id	sid	tid	token	lemma	pos
bracis-28433	1	1	investigation	investigation	NOUN
bracis-28433	1	2	of	of	ADP
bracis-28433	1	3	 	 	SPACE
bracis-28433	1	4	deep	deep	ADJ
bracis-28433	1	5	active	active	ADJ
bracis-28433	1	6	self	self	NOUN
bracis-28433	1	7	-	-	PUNCT
bracis-28433	1	8	learning	learn	VERB
bracis-28433	1	9	algorithms	algorithm	NOUN
bracis-28433	1	10	applied	apply	VERB
bracis-28433	1	11	to	to	ADP
bracis-28433	1	12	 	 	SPACE
bracis-28433	1	13	named	name	VERB
bracis-28433	1	14	entity	entity	NOUN
bracis-28433	1	15	recognition	recognition	NOUN
bracis-28433	1	16	|	|	NOUN
bracis-28433	1	17	springer	springer	NOUN
bracis-28433	1	18	nature	nature	PROPN
bracis-28433	1	19	link	link	PROPN
bracis-28433	1	20	(	(	PUNCT
bracis-28433	1	21	formerly	formerly	ADV
bracis-28433	1	22	springerlink	springerlink	NOUN
bracis-28433	1	23	)	)	PUNCT
bracis-28433	1	24	skip	skip	VERB
bracis-28433	1	25	to	to	ADP
bracis-28433	1	26	main	main	ADJ
bracis-28433	1	27	content	content	NOUN
bracis-28433	1	28	advertisement	advertisement	NOUN
bracis-28433	1	29	log	log	NOUN
bracis-28433	1	30	in	in	ADP
bracis-28433	1	31	menu	menu	NOUN
bracis-28433	1	32	find	find	VERB
bracis-28433	1	33	a	a	DET
bracis-28433	1	34	journal	journal	NOUN
bracis-28433	1	35	publish	publish	VERB
bracis-28433	1	36	with	with	ADP
bracis-28433	1	37	us	we	PRON
bracis-28433	1	38	track	track	VERB
bracis-28433	1	39	your	your	PRON
bracis-28433	1	40	research	research	NOUN
bracis-28433	1	41	search	search	NOUN
bracis-28433	1	42	cart	cart	NOUN
bracis-28433	1	43	home	home	NOUN
bracis-28433	1	44	intelligent	intelligent	ADJ
bracis-28433	1	45	systems	system	NOUN
bracis-28433	1	46	conference	conference	NOUN
bracis-28433	1	47	paper	paper	NOUN
bracis-28433	1	48	investigation	investigation	NOUN
bracis-28433	1	49	of	of	ADP
bracis-28433	1	50	 	 	SPACE
bracis-28433	1	51	deep	deep	ADJ
bracis-28433	1	52	active	active	ADJ
bracis-28433	1	53	self	self	NOUN
bracis-28433	1	54	-	-	PUNCT
bracis-28433	1	55	learning	learn	VERB
bracis-28433	1	56	algorithms	algorithm	NOUN
bracis-28433	1	57	applied	apply	VERB
bracis-28433	1	58	to	to	ADP
bracis-28433	1	59	 	 	SPACE
bracis-28433	1	60	named	name	VERB
bracis-28433	1	61	entity	entity	NOUN
bracis-28433	1	62	recognition	recognition	NOUN
bracis-28433	1	63	conference	conference	NOUN
bracis-28433	1	64	paper	paper	NOUN
bracis-28433	1	65	first	first	ADV
bracis-28433	1	66	online	online	ADV
bracis-28433	1	67	:	:	PUNCT
bracis-28433	1	68	12	12	NUM
bracis-28433	1	69	october	october	NOUN
bracis-28433	1	70	2023	2023	NUM
bracis-28433	1	71	pp	pp	ADP
bracis-28433	1	72	470–484	470–484	NUM
bracis-28433	1	73	cite	cite	VERB
bracis-28433	1	74	this	this	DET
bracis-28433	1	75	conference	conference	NOUN
bracis-28433	1	76	paper	paper	NOUN
bracis-28433	1	77	access	access	NOUN
bracis-28433	1	78	provided	provide	VERB
bracis-28433	1	79	by	by	ADP
bracis-28433	1	80	university	university	PROPN
bracis-28433	1	81	of	of	ADP
bracis-28433	1	82	notre	notre	PROPN
bracis-28433	1	83	dame	dame	PROPN
bracis-28433	1	84	hesburgh	hesburgh	PROPN
bracis-28433	1	85	library	library	PROPN
bracis-28433	1	86	download	download	PROPN
bracis-28433	1	87	book	book	NOUN
bracis-28433	1	88	pdf	pdf	PROPN
bracis-28433	1	89	download	download	NOUN
bracis-28433	1	90	book	book	NOUN
bracis-28433	1	91	epub	epub	PROPN
bracis-28433	1	92	intelligent	intelligent	ADJ
bracis-28433	1	93	systems	system	NOUN
bracis-28433	1	94	(	(	PUNCT
bracis-28433	1	95	bracis	bracis	NOUN
bracis-28433	1	96	2023	2023	NUM
bracis-28433	1	97	)	)	PUNCT
bracis-28433	1	98	investigation	investigation	NOUN
bracis-28433	1	99	of	of	ADP
bracis-28433	1	100	 	 	SPACE
bracis-28433	1	101	deep	deep	ADJ
bracis-28433	1	102	active	active	ADJ
bracis-28433	1	103	self	self	NOUN
bracis-28433	1	104	-	-	PUNCT
bracis-28433	1	105	learning	learn	VERB
bracis-28433	1	106	algorithms	algorithm	NOUN
bracis-28433	1	107	applied	apply	VERB
bracis-28433	1	108	to	to	ADP
bracis-28433	1	109	 	 	SPACE
bracis-28433	1	110	named	name	VERB
bracis-28433	1	111	entity	entity	NOUN
bracis-28433	1	112	recognition	recognition	NOUN
bracis-28433	1	113	download	download	NOUN
bracis-28433	1	114	book	book	NOUN
bracis-28433	1	115	pdf	pdf	PROPN
bracis-28433	1	116	download	download	NOUN
bracis-28433	1	117	book	book	PROPN
bracis-28433	1	118	epub	epub	PROPN
bracis-28433	1	119	josé	josé	PROPN
bracis-28433	1	120	reinaldo	reinaldo	PROPN
bracis-28433	1	121	cunha	cunha	PROPN
bracis-28433	1	122	santos	santos	PROPN
bracis-28433	1	123	a.	a.	PROPN
bracis-28433	1	124	v.	v.	PROPN
bracis-28433	1	125	silva	silva	PROPN
bracis-28433	1	126	neto9	neto9	PROPN
bracis-28433	1	127	&	&	CCONJ
bracis-28433	1	128	thiago	thiago	PROPN
bracis-28433	1	129	de	de	PROPN
bracis-28433	1	130	paulo	paulo	PROPN
bracis-28433	1	131	faleiros10	faleiros10	NOUN
bracis-28433	1	132	  	  	SPACE
bracis-28433	1	133	part	part	NOUN
bracis-28433	1	134	of	of	ADP
bracis-28433	1	135	the	the	DET
bracis-28433	1	136	book	book	NOUN
bracis-28433	1	137	series	series	NOUN
bracis-28433	1	138	:	:	PUNCT
bracis-28433	1	139	lecture	lecture	NOUN
bracis-28433	1	140	notes	note	NOUN
bracis-28433	1	141	in	in	ADP
bracis-28433	1	142	computer	computer	NOUN
bracis-28433	1	143	science	science	NOUN
bracis-28433	1	144	(	(	PUNCT
bracis-28433	1	145	(	(	PUNCT
bracis-28433	1	146	lnai	lnai	ADJ
bracis-28433	1	147	,	,	PUNCT
bracis-28433	1	148	volume	volume	NOUN
bracis-28433	1	149	14197	14197	NUM
bracis-28433	1	150	)	)	PUNCT
bracis-28433	1	151	)	)	PUNCT
bracis-28433	1	152	included	include	VERB
bracis-28433	1	153	in	in	ADP
bracis-28433	1	154	the	the	DET
bracis-28433	1	155	following	follow	VERB
bracis-28433	1	156	conference	conference	NOUN
bracis-28433	1	157	series	series	NOUN
bracis-28433	1	158	:	:	PUNCT
bracis-28433	1	159	brazilian	brazilian	ADJ
bracis-28433	1	160	conference	conference	NOUN
bracis-28433	1	161	on	on	ADP
bracis-28433	1	162	intelligent	intelligent	ADJ
bracis-28433	1	163	systems	system	NOUN
bracis-28433	1	164	514	514	NUM
bracis-28433	1	165	accesses	access	NOUN
bracis-28433	1	166	abstract	abstract	ADJ
bracis-28433	1	167	active	active	ADJ
bracis-28433	1	168	self	self	NOUN
bracis-28433	1	169	-	-	PUNCT
bracis-28433	1	170	learning	learn	VERB
bracis-28433	1	171	algorithms	algorithm	NOUN
bracis-28433	1	172	reduce	reduce	VERB
bracis-28433	1	173	the	the	DET
bracis-28433	1	174	labeled	label	VERB
bracis-28433	1	175	data	datum	NOUN
bracis-28433	1	176	required	require	VERB
bracis-28433	1	177	to	to	PART
bracis-28433	1	178	train	train	VERB
bracis-28433	1	179	a	a	DET
bracis-28433	1	180	machine	machine	NOUN
bracis-28433	1	181	learning	learn	VERB
bracis-28433	1	182	model	model	NOUN
bracis-28433	1	183	through	through	ADP
bracis-28433	1	184	supervised	supervised	ADJ
bracis-28433	1	185	training	training	NOUN
bracis-28433	1	186	.	.	PUNCT
bracis-28433	2	1	this	this	DET
bracis-28433	2	2	paper	paper	NOUN
bracis-28433	2	3	explores	explore	VERB
bracis-28433	2	4	various	various	ADJ
bracis-28433	2	5	active	active	ADJ
bracis-28433	2	6	self	self	NOUN
bracis-28433	2	7	-	-	PUNCT
bracis-28433	2	8	learning	learn	VERB
bracis-28433	2	9	algorithms	algorithm	NOUN
bracis-28433	2	10	for	for	ADP
bracis-28433	2	11	named	name	VERB
bracis-28433	2	12	entity	entity	NOUN
bracis-28433	2	13	recognition	recognition	NOUN
bracis-28433	2	14	tasks	task	NOUN
bracis-28433	2	15	.	.	PUNCT
bracis-28433	3	1	firstly	firstly	ADV
bracis-28433	3	2	,	,	PUNCT
bracis-28433	3	3	we	we	PRON
bracis-28433	3	4	investigate	investigate	VERB
bracis-28433	3	5	the	the	DET
bracis-28433	3	6	impact	impact	NOUN
bracis-28433	3	7	of	of	ADP
bracis-28433	3	8	different	different	ADJ
bracis-28433	3	9	self	self	NOUN
bracis-28433	3	10	-	-	PUNCT
bracis-28433	3	11	training	training	NOUN
bracis-28433	3	12	techniques	technique	NOUN
bracis-28433	3	13	on	on	ADP
bracis-28433	3	14	active	active	ADJ
bracis-28433	3	15	self	self	NOUN
bracis-28433	3	16	-	-	PUNCT
bracis-28433	3	17	learning	learn	VERB
bracis-28433	3	18	algorithms	algorithm	NOUN
bracis-28433	3	19	.	.	PUNCT
bracis-28433	4	1	secondly	secondly	ADV
bracis-28433	4	2	,	,	PUNCT
bracis-28433	4	3	we	we	PRON
bracis-28433	4	4	propose	propose	VERB
bracis-28433	4	5	a	a	DET
bracis-28433	4	6	novel	novel	ADJ
bracis-28433	4	7	token	token	VERB
bracis-28433	4	8	-	-	PUNCT
bracis-28433	4	9	level	level	NOUN
bracis-28433	4	10	active	active	ADJ
bracis-28433	4	11	self	self	NOUN
bracis-28433	4	12	-	-	PUNCT
bracis-28433	4	13	learning	learn	VERB
bracis-28433	4	14	algorithm	algorithm	NOUN
bracis-28433	4	15	that	that	PRON
bracis-28433	4	16	achieves	achieve	VERB
bracis-28433	4	17	near	near	ADJ
bracis-28433	4	18	-	-	PUNCT
bracis-28433	4	19	peak	peak	NOUN
bracis-28433	4	20	performance	performance	NOUN
bracis-28433	4	21	using	use	VERB
bracis-28433	4	22	fewer	few	ADJ
bracis-28433	4	23	hand	hand	NOUN
bracis-28433	4	24	-	-	PUNCT
bracis-28433	4	25	annotated	annotate	VERB
bracis-28433	4	26	tokens	token	NOUN
bracis-28433	4	27	compared	compare	VERB
bracis-28433	4	28	to	to	ADP
bracis-28433	4	29	existing	exist	VERB
bracis-28433	4	30	works	work	NOUN
bracis-28433	4	31	.	.	PUNCT
bracis-28433	5	1	through	through	ADP
bracis-28433	5	2	numerous	numerous	ADJ
bracis-28433	5	3	experiments	experiment	NOUN
bracis-28433	5	4	,	,	PUNCT
bracis-28433	5	5	we	we	PRON
bracis-28433	5	6	found	find	VERB
bracis-28433	5	7	that	that	SCONJ
bracis-28433	5	8	the	the	DET
bracis-28433	5	9	sentence	sentence	NOUN
bracis-28433	5	10	-	-	PUNCT
bracis-28433	5	11	level	level	NOUN
bracis-28433	5	12	active	active	ADJ
bracis-28433	5	13	self	self	NOUN
bracis-28433	5	14	-	-	PUNCT
bracis-28433	5	15	learning	learn	VERB
bracis-28433	5	16	algorithm	algorithm	NOUN
bracis-28433	5	17	did	do	AUX
bracis-28433	5	18	not	not	PART
bracis-28433	5	19	consistently	consistently	ADV
bracis-28433	5	20	yield	yield	VERB
bracis-28433	5	21	significant	significant	ADJ
bracis-28433	5	22	results	result	NOUN
bracis-28433	5	23	compared	compare	VERB
bracis-28433	5	24	to	to	ADP
bracis-28433	5	25	pure	pure	ADJ
bracis-28433	5	26	active	active	ADJ
bracis-28433	5	27	learning	learning	NOUN
bracis-28433	5	28	.	.	PUNCT
bracis-28433	6	1	however	however	ADV
bracis-28433	6	2	,	,	PUNCT
bracis-28433	6	3	our	our	PRON
bracis-28433	6	4	proposed	propose	VERB
bracis-28433	6	5	token	token	VERB
bracis-28433	6	6	-	-	PUNCT
bracis-28433	6	7	level	level	NOUN
bracis-28433	6	8	active	active	ADJ
bracis-28433	6	9	self	self	NOUN
bracis-28433	6	10	-	-	PUNCT
bracis-28433	6	11	learning	learn	VERB
bracis-28433	6	12	algorithm	algorithm	NOUN
bracis-28433	6	13	showed	show	VERB
bracis-28433	6	14	promising	promising	ADJ
bracis-28433	6	15	performance	performance	NOUN
bracis-28433	6	16	,	,	PUNCT
bracis-28433	6	17	training	train	VERB
bracis-28433	6	18	a	a	DET
bracis-28433	6	19	neural	neural	ADJ
bracis-28433	6	20	model	model	NOUN
bracis-28433	6	21	to	to	ADP
bracis-28433	6	22	nearly	nearly	ADV
bracis-28433	6	23	peak	peak	VERB
bracis-28433	6	24	accuracy	accuracy	NOUN
bracis-28433	6	25	with	with	ADP
bracis-28433	6	26	fewer	few	ADJ
bracis-28433	6	27	human	human	ADJ
bracis-28433	6	28	-	-	PUNCT
bracis-28433	6	29	annotated	annotate	VERB
bracis-28433	6	30	tokens	token	NOUN
bracis-28433	6	31	compared	compare	VERB
bracis-28433	6	32	to	to	ADP
bracis-28433	6	33	state	state	NOUN
bracis-28433	6	34	-	-	PUNCT
bracis-28433	6	35	of	of	ADP
bracis-28433	6	36	-	-	PUNCT
bracis-28433	6	37	the	the	DET
bracis-28433	6	38	-	-	PUNCT
bracis-28433	6	39	art	art	NOUN
bracis-28433	6	40	active	active	ADJ
bracis-28433	6	41	learning	learn	VERB
bracis-28433	6	42	baseline	baseline	NOUN
bracis-28433	6	43	algorithms	algorithm	NOUN
bracis-28433	6	44	.	.	PUNCT
bracis-28433	7	1	the	the	DET
bracis-28433	7	2	experimental	experimental	ADJ
bracis-28433	7	3	results	result	NOUN
bracis-28433	7	4	are	be	AUX
bracis-28433	7	5	presented	present	VERB
bracis-28433	7	6	and	and	CCONJ
bracis-28433	7	7	discussed	discuss	VERB
bracis-28433	7	8	,	,	PUNCT
bracis-28433	7	9	demonstrating	demonstrate	VERB
bracis-28433	7	10	the	the	DET
bracis-28433	7	11	superior	superior	ADJ
bracis-28433	7	12	performance	performance	NOUN
bracis-28433	7	13	of	of	ADP
bracis-28433	7	14	the	the	DET
bracis-28433	7	15	token	token	ADJ
bracis-28433	7	16	-	-	PUNCT
bracis-28433	7	17	level	level	NOUN
bracis-28433	7	18	active	active	ADJ
bracis-28433	7	19	self	self	NOUN
bracis-28433	7	20	-	-	PUNCT
bracis-28433	7	21	learning	learn	VERB
bracis-28433	7	22	algorithm	algorithm	NOUN
bracis-28433	7	23	j.	j.	PROPN
bracis-28433	7	24	r.	r.	PROPN
bracis-28433	7	25	c.	c.	PROPN
bracis-28433	7	26	s.	s.	PROPN
bracis-28433	7	27	a.	a.	PROPN
bracis-28433	7	28	v.	v.	PROPN
bracis-28433	7	29	s.	s.	PROPN
bracis-28433	7	30	neto	neto	PROPN
bracis-28433	7	31	—	—	PUNCT
bracis-28433	7	32	research	research	NOUN
bracis-28433	7	33	performed	perform	VERB
bracis-28433	7	34	during	during	ADP
bracis-28433	7	35	the	the	DET
bracis-28433	7	36	author	author	NOUN
bracis-28433	7	37	’s	’s	PART
bracis-28433	7	38	masters	master	NOUN
bracis-28433	7	39	undertaking	undertake	VERB
bracis-28433	7	40	at	at	ADP
bracis-28433	7	41	the	the	DET
bracis-28433	7	42	university	university	PROPN
bracis-28433	7	43	of	of	ADP
bracis-28433	7	44	brasilia	brasilia	PROPN
bracis-28433	7	45	(	(	PUNCT
bracis-28433	7	46	unb	unb	PROPN
bracis-28433	7	47	)	)	PUNCT
bracis-28433	7	48	.	.	PUNCT
bracis-28433	8	1	access	access	NOUN
bracis-28433	8	2	provided	provide	VERB
bracis-28433	8	3	by	by	ADP
bracis-28433	8	4	university	university	PROPN
bracis-28433	8	5	of	of	ADP
bracis-28433	8	6	notre	notre	PROPN
bracis-28433	8	7	dame	dame	PROPN
bracis-28433	8	8	hesburgh	hesburgh	PROPN
bracis-28433	8	9	library	library	PROPN
bracis-28433	8	10	.	.	PUNCT
bracis-28433	9	1	download	download	PROPN
bracis-28433	9	2	conference	conference	NOUN
bracis-28433	9	3	paper	paper	NOUN
bracis-28433	9	4	pdf	pdf	NOUN
bracis-28433	9	5	similar	similar	ADJ
bracis-28433	9	6	content	content	NOUN
bracis-28433	9	7	being	be	AUX
bracis-28433	9	8	viewed	view	VERB
bracis-28433	9	9	by	by	ADP
bracis-28433	9	10	others	other	NOUN
bracis-28433	9	11	deep	deep	ADJ
bracis-28433	9	12	active	active	ADJ
bracis-28433	9	13	-	-	PUNCT
bracis-28433	9	14	self	self	NOUN
bracis-28433	9	15	learning	learning	NOUN
bracis-28433	9	16	applied	apply	VERB
bracis-28433	9	17	to	to	ADP
bracis-28433	9	18	named	name	VERB
bracis-28433	9	19	entity	entity	NOUN
bracis-28433	9	20	recognition	recognition	NOUN
bracis-28433	9	21	chapter	chapter	NOUN
bracis-28433	9	22	©	©	PROPN
bracis-28433	9	23	2021	2021	NUM
bracis-28433	9	24	dropout	dropout	NOUN
bracis-28433	9	25	-	-	PUNCT
bracis-28433	9	26	based	base	VERB
bracis-28433	9	27	active	active	ADJ
bracis-28433	9	28	learning	learning	NOUN
bracis-28433	9	29	for	for	ADP
bracis-28433	9	30	regression	regression	NOUN
bracis-28433	9	31	chapter	chapter	NOUN
bracis-28433	9	32	©	©	PROPN
bracis-28433	9	33	2018	2018	NUM
bracis-28433	9	34	a	a	DET
bracis-28433	9	35	self	self	NOUN
bracis-28433	9	36	-	-	PUNCT
bracis-28433	9	37	training	training	NOUN
bracis-28433	9	38	approach	approach	NOUN
bracis-28433	9	39	for	for	ADP
bracis-28433	9	40	few	few	ADJ
bracis-28433	9	41	-	-	PUNCT
bracis-28433	9	42	shot	shot	NOUN
bracis-28433	9	43	named	name	VERB
bracis-28433	9	44	entity	entity	NOUN
bracis-28433	9	45	recognition	recognition	NOUN
bracis-28433	9	46	chapter	chapter	NOUN
bracis-28433	9	47	©	©	PROPN
bracis-28433	9	48	2023	2023	NUM
bracis-28433	9	49	explore	explore	VERB
bracis-28433	9	50	related	relate	VERB
bracis-28433	9	51	subjects	subject	NOUN
bracis-28433	9	52	discover	discover	VERB
bracis-28433	9	53	the	the	DET
bracis-28433	9	54	latest	late	ADJ
bracis-28433	9	55	articles	article	NOUN
bracis-28433	9	56	,	,	PUNCT
bracis-28433	9	57	books	book	NOUN
bracis-28433	9	58	and	and	CCONJ
bracis-28433	9	59	news	news	NOUN
bracis-28433	9	60	in	in	ADP
bracis-28433	9	61	related	related	ADJ
bracis-28433	9	62	subjects	subject	NOUN
bracis-28433	9	63	,	,	PUNCT
bracis-28433	9	64	suggested	suggest	VERB
bracis-28433	9	65	using	use	VERB
bracis-28433	9	66	machine	machine	NOUN
bracis-28433	9	67	learning	learning	NOUN
bracis-28433	9	68	.	.	PUNCT
bracis-28433	10	1	algorithms	algorithm	NOUN
bracis-28433	10	2	automated	automate	VERB
bracis-28433	10	3	pattern	pattern	NOUN
bracis-28433	10	4	recognition	recognition	NOUN
bracis-28433	10	5	categorization	categorization	NOUN
bracis-28433	10	6	learning	learn	VERB
bracis-28433	10	7	algorithms	algorithm	NOUN
bracis-28433	10	8	machine	machine	NOUN
bracis-28433	10	9	learning	learn	VERB
bracis-28433	10	10	sequence	sequence	NOUN
bracis-28433	10	11	annotation	annotation	NOUN
bracis-28433	10	12	1	1	NUM
bracis-28433	10	13	introduction	introduction	NOUN
bracis-28433	10	14	active	active	ADJ
bracis-28433	10	15	learning	learning	NOUN
bracis-28433	10	16	(	(	PUNCT
bracis-28433	10	17	al	al	PROPN
bracis-28433	10	18	)	)	PUNCT
bracis-28433	10	19	algorithms	algorithm	NOUN
bracis-28433	10	20	reduce	reduce	VERB
bracis-28433	10	21	the	the	DET
bracis-28433	10	22	labeled	label	VERB
bracis-28433	10	23	data	datum	NOUN
bracis-28433	10	24	required	require	VERB
bracis-28433	10	25	to	to	PART
bracis-28433	10	26	train	train	VERB
bracis-28433	10	27	a	a	DET
