may	may	AUX
dad-10683	718	14	be	be	AUX
dad-10683	718	15	necessary	necessary	ADJ
dad-10683	718	16	in	in	ADP
dad-10683	718	17	order	order	NOUN
dad-10683	718	18	to	to	PART
dad-10683	718	19	develop	develop	VERB
dad-10683	718	20	systems	system	NOUN
dad-10683	718	21	that	that	PRON
dad-10683	718	22	can	can	AUX
dad-10683	718	23	solve	solve	VERB
dad-10683	718	24	users	user	NOUN
dad-10683	718	25	’	’	PART
dad-10683	718	26	problems	problem	NOUN
dad-10683	718	27	in	in	ADP
dad-10683	718	28	a	a	DET
dad-10683	718	29	meaningful	meaningful	ADJ
dad-10683	718	30	way	way	NOUN
dad-10683	718	31	in	in	ADP
dad-10683	718	32	a	a	DET
dad-10683	718	33	domain	domain	NOUN
dad-10683	718	34	as	as	ADV
dad-10683	718	35	complex	complex	ADJ
dad-10683	718	36	as	as	ADP
dad-10683	718	37	ubuntu	ubuntu	NOUN
dad-10683	718	38	.	.	PUNCT
dad-10683	719	1	acknowledgments	acknowledgment	NOUN
dad-10683	719	2	the	the	DET
dad-10683	719	3	authors	author	NOUN
dad-10683	719	4	would	would	AUX
dad-10683	719	5	like	like	VERB
dad-10683	719	6	to	to	PART
dad-10683	719	7	profusely	profusely	ADV
dad-10683	719	8	thank	thank	VERB
dad-10683	719	9	rudolf	rudolf	NOUN
dad-10683	719	10	kadlec	kadlec	PROPN
dad-10683	719	11	and	and	CCONJ
dad-10683	719	12	martin	martin	PROPN
dad-10683	719	13	schmid	schmid	PROPN
dad-10683	719	14	,	,	PUNCT
dad-10683	719	15	who	who	PRON
dad-10683	719	16	produced	produce	VERB
dad-10683	719	17	the	the	DET
dad-10683	719	18	script	script	NOUN
dad-10683	719	19	for	for	ADP
dad-10683	719	20	generating	generate	VERB
dad-10683	719	21	the	the	DET
dad-10683	719	22	updated	update	VERB
dad-10683	719	23	version	version	NOUN
dad-10683	719	24	of	of	ADP
dad-10683	719	25	the	the	DET
dad-10683	719	26	ubuntu	ubuntu	ADJ
dad-10683	719	27	dialogue	dialogue	NOUN
dad-10683	719	28	corpus	corpus	NOUN
dad-10683	719	29	.	.	PUNCT
dad-10683	720	1	we	we	PRON
dad-10683	720	2	gratefully	gratefully	ADV
dad-10683	720	3	acknowledge	acknowledge	VERB
dad-10683	720	4	financial	financial	ADJ
dad-10683	720	5	support	support	NOUN
dad-10683	720	6	for	for	ADP
dad-10683	720	7	this	this	DET
dad-10683	720	8	work	work	NOUN
dad-10683	720	9	by	by	ADP
dad-10683	720	10	the	the	DET
dad-10683	720	11	samsung	samsung	PROPN
dad-10683	720	12	advanced	advanced	PROPN
dad-10683	720	13	institute	institute	PROPN
dad-10683	720	14	of	of	ADP
dad-10683	720	15	technology	technology	NOUN
dad-10683	720	16	(	(	PUNCT
dad-10683	720	17	sait	sait	ADJ
dad-10683	720	18	)	)	PUNCT
dad-10683	720	19	and	and	CCONJ
dad-10683	720	20	the	the	DET
dad-10683	720	21	natural	natural	ADJ
dad-10683	720	22	sciences	sciences	PROPN
dad-10683	720	23	and	and	CCONJ
dad-10683	720	24	engineering	engineering	NOUN
dad-10683	720	25	research	research	NOUN
dad-10683	720	26	council	council	PROPN
dad-10683	720	27	of	of	ADP
dad-10683	720	28	canada	canada	PROPN
dad-10683	720	29	(	(	PUNCT
dad-10683	720	30	nserc	nserc	PROPN
dad-10683	720	31	)	)	PUNCT
dad-10683	720	32	.	.	PUNCT
dad-10683	721	1	we	we	PRON
dad-10683	721	2	would	would	AUX
dad-10683	721	3	like	like	VERB
dad-10683	721	4	to	to	PART
dad-10683	721	5	thank	thank	VERB
dad-10683	721	6	michael	michael	PROPN
dad-10683	721	7	noseworthy	noseworthy	PROPN
dad-10683	721	8	and	and	CCONJ
dad-10683	721	9	nicolas	nicolas	PROPN
dad-10683	721	10	angelard	angelard	PROPN
dad-10683	721	11	-	-	PUNCT
dad-10683	721	12	gontier	gontier	NOUN
dad-10683	721	13	for	for	ADP
dad-10683	721	14	their	their	PRON
dad-10683	721	15	input	input	NOUN
dad-10683	721	16	into	into	ADP
dad-10683	721	17	this	this	DET
dad-10683	721	18	paper	paper	NOUN
dad-10683	721	19	,	,	PUNCT
dad-10683	721	20	and	and	CCONJ
dad-10683	721	21	the	the	DET
dad-10683	721	22	anonymous	anonymous	ADJ
dad-10683	721	23	reviewers	reviewer	NOUN
dad-10683	721	24	for	for	ADP
dad-10683	721	25	their	their	PRON
dad-10683	721	26	helpful	helpful	ADJ
dad-10683	721	27	feedback	feedback	NOUN
dad-10683	721	28	.	.	PUNCT
dad-10683	722	1	references	reference	NOUN
dad-10683	722	2	p.	p.	NOUN
dad-10683	722	3	baudiš	baudiš	PROPN
dad-10683	723	1	and	and	CCONJ
dad-10683	723	2	j.	j.	PROPN
dad-10683	723	3	šedivỳ.	šedivỳ.	PROPN
dad-10683	723	4	sentence	sentence	NOUN
dad-10683	723	5	pair	pair	NOUN
dad-10683	723	6	scoring	scoring	NOUN
dad-10683	723	7	:	:	PUNCT
dad-10683	723	8	towards	towards	ADP
dad-10683	723	9	unified	unify	VERB
dad-10683	723	10	framework	framework	NOUN
dad-10683	723	11	for	for	ADP
dad-10683	723	12	text	text	NOUN
dad-10683	723	13	comprehension	comprehension	NOUN
dad-10683	723	14	.	.	PUNCT
dad-10683	724	1	arxiv	arxiv	PROPN
dad-10683	724	2	preprint	preprint	NOUN
dad-10683	724	3	arxiv:1603.06127	arxiv:1603.06127	NOUN
dad-10683	724	4	,	,	PUNCT
dad-10683	724	5	2016	2016	NUM
dad-10683	724	6	.	.	PUNCT
dad-10683	725	1	y.	y.	PROPN
dad-10683	725	2	bengio	bengio	PROPN
dad-10683	725	3	,	,	PUNCT
dad-10683	725	4	p.	p.	NOUN
dad-10683	725	5	simard	simard	PROPN
dad-10683	725	6	,	,	PUNCT
dad-10683	725	7	and	and	CCONJ
dad-10683	725	8	p.	p.	NOUN
dad-10683	725	9	frasconi	frasconi	NOUN
dad-10683	725	10	.	.	PUNCT
dad-10683	726	1	learning	learn	VERB
dad-10683	726	2	long	long	ADJ
dad-10683	726	3	-	-	PUNCT
dad-10683	726	4	term	term	NOUN
dad-10683	726	5	dependencies	dependency	NOUN
dad-10683	726	6	with	with	ADP
dad-10683	726	7	gradient	gradient	ADJ
dad-10683	726	8	descent	descent	NOUN
dad-10683	726	9	is	be	AUX
dad-10683	726	10	difficult	difficult	ADJ
dad-10683	726	11	.	.	PUNCT
dad-10683	727	1	neural	neural	ADJ
dad-10683	727	2	networks	network	NOUN
dad-10683	727	3	,	,	PUNCT
dad-10683	727	4	ieee	ieee	NOUN
dad-10683	727	5	transactions	transaction	NOUN
dad-10683	727	6	on	on	ADP
dad-10683	727	7	,	,	PUNCT
dad-10683	727	8	5(2):157–166	5(2):157–166	NUM
dad-10683	727	9	,	,	PUNCT
dad-10683	727	10	1994	1994	NUM
dad-10683	727	11	.	.	PUNCT
dad-10683	728	1	a.	a.	PROPN
dad-10683	728	2	bordes	bordes	PROPN
dad-10683	728	3	,	,	PUNCT
dad-10683	728	4	j.	j.	PROPN
dad-10683	728	5	weston	weston	PROPN
dad-10683	728	6	,	,	PUNCT
dad-10683	728	7	and	and	CCONJ
dad-10683	728	8	n.	n.	PROPN
dad-10683	728	9	usunier	usunier	PROPN
dad-10683	728	10	.	.	PUNCT
dad-10683	729	1	open	open	ADJ
dad-10683	729	2	question	question	NOUN
dad-10683	729	3	answering	answer	VERB
dad-10683	729	4	with	with	ADP
dad-10683	729	5	weakly	weakly	ADJ
dad-10683	729	6	supervised	supervised	ADJ
dad-10683	729	7	embedding	embed	VERB
dad-10683	729	8	models	model	NOUN
dad-10683	729	9	.	.	PUNCT
dad-10683	730	1	in	in	ADP
dad-10683	730	2	proceedings	proceeding	NOUN
dad-10683	730	3	of	of	ADP
dad-10683	730	4	the	the	DET
dad-10683	730	5	meeting	meeting	NOUN
dad-10683	730	6	on	on	ADP
dad-10683	730	7	machine	machine	NOUN
dad-10683	730	8	learning	learning	NOUN
dad-10683	730	9	and	and	CCONJ
dad-10683	730	10	knowledge	knowledge	NOUN
dad-10683	730	11	discovery	discovery	NOUN
dad-10683	730	12	in	in	ADP
dad-10683	730	13	databases	database	NOUN
dad-10683	730	14	,	,	PUNCT
dad-10683	730	15	2014	2014	NUM
dad-10683	730	16	.	.	PUNCT
dad-10683	731	1	j.	j.	PROPN
dad-10683	731	2	boyd	boyd	PROPN
dad-10683	731	3	-	-	PUNCT
dad-10683	731	4	graber	graber	PROPN
dad-10683	731	5	,	,	PUNCT
dad-10683	731	6	b.	b.	PROPN
dad-10683	731	7	satinoff	satinoff	PROPN
dad-10683	731	8	,	,	PUNCT
dad-10683	731	9	h.	h.	PROPN
dad-10683	731	10	he	he	PRON
dad-10683	731	11	,	,	PUNCT
dad-10683	731	12	and	and	CCONJ
dad-10683	731	13	h.	h.	PROPN
dad-10683	731	14	daume	daume	PROPN
dad-10683	731	15	.	.	PUNCT
dad-10683	732	1	besting	best	VERB
dad-10683	732	2	the	the	DET
dad-10683	732	3	quiz	quiz	NOUN
dad-10683	732	4	master	master	NOUN
dad-10683	732	5	:	:	PUNCT
dad-10683	732	6	crowdsourcing	crowdsource	VERB
dad-10683	732	7	incremental	incremental	ADJ
dad-10683	732	8	classification	classification	NOUN
dad-10683	732	9	games	game	NOUN
dad-10683	732	10	.	.	PUNCT
dad-10683	733	1	in	in	ADP
dad-10683	733	2	proceedings	proceeding	NOUN
dad-10683	733	3	of	of	ADP
dad-10683	733	4	the	the	DET
dad-10683	733	5	conference	conference	NOUN
dad-10683	733	6	on	on	ADP
dad-10683	733	7	empirical	empirical	ADJ
dad-10683	733	8	methods	method	NOUN
dad-10683	733	9	in	in	ADP
dad-10683	733	10	natural	natural	ADJ
dad-10683	733	11	language	language	NOUN
dad-10683	733	12	processing	processing	NOUN
dad-10683	733	13	,	,	PUNCT
dad-10683	733	14	2012	2012	NUM
dad-10683	733	15	.	.	PUNCT
dad-10683	734	1	k.	k.	PROPN
dad-10683	734	2	cho	cho	PROPN
dad-10683	734	3	,	,	PUNCT
dad-10683	734	4	b.	b.	PROPN
dad-10683	734	5	van	van	PROPN
dad-10683	734	6	merrienboer	merrienboer	PROPN
dad-10683	734	7	,	,	PUNCT
dad-10683	734	8	c.	c.	PROPN
dad-10683	734	9	gulcehre	gulcehre	PROPN
dad-10683	734	10	,	,	PUNCT
dad-10683	734	11	f.	f.	PROPN
dad-10683	734	12	bougares	bougares	PROPN
dad-10683	734	13	,	,	PUNCT
dad-10683	734	14	h.	h.	PROPN
dad-10683	734	15	schwenk	schwenk	PROPN
dad-10683	734	16	,	,	PUNCT
dad-10683	734	17	and	and	CCONJ
dad-10683	734	18	y.	y.	PROPN
dad-10683	734	19	bengio	bengio	PROPN
dad-10683	734	20	.	.	PUNCT
dad-10683	735	1	learning	learn	VERB
dad-10683	735	2	phrase	phrase	NOUN
dad-10683	735	3	representations	representation	NOUN
dad-10683	735	4	using	use	VERB
dad-10683	735	5	rnn	rnn	NOUN
dad-10683	735	6	encoder	encoder	NOUN
dad-10683	735	7	-	-	PUNCT
dad-10683	735	8	decoder	decoder	NOUN
dad-10683	735	9	for	for	ADP
dad-10683	735	10	statistical	statistical	ADJ
dad-10683	735	11	machine	machine	NOUN
dad-10683	735	12	translation	translation	NOUN
dad-10683	735	13	.	.	PUNCT
dad-10683	736	1	in	in	ADP
dad-10683	736	2	proceedings	proceeding	NOUN
dad-10683	736	3	of	of	ADP
dad-10683	736	4	the	the	DET
dad-10683	736	5	conference	conference	NOUN
dad-10683	736	6	on	on	ADP
dad-10683	736	7	empirical	empirical	ADJ
dad-10683	736	8	methods	method	NOUN
dad-10683	736	9	in	in	ADP
dad-10683	736	10	natural	natural	ADJ
dad-10683	736	11	language	language	NOUN
dad-10683	736	12	processing	processing	NOUN
dad-10683	736	13	,	,	PUNCT
dad-10683	736	14	2014	2014	NUM
dad-10683	736	15	.	.	PUNCT
dad-10683	737	1	j.	j.	PROPN
dad-10683	737	2	deng	deng	PROPN
dad-10683	737	3	,	,	PUNCT
dad-10683	737	4	w.	w.	PROPN
dad-10683	737	5	dong	dong	PROPN
dad-10683	737	6	,	,	PUNCT
dad-10683	737	7	r.	r.	PROPN
dad-10683	737	8	socher	socher	PROPN
dad-10683	737	9	,	,	PUNCT
dad-10683	737	10	l.j	l.j	PROPN
dad-10683	737	11	.	.	PROPN
dad-10683	737	12	li	li	PROPN
dad-10683	737	13	,	,	PUNCT
dad-10683	737	14	k.	k.	PROPN
dad-10683	737	15	li	li	PROPN
dad-10683	737	16	,	,	PUNCT
dad-10683	737	17	and	and	CCONJ
dad-10683	737	18	l.	l.	PROPN
dad-10683	737	19	fei	fei	PROPN
dad-10683	737	20	-	-	PUNCT
dad-10683	737	21	fei	fei	PROPN
dad-10683	737	22	.	.	PUNCT
dad-10683	738	1	imagenet	imagenet	PROPN
dad-10683	738	2	:	:	PUNCT
dad-10683	738	3	a	a	DET
dad-10683	738	4	large	large	ADJ
dad-10683	738	5	-	-	PUNCT
dad-10683	738	6	scale	scale	NOUN
dad-10683	738	7	hierarchical	hierarchical	ADJ
dad-10683	738	8	image	image	NOUN
dad-10683	738	9	database	database	NOUN
dad-10683	738	10	.	.	PUNCT
dad-10683	739	1	in	in	ADP
dad-10683	739	2	proceedings	proceeding	NOUN
dad-10683	739	3	of	of	ADP
dad-10683	739	4	the	the	DET
dad-10683	739	5	conference	conference	NOUN
dad-10683	739	6	on	on	ADP
dad-10683	739	7	computer	computer	NOUN
dad-10683	739	8	vision	vision	NOUN
dad-10683	739	9	and	and	CCONJ
dad-10683	739	10	pattern	pattern	NOUN
dad-10683	739	11	recognition	recognition	NOUN
dad-10683	739	12	,	,	PUNCT
dad-10683	739	13	2009	2009	NUM
dad-10683	739	14	.	.	PUNCT
dad-10683	740	1	j.	j.	PROPN
dad-10683	740	2	dodge	dodge	PROPN
dad-10683	740	3	,	,	PUNCT
dad-10683	740	4	a.	a.	NOUN
dad-10683	740	5	gane	gane	NOUN
dad-10683	740	6	,	,	PUNCT
dad-10683	740	7	x.	x.	PROPN
dad-10683	740	8	zhang	zhang	PROPN
dad-10683	740	9	,	,	PUNCT
dad-10683	740	10	a.	a.	PROPN
dad-10683	740	11	bordes	bordes	PROPN
dad-10683	740	12	,	,	PUNCT
dad-10683	740	13	s.	s.	PROPN
dad-10683	740	14	chopra	chopra	PROPN
dad-10683	740	15	,	,	PUNCT
dad-10683	740	16	a.	a.	PROPN
dad-10683	740	17	miller	miller	PROPN
dad-10683	740	18	,	,	PUNCT
dad-10683	740	19	a.	a.	PROPN
dad-10683	740	20	szlam	szlam	PROPN
dad-10683	740	21	,	,	PUNCT
dad-10683	740	22	and	and	CCONJ
dad-10683	740	23	j.	j.	PROPN
dad-10683	740	24	weston	weston	PROPN
dad-10683	740	25	.	.	PUNCT
dad-10683	741	1	evaluating	evaluate	VERB
dad-10683	741	2	prerequisite	prerequisite	NOUN
dad-10683	741	3	qualities	quality	NOUN
dad-10683	741	4	for	for	ADP
dad-10683	741	5	learning	learn	VERB
dad-10683	741	6	end	end	NOUN
dad-10683	741	7	-	-	PUNCT
dad-10683	741	8	to	to	ADP
dad-10683	741	9	-	-	PUNCT
dad-10683	741	10	end	end	NOUN
dad-10683	741	11	dialog	dialog	NOUN
dad-10683	741	12	systems	system	NOUN
dad-10683	741	13	.	.	PUNCT
dad-10683	742	1	in	in	ADP
dad-10683	742	2	proceedings	proceeding	NOUN
dad-10683	742	3	of	of	ADP
dad-10683	742	4	the	the	DET
dad-10683	742	5	international	international	ADJ
dad-10683	742	6	conference	conference	NOUN
dad-10683	742	7	on	on	ADP
dad-10683	742	8	learning	learn	VERB
dad-10683	742	9	representations	representation	NOUN
dad-10683	742	10	,	,	PUNCT
dad-10683	742	11	2015	2015	NUM
dad-10683	742	12	.	.	PUNCT
dad-10683	743	1	61	61	NUM
dad-10683	743	2	lowe	lowe	PROPN
dad-10683	743	3	,	,	PUNCT
dad-10683	743	4	pow	pow	PROPN
dad-10683	743	5	,	,	PUNCT
dad-10683	743	6	serban	serban	PROPN
dad-10683	743	7	,	,	PUNCT
dad-10683	743	8	charlinn	charlinn	PROPN
dad-10683	743	9	,	,	PUNCT
dad-10683	743	10	liu	liu	PROPN
dad-10683	743	11	and	and	CCONJ
dad-10683	743	12	pineau	pineau	PROPN
dad-10683	743	13	j.	j.	PROPN
dad-10683	743	14	l.	l.	PROPN
dad-10683	743	15	elman	elman	PROPN
dad-10683	743	16	.	.	PUNCT
dad-10683	743	17	finding	find	VERB
dad-10683	743	18	structure	structure	NOUN
dad-10683	743	19	in	in	ADP
dad-10683	743	20	time	time	NOUN
dad-10683	743	21	.	.	PUNCT
dad-10683	744	1	cognitive	cognitive	ADJ
dad-10683	744	2	science	science	NOUN
dad-10683	744	3	,	,	PUNCT
dad-10683	744	4	14(2):179–211	14(2):179–211	NUM
dad-10683	744	5	,	,	PUNCT
dad-10683	744	6	1990	1990	NUM
dad-10683	744	7	.	.	PUNCT
dad-10683	745	1	m.	m.	NOUN
dad-10683	745	2	elsner	elsner	NOUN
dad-10683	745	3	and	and	CCONJ
dad-10683	745	4	e.	e.	PROPN
dad-10683	745	5	charniak	charniak	PROPN
dad-10683	745	6	.	.	PUNCT
dad-10683	746	1	you	you	PRON
dad-10683	746	2	talking	talk	VERB
dad-10683	746	3	to	to	ADP
dad-10683	746	4	me	i	PRON
dad-10683	746	5	?	?	PUNCT
dad-10683	747	1	a	a	DET
dad-10683	747	2	corpus	corpus	NOUN
dad-10683	747	3	and	and	CCONJ
dad-10683	747	4	algorithm	algorithm	NOUN
dad-10683	747	5	for	for	ADP
dad-10683	747	6	conversation	conversation	NOUN
dad-10683	747	7	disentanglement	disentanglement	NOUN
dad-10683	747	8	.	.	PUNCT
dad-10683	748	1	in	in	ADP
dad-10683	748	2	proceedings	proceeding	NOUN
dad-10683	748	3	of	of	ADP
dad-10683	748	4	the	the	DET
dad-10683	748	5	annual	annual	ADJ
dad-10683	748	6	meeting	meeting	NOUN
dad-10683	748	7	of	of	ADP
dad-10683	748	8	the	the	DET
dad-10683	748	9	association	association	NOUN
dad-10683	748	10	for	for	ADP
dad-10683	748	11	computational	computational	ADJ
dad-10683	748	12	linguistics	linguistic	NOUN
dad-10683	748	13	,	,	PUNCT
dad-10683	748	14	2008	2008	NUM
dad-10683	748	15	.	.	PUNCT
dad-10683	749	1	g.	g.	PROPN
dad-10683	749	2	forgues	forgues	PROPN
dad-10683	749	3	,	,	PUNCT
dad-10683	749	4	j.	j.	PROPN
dad-10683	749	5	pineau	pineau	PROPN
dad-10683	749	6	,	,	PUNCT
dad-10683	749	7	j.m	j.m	PROPN
dad-10683	749	8	.	.	PROPN
dad-10683	749	9	larchevêque	larchevêque	PROPN
dad-10683	749	10	,	,	PUNCT
dad-10683	749	11	and	and	CCONJ
dad-10683	749	12	r.	r.	PROPN
dad-10683	749	13	tremblay	tremblay	PROPN
dad-10683	749	14	.	.	PUNCT
dad-10683	750	1	bootstrapping	bootstrappe	VERB
dad-10683	750	2	dialog	dialog	NOUN
dad-10683	750	3	systems	system	NOUN
dad-10683	750	4	with	with	ADP
dad-10683	750	5	word	word	NOUN
dad-10683	750	6	embeddings	embedding	NOUN
dad-10683	750	7	.	.	PUNCT
dad-10683	751	1	in	in	ADP
dad-10683	751	2	proceedings	proceeding	NOUN
dad-10683	751	3	of	of	ADP
dad-10683	751	4	the	the	DET
dad-10683	751	5	workshop	workshop	NOUN
dad-10683	751	6	on	on	ADP
dad-10683	751	7	modern	modern	ADJ
dad-10683	751	8	machine	machine	NOUN
dad-10683	751	9	learning	learning	NOUN
dad-10683	751	10	and	and	CCONJ
dad-10683	751	11	nlp	nlp	NOUN
dad-10683	751	12	,	,	PUNCT
dad-10683	751	13	nips	nip	NOUN
dad-10683	751	14	,	,	PUNCT
dad-10683	751	15	2014	2014	NUM
dad-10683	751	16	.	.	PUNCT
dad-10683	752	1	m.	m.	NOUN
dad-10683	752	2	galley	galley	NOUN
dad-10683	752	3	,	,	PUNCT
dad-10683	752	4	c.	c.	PROPN
dad-10683	752	5	brockett	brockett	PROPN
dad-10683	752	6	,	,	PUNCT
dad-10683	752	7	a.	a.	PROPN
dad-10683	752	8	sordoni	sordoni	PROPN
dad-10683	752	9	,	,	PUNCT
dad-10683	752	10	y.	y.	PROPN
dad-10683	752	11	ji	ji	PROPN
dad-10683	752	12	,	,	PUNCT
dad-10683	752	13	m.	m.	NOUN
dad-10683	752	14	auli	auli	PROPN
dad-10683	752	15	,	,	PUNCT
dad-10683	752	16	c.	c.	PROPN
dad-10683	752	17	quirk	quirk	PROPN
dad-10683	752	18	,	,	PUNCT
dad-10683	752	19	m.	m.	PROPN
dad-10683	752	20	mitchell	mitchell	PROPN
dad-10683	752	21	,	,	PUNCT
dad-10683	752	22	j.	j.	PROPN
dad-10683	752	23	gao	gao	PROPN
dad-10683	752	24	,	,	PUNCT
dad-10683	752	25	and	and	CCONJ
dad-10683	752	26	b.	b.	PROPN
dad-10683	752	27	dolan	dolan	PROPN
dad-10683	752	28	.	.	PUNCT
dad-10683	753	1	deltableu	deltableu	PROPN
dad-10683	753	2	:	:	PUNCT
dad-10683	753	3	a	a	DET
dad-10683	753	4	discriminative	discriminative	NOUN
dad-10683	753	5	metric	metric	ADJ
dad-10683	753	6	for	for	ADP
dad-10683	753	7	generation	generation	NOUN
dad-10683	753	8	tasks	task	NOUN
dad-10683	753	9	with	with	ADP
dad-10683	753	10	intrinsically	intrinsically	ADV
dad-10683	753	11	diverse	diverse	ADJ
dad-10683	753	12	targets	target	NOUN
dad-10683	753	13	.	.	PUNCT
dad-10683	754	1	arxiv	arxiv	PROPN
dad-10683	754	2	preprint	preprint	PROPN
dad-10683	754	3	arxiv:1506.06863	arxiv:1506.06863	PROPN
dad-10683	754	4	,	,	PUNCT
dad-10683	754	5	2015	2015	NUM
dad-10683	754	6	.	.	PUNCT
dad-10683	755	1	j.j	j.j	PROPN
dad-10683	755	2	.	.	PROPN
dad-10683	755	3	godfrey	godfrey	PROPN
dad-10683	755	4	,	,	PUNCT
dad-10683	755	5	e.c	e.c	PROPN
dad-10683	755	6	.	.	PROPN
dad-10683	755	7	holliman	holliman	PROPN
dad-10683	755	8	,	,	PUNCT
dad-10683	755	9	and	and	CCONJ
dad-10683	755	10	j.	j.	PROPN
dad-10683	755	11	mcdaniel	mcdaniel	PROPN
dad-10683	755	12	.	.	PROPN
dad-10683	755	13	switchboard	switchboard	PROPN
dad-10683	755	14	:	:	PUNCT
dad-10683	755	15	telephone	telephone	NOUN
dad-10683	755	16	speech	speech	NOUN
dad-10683	755	17	corpus	corpus	NOUN
dad-10683	755	18	for	for	ADP
dad-10683	755	19	research	research	NOUN
dad-10683	755	20	and	and	CCONJ
dad-10683	755	21	development	development	NOUN
dad-10683	755	22	.	.	PUNCT
dad-10683	756	1	in	in	ADP
dad-10683	756	2	proceedings	proceeding	NOUN
dad-10683	756	3	of	of	ADP
dad-10683	756	4	the	the	DET
dad-10683	756	5	international	international	ADJ
dad-10683	756	6	conference	conference	NOUN
dad-10683	756	7	on	on	ADP
dad-10683	756	8	acoustics	acoustic	NOUN
dad-10683	756	9	,	,	PUNCT
dad-10683	756	10	speech	speech	NOUN
dad-10683	756	11	and	and	CCONJ
dad-10683	756	12	signal	signal	NOUN
dad-10683	756	13	processing	processing	NOUN
dad-10683	756	14	,	,	PUNCT
dad-10683	756	15	1992	1992	NUM
dad-10683	756	16	.	.	PUNCT
dad-10683	757	1	helen	helen	PROPN
dad-10683	757	2	hastie	hastie	PROPN
dad-10683	757	3	.	.	PUNCT
dad-10683	758	1	metrics	metric	NOUN
dad-10683	758	2	and	and	CCONJ
dad-10683	758	3	evaluation	evaluation	NOUN
dad-10683	758	4	of	of	ADP
dad-10683	758	5	spoken	spoken	ADJ
dad-10683	758	6	dialogue	dialogue	NOUN
dad-10683	758	7	systems	system	NOUN
dad-10683	758	8	.	.	PUNCT
dad-10683	759	1	in	in	ADP
dad-10683	759	2	data	data	NOUN
dad-10683	759	3	-	-	PUNCT
dad-10683	759	4	driven	drive	VERB
dad-10683	759	5	methods	method	NOUN
dad-10683	759	6	for	for	ADP
dad-10683	759	7	adaptive	adaptive	ADJ
dad-10683	759	8	spoken	spoken	ADJ
dad-10683	759	9	dialogue	dialogue	NOUN
dad-10683	759	10	systems	system	NOUN
dad-10683	759	11	,	,	PUNCT
dad-10683	759	12	pages	page	NOUN
dad-10683	759	13	131–150	131–150	NUM
dad-10683	759	14	.	.	PUNCT
dad-10683	759	15	springer	springer	NOUN
dad-10683	759	16	,	,	PUNCT
dad-10683	759	17	2012	2012	NUM
dad-10683	759	18	.	.	PUNCT
dad-10683	760	1	m.	m.	PROPN
dad-10683	760	2	henderson	henderson	PROPN
dad-10683	760	3	,	,	PUNCT
dad-10683	760	4	b.	b.	PROPN
dad-10683	760	5	thomson	thomson	PROPN
dad-10683	760	6	,	,	PUNCT
dad-10683	760	7	and	and	CCONJ
dad-10683	760	8	j.	j.	PROPN
dad-10683	760	9	williams	williams	PROPN
dad-10683	760	10	.	.	PUNCT
dad-10683	761	1	the	the	DET
dad-10683	761	2	second	second	ADJ
dad-10683	761	3	dialog	dialog	NOUN