bracis-28433	10	28	machine	machine	NOUN
bracis-28433	10	29	learning	learn	VERB
bracis-28433	10	30	model	model	NOUN
bracis-28433	10	31	through	through	ADP
bracis-28433	10	32	supervised	supervised	ADJ
bracis-28433	10	33	training	training	NOUN
bracis-28433	10	34	.	.	PUNCT
bracis-28433	11	1	these	these	DET
bracis-28433	11	2	algorithms	algorithm	NOUN
bracis-28433	11	3	can	can	AUX
bracis-28433	11	4	be	be	AUX
bracis-28433	11	5	generally	generally	ADV
bracis-28433	11	6	separated	separate	VERB
bracis-28433	11	7	into	into	ADP
bracis-28433	11	8	three	three	NUM
bracis-28433	11	9	classes	class	NOUN
bracis-28433	11	10	:	:	PUNCT
bracis-28433	11	11	(	(	PUNCT
bracis-28433	11	12	1	1	X
bracis-28433	11	13	)	)	PUNCT
bracis-28433	11	14	pool	pool	NOUN
bracis-28433	11	15	-	-	PUNCT
bracis-28433	11	16	based	base	VERB
bracis-28433	11	17	,	,	PUNCT
bracis-28433	11	18	(	(	PUNCT
bracis-28433	11	19	2	2	NUM
bracis-28433	11	20	)	)	PUNCT
bracis-28433	11	21	stream	stream	NOUN
bracis-28433	11	22	-	-	PUNCT
bracis-28433	11	23	based	base	VERB
bracis-28433	11	24	,	,	PUNCT
bracis-28433	11	25	and	and	CCONJ
bracis-28433	11	26	(	(	PUNCT
bracis-28433	11	27	3	3	X
bracis-28433	11	28	)	)	PUNCT
bracis-28433	11	29	query	query	NOUN
bracis-28433	11	30	synthesis	synthesis	NOUN
bracis-28433	11	31	.	.	PUNCT
bracis-28433	12	1	for	for	ADP
bracis-28433	12	2	this	this	DET
bracis-28433	12	3	paper	paper	NOUN
bracis-28433	12	4	,	,	PUNCT
bracis-28433	12	5	we	we	PRON
bracis-28433	12	6	focus	focus	VERB
bracis-28433	12	7	on	on	ADP
bracis-28433	12	8	the	the	DET
bracis-28433	12	9	pool	pool	NOUN
bracis-28433	12	10	-	-	PUNCT
bracis-28433	12	11	based	base	VERB
bracis-28433	12	12	class	class	NOUN
bracis-28433	12	13	which	which	PRON
bracis-28433	12	14	thrives	thrive	VERB
bracis-28433	12	15	in	in	ADP
bracis-28433	12	16	scenarios	scenario	NOUN
bracis-28433	12	17	where	where	SCONJ
bracis-28433	12	18	large	large	ADJ
bracis-28433	12	19	amounts	amount	NOUN
bracis-28433	12	20	of	of	ADP
bracis-28433	12	21	data	datum	NOUN
bracis-28433	12	22	are	be	AUX
bracis-28433	12	23	available	available	ADJ
bracis-28433	12	24	,	,	PUNCT
bracis-28433	12	25	but	but	CCONJ
bracis-28433	12	26	the	the	DET
bracis-28433	12	27	annotation	annotation	NOUN
bracis-28433	12	28	process	process	NOUN
bracis-28433	12	29	for	for	ADP
bracis-28433	12	30	the	the	DET
bracis-28433	12	31	complete	complete	ADJ
bracis-28433	12	32	pool	pool	NOUN
bracis-28433	12	33	of	of	ADP
bracis-28433	12	34	data	datum	NOUN
bracis-28433	12	35	is	be	AUX
bracis-28433	12	36	costly	costly	ADJ
bracis-28433	12	37	.	.	PUNCT
bracis-28433	13	1	the	the	DET
bracis-28433	13	2	objective	objective	NOUN
bracis-28433	13	3	of	of	ADP
bracis-28433	13	4	the	the	DET
bracis-28433	13	5	pool	pool	NOUN
bracis-28433	13	6	-	-	PUNCT
bracis-28433	13	7	based	base	VERB
bracis-28433	13	8	al	al	PROPN
bracis-28433	13	9	technique	technique	NOUN
bracis-28433	13	10	is	be	AUX
bracis-28433	13	11	to	to	PART
bracis-28433	13	12	identify	identify	VERB
bracis-28433	13	13	,	,	PUNCT
bracis-28433	13	14	given	give	VERB
bracis-28433	13	15	a	a	DET
bracis-28433	13	16	considerable	considerable	ADJ
bracis-28433	13	17	pool	pool	NOUN
bracis-28433	13	18	of	of	ADP
bracis-28433	13	19	unlabeled	unlabeled	ADJ
bracis-28433	13	20	data	datum	NOUN
bracis-28433	13	21	,	,	PUNCT
bracis-28433	13	22	a	a	DET
bracis-28433	13	23	smaller	small	ADJ
bracis-28433	13	24	subset	subset	NOUN
bracis-28433	13	25	of	of	ADP
bracis-28433	13	26	data	datum	NOUN
bracis-28433	13	27	samples	sample	NOUN
bracis-28433	13	28	that	that	PRON
bracis-28433	13	29	represent	represent	VERB
bracis-28433	13	30	the	the	DET
bracis-28433	13	31	whole	whole	ADJ
bracis-28433	13	32	data	datum	NOUN
bracis-28433	13	33	distribution	distribution	NOUN
bracis-28433	13	34	well	well	ADV
bracis-28433	13	35	.	.	PUNCT
bracis-28433	14	1	by	by	ADP
bracis-28433	14	2	annotating	annotate	VERB
bracis-28433	14	3	this	this	DET
bracis-28433	14	4	smaller	small	ADJ
bracis-28433	14	5	subset	subset	NOUN
bracis-28433	14	6	of	of	ADP
bracis-28433	14	7	data	datum	NOUN
bracis-28433	14	8	samples	sample	NOUN
bracis-28433	14	9	and	and	CCONJ
bracis-28433	14	10	training	train	VERB
bracis-28433	14	11	a	a	DET
bracis-28433	14	12	model	model	NOUN
bracis-28433	14	13	through	through	ADP
bracis-28433	14	14	supervised	supervised	ADJ
bracis-28433	14	15	training	training	NOUN
bracis-28433	14	16	,	,	PUNCT
bracis-28433	14	17	it	it	PRON
bracis-28433	14	18	is	be	AUX
bracis-28433	14	19	possible	possible	ADJ
bracis-28433	14	20	to	to	PART
bracis-28433	14	21	achieve	achieve	VERB
bracis-28433	14	22	good	good	ADJ
bracis-28433	14	23	model	model	NOUN
bracis-28433	14	24	performance	performance	NOUN
bracis-28433	14	25	with	with	ADP
bracis-28433	14	26	a	a	DET
bracis-28433	14	27	significant	significant	ADJ
bracis-28433	14	28	decrease	decrease	NOUN
bracis-28433	14	29	in	in	ADP
bracis-28433	14	30	data	data	NOUN
bracis-28433	14	31	annotation	annotation	NOUN
bracis-28433	14	32	costs	cost	NOUN
bracis-28433	14	33	.	.	PUNCT
bracis-28433	15	1	in	in	ADP
bracis-28433	15	2	active	active	ADJ
bracis-28433	15	3	learning	learning	NOUN
bracis-28433	15	4	for	for	ADP
bracis-28433	15	5	named	name	VERB
bracis-28433	15	6	entity	entity	NOUN
bracis-28433	15	7	recognition	recognition	NOUN
bracis-28433	15	8	(	(	PUNCT
bracis-28433	15	9	ner	ner	NOUN
bracis-28433	15	10	)	)	PUNCT
bracis-28433	15	11	,	,	PUNCT
bracis-28433	15	12	most	most	ADJ
bracis-28433	15	13	research	research	NOUN
bracis-28433	15	14	proposes	propose	VERB
bracis-28433	15	15	using	use	VERB
bracis-28433	15	16	sentence	sentence	NOUN
bracis-28433	15	17	-	-	PUNCT
bracis-28433	15	18	level	level	NOUN
bracis-28433	15	19	querying	query	VERB
bracis-28433	15	20	strategies	strategy	NOUN
bracis-28433	15	21	.	.	PUNCT
bracis-28433	16	1	sentence	sentence	NOUN
bracis-28433	16	2	-	-	PUNCT
bracis-28433	16	3	level	level	NOUN
bracis-28433	16	4	querying	querying	NOUN
bracis-28433	16	5	means	mean	VERB
bracis-28433	16	6	that	that	SCONJ
bracis-28433	16	7	the	the	DET
bracis-28433	16	8	whole	whole	ADJ
bracis-28433	16	9	unlabeled	unlabeled	ADJ
bracis-28433	16	10	sentence	sentence	NOUN
bracis-28433	16	11	is	be	AUX
bracis-28433	16	12	queried	query	VERB
bracis-28433	16	13	to	to	ADP
bracis-28433	16	14	the	the	DET
bracis-28433	16	15	oracle	oracle	NOUN
bracis-28433	16	16	,	,	PUNCT
bracis-28433	16	17	which	which	PRON
bracis-28433	16	18	is	be	AUX
bracis-28433	16	19	expected	expect	VERB
bracis-28433	16	20	to	to	PART
bracis-28433	16	21	provide	provide	VERB
bracis-28433	16	22	labels	label	NOUN
bracis-28433	16	23	for	for	ADP
bracis-28433	16	24	all	all	DET
bracis-28433	16	25	tokens	token	NOUN
bracis-28433	16	26	.	.	PUNCT
bracis-28433	17	1	it	it	PRON
bracis-28433	17	2	provides	provide	VERB
bracis-28433	17	3	more	more	ADJ
bracis-28433	17	4	context	context	NOUN
bracis-28433	17	5	and	and	CCONJ
bracis-28433	17	6	can	can	AUX
bracis-28433	17	7	lead	lead	VERB
bracis-28433	17	8	to	to	ADP
bracis-28433	17	9	more	more	ADV
bracis-28433	17	10	accurate	accurate	ADJ
bracis-28433	17	11	annotation	annotation	NOUN
bracis-28433	17	12	decisions	decision	NOUN
bracis-28433	17	13	,	,	PUNCT
bracis-28433	17	14	but	but	CCONJ
bracis-28433	17	15	it	it	PRON
bracis-28433	17	16	can	can	AUX
bracis-28433	17	17	be	be	AUX
bracis-28433	17	18	more	more	ADV
bracis-28433	17	19	computationally	computationally	ADV
bracis-28433	17	20	expensive	expensive	ADJ
bracis-28433	17	21	and	and	CCONJ
bracis-28433	17	22	time	time	NOUN
bracis-28433	17	23	-	-	PUNCT
bracis-28433	17	24	consuming	consume	VERB
bracis-28433	17	25	.	.	PUNCT
bracis-28433	18	1	some	some	DET
bracis-28433	18	2	works	work	VERB
bracis-28433	18	3	propose	propose	VERB
bracis-28433	18	4	alternative	alternative	ADJ
bracis-28433	18	5	strategies	strategy	NOUN
bracis-28433	18	6	such	such	ADJ
bracis-28433	18	7	as	as	ADP
bracis-28433	18	8	token	token	VERB
bracis-28433	18	9	-	-	PUNCT
bracis-28433	18	10	level	level	NOUN
bracis-28433	18	11	[	[	X
bracis-28433	18	12	7	7	NUM
bracis-28433	18	13	]	]	PUNCT
bracis-28433	18	14	or	or	CCONJ
bracis-28433	18	15	subsentence	subsentence	NOUN
bracis-28433	18	16	-	-	PUNCT
bracis-28433	18	17	level	level	NOUN
bracis-28433	18	18	[	[	X
bracis-28433	18	19	16	16	NUM
bracis-28433	18	20	]	]	PUNCT
bracis-28433	18	21	querying	query	VERB
bracis-28433	18	22	.	.	PUNCT
bracis-28433	19	1	these	these	DET
bracis-28433	19	2	strategies	strategy	NOUN
bracis-28433	19	3	aim	aim	VERB
bracis-28433	19	4	to	to	PART
bracis-28433	19	5	make	make	VERB
bracis-28433	19	6	the	the	DET
bracis-28433	19	7	annotation	annotation	NOUN
bracis-28433	19	8	process	process	NOUN
bracis-28433	19	9	more	more	ADV
bracis-28433	19	10	efficient	efficient	ADJ
bracis-28433	19	11	by	by	ADP
bracis-28433	19	12	reducing	reduce	VERB
bracis-28433	19	13	the	the	DET
bracis-28433	19	14	number	number	NOUN
bracis-28433	19	15	of	of	ADP
bracis-28433	19	16	tokens	token	NOUN
bracis-28433	19	17	that	that	PRON
bracis-28433	19	18	need	need	VERB
bracis-28433	19	19	to	to	PART
bracis-28433	19	20	be	be	AUX
bracis-28433	19	21	manually	manually	ADV
bracis-28433	19	22	labeled	label	VERB
bracis-28433	19	23	.	.	PUNCT
bracis-28433	20	1	this	this	DET
bracis-28433	20	2	paper	paper	NOUN
bracis-28433	20	3	proposes	propose	VERB
bracis-28433	20	4	cooperative	cooperative	ADJ
bracis-28433	20	5	approaches	approach	NOUN
bracis-28433	20	6	for	for	ADP
bracis-28433	20	7	reducing	reduce	VERB
bracis-28433	20	8	the	the	DET
bracis-28433	20	9	cost	cost	NOUN
bracis-28433	20	10	of	of	ADP
bracis-28433	20	11	annotation	annotation	NOUN
bracis-28433	20	12	by	by	ADP
bracis-28433	20	13	an	an	DET
bracis-28433	20	14	oracle	oracle	NOUN
bracis-28433	20	15	and	and	CCONJ
bracis-28433	20	16	speeding	speed	VERB
bracis-28433	20	17	up	up	ADP
bracis-28433	20	18	the	the	DET
bracis-28433	20	19	annotation	annotation	NOUN
bracis-28433	20	20	process	process	NOUN
bracis-28433	20	21	.	.	PUNCT
bracis-28433	21	1	we	we	PRON
bracis-28433	21	2	focus	focus	VERB
bracis-28433	21	3	on	on	ADP
bracis-28433	21	4	active	active	ADJ
bracis-28433	21	5	self	self	NOUN
bracis-28433	21	6	-	-	PUNCT
bracis-28433	21	7	learning	learn	VERB
bracis-28433	21	8	(	(	PUNCT
bracis-28433	21	9	asl	asl	NOUN
bracis-28433	21	10	)	)	PUNCT
bracis-28433	21	11	algorithms	algorithm	NOUN
bracis-28433	21	12	for	for	ADP
bracis-28433	21	13	that	that	PRON
bracis-28433	21	14	.	.	PUNCT
bracis-28433	22	1	the	the	DET
bracis-28433	22	2	most	most	ADV
bracis-28433	22	3	straightforward	straightforward	ADJ
bracis-28433	22	4	implementation	implementation	NOUN
bracis-28433	22	5	of	of	ADP
bracis-28433	22	6	an	an	DET
bracis-28433	22	7	active	active	ADJ
bracis-28433	22	8	self	self	NOUN
bracis-28433	22	9	-	-	PUNCT
bracis-28433	22	10	learning	learn	VERB
bracis-28433	22	11	algorithm	algorithm	NOUN
bracis-28433	22	12	uses	use	VERB
bracis-28433	22	13	the	the	DET
bracis-28433	22	14	machine	machine	NOUN
bracis-28433	22	15	learning	learn	VERB
bracis-28433	22	16	model	model	NOUN
bracis-28433	22	17	to	to	PART
bracis-28433	22	18	predict	predict	VERB
bracis-28433	22	19	classes	class	NOUN
bracis-28433	22	20	for	for	ADP
bracis-28433	22	21	the	the	DET
bracis-28433	22	22	unlabeled	unlabeled	ADJ
bracis-28433	22	23	data	datum	NOUN
bracis-28433	22	24	.	.	PUNCT
bracis-28433	23	1	the	the	DET
bracis-28433	23	2	predicted	predict	VERB
bracis-28433	23	3	labels	label	NOUN
bracis-28433	23	4	are	be	AUX
bracis-28433	23	5	then	then	ADV
bracis-28433	23	6	split	split	VERB
bracis-28433	23	7	into	into	ADP
bracis-28433	23	8	high	high	ADJ
bracis-28433	23	9	-	-	PUNCT
bracis-28433	23	10	confidence	confidence	NOUN
bracis-28433	23	11	and	and	CCONJ
bracis-28433	23	12	low	low	ADJ
bracis-28433	23	13	-	-	PUNCT
bracis-28433	23	14	confidence	confidence	NOUN
bracis-28433	23	15	,	,	PUNCT
bracis-28433	23	16	depending	depend	VERB
bracis-28433	23	17	on	on	ADP
bracis-28433	23	18	the	the	DET
bracis-28433	23	19	model	model	NOUN
bracis-28433	23	20	’s	’s	PART
bracis-28433	23	21	confidence	confidence	NOUN
bracis-28433	23	22	in	in	ADP
bracis-28433	23	23	its	its	PRON
bracis-28433	23	24	predictions	prediction	NOUN
bracis-28433	23	25	.	.	PUNCT
bracis-28433	24	1	high	high	ADJ
bracis-28433	24	2	-	-	PUNCT
bracis-28433	24	3	confidence	confidence	NOUN
bracis-28433	24	4	samples	sample	NOUN
bracis-28433	24	5	are	be	AUX
bracis-28433	24	6	labeled	label	VERB
bracis-28433	24	7	automatically	automatically	ADV
bracis-28433	24	8	using	use	VERB
bracis-28433	24	9	the	the	DET
bracis-28433	24	10	model	model	NOUN
bracis-28433	24	11	’s	’s	PART
bracis-28433	24	12	predictions	prediction	NOUN
bracis-28433	24	13	,	,	PUNCT
bracis-28433	24	14	while	while	SCONJ
bracis-28433	24	15	low	low	ADJ
bracis-28433	24	16	-	-	PUNCT
bracis-28433	24	17	confidence	confidence	NOUN
bracis-28433	24	18	samples	sample	NOUN
bracis-28433	24	19	are	be	AUX
bracis-28433	24	20	queried	query	VERB
bracis-28433	24	21	for	for	SCONJ
bracis-28433	24	22	the	the	DET
bracis-28433	24	23	human	human	ADJ
bracis-28433	24	24	annotator	annotator	NOUN
bracis-28433	24	25	to	to	PART
bracis-28433	24	26	be	be	AUX
bracis-28433	24	27	labeled	label	VERB
bracis-28433	24	28	manually	manually	ADV
bracis-28433	24	29	.	.	PUNCT
bracis-28433	25	1	labeled	label	VERB
bracis-28433	25	2	samples	sample	NOUN
bracis-28433	25	3	are	be	AUX
bracis-28433	25	4	added	add	VERB
bracis-28433	25	5	to	to	ADP
bracis-28433	25	6	the	the	DET
bracis-28433	25	7	labeled	label	VERB
bracis-28433	25	8	dataset	dataset	NOUN
bracis-28433	25	9	,	,	PUNCT
bracis-28433	25	10	which	which	PRON
bracis-28433	25	11	is	be	AUX
bracis-28433	25	12	used	use	VERB
bracis-28433	25	13	to	to	PART
bracis-28433	25	14	train	train	VERB
bracis-28433	25	15	the	the	DET
bracis-28433	25	16	machine	machine	NOUN
bracis-28433	25	17	-	-	PUNCT
bracis-28433	25	18	learning	learn	VERB
bracis-28433	25	19	model	model	NOUN
bracis-28433	25	20	further	far	ADV
bracis-28433	25	21	.	.	PUNCT
bracis-28433	26	1	this	this	DET
bracis-28433	26	2	process	process	NOUN
bracis-28433	26	3	is	be	AUX
bracis-28433	26	4	repeated	repeat	VERB
bracis-28433	26	5	until	until	SCONJ
bracis-28433	26	6	the	the	DET
bracis-28433	26	7	desired	desire	VERB
bracis-28433	26	8	level	level	NOUN
bracis-28433	26	9	of	of	ADP
bracis-28433	26	10	model	model	NOUN
bracis-28433	26	11	performance	performance	NOUN
bracis-28433	26	12	is	be	AUX
bracis-28433	26	13	achieved	achieve	VERB
bracis-28433	26	14	,	,	PUNCT
bracis-28433	26	15	or	or	CCONJ
bracis-28433	26	16	a	a	DET
bracis-28433	26	17	stopping	stop	VERB
bracis-28433	26	18	criterion	criterion	NOUN
bracis-28433	26	19	is	be	AUX
bracis-28433	26	20	reached	reach	VERB
bracis-28433	26	21	.	.	PUNCT
bracis-28433	27	1	in	in	ADP
bracis-28433	27	2	active	active	ADJ
bracis-28433	27	3	self	self	NOUN
bracis-28433	27	4	-	-	PUNCT
bracis-28433	27	5	learning	learn	VERB
bracis-28433	27	6	algorithms	algorithm	NOUN
bracis-28433	27	7	,	,	PUNCT
bracis-28433	27	8	the	the	DET
bracis-28433	27	9	active	active	ADJ
bracis-28433	27	10	learning	learning	NOUN
bracis-28433	27	11	part	part	NOUN
bracis-28433	27	12	corresponds	correspond	VERB
bracis-28433	27	13	to	to	ADP
bracis-28433	27	14	the	the	DET
bracis-28433	27	15	query	query	NOUN
bracis-28433	27	16	to	to	ADP
bracis-28433	27	17	the	the	DET
bracis-28433	27	18	human	human	ADJ
bracis-28433	27	19	annotator	annotator	NOUN
bracis-28433	27	20	.	.	PUNCT
bracis-28433	28	1	in	in	ADP
bracis-28433	28	2	contrast	contrast	NOUN
bracis-28433	28	3	,	,	PUNCT
bracis-28433	28	4	the	the	DET
bracis-28433	28	5	self	self	NOUN
bracis-28433	28	6	-	-	PUNCT
bracis-28433	28	7	learning	learn	VERB
bracis-28433	28	8	part	part	NOUN
bracis-28433	28	9	corresponds	correspond	VERB
bracis-28433	28	10	to	to	ADP
bracis-28433	28	11	the	the	DET
bracis-28433	28	12	automatic	automatic	ADJ
bracis-28433	28	13	labeling	labeling	NOUN
bracis-28433	28	14	done	do	VERB
bracis-28433	28	15	by	by	ADP
bracis-28433	28	16	the	the	DET
bracis-28433	28	17	trained	train	VERB
bracis-28433	28	18	model	model	NOUN
bracis-28433	28	19	.	.	PUNCT
bracis-28433	29	1	the	the	DET
bracis-28433	29	2	asl	asl	NOUN
bracis-28433	29	3	creates	create	VERB
bracis-28433	29	4	a	a	DET
bracis-28433	29	5	cooperative	cooperative	ADJ
bracis-28433	29	6	scenario	scenario	NOUN
bracis-28433	29	7	where	where	SCONJ
bracis-28433	29	8	humans	human	NOUN
bracis-28433	29	9	and	and	CCONJ
bracis-28433	29	10	models	model	NOUN
bracis-28433	29	11	annotate	annotate	VERB
bracis-28433	29	12	data	datum	NOUN
bracis-28433	29	13	together	together	ADV
bracis-28433	29	14	and	and	CCONJ
bracis-28433	29	15	can	can	AUX
bracis-28433	29	16	potentially	potentially	ADV
bracis-28433	29	17	reduce	reduce	VERB
bracis-28433	29	18	annotation	annotation	NOUN
bracis-28433	29	19	costs	cost	NOUN
bracis-28433	29	20	compared	compare	VERB
bracis-28433	29	21	to	to	ADP
bracis-28433	29	22	using	use	VERB
bracis-28433	29	23	only	only	ADV
bracis-28433	29	24	active	active	ADJ
bracis-28433	29	25	learning	learning	NOUN
bracis-28433	29	26	strategies	strategy	NOUN
bracis-28433	29	27	[	[	X
bracis-28433	29	28	21	21	NUM
bracis-28433	29	29	]	]	PUNCT
bracis-28433	29	30	.	.	PUNCT