dad-10683	761	4	state	state	NOUN
dad-10683	761	5	tracking	tracking	NOUN
dad-10683	761	6	challenge	challenge	NOUN
dad-10683	761	7	.	.	PUNCT
dad-10683	762	1	in	in	ADP
dad-10683	762	2	proceedings	proceeding	NOUN
dad-10683	762	3	of	of	ADP
dad-10683	762	4	the	the	DET
dad-10683	762	5	meeting	meeting	NOUN
dad-10683	762	6	of	of	ADP
dad-10683	762	7	the	the	DET
dad-10683	762	8	special	special	ADJ
dad-10683	762	9	interest	interest	NOUN
dad-10683	762	10	group	group	NOUN
dad-10683	762	11	on	on	ADP
dad-10683	762	12	dialogue	dialogue	NOUN
dad-10683	762	13	and	and	CCONJ
dad-10683	762	14	discourse	discourse	NOUN
dad-10683	762	15	(	(	PUNCT
dad-10683	762	16	sigdial	sigdial	ADJ
dad-10683	762	17	)	)	PUNCT
dad-10683	762	18	,	,	PUNCT
dad-10683	762	19	page	page	NOUN
dad-10683	762	20	263	263	NUM
dad-10683	762	21	,	,	PUNCT
dad-10683	762	22	2014a	2014a	NUM
dad-10683	762	23	.	.	PUNCT
dad-10683	763	1	m.	m.	PROPN
dad-10683	763	2	henderson	henderson	PROPN
dad-10683	763	3	,	,	PUNCT
dad-10683	763	4	b.	b.	PROPN
dad-10683	763	5	thomson	thomson	PROPN
dad-10683	763	6	,	,	PUNCT
dad-10683	763	7	and	and	CCONJ
dad-10683	763	8	s.	s.	PROPN
dad-10683	763	9	young	young	PROPN
dad-10683	763	10	.	.	PUNCT
dad-10683	764	1	word	word	NOUN
dad-10683	764	2	-	-	PUNCT
dad-10683	764	3	based	base	VERB
dad-10683	764	4	dialog	dialog	NOUN
dad-10683	764	5	state	state	NOUN
dad-10683	764	6	tracking	tracking	NOUN
dad-10683	764	7	with	with	ADP
dad-10683	764	8	recurrent	recurrent	ADJ
dad-10683	764	9	neural	neural	ADJ
dad-10683	764	10	networks	network	NOUN
dad-10683	764	11	.	.	PUNCT
dad-10683	765	1	in	in	ADP
dad-10683	765	2	proceedings	proceeding	NOUN
dad-10683	765	3	of	of	ADP
dad-10683	765	4	the	the	DET
dad-10683	765	5	meeting	meeting	NOUN
dad-10683	765	6	of	of	ADP
dad-10683	765	7	the	the	DET
dad-10683	765	8	special	special	ADJ
dad-10683	765	9	interest	interest	NOUN
dad-10683	765	10	group	group	NOUN
dad-10683	765	11	on	on	ADP
dad-10683	765	12	dialogue	dialogue	NOUN
dad-10683	765	13	and	and	CCONJ
dad-10683	765	14	discourse	discourse	NOUN
dad-10683	765	15	(	(	PUNCT
dad-10683	765	16	sigdial	sigdial	ADJ
dad-10683	765	17	)	)	PUNCT
dad-10683	765	18	,	,	PUNCT
dad-10683	765	19	2014b	2014b	NUM
dad-10683	765	20	.	.	PUNCT
dad-10683	766	1	k.	k.	PROPN
dad-10683	766	2	hermann	hermann	PROPN
dad-10683	766	3	,	,	PUNCT
dad-10683	766	4	t.	t.	PROPN
dad-10683	766	5	kocisky	kocisky	PROPN
dad-10683	766	6	,	,	PUNCT
dad-10683	766	7	e.	e.	PROPN
dad-10683	766	8	grefenstette	grefenstette	PROPN
dad-10683	766	9	,	,	PUNCT
dad-10683	766	10	l.	l.	PROPN
dad-10683	766	11	espeholt	espeholt	PROPN
dad-10683	766	12	,	,	PUNCT
dad-10683	766	13	w.	w.	PROPN
dad-10683	766	14	kay	kay	PROPN
dad-10683	766	15	,	,	PUNCT
dad-10683	766	16	m.	m.	NOUN
dad-10683	766	17	suleyman	suleyman	NOUN
dad-10683	766	18	,	,	PUNCT
dad-10683	766	19	and	and	CCONJ
dad-10683	766	20	p.	p.	NOUN
dad-10683	766	21	blunsom	blunsom	NOUN
dad-10683	766	22	.	.	PUNCT
dad-10683	767	1	teaching	teaching	NOUN
dad-10683	767	2	machines	machine	NOUN
dad-10683	767	3	to	to	PART
dad-10683	767	4	read	read	VERB
dad-10683	767	5	and	and	CCONJ
dad-10683	767	6	comprehend	comprehend	VERB
dad-10683	767	7	.	.	PUNCT
dad-10683	768	1	in	in	ADP
dad-10683	768	2	proceedings	proceeding	NOUN
dad-10683	768	3	of	of	ADP
dad-10683	768	4	the	the	DET
dad-10683	768	5	annual	annual	ADJ
dad-10683	768	6	conference	conference	NOUN
dad-10683	768	7	on	on	ADP
dad-10683	768	8	neural	neural	ADJ
dad-10683	768	9	information	information	NOUN
dad-10683	768	10	processing	processing	NOUN
dad-10683	768	11	systems	system	NOUN
dad-10683	768	12	,	,	PUNCT
dad-10683	768	13	2015	2015	NUM
dad-10683	768	14	.	.	PUNCT
dad-10683	769	1	s.	s.	PROPN
dad-10683	769	2	hochreiter	hochreiter	PROPN
dad-10683	769	3	and	and	CCONJ
dad-10683	769	4	j.	j.	PROPN
dad-10683	769	5	schmidhuber	schmidhuber	PROPN
dad-10683	769	6	.	.	PUNCT
dad-10683	770	1	long	long	ADJ
dad-10683	770	2	short	short	ADJ
dad-10683	770	3	-	-	PUNCT
dad-10683	770	4	term	term	NOUN
dad-10683	770	5	memory	memory	NOUN
dad-10683	770	6	.	.	PUNCT
dad-10683	771	1	neural	neural	ADJ
dad-10683	771	2	computation	computation	NOUN
dad-10683	771	3	,	,	PUNCT
dad-10683	771	4	9(8):1735–1780	9(8):1735–1780	NUM
dad-10683	771	5	,	,	PUNCT
dad-10683	771	6	1997	1997	NUM
dad-10683	771	7	.	.	PUNCT
dad-10683	772	1	michimasa	michimasa	PROPN
dad-10683	772	2	inaba	inaba	PROPN
dad-10683	772	3	and	and	CCONJ
dad-10683	772	4	kenichi	kenichi	PROPN
dad-10683	772	5	takahashi	takahashi	PROPN
dad-10683	772	6	.	.	PUNCT
dad-10683	773	1	neural	neural	ADJ
dad-10683	773	2	utterance	utterance	VERB
dad-10683	773	3	ranking	rank	VERB
dad-10683	773	4	model	model	NOUN
dad-10683	773	5	for	for	ADP
dad-10683	773	6	conversational	conversational	ADJ
dad-10683	773	7	dialogue	dialogue	NOUN
dad-10683	773	8	systems	system	NOUN
dad-10683	773	9	.	.	PUNCT
dad-10683	774	1	in	in	ADP
dad-10683	774	2	proceedings	proceeding	NOUN
dad-10683	774	3	of	of	ADP
dad-10683	774	4	the	the	DET
dad-10683	774	5	meeting	meeting	NOUN
dad-10683	774	6	of	of	ADP
dad-10683	774	7	the	the	DET
dad-10683	774	8	special	special	ADJ
dad-10683	774	9	interest	interest	NOUN
dad-10683	774	10	group	group	NOUN
dad-10683	774	11	on	on	ADP
dad-10683	774	12	discourse	discourse	NOUN
dad-10683	774	13	and	and	CCONJ
dad-10683	774	14	dialogue	dialogue	NOUN
dad-10683	774	15	(	(	PUNCT
dad-10683	774	16	sigdial	sigdial	ADJ
dad-10683	774	17	)	)	PUNCT
dad-10683	774	18	,	,	PUNCT
dad-10683	774	19	2016	2016	NUM
dad-10683	774	20	.	.	PUNCT
dad-10683	775	1	s.	s.	PROPN
dad-10683	775	2	jafarpour	jafarpour	PROPN
dad-10683	775	3	,	,	PUNCT
dad-10683	775	4	c.	c.	PROPN
dad-10683	775	5	burges	burge	NOUN
dad-10683	775	6	,	,	PUNCT
dad-10683	775	7	and	and	CCONJ
dad-10683	775	8	a.	a.	NOUN
dad-10683	775	9	ritter	ritter	PROPN
dad-10683	775	10	.	.	PUNCT
dad-10683	776	1	filter	filter	PROPN
dad-10683	776	2	,	,	PUNCT
dad-10683	776	3	rank	rank	NOUN
dad-10683	776	4	,	,	PUNCT
dad-10683	776	5	and	and	CCONJ
dad-10683	776	6	transfer	transfer	VERB
dad-10683	776	7	the	the	DET
dad-10683	776	8	knowledge	knowledge	NOUN
dad-10683	776	9	:	:	PUNCT
dad-10683	776	10	learning	learn	VERB
dad-10683	776	11	to	to	PART
dad-10683	776	12	chat	chat	VERB
dad-10683	776	13	.	.	PUNCT
dad-10683	777	1	advances	advance	NOUN
dad-10683	777	2	in	in	ADP
dad-10683	777	3	ranking	ranking	ADJ
dad-10683	777	4	,	,	PUNCT
dad-10683	777	5	10	10	NUM
dad-10683	777	6	,	,	PUNCT
dad-10683	777	7	2010	2010	NUM
dad-10683	777	8	.	.	PUNCT
dad-10683	778	1	k.	k.	PROPN
dad-10683	778	2	jokinen	jokinen	PROPN
dad-10683	778	3	and	and	CCONJ
dad-10683	778	4	m.	m.	NOUN
dad-10683	778	5	mctear	mctear	PROPN
dad-10683	778	6	.	.	PUNCT
dad-10683	779	1	spoken	speak	VERB
dad-10683	779	2	dialogue	dialogue	NOUN
dad-10683	779	3	systems	system	NOUN
dad-10683	779	4	.	.	PUNCT
dad-10683	780	1	morgan	morgan	PROPN
dad-10683	780	2	claypool	claypool	PROPN
dad-10683	780	3	,	,	PUNCT
dad-10683	780	4	2009	2009	NUM
dad-10683	780	5	.	.	PUNCT
dad-10683	781	1	r.	r.	PROPN
dad-10683	781	2	kadlec	kadlec	PROPN
dad-10683	781	3	,	,	PUNCT
dad-10683	781	4	m.	m.	NOUN
dad-10683	781	5	schmid	schmid	PROPN
dad-10683	781	6	,	,	PUNCT
dad-10683	781	7	and	and	CCONJ
dad-10683	781	8	j.	j.	PROPN
dad-10683	781	9	kleindienst	kleindienst	PROPN
dad-10683	781	10	.	.	PUNCT
dad-10683	781	11	improved	improve	VERB
dad-10683	781	12	deep	deep	ADJ
dad-10683	781	13	learning	learn	VERB
dad-10683	781	14	baselines	baseline	NOUN
dad-10683	781	15	for	for	ADP
dad-10683	781	16	ubuntu	ubuntu	ADJ
dad-10683	781	17	corpus	corpus	NOUN
dad-10683	781	18	dialogs	dialog	NOUN
dad-10683	781	19	.	.	PUNCT
dad-10683	782	1	in	in	ADP
dad-10683	782	2	proceedings	proceeding	NOUN
dad-10683	782	3	of	of	ADP
dad-10683	782	4	the	the	DET
dad-10683	782	5	workshop	workshop	NOUN
dad-10683	782	6	on	on	ADP
dad-10683	782	7	spoken	spoken	ADJ
dad-10683	782	8	language	language	NOUN
dad-10683	782	9	understanding	understanding	NOUN
dad-10683	782	10	,	,	PUNCT
dad-10683	782	11	nips	nip	NOUN
dad-10683	782	12	,	,	PUNCT
dad-10683	782	13	2015	2015	NUM
dad-10683	782	14	.	.	PUNCT
dad-10683	783	1	d.p	d.p	PROPN
dad-10683	783	2	.	.	PROPN
dad-10683	783	3	kingma	kingma	PROPN
dad-10683	783	4	and	and	CCONJ
dad-10683	783	5	j.	j.	PROPN
dad-10683	783	6	ba	ba	PROPN
dad-10683	783	7	.	.	PUNCT
dad-10683	784	1	adam	adam	PROPN
dad-10683	784	2	:	:	PUNCT
dad-10683	784	3	a	a	DET
dad-10683	784	4	method	method	NOUN
dad-10683	784	5	for	for	ADP
dad-10683	784	6	stochastic	stochastic	ADJ
dad-10683	784	7	optimization	optimization	NOUN
dad-10683	784	8	.	.	PUNCT
dad-10683	785	1	proceedings	proceeding	NOUN
dad-10683	785	2	of	of	ADP
dad-10683	785	3	the	the	DET
dad-10683	785	4	international	international	ADJ
dad-10683	785	5	conference	conference	NOUN
dad-10683	785	6	on	on	ADP
dad-10683	785	7	learning	learn	VERB
dad-10683	785	8	representations	representation	NOUN
dad-10683	785	9	,	,	PUNCT
dad-10683	785	10	2014	2014	NUM
dad-10683	785	11	.	.	PUNCT
dad-10683	786	1	62	62	NUM
dad-10683	786	2	training	training	NOUN
dad-10683	786	3	end	end	NOUN
dad-10683	786	4	-	-	PUNCT
dad-10683	786	5	to	to	ADP
dad-10683	786	6	-	-	PUNCT
dad-10683	786	7	end	end	NOUN
dad-10683	786	8	dialogue	dialogue	NOUN
dad-10683	786	9	systems	systems	PROPN
dad-10683	786	10	cheongjae	cheongjae	PROPN
dad-10683	786	11	lee	lee	PROPN
dad-10683	786	12	,	,	PUNCT
dad-10683	786	13	sangkeun	sangkeun	PROPN
dad-10683	786	14	jung	jung	PROPN
dad-10683	786	15	,	,	PUNCT
dad-10683	786	16	seokhwan	seokhwan	PROPN
dad-10683	786	17	kim	kim	PROPN
dad-10683	786	18	,	,	PUNCT
dad-10683	786	19	and	and	CCONJ
dad-10683	786	20	gary	gary	PROPN
dad-10683	786	21	geunbae	geunbae	PROPN
dad-10683	786	22	lee	lee	PROPN
dad-10683	786	23	.	.	PROPN
dad-10683	786	24	example	example	NOUN
dad-10683	786	25	-	-	PUNCT
dad-10683	786	26	based	base	VERB
dad-10683	786	27	dialog	dialog	NOUN
dad-10683	786	28	modeling	modeling	NOUN
dad-10683	786	29	for	for	ADP
dad-10683	786	30	practical	practical	ADJ
dad-10683	786	31	multi	multi	ADJ
dad-10683	786	32	-	-	ADJ
dad-10683	786	33	domain	domain	ADJ
dad-10683	786	34	dialog	dialog	NOUN
dad-10683	786	35	system	system	NOUN
dad-10683	786	36	.	.	PUNCT
dad-10683	787	1	speech	speech	NOUN
dad-10683	787	2	communication	communication	NOUN
dad-10683	787	3	,	,	PUNCT