bracis-28433	30	1	our	our	PRON
bracis-28433	30	2	2	2	NUM
bracis-28433	30	3	main	main	ADJ
bracis-28433	30	4	contributions	contribution	NOUN
bracis-28433	30	5	in	in	ADP
bracis-28433	30	6	this	this	DET
bracis-28433	30	7	paper	paper	NOUN
bracis-28433	30	8	are	be	AUX
bracis-28433	30	9	:	:	PUNCT
bracis-28433	30	10	(	(	PUNCT
bracis-28433	30	11	1	1	X
bracis-28433	30	12	)	)	PUNCT
bracis-28433	30	13	we	we	PRON
bracis-28433	30	14	investigate	investigate	VERB
bracis-28433	30	15	the	the	DET
bracis-28433	30	16	impact	impact	NOUN
bracis-28433	30	17	of	of	ADP
bracis-28433	30	18	different	different	ADJ
bracis-28433	30	19	self	self	NOUN
bracis-28433	30	20	-	-	PUNCT
bracis-28433	30	21	training	training	NOUN
bracis-28433	30	22	techniques	technique	NOUN
bracis-28433	30	23	on	on	ADP
bracis-28433	30	24	active	active	ADJ
bracis-28433	30	25	self	self	NOUN
bracis-28433	30	26	-	-	PUNCT
bracis-28433	30	27	learning	learn	VERB
bracis-28433	30	28	algorithms	algorithm	NOUN
bracis-28433	30	29	;	;	PUNCT
bracis-28433	30	30	and	and	CCONJ
bracis-28433	30	31	(	(	PUNCT
bracis-28433	30	32	2	2	X
bracis-28433	30	33	)	)	PUNCT
bracis-28433	30	34	we	we	PRON
bracis-28433	30	35	propose	propose	VERB
bracis-28433	30	36	a	a	DET
bracis-28433	30	37	novel	novel	ADJ
bracis-28433	30	38	active	active	ADJ
bracis-28433	30	39	self	self	NOUN
bracis-28433	30	40	-	-	PUNCT
bracis-28433	30	41	learning	learn	VERB
bracis-28433	30	42	algorithm	algorithm	NOUN
bracis-28433	30	43	based	base	VERB
bracis-28433	30	44	on	on	ADP
bracis-28433	30	45	token	token	VERB
bracis-28433	30	46	-	-	PUNCT
bracis-28433	30	47	level	level	NOUN
bracis-28433	30	48	querying	querying	NOUN
bracis-28433	30	49	that	that	PRON
bracis-28433	30	50	achieves	achieve	VERB
bracis-28433	30	51	peak	peak	ADJ
bracis-28433	30	52	performance	performance	NOUN
bracis-28433	30	53	with	with	ADP
bracis-28433	30	54	less	less	ADJ
bracis-28433	30	55	hand	hand	NOUN
bracis-28433	30	56	-	-	PUNCT
bracis-28433	30	57	annotated	annotate	VERB
bracis-28433	30	58	data	datum	NOUN
bracis-28433	30	59	than	than	ADP
bracis-28433	30	60	previous	previous	ADJ
bracis-28433	30	61	works	work	NOUN
bracis-28433	30	62	.	.	PUNCT
bracis-28433	31	1	the	the	DET
bracis-28433	31	2	overview	overview	NOUN
bracis-28433	31	3	of	of	ADP
bracis-28433	31	4	the	the	DET
bracis-28433	31	5	paper	paper	NOUN
bracis-28433	31	6	is	be	AUX
bracis-28433	31	7	as	as	SCONJ
bracis-28433	31	8	follows	follow	VERB
bracis-28433	31	9	.	.	PUNCT
bracis-28433	32	1	in	in	ADP
bracis-28433	32	2	sect	sect	NOUN
bracis-28433	32	3	.	.	PUNCT
bracis-28433	32	4	 	 	SPACE
bracis-28433	32	5	2	2	NUM
bracis-28433	32	6	,	,	PUNCT
bracis-28433	32	7	we	we	PRON
bracis-28433	32	8	present	present	VERB
bracis-28433	32	9	relevant	relevant	ADJ
bracis-28433	32	10	works	work	NOUN
bracis-28433	32	11	from	from	ADP
bracis-28433	32	12	the	the	DET
bracis-28433	32	13	literature	literature	NOUN
bracis-28433	32	14	on	on	ADP
bracis-28433	32	15	both	both	CCONJ
bracis-28433	32	16	active	active	ADJ
bracis-28433	32	17	learning	learning	NOUN
bracis-28433	32	18	and	and	CCONJ
bracis-28433	32	19	active	active	ADJ
bracis-28433	32	20	self	self	NOUN
bracis-28433	32	21	-	-	PUNCT
bracis-28433	32	22	learning	learn	VERB
bracis-28433	32	23	algorithms	algorithm	NOUN
bracis-28433	32	24	applied	apply	VERB
bracis-28433	32	25	to	to	ADP
bracis-28433	32	26	named	name	VERB
bracis-28433	32	27	entity	entity	NOUN
bracis-28433	32	28	recognition	recognition	NOUN
bracis-28433	32	29	tasks	task	NOUN
bracis-28433	32	30	.	.	PUNCT
bracis-28433	33	1	section	section	NOUN
bracis-28433	33	2	 	 	SPACE
bracis-28433	33	3	3	3	NUM
bracis-28433	33	4	briefly	briefly	NOUN
bracis-28433	33	5	reviews	review	VERB
bracis-28433	33	6	the	the	DET
bracis-28433	33	7	active	active	ADJ
bracis-28433	33	8	self	self	NOUN
bracis-28433	33	9	-	-	PUNCT
bracis-28433	33	10	learning	learn	VERB
bracis-28433	33	11	algorithm	algorithm	NOUN
bracis-28433	33	12	from	from	ADP
bracis-28433	33	13	[	[	X
bracis-28433	33	14	2	2	NUM
bracis-28433	33	15	]	]	PUNCT
bracis-28433	33	16	based	base	VERB
bracis-28433	33	17	on	on	ADP
bracis-28433	33	18	sentence	sentence	NOUN
bracis-28433	33	19	-	-	PUNCT
bracis-28433	33	20	level	level	NOUN
bracis-28433	33	21	querying	querying	NOUN
bracis-28433	33	22	.	.	PUNCT
bracis-28433	34	1	in	in	ADP
bracis-28433	34	2	sect	sect	NOUN
bracis-28433	34	3	.	.	PUNCT
bracis-28433	34	4	 	 	SPACE
bracis-28433	34	5	4	4	NUM
bracis-28433	34	6	,	,	PUNCT
bracis-28433	34	7	we	we	PRON
bracis-28433	34	8	investigate	investigate	VERB
bracis-28433	34	9	the	the	DET
bracis-28433	34	10	impact	impact	NOUN
bracis-28433	34	11	of	of	ADP
bracis-28433	34	12	different	different	ADJ
bracis-28433	34	13	self	self	NOUN
bracis-28433	34	14	-	-	PUNCT
bracis-28433	34	15	learning	learn	VERB
bracis-28433	34	16	techniques	technique	NOUN
bracis-28433	34	17	on	on	ADP
bracis-28433	34	18	the	the	DET
bracis-28433	34	19	proposed	propose	VERB
bracis-28433	34	20	active	active	ADJ
bracis-28433	34	21	self	self	NOUN
bracis-28433	34	22	-	-	PUNCT
bracis-28433	34	23	learning	learn	VERB
bracis-28433	34	24	algorithm	algorithm	NOUN
bracis-28433	34	25	from	from	ADP
bracis-28433	34	26	[	[	X
bracis-28433	34	27	2	2	NUM
bracis-28433	34	28	]	]	PUNCT
bracis-28433	34	29	.	.	PUNCT
bracis-28433	35	1	in	in	ADP
bracis-28433	35	2	sect	sect	NOUN
bracis-28433	35	3	.	.	PUNCT
bracis-28433	35	4	 	 	SPACE
bracis-28433	35	5	5	5	NUM
bracis-28433	35	6	,	,	PUNCT
bracis-28433	35	7	we	we	PRON
bracis-28433	35	8	propose	propose	VERB
bracis-28433	35	9	a	a	DET
bracis-28433	35	10	novel	novel	ADJ
bracis-28433	35	11	active	active	ADJ
bracis-28433	35	12	self	self	NOUN
bracis-28433	35	13	-	-	PUNCT
bracis-28433	35	14	learning	learn	VERB
bracis-28433	35	15	algorithm	algorithm	NOUN
bracis-28433	35	16	based	base	VERB
bracis-28433	35	17	on	on	ADP
bracis-28433	35	18	token	token	VERB
bracis-28433	35	19	-	-	PUNCT
bracis-28433	35	20	level	level	NOUN
bracis-28433	35	21	querying	querying	NOUN
bracis-28433	35	22	,	,	PUNCT
bracis-28433	35	23	where	where	SCONJ
bracis-28433	35	24	both	both	CCONJ
bracis-28433	35	25	the	the	DET
bracis-28433	35	26	human	human	ADJ
bracis-28433	35	27	annotator	annotator	NOUN
bracis-28433	35	28	and	and	CCONJ
bracis-28433	35	29	machine	machine	NOUN
bracis-28433	35	30	learning	learning	NOUN
bracis-28433	35	31	model	model	NOUN
bracis-28433	35	32	cooperatively	cooperatively	ADV
bracis-28433	35	33	annotate	annotate	VERB
bracis-28433	35	34	tokens	token	NOUN
bracis-28433	35	35	from	from	ADP
bracis-28433	35	36	the	the	DET
bracis-28433	35	37	same	same	ADJ
bracis-28433	35	38	sentence	sentence	NOUN
bracis-28433	35	39	.	.	PUNCT
bracis-28433	36	1	section	section	NOUN
bracis-28433	36	2	 	 	SPACE
bracis-28433	36	3	7	7	NUM
bracis-28433	36	4	we	we	PRON
bracis-28433	36	5	present	present	VERB
bracis-28433	36	6	the	the	DET
bracis-28433	36	7	results	result	NOUN
bracis-28433	36	8	of	of	ADP
bracis-28433	36	9	our	our	PRON
bracis-28433	36	10	investigations	investigation	NOUN
bracis-28433	36	11	on	on	ADP
bracis-28433	36	12	the	the	DET
bracis-28433	36	13	asl	asl	NOUN
bracis-28433	36	14	algorithms	algorithm	NOUN
bracis-28433	36	15	proposed	propose	VERB
bracis-28433	36	16	in	in	ADP
bracis-28433	36	17	sects	sect	NOUN
bracis-28433	36	18	.	.	PUNCT
bracis-28433	36	19	 	 	SPACE
bracis-28433	36	20	4	4	NUM
bracis-28433	36	21	and	and	CCONJ
bracis-28433	36	22	 	 	SPACE
bracis-28433	36	23	5	5	NUM
bracis-28433	36	24	.	.	PUNCT
bracis-28433	37	1	finally	finally	ADV
bracis-28433	37	2	,	,	PUNCT
bracis-28433	37	3	in	in	ADP
bracis-28433	37	4	sect	sect	NOUN
bracis-28433	37	5	.	.	PUNCT
bracis-28433	37	6	 	 	SPACE
bracis-28433	37	7	8	8	NUM
bracis-28433	37	8	,	,	PUNCT
bracis-28433	37	9	we	we	PRON
bracis-28433	37	10	summarize	summarize	VERB
bracis-28433	37	11	the	the	DET
bracis-28433	37	12	results	result	NOUN
bracis-28433	37	13	from	from	ADP
bracis-28433	37	14	both	both	DET
bracis-28433	37	15	sentence	sentence	NOUN
bracis-28433	37	16	-	-	PUNCT
bracis-28433	37	17	level	level	NOUN
bracis-28433	37	18	and	and	CCONJ
bracis-28433	37	19	token	token	ADJ
bracis-28433	37	20	-	-	PUNCT
bracis-28433	37	21	level	level	NOUN
bracis-28433	37	22	active	active	ADJ
bracis-28433	37	23	self	self	NOUN
bracis-28433	37	24	-	-	PUNCT
bracis-28433	37	25	learning	learn	VERB
bracis-28433	37	26	algorithms	algorithm	NOUN
bracis-28433	37	27	.	.	PUNCT
bracis-28433	38	1	2	2	NUM
bracis-28433	38	2	related	relate	VERB
bracis-28433	38	3	works	work	VERB
bracis-28433	38	4	many	many	ADJ
bracis-28433	38	5	studies	study	NOUN
bracis-28433	38	6	in	in	ADP
bracis-28433	38	7	the	the	DET
bracis-28433	38	8	current	current	ADJ
bracis-28433	38	9	literature	literature	NOUN
bracis-28433	38	10	focusing	focus	VERB
bracis-28433	38	11	on	on	ADP
bracis-28433	38	12	active	active	ADJ
bracis-28433	38	13	learning	learning	NOUN
bracis-28433	38	14	algorithms	algorithm	NOUN
bracis-28433	38	15	for	for	ADP
bracis-28433	38	16	ner	ner	NOUN
bracis-28433	38	17	utilize	utilize	VERB
bracis-28433	38	18	sentence	sentence	NOUN
bracis-28433	38	19	-	-	PUNCT
bracis-28433	38	20	level	level	NOUN
bracis-28433	38	21	querying	querying	NOUN
bracis-28433	38	22	.	.	PUNCT
bracis-28433	39	1	shen	shen	PROPN
bracis-28433	39	2	et	et	PROPN
bracis-28433	39	3	al	al	PROPN
bracis-28433	39	4	.	.	PUNCT
bracis-28433	40	1	[	[	X
bracis-28433	40	2	18	18	NUM
bracis-28433	40	3	]	]	PUNCT
bracis-28433	40	4	proposes	propose	VERB
bracis-28433	40	5	,	,	PUNCT
bracis-28433	40	6	to	to	ADP
bracis-28433	40	7	the	the	DET
bracis-28433	40	8	best	good	ADJ
bracis-28433	40	9	of	of	ADP
bracis-28433	40	10	our	our	PRON
bracis-28433	40	11	knowledge	knowledge	NOUN
bracis-28433	40	12	,	,	PUNCT
bracis-28433	40	13	the	the	DET
bracis-28433	40	14	first	first	ADJ
bracis-28433	40	15	active	active	ADJ
bracis-28433	40	16	learning	learn	VERB
bracis-28433	40	17	algorithm	algorithm	NOUN
bracis-28433	40	18	based	base	VERB
bracis-28433	40	19	on	on	ADP
bracis-28433	40	20	neural	neural	ADJ
bracis-28433	40	21	networks	network	NOUN
bracis-28433	40	22	for	for	ADP
bracis-28433	40	23	tasks	task	NOUN
bracis-28433	40	24	of	of	ADP
bracis-28433	40	25	sequence	sequence	NOUN
bracis-28433	40	26	tagging	tagging	NOUN
bracis-28433	40	27	(	(	PUNCT
bracis-28433	40	28	e.g.	e.g.	ADV
bracis-28433	40	29	ner	ner	NOUN
bracis-28433	40	30	and	and	CCONJ
bracis-28433	40	31	part	part	NOUN
bracis-28433	40	32	-	-	PUNCT
bracis-28433	40	33	of	of	ADP
bracis-28433	40	34	-	-	PUNCT
bracis-28433	40	35	speech	speech	NOUN
bracis-28433	40	36	tagging	tagging	NOUN
bracis-28433	40	37	)	)	PUNCT
bracis-28433	40	38	.	.	PUNCT
bracis-28433	41	1	their	their	PRON
bracis-28433	41	2	proposed	propose	VERB
bracis-28433	41	3	neural	neural	ADJ
bracis-28433	41	4	model	model	NOUN
bracis-28433	41	5	uses	use	VERB
bracis-28433	41	6	convolutional	convolutional	ADJ
bracis-28433	41	7	layers	layer	NOUN
bracis-28433	41	8	for	for	ADP
bracis-28433	41	9	character	character	NOUN
bracis-28433	41	10	and	and	CCONJ
bracis-28433	41	11	word	word	NOUN
bracis-28433	41	12	-	-	PUNCT
bracis-28433	41	13	level	level	NOUN
bracis-28433	41	14	feature	feature	NOUN
bracis-28433	41	15	encoding	encoding	NOUN
bracis-28433	41	16	and	and	CCONJ
bracis-28433	41	17	an	an	DET
bracis-28433	41	18	lstm	lstm	NOUN
bracis-28433	41	19	layer	layer	NOUN
bracis-28433	41	20	with	with	ADP
bracis-28433	41	21	greedy	greedy	ADJ
bracis-28433	41	22	decoding	decode	VERB
bracis-28433	41	23	as	as	ADP
bracis-28433	41	24	the	the	DET
bracis-28433	41	25	tag	tag	NOUN
bracis-28433	41	26	decoder	decoder	NOUN
bracis-28433	41	27	.	.	PUNCT
bracis-28433	42	1	the	the	DET
bracis-28433	42	2	authors	author	NOUN
bracis-28433	42	3	also	also	ADV
bracis-28433	42	4	proposed	propose	VERB
bracis-28433	42	5	the	the	DET
bracis-28433	42	6	maximum	maximum	ADV
bracis-28433	42	7	normalized	normalize	VERB
bracis-28433	42	8	log	log	NOUN
bracis-28433	42	9	-	-	PUNCT
bracis-28433	42	10	probability	probability	NOUN
bracis-28433	42	11	(	(	PUNCT
bracis-28433	42	12	mnlp	mnlp	ADJ
bracis-28433	42	13	)	)	PUNCT
bracis-28433	42	14	sentence	sentence	NOUN
bracis-28433	42	15	-	-	PUNCT
bracis-28433	42	16	level	level	NOUN
bracis-28433	42	17	query	query	NOUN
bracis-28433	42	18	function	function	NOUN
bracis-28433	42	19	,	,	PUNCT
bracis-28433	42	20	which	which	PRON
bracis-28433	42	21	normalizes	normalize	VERB
bracis-28433	42	22	the	the	DET
bracis-28433	42	23	model	model	NOUN
bracis-28433	42	24	’s	’s	PART
bracis-28433	42	25	confidence	confidence	NOUN
bracis-28433	42	26	for	for	ADP
bracis-28433	42	27	a	a	DET
bracis-28433	42	28	given	give	VERB
bracis-28433	42	29	unlabeled	unlabeled	ADJ
bracis-28433	42	30	sentence	sentence	NOUN
bracis-28433	42	31	.	.	PUNCT
bracis-28433	43	1	their	their	PRON
bracis-28433	43	2	experiments	experiment	NOUN
bracis-28433	43	3	show	show	VERB
bracis-28433	43	4	that	that	SCONJ
bracis-28433	43	5	the	the	DET
bracis-28433	43	6	proposed	propose	VERB
bracis-28433	43	7	algorithm	algorithm	NOUN
bracis-28433	43	8	trains	train	VERB
bracis-28433	43	9	the	the	DET
bracis-28433	43	10	model	model	NOUN
bracis-28433	43	11	to	to	PART
bracis-28433	43	12	peak	peak	VERB
bracis-28433	43	13	performance	performance	NOUN
bracis-28433	43	14	using	use	VERB
bracis-28433	43	15	only	only	ADV
bracis-28433	43	16	25	25	NUM
bracis-28433	43	17	%	%	NOUN
bracis-28433	43	18	of	of	ADP
bracis-28433	43	19	the	the	DET
bracis-28433	43	20	training	training	NOUN
bracis-28433	43	21	set	set	NOUN
bracis-28433	43	22	of	of	ADP
bracis-28433	43	23	the	the	DET
bracis-28433	43	24	ontonotes5.0	ontonotes5.0	NUM
bracis-28433	43	25	dataset	dataset	VERB
bracis-28433	43	26	[	[	X
bracis-28433	43	27	15	15	NUM
bracis-28433	43	28	]	]	PUNCT
bracis-28433	43	29	.	.	PUNCT
bracis-28433	44	1	siddhant	siddhant	PROPN
bracis-28433	44	2	and	and	CCONJ
bracis-28433	44	3	lipton	lipton	PROPN
bracis-28433	45	1	[	[	X
bracis-28433	45	2	19	19	NUM
bracis-28433	45	3	]	]	PUNCT
bracis-28433	45	4	extend	extend	VERB
bracis-28433	45	5	previous	previous	ADJ
bracis-28433	45	6	work	work	NOUN
bracis-28433	45	7	by	by	ADP
bracis-28433	45	8	shen	shen	PROPN
bracis-28433	45	9	et	et	PROPN
bracis-28433	45	10	al	al	PROPN
bracis-28433	45	11	.	.	PUNCT
bracis-28433	46	1	they	they	PRON
bracis-28433	46	2	use	use	VERB
bracis-28433	46	3	the	the	DET
bracis-28433	46	4	bayesian	bayesian	NOUN
bracis-28433	46	5	active	active	ADJ
bracis-28433	46	6	learning	learn	VERB
bracis-28433	46	7	through	through	ADP
bracis-28433	46	8	disagreement	disagreement	NOUN
bracis-28433	46	9	[	[	X
bracis-28433	46	10	6	6	NUM
bracis-28433	46	11	]	]	X
bracis-28433	46	12	(	(	PUNCT
bracis-28433	46	13	bald	bald	ADJ
bracis-28433	46	14	)	)	PUNCT
bracis-28433	46	15	framework	framework	NOUN
bracis-28433	46	16	by	by	ADP
bracis-28433	46	17	querying	query	VERB
bracis-28433	46	18	the	the	DET
bracis-28433	46	19	unlabeled	unlabeled	ADJ
bracis-28433	46	20	sentences	sentence	NOUN
bracis-28433	46	21	that	that	PRON
bracis-28433	46	22	generate	generate	VERB
bracis-28433	46	23	the	the	DET
bracis-28433	46	24	most	most	ADV
bracis-28433	46	25	disagreement	disagreement	NOUN
bracis-28433	46	26	over	over	ADP
bracis-28433	46	27	multiple	multiple	ADJ
bracis-28433	46	28	passes	pass	NOUN
bracis-28433	46	29	on	on	ADP
bracis-28433	46	30	neural	neural	ADJ
bracis-28433	46	31	models	model	NOUN
bracis-28433	46	32	.	.	PUNCT
bracis-28433	47	1	to	to	PART
bracis-28433	47	2	introduce	introduce	VERB
bracis-28433	47	3	stochastic	stochastic	ADJ
bracis-28433	47	4	behavior	behavior	NOUN
bracis-28433	47	5	in	in	ADP
bracis-28433	47	6	the	the	DET
bracis-28433	47	7	neural	neural	ADJ
bracis-28433	47	8	models	model	NOUN
bracis-28433	47	9	,	,	PUNCT
bracis-28433	47	10	they	they	PRON
bracis-28433	47	11	propose	propose	VERB
bracis-28433	47	12	two	two	NUM
bracis-28433	47	13	solutions	solution	NOUN
bracis-28433	47	14	.	.	PUNCT
bracis-28433	48	1	the	the	DET
bracis-28433	48	2	first	first	ADJ
bracis-28433	48	3	solution	solution	NOUN
bracis-28433	48	4	is	be	AUX
bracis-28433	48	5	the	the	DET
bracis-28433	48	6	monte	monte	PROPN
bracis-28433	48	7	carlo	carlo	PROPN
bracis-28433	48	8	dropout	dropout	PROPN
bracis-28433	48	9	,	,	PUNCT
bracis-28433	48	10	where	where	SCONJ
bracis-28433	48	11	the	the	DET
bracis-28433	48	12	model	model	NOUN
bracis-28433	48	13	makes	make	VERB
bracis-28433	48	14	multiple	multiple	ADJ
bracis-28433	48	15	predictions	prediction	NOUN
bracis-28433	48	16	for	for	ADP
bracis-28433	48	17	the	the	DET
bracis-28433	48	18	same	same	ADJ
bracis-28433	48	19	sentence	sentence	NOUN
bracis-28433	48	20	but	but	CCONJ
bracis-28433	48	21	with	with	ADP
bracis-28433	48	22	a	a	DET
bracis-28433	48	23	different	different	ADJ
bracis-28433	48	24	dropout	dropout	NOUN
bracis-28433	48	25	mask	mask	NOUN
bracis-28433	48	26	at	at	ADP
bracis-28433	48	27	a	a	DET
bracis-28433	48	28	time	time	NOUN
bracis-28433	48	29	.	.	PUNCT
bracis-28433	49	1	the	the	DET
bracis-28433	49	2	second	second	ADJ
bracis-28433	49	3	solution	solution	NOUN
bracis-28433	49	4	is	be	AUX
bracis-28433	49	5	the	the	DET
bracis-28433	49	6	bayes	bayes	NOUN
bracis-28433	49	7	by	by	ADP
bracis-28433	49	8	backpropagation	backpropagation	NOUN
bracis-28433	49	9	,	,	PUNCT