dad-10683	787	4	51(5):466–484	51(5):466–484	PROPN
dad-10683	787	5	,	,	PUNCT
dad-10683	787	6	2009	2009	NUM
dad-10683	787	7	.	.	PUNCT
dad-10683	788	1	j.	j.	PROPN
dad-10683	788	2	li	li	PROPN
dad-10683	788	3	,	,	PUNCT
dad-10683	788	4	m.	m.	NOUN
dad-10683	788	5	galley	galley	NOUN
dad-10683	788	6	,	,	PUNCT
dad-10683	788	7	c.	c.	PROPN
dad-10683	788	8	brockett	brockett	PROPN
dad-10683	788	9	,	,	PUNCT
dad-10683	788	10	j.	j.	PROPN
dad-10683	788	11	gao	gao	PROPN
dad-10683	788	12	,	,	PUNCT
dad-10683	788	13	and	and	CCONJ
dad-10683	788	14	b.	b.	PROPN
dad-10683	788	15	dolan	dolan	PROPN
dad-10683	788	16	.	.	PUNCT
dad-10683	789	1	a	a	DET
dad-10683	789	2	diversity	diversity	NOUN
dad-10683	789	3	-	-	PUNCT
dad-10683	789	4	promoting	promote	VERB
dad-10683	789	5	objective	objective	ADJ
dad-10683	789	6	function	function	NOUN
dad-10683	789	7	for	for	ADP
dad-10683	789	8	neural	neural	ADJ
dad-10683	789	9	conversation	conversation	NOUN
dad-10683	789	10	models	model	NOUN
dad-10683	789	11	.	.	PUNCT
dad-10683	790	1	in	in	ADP
dad-10683	790	2	proceedings	proceeding	NOUN
dad-10683	790	3	of	of	ADP
dad-10683	790	4	the	the	DET
dad-10683	790	5	meeting	meeting	NOUN
dad-10683	790	6	of	of	ADP
dad-10683	790	7	the	the	DET
dad-10683	790	8	north	north	ADJ
dad-10683	790	9	american	american	ADJ
dad-10683	790	10	chapter	chapter	NOUN
dad-10683	790	11	of	of	ADP
dad-10683	790	12	the	the	DET
dad-10683	790	13	association	association	NOUN
dad-10683	790	14	for	for	ADP
dad-10683	790	15	computational	computational	ADJ
dad-10683	790	16	linguistics	linguistic	NOUN
dad-10683	790	17	,	,	PUNCT
dad-10683	790	18	2015	2015	NUM
dad-10683	790	19	.	.	PUNCT
dad-10683	791	1	j.	j.	PROPN
dad-10683	791	2	li	li	PROPN
dad-10683	791	3	,	,	PUNCT
dad-10683	791	4	m.	m.	NOUN
dad-10683	791	5	galley	galley	NOUN
dad-10683	791	6	,	,	PUNCT
dad-10683	791	7	c.	c.	PROPN
dad-10683	791	8	brockett	brockett	PROPN
dad-10683	791	9	,	,	PUNCT
dad-10683	791	10	j.	j.	PROPN
dad-10683	791	11	gao	gao	PROPN
dad-10683	791	12	,	,	PUNCT
dad-10683	791	13	and	and	CCONJ
dad-10683	791	14	b.	b.	PROPN
dad-10683	791	15	dolan	dolan	PROPN
dad-10683	791	16	.	.	PUNCT
dad-10683	792	1	a	a	DET
dad-10683	792	2	persona	persona	NOUN
dad-10683	792	3	-	-	PUNCT
dad-10683	792	4	based	base	VERB
dad-10683	792	5	neural	neural	ADJ
dad-10683	792	6	conversation	conversation	NOUN
dad-10683	792	7	model	model	NOUN
dad-10683	792	8	.	.	PUNCT
dad-10683	793	1	in	in	ADP
dad-10683	793	2	proceedings	proceeding	NOUN
dad-10683	793	3	of	of	ADP
dad-10683	793	4	the	the	DET
dad-10683	793	5	association	association	NOUN
dad-10683	793	6	for	for	ADP
dad-10683	793	7	computational	computational	ADJ
dad-10683	793	8	linguistics	linguistic	NOUN
dad-10683	793	9	,	,	PUNCT
dad-10683	793	10	2016a	2016a	NUM
dad-10683	793	11	.	.	PUNCT
dad-10683	794	1	j.	j.	PROPN
dad-10683	794	2	li	li	PROPN
dad-10683	794	3	,	,	PUNCT
dad-10683	794	4	w.	w.	PROPN
dad-10683	794	5	monroe	monroe	PROPN
dad-10683	794	6	,	,	PUNCT
dad-10683	794	7	a.	a.	NOUN
dad-10683	794	8	ritter	ritter	PROPN
dad-10683	794	9	,	,	PUNCT
dad-10683	794	10	and	and	CCONJ
dad-10683	794	11	d.	d.	PROPN
dad-10683	794	12	jurafsky	jurafsky	PROPN
dad-10683	794	13	.	.	PUNCT
dad-10683	795	1	deep	deep	ADJ
dad-10683	795	2	reinforcement	reinforcement	NOUN
dad-10683	795	3	learning	learning	NOUN
dad-10683	795	4	for	for	ADP
dad-10683	795	5	dialogue	dialogue	NOUN
dad-10683	795	6	generation	generation	NOUN
dad-10683	795	7	.	.	PUNCT
dad-10683	796	1	in	in	ADP
dad-10683	796	2	proceedings	proceeding	NOUN
dad-10683	796	3	of	of	ADP
dad-10683	796	4	the	the	DET
dad-10683	796	5	conference	conference	NOUN
dad-10683	796	6	on	on	ADP
dad-10683	796	7	empirical	empirical	ADJ
dad-10683	796	8	methods	method	NOUN
dad-10683	796	9	in	in	ADP
dad-10683	796	10	natural	natural	ADJ
dad-10683	796	11	language	language	NOUN
dad-10683	796	12	processing	processing	NOUN
dad-10683	796	13	,	,	PUNCT
dad-10683	796	14	2016b	2016b	NUM
dad-10683	796	15	.	.	PUNCT
dad-10683	797	1	c.-w	c.-w	PROPN
dad-10683	797	2	.	.	PUNCT
dad-10683	798	1	liu	liu	PROPN
dad-10683	798	2	,	,	PUNCT
dad-10683	798	3	r.	r.	PROPN
dad-10683	798	4	lowe	lowe	PROPN
dad-10683	798	5	,	,	PUNCT
dad-10683	798	6	i.v	i.v	PROPN
dad-10683	798	7	.	.	PROPN
dad-10683	798	8	serban	serban	PROPN
dad-10683	798	9	,	,	PUNCT
dad-10683	798	10	m.	m.	NOUN
dad-10683	798	11	noseworthy	noseworthy	PROPN
dad-10683	798	12	,	,	PUNCT
dad-10683	798	13	l.	l.	PROPN
dad-10683	798	14	charlin	charlin	PROPN
dad-10683	798	15	,	,	PUNCT
dad-10683	798	16	and	and	CCONJ
dad-10683	798	17	j.	j.	PROPN
dad-10683	798	18	pineau	pineau	PROPN
dad-10683	798	19	.	.	PUNCT
dad-10683	799	1	how	how	SCONJ
dad-10683	799	2	not	not	PART
dad-10683	799	3	to	to	PART
dad-10683	799	4	evaluate	evaluate	VERB
dad-10683	799	5	your	your	PRON
dad-10683	799	6	dialogue	dialogue	NOUN
dad-10683	799	7	system	system	NOUN
dad-10683	799	8	:	:	PUNCT
dad-10683	799	9	an	an	DET
dad-10683	799	10	empirical	empirical	ADJ
dad-10683	799	11	study	study	NOUN
dad-10683	799	12	of	of	ADP
dad-10683	799	13	unsupervised	unsupervised	ADJ
dad-10683	799	14	evaluation	evaluation	NOUN
dad-10683	799	15	metrics	metric	NOUN
dad-10683	799	16	for	for	ADP
dad-10683	799	17	dialogue	dialogue	NOUN
dad-10683	799	18	response	response	NOUN
dad-10683	799	19	generation	generation	NOUN
dad-10683	799	20	.	.	PUNCT
dad-10683	800	1	in	in	ADP
dad-10683	800	2	proceedings	proceeding	NOUN
dad-10683	800	3	of	of	ADP
dad-10683	800	4	the	the	DET
dad-10683	800	5	conference	conference	NOUN
dad-10683	800	6	on	on	ADP
dad-10683	800	7	empirical	empirical	ADJ
dad-10683	800	8	methods	method	NOUN
dad-10683	800	9	in	in	ADP
dad-10683	800	10	natural	natural	ADJ
dad-10683	800	11	language	language	NOUN
dad-10683	800	12	processing	processing	NOUN
dad-10683	800	13	,	,	PUNCT
dad-10683	800	14	2016	2016	NUM
dad-10683	800	15	.	.	PUNCT
dad-10683	801	1	r.	r.	PROPN
dad-10683	801	2	lowe	lowe	PROPN
dad-10683	801	3	,	,	PUNCT
dad-10683	801	4	n.	n.	PROPN
dad-10683	801	5	pow	pow	PROPN
dad-10683	801	6	,	,	PUNCT
dad-10683	801	7	i.	i.	PROPN
dad-10683	801	8	serban	serban	PROPN
dad-10683	801	9	,	,	PUNCT
dad-10683	801	10	and	and	CCONJ
dad-10683	801	11	j.	j.	PROPN
dad-10683	801	12	pineau	pineau	PROPN
dad-10683	801	13	.	.	PUNCT
dad-10683	802	1	the	the	DET
dad-10683	802	2	ubuntu	ubuntu	ADJ
dad-10683	802	3	dialogue	dialogue	NOUN
dad-10683	802	4	corpus	corpus	PROPN
dad-10683	802	5	:	:	PUNCT
dad-10683	802	6	a	a	DET
dad-10683	802	7	large	large	ADJ
dad-10683	802	8	dataset	dataset	NOUN
dad-10683	802	9	for	for	ADP
dad-10683	802	10	research	research	NOUN
dad-10683	802	11	in	in	ADP
dad-10683	802	12	unstructured	unstructured	ADJ
dad-10683	802	13	multi	multi	ADJ
dad-10683	802	14	-	-	ADJ
dad-10683	802	15	turn	turn	ADJ
dad-10683	802	16	dialogue	dialogue	NOUN
dad-10683	802	17	systems	system	NOUN
dad-10683	802	18	.	.	PUNCT
dad-10683	803	1	in	in	ADP
dad-10683	803	2	proceedings	proceeding	NOUN
dad-10683	803	3	of	of	ADP
dad-10683	803	4	the	the	DET
dad-10683	803	5	meeting	meeting	NOUN
dad-10683	803	6	of	of	ADP
dad-10683	803	7	the	the	DET
dad-10683	803	8	special	special	ADJ
dad-10683	803	9	interest	interest	NOUN
dad-10683	803	10	group	group	NOUN
dad-10683	803	11	on	on	ADP
dad-10683	803	12	dialogue	dialogue	NOUN
dad-10683	803	13	and	and	CCONJ
dad-10683	803	14	discourse	discourse	NOUN
dad-10683	803	15	(	(	PUNCT
dad-10683	803	16	sigdial	sigdial	ADJ
dad-10683	803	17	)	)	PUNCT
dad-10683	803	18	,	,	PUNCT
dad-10683	803	19	2015	2015	NUM
dad-10683	803	20	.	.	PUNCT
dad-10683	804	1	r.	r.	PROPN
dad-10683	804	2	lowe	lowe	PROPN
dad-10683	804	3	,	,	PUNCT
dad-10683	804	4	i.	i.	PROPN
dad-10683	804	5	serban	serban	PROPN
dad-10683	804	6	,	,	PUNCT
dad-10683	804	7	m.	m.	NOUN
dad-10683	804	8	noseworthy	noseworthy	PROPN
dad-10683	804	9	,	,	PUNCT
dad-10683	804	10	l.	l.	PROPN
dad-10683	804	11	charlin	charlin	PROPN
dad-10683	804	12	,	,	PUNCT
dad-10683	804	13	and	and	CCONJ
dad-10683	804	14	j.	j.	PROPN
dad-10683	804	15	pineau	pineau	PROPN
dad-10683	804	16	.	.	PUNCT
dad-10683	805	1	on	on	ADP
dad-10683	805	2	the	the	DET
dad-10683	805	3	evaluation	evaluation	NOUN
dad-10683	805	4	of	of	ADP
dad-10683	805	5	dialogue	dialogue	NOUN
dad-10683	805	6	systems	system	NOUN
dad-10683	805	7	with	with	ADP
dad-10683	805	8	next	next	ADJ
dad-10683	805	9	utterance	utterance	NOUN
dad-10683	805	10	classification	classification	NOUN
dad-10683	805	11	.	.	PUNCT
dad-10683	806	1	in	in	ADP
dad-10683	806	2	proceedings	proceeding	NOUN
dad-10683	806	3	of	of	ADP
dad-10683	806	4	the	the	DET
dad-10683	806	5	meeting	meeting	NOUN
dad-10683	806	6	of	of	ADP
dad-10683	806	7	the	the	DET
dad-10683	806	8	special	special	ADJ
dad-10683	806	9	interest	interest	NOUN
dad-10683	806	10	group	group	NOUN
dad-10683	806	11	on	on	ADP
dad-10683	806	12	dialogue	dialogue	NOUN
dad-10683	806	13	and	and	CCONJ
dad-10683	806	14	discourse	discourse	NOUN
dad-10683	806	15	(	(	PUNCT
dad-10683	806	16	sigdial	sigdial	ADJ
dad-10683	806	17	)	)	PUNCT
dad-10683	806	18	,	,	PUNCT
dad-10683	806	19	2016	2016	NUM
dad-10683	806	20	.	.	PUNCT
dad-10683	807	1	y.	y.	PROPN
dad-10683	807	2	luan	luan	PROPN
dad-10683	807	3	,	,	PUNCT
dad-10683	807	4	y.	y.	PROPN
dad-10683	807	5	ji	ji	PROPN
dad-10683	807	6	,	,	PUNCT
dad-10683	807	7	and	and	CCONJ
dad-10683	807	8	m.	m.	PROPN
dad-10683	807	9	ostendorf	ostendorf	PROPN
dad-10683	807	10	.	.	PUNCT
dad-10683	808	1	lstm	lstm	PROPN
dad-10683	808	2	based	base	VERB
dad-10683	808	3	conversation	conversation	NOUN
dad-10683	808	4	models	model	NOUN
dad-10683	808	5	.	.	PUNCT
dad-10683	809	1	arxiv	arxiv	PROPN
dad-10683	809	2	preprint	preprint	PROPN
dad-10683	809	3	arxiv:1603.09457	arxiv:1603.09457	PROPN
dad-10683	809	4	,	,	PUNCT
dad-10683	809	5	2016	2016	NUM
dad-10683	809	6	.	.	PUNCT
dad-10683	810	1	t.	t.	PROPN
dad-10683	810	2	mikolov	mikolov	PROPN
dad-10683	810	3	,	,	PUNCT
dad-10683	810	4	m.	m.	NOUN
dad-10683	810	5	karafiát	karafiát	PROPN
dad-10683	810	6	,	,	PUNCT
dad-10683	810	7	l.	l.	PROPN
dad-10683	810	8	burget	burget	PROPN
dad-10683	810	9	,	,	PUNCT
dad-10683	810	10	j.	j.	PROPN
dad-10683	810	11	cernockỳ	cernockỳ	PROPN
dad-10683	810	12	,	,	PUNCT
dad-10683	810	13	and	and	CCONJ
dad-10683	810	14	s.	s.	PROPN
dad-10683	810	15	khudanpur	khudanpur	PROPN
dad-10683	810	16	.	.	PUNCT
dad-10683	811	1	recurrent	recurrent	ADJ
dad-10683	811	2	neural	neural	ADJ
dad-10683	811	3	network	network	NOUN
dad-10683	811	4	based	base	VERB
dad-10683	811	5	language	language	NOUN
dad-10683	811	6	model	model	NOUN
dad-10683	811	7	.	.	PUNCT
dad-10683	812	1	in	in	ADP
dad-10683	812	2	proceedings	proceeding	NOUN
dad-10683	812	3	of	of	ADP
dad-10683	812	4	interspeech	interspeech	NOUN
dad-10683	812	5	,	,	PUNCT
dad-10683	812	6	2010	2010	NUM
dad-10683	812	7	.	.	PUNCT
dad-10683	813	1	t.	t.	PROPN
dad-10683	813	2	mikolov	mikolov	PROPN
dad-10683	813	3	,	,	PUNCT
dad-10683	813	4	i.	i.	PROPN
dad-10683	813	5	sutskever	sutskever	PROPN
dad-10683	813	6	,	,	PUNCT
dad-10683	813	7	k.	k.	PROPN
dad-10683	813	8	chen	chen	PROPN
dad-10683	813	9	,	,	PUNCT
dad-10683	813	10	g.	g.	PROPN
dad-10683	813	11	s.	s.	PROPN
dad-10683	813	12	corrado	corrado	PROPN
dad-10683	813	13	,	,	PUNCT
dad-10683	813	14	and	and	CCONJ
dad-10683	813	15	j.	j.	PROPN
dad-10683	813	16	dean	dean	PROPN
dad-10683	813	17	.	.	PUNCT
dad-10683	814	1	distributed	distribute	VERB
dad-10683	814	2	representations	representation	NOUN
dad-10683	814	3	of	of	ADP
dad-10683	814	4	words	word	NOUN
dad-10683	814	5	and	and	CCONJ
dad-10683	814	6	phrases	phrase	NOUN
dad-10683	814	7	and	and	CCONJ
dad-10683	814	8	their	their	PRON
dad-10683	814	9	compositionality	compositionality	NOUN
dad-10683	814	10	.	.	PUNCT
dad-10683	815	1	in	in	ADP
dad-10683	815	2	proceedings	proceeding	NOUN
dad-10683	815	3	of	of	ADP
dad-10683	815	4	the	the	DET
dad-10683	815	5	annual	annual	ADJ
dad-10683	815	6	conference	conference	NOUN
dad-10683	815	7	on	on	ADP
dad-10683	815	8	neural	neural	ADJ
dad-10683	815	9	information	information	NOUN
dad-10683	815	10	processing	processing	NOUN
dad-10683	815	11	systems	system	NOUN
dad-10683	815	12	,	,	PUNCT
dad-10683	815	13	2013	2013	NUM
dad-10683	815	14	.	.	PUNCT
dad-10683	816	1	s.	s.	PROPN
dad-10683	816	2	möller	möller	PROPN
dad-10683	816	3	,	,	PUNCT
dad-10683	816	4	r.	r.	PROPN
dad-10683	816	5	englert	englert	PROPN
dad-10683	816	6	,	,	PUNCT
dad-10683	816	7	k.	k.	PROPN
dad-10683	816	8	engelbrecht	engelbrecht	PROPN
dad-10683	816	9	,	,	PUNCT
dad-10683	816	10	v.	v.	PROPN
dad-10683	816	11	hafner	hafner	PROPN
dad-10683	816	12	,	,	PUNCT
dad-10683	816	13	a.	a.	PROPN
dad-10683	816	14	jameson	jameson	PROPN
dad-10683	816	15	,	,	PUNCT
dad-10683	816	16	a.	a.	NOUN
dad-10683	816	17	oulasvirta	oulasvirta	PROPN
dad-10683	816	18	,	,	PUNCT
dad-10683	816	19	a.	a.	NOUN
dad-10683	816	20	raake	raake	NOUN
dad-10683	816	21	,	,	PUNCT
dad-10683	816	22	and	and	CCONJ
dad-10683	816	23	n.	n.	NOUN
dad-10683	816	24	reithinger	reithinger	NOUN
dad-10683	816	25	.	.	PUNCT
dad-10683	817	1	memo	memo	NOUN
dad-10683	817	2	:	:	PUNCT
dad-10683	817	3	towards	towards	ADP
dad-10683	817	4	automatic	automatic	ADJ
dad-10683	817	5	usability	usability	NOUN
dad-10683	817	6	evaluation	evaluation	NOUN
dad-10683	817	7	of	of	ADP
dad-10683	817	8	spoken	spoken	ADJ
dad-10683	817	9	dialogue	dialogue	NOUN
dad-10683	817	10	services	service	NOUN
dad-10683	817	11	by	by	ADP
dad-10683	817	12	user	user	NOUN
dad-10683	817	13	error	error	NOUN
dad-10683	817	14	simulations	simulation	NOUN
dad-10683	817	15	.	.	PUNCT
dad-10683	818	1	in	in	ADP
dad-10683	818	2	proceedings	proceeding	NOUN
dad-10683	818	3	of	of	ADP
dad-10683	818	4	interspeech	interspeech	NOUN
dad-10683	818	5	,	,	PUNCT
dad-10683	818	6	2006	2006	NUM
dad-10683	818	7	.	.	PUNCT
dad-10683	819	1	lasguido	lasguido	PROPN
dad-10683	819	2	nio	nio	PROPN
dad-10683	819	3	,	,	PUNCT
dad-10683	819	4	sakriani	sakriani	PROPN
dad-10683	819	5	sakti	sakti	PROPN
dad-10683	819	6	,	,	PUNCT
dad-10683	819	7	graham	graham	PROPN
dad-10683	819	8	neubig	neubig	PROPN
dad-10683	819	9	,	,	PUNCT
dad-10683	819	10	tomoki	tomoki	NOUN
dad-10683	819	11	toda	toda	PROPN
dad-10683	819	12	,	,	PUNCT
dad-10683	819	13	mirna	mirna	PROPN
dad-10683	819	14	adriani	adriani	PROPN
dad-10683	819	15	,	,	PUNCT
dad-10683	819	16	and	and	CCONJ
dad-10683	819	17	satoshi	satoshi	PROPN
dad-10683	819	18	nakamura	nakamura	PROPN
dad-10683	819	19	.	.	PUNCT
dad-10683	820	1	developing	develop	VERB
dad-10683	820	2	non	non	ADJ
dad-10683	820	3	-	-	ADJ
dad-10683	820	4	goal	goal	ADJ
dad-10683	820	5	dialog	dialog	NOUN
dad-10683	820	6	system	system	NOUN
dad-10683	820	7	based	base	VERB
dad-10683	820	8	on	on	ADP
dad-10683	820	9	examples	example	NOUN
dad-10683	820	10	of	of	ADP
dad-10683	820	11	drama	drama	NOUN
dad-10683	820	12	television	television	NOUN
dad-10683	820	13	.	.	PUNCT
dad-10683	821	1	in	in	ADP
dad-10683	821	2	natural	natural	ADJ
dad-10683	821	3	interaction	interaction	NOUN
dad-10683	821	4	with	with	ADP
dad-10683	821	5	robots	robot	NOUN
dad-10683	821	6	,	,	PUNCT
dad-10683	821	7	knowbots	knowbot	NOUN
dad-10683	821	8	and	and	CCONJ
dad-10683	821	9	smartphones	smartphone	NOUN
dad-10683	821	10	,	,	PUNCT
dad-10683	821	11	pages	page	NOUN
dad-10683	821	12	355–361	355–361	NUM
dad-10683	821	13	.	.	PUNCT
dad-10683	821	14	springer	springer	NOUN
dad-10683	821	15	,	,	PUNCT
dad-10683	821	16	2014	2014	NUM
dad-10683	821	17	.	.	PUNCT
dad-10683	822	1	j.	j.	PROPN
dad-10683	822	2	pennington	pennington	PROPN
dad-10683	822	3	,	,	PUNCT
dad-10683	822	4	r.	r.	PROPN
dad-10683	822	5	socher	socher	PROPN
dad-10683	822	6	,	,	PUNCT
dad-10683	822	7	and	and	CCONJ
dad-10683	822	8	c.d	c.d	PROPN
dad-10683	822	9	.	.	PROPN
dad-10683	822	10	manning	manning	PROPN
dad-10683	822	11	.	.	PUNCT
dad-10683	823	1	glove	glove	NOUN
dad-10683	823	2	:	:	PUNCT
dad-10683	823	3	global	global	ADJ
dad-10683	823	4	vectors	vector	NOUN
dad-10683	823	5	for	for	ADP
dad-10683	823	6	word	word	NOUN
dad-10683	823	7	representation	representation	NOUN
dad-10683	823	8	.	.	PUNCT
dad-10683	824	1	in	in	ADP
dad-10683	824	2	proceedings	proceeding	NOUN
dad-10683	824	3	of	of	ADP
dad-10683	824	4	the	the	DET
dad-10683	824	5	conference	conference	NOUN
dad-10683	824	6	on	on	ADP
dad-10683	824	7	empirical	empirical	ADJ
dad-10683	824	8	methods	method	NOUN
dad-10683	824	9	in	in	ADP
dad-10683	824	10	natural	natural	ADJ
dad-10683	824	11	language	language	NOUN
dad-10683	824	12	processing	processing	NOUN
dad-10683	824	13	,	,	PUNCT
dad-10683	824	14	2014	2014	NUM
dad-10683	824	15	.	.	PUNCT
dad-10683	825	1	o.	o.	PROPN
dad-10683	825	2	pietquin	pietquin	PROPN
dad-10683	825	3	and	and	CCONJ
dad-10683	825	4	h.	h.	PROPN
dad-10683	825	5	hastie	hastie	PROPN
dad-10683	825	6	.	.	PUNCT
dad-10683	826	1	a	a	DET
dad-10683	826	2	survey	survey	NOUN
dad-10683	826	3	on	on	ADP
dad-10683	826	4	metrics	metric	NOUN
dad-10683	826	5	for	for	ADP
dad-10683	826	6	the	the	DET
dad-10683	826	7	evaluation	evaluation	NOUN
dad-10683	826	8	of	of	ADP
dad-10683	826	9	user	user	NOUN
dad-10683	826	10	simulations	simulation	NOUN
dad-10683	826	11	.	.	PUNCT
dad-10683	827	1	the	the	DET
dad-10683	827	2	knowledge	knowledge	PROPN
dad-10683	827	3	engineering	engineering	PROPN
dad-10683	827	4	review	review	PROPN
dad-10683	827	5	,	,	PUNCT
dad-10683	827	6	28(01):59–73	28(01):59–73	NUM
dad-10683	827	7	,	,	PUNCT
dad-10683	827	8	2013	2013	NUM
dad-10683	827	9	.	.	PUNCT
dad-10683	828	1	63	63	NUM
dad-10683	828	2	lowe	lowe	PROPN
dad-10683	828	3	,	,	PUNCT
dad-10683	828	4	pow	pow	PROPN
dad-10683	828	5	,	,	PUNCT
dad-10683	828	6	serban	serban	PROPN
dad-10683	828	7	,	,	PUNCT
dad-10683	828	8	charlinn	charlinn	PROPN
dad-10683	828	9	,	,	PUNCT
dad-10683	828	10	liu	liu	PROPN
dad-10683	828	11	and	and	CCONJ
dad-10683	828	12	pineau	pineau	PROPN
dad-10683	828	13	j.	j.	PROPN
dad-10683	828	14	ramos	ramos	PROPN
dad-10683	828	15	.	.	PUNCT
dad-10683	829	1	using	use	VERB
dad-10683	829	2	tf	tf	PROPN
dad-10683	829	3	-	-	PUNCT
dad-10683	829	4	idf	idf	NOUN
dad-10683	829	5	to	to	PART
dad-10683	829	6	determine	determine	VERB
dad-10683	829	7	word	word	NOUN
dad-10683	829	8	relevance	relevance	NOUN
dad-10683	829	9	in	in	ADP
dad-10683	829	10	document	document	NOUN
dad-10683	829	11	queries	query	NOUN
dad-10683	829	12	.	.	PUNCT
dad-10683	830	1	in	in	ADP
dad-10683	830	2	proceedings	proceeding	NOUN
dad-10683	830	3	of	of	ADP
dad-10683	830	4	the	the	DET
dad-10683	830	5	international	international	ADJ
dad-10683	830	6	conference	conference	NOUN
dad-10683	830	7	on	on	ADP
dad-10683	830	8	machine	machine	NOUN
dad-10683	830	9	learning	learning	NOUN
dad-10683	830	10	,	,	PUNCT
dad-10683	830	11	2003	2003	NUM
dad-10683	830	12	.	.	PUNCT
dad-10683	831	1	a.	a.	NOUN
dad-10683	831	2	ritter	ritter	PROPN
dad-10683	831	3	,	,	PUNCT
dad-10683	831	4	c.	c.	PROPN
dad-10683	831	5	cherry	cherry	PROPN
dad-10683	831	6	,	,	PUNCT
dad-10683	831	7	and	and	CCONJ
dad-10683	831	8	w.	w.	PROPN
dad-10683	831	9	dolan	dolan	PROPN
dad-10683	831	10	.	.	PUNCT
dad-10683	832	1	unsupervised	unsupervised	ADJ
dad-10683	832	2	modeling	modeling	NOUN
dad-10683	832	3	of	of	ADP
dad-10683	832	4	twitter	twitter	NOUN
dad-10683	832	5	conversations	conversation	NOUN
dad-10683	832	6	.	.	PUNCT
dad-10683	833	1	in	in	ADP
dad-10683	833	2	proceedings	proceeding	NOUN
dad-10683	833	3	of	of	ADP
dad-10683	833	4	the	the	DET
dad-10683	833	5	meeting	meeting	NOUN
dad-10683	833	6	of	of	ADP
dad-10683	833	7	the	the	DET
dad-10683	833	8	north	north	ADJ
dad-10683	833	9	american	american	ADJ
dad-10683	833	10	chapter	chapter	NOUN
dad-10683	833	11	of	of	ADP
dad-10683	833	12	the	the	DET
dad-10683	833	13	association	association	NOUN
dad-10683	833	14	for	for	ADP
dad-10683	833	15	computational	computational	ADJ
dad-10683	833	16	linguistics	linguistic	NOUN
dad-10683	833	17	,	,	PUNCT
dad-10683	833	18	2010	2010	NUM
dad-10683	833	19	.	.	PUNCT
dad-10683	834	1	a.	a.	NOUN
dad-10683	834	2	ritter	ritter	PROPN
dad-10683	834	3	,	,	PUNCT
dad-10683	834	4	c.	c.	PROPN
dad-10683	834	5	cherry	cherry	PROPN
dad-10683	834	6	,	,	PUNCT
dad-10683	834	7	and	and	CCONJ
dad-10683	834	8	w.	w.	PROPN
dad-10683	834	9	dolan	dolan	PROPN
dad-10683	834	10	.	.	PUNCT
dad-10683	835	1	data	data	NOUN
dad-10683	835	2	-	-	PUNCT
dad-10683	835	3	driven	drive	VERB
dad-10683	835	4	response	response	NOUN
dad-10683	835	5	generation	generation	NOUN
dad-10683	835	6	in	in	ADP
dad-10683	835	7	social	social	ADJ
dad-10683	835	8	media	medium	NOUN
dad-10683	835	9	.	.	PUNCT
dad-10683	836	1	in	in	ADP
dad-10683	836	2	proceedings	proceeding	NOUN
dad-10683	836	3	of	of	ADP
dad-10683	836	4	the	the	DET