bracis-28433	49	10	where	where	SCONJ
bracis-28433	49	11	some	some	DET
bracis-28433	49	12	layers	layer	NOUN
bracis-28433	49	13	have	have	VERB
bracis-28433	49	14	their	their	PRON
bracis-28433	49	15	deterministic	deterministic	ADJ
bracis-28433	49	16	parameters	parameter	NOUN
bracis-28433	49	17	replaced	replace	VERB
bracis-28433	49	18	by	by	ADP
bracis-28433	49	19	stochastic	stochastic	ADJ
bracis-28433	49	20	parameters	parameter	NOUN
bracis-28433	49	21	.	.	PUNCT
bracis-28433	50	1	experiments	experiment	NOUN
bracis-28433	50	2	showed	show	VERB
bracis-28433	50	3	that	that	SCONJ
bracis-28433	50	4	the	the	DET
bracis-28433	50	5	proposed	propose	VERB
bracis-28433	50	6	query	query	NOUN
bracis-28433	50	7	strategies	strategy	NOUN
bracis-28433	50	8	performed	perform	VERB
bracis-28433	50	9	consistently	consistently	ADV
bracis-28433	50	10	better	well	ADV
bracis-28433	50	11	than	than	ADP
bracis-28433	50	12	the	the	DET
bracis-28433	50	13	previous	previous	ADJ
bracis-28433	50	14	mnlp	mnlp	ADJ
bracis-28433	50	15	,	,	PUNCT
bracis-28433	50	16	but	but	CCONJ
bracis-28433	50	17	this	this	PRON
bracis-28433	50	18	came	come	VERB
bracis-28433	50	19	with	with	ADP
bracis-28433	50	20	more	more	ADV
bracis-28433	50	21	complex	complex	ADJ
bracis-28433	50	22	and	and	CCONJ
bracis-28433	50	23	costly	costly	ADJ
bracis-28433	50	24	compute	compute	NOUN
bracis-28433	50	25	query	query	NOUN
bracis-28433	50	26	functions	function	NOUN
bracis-28433	50	27	.	.	PUNCT
bracis-28433	51	1	in	in	ADP
bracis-28433	51	2	the	the	DET
bracis-28433	51	3	literature	literature	NOUN
bracis-28433	51	4	on	on	ADP
bracis-28433	51	5	active	active	ADJ
bracis-28433	51	6	learning	learning	NOUN
bracis-28433	51	7	applied	apply	VERB
bracis-28433	51	8	to	to	ADP
bracis-28433	51	9	named	name	VERB
bracis-28433	51	10	entity	entity	NOUN
bracis-28433	51	11	recognition	recognition	NOUN
bracis-28433	51	12	,	,	PUNCT
bracis-28433	51	13	sub	sub	NOUN
bracis-28433	51	14	-	-	ADJ
bracis-28433	51	15	sentence	sentence	NOUN
bracis-28433	51	16	-	-	PUNCT
bracis-28433	51	17	level	level	NOUN
bracis-28433	51	18	querying	querying	NOUN
bracis-28433	51	19	strategies	strategy	NOUN
bracis-28433	51	20	are	be	AUX
bracis-28433	51	21	found	find	VERB
bracis-28433	51	22	less	less	ADV
bracis-28433	51	23	often	often	ADV
bracis-28433	51	24	,	,	PUNCT
bracis-28433	51	25	including	include	VERB
bracis-28433	51	26	token	token	VERB
bracis-28433	51	27	-	-	PUNCT
bracis-28433	51	28	level	level	NOUN
bracis-28433	51	29	querying	query	VERB
bracis-28433	51	30	strategies	strategy	NOUN
bracis-28433	51	31	.	.	PUNCT
bracis-28433	52	1	the	the	DET
bracis-28433	52	2	main	main	ADJ
bracis-28433	52	3	challenge	challenge	NOUN
bracis-28433	52	4	is	be	AUX
bracis-28433	52	5	how	how	SCONJ
bracis-28433	52	6	to	to	PART
bracis-28433	52	7	implement	implement	VERB
bracis-28433	52	8	the	the	DET
bracis-28433	52	9	model	model	NOUN
bracis-28433	52	10	training	training	NOUN
bracis-28433	52	11	routine	routine	NOUN
bracis-28433	52	12	with	with	ADP
bracis-28433	52	13	sentences	sentence	NOUN
bracis-28433	52	14	that	that	PRON
bracis-28433	52	15	are	be	AUX
bracis-28433	52	16	partially	partially	ADV
bracis-28433	52	17	labeled	label	VERB
bracis-28433	52	18	.	.	PUNCT
bracis-28433	53	1	kobayashi	kobayashi	PROPN
bracis-28433	53	2	and	and	CCONJ
bracis-28433	53	3	wakabayashi	wakabayashi	NOUN
bracis-28433	54	1	[	[	X
bracis-28433	54	2	7	7	NUM
bracis-28433	54	3	]	]	X
bracis-28433	54	4	propose	propose	VERB
bracis-28433	54	5	the	the	DET
bracis-28433	54	6	use	use	NOUN
bracis-28433	54	7	of	of	ADP
bracis-28433	54	8	a	a	DET
bracis-28433	54	9	multi	multi	ADJ
bracis-28433	54	10	-	-	ADJ
bracis-28433	54	11	class	class	ADJ
bracis-28433	54	12	logistic	logistic	ADJ
bracis-28433	54	13	regression	regression	NOUN
bracis-28433	54	14	model	model	NOUN
bracis-28433	54	15	with	with	ADP
bracis-28433	54	16	point	point	NOUN
bracis-28433	54	17	-	-	PUNCT
bracis-28433	54	18	wise	wise	ADJ
bracis-28433	54	19	predictions	prediction	NOUN
bracis-28433	54	20	[	[	X
bracis-28433	54	21	12	12	NUM
bracis-28433	54	22	]	]	PUNCT
bracis-28433	54	23	.	.	PUNCT
bracis-28433	55	1	they	they	PRON
bracis-28433	55	2	can	can	AUX
bracis-28433	55	3	use	use	VERB
bracis-28433	55	4	token	token	VERB
bracis-28433	55	5	-	-	PUNCT
bracis-28433	55	6	level	level	NOUN
bracis-28433	55	7	queries	query	NOUN
bracis-28433	55	8	to	to	PART
bracis-28433	55	9	train	train	VERB
bracis-28433	55	10	their	their	PRON
bracis-28433	55	11	models	model	NOUN
bracis-28433	55	12	through	through	ADP
bracis-28433	55	13	this	this	DET
bracis-28433	55	14	method	method	NOUN
bracis-28433	55	15	.	.	PUNCT
bracis-28433	56	1	radmard	radmard	NOUN
bracis-28433	56	2	et	et	PROPN
bracis-28433	56	3	al	al	PROPN
bracis-28433	56	4	.	.	PUNCT
bracis-28433	57	1	[	[	X
bracis-28433	57	2	16	16	NUM
bracis-28433	57	3	]	]	PUNCT
bracis-28433	57	4	proposes	propose	VERB
bracis-28433	57	5	a	a	DET
bracis-28433	57	6	sub	sub	NOUN
bracis-28433	57	7	-	-	NOUN
bracis-28433	57	8	sentence	sentence	NOUN
bracis-28433	57	9	-	-	PUNCT
bracis-28433	57	10	based	base	VERB
bracis-28433	57	11	query	query	NOUN
bracis-28433	57	12	strategy	strategy	NOUN
bracis-28433	57	13	,	,	PUNCT
bracis-28433	57	14	where	where	SCONJ
bracis-28433	57	15	only	only	ADV
bracis-28433	57	16	the	the	DET
bracis-28433	57	17	most	most	ADV
bracis-28433	57	18	essential	essential	ADJ
bracis-28433	57	19	,	,	PUNCT
bracis-28433	57	20	non	non	ADJ
bracis-28433	57	21	-	-	ADJ
bracis-28433	57	22	overlapping	overlapping	ADJ
bracis-28433	57	23	sub	sub	NOUN
bracis-28433	57	24	-	-	NOUN
bracis-28433	57	25	sentence	sentence	NOUN
bracis-28433	57	26	are	be	AUX
bracis-28433	57	27	queried	query	VERB
bracis-28433	57	28	instead	instead	ADV
bracis-28433	57	29	of	of	ADP
bracis-28433	57	30	the	the	DET
bracis-28433	57	31	whole	whole	ADJ
bracis-28433	57	32	sentence	sentence	NOUN
bracis-28433	57	33	.	.	PUNCT
bracis-28433	58	1	the	the	DET
bracis-28433	58	2	strategy	strategy	NOUN
bracis-28433	58	3	is	be	AUX
bracis-28433	58	4	to	to	PART
bracis-28433	58	5	query	query	VERB
bracis-28433	58	6	sub	sub	NOUN
bracis-28433	58	7	-	-	NOUN
bracis-28433	58	8	sentence	sentence	NOUN
bracis-28433	58	9	without	without	ADP
bracis-28433	58	10	their	their	PRON
bracis-28433	58	11	surrounding	surround	VERB
bracis-28433	58	12	context	context	NOUN
bracis-28433	58	13	and	and	CCONJ
bracis-28433	58	14	store	store	VERB
bracis-28433	58	15	the	the	DET
bracis-28433	58	16	annotation	annotation	NOUN
bracis-28433	58	17	in	in	ADP
bracis-28433	58	18	a	a	DET
bracis-28433	58	19	dictionary	dictionary	NOUN
bracis-28433	58	20	that	that	PRON
bracis-28433	58	21	associates	associate	VERB
bracis-28433	58	22	each	each	DET
bracis-28433	58	23	sub	sub	NOUN
bracis-28433	58	24	-	-	NOUN
bracis-28433	58	25	sentence	sentence	NOUN
bracis-28433	58	26	with	with	ADP
bracis-28433	58	27	its	its	PRON
bracis-28433	58	28	corresponding	corresponding	ADJ
bracis-28433	58	29	accurate	accurate	ADJ
bracis-28433	58	30	labels	label	NOUN
bracis-28433	58	31	.	.	PUNCT
bracis-28433	59	1	annotations	annotation	NOUN
bracis-28433	59	2	for	for	ADP
bracis-28433	59	3	a	a	DET
bracis-28433	59	4	sub	sub	NOUN
bracis-28433	59	5	-	-	NOUN
bracis-28433	59	6	sentence	sentence	NOUN
bracis-28433	59	7	are	be	AUX
bracis-28433	59	8	propagated	propagate	VERB
bracis-28433	59	9	onto	onto	ADP
bracis-28433	59	10	the	the	DET
bracis-28433	59	11	unlabeled	unlabele	VERB
bracis-28433	59	12	dataset	dataset	NOUN
bracis-28433	59	13	,	,	PUNCT
bracis-28433	59	14	meaning	mean	VERB
bracis-28433	59	15	that	that	SCONJ
bracis-28433	59	16	all	all	DET
bracis-28433	59	17	occurrences	occurrence	NOUN
bracis-28433	59	18	of	of	ADP
bracis-28433	59	19	a	a	PRON
bracis-28433	59	20	particular	particular	ADJ
bracis-28433	59	21	sub	sub	NOUN
bracis-28433	59	22	-	-	NOUN
bracis-28433	59	23	sentence	sentence	NOUN
bracis-28433	59	24	are	be	AUX
bracis-28433	59	25	assigned	assign	VERB
bracis-28433	59	26	the	the	DET
bracis-28433	59	27	same	same	ADJ
bracis-28433	59	28	labels	label	NOUN
bracis-28433	59	29	.	.	PUNCT
bracis-28433	60	1	they	they	PRON
bracis-28433	60	2	then	then	ADV
bracis-28433	60	3	train	train	VERB
bracis-28433	60	4	a	a	DET
bracis-28433	60	5	neural	neural	ADJ
bracis-28433	60	6	model	model	NOUN
bracis-28433	60	7	using	use	VERB
bracis-28433	60	8	a	a	DET
bracis-28433	60	9	loss	loss	NOUN
bracis-28433	60	10	function	function	NOUN
bracis-28433	60	11	computed	compute	VERB
bracis-28433	60	12	only	only	ADV
bracis-28433	60	13	on	on	ADP
bracis-28433	60	14	labeled	label	VERB
bracis-28433	60	15	tokens	token	NOUN
bracis-28433	60	16	.	.	PUNCT
bracis-28433	61	1	active	active	ADJ
bracis-28433	61	2	self	self	NOUN
bracis-28433	61	3	-	-	PUNCT
bracis-28433	61	4	learning	learn	VERB
bracis-28433	61	5	algorithms	algorithm	NOUN
bracis-28433	61	6	have	have	AUX
bracis-28433	61	7	received	receive	VERB
bracis-28433	61	8	considerably	considerably	ADV
bracis-28433	61	9	less	less	ADJ
bracis-28433	61	10	attention	attention	NOUN
bracis-28433	61	11	in	in	ADP
bracis-28433	61	12	the	the	DET
bracis-28433	61	13	ner	ner	NOUN
bracis-28433	61	14	literature	literature	NOUN
bracis-28433	61	15	when	when	SCONJ
bracis-28433	61	16	compared	compare	VERB
bracis-28433	61	17	to	to	ADP
bracis-28433	61	18	purely	purely	ADV
bracis-28433	61	19	active	active	ADJ
bracis-28433	61	20	learning	learning	NOUN
bracis-28433	61	21	algorithms	algorithm	NOUN
bracis-28433	61	22	.	.	PUNCT
bracis-28433	62	1	tran	tran	PROPN
bracis-28433	62	2	et	et	PROPN
bracis-28433	62	3	al	al	PROPN
bracis-28433	62	4	.	.	PUNCT
bracis-28433	63	1	[	[	X
bracis-28433	63	2	21	21	NUM
bracis-28433	63	3	]	]	PUNCT
bracis-28433	63	4	proposes	propose	VERB
bracis-28433	63	5	an	an	DET
bracis-28433	63	6	active	active	ADJ
bracis-28433	63	7	self	self	NOUN
bracis-28433	63	8	-	-	PUNCT
bracis-28433	63	9	learning	learn	VERB
bracis-28433	63	10	algorithm	algorithm	NOUN
bracis-28433	63	11	based	base	VERB
bracis-28433	63	12	on	on	ADP
bracis-28433	63	13	conditional	conditional	ADJ
bracis-28433	63	14	random	random	ADJ
bracis-28433	63	15	fields	field	NOUN
bracis-28433	63	16	(	(	PUNCT
bracis-28433	63	17	crf	crf	NOUN
bracis-28433	63	18	)	)	PUNCT
bracis-28433	63	19	models	model	NOUN
bracis-28433	63	20	.	.	PUNCT
bracis-28433	64	1	the	the	DET
bracis-28433	64	2	active	active	ADJ
bracis-28433	64	3	learning	learning	NOUN
bracis-28433	64	4	process	process	NOUN
bracis-28433	64	5	is	be	AUX
bracis-28433	64	6	related	relate	VERB
bracis-28433	64	7	to	to	ADP
bracis-28433	64	8	using	use	VERB
bracis-28433	64	9	a	a	DET
bracis-28433	64	10	diversity	diversity	NOUN
bracis-28433	64	11	measure	measure	NOUN
bracis-28433	64	12	to	to	PART
bracis-28433	64	13	query	query	VERB
bracis-28433	64	14	the	the	DET
bracis-28433	64	15	most	most	ADV
bracis-28433	64	16	informative	informative	ADJ
bracis-28433	64	17	samples	sample	NOUN
bracis-28433	64	18	to	to	ADP
bracis-28433	64	19	the	the	DET
bracis-28433	64	20	oracle	oracle	NOUN
bracis-28433	64	21	.	.	PUNCT
bracis-28433	65	1	at	at	ADP
bracis-28433	65	2	the	the	DET
bracis-28433	65	3	same	same	ADJ
bracis-28433	65	4	time	time	NOUN
bracis-28433	65	5	,	,	PUNCT
bracis-28433	65	6	the	the	DET
bracis-28433	65	7	self	self	NOUN
bracis-28433	65	8	-	-	PUNCT
bracis-28433	65	9	learning	learn	VERB
bracis-28433	65	10	process	process	NOUN
bracis-28433	65	11	is	be	AUX
bracis-28433	65	12	related	relate	VERB
bracis-28433	65	13	to	to	ADP
bracis-28433	65	14	using	use	VERB
bracis-28433	65	15	the	the	DET
bracis-28433	65	16	trained	train	VERB
bracis-28433	65	17	crf	crf	NOUN
bracis-28433	65	18	model	model	NOUN
bracis-28433	65	19	from	from	ADP
bracis-28433	65	20	the	the	DET
bracis-28433	65	21	previous	previous	ADJ
bracis-28433	65	22	iteration	iteration	NOUN
bracis-28433	65	23	to	to	PART
bracis-28433	65	24	annotate	annotate	VERB
bracis-28433	65	25	unlabeled	unlabeled	ADJ
bracis-28433	65	26	sentences	sentence	NOUN
bracis-28433	65	27	with	with	ADP
bracis-28433	65	28	high	high	ADJ
bracis-28433	65	29	confidence	confidence	NOUN
bracis-28433	65	30	in	in	ADP
bracis-28433	65	31	its	its	PRON
bracis-28433	65	32	predictions	prediction	NOUN
bracis-28433	65	33	.	.	PUNCT
bracis-28433	66	1	the	the	DET
bracis-28433	66	2	dataset	dataset	NOUN
bracis-28433	66	3	used	use	VERB
bracis-28433	66	4	was	be	AUX
bracis-28433	66	5	composed	compose	VERB
bracis-28433	66	6	of	of	ADP
bracis-28433	66	7	sentences	sentence	NOUN
bracis-28433	66	8	extracted	extract	VERB
bracis-28433	66	9	from	from	ADP
bracis-28433	66	10	twitter	twitter	NOUN
bracis-28433	66	11	.	.	PUNCT
bracis-28433	67	1	the	the	DET
bracis-28433	67	2	experiments	experiment	NOUN
bracis-28433	67	3	compared	compare	VERB
bracis-28433	67	4	uncertainty	uncertainty	NOUN
bracis-28433	67	5	,	,	PUNCT
bracis-28433	67	6	diversity	diversity	NOUN
bracis-28433	67	7	query	query	NOUN
bracis-28433	67	8	strategies	strategy	NOUN
bracis-28433	67	9	,	,	PUNCT
bracis-28433	67	10	and	and	CCONJ
bracis-28433	67	11	active	active	ADJ
bracis-28433	67	12	learning	learning	NOUN
bracis-28433	67	13	algorithms	algorithm	NOUN
bracis-28433	67	14	with	with	ADP
bracis-28433	67	15	and	and	CCONJ
bracis-28433	67	16	without	without	ADP
bracis-28433	67	17	self	self	NOUN
bracis-28433	67	18	-	-	PUNCT
bracis-28433	67	19	learning	learning	NOUN
bracis-28433	67	20	.	.	PUNCT
bracis-28433	68	1	results	result	NOUN
bracis-28433	68	2	have	have	AUX
bracis-28433	68	3	shown	show	VERB
bracis-28433	68	4	that	that	SCONJ
bracis-28433	68	5	active	active	ADJ
bracis-28433	68	6	self	self	NOUN
bracis-28433	68	7	-	-	PUNCT
bracis-28433	68	8	learning	learn	VERB
bracis-28433	68	9	algorithms	algorithm	NOUN
bracis-28433	68	10	achieved	achieve	VERB
bracis-28433	68	11	,	,	PUNCT
bracis-28433	68	12	in	in	ADP
bracis-28433	68	13	general	general	ADJ
bracis-28433	68	14	,	,	PUNCT
bracis-28433	68	15	better	well	ADJ
bracis-28433	68	16	results	result	NOUN
bracis-28433	68	17	than	than	ADP
bracis-28433	68	18	the	the	DET
bracis-28433	68	19	purely	purely	ADV
bracis-28433	68	20	active	active	ADJ
bracis-28433	68	21	learning	learning	NOUN
bracis-28433	68	22	algorithm	algorithm	NOUN
bracis-28433	68	23	.	.	PUNCT
bracis-28433	69	1	inspired	inspire	VERB
bracis-28433	69	2	by	by	ADP
bracis-28433	69	3	the	the	DET
bracis-28433	69	4	work	work	NOUN
bracis-28433	69	5	of	of	ADP
bracis-28433	69	6	tran	tran	PROPN
bracis-28433	69	7	et	et	PROPN
bracis-28433	69	8	al	al	PROPN
bracis-28433	69	9	.	.	PUNCT
bracis-28433	70	1	[	[	X
bracis-28433	70	2	21	21	NUM
bracis-28433	70	3	]	]	PUNCT
bracis-28433	70	4	,	,	PUNCT
bracis-28433	70	5	cunha	cunha	NOUN
bracis-28433	70	6	and	and	CCONJ
bracis-28433	70	7	faleiros	faleiro	VERB
bracis-28433	71	1	[	[	X
bracis-28433	71	2	2	2	NUM
bracis-28433	71	3	]	]	PUNCT
bracis-28433	71	4	presents	present	VERB
bracis-28433	71	5	another	another	DET
bracis-28433	71	6	active	active	ADJ
bracis-28433	71	7	self	self	NOUN
bracis-28433	71	8	-	-	PUNCT
bracis-28433	71	9	learning	learn	VERB
bracis-28433	71	10	algorithm	algorithm	NOUN
bracis-28433	71	11	.	.	PUNCT
bracis-28433	72	1	they	they	PRON
bracis-28433	72	2	made	make	VERB
bracis-28433	72	3	specific	specific	ADJ
bracis-28433	72	4	changes	change	NOUN
bracis-28433	72	5	to	to	PART
bracis-28433	72	6	address	address	VERB
bracis-28433	72	7	the	the	DET
bracis-28433	72	8	use	use	NOUN
bracis-28433	72	9	of	of	ADP
bracis-28433	72	10	deep	deep	ADJ
bracis-28433	72	11	neural	neural	ADJ
bracis-28433	72	12	models	model	NOUN
bracis-28433	72	13	instead	instead	ADV
bracis-28433	72	14	of	of	ADP
bracis-28433	72	15	crf	crf	PROPN
bracis-28433	72	16	.	.	PUNCT
bracis-28433	73	1	the	the	DET
bracis-28433	73	2	proposed	propose	VERB
bracis-28433	73	3	asl	asl	PROPN
bracis-28433	73	4	algorithm	algorithm	NOUN
bracis-28433	73	5	has	have	AUX
bracis-28433	73	6	been	be	AUX
bracis-28433	73	7	shown	show	VERB
bracis-28433	73	8	to	to	PART
bracis-28433	73	9	be	be	AUX
bracis-28433	73	10	less	less	ADV
bracis-28433	73	11	sensitive	sensitive	ADJ
bracis-28433	73	12	to	to	ADP
bracis-28433	73	13	the	the	DET
bracis-28433	73	14	quality	quality	NOUN
bracis-28433	73	15	of	of	ADP
bracis-28433	73	16	the	the	DET
bracis-28433	73	17	labeled	label	VERB
bracis-28433	73	18	set	set	NOUN
bracis-28433	73	19	in	in	ADP
bracis-28433	73	20	initial	initial	ADJ
bracis-28433	73	21	iterations	iteration	NOUN
bracis-28433	73	22	,	,	PUNCT
bracis-28433	73	23	when	when	SCONJ
bracis-28433	73	24	compared	compare	VERB
bracis-28433	73	25	to	to	ADP
bracis-28433	73	26	the	the	DET
bracis-28433	73	27	previous	previous	ADJ
bracis-28433	73	28	active	active	ADJ
bracis-28433	73	29	self	self	NOUN
bracis-28433	73	30	-	-	PUNCT
bracis-28433	73	31	learning	learn	VERB
bracis-28433	73	32	algorithm	algorithm	NOUN
bracis-28433	73	33	from	from	ADP
bracis-28433	73	34	the	the	DET
bracis-28433	73	35	literature	literature	NOUN
bracis-28433	73	36	[	[	X
bracis-28433	73	37	1	1	NUM
bracis-28433	73	38	]	]	PUNCT
bracis-28433	73	39	.	.	PUNCT
bracis-28433	74	1	3	3	NUM
bracis-28433	74	2	deep	deep	ADJ
bracis-28433	74	3	active	active	ADJ
bracis-28433	74	4	self	self	NOUN
bracis-28433	74	5	-	-	PUNCT
bracis-28433	74	6	learning	learn	VERB