dad-10683	836	5	conference	conference	NOUN
dad-10683	836	6	on	on	ADP
dad-10683	836	7	empirical	empirical	ADJ
dad-10683	836	8	methods	method	NOUN
dad-10683	836	9	in	in	ADP
dad-10683	836	10	natural	natural	ADJ
dad-10683	836	11	language	language	NOUN
dad-10683	836	12	processing	processing	NOUN
dad-10683	836	13	,	,	PUNCT
dad-10683	836	14	2011	2011	NUM
dad-10683	836	15	.	.	PUNCT
dad-10683	837	1	v.	v.	ADP
dad-10683	837	2	rus	rus	NOUN
dad-10683	837	3	and	and	CCONJ
dad-10683	837	4	m.	m.	NOUN
dad-10683	837	5	lintean	lintean	PROPN
dad-10683	837	6	.	.	PUNCT
dad-10683	838	1	a	a	DET
dad-10683	838	2	comparison	comparison	NOUN
dad-10683	838	3	of	of	ADP
dad-10683	838	4	greedy	greedy	ADJ
dad-10683	838	5	and	and	CCONJ
dad-10683	838	6	optimal	optimal	ADJ
dad-10683	838	7	assessment	assessment	NOUN
dad-10683	838	8	of	of	ADP
dad-10683	838	9	natural	natural	ADJ
dad-10683	838	10	language	language	NOUN
dad-10683	838	11	student	student	NOUN
dad-10683	838	12	input	input	NOUN
dad-10683	838	13	using	use	VERB
dad-10683	838	14	word	word	NOUN
dad-10683	838	15	-	-	PUNCT
dad-10683	838	16	to	to	ADP
dad-10683	838	17	-	-	PUNCT
dad-10683	838	18	word	word	NOUN
dad-10683	838	19	similarity	similarity	NOUN
dad-10683	838	20	metrics	metric	NOUN
dad-10683	838	21	.	.	PUNCT
dad-10683	839	1	in	in	ADP
dad-10683	839	2	proceedings	proceeding	NOUN
dad-10683	839	3	of	of	ADP
dad-10683	839	4	the	the	DET
dad-10683	839	5	seventh	seventh	ADJ
dad-10683	839	6	workshop	workshop	NOUN
dad-10683	839	7	on	on	ADP
dad-10683	839	8	building	build	VERB
dad-10683	839	9	educational	educational	ADJ
dad-10683	839	10	applications	application	NOUN
dad-10683	839	11	using	use	VERB
dad-10683	839	12	nlp	nlp	NOUN
dad-10683	839	13	,	,	PUNCT
dad-10683	839	14	2012	2012	NUM
dad-10683	839	15	.	.	PUNCT
dad-10683	840	1	a.m.	a.m.	PROPN
dad-10683	840	2	saxe	saxe	PROPN
dad-10683	840	3	,	,	PUNCT
dad-10683	840	4	j.l	j.l	PROPN
dad-10683	840	5	.	.	PROPN
dad-10683	840	6	mcclelland	mcclelland	PROPN
dad-10683	840	7	,	,	PUNCT
dad-10683	840	8	and	and	CCONJ
dad-10683	840	9	s.	s.	PROPN
dad-10683	840	10	ganguli	ganguli	PROPN
dad-10683	840	11	.	.	PUNCT
dad-10683	841	1	exact	exact	ADJ
dad-10683	841	2	solutions	solution	NOUN
dad-10683	841	3	to	to	ADP
dad-10683	841	4	the	the	DET
dad-10683	841	5	nonlinear	nonlinear	ADJ
dad-10683	841	6	dynamics	dynamic	NOUN
dad-10683	841	7	of	of	ADP
dad-10683	841	8	learning	learn	VERB
dad-10683	841	9	in	in	ADP
dad-10683	841	10	deep	deep	ADJ
dad-10683	841	11	linear	linear	ADJ
dad-10683	841	12	neural	neural	ADJ
dad-10683	841	13	networks	network	NOUN
dad-10683	841	14	.	.	PUNCT
dad-10683	842	1	arxiv	arxiv	PROPN
dad-10683	842	2	preprint	preprint	PROPN
dad-10683	842	3	arxiv:1312.6120	arxiv:1312.6120	NUM
dad-10683	842	4	,	,	PUNCT
dad-10683	842	5	2013	2013	NUM
dad-10683	842	6	.	.	PUNCT
dad-10683	843	1	j.	j.	PROPN
dad-10683	843	2	schatzmann	schatzmann	PROPN
dad-10683	843	3	,	,	PUNCT
dad-10683	843	4	k.	k.	PROPN
dad-10683	843	5	georgila	georgila	PROPN
dad-10683	843	6	,	,	PUNCT
dad-10683	843	7	and	and	CCONJ
dad-10683	843	8	s.	s.	PROPN
dad-10683	843	9	young	young	PROPN
dad-10683	843	10	.	.	PUNCT
dad-10683	844	1	quantitative	quantitative	ADJ
dad-10683	844	2	evaluation	evaluation	NOUN
dad-10683	844	3	of	of	ADP
dad-10683	844	4	user	user	NOUN
dad-10683	844	5	simulation	simulation	NOUN
dad-10683	844	6	techniques	technique	NOUN
dad-10683	844	7	for	for	ADP
dad-10683	844	8	spoken	spoken	ADJ
dad-10683	844	9	dialogue	dialogue	NOUN
dad-10683	844	10	systems	system	NOUN
dad-10683	844	11	.	.	PUNCT
dad-10683	845	1	in	in	ADP
dad-10683	845	2	proceedings	proceeding	NOUN
dad-10683	845	3	of	of	ADP
dad-10683	845	4	the	the	DET
dad-10683	845	5	meeting	meeting	NOUN
dad-10683	845	6	of	of	ADP
dad-10683	845	7	the	the	DET
dad-10683	845	8	special	special	ADJ
dad-10683	845	9	interest	interest	NOUN
dad-10683	845	10	group	group	NOUN
dad-10683	845	11	on	on	ADP
dad-10683	845	12	dialogue	dialogue	NOUN
dad-10683	845	13	and	and	CCONJ
dad-10683	845	14	discourse	discourse	NOUN
dad-10683	845	15	(	(	PUNCT
dad-10683	845	16	sigdial	sigdial	ADJ
dad-10683	845	17	)	)	PUNCT
dad-10683	845	18	,	,	PUNCT
dad-10683	845	19	2005	2005	NUM
dad-10683	845	20	.	.	PUNCT
dad-10683	846	1	i.	i.	PROPN
dad-10683	846	2	v.	v.	PROPN
dad-10683	846	3	serban	serban	PROPN
dad-10683	846	4	,	,	PUNCT
dad-10683	846	5	r.	r.	PROPN
dad-10683	846	6	lowe	lowe	PROPN
dad-10683	846	7	,	,	PUNCT
dad-10683	846	8	l.	l.	PROPN
dad-10683	846	9	charlin	charlin	PROPN
dad-10683	846	10	,	,	PUNCT
dad-10683	846	11	and	and	CCONJ
dad-10683	846	12	j.	j.	PROPN
dad-10683	846	13	pineau	pineau	PROPN
dad-10683	846	14	.	.	PUNCT
dad-10683	847	1	a	a	DET
dad-10683	847	2	survey	survey	NOUN
dad-10683	847	3	of	of	ADP
dad-10683	847	4	available	available	ADJ
dad-10683	847	5	corpora	corpora	PROPN
dad-10683	847	6	for	for	ADP
dad-10683	847	7	building	build	VERB
dad-10683	847	8	datadriven	datadriven	ADJ
dad-10683	847	9	dialogue	dialogue	NOUN
dad-10683	847	10	systems	system	NOUN
dad-10683	847	11	.	.	PUNCT
dad-10683	848	1	arxiv	arxiv	PROPN
dad-10683	848	2	preprint	preprint	VERB
dad-10683	848	3	arxiv:1512.05742	arxiv:1512.05742	PROPN
dad-10683	848	4	,	,	PUNCT
dad-10683	848	5	2015	2015	NUM
dad-10683	848	6	.	.	PUNCT
dad-10683	849	1	i.	i.	PROPN
dad-10683	849	2	v.	v.	PROPN
dad-10683	849	3	serban	serban	PROPN
dad-10683	849	4	,	,	PUNCT
dad-10683	849	5	a.	a.	PROPN
dad-10683	849	6	sordoni	sordoni	PROPN
dad-10683	849	7	,	,	PUNCT
dad-10683	849	8	y.	y.	PROPN
dad-10683	849	9	bengio	bengio	PROPN
dad-10683	849	10	,	,	PUNCT
dad-10683	849	11	a.	a.	PROPN
dad-10683	849	12	c.	c.	PROPN
dad-10683	849	13	courville	courville	PROPN
dad-10683	849	14	,	,	PUNCT
dad-10683	849	15	and	and	CCONJ
dad-10683	849	16	j.	j.	PROPN
dad-10683	849	17	pineau	pineau	PROPN
dad-10683	849	18	.	.	PUNCT
dad-10683	850	1	building	build	VERB
dad-10683	850	2	end	end	NOUN
dad-10683	850	3	-	-	PUNCT
dad-10683	850	4	to	to	ADP
dad-10683	850	5	-	-	PUNCT
dad-10683	850	6	end	end	NOUN
dad-10683	850	7	dialogue	dialogue	NOUN
dad-10683	850	8	systems	system	NOUN
dad-10683	850	9	using	use	VERB
dad-10683	850	10	generative	generative	ADJ
dad-10683	850	11	hierarchical	hierarchical	ADJ
dad-10683	850	12	neural	neural	ADJ
dad-10683	850	13	network	network	NOUN
dad-10683	850	14	models	model	NOUN
dad-10683	850	15	.	.	PUNCT
dad-10683	851	1	in	in	ADP
dad-10683	851	2	proceedings	proceeding	NOUN
dad-10683	851	3	of	of	ADP
dad-10683	851	4	the	the	DET
dad-10683	851	5	thirtieth	thirtieth	ADJ
dad-10683	851	6	aaai	aaai	PROPN
dad-10683	851	7	conference	conference	NOUN
dad-10683	851	8	on	on	ADP
dad-10683	851	9	artificial	artificial	ADJ
dad-10683	851	10	intelligence	intelligence	NOUN
dad-10683	851	11	,	,	PUNCT
dad-10683	851	12	2016	2016	NUM
dad-10683	851	13	.	.	PUNCT
dad-10683	852	1	l.	l.	PROPN
dad-10683	852	2	shang	shang	PROPN
dad-10683	852	3	,	,	PUNCT
dad-10683	852	4	z.	z.	PROPN
dad-10683	852	5	lu	lu	PROPN
dad-10683	852	6	,	,	PUNCT
dad-10683	852	7	and	and	CCONJ
dad-10683	852	8	h.	h.	PROPN
dad-10683	852	9	li	li	PROPN
dad-10683	852	10	.	.	PROPN
dad-10683	852	11	neural	neural	ADJ
dad-10683	852	12	responding	respond	VERB
dad-10683	852	13	machine	machine	NOUN
dad-10683	852	14	for	for	ADP
dad-10683	852	15	short	short	ADJ
dad-10683	852	16	-	-	PUNCT
dad-10683	852	17	text	text	NOUN
dad-10683	852	18	conversation	conversation	NOUN
dad-10683	852	19	.	.	PUNCT
dad-10683	853	1	arxiv	arxiv	PROPN
dad-10683	853	2	preprint	preprint	PROPN
dad-10683	853	3	arxiv:1503.02364	arxiv:1503.02364	NOUN
dad-10683	853	4	,	,	PUNCT
dad-10683	853	5	2015	2015	NUM
dad-10683	853	6	.	.	PUNCT
dad-10683	854	1	s.	s.	PROPN
dad-10683	854	2	singh	singh	PROPN
dad-10683	854	3	,	,	PUNCT
dad-10683	854	4	d.	d.	PROPN
dad-10683	854	5	litman	litman	PROPN
dad-10683	854	6	,	,	PUNCT
dad-10683	854	7	m.	m.	NOUN
dad-10683	854	8	kearns	kearn	NOUN
dad-10683	854	9	,	,	PUNCT
dad-10683	854	10	and	and	CCONJ
dad-10683	854	11	m.	m.	PROPN
dad-10683	854	12	walker	walker	PROPN
dad-10683	854	13	.	.	PUNCT
dad-10683	855	1	optimizing	optimize	VERB
dad-10683	855	2	dialogue	dialogue	NOUN
dad-10683	855	3	management	management	NOUN
dad-10683	855	4	with	with	ADP
dad-10683	855	5	reinforcement	reinforcement	NOUN
dad-10683	855	6	learning	learning	NOUN
dad-10683	855	7	:	:	PUNCT
dad-10683	855	8	experiments	experiment	NOUN
dad-10683	855	9	with	with	ADP
dad-10683	855	10	the	the	DET
dad-10683	855	11	njfun	njfun	NOUN
dad-10683	855	12	system	system	NOUN
dad-10683	855	13	.	.	PUNCT
dad-10683	856	1	journal	journal	NOUN
dad-10683	856	2	of	of	ADP
dad-10683	856	3	artificial	artificial	ADJ
dad-10683	856	4	intelligence	intelligence	NOUN
dad-10683	856	5	research	research	NOUN
dad-10683	856	6	,	,	PUNCT
dad-10683	856	7	16:105–133	16:105–133	NUM
dad-10683	856	8	,	,	PUNCT
dad-10683	856	9	2002	2002	NUM
dad-10683	856	10	.	.	PUNCT
dad-10683	857	1	a.	a.	PROPN
dad-10683	857	2	sordoni	sordoni	PROPN
dad-10683	857	3	,	,	PUNCT
dad-10683	857	4	y.	y.	PROPN
dad-10683	857	5	bengio	bengio	PROPN
dad-10683	857	6	,	,	PUNCT
dad-10683	857	7	h.	h.	PROPN
dad-10683	857	8	vahabi	vahabi	PROPN
dad-10683	857	9	,	,	PUNCT
dad-10683	857	10	c.	c.	PROPN
dad-10683	857	11	lioma	lioma	NOUN
dad-10683	857	12	,	,	PUNCT
dad-10683	857	13	jakob	jakob	PROPN
dad-10683	857	14	grue	grue	PROPN
dad-10683	857	15	s.	s.	PROPN
dad-10683	857	16	,	,	PUNCT
dad-10683	857	17	and	and	CCONJ
dad-10683	857	18	j.	j.	PROPN
dad-10683	857	19	y.	y.	PROPN
dad-10683	857	20	nie	nie	PROPN
dad-10683	857	21	.	.	PUNCT
dad-10683	858	1	a	a	DET
dad-10683	858	2	hierarchical	hierarchical	ADJ
dad-10683	858	3	recurrent	recurrent	ADJ
dad-10683	858	4	encoder	encoder	NOUN
dad-10683	858	5	-	-	PUNCT
dad-10683	858	6	decoder	decoder	NOUN
dad-10683	858	7	for	for	ADP
dad-10683	858	8	generative	generative	ADJ
dad-10683	858	9	context	context	NOUN
dad-10683	858	10	-	-	PUNCT
dad-10683	858	11	aware	aware	ADJ
dad-10683	858	12	query	query	NOUN
dad-10683	858	13	suggestion	suggestion	NOUN
dad-10683	858	14	.	.	PUNCT
dad-10683	859	1	in	in	ADP
dad-10683	859	2	proceedings	proceeding	NOUN
dad-10683	859	3	of	of	ADP
dad-10683	859	4	the	the	DET
dad-10683	859	5	acm	acm	PROPN
dad-10683	859	6	international	international	PROPN
dad-10683	859	7	on	on	ADP
dad-10683	859	8	conference	conference	NOUN