bracis-28433	74	7	algorithm	algorithm	NOUN
bracis-28433	74	8	this	this	DET
bracis-28433	74	9	section	section	NOUN
bracis-28433	74	10	briefly	briefly	ADV
bracis-28433	74	11	describes	describe	VERB
bracis-28433	74	12	the	the	DET
bracis-28433	74	13	active	active	ADJ
bracis-28433	74	14	self	self	NOUN
bracis-28433	74	15	-	-	PUNCT
bracis-28433	74	16	learning	learn	VERB
bracis-28433	74	17	algorithm	algorithm	NOUN
bracis-28433	74	18	proposed	propose	VERB
bracis-28433	74	19	by	by	ADP
bracis-28433	74	20	cunha	cunha	NOUN
bracis-28433	74	21	and	and	CCONJ
bracis-28433	74	22	faleiros	faleiro	VERB
bracis-28433	75	1	[	[	X
bracis-28433	75	2	2	2	NUM
bracis-28433	75	3	]	]	PUNCT
bracis-28433	75	4	.	.	PUNCT
bracis-28433	76	1	the	the	DET
bracis-28433	76	2	active	active	ADJ
bracis-28433	76	3	self	self	NOUN
bracis-28433	76	4	-	-	PUNCT
bracis-28433	76	5	learning	learn	VERB
bracis-28433	76	6	algorithms	algorithm	NOUN
bracis-28433	76	7	described	describe	VERB
bracis-28433	76	8	in	in	ADP
bracis-28433	76	9	existing	exist	VERB
bracis-28433	76	10	literature	literature	NOUN
bracis-28433	76	11	require	require	VERB
bracis-28433	76	12	collaboration	collaboration	NOUN
bracis-28433	76	13	between	between	ADP
bracis-28433	76	14	a	a	DET
bracis-28433	76	15	human	human	ADJ
bracis-28433	76	16	annotator	annotator	NOUN
bracis-28433	76	17	and	and	CCONJ
bracis-28433	76	18	a	a	DET
bracis-28433	76	19	trained	train	VERB
bracis-28433	76	20	model	model	NOUN
bracis-28433	76	21	to	to	PART
bracis-28433	76	22	annotate	annotate	VERB
bracis-28433	76	23	samples	sample	NOUN
bracis-28433	76	24	from	from	ADP
bracis-28433	76	25	an	an	DET
bracis-28433	76	26	unlabeled	unlabele	VERB
bracis-28433	76	27	database	database	NOUN
bracis-28433	76	28	.	.	PUNCT
bracis-28433	77	1	however	however	ADV
bracis-28433	77	2	,	,	PUNCT
bracis-28433	77	3	this	this	DET
bracis-28433	77	4	process	process	NOUN
bracis-28433	77	5	is	be	AUX
bracis-28433	77	6	sensitive	sensitive	ADJ
bracis-28433	77	7	to	to	ADP
bracis-28433	77	8	the	the	DET
bracis-28433	77	9	initial	initial	ADJ
bracis-28433	77	10	labeled	label	VERB
bracis-28433	77	11	set	set	NOUN
bracis-28433	77	12	used	use	VERB
bracis-28433	77	13	to	to	PART
bracis-28433	77	14	train	train	VERB
bracis-28433	77	15	the	the	DET
bracis-28433	77	16	machine	machine	NOUN
bracis-28433	77	17	learning	learning	NOUN
bracis-28433	77	18	model	model	NOUN
bracis-28433	77	19	[	[	X
bracis-28433	77	20	1	1	NUM
bracis-28433	77	21	]	]	PUNCT
bracis-28433	77	22	.	.	PUNCT
bracis-28433	78	1	we	we	PRON
bracis-28433	78	2	argue	argue	VERB
bracis-28433	78	3	that	that	SCONJ
bracis-28433	78	4	this	this	DET
bracis-28433	78	5	sensitivity	sensitivity	NOUN
bracis-28433	78	6	arises	arise	VERB
bracis-28433	78	7	from	from	ADP
bracis-28433	78	8	poorly	poorly	ADV
bracis-28433	78	9	annotated	annotate	VERB
bracis-28433	78	10	samples	sample	NOUN
bracis-28433	78	11	selected	select	VERB
bracis-28433	78	12	by	by	ADP
bracis-28433	78	13	the	the	DET
bracis-28433	78	14	model	model	NOUN
bracis-28433	78	15	in	in	ADP
bracis-28433	78	16	the	the	DET
bracis-28433	78	17	early	early	ADJ
bracis-28433	78	18	rounds	round	NOUN
bracis-28433	78	19	of	of	ADP
bracis-28433	78	20	the	the	DET
bracis-28433	78	21	active	active	ADJ
bracis-28433	78	22	self	self	NOUN
bracis-28433	78	23	-	-	PUNCT
bracis-28433	78	24	learning	learn	VERB
bracis-28433	78	25	algorithm	algorithm	NOUN
bracis-28433	78	26	.	.	PUNCT
bracis-28433	79	1	this	this	DET
bracis-28433	79	2	poorly	poorly	ADV
bracis-28433	79	3	annotated	annotate	VERB
bracis-28433	79	4	data	datum	NOUN
bracis-28433	79	5	may	may	AUX
bracis-28433	79	6	introduce	introduce	VERB
bracis-28433	79	7	permanent	permanent	ADJ
bracis-28433	79	8	bias	bias	NOUN
bracis-28433	79	9	to	to	ADP
bracis-28433	79	10	the	the	DET
bracis-28433	79	11	labeled	label	VERB
bracis-28433	79	12	set	set	NOUN
bracis-28433	79	13	.	.	PUNCT
bracis-28433	80	1	to	to	PART
bracis-28433	80	2	address	address	VERB
bracis-28433	80	3	this	this	DET
bracis-28433	80	4	issue	issue	NOUN
bracis-28433	80	5	,	,	PUNCT
bracis-28433	80	6	the	the	DET
bracis-28433	80	7	work	work	NOUN
bracis-28433	80	8	of	of	ADP
bracis-28433	80	9	[	[	X
bracis-28433	80	10	2	2	NUM
bracis-28433	80	11	]	]	PUNCT
bracis-28433	80	12	proposed	propose	VERB
bracis-28433	80	13	an	an	DET
bracis-28433	80	14	active	active	ADJ
bracis-28433	80	15	self	self	NOUN
bracis-28433	80	16	-	-	PUNCT
bracis-28433	80	17	learning	learn	VERB
bracis-28433	80	18	algorithm	algorithm	NOUN
bracis-28433	80	19	that	that	PRON
bracis-28433	80	20	distinguishes	distinguish	VERB
bracis-28433	80	21	between	between	ADP
bracis-28433	80	22	samples	sample	NOUN
bracis-28433	80	23	labeled	label	VERB
bracis-28433	80	24	by	by	ADP
bracis-28433	80	25	the	the	DET
bracis-28433	80	26	model	model	NOUN
bracis-28433	80	27	and	and	CCONJ
bracis-28433	80	28	those	those	PRON
bracis-28433	80	29	labeled	label	VERB
bracis-28433	80	30	by	by	ADP
bracis-28433	80	31	the	the	DET
bracis-28433	80	32	human	human	ADJ
bracis-28433	80	33	annotator	annotator	NOUN
bracis-28433	80	34	.	.	PUNCT
bracis-28433	81	1	the	the	DET
bracis-28433	81	2	former	former	NOUN
bracis-28433	81	3	has	have	VERB
bracis-28433	81	4	less	less	ADJ
bracis-28433	81	5	impact	impact	NOUN
bracis-28433	81	6	on	on	ADP
bracis-28433	81	7	the	the	DET
bracis-28433	81	8	model	model	NOUN
bracis-28433	81	9	’s	’s	PART
bracis-28433	81	10	parameters	parameter	NOUN
bracis-28433	81	11	during	during	ADP
bracis-28433	81	12	training	training	NOUN
bracis-28433	81	13	.	.	PUNCT
bracis-28433	82	1	additionally	additionally	ADV
bracis-28433	82	2	,	,	PUNCT
bracis-28433	82	3	samples	sample	NOUN
bracis-28433	82	4	labeled	label	VERB
bracis-28433	82	5	automatically	automatically	ADV
bracis-28433	82	6	by	by	ADP
bracis-28433	82	7	the	the	DET
bracis-28433	82	8	model	model	NOUN
bracis-28433	82	9	were	be	AUX
bracis-28433	82	10	returned	return	VERB
bracis-28433	82	11	to	to	ADP
bracis-28433	82	12	the	the	DET
bracis-28433	82	13	unlabeled	unlabele	VERB
bracis-28433	82	14	set	set	VERB
bracis-28433	82	15	after	after	ADP
bracis-28433	82	16	each	each	DET
bracis-28433	82	17	iteration	iteration	NOUN
bracis-28433	82	18	of	of	ADP
bracis-28433	82	19	the	the	DET
bracis-28433	82	20	active	active	ADJ
bracis-28433	82	21	self	self	NOUN
bracis-28433	82	22	-	-	PUNCT
bracis-28433	82	23	learning	learn	VERB
bracis-28433	82	24	algorithm	algorithm	NOUN
bracis-28433	82	25	.	.	PUNCT
bracis-28433	83	1	these	these	DET
bracis-28433	83	2	modifications	modification	NOUN
bracis-28433	83	3	seem	seem	VERB
bracis-28433	83	4	to	to	PART
bracis-28433	83	5	mitigate	mitigate	VERB
bracis-28433	83	6	the	the	DET
bracis-28433	83	7	risk	risk	NOUN
bracis-28433	83	8	of	of	ADP
bracis-28433	83	9	introducing	introduce	VERB
bracis-28433	83	10	permanent	permanent	ADJ
bracis-28433	83	11	bias	bias	NOUN
bracis-28433	83	12	to	to	ADP
bracis-28433	83	13	the	the	DET
bracis-28433	83	14	trained	train	VERB
bracis-28433	83	15	model	model	NOUN
bracis-28433	83	16	and	and	CCONJ
bracis-28433	83	17	labeled	label	VERB
bracis-28433	83	18	dataset	dataset	NOUN
bracis-28433	83	19	.	.	PUNCT
bracis-28433	84	1	algorithm	algorithm	PROPN
bracis-28433	84	2	 	 	SPACE
bracis-28433	84	3	1	1	NUM
bracis-28433	84	4	presents	present	VERB
bracis-28433	84	5	a	a	DET
bracis-28433	84	6	more	more	ADV
bracis-28433	84	7	detailed	detailed	ADJ
bracis-28433	84	8	explanation	explanation	NOUN
bracis-28433	84	9	of	of	ADP
bracis-28433	84	10	active	active	ADJ
bracis-28433	84	11	self	self	NOUN
bracis-28433	84	12	-	-	PUNCT
bracis-28433	84	13	learning	learn	VERB
bracis-28433	84	14	algorithm	algorithm	NOUN
bracis-28433	84	15	.	.	PUNCT
bracis-28433	85	1	algorithm	algorithm	NOUN
bracis-28433	85	2	1	1	NUM
bracis-28433	85	3	.	.	PUNCT
bracis-28433	86	1	active	active	ADJ
bracis-28433	86	2	self	self	NOUN
bracis-28433	86	3	-	-	PUNCT
bracis-28433	86	4	learning	learn	VERB
bracis-28433	86	5	algorithm	algorithm	NOUN
bracis-28433	86	6	full	full	ADJ
bracis-28433	86	7	size	size	NOUN
bracis-28433	86	8	image	image	NOUN
bracis-28433	86	9	note	note	VERB
bracis-28433	86	10	that	that	SCONJ
bracis-28433	86	11	in	in	ADP
bracis-28433	86	12	algorithm	algorithm	NOUN
bracis-28433	86	13	 	 	SPACE
bracis-28433	86	14	1	1	NUM
bracis-28433	86	15	,	,	PUNCT
bracis-28433	86	16	m	m	VERB
bracis-28433	86	17	represents	represent	VERB
bracis-28433	86	18	the	the	DET
bracis-28433	86	19	machine	machine	NOUN
bracis-28433	86	20	learning	learning	NOUN
bracis-28433	86	21	model	model	NOUN
bracis-28433	86	22	,	,	PUNCT
bracis-28433	86	23	q	q	PROPN
bracis-28433	86	24	is	be	AUX
bracis-28433	86	25	the	the	DET
bracis-28433	86	26	query	query	NOUN
bracis-28433	86	27	budget	budget	NOUN
bracis-28433	86	28	,	,	PUNCT
bracis-28433	86	29	\(min\_confidence\	\(min\_confidence\	PROPN
bracis-28433	86	30	)	)	PUNCT
bracis-28433	86	31	is	be	AUX
bracis-28433	86	32	the	the	DET
bracis-28433	86	33	minimum	minimum	ADJ
bracis-28433	86	34	confidence	confidence	NOUN
bracis-28433	86	35	level	level	NOUN
bracis-28433	86	36	required	require	VERB
bracis-28433	86	37	for	for	ADP
bracis-28433	86	38	the	the	DET
bracis-28433	86	39	model	model	NOUN
bracis-28433	86	40	to	to	PART
bracis-28433	86	41	annotate	annotate	VERB
bracis-28433	86	42	an	an	DET
bracis-28433	86	43	unlabeled	unlabeled	ADJ
bracis-28433	86	44	sample	sample	NOUN
bracis-28433	86	45	,	,	PUNCT
bracis-28433	86	46	a.l	a.l	PROPN
bracis-28433	86	47	.	.	PROPN
bracis-28433	86	48	denotes	denote	VERB
bracis-28433	86	49	the	the	DET
bracis-28433	86	50	active	active	ADJ
bracis-28433	86	51	labeled	label	VERB
bracis-28433	86	52	set	set	NOUN
bracis-28433	86	53	that	that	PRON
bracis-28433	86	54	contains	contain	VERB
bracis-28433	86	55	samples	sample	NOUN
bracis-28433	86	56	annotated	annotate	VERB
bracis-28433	86	57	by	by	ADP
bracis-28433	86	58	the	the	DET
bracis-28433	86	59	oracle	oracle	NOUN
bracis-28433	86	60	,	,	PUNCT
bracis-28433	86	61	s.l	s.l	PROPN
bracis-28433	86	62	.	.	PROPN
bracis-28433	86	63	contains	contain	VERB
bracis-28433	86	64	samples	sample	NOUN
bracis-28433	86	65	labeled	label	VERB
bracis-28433	86	66	by	by	ADP
bracis-28433	86	67	the	the	DET
bracis-28433	86	68	trained	train	VERB
bracis-28433	86	69	model	model	NOUN
bracis-28433	86	70	,	,	PUNCT
bracis-28433	86	71	and	and	CCONJ
bracis-28433	86	72	u	u	NOUN
bracis-28433	86	73	represents	represent	VERB
bracis-28433	86	74	the	the	DET
bracis-28433	86	75	set	set	NOUN
bracis-28433	86	76	of	of	ADP
bracis-28433	86	77	unlabeled	unlabeled	ADJ
bracis-28433	86	78	data	datum	NOUN
bracis-28433	86	79	.	.	PUNCT
bracis-28433	87	1	the	the	DET
bracis-28433	87	2	active	active	ADJ
bracis-28433	87	3	learning	learning	NOUN
bracis-28433	87	4	procedure	procedure	NOUN
bracis-28433	87	5	,	,	PUNCT
bracis-28433	87	6	represented	represent	VERB
bracis-28433	87	7	by	by	ADP
bracis-28433	87	8	the	the	DET
bracis-28433	87	9	\(active\_learning\_query(\cdot	\(active\_learning\_query(\cdot	NOUN
bracis-28433	87	10	)	)	PUNCT
bracis-28433	87	11	\	\	ADJ
bracis-28433	87	12	)	)	PUNCT
bracis-28433	87	13	function	function	NOUN
bracis-28433	87	14	,	,	PUNCT
bracis-28433	87	15	identifies	identify	VERB
bracis-28433	87	16	the	the	DET
bracis-28433	87	17	most	most	ADV
bracis-28433	87	18	informative	informative	ADJ
bracis-28433	87	19	unlabeled	unlabele	VERB
bracis-28433	87	20	samples	sample	NOUN
bracis-28433	87	21	for	for	ADP
bracis-28433	87	22	annotation	annotation	NOUN
bracis-28433	87	23	by	by	ADP
bracis-28433	87	24	the	the	DET
bracis-28433	87	25	oracle	oracle	NOUN
bracis-28433	87	26	.	.	PUNCT
bracis-28433	88	1	the	the	DET
bracis-28433	88	2	self	self	NOUN
bracis-28433	88	3	-	-	PUNCT
bracis-28433	88	4	learning	learn	VERB
bracis-28433	88	5	procedure	procedure	NOUN
bracis-28433	88	6	,	,	PUNCT
bracis-28433	88	7	represented	represent	VERB
bracis-28433	88	8	by	by	ADP
bracis-28433	88	9	the	the	DET
bracis-28433	88	10	\(self\_learning\_query(\cdot	\(self\_learning\_query(\cdot	NOUN
bracis-28433	88	11	)	)	PUNCT
bracis-28433	88	12	\	\	ADJ
bracis-28433	88	13	)	)	PUNCT
bracis-28433	88	14	function	function	NOUN
bracis-28433	88	15	,	,	PUNCT
bracis-28433	88	16	corresponds	correspond	VERB
bracis-28433	88	17	to	to	ADP
bracis-28433	88	18	the	the	DET
bracis-28433	88	19	different	different	ADJ
bracis-28433	88	20	self	self	NOUN
bracis-28433	88	21	-	-	PUNCT
bracis-28433	88	22	learning	learn	VERB
bracis-28433	88	23	strategies	strategy	NOUN
bracis-28433	88	24	that	that	PRON
bracis-28433	88	25	will	will	AUX
bracis-28433	88	26	be	be	AUX
bracis-28433	88	27	applied	apply	VERB
bracis-28433	88	28	.	.	PUNCT
bracis-28433	89	1	we	we	PRON
bracis-28433	89	2	investigate	investigate	VERB
bracis-28433	89	3	the	the	DET
bracis-28433	89	4	impact	impact	NOUN
bracis-28433	89	5	of	of	ADP
bracis-28433	89	6	different	different	ADJ
bracis-28433	89	7	self	self	NOUN
bracis-28433	89	8	-	-	PUNCT
bracis-28433	89	9	learning	learn	VERB
bracis-28433	89	10	techniques	technique	NOUN
bracis-28433	89	11	based	base	VERB
bracis-28433	89	12	on	on	ADP
bracis-28433	89	13	sentence	sentence	NOUN
bracis-28433	89	14	-	-	PUNCT
bracis-28433	89	15	level	level	NOUN
bracis-28433	89	16	and	and	CCONJ
bracis-28433	89	17	word	word	NOUN
bracis-28433	89	18	-	-	PUNCT
bracis-28433	89	19	level	level	NOUN
bracis-28433	89	20	querying	query	VERB
bracis-28433	89	21	strategies	strategy	NOUN
bracis-28433	89	22	.	.	PUNCT
bracis-28433	90	1	4	4	NUM
bracis-28433	90	2	sentence	sentence	NOUN
bracis-28433	90	3	-	-	PUNCT
bracis-28433	90	4	level	level	NOUN
bracis-28433	90	5	active	active	ADJ
bracis-28433	90	6	self	self	NOUN
bracis-28433	90	7	-	-	PUNCT
bracis-28433	90	8	learning	learn	VERB
bracis-28433	90	9	algorithm	algorithm	NOUN
bracis-28433	90	10	this	this	DET
bracis-28433	90	11	section	section	NOUN
bracis-28433	90	12	introduces	introduce	VERB
bracis-28433	90	13	our	our	PRON
bracis-28433	90	14	proposed	propose	VERB
bracis-28433	90	15	active	active	ADJ
bracis-28433	90	16	self	self	NOUN
bracis-28433	90	17	-	-	PUNCT
bracis-28433	90	18	learning	learn	VERB
bracis-28433	90	19	algorithms	algorithm	NOUN
bracis-28433	90	20	that	that	PRON
bracis-28433	90	21	utilize	utilize	VERB
bracis-28433	90	22	sentence	sentence	NOUN
bracis-28433	90	23	-	-	PUNCT
bracis-28433	90	24	level	level	NOUN
bracis-28433	90	25	querying	querying	NOUN
bracis-28433	90	26	.	.	PUNCT
bracis-28433	91	1	this	this	DET
bracis-28433	91	2	algorithm	algorithm	NOUN
bracis-28433	91	3	is	be	AUX
bracis-28433	91	4	built	build	VERB
bracis-28433	91	5	upon	upon	SCONJ
bracis-28433	91	6	the	the	DET
bracis-28433	91	7	work	work	NOUN
bracis-28433	91	8	of	of	ADP
bracis-28433	91	9	[	[	X
bracis-28433	91	10	2	2	NUM
bracis-28433	91	11	]	]	PUNCT
bracis-28433	91	12	by	by	ADP
bracis-28433	91	13	incorporating	incorporate	VERB
bracis-28433	91	14	various	various	ADJ
bracis-28433	91	15	self	self	NOUN
bracis-28433	91	16	-	-	PUNCT
bracis-28433	91	17	learning	learn	VERB
bracis-28433	91	18	techniques	technique	NOUN
bracis-28433	91	19	.	.	PUNCT
bracis-28433	92	1	the	the	DET
bracis-28433	92	2	diagram	diagram	NOUN
bracis-28433	92	3	of	of	ADP
bracis-28433	92	4	the	the	DET
bracis-28433	92	5	active	active	ADJ
bracis-28433	92	6	self	self	NOUN
bracis-28433	92	7	-	-	PUNCT
bracis-28433	92	8	learning	learn	VERB
bracis-28433	92	9	algorithm	algorithm	NOUN
bracis-28433	92	10	proposed	propose	VERB
bracis-28433	92	11	is	be	AUX
bracis-28433	92	12	shown	show	VERB
bracis-28433	92	13	in	in	ADP
bracis-28433	92	14	fig	fig	NOUN
bracis-28433	92	15	.	.	PUNCT
bracis-28433	92	16	 	 	SPACE
bracis-28433	93	1	1	1	NUM
bracis-28433	93	2	.	.	X
bracis-28433	93	3	two	two	NUM
bracis-28433	93	4	main	main	ADJ
bracis-28433	93	5	characteristics	characteristic	NOUN
bracis-28433	93	6	make	make	VERB
bracis-28433	93	7	this	this	DET
bracis-28433	93	8	algorithm	algorithm	NOUN
bracis-28433	93	9	different	different	ADJ
bracis-28433	93	10	from	from	ADP
bracis-28433	93	11	previous	previous	ADJ
bracis-28433	93	12	algorithms	algorithm	NOUN
bracis-28433	93	13	in	in	ADP
bracis-28433	93	14	the	the	DET
bracis-28433	93	15	literature	literature	NOUN
bracis-28433	93	16	.	.	PUNCT
bracis-28433	94	1	(	(	PUNCT
bracis-28433	94	2	1	1	X
bracis-28433	94	3	)	)	PUNCT
bracis-28433	94	4	the	the	DET
bracis-28433	94	5	first	first	ADJ
bracis-28433	94	6	change	change	NOUN
bracis-28433	94	7	is	be	AUX
bracis-28433	94	8	that	that	SCONJ
bracis-28433	94	9	we	we	PRON
bracis-28433	94	10	now	now	ADV
bracis-28433	94	11	have	have	VERB
bracis-28433	94	12	two	two	NUM
bracis-28433	94	13	individual	individual	ADJ
bracis-28433	94	14	labeled	label	VERB
bracis-28433	94	15	sets	set	NOUN
bracis-28433	94	16	(	(	PUNCT
bracis-28433	94	17	shown	show	VERB
bracis-28433	94	18	in	in	ADP
bracis-28433	94	19	orange	orange	NOUN
bracis-28433	94	20	in	in	ADP
bracis-28433	94	21	fig	fig	NOUN
bracis-28433	94	22	.	.	PUNCT
bracis-28433	94	23	 	 	SPACE
bracis-28433	94	24	1	1	NUM
bracis-28433	94	25	)	)	PUNCT
bracis-28433	94	26	.	.	PUNCT
bracis-28433	95	1	one	one	NUM
bracis-28433	95	2	of	of	ADP
bracis-28433	95	3	these	these	DET
bracis-28433	95	4	labeled	label	VERB
bracis-28433	95	5	datasets	dataset	NOUN
bracis-28433	95	6	is	be	AUX