dad-10683	859	9	on	on	ADP
dad-10683	859	10	information	information	NOUN
dad-10683	859	11	and	and	CCONJ
dad-10683	859	12	knowledge	knowledge	NOUN
dad-10683	859	13	management	management	NOUN
dad-10683	859	14	,	,	PUNCT
dad-10683	859	15	pages	page	NOUN
dad-10683	859	16	553–562	553–562	NUM
dad-10683	859	17	,	,	PUNCT
dad-10683	859	18	2015a	2015a	NUM
dad-10683	859	19	.	.	PUNCT
dad-10683	860	1	a.	a.	PROPN
dad-10683	860	2	sordoni	sordoni	PROPN
dad-10683	860	3	,	,	PUNCT
dad-10683	860	4	m.	m.	NOUN
dad-10683	860	5	galley	galley	NOUN
dad-10683	860	6	,	,	PUNCT
dad-10683	860	7	m.	m.	NOUN
dad-10683	860	8	auli	auli	PROPN
dad-10683	860	9	,	,	PUNCT
dad-10683	860	10	c.	c.	PROPN
dad-10683	860	11	brockett	brockett	PROPN
dad-10683	860	12	,	,	PUNCT
dad-10683	860	13	y.	y.	PROPN
dad-10683	860	14	ji	ji	PROPN
dad-10683	860	15	,	,	PUNCT
dad-10683	860	16	m.	m.	PROPN
dad-10683	860	17	mitchell	mitchell	PROPN
dad-10683	860	18	,	,	PUNCT
dad-10683	860	19	j.y	j.y	PROPN
dad-10683	860	20	.	.	PROPN
dad-10683	860	21	nie	nie	PROPN
dad-10683	860	22	,	,	PUNCT
dad-10683	860	23	j.	j.	PROPN
dad-10683	860	24	gao	gao	PROPN
dad-10683	860	25	,	,	PUNCT
dad-10683	860	26	and	and	CCONJ
dad-10683	860	27	w.	w.	PROPN
dad-10683	860	28	dolan	dolan	PROPN
dad-10683	860	29	.	.	PUNCT
dad-10683	861	1	a	a	DET
dad-10683	861	2	neural	neural	ADJ
dad-10683	861	3	network	network	NOUN
dad-10683	861	4	approach	approach	NOUN
dad-10683	861	5	to	to	ADP
dad-10683	861	6	context	context	NOUN
dad-10683	861	7	-	-	PUNCT
dad-10683	861	8	sensitive	sensitive	ADJ
dad-10683	861	9	generation	generation	NOUN
dad-10683	861	10	of	of	ADP
dad-10683	861	11	conversational	conversational	ADJ
dad-10683	861	12	responses	response	NOUN
dad-10683	861	13	.	.	PUNCT
dad-10683	862	1	in	in	ADP
dad-10683	862	2	proceedings	proceeding	NOUN
dad-10683	862	3	of	of	ADP
dad-10683	862	4	the	the	DET
dad-10683	862	5	meeting	meeting	NOUN
dad-10683	862	6	of	of	ADP
dad-10683	862	7	the	the	DET
dad-10683	862	8	north	north	ADJ
dad-10683	862	9	american	american	ADJ
dad-10683	862	10	chapter	chapter	NOUN
dad-10683	862	11	of	of	ADP
dad-10683	862	12	the	the	DET
dad-10683	862	13	association	association	NOUN
dad-10683	862	14	for	for	ADP
dad-10683	862	15	computational	computational	ADJ
dad-10683	862	16	linguistics	linguistic	NOUN
dad-10683	862	17	,	,	PUNCT
dad-10683	862	18	2015b	2015b	NUM
dad-10683	862	19	.	.	PUNCT
dad-10683	863	1	i.	i.	PROPN
dad-10683	863	2	sutskever	sutskever	PROPN
dad-10683	863	3	,	,	PUNCT
dad-10683	863	4	o.	o.	PROPN
dad-10683	863	5	vinyals	vinyal	NOUN
dad-10683	863	6	,	,	PUNCT
dad-10683	863	7	and	and	CCONJ
dad-10683	863	8	q.	q.	PROPN
dad-10683	863	9	le	le	PROPN
dad-10683	863	10	.	.	PROPN
dad-10683	863	11	sequence	sequence	NOUN
dad-10683	863	12	to	to	ADP
dad-10683	863	13	sequence	sequence	NOUN
dad-10683	863	14	learning	learn	VERB
dad-10683	863	15	with	with	ADP
dad-10683	863	16	neural	neural	ADJ
dad-10683	863	17	networks	network	NOUN
dad-10683	863	18	.	.	PUNCT
dad-10683	864	1	in	in	ADP
dad-10683	864	2	proceedings	proceeding	NOUN
dad-10683	864	3	of	of	ADP
dad-10683	864	4	the	the	DET
dad-10683	864	5	annual	annual	ADJ
dad-10683	864	6	conference	conference	NOUN
dad-10683	864	7	on	on	ADP
dad-10683	864	8	neural	neural	ADJ
dad-10683	864	9	information	information	NOUN
dad-10683	864	10	processing	processing	NOUN
dad-10683	864	11	systems	system	NOUN
dad-10683	864	12	,	,	PUNCT
dad-10683	864	13	2014	2014	NUM
dad-10683	864	14	.	.	PUNCT
dad-10683	865	1	d.	d.	PROPN
dad-10683	865	2	traum	traum	PROPN
dad-10683	865	3	,	,	PUNCT
dad-10683	865	4	k.	k.	PROPN
dad-10683	865	5	georgila	georgila	PROPN
dad-10683	865	6	,	,	PUNCT
dad-10683	865	7	r.	r.	PROPN
dad-10683	865	8	artstein	artstein	ADV
dad-10683	865	9	,	,	PUNCT
dad-10683	865	10	and	and	CCONJ
dad-10683	865	11	a.	a.	NOUN
dad-10683	865	12	leuski	leuski	NOUN
dad-10683	865	13	.	.	PUNCT
dad-10683	866	1	evaluating	evaluate	VERB
dad-10683	866	2	spoken	spoken	ADJ
dad-10683	866	3	dialogue	dialogue	NOUN
dad-10683	866	4	processing	processing	NOUN
dad-10683	866	5	for	for	ADP
dad-10683	866	6	timeoffset	timeoffset	NOUN
dad-10683	866	7	interaction	interaction	NOUN
dad-10683	866	8	.	.	PUNCT
dad-10683	867	1	in	in	ADP
dad-10683	867	2	proceedings	proceeding	NOUN
dad-10683	867	3	of	of	ADP
dad-10683	867	4	the	the	DET
dad-10683	867	5	meeting	meeting	NOUN
dad-10683	867	6	of	of	ADP
dad-10683	867	7	the	the	DET
dad-10683	867	8	special	special	ADJ
dad-10683	867	9	interest	interest	NOUN
dad-10683	867	10	group	group	NOUN
dad-10683	867	11	on	on	ADP
dad-10683	867	12	dialogue	dialogue	NOUN
dad-10683	867	13	and	and	CCONJ
dad-10683	867	14	discourse	discourse	NOUN
dad-10683	867	15	(	(	PUNCT
dad-10683	867	16	sigdial	sigdial	ADJ
dad-10683	867	17	)	)	PUNCT
dad-10683	867	18	,	,	PUNCT
dad-10683	867	19	2015	2015	NUM
dad-10683	867	20	.	.	PUNCT
dad-10683	868	1	64	64	NUM
dad-10683	868	2	training	training	NOUN
dad-10683	868	3	end	end	NOUN
dad-10683	868	4	-	-	PUNCT
dad-10683	868	5	to	to	ADP
dad-10683	868	6	-	-	PUNCT
dad-10683	868	7	end	end	NOUN
dad-10683	868	8	dialogue	dialogue	NOUN
dad-10683	868	9	systems	systems	PROPN
dad-10683	868	10	d.c	d.c	PROPN
dad-10683	868	11	.	.	PUNCT
dad-10683	868	12	uthus	uthus	PROPN
dad-10683	868	13	and	and	CCONJ
dad-10683	868	14	d.w	d.w	PROPN
dad-10683	868	15	.	.	PROPN
dad-10683	868	16	aha	aha	INTJ
dad-10683	868	17	.	.	PUNCT
dad-10683	869	1	extending	extend	VERB
dad-10683	869	2	word	word	NOUN
dad-10683	869	3	highlighting	highlighting	NOUN
dad-10683	869	4	in	in	ADP
dad-10683	869	5	multiparticipant	multiparticipant	ADJ
dad-10683	869	6	chat	chat	PROPN
dad-10683	869	7	.	.	PUNCT
dad-10683	870	1	in	in	ADP
dad-10683	870	2	proceedings	proceeding	NOUN
dad-10683	870	3	of	of	ADP
dad-10683	870	4	the	the	DET
dad-10683	870	5	florida	florida	PROPN
dad-10683	870	6	artificial	artificial	PROPN
dad-10683	870	7	intelligence	intelligence	PROPN
dad-10683	870	8	research	research	NOUN
dad-10683	870	9	society	society	NOUN
dad-10683	870	10	conference	conference	NOUN
dad-10683	870	11	,	,	PUNCT
dad-10683	870	12	2013a	2013a	NUM
dad-10683	870	13	.	.	PUNCT
dad-10683	871	1	d.c	d.c	PROPN
dad-10683	871	2	.	.	PUNCT
dad-10683	871	3	uthus	uthus	PROPN
dad-10683	871	4	and	and	CCONJ
dad-10683	871	5	d.w	d.w	PROPN
dad-10683	871	6	.	.	PROPN
dad-10683	872	1	aha	aha	INTJ
dad-10683	872	2	.	.	PUNCT
dad-10683	873	1	the	the	DET
dad-10683	873	2	ubuntu	ubuntu	ADJ
dad-10683	873	3	chat	chat	NOUN
dad-10683	873	4	corpus	corpu	NOUN
dad-10683	873	5	for	for	ADP
dad-10683	873	6	multiparticipant	multiparticipant	NOUN
dad-10683	873	7	chat	chat	NOUN
dad-10683	873	8	analysis	analysis	NOUN
dad-10683	873	9	.	.	PUNCT
dad-10683	874	1	in	in	ADP
dad-10683	874	2	proceedings	proceeding	NOUN
dad-10683	874	3	of	of	ADP
dad-10683	874	4	the	the	DET
dad-10683	874	5	aaai	aaai	PROPN
dad-10683	874	6	spring	spring	NOUN
dad-10683	874	7	symposium	symposium	NOUN
dad-10683	874	8	on	on	ADP
dad-10683	874	9	analyzing	analyze	VERB
dad-10683	874	10	microtext	microtext	ADJ
dad-10683	874	11	,	,	PUNCT
dad-10683	874	12	2013b	2013b	NUM
dad-10683	874	13	.	.	PUNCT
dad-10683	875	1	o.	o.	NOUN
dad-10683	875	2	vinyals	vinyal	NOUN
dad-10683	875	3	and	and	CCONJ
dad-10683	875	4	q.	q.	PROPN
dad-10683	875	5	le	le	PROPN
dad-10683	875	6	.	.	PUNCT
dad-10683	876	1	a	a	DET
dad-10683	876	2	neural	neural	ADJ
dad-10683	876	3	conversational	conversational	ADJ
dad-10683	876	4	model	model	NOUN
dad-10683	876	5	.	.	PUNCT
dad-10683	877	1	arxiv	arxiv	PROPN
dad-10683	877	2	preprint	preprint	VERB
dad-10683	877	3	arxiv:1506.05869	arxiv:1506.05869	PROPN
dad-10683	877	4	,	,	PUNCT
dad-10683	877	5	2015	2015	NUM
dad-10683	877	6	.	.	PUNCT
dad-10683	878	1	m.	m.	PROPN
dad-10683	878	2	walker	walker	PROPN
dad-10683	878	3	,	,	PUNCT
dad-10683	878	4	d.	d.	PROPN
dad-10683	878	5	litman	litman	PROPN
dad-10683	878	6	,	,	PUNCT
dad-10683	878	7	c.	c.	PROPN
dad-10683	878	8	kamm	kamm	PROPN
dad-10683	878	9	,	,	PUNCT
dad-10683	878	10	and	and	CCONJ
dad-10683	878	11	a.	a.	NOUN
dad-10683	878	12	abella	abella	PROPN
dad-10683	878	13	.	.	PUNCT
dad-10683	879	1	paradise	paradise	PROPN
dad-10683	879	2	:	:	PUNCT
dad-10683	879	3	a	a	DET
dad-10683	879	4	framework	framework	NOUN
dad-10683	879	5	for	for	ADP
dad-10683	879	6	evaluating	evaluate	VERB
dad-10683	879	7	spoken	spoken	ADJ
dad-10683	879	8	dialogue	dialogue	NOUN
dad-10683	879	9	agents	agent	NOUN
dad-10683	879	10	.	.	PUNCT
dad-10683	880	1	in	in	ADP
dad-10683	880	2	proceedings	proceeding	NOUN
dad-10683	880	3	of	of	ADP
dad-10683	880	4	the	the	DET
dad-10683	880	5	meeting	meeting	NOUN
dad-10683	880	6	of	of	ADP
dad-10683	880	7	the	the	DET
dad-10683	880	8	european	european	ADJ
dad-10683	880	9	chapter	chapter	NOUN
dad-10683	880	10	of	of	ADP
dad-10683	880	11	the	the	DET
dad-10683	880	12	association	association	NOUN
dad-10683	880	13	for	for	ADP
dad-10683	880	14	computational	computational	ADJ
dad-10683	880	15	linguistics	linguistic	NOUN
dad-10683	880	16	,	,	PUNCT
dad-10683	880	17	1997	1997	NUM
dad-10683	880	18	.	.	PUNCT
dad-10683	881	1	h.	h.	PROPN
dad-10683	881	2	wang	wang	PROPN
dad-10683	881	3	,	,	PUNCT
dad-10683	881	4	z.	z.	PROPN
dad-10683	881	5	lu	lu	PROPN
dad-10683	881	6	,	,	PUNCT
dad-10683	881	7	h.	h.	PROPN
dad-10683	881	8	li	li	PROPN
dad-10683	881	9	,	,	PUNCT
dad-10683	881	10	and	and	CCONJ
dad-10683	881	11	e.	e.	PROPN
dad-10683	881	12	chen	chen	PROPN
dad-10683	881	13	.	.	PUNCT
dad-10683	882	1	a	a	DET
dad-10683	882	2	dataset	dataset	NOUN
dad-10683	882	3	for	for	ADP
dad-10683	882	4	research	research	NOUN
dad-10683	882	5	on	on	ADP
dad-10683	882	6	short	short	ADJ
dad-10683	882	7	-	-	PUNCT
dad-10683	882	8	text	text	NOUN
dad-10683	882	9	conversations	conversation	NOUN
dad-10683	882	10	.	.	PUNCT
dad-10683	883	1	in	in	ADP
dad-10683	883	2	proceedings	proceeding	NOUN
dad-10683	883	3	of	of	ADP
dad-10683	883	4	the	the	DET
dad-10683	883	5	conference	conference	NOUN
dad-10683	883	6	on	on	ADP
dad-10683	883	7	empirical	empirical	ADJ
dad-10683	883	8	methods	method	NOUN
dad-10683	883	9	in	in	ADP
dad-10683	883	10	natural	natural	ADJ
dad-10683	883	11	language	language	NOUN
dad-10683	883	12	processing	processing	NOUN
dad-10683	883	13	,	,	PUNCT
dad-10683	883	14	2013	2013	NUM
dad-10683	883	15	.	.	PUNCT
dad-10683	884	1	m.	m.	PROPN
dad-10683	884	2	wen	wen	PROPN
dad-10683	884	3	,	,	PUNCT
dad-10683	884	4	t.and	t.and	ADV
dad-10683	884	5	gasic	gasic	NOUN
dad-10683	884	6	,	,	PUNCT
dad-10683	884	7	n.	n.	PROPN
dad-10683	884	8	mrksic	mrksic	NOUN
dad-10683	884	9	,	,	PUNCT
dad-10683	884	10	l.	l.	PROPN
dad-10683	884	11	rojas	rojas	PROPN
dad-10683	884	12	-	-	PUNCT
dad-10683	884	13	barahona	barahona	PROPN
dad-10683	884	14	,	,	PUNCT
dad-10683	884	15	p.	p.	PROPN
dad-10683	884	16	su	su	PROPN
dad-10683	884	17	,	,	PUNCT
dad-10683	884	18	s.	s.	PROPN
dad-10683	884	19	ultes	ultes	PROPN
dad-10683	884	20	,	,	PUNCT
dad-10683	884	21	d.	d.	PROPN
dad-10683	884	22	vandyke	vandyke	PROPN
dad-10683	884	23	,	,	PUNCT
dad-10683	884	24	and	and	CCONJ
dad-10683	884	25	s.	s.	PROPN
dad-10683	884	26	young	young	PROPN
dad-10683	884	27	.	.	PUNCT
dad-10683	885	1	a	a	DET
dad-10683	885	2	network	network	NOUN
dad-10683	885	3	-	-	PUNCT
dad-10683	885	4	based	base	VERB
dad-10683	885	5	end	end	NOUN
dad-10683	885	6	-	-	PUNCT
dad-10683	885	7	to	to	ADP
dad-10683	885	8	-	-	PUNCT
dad-10683	885	9	end	end	NOUN
dad-10683	885	10	trainable	trainable	ADJ
dad-10683	885	11	task	task	NOUN
dad-10683	885	12	-	-	PUNCT
dad-10683	885	13	oriented	orient	VERB
dad-10683	885	14	dialogue	dialogue	NOUN
dad-10683	885	15	system	system	NOUN
dad-10683	885	16	.	.	PUNCT
dad-10683	886	1	arxiv	arxiv	PROPN
dad-10683	886	2	preprint	preprint	NOUN
dad-10683	886	3	arxiv:1604.04562	arxiv:1604.04562	NOUN
dad-10683	886	4	,	,	PUNCT
dad-10683	886	5	2016	2016	NUM
dad-10683	886	6	.	.	PUNCT
dad-10683	887	1	j.	j.	PROPN
dad-10683	887	2	williams	williams	PROPN
dad-10683	887	3	,	,	PUNCT
dad-10683	887	4	a.	a.	PROPN
dad-10683	887	5	raux	raux	PROPN
dad-10683	887	6	,	,	PUNCT
dad-10683	887	7	d.	d.	PROPN
dad-10683	887	8	ramachandran	ramachandran	PROPN
dad-10683	887	9	,	,	PUNCT
dad-10683	887	10	and	and	CCONJ
dad-10683	887	11	a.	a.	NOUN
dad-10683	887	12	black	black	NOUN
dad-10683	887	13	.	.	PUNCT
dad-10683	888	1	the	the	DET
dad-10683	888	2	dialog	dialog	NOUN
dad-10683	888	3	state	state	NOUN
dad-10683	888	4	tracking	tracking	NOUN
dad-10683	888	5	challenge	challenge	NOUN
dad-10683	888	6	.	.	PUNCT
dad-10683	889	1	in	in	ADP
dad-10683	889	2	proceedings	proceeding	NOUN
dad-10683	889	3	of	of	ADP
dad-10683	889	4	the	the	DET
dad-10683	889	5	meeting	meeting	NOUN
dad-10683	889	6	of	of	ADP
dad-10683	889	7	the	the	DET
dad-10683	889	8	special	special	ADJ
dad-10683	889	9	interest	interest	NOUN
dad-10683	889	10	group	group	NOUN
dad-10683	889	11	on	on	ADP
dad-10683	889	12	dialogue	dialogue	NOUN
dad-10683	889	13	and	and	CCONJ
dad-10683	889	14	discourse	discourse	NOUN
dad-10683	889	15	(	(	PUNCT
dad-10683	889	16	sigdial	sigdial	ADJ
dad-10683	889	17	)	)	PUNCT
dad-10683	889	18	,	,	PUNCT
dad-10683	889	19	2013	2013	NUM
dad-10683	889	20	.	.	PUNCT
dad-10683	890	1	j.	j.	PROPN
dad-10683	890	2	williams	williams	PROPN
dad-10683	890	3	,	,	PUNCT
dad-10683	890	4	a.	a.	PROPN
dad-10683	890	5	raux	raux	PROPN
dad-10683	890	6	,	,	PUNCT
dad-10683	890	7	and	and	CCONJ
dad-10683	890	8	m.	m.	PROPN
dad-10683	890	9	henderson	henderson	PROPN
dad-10683	890	10	.	.	PUNCT
dad-10683	891	1	the	the	DET
dad-10683	891	2	dialog	dialog	NOUN
dad-10683	891	3	state	state	NOUN
dad-10683	891	4	tracking	tracking	NOUN
dad-10683	891	5	challenge	challenge	NOUN
dad-10683	891	6	series	series	NOUN
dad-10683	891	7	:	:	PUNCT
dad-10683	891	8	a	a	DET
dad-10683	891	9	review	review	NOUN
dad-10683	891	10	.	.	PUNCT
dad-10683	892	1	dialogue	dialogue	NOUN
dad-10683	892	2	&	&	CCONJ
dad-10683	892	3	discourse	discourse	PROPN
dad-10683	892	4	,	,	PUNCT
dad-10683	892	5	7(3):4–33	7(3):4–33	PROPN
dad-10683	892	6	,	,	PUNCT
dad-10683	892	7	2016	2016	NUM
dad-10683	892	8	.	.	PUNCT
dad-10683	893	1	j.	j.	PROPN
dad-10683	893	2	d.	d.	PROPN
dad-10683	893	3	williams	williams	PROPN
dad-10683	893	4	.	.	PUNCT
dad-10683	894	1	web	web	NOUN
dad-10683	894	2	-	-	NOUN
dad-10683	894	3	style	style	NOUN
dad-10683	894	4	ranking	ranking	NOUN
dad-10683	894	5	and	and	CCONJ
dad-10683	894	6	slu	slu	NOUN
dad-10683	894	7	combination	combination	NOUN
dad-10683	894	8	for	for	ADP
dad-10683	894	9	dialog	dialog	NOUN
dad-10683	894	10	state	state	NOUN
dad-10683	894	11	tracking	tracking	NOUN
dad-10683	894	12	.	.	PUNCT
dad-10683	895	1	in	in	ADP
dad-10683	895	2	proceedings	proceeding	NOUN
dad-10683	895	3	of	of	ADP
dad-10683	895	4	the	the	DET
dad-10683	895	5	meeting	meeting	NOUN
dad-10683	895	6	of	of	ADP
dad-10683	895	7	the	the	DET
dad-10683	895	8	special	special	ADJ
dad-10683	895	9	interest	interest	NOUN
dad-10683	895	10	group	group	NOUN
dad-10683	895	11	on	on	ADP
dad-10683	895	12	discourse	discourse	NOUN
dad-10683	895	13	and	and	CCONJ
dad-10683	895	14	dialogue	dialogue	NOUN
dad-10683	895	15	(	(	PUNCT
dad-10683	895	16	sigdial	sigdial	ADJ
dad-10683	895	17	)	)	PUNCT
dad-10683	895	18	,	,	PUNCT
dad-10683	895	19	2014	2014	NUM
dad-10683	895	20	.	.	PUNCT
dad-10683	896	1	z.	z.	PROPN
dad-10683	896	2	xu	xu	PROPN
dad-10683	896	3	,	,	PUNCT
dad-10683	896	4	b.	b.	PROPN
dad-10683	896	5	liu	liu	PROPN
dad-10683	896	6	,	,	PUNCT
dad-10683	896	7	b.	b.	PROPN
dad-10683	896	8	wang	wang	PROPN
dad-10683	896	9	,	,	PUNCT
dad-10683	896	10	c.	c.	PROPN
dad-10683	896	11	sun	sun	PROPN
dad-10683	896	12	,	,	PUNCT
dad-10683	896	13	and	and	CCONJ
dad-10683	896	14	x.	x.	PROPN
dad-10683	896	15	wang	wang	PROPN
dad-10683	896	16	.	.	PUNCT
dad-10683	897	1	incorporating	incorporate	VERB
dad-10683	897	2	loose	loose	ADV
dad-10683	897	3	-	-	PUNCT
dad-10683	897	4	structured	structure	VERB
dad-10683	897	5	knowledge	knowledge	NOUN
dad-10683	897	6	into	into	ADP
dad-10683	897	7	lstm	lstm	NOUN
dad-10683	897	8	with	with	ADP
dad-10683	897	9	recall	recall	NOUN
dad-10683	897	10	gate	gate	NOUN
dad-10683	897	11	for	for	ADP
dad-10683	897	12	conversation	conversation	NOUN
dad-10683	897	13	modeling	modeling	NOUN
dad-10683	897	14	.	.	PUNCT
dad-10683	898	1	arxiv	arxiv	PROPN
dad-10683	898	2	preprint	preprint	PROPN
dad-10683	898	3	arxiv:1605.05110	arxiv:1605.05110	PROPN
dad-10683	898	4	,	,	PUNCT
dad-10683	898	5	2016	2016	NUM
dad-10683	898	6	.	.	PUNCT
dad-10683	899	1	s.	s.	PROPN
dad-10683	899	2	j.	j.	PROPN
dad-10683	899	3	young	young	PROPN
dad-10683	899	4	.	.	PUNCT
dad-10683	900	1	probabilistic	probabilistic	ADJ
dad-10683	900	2	methods	method	NOUN
dad-10683	900	3	in	in	ADP
dad-10683	900	4	spoken	spoken	ADJ
dad-10683	900	5	–	–	PUNCT
dad-10683	900	6	dialogue	dialogue	NOUN
dad-10683	900	7	systems	system	NOUN
dad-10683	900	8	.	.	PUNCT
dad-10683	901	1	philosophical	philosophical	ADJ
dad-10683	901	2	transactions	transaction	NOUN
dad-10683	901	3	of	of	ADP
dad-10683	901	4	the	the	DET
dad-10683	901	5	royal	royal	ADJ
dad-10683	901	6	society	society	NOUN
dad-10683	901	7	of	of	ADP
dad-10683	901	8	london	london	PROPN
dad-10683	901	9	a	a	DET
dad-10683	901	10	:	:	PUNCT
dad-10683	901	11	mathematical	mathematical	ADJ
dad-10683	901	12	,	,	PUNCT
dad-10683	901	13	physical	physical	ADJ
dad-10683	901	14	and	and	CCONJ
dad-10683	901	15	engineering	engineering	NOUN
dad-10683	901	16	sciences	science	NOUN
dad-10683	901	17	,	,	PUNCT
dad-10683	901	18	358(1769	358(1769	NUM
dad-10683	901	19	):	):	PUNCT
dad-10683	901	20	1389–1402	1389–1402	NUM
dad-10683	901	21	,	,	PUNCT
dad-10683	901	22	2000	2000	NUM
dad-10683	901	23	.	.	PUNCT
dad-10683	902	1	l.	l.	PROPN
dad-10683	902	2	yu	yu	PROPN
dad-10683	902	3	,	,	PUNCT
dad-10683	902	4	k.	k.	PROPN
dad-10683	902	5	m.	m.	PROPN
dad-10683	902	6	hermann	hermann	PROPN
dad-10683	902	7	,	,	PUNCT
dad-10683	902	8	p.	p.	NOUN
dad-10683	902	9	blunsom	blunsom	NOUN
dad-10683	902	10	,	,	PUNCT
dad-10683	902	11	and	and	CCONJ
dad-10683	902	12	s.	s.	PROPN
dad-10683	902	13	pulman	pulman	PROPN
dad-10683	902	14	.	.	PUNCT
dad-10683	903	1	deep	deep	ADJ
dad-10683	903	2	learning	learning	NOUN
dad-10683	903	3	for	for	ADP
dad-10683	903	4	answer	answer	NOUN
dad-10683	903	5	sentence	sentence	NOUN
dad-10683	903	6	selection	selection	NOUN
dad-10683	903	7	.	.	PUNCT
dad-10683	904	1	arxiv	arxiv	PROPN
dad-10683	904	2	preprint	preprint	PROPN
dad-10683	904	3	arxiv:1412.1632	arxiv:1412.1632	NOUN
dad-10683	904	4	,	,	PUNCT
dad-10683	904	5	2014	2014	NUM
dad-10683	904	6	.	.	PUNCT
dad-10683	905	1	m.d	m.d	PROPN
dad-10683	905	2	.	.	PROPN
dad-10683	905	3	zeiler	zeiler	PROPN
dad-10683	905	4	.	.	PUNCT
dad-10683	906	1	adadelta	adadelta	ADJ
dad-10683	906	2	:	:	PUNCT
dad-10683	906	3	an	an	DET
dad-10683	906	4	adaptive	adaptive	ADJ
dad-10683	906	5	learning	learning	NOUN
dad-10683	906	6	rate	rate	NOUN
dad-10683	906	7	method	method	NOUN
dad-10683	906	8	.	.	PUNCT
dad-10683	907	1	arxiv	arxiv	PROPN
dad-10683	907	2	preprint	preprint	PROPN
dad-10683	907	3	arxiv:1212.5701	arxiv:1212.5701	PROPN
dad-10683	907	4	,	,	PUNCT
dad-10683	907	5	2012	2012	NUM
dad-10683	907	6	.	.	PUNCT
dad-10683	908	1	tiancheng	tiancheng	PROPN
dad-10683	908	2	zhao	zhao	PROPN
dad-10683	908	3	and	and	CCONJ
dad-10683	908	4	maxine	maxine	PROPN
dad-10683	908	5	eskenazi	eskenazi	PROPN
dad-10683	908	6	.	.	PUNCT
dad-10683	909	1	towards	towards	ADP
dad-10683	909	2	end	end	NOUN
dad-10683	909	3	-	-	PUNCT
dad-10683	909	4	to	to	ADP
dad-10683	909	5	-	-	PUNCT
dad-10683	909	6	end	end	NOUN
dad-10683	909	7	learning	learning	NOUN
dad-10683	909	8	for	for	ADP
dad-10683	909	9	dialog	dialog	NOUN
dad-10683	909	10	state	state	NOUN
dad-10683	909	11	tracking	tracking	NOUN
dad-10683	909	12	and	and	CCONJ
dad-10683	909	13	management	management	NOUN
dad-10683	909	14	using	use	VERB
dad-10683	909	15	deep	deep	ADJ
dad-10683	909	16	reinforcement	reinforcement	NOUN
dad-10683	909	17	learning	learning	NOUN
dad-10683	909	18	.	.	PUNCT
dad-10683	910	1	in	in	ADP
dad-10683	910	2	proceedings	proceeding	NOUN
dad-10683	910	3	of	of	ADP
dad-10683	910	4	the	the	DET
dad-10683	910	5	meeting	meeting	NOUN
dad-10683	910	6	of	of	ADP
dad-10683	910	7	the	the	DET
dad-10683	910	8	special	special	ADJ
dad-10683	910	9	interest	interest	NOUN
dad-10683	910	10	group	group	NOUN
dad-10683	910	11	on	on	ADP
dad-10683	910	12	discourse	discourse	NOUN
dad-10683	910	13	and	and	CCONJ
dad-10683	910	14	dialogue	dialogue	NOUN
dad-10683	910	15	(	(	PUNCT
dad-10683	910	16	sigdial	sigdial	ADJ
dad-10683	910	17	)	)	PUNCT
dad-10683	910	18	,	,	PUNCT
dad-10683	910	19	2016	2016	NUM
dad-10683	910	20	.	.	PUNCT
dad-10683	911	1	65	65	NUM