bracis-28433	95	7	responsible	responsible	ADJ
bracis-28433	95	8	for	for	ADP
bracis-28433	95	9	keeping	keep	VERB
bracis-28433	95	10	sentences	sentence	NOUN
bracis-28433	95	11	hand	hand	NOUN
bracis-28433	95	12	-	-	PUNCT
bracis-28433	95	13	annotated	annotate	VERB
bracis-28433	95	14	by	by	ADP
bracis-28433	95	15	the	the	DET
bracis-28433	95	16	oracle	oracle	NOUN
bracis-28433	95	17	,	,	PUNCT
bracis-28433	95	18	thus	thus	ADV
bracis-28433	95	19	having	have	VERB
bracis-28433	95	20	highly	highly	ADV
bracis-28433	95	21	reliable	reliable	ADJ
bracis-28433	95	22	labels	label	NOUN
bracis-28433	95	23	.	.	PUNCT
bracis-28433	96	1	at	at	ADP
bracis-28433	96	2	the	the	DET
bracis-28433	96	3	same	same	ADJ
bracis-28433	96	4	time	time	NOUN
bracis-28433	96	5	,	,	PUNCT
bracis-28433	96	6	the	the	DET
bracis-28433	96	7	other	other	ADJ
bracis-28433	96	8	is	be	AUX
bracis-28433	96	9	responsible	responsible	ADJ
bracis-28433	96	10	for	for	ADP
bracis-28433	96	11	keeping	keep	VERB
bracis-28433	96	12	sentences	sentence	NOUN
bracis-28433	96	13	labeled	label	VERB
bracis-28433	96	14	by	by	ADP
bracis-28433	96	15	the	the	DET
bracis-28433	96	16	machine	machine	NOUN
bracis-28433	96	17	learning	learn	VERB
bracis-28433	96	18	model	model	NOUN
bracis-28433	96	19	with	with	ADP
bracis-28433	96	20	less	less	ADV
bracis-28433	96	21	reliable	reliable	ADJ
bracis-28433	96	22	labels	label	NOUN
bracis-28433	96	23	.	.	PUNCT
bracis-28433	97	1	this	this	DET
bracis-28433	97	2	separation	separation	NOUN
bracis-28433	97	3	of	of	ADP
bracis-28433	97	4	highly	highly	ADV
bracis-28433	97	5	reliable	reliable	ADJ
bracis-28433	97	6	and	and	CCONJ
bracis-28433	97	7	less	less	ADV
bracis-28433	97	8	reliable	reliable	ADJ
bracis-28433	97	9	labeled	label	VERB
bracis-28433	97	10	data	datum	NOUN
bracis-28433	97	11	allows	allow	VERB
bracis-28433	97	12	us	we	PRON
bracis-28433	97	13	to	to	PART
bracis-28433	97	14	use	use	VERB
bracis-28433	97	15	them	they	PRON
bracis-28433	97	16	differently	differently	ADV
bracis-28433	97	17	during	during	ADP
bracis-28433	97	18	training	training	NOUN
bracis-28433	97	19	.	.	PUNCT
bracis-28433	98	1	we	we	PRON
bracis-28433	98	2	employ	employ	VERB
bracis-28433	98	3	hand	hand	NOUN
bracis-28433	98	4	-	-	PUNCT
bracis-28433	98	5	annotated	annotate	VERB
bracis-28433	98	6	data	datum	NOUN
bracis-28433	98	7	for	for	ADP
bracis-28433	98	8	traditional	traditional	ADJ
bracis-28433	98	9	supervised	supervised	ADJ
bracis-28433	98	10	learning	learning	NOUN
bracis-28433	98	11	and	and	CCONJ
bracis-28433	98	12	utilize	utilize	VERB
bracis-28433	98	13	data	datum	NOUN
bracis-28433	98	14	labeled	label	VERB
bracis-28433	98	15	by	by	ADP
bracis-28433	98	16	the	the	DET
bracis-28433	98	17	model	model	NOUN
bracis-28433	98	18	for	for	ADP
bracis-28433	98	19	self	self	NOUN
bracis-28433	98	20	-	-	PUNCT
bracis-28433	98	21	learning	learning	NOUN
bracis-28433	98	22	.	.	PUNCT
bracis-28433	99	1	(	(	PUNCT
bracis-28433	99	2	2	2	X
bracis-28433	99	3	)	)	PUNCT
bracis-28433	99	4	the	the	DET
bracis-28433	99	5	second	second	ADJ
bracis-28433	99	6	change	change	NOUN
bracis-28433	99	7	is	be	AUX
bracis-28433	99	8	that	that	SCONJ
bracis-28433	99	9	after	after	ADP
bracis-28433	99	10	training	train	VERB
bracis-28433	99	11	the	the	DET
bracis-28433	99	12	machine	machine	NOUN
bracis-28433	99	13	learning	learn	VERB
bracis-28433	99	14	model	model	NOUN
bracis-28433	99	15	in	in	ADP
bracis-28433	99	16	any	any	DET
bracis-28433	99	17	given	give	VERB
bracis-28433	99	18	algorithm	algorithm	NOUN
bracis-28433	99	19	iteration	iteration	NOUN
bracis-28433	99	20	,	,	PUNCT
bracis-28433	99	21	the	the	DET
bracis-28433	99	22	data	data	NOUN
bracis-28433	99	23	self	self	NOUN
bracis-28433	99	24	-	-	PUNCT
bracis-28433	99	25	labeled	label	VERB
bracis-28433	99	26	by	by	ADP
bracis-28433	99	27	the	the	DET
bracis-28433	99	28	machine	machine	NOUN
bracis-28433	99	29	learning	learn	VERB
bracis-28433	99	30	model	model	NOUN
bracis-28433	99	31	returns	return	NOUN
bracis-28433	99	32	to	to	ADP
bracis-28433	99	33	the	the	DET
bracis-28433	99	34	unlabeled	unlabele	VERB
bracis-28433	99	35	data	data	NOUN
bracis-28433	99	36	pool	pool	NOUN
bracis-28433	99	37	.	.	PUNCT
bracis-28433	100	1	these	these	DET
bracis-28433	100	2	alterations	alteration	NOUN
bracis-28433	100	3	allow	allow	VERB
bracis-28433	100	4	a	a	DET
bracis-28433	100	5	better	well	ADV
bracis-28433	100	6	-	-	PUNCT
bracis-28433	100	7	trained	train	VERB
bracis-28433	100	8	model	model	NOUN
bracis-28433	100	9	to	to	PART
bracis-28433	100	10	re	re	VERB
bracis-28433	100	11	-	-	VERB
bracis-28433	100	12	annotate	annotate	VERB
bracis-28433	100	13	samples	sample	NOUN
bracis-28433	100	14	in	in	ADP
bracis-28433	100	15	later	late	ADJ
bracis-28433	100	16	iterations	iteration	NOUN
bracis-28433	100	17	.	.	PUNCT
bracis-28433	101	1	fig	fig	NOUN
bracis-28433	101	2	.	.	PUNCT
bracis-28433	102	1	1	1	NUM
bracis-28433	102	2	.	.	X
bracis-28433	102	3	annotation	annotation	NOUN
bracis-28433	102	4	process	process	NOUN
bracis-28433	102	5	for	for	ADP
bracis-28433	102	6	a	a	DET
bracis-28433	102	7	sentence	sentence	NOUN
bracis-28433	102	8	-	-	PUNCT
bracis-28433	102	9	level	level	NOUN
bracis-28433	102	10	active	active	ADJ
bracis-28433	102	11	-	-	PUNCT
bracis-28433	102	12	self	self	NOUN
bracis-28433	102	13	learning	learn	VERB
bracis-28433	102	14	algorithm	algorithm	NOUN
bracis-28433	102	15	.	.	PUNCT
bracis-28433	103	1	(	(	PUNCT
bracis-28433	103	2	color	color	NOUN
bracis-28433	103	3	figure	figure	NOUN
bracis-28433	103	4	online	online	ADV
bracis-28433	103	5	)	)	PUNCT
bracis-28433	103	6	full	full	ADJ
bracis-28433	103	7	size	size	NOUN
bracis-28433	103	8	image	image	NOUN
bracis-28433	103	9	the	the	DET
bracis-28433	103	10	algorithm	algorithm	NOUN
bracis-28433	103	11	proposed	propose	VERB
bracis-28433	103	12	in	in	ADP
bracis-28433	103	13	[	[	X
bracis-28433	103	14	2	2	NUM
bracis-28433	103	15	]	]	PUNCT
bracis-28433	103	16	adopted	adopt	VERB
bracis-28433	103	17	pseudo	pseudo	NOUN
bracis-28433	103	18	-	-	NOUN
bracis-28433	103	19	labeling	labeling	NOUN
bracis-28433	103	20	as	as	ADP
bracis-28433	103	21	the	the	DET
bracis-28433	103	22	self	self	NOUN
bracis-28433	103	23	-	-	PUNCT
bracis-28433	103	24	learning	learn	VERB
bracis-28433	103	25	strategy	strategy	NOUN
bracis-28433	103	26	.	.	PUNCT
bracis-28433	104	1	here	here	ADV
bracis-28433	104	2	,	,	PUNCT
bracis-28433	104	3	our	our	PRON
bracis-28433	104	4	alternate	alternate	ADJ
bracis-28433	104	5	version	version	NOUN
bracis-28433	104	6	replaces	replace	VERB
bracis-28433	104	7	the	the	DET
bracis-28433	104	8	pseudo	pseudo	NOUN
bracis-28433	104	9	-	-	ADJ
bracis-28433	104	10	labeling	labeling	NOUN
bracis-28433	104	11	technique	technique	NOUN
bracis-28433	104	12	with	with	ADP
bracis-28433	104	13	self	self	NOUN
bracis-28433	104	14	-	-	PUNCT
bracis-28433	104	15	learning	learn	VERB
bracis-28433	104	16	methods	method	NOUN
bracis-28433	104	17	.	.	PUNCT
bracis-28433	105	1	we	we	PRON
bracis-28433	105	2	test	test	VERB
bracis-28433	105	3	three	three	NUM
bracis-28433	105	4	different	different	ADJ
bracis-28433	105	5	self	self	NOUN
bracis-28433	105	6	-	-	PUNCT
bracis-28433	105	7	learning	learn	VERB
bracis-28433	105	8	techniques	technique	NOUN
bracis-28433	105	9	,	,	PUNCT
bracis-28433	105	10	namely	namely	ADV
bracis-28433	105	11	:	:	PUNCT
bracis-28433	105	12	(	(	PUNCT
bracis-28433	105	13	1	1	X
bracis-28433	105	14	)	)	PUNCT
bracis-28433	105	15	word	word	NOUN
bracis-28433	105	16	dropout	dropout	NOUN
bracis-28433	105	17	,	,	PUNCT
bracis-28433	105	18	(	(	PUNCT
bracis-28433	105	19	2	2	X
bracis-28433	105	20	)	)	PUNCT
bracis-28433	105	21	virtual	virtual	ADJ
bracis-28433	105	22	adversarial	adversarial	ADJ
bracis-28433	105	23	training	training	NOUN
bracis-28433	105	24	,	,	PUNCT
bracis-28433	105	25	and	and	CCONJ
bracis-28433	105	26	(	(	PUNCT
bracis-28433	105	27	3	3	X
bracis-28433	105	28	)	)	PUNCT
bracis-28433	105	29	cross	cross	ADJ
bracis-28433	105	30	-	-	ADJ
bracis-28433	105	31	view	view	ADJ
bracis-28433	105	32	training	training	NOUN
bracis-28433	105	33	.	.	PUNCT
bracis-28433	106	1	next	next	ADV
bracis-28433	106	2	,	,	PUNCT
bracis-28433	106	3	we	we	PRON
bracis-28433	106	4	will	will	AUX
bracis-28433	106	5	explain	explain	VERB
bracis-28433	106	6	how	how	SCONJ
bracis-28433	106	7	our	our	PRON
bracis-28433	106	8	algorithm	algorithm	NOUN
bracis-28433	106	9	utilized	utilize	VERB
bracis-28433	106	10	pseudo	pseudo	NOUN
bracis-28433	106	11	-	-	NOUN
bracis-28433	106	12	labeling	labeling	NOUN
bracis-28433	106	13	and	and	CCONJ
bracis-28433	106	14	integrated	integrate	VERB
bracis-28433	106	15	these	these	DET
bracis-28433	106	16	three	three	NUM
bracis-28433	106	17	techniques	technique	NOUN
bracis-28433	106	18	:	:	PUNCT
bracis-28433	106	19	the	the	DET
bracis-28433	106	20	pseudo	pseudo	NOUN
bracis-28433	106	21	-	-	NOUN
bracis-28433	106	22	labeling	labeling	NOUN
bracis-28433	106	23	(	(	PUNCT
bracis-28433	106	24	pl	pl	NOUN
bracis-28433	106	25	)	)	PUNCT
bracis-28433	106	26	technique	technique	NOUN
bracis-28433	106	27	identifies	identify	VERB
bracis-28433	106	28	highly	highly	ADV
bracis-28433	106	29	reliable	reliable	ADJ
bracis-28433	106	30	unlabeled	unlabele	VERB
bracis-28433	106	31	samples	sample	NOUN
bracis-28433	106	32	to	to	PART
bracis-28433	106	33	be	be	AUX
bracis-28433	106	34	automatically	automatically	ADV
bracis-28433	106	35	annotated	annotate	VERB
bracis-28433	106	36	by	by	ADP
bracis-28433	106	37	the	the	DET
bracis-28433	106	38	model	model	NOUN
bracis-28433	106	39	.	.	PUNCT
bracis-28433	107	1	in	in	ADP
bracis-28433	107	2	this	this	DET
bracis-28433	107	3	case	case	NOUN
bracis-28433	107	4	,	,	PUNCT
bracis-28433	107	5	highly	highly	ADV
bracis-28433	107	6	reliable	reliable	ADJ
bracis-28433	107	7	means	mean	VERB
bracis-28433	107	8	that	that	SCONJ
bracis-28433	107	9	the	the	DET
bracis-28433	107	10	model	model	NOUN
bracis-28433	107	11	’s	’s	PART
bracis-28433	107	12	confidence	confidence	NOUN
bracis-28433	107	13	in	in	ADP
bracis-28433	107	14	its	its	PRON
bracis-28433	107	15	predictions	prediction	NOUN
bracis-28433	107	16	is	be	AUX
bracis-28433	107	17	above	above	ADP
bracis-28433	107	18	a	a	DET
bracis-28433	107	19	predefined	predefine	VERB
bracis-28433	107	20	threshold	threshold	NOUN
bracis-28433	107	21	.	.	PUNCT
bracis-28433	108	1	our	our	PRON
bracis-28433	108	2	strategy	strategy	NOUN
bracis-28433	108	3	consists	consist	VERB
bracis-28433	108	4	of	of	ADP
bracis-28433	108	5	using	use	VERB
bracis-28433	108	6	the	the	DET
bracis-28433	108	7	machine	machine	NOUN
bracis-28433	108	8	learning	learn	VERB
bracis-28433	108	9	model	model	NOUN
bracis-28433	108	10	’s	’s	PART
bracis-28433	108	11	predictions	prediction	NOUN
bracis-28433	108	12	as	as	ADP
bracis-28433	108	13	labels	label	NOUN
bracis-28433	108	14	.	.	PUNCT
bracis-28433	109	1	then	then	ADV
bracis-28433	109	2	,	,	PUNCT
bracis-28433	109	3	apply	apply	VERB
bracis-28433	109	4	supervised	supervised	ADJ
bracis-28433	109	5	training	training	NOUN
bracis-28433	109	6	with	with	ADP
bracis-28433	109	7	the	the	DET
bracis-28433	109	8	pseudo	pseudo	NOUN
bracis-28433	109	9	-	-	VERB
bracis-28433	109	10	labeled	label	VERB
bracis-28433	109	11	and	and	CCONJ
bracis-28433	109	12	the	the	DET
bracis-28433	109	13	hand	hand	NOUN
bracis-28433	109	14	-	-	PUNCT
bracis-28433	109	15	labeled	label	VERB
bracis-28433	109	16	data	datum	NOUN
bracis-28433	109	17	.	.	PUNCT
bracis-28433	110	1	the	the	DET
bracis-28433	110	2	cross	cross	ADJ
bracis-28433	110	3	-	-	ADJ
bracis-28433	110	4	view	view	ADJ
bracis-28433	110	5	training	training	NOUN
bracis-28433	110	6	(	(	PUNCT
bracis-28433	110	7	cvt	cvt	PROPN
bracis-28433	110	8	)	)	PUNCT
bracis-28433	111	1	[	[	X
bracis-28433	111	2	3	3	X
bracis-28433	111	3	]	]	PUNCT
bracis-28433	111	4	aims	aim	VERB
bracis-28433	111	5	to	to	PART
bracis-28433	111	6	improve	improve	VERB
bracis-28433	111	7	the	the	DET
bracis-28433	111	8	representation	representation	NOUN
bracis-28433	111	9	capabilities	capability	NOUN
bracis-28433	111	10	of	of	ADP
bracis-28433	111	11	the	the	DET
bracis-28433	111	12	model	model	NOUN
bracis-28433	111	13	by	by	ADP
bracis-28433	111	14	forcing	force	VERB
bracis-28433	111	15	it	it	PRON
bracis-28433	111	16	to	to	PART
bracis-28433	111	17	output	output	VERB
bracis-28433	111	18	similar	similar	ADJ
bracis-28433	111	19	predictions	prediction	NOUN
bracis-28433	111	20	with	with	ADP
bracis-28433	111	21	different	different	ADJ
bracis-28433	111	22	views	view	NOUN
bracis-28433	111	23	of	of	ADP
bracis-28433	111	24	the	the	DET
bracis-28433	111	25	same	same	ADJ
bracis-28433	111	26	input	input	NOUN
bracis-28433	111	27	data	datum	NOUN
bracis-28433	111	28	.	.	PUNCT
bracis-28433	112	1	the	the	DET
bracis-28433	112	2	cvt	cvt	PROPN
bracis-28433	112	3	algorithm	algorithm	NOUN
bracis-28433	112	4	uses	use	VERB
bracis-28433	112	5	neural	neural	ADJ
bracis-28433	112	6	models	model	NOUN
bracis-28433	112	7	with	with	ADP
bracis-28433	112	8	auxiliary	auxiliary	ADJ
bracis-28433	112	9	classification	classification	NOUN
bracis-28433	112	10	heads	head	NOUN
bracis-28433	112	11	.	.	PUNCT
bracis-28433	113	1	each	each	DET
bracis-28433	113	2	head	head	NOUN
bracis-28433	113	3	learns	learn	VERB
bracis-28433	113	4	to	to	PART
bracis-28433	113	5	predict	predict	VERB
bracis-28433	113	6	tokens	token	NOUN
bracis-28433	113	7	given	give	VERB
bracis-28433	113	8	a	a	DET
bracis-28433	113	9	limited	limited	ADJ
bracis-28433	113	10	view	view	NOUN
bracis-28433	113	11	of	of	ADP
bracis-28433	113	12	the	the	DET
bracis-28433	113	13	input	input	NOUN
bracis-28433	113	14	sentence	sentence	NOUN
bracis-28433	113	15	(	(	PUNCT
bracis-28433	113	16	e.g.	e.g.	ADV
bracis-28433	113	17	,	,	PUNCT
bracis-28433	113	18	left	left	ADJ
bracis-28433	113	19	-	-	PUNCT
bracis-28433	113	20	context	context	NOUN
bracis-28433	113	21	only	only	ADV
bracis-28433	113	22	,	,	PUNCT
bracis-28433	113	23	right	right	ADJ
bracis-28433	113	24	-	-	PUNCT
bracis-28433	113	25	context	context	NOUN
bracis-28433	113	26	only	only	ADV
bracis-28433	113	27	)	)	PUNCT
bracis-28433	113	28	.	.	PUNCT
bracis-28433	114	1	the	the	DET
bracis-28433	114	2	auxiliary	auxiliary	ADJ
bracis-28433	114	3	classification	classification	NOUN
bracis-28433	114	4	head	head	NOUN
bracis-28433	114	5	comprises	comprise	VERB
bracis-28433	114	6	a	a	DET
bracis-28433	114	7	fully	fully	ADV
bracis-28433	114	8	-	-	PUNCT
bracis-28433	114	9	connected	connect	VERB
bracis-28433	114	10	layer	layer	NOUN
bracis-28433	114	11	with	with	ADP
bracis-28433	114	12	relu	relu	NOUN
bracis-28433	114	13	[	[	X
bracis-28433	114	14	11	11	NUM
bracis-28433	114	15	]	]	PUNCT
bracis-28433	114	16	activation	activation	NOUN
bracis-28433	114	17	,	,	PUNCT
bracis-28433	114	18	followed	follow	VERB
bracis-28433	114	19	by	by	ADP
bracis-28433	114	20	a	a	DET
bracis-28433	114	21	softmax	softmax	NOUN
bracis-28433	114	22	function	function	NOUN
bracis-28433	114	23	to	to	PART
bracis-28433	114	24	generate	generate	VERB
bracis-28433	114	25	a	a	DET
bracis-28433	114	26	distribution	distribution	NOUN
bracis-28433	114	27	over	over	ADP
bracis-28433	114	28	predicted	predict	VERB
bracis-28433	114	29	classes	class	NOUN
bracis-28433	114	30	.	.	PUNCT
bracis-28433	115	1	the	the	DET
bracis-28433	115	2	cvt	cvt	PROPN
bracis-28433	115	3	loss	loss	NOUN
bracis-28433	115	4	is	be	AUX
bracis-28433	115	5	the	the	DET
bracis-28433	115	6	kullback	kullback	NOUN
bracis-28433	115	7	-	-	PUNCT
bracis-28433	115	8	leibler	leibler	NOUN
bracis-28433	115	9	(	(	PUNCT
bracis-28433	115	10	kl	kl	NOUN
bracis-28433	115	11	)	)	PUNCT
bracis-28433	115	12	divergence	divergence	NOUN
bracis-28433	115	13	between	between	ADP
bracis-28433	115	14	the	the	DET
bracis-28433	115	15	outputs	output	NOUN
bracis-28433	115	16	of	of	ADP
bracis-28433	115	17	the	the	DET
bracis-28433	115	18	main	main	ADJ
bracis-28433	115	19	classification	classification	NOUN
bracis-28433	115	20	head	head	NOUN
bracis-28433	115	21	,	,	PUNCT
bracis-28433	115	22	which	which	PRON
bracis-28433	115	23	sees	see	VERB
bracis-28433	115	24	the	the	DET
bracis-28433	115	25	complete	complete	ADJ
bracis-28433	115	26	input	input	NOUN
bracis-28433	115	27	,	,	PUNCT
bracis-28433	115	28	and	and	CCONJ
bracis-28433	115	29	the	the	DET
bracis-28433	115	30	auxiliary	auxiliary	ADJ
bracis-28433	115	31	heads	head	NOUN
bracis-28433	115	32	,	,	PUNCT
bracis-28433	115	33	which	which	PRON
bracis-28433	115	34	see	see	VERB
bracis-28433	115	35	partial	partial	ADJ
bracis-28433	115	36	views	view	NOUN
bracis-28433	115	37	of	of	ADP
bracis-28433	115	38	the	the	DET
bracis-28433	115	39	input	input	NOUN
bracis-28433	115	40	.	.	PUNCT
bracis-28433	116	1	the	the	DET
bracis-28433	116	2	word	word	NOUN
bracis-28433	116	3	dropout	dropout	NOUN
bracis-28433	116	4	(	(	PUNCT
bracis-28433	116	5	wd	wd	ADJ
bracis-28433	116	6	)	)	PUNCT
bracis-28433	116	7	technique	technique	NOUN
bracis-28433	116	8	[	[	X
bracis-28433	116	9	3	3	NUM
bracis-28433	116	10	,	,	PUNCT
bracis-28433	116	11	8	8	NUM
bracis-28433	116	12	]	]	PUNCT
bracis-28433	116	13	consists	consist	VERB
bracis-28433	116	14	in	in	ADP
bracis-28433	116	15	replacing	replace	VERB
bracis-28433	116	16	random	random	ADJ
bracis-28433	116	17	words	word	NOUN
bracis-28433	116	18	in	in	ADP
bracis-28433	116	19	a	a	DET
bracis-28433	116	20	sentence	sentence	NOUN
bracis-28433	116	21	with	with	ADP
bracis-28433	116	22	special	special	ADJ
bracis-28433	116	23	tokens	token	NOUN
bracis-28433	116	24	,	,	PUNCT
bracis-28433	116	25	such	such	ADJ
bracis-28433	116	26	as	as	ADP
bracis-28433	116	27	\(<removed>\	\(<removed>\	PROPN
bracis-28433	116	28	)	)	PUNCT
bracis-28433	116	29	or	or	CCONJ
bracis-28433	116	30	\(<unk>\	\(<unk>\	NOUN
bracis-28433	116	31	)	)	PUNCT
bracis-28433	116	32	,	,	PUNCT
bracis-28433	116	33	and	and	CCONJ
bracis-28433	116	34	training	train	VERB
bracis-28433	116	35	the	the	DET
bracis-28433	116	36	model	model	NOUN
bracis-28433	116	37	to	to	PART
bracis-28433	116	38	produce	produce	VERB
bracis-28433	116	39	a	a	DET
bracis-28433	116	40	similar	similar	ADJ
bracis-28433	116	41	output	output	NOUN
bracis-28433	116	42	distribution	distribution	NOUN
bracis-28433	116	43	as	as	SCONJ
bracis-28433	116	44	when	when	SCONJ
bracis-28433	116	45	the	the	DET
bracis-28433	116	46	word	word	NOUN
bracis-28433	116	47	was	be	AUX
bracis-28433	116	48	unmasked	unmasked	ADJ
bracis-28433	116	49	.	.	PUNCT
bracis-28433	117	1	the	the	DET
bracis-28433	117	2	virtual	virtual	ADJ
bracis-28433	117	3	adversarial	adversarial	ADJ
bracis-28433	117	4	technique	technique	NOUN
bracis-28433	117	5	(	(	PUNCT
bracis-28433	117	6	vat	vat	NOUN
bracis-28433	117	7	)	)	PUNCT
bracis-28433	117	8	is	be	AUX
bracis-28433	117	9	a	a	DET
bracis-28433	117	10	self	self	NOUN
bracis-28433	117	11	-	-	PUNCT
bracis-28433	117	12	learning	learn	VERB
bracis-28433	117	13	technique	technique	NOUN
bracis-28433	117	14	proposed	propose	VERB
bracis-28433	117	15	by	by	ADP
bracis-28433	117	16	miyato	miyato	PROPN
bracis-28433	117	17	et	et	PROPN
bracis-28433	117	18	al	al	PROPN
bracis-28433	117	19	.	.	PUNCT
bracis-28433	118	1	[	[	X
bracis-28433	118	2	10	10	NUM
bracis-28433	118	3	]	]	PUNCT
bracis-28433	118	4	for	for	ADP
bracis-28433	118	5	text	text	NOUN
bracis-28433	118	6	classification	classification	NOUN
bracis-28433	118	7	and	and	CCONJ
bracis-28433	118	8	applied	apply	VERB
bracis-28433	118	9	to	to	ADP
bracis-28433	118	10	ner	ner	NOUN
bracis-28433	118	11	by	by	ADP
bracis-28433	118	12	clark	clark	PROPN
bracis-28433	118	13	et	et	PROPN
bracis-28433	118	14	al	al	PROPN
bracis-28433	118	15	.	.	PUNCT
bracis-28433	119	1	[	[	X
bracis-28433	119	2	3	3	NUM
bracis-28433	119	3	]	]	PUNCT
bracis-28433	119	4	.	.	PUNCT
bracis-28433	120	1	this	this	DET
bracis-28433	120	2	technique	technique	NOUN
bracis-28433	120	3	extends	extend	VERB
bracis-28433	120	4	the	the	DET
bracis-28433	120	5	adversarial	adversarial	ADJ
bracis-28433	120	6	training	training	NOUN
bracis-28433	120	7	[	[	X
bracis-28433	120	8	4	4	NUM
bracis-28433	120	9	]	]	PUNCT
bracis-28433	120	10	,	,	PUNCT
bracis-28433	120	11	where	where	SCONJ
bracis-28433	120	12	a	a	DET
bracis-28433	120	13	sample	sample	NOUN
bracis-28433	120	14	of	of	ADP
bracis-28433	120	15	input	input	NOUN
bracis-28433	120	16	data	datum	NOUN
bracis-28433	120	17	is	be	AUX
bracis-28433	120	18	perturbed	perturb	VERB
bracis-28433	120	19	with	with	ADP
bracis-28433	120	20	specially	specially	ADV
bracis-28433	120	21	crafted	craft	VERB
bracis-28433	120	22	noise	noise	NOUN
bracis-28433	120	23	designed	design	VERB
bracis-28433	120	24	to	to	PART
bracis-28433	120	25	fool	fool	VERB
bracis-28433	120	26	the	the	DET
bracis-28433	120	27	model	model	NOUN
bracis-28433	120	28	.	.	PUNCT
bracis-28433	121	1	this	this	PRON
bracis-28433	121	2	helps	help	VERB
bracis-28433	121	3	the	the	DET
bracis-28433	121	4	model	model	NOUN
bracis-28433	121	5	to	to	PART
bracis-28433	121	6	be	be	AUX
bracis-28433	121	7	more	more	ADV
bracis-28433	121	8	robust	robust	ADJ
bracis-28433	121	9	to	to	ADP
bracis-28433	121	10	small	small	ADJ
bracis-28433	121	11	perturbations	perturbation	NOUN
bracis-28433	121	12	in	in	ADP
bracis-28433	121	13	the	the	DET
bracis-28433	121	14	input	input	NOUN
bracis-28433	121	15	data	datum	NOUN
bracis-28433	121	16	.	.	PUNCT
bracis-28433	122	1	5	5	NUM
bracis-28433	122	2	token	token	VERB
bracis-28433	122	3	-	-	PUNCT
bracis-28433	122	4	level	level	NOUN
bracis-28433	122	5	active	active	ADJ
bracis-28433	122	6	-	-	PUNCT
bracis-28433	122	7	self	self	NOUN
bracis-28433	122	8	learning	learn	VERB
bracis-28433	122	9	algorithm	algorithm	NOUN
bracis-28433	122	10	the	the	DET
bracis-28433	122	11	model	model	NOUN
bracis-28433	122	12	’s	’s	PART
bracis-28433	122	13	overall	overall	ADJ
bracis-28433	122	14	confidence	confidence	NOUN
bracis-28433	122	15	for	for	ADP
bracis-28433	122	16	the	the	DET
bracis-28433	122	17	entire	entire	ADJ
bracis-28433	122	18	sentence	sentence	NOUN
bracis-28433	122	19	is	be	AUX
bracis-28433	122	20	considered	consider	VERB
bracis-28433	122	21	when	when	SCONJ
bracis-28433	122	22	using	use	VERB
bracis-28433	122	23	sentence	sentence	NOUN
bracis-28433	122	24	-	-	PUNCT
bracis-28433	122	25	level	level	NOUN
bracis-28433	122	26	querying	querying	NOUN
bracis-28433	122	27	.	.	PUNCT
bracis-28433	123	1	this	this	PRON
bracis-28433	123	2	implies	imply	VERB
bracis-28433	123	3	that	that	SCONJ
bracis-28433	123	4	the	the	DET
bracis-28433	123	5	model	model	NOUN
bracis-28433	123	6	assumes	assume	VERB
bracis-28433	123	7	average	average	ADJ
bracis-28433	123	8	confidence	confidence	NOUN
bracis-28433	123	9	in	in	ADP
bracis-28433	123	10	its	its	PRON
bracis-28433	123	11	predictions	prediction	NOUN
bracis-28433	123	12	for	for	ADP
bracis-28433	123	13	all	all	DET
bracis-28433	123	14	the	the	DET
bracis-28433	123	15	words	word	NOUN
bracis-28433	123	16	/	/	SYM
bracis-28433	123	17	tokens	token	NOUN
bracis-28433	123	18	within	within	ADP
bracis-28433	123	19	the	the	DET
bracis-28433	123	20	sentence	sentence	NOUN
bracis-28433	123	21	.	.	PUNCT
bracis-28433	124	1	however	however	ADV
bracis-28433	124	2	,	,	PUNCT
bracis-28433	124	3	hard	hard	ADJ
bracis-28433	124	4	-	-	PUNCT
bracis-28433	124	5	to	to	PART
bracis-28433	124	6	-	-	PUNCT
bracis-28433	124	7	predict	predict	VERB
bracis-28433	124	8	entities	entity	NOUN
bracis-28433	124	9	may	may	AUX
bracis-28433	124	10	be	be	AUX
bracis-28433	124	11	surrounded	surround	VERB
bracis-28433	124	12	by	by	ADP
bracis-28433	124	13	various	various	ADJ
bracis-28433	124	14	easy	easy	ADJ
bracis-28433	124	15	-	-	PUNCT
bracis-28433	124	16	to	to	PART
bracis-28433	124	17	-	-	PUNCT
bracis-28433	124	18	predict	predict	VERB
bracis-28433	124	19	tokens	token	NOUN
bracis-28433	124	20	.	.	PUNCT
bracis-28433	125	1	we	we	PRON
bracis-28433	125	2	argue	argue	VERB
bracis-28433	125	3	this	this	PRON
bracis-28433	125	4	may	may	AUX
bracis-28433	125	5	lead	lead	VERB
bracis-28433	125	6	to	to	ADP
bracis-28433	125	7	an	an	DET
bracis-28433	125	8	overall	overall	ADJ
bracis-28433	125	9	overconfidence	overconfidence	NOUN
bracis-28433	125	10	in	in	ADP
bracis-28433	125	11	the	the	DET
bracis-28433	125	12	model	model	NOUN
bracis-28433	125	13	for	for	ADP
bracis-28433	125	14	complex	complex	ADJ
bracis-28433	125	15	tokens	token	NOUN
bracis-28433	125	16	,	,	PUNCT
bracis-28433	125	17	leading	lead	VERB
bracis-28433	125	18	to	to	ADP
bracis-28433	125	19	poor	poor	ADJ
bracis-28433	125	20	self	self	NOUN
bracis-28433	125	21	-	-	PUNCT
bracis-28433	125	22	annotations	annotation	NOUN
bracis-28433	125	23	that	that	PRON
bracis-28433	125	24	may	may	AUX
bracis-28433	125	25	hamper	hamper	VERB
bracis-28433	125	26	the	the	DET
bracis-28433	125	27	active	active	ADJ
bracis-28433	125	28	self	self	NOUN
bracis-28433	125	29	-	-	PUNCT
bracis-28433	125	30	learning	learn	VERB
bracis-28433	125	31	process	process	NOUN
bracis-28433	125	32	.	.	PUNCT
bracis-28433	126	1	we	we	PRON
bracis-28433	126	2	leverage	leverage	VERB
bracis-28433	126	3	this	this	DET
bracis-28433	126	4	idea	idea	NOUN
bracis-28433	126	5	to	to	PART
bracis-28433	126	6	propose	propose	VERB
bracis-28433	126	7	a	a	DET
bracis-28433	126	8	token	token	ADJ
bracis-28433	126	9	-	-	PUNCT
bracis-28433	126	10	level	level	NOUN
bracis-28433	126	11	active	active	ADJ
bracis-28433	126	12	self	self	NOUN
bracis-28433	126	13	-	-	PUNCT
bracis-28433	126	14	learning	learn	VERB
bracis-28433	126	15	algorithm	algorithm	NOUN
bracis-28433	126	16	.	.	PUNCT
bracis-28433	127	1	the	the	DET
bracis-28433	127	2	token	token	VERB
bracis-28433	127	3	-	-	PUNCT
bracis-28433	127	4	level	level	NOUN
bracis-28433	127	5	active	active	ADJ
bracis-28433	127	6	self	self	NOUN
bracis-28433	127	7	-	-	PUNCT
bracis-28433	127	8	learning	learn	VERB
bracis-28433	127	9	algorithm	algorithm	NOUN
bracis-28433	127	10	proposed	propose	VERB
bracis-28433	127	11	is	be	AUX
bracis-28433	127	12	a	a	DET
bracis-28433	127	13	modified	modify	VERB
bracis-28433	127	14	version	version	NOUN
bracis-28433	127	15	of	of	ADP
bracis-28433	127	16	the	the	DET
bracis-28433	127	17	deep	deep	ADJ
bracis-28433	127	18	active	active	ADJ
bracis-28433	127	19	learning	learning	NOUN
bracis-28433	127	20	(	(	PUNCT
bracis-28433	127	21	dal	dal	NOUN
bracis-28433	127	22	)	)	PUNCT
bracis-28433	127	23	algorithm	algorithm	NOUN
bracis-28433	127	24	presented	present	VERB
bracis-28433	127	25	by	by	ADP
bracis-28433	127	26	shen	shen	PROPN
bracis-28433	127	27	et	et	PROPN
bracis-28433	127	28	al	al	PROPN
bracis-28433	127	29	.	.	PUNCT
bracis-28433	128	1	[	[	X
bracis-28433	128	2	18	18	NUM
bracis-28433	128	3	]	]	PUNCT
bracis-28433	128	4	.	.	PUNCT
bracis-28433	129	1	the	the	DET
bracis-28433	129	2	main	main	ADJ
bracis-28433	129	3	difference	difference	NOUN
bracis-28433	129	4	between	between	ADP
bracis-28433	129	5	the	the	DET
bracis-28433	129	6	original	original	ADJ
bracis-28433	129	7	dal	dal	NOUN
bracis-28433	129	8	and	and	CCONJ
bracis-28433	129	9	our	our	PRON
bracis-28433	129	10	proposed	propose	VERB
bracis-28433	129	11	algorithm	algorithm	NOUN
bracis-28433	129	12	is	be	AUX
bracis-28433	129	13	its	its	PRON
bracis-28433	129	14	labeling	labeling	NOUN
bracis-28433	129	15	process	process	NOUN
bracis-28433	129	16	.	.	PUNCT
bracis-28433	130	1	the	the	DET
bracis-28433	130	2	original	original	ADJ
bracis-28433	130	3	algorithm	algorithm	NOUN
bracis-28433	130	4	queried	query	VERB
bracis-28433	130	5	the	the	DET
bracis-28433	130	6	most	most	ADV
bracis-28433	130	7	uncertain	uncertain	ADJ
bracis-28433	130	8	sentences	sentence	NOUN
bracis-28433	130	9	and	and	CCONJ
bracis-28433	130	10	asked	ask	VERB
bracis-28433	130	11	the	the	DET
bracis-28433	130	12	oracle	oracle	NOUN
bracis-28433	130	13	to	to	PART
bracis-28433	130	14	annotate	annotate	VERB
bracis-28433	130	15	them	they	PRON
bracis-28433	130	16	.	.	PUNCT
bracis-28433	131	1	our	our	PRON
bracis-28433	131	2	proposed	propose	VERB
bracis-28433	131	3	algorithm	algorithm	NOUN
bracis-28433	131	4	takes	take	VERB
bracis-28433	131	5	the	the	DET
bracis-28433	131	6	queried	query	VERB
bracis-28433	131	7	sentences	sentence	NOUN
bracis-28433	131	8	,	,	PUNCT
bracis-28433	131	9	identifies	identify	VERB
bracis-28433	131	10	the	the	DET
bracis-28433	131	11	tokens	token	NOUN
bracis-28433	131	12	with	with	ADP
bracis-28433	131	13	low	low	ADJ
bracis-28433	131	14	confidence	confidence	NOUN
bracis-28433	131	15	,	,	PUNCT
bracis-28433	131	16	and	and	CCONJ
bracis-28433	131	17	asks	ask	VERB
bracis-28433	131	18	the	the	DET
bracis-28433	131	19	oracle	oracle	NOUN
bracis-28433	131	20	to	to	PART
bracis-28433	131	21	annotate	annotate	VERB
bracis-28433	131	22	them	they	PRON
bracis-28433	131	23	.	.	PUNCT
bracis-28433	132	1	the	the	DET
bracis-28433	132	2	remaining	remain	VERB
bracis-28433	132	3	unlabeled	unlabeled	ADJ
bracis-28433	132	4	tokens	token	NOUN
bracis-28433	132	5	from	from	ADP
bracis-28433	132	6	the	the	DET
bracis-28433	132	7	queried	query	VERB
bracis-28433	132	8	sentence	sentence	NOUN
bracis-28433	132	9	are	be	AUX
bracis-28433	132	10	automatically	automatically	ADV
bracis-28433	132	11	labeled	label	VERB
bracis-28433	132	12	using	use	VERB
bracis-28433	132	13	the	the	DET
bracis-28433	132	14	model	model	NOUN
bracis-28433	132	15	’s	’s	PART
bracis-28433	132	16	output	output	NOUN
bracis-28433	132	17	predictions	prediction	NOUN
bracis-28433	132	18	.	.	PUNCT
bracis-28433	133	1	we	we	PRON
bracis-28433	133	2	also	also	ADV
bracis-28433	133	3	improve	improve	VERB
bracis-28433	133	4	the	the	DET
bracis-28433	133	5	accuracy	accuracy	NOUN
bracis-28433	133	6	of	of	ADP
bracis-28433	133	7	our	our	PRON
bracis-28433	133	8	self	self	NOUN
bracis-28433	133	9	-	-	PUNCT
bracis-28433	133	10	labeling	labeling	NOUN
bracis-28433	133	11	process	process	NOUN
bracis-28433	133	12	by	by	ADP
bracis-28433	133	13	using	use	VERB
bracis-28433	133	14	hand	hand	NOUN
bracis-28433	133	15	-	-	PUNCT
bracis-28433	133	16	annotated	annotate	VERB
bracis-28433	133	17	labels	label	NOUN
bracis-28433	133	18	to	to	PART
bracis-28433	133	19	refine	refine	VERB
bracis-28433	133	20	the	the	DET
bracis-28433	133	21	predictions	prediction	NOUN
bracis-28433	133	22	made	make	VERB
bracis-28433	133	23	by	by	ADP
bracis-28433	133	24	the	the	DET
bracis-28433	133	25	model	model	NOUN
bracis-28433	133	26	.	.	PUNCT
bracis-28433	134	1	this	this	DET
bracis-28433	134	2	cooperative	cooperative	ADJ
bracis-28433	134	3	scenario	scenario	NOUN
bracis-28433	134	4	can	can	AUX
bracis-28433	134	5	alleviate	alleviate	VERB
bracis-28433	134	6	the	the	DET
bracis-28433	134	7	cost	cost	NOUN
bracis-28433	134	8	of	of	ADP
bracis-28433	134	9	manual	manual	ADJ
bracis-28433	134	10	annotation	annotation	NOUN
bracis-28433	134	11	for	for	ADP
bracis-28433	134	12	the	the	DET
bracis-28433	134	13	oracle	oracle	NOUN
bracis-28433	134	14	and	and	CCONJ
bracis-28433	134	15	speed	speed	VERB
bracis-28433	134	16	up	up	ADP
bracis-28433	134	17	the	the	DET
bracis-28433	134	18	annotation	annotation	NOUN
bracis-28433	134	19	process	process	NOUN
bracis-28433	134	20	by	by	ADP
bracis-28433	134	21	highlighting	highlight	VERB
bracis-28433	134	22	specific	specific	ADJ
bracis-28433	134	23	words	word	NOUN
bracis-28433	134	24	in	in	ADP
bracis-28433	134	25	a	a	DET
bracis-28433	134	26	sentence	sentence	NOUN
bracis-28433	134	27	that	that	PRON
bracis-28433	134	28	must	must	AUX
bracis-28433	134	29	be	be	AUX
bracis-28433	134	30	hand	hand	NOUN
bracis-28433	134	31	-	-	PUNCT
bracis-28433	134	32	annotated	annotate	VERB
bracis-28433	134	33	.	.	PUNCT
bracis-28433	135	1	our	our	PRON
bracis-28433	135	2	proposed	propose	VERB
bracis-28433	135	3	token	token	VERB
bracis-28433	135	4	-	-	PUNCT
bracis-28433	135	5	level	level	NOUN
bracis-28433	135	6	active	active	ADJ
bracis-28433	135	7	self	self	NOUN
bracis-28433	135	8	-	-	PUNCT
bracis-28433	135	9	learning	learn	VERB
bracis-28433	135	10	algorithm	algorithm	NOUN
bracis-28433	135	11	is	be	AUX
bracis-28433	135	12	illustrated	illustrate	VERB
bracis-28433	135	13	in	in	ADP
bracis-28433	135	14	fig	fig	NOUN
bracis-28433	135	15	.	.	PUNCT
bracis-28433	135	16	 	 	SPACE
bracis-28433	136	1	2	2	NUM
bracis-28433	136	2	,	,	PUNCT
bracis-28433	136	3	which	which	PRON
bracis-28433	136	4	shows	show	VERB
bracis-28433	136	5	the	the	DET
bracis-28433	136	6	annotation	annotation	NOUN
bracis-28433	136	7	process	process	NOUN
bracis-28433	136	8	.	.	PUNCT
bracis-28433	137	1	fig	fig	NOUN
bracis-28433	137	2	.	.	PUNCT
bracis-28433	138	1	2	2	X
bracis-28433	138	2	.	.	X
bracis-28433	138	3	an	an	DET
bracis-28433	138	4	illustration	illustration	NOUN
bracis-28433	138	5	of	of	ADP
bracis-28433	138	6	the	the	DET
bracis-28433	138	7	collaborative	collaborative	ADJ
bracis-28433	138	8	configuration	configuration	NOUN
bracis-28433	138	9	where	where	SCONJ
bracis-28433	138	10	an	an	DET
bracis-28433	138	11	oracle	oracle	NOUN
bracis-28433	138	12	and	and	CCONJ
bracis-28433	138	13	machine	machine	NOUN
bracis-28433	138	14	learning	learning	NOUN
bracis-28433	138	15	model	model	NOUN
bracis-28433	138	16	annotate	annotate	VERB
bracis-28433	138	17	the	the	DET
bracis-28433	138	18	same	same	ADJ
bracis-28433	138	19	sentence	sentence	NOUN
bracis-28433	138	20	jointly	jointly	ADV
bracis-28433	138	21	.	.	PUNCT
bracis-28433	139	1	full	full	ADJ
bracis-28433	139	2	size	size	NOUN
bracis-28433	139	3	image	image	NOUN
bracis-28433	139	4	our	our	PRON
bracis-28433	139	5	proposed	propose	VERB
bracis-28433	139	6	algorithm	algorithm	NOUN
bracis-28433	139	7	has	have	VERB
bracis-28433	139	8	three	three	NUM
bracis-28433	139	9	main	main	ADJ
bracis-28433	139	10	procedures	procedure	NOUN
bracis-28433	139	11	:	:	PUNCT
bracis-28433	139	12	1	1	X
bracis-28433	139	13	)	)	PUNCT
bracis-28433	139	14	predict	predict	VERB
bracis-28433	139	15	the	the	DET
bracis-28433	139	16	highest	high	ADJ
bracis-28433	139	17	confidence	confidence	NOUN
bracis-28433	139	18	tokens	token	NOUN
bracis-28433	139	19	to	to	ADP
bracis-28433	139	20	automatic	automatic	ADJ
bracis-28433	139	21	annotation	annotation	NOUN
bracis-28433	139	22	;	;	PUNCT
bracis-28433	139	23	2	2	X
bracis-28433	139	24	)	)	PUNCT
bracis-28433	139	25	query	query	NOUN
bracis-28433	139	26	low	low	ADJ
bracis-28433	139	27	-	-	PUNCT
bracis-28433	139	28	confidence	confidence	NOUN
bracis-28433	139	29	tokens	token	NOUN
bracis-28433	139	30	to	to	ADP
bracis-28433	139	31	the	the	DET
bracis-28433	139	32	oracle	oracle	NOUN
bracis-28433	139	33	for	for	ADP
bracis-28433	139	34	manual	manual	ADJ
bracis-28433	139	35	annotation	annotation	NOUN
bracis-28433	139	36	and	and	CCONJ
bracis-28433	139	37	3	3	X
bracis-28433	139	38	)	)	PUNCT
bracis-28433	139	39	self	self	NOUN
bracis-28433	139	40	-	-	PUNCT
bracis-28433	139	41	labeling	labeling	NOUN
bracis-28433	139	42	refinement	refinement	NOUN
bracis-28433	139	43	.	.	PUNCT
bracis-28433	140	1	1	1	X
bracis-28433	140	2	.	.	X
bracis-28433	140	3	identifying	identify	VERB
bracis-28433	140	4	low	low	ADJ
bracis-28433	140	5	-	-	PUNCT
bracis-28433	140	6	confidence	confidence	NOUN
bracis-28433	140	7	tokens	token	NOUN
bracis-28433	140	8	:	:	PUNCT
bracis-28433	140	9	suppose	suppose	VERB
bracis-28433	140	10	a	a	DET
bracis-28433	140	11	neural	neural	ADJ
bracis-28433	140	12	network	network	NOUN
bracis-28433	140	13	produces	produce	VERB
bracis-28433	140	14	a	a	DET
bracis-28433	140	15	probability	probability	NOUN
bracis-28433	140	16	distribution	distribution	NOUN
bracis-28433	140	17	for	for	ADP
bracis-28433	140	18	an	an	DET
bracis-28433	140	19	input	input	NOUN
bracis-28433	140	20	token	token	VERB
bracis-28433	140	21	,	,	PUNCT
bracis-28433	140	22	which	which	PRON
bracis-28433	140	23	represents	represent	VERB
bracis-28433	140	24	the	the	DET
bracis-28433	140	25	likelihood	likelihood	NOUN
bracis-28433	140	26	of	of	ADP
bracis-28433	140	27	the	the	DET
bracis-28433	140	28	token	token	NOUN
bracis-28433	140	29	belonging	belong	VERB
bracis-28433	140	30	to	to	ADP
bracis-28433	140	31	one	one	NUM
bracis-28433	140	32	of	of	ADP
bracis-28433	140	33	the	the	DET
bracis-28433	140	34	classes	class	NOUN
bracis-28433	140	35	of	of	ADP
bracis-28433	140	36	named	name	VERB
bracis-28433	140	37	entities	entity	NOUN
bracis-28433	140	38	.	.	PUNCT
bracis-28433	141	1	in	in	ADP
bracis-28433	141	2	this	this	DET
bracis-28433	141	3	context	context	NOUN
bracis-28433	141	4	,	,	PUNCT
bracis-28433	141	5	we	we	PRON
bracis-28433	141	6	define	define	VERB
bracis-28433	141	7	low	low	ADJ
bracis-28433	141	8	-	-	PUNCT
bracis-28433	141	9	confidence	confidence	NOUN
bracis-28433	141	10	tokens	token	NOUN
bracis-28433	141	11	as	as	ADP
bracis-28433	141	12	those	those	PRON
bracis-28433	141	13	for	for	ADP
bracis-28433	141	14	which	which	PRON
bracis-28433	141	15	the	the	DET
bracis-28433	141	16	model	model	NOUN
bracis-28433	141	17	has	have	VERB
bracis-28433	141	18	confidence	confidence	NOUN
bracis-28433	141	19	in	in	ADP
bracis-28433	141	20	its	its	PRON
bracis-28433	141	21	prediction	prediction	NOUN
bracis-28433	141	22	lower	low	ADJ
bracis-28433	141	23	than	than	ADP
bracis-28433	141	24	a	a	DET
bracis-28433	141	25	predefined	predefine	VERB
bracis-28433	141	26	threshold	threshold	NOUN
bracis-28433	141	27	.	.	PUNCT
bracis-28433	142	1	we	we	PRON
bracis-28433	142	2	empirically	empirically	ADV
bracis-28433	142	3	selected	select	VERB
bracis-28433	142	4	a	a	DET
bracis-28433	142	5	threshold	threshold	NOUN
bracis-28433	142	6	confidence	confidence	NOUN
bracis-28433	142	7	of	of	ADP
bracis-28433	142	8	99	99	NUM
bracis-28433	142	9	%	%	NOUN
bracis-28433	142	10	,	,	PUNCT
bracis-28433	142	11	implying	imply	VERB
bracis-28433	142	12	that	that	SCONJ
bracis-28433	142	13	tokens	token	NOUN
bracis-28433	142	14	with	with	ADP
bracis-28433	142	15	less	less	ADJ
bracis-28433	142	16	than	than	ADP
bracis-28433	142	17	99	99	NUM
bracis-28433	142	18	%	%	NOUN
bracis-28433	142	19	of	of	ADP
bracis-28433	142	20	confidence	confidence	NOUN
bracis-28433	142	21	in	in	ADP
bracis-28433	142	22	their	their	PRON
bracis-28433	142	23	predicted	predict	VERB
bracis-28433	142	24	class	class	NOUN
bracis-28433	142	25	will	will	AUX
bracis-28433	142	26	be	be	AUX
bracis-28433	142	27	labeled	label	VERB
bracis-28433	142	28	by	by	ADP
bracis-28433	142	29	the	the	DET
bracis-28433	142	30	oracle	oracle	NOUN
bracis-28433	142	31	(	(	PUNCT
bracis-28433	142	32	i.e.	i.e.	X
bracis-28433	142	33	,	,	PUNCT
bracis-28433	142	34	human	human	ADJ
bracis-28433	142	35	annotator	annotator	NOUN
bracis-28433	142	36	)	)	PUNCT
bracis-28433	142	37	.	.	PUNCT
bracis-28433	143	1	in	in	ADP
bracis-28433	143	2	contrast	contrast	NOUN
bracis-28433	143	3	,	,	PUNCT
bracis-28433	143	4	the	the	DET
bracis-28433	143	5	remaining	remain	VERB
bracis-28433	143	6	tokens	token	NOUN
bracis-28433	143	7	will	will	AUX
bracis-28433	143	8	be	be	AUX
bracis-28433	143	9	self	self	NOUN
bracis-28433	143	10	-	-	PUNCT
bracis-28433	143	11	labeled	label	VERB
bracis-28433	143	12	by	by	ADP
bracis-28433	143	13	the	the	DET
bracis-28433	143	14	model	model	NOUN
bracis-28433	143	15	.	.	PUNCT
bracis-28433	144	1	2	2	X
bracis-28433	144	2	.	.	NUM
bracis-28433	144	3	query	query	NOUN
bracis-28433	144	4	to	to	ADP
bracis-28433	144	5	the	the	DET
bracis-28433	144	6	oracle	oracle	NOUN
bracis-28433	144	7	:	:	PUNCT
bracis-28433	144	8	it	it	PRON
bracis-28433	144	9	is	be	AUX
bracis-28433	144	10	the	the	DET
bracis-28433	144	11	traditional	traditional	ADJ
bracis-28433	144	12	active	active	ADJ
bracis-28433	144	13	learning	learning	NOUN
bracis-28433	144	14	technique	technique	NOUN
bracis-28433	144	15	implemented	implement	VERB
bracis-28433	144	16	at	at	ADP
bracis-28433	144	17	a	a	DET
bracis-28433	144	18	token	token	ADJ
bracis-28433	144	19	level	level	NOUN
bracis-28433	144	20	,	,	PUNCT
bracis-28433	144	21	where	where	SCONJ
bracis-28433	144	22	only	only	ADV
bracis-28433	144	23	the	the	DET
bracis-28433	144	24	low	low	ADJ
bracis-28433	144	25	-	-	PUNCT
bracis-28433	144	26	confidence	confidence	NOUN
bracis-28433	144	27	tokens	token	NOUN
bracis-28433	144	28	in	in	ADP
bracis-28433	144	29	the	the	DET
bracis-28433	144	30	selected	select	VERB
bracis-28433	144	31	sentences	sentence	NOUN
bracis-28433	144	32	are	be	AUX
bracis-28433	144	33	queried	query	VERB
bracis-28433	144	34	to	to	ADP
bracis-28433	144	35	the	the	DET
bracis-28433	144	36	oracle	oracle	NOUN
bracis-28433	144	37	.	.	PUNCT
bracis-28433	145	1	3	3	X
bracis-28433	145	2	.	.	X
bracis-28433	145	3	self	self	NOUN
bracis-28433	145	4	-	-	PUNCT
bracis-28433	145	5	labeling	labeling	NOUN
bracis-28433	145	6	refinement	refinement	NOUN
bracis-28433	145	7	:	:	PUNCT
bracis-28433	145	8	once	once	SCONJ
bracis-28433	145	9	an	an	DET
bracis-28433	145	10	oracle	oracle	NOUN
bracis-28433	145	11	has	have	AUX
bracis-28433	145	12	labeled	label	VERB
bracis-28433	145	13	the	the	DET
bracis-28433	145	14	low	low	ADJ
bracis-28433	145	15	-	-	PUNCT
bracis-28433	145	16	confidence	confidence	NOUN
bracis-28433	145	17	tokens	token	NOUN
bracis-28433	145	18	,	,	PUNCT
bracis-28433	145	19	we	we	PRON
bracis-28433	145	20	can	can	AUX
bracis-28433	145	21	use	use	VERB
bracis-28433	145	22	self	self	NOUN
bracis-28433	145	23	-	-	PUNCT
bracis-28433	145	24	labeling	labeling	NOUN
bracis-28433	145	25	to	to	PART
bracis-28433	145	26	label	label	VERB
bracis-28433	145	27	the	the	DET
bracis-28433	145	28	remaining	remain	VERB
bracis-28433	145	29	unlabeled	unlabeled	ADJ
bracis-28433	145	30	tokens	token	NOUN
bracis-28433	145	31	from	from	ADP
bracis-28433	145	32	the	the	DET
bracis-28433	145	33	queried	query	VERB
bracis-28433	145	34	samples	sample	NOUN
bracis-28433	145	35	.	.	PUNCT
bracis-28433	146	1	a	a	DET
bracis-28433	146	2	simple	simple	ADJ
bracis-28433	146	3	approach	approach	NOUN
bracis-28433	146	4	is	be	AUX
bracis-28433	146	5	to	to	PART
bracis-28433	146	6	predict	predict	VERB
bracis-28433	146	7	the	the	DET
bracis-28433	146	8	classes	class	NOUN
bracis-28433	146	9	for	for	ADP
bracis-28433	146	10	all	all	DET
bracis-28433	146	11	tokens	token	NOUN
bracis-28433	146	12	and	and	CCONJ
bracis-28433	146	13	use	use	VERB
bracis-28433	146	14	the	the	DET
bracis-28433	146	15	predictions	prediction	NOUN
bracis-28433	146	16	to	to	PART
bracis-28433	146	17	label	label	VERB
bracis-28433	146	18	high	high	ADJ
bracis-28433	146	19	-	-	PUNCT
bracis-28433	146	20	confidence	confidence	NOUN
bracis-28433	146	21	tokens	token	NOUN
bracis-28433	146	22	.	.	PUNCT
bracis-28433	147	1	however	however	ADV
bracis-28433	147	2	,	,	PUNCT
bracis-28433	147	3	we	we	PRON
bracis-28433	147	4	can	can	AUX
bracis-28433	147	5	leverage	leverage	VERB
bracis-28433	147	6	these	these	DET
bracis-28433	147	7	labels	label	NOUN
bracis-28433	147	8	to	to	PART
bracis-28433	147	9	improve	improve	VERB
bracis-28433	147	10	our	our	PRON
bracis-28433	147	11	predictions	prediction	NOUN
bracis-28433	147	12	since	since	SCONJ
bracis-28433	147	13	we	we	PRON
bracis-28433	147	14	have	have	VERB
bracis-28433	147	15	the	the	DET
bracis-28433	147	16	proper	proper	ADJ
bracis-28433	147	17	labels	label	NOUN
bracis-28433	147	18	of	of	ADP
bracis-28433	147	19	the	the	DET
bracis-28433	147	20	low	low	ADJ
bracis-28433	147	21	-	-	PUNCT
bracis-28433	147	22	confidence	confidence	NOUN
bracis-28433	147	23	tokens	token	NOUN
bracis-28433	147	24	,	,	PUNCT
bracis-28433	147	25	which	which	PRON
bracis-28433	147	26	were	be	AUX
bracis-28433	147	27	labeled	label	VERB
bracis-28433	147	28	by	by	ADP
bracis-28433	147	29	the	the	DET
bracis-28433	147	30	oracle	oracle	NOUN
bracis-28433	147	31	.	.	PUNCT
bracis-28433	148	1	for	for	ADP
bracis-28433	148	2	example	example	NOUN
bracis-28433	148	3	,	,	PUNCT
bracis-28433	148	4	the	the	DET
bracis-28433	148	5	cnn	cnn	PROPN
bracis-28433	148	6	-	-	PUNCT
bracis-28433	148	7	cnn	cnn	PROPN
bracis-28433	148	8	-	-	PUNCT
bracis-28433	148	9	lstm	lstm	ADJ
bracis-28433	148	10	model	model	NOUN
bracis-28433	148	11	[	[	X
bracis-28433	148	12	18	18	NUM
bracis-28433	148	13	]	]	PUNCT
bracis-28433	148	14	uses	use	VERB
bracis-28433	148	15	a	a	DET
bracis-28433	148	16	greedy	greedy	ADJ
bracis-28433	148	17	decoding	decode	VERB
bracis-28433	148	18	approach	approach	NOUN
bracis-28433	148	19	where	where	SCONJ
bracis-28433	148	20	it	it	PRON
bracis-28433	148	21	receives	receive	VERB
bracis-28433	148	22	the	the	DET
bracis-28433	148	23	predicted	predict	VERB
bracis-28433	148	24	label	label	NOUN
bracis-28433	148	25	of	of	ADP
bracis-28433	148	26	the	the	DET
bracis-28433	148	27	previous	previous	ADJ
bracis-28433	148	28	token	token	NOUN
bracis-28433	148	29	as	as	ADP
bracis-28433	148	30	an	an	DET
bracis-28433	148	31	additional	additional	ADJ
bracis-28433	148	32	input	input	NOUN
bracis-28433	148	33	to	to	PART
bracis-28433	148	34	help	help	VERB
bracis-28433	148	35	to	to	PART
bracis-28433	148	36	predict	predict	VERB
bracis-28433	148	37	the	the	DET
bracis-28433	148	38	current	current	ADJ
bracis-28433	148	39	token	token	PROPN
bracis-28433	148	40	’s	’s	PART
bracis-28433	148	41	class	class	NOUN
bracis-28433	148	42	.	.	PUNCT
bracis-28433	149	1	if	if	SCONJ
bracis-28433	149	2	the	the	DET
bracis-28433	149	3	previous	previous	ADJ
bracis-28433	149	4	token	token	NOUN
bracis-28433	149	5	had	have	VERB
bracis-28433	149	6	low	low	ADJ
bracis-28433	149	7	confidence	confidence	NOUN
bracis-28433	149	8	and	and	CCONJ
bracis-28433	149	9	was	be	AUX
bracis-28433	149	10	manually	manually	ADV
bracis-28433	149	11	labeled	label	VERB
bracis-28433	149	12	by	by	ADP
bracis-28433	149	13	the	the	DET
bracis-28433	149	14	oracle	oracle	NOUN
bracis-28433	149	15	,	,	PUNCT
bracis-28433	149	16	we	we	PRON
bracis-28433	149	17	could	could	AUX
bracis-28433	149	18	modify	modify	VERB
bracis-28433	149	19	this	this	DET
bracis-28433	149	20	approach	approach	NOUN
bracis-28433	149	21	by	by	ADP
bracis-28433	149	22	using	use	VERB
bracis-28433	149	23	the	the	DET
bracis-28433	149	24	oracle	oracle	NOUN
bracis-28433	149	25	-	-	PUNCT
bracis-28433	149	26	assigned	assign	VERB
bracis-28433	149	27	label	label	NOUN
bracis-28433	149	28	instead	instead	ADV
bracis-28433	149	29	of	of	ADP
bracis-28433	149	30	the	the	DET
bracis-28433	149	31	label	label	NOUN
bracis-28433	149	32	previously	previously	ADV
bracis-28433	149	33	predicted	predict	VERB
bracis-28433	149	34	by	by	ADP
bracis-28433	149	35	the	the	DET
bracis-28433	149	36	model	model	NOUN
bracis-28433	149	37	.	.	PUNCT
bracis-28433	150	1	this	this	DET
bracis-28433	150	2	process	process	NOUN
bracis-28433	150	3	can	can	AUX
bracis-28433	150	4	be	be	AUX
bracis-28433	150	5	repeated	repeat	VERB
bracis-28433	150	6	a	a	DET
bracis-28433	150	7	predefined	predefine	VERB
bracis-28433	150	8	number	number	NOUN
bracis-28433	150	9	of	of	ADP
bracis-28433	150	10	times	time	NOUN
bracis-28433	150	11	,	,	PUNCT
bracis-28433	150	12	iteratively	iteratively	ADV
bracis-28433	150	13	,	,	PUNCT
bracis-28433	150	14	with	with	ADP
bracis-28433	150	15	a	a	DET
bracis-28433	150	16	decreasing	decrease	VERB
bracis-28433	150	17	number	number	NOUN
bracis-28433	150	18	of	of	ADP
bracis-28433	150	19	tokens	token	NOUN
bracis-28433	150	20	replacement	replacement	NOUN
bracis-28433	150	21	in	in	ADP
bracis-28433	150	22	each	each	DET
bracis-28433	150	23	iteration	iteration	NOUN
bracis-28433	150	24	.	.	PUNCT
bracis-28433	151	1	the	the	DET
bracis-28433	151	2	refinement	refinement	NOUN
bracis-28433	151	3	step	step	NOUN
bracis-28433	151	4	was	be	AUX
bracis-28433	151	5	inspired	inspire	VERB
bracis-28433	151	6	by	by	ADP
bracis-28433	151	7	the	the	DET
bracis-28433	151	8	iterative	iterative	NOUN
bracis-28433	151	9	algorithm	algorithm	NOUN
bracis-28433	151	10	proposed	propose	VERB
bracis-28433	151	11	by	by	ADP
bracis-28433	151	12	park	park	NOUN
bracis-28433	151	13	et	et	PROPN
bracis-28433	151	14	al	al	PROPN
bracis-28433	151	15	.	.	PUNCT
bracis-28433	151	16	 	 	SPACE
bracis-28433	152	1	[	[	X
bracis-28433	152	2	13	13	NUM
bracis-28433	152	3	]	]	PUNCT
bracis-28433	152	4	.	.	PUNCT
bracis-28433	153	1	they	they	PRON
bracis-28433	153	2	use	use	VERB
bracis-28433	153	3	iterative	iterative	NOUN
bracis-28433	153	4	refinement	refinement	NOUN
bracis-28433	153	5	to	to	PART
bracis-28433	153	6	identify	identify	VERB
bracis-28433	153	7	synonyms	synonym	NOUN
bracis-28433	153	8	to	to	PART
bracis-28433	153	9	substitute	substitute	VERB
bracis-28433	153	10	specific	specific	ADJ
bracis-28433	153	11	tokens	token	NOUN
bracis-28433	153	12	from	from	ADP
bracis-28433	153	13	a	a	DET
bracis-28433	153	14	sentence	sentence	NOUN
bracis-28433	153	15	,	,	PUNCT
bracis-28433	153	16	with	with	ADP
bracis-28433	153	17	a	a	DET
bracis-28433	153	18	low	low	ADJ
bracis-28433	153	19	impact	impact	NOUN
bracis-28433	153	20	on	on	ADP
bracis-28433	153	21	its	its	PRON
bracis-28433	153	22	coherence	coherence	NOUN
bracis-28433	153	23	.	.	PUNCT
bracis-28433	154	1	the	the	DET
bracis-28433	154	2	idea	idea	NOUN
bracis-28433	154	3	is	be	AUX
bracis-28433	154	4	to	to	PART
bracis-28433	154	5	identify	identify	VERB
bracis-28433	154	6	potential	potential	ADJ
bracis-28433	154	7	synonyms	synonym	NOUN
bracis-28433	154	8	for	for	ADP
bracis-28433	154	9	specific	specific	ADJ
bracis-28433	154	10	tokens	token	NOUN
bracis-28433	154	11	,	,	PUNCT
bracis-28433	154	12	replace	replace	VERB
bracis-28433	154	13	them	they	PRON
bracis-28433	154	14	,	,	PUNCT
bracis-28433	154	15	and	and	CCONJ
bracis-28433	154	16	verify	verify	VERB
bracis-28433	154	17	if	if	SCONJ
bracis-28433	154	18	a	a	DET
bracis-28433	154	19	masked	mask	VERB
bracis-28433	154	20	language	language	NOUN
bracis-28433	154	21	model	model	NOUN
bracis-28433	154	22	predicts	predict	VERB
bracis-28433	154	23	the	the	DET
bracis-28433	154	24	synonyms	synonym	NOUN
bracis-28433	154	25	with	with	ADP
bracis-28433	154	26	high	high	ADJ
bracis-28433	154	27	confidence	confidence	NOUN
bracis-28433	154	28	.	.	PUNCT
bracis-28433	155	1	other	other	ADJ
bracis-28433	155	2	synonyms	synonym	NOUN
bracis-28433	155	3	replace	replace	VERB
bracis-28433	155	4	the	the	DET
bracis-28433	155	5	synonyms	synonym	NOUN
bracis-28433	155	6	with	with	ADP
bracis-28433	155	7	the	the	DET
bracis-28433	155	8	lowest	low	ADJ
bracis-28433	155	9	masked	mask	VERB
bracis-28433	155	10	language	language	NOUN
bracis-28433	155	11	model	model	NOUN
bracis-28433	155	12	scores	score	NOUN
bracis-28433	155	13	.	.	PUNCT
bracis-28433	156	1	for	for	ADP
bracis-28433	156	2	our	our	PRON
bracis-28433	156	3	refinement	refinement	NOUN
bracis-28433	156	4	step	step	NOUN
bracis-28433	156	5	,	,	PUNCT
bracis-28433	156	6	however	however	ADV
bracis-28433	156	7	,	,	PUNCT
bracis-28433	156	8	we	we	PRON
bracis-28433	156	9	use	use	VERB
bracis-28433	156	10	reliable	reliable	ADJ
bracis-28433	156	11	oracle	oracle	NOUN
bracis-28433	156	12	annotated	annotate	VERB
bracis-28433	156	13	tokens	token	NOUN
bracis-28433	156	14	to	to	PART
bracis-28433	156	15	enhance	enhance	VERB
bracis-28433	156	16	the	the	DET
bracis-28433	156	17	model	model	NOUN
bracis-28433	156	18	predictions	prediction	NOUN
bracis-28433	156	19	for	for	ADP
bracis-28433	156	20	the	the	DET
bracis-28433	156	21	unlabeled	unlabeled	ADJ
bracis-28433	156	22	tokens	token	NOUN
bracis-28433	156	23	.	.	PUNCT
bracis-28433	157	1	6	6	NUM
bracis-28433	157	2	experimental	experimental	ADJ
bracis-28433	157	3	design	design	NOUN
bracis-28433	157	4	we	we	PRON
bracis-28433	157	5	use	use	VERB
bracis-28433	157	6	two	two	NUM
bracis-28433	157	7	consolidated	consolidated	ADJ
bracis-28433	157	8	english	english	ADJ
bracis-28433	157	9	ner	ner	NOUN
bracis-28433	157	10	datasets	dataset	NOUN
bracis-28433	157	11	,	,	PUNCT
bracis-28433	157	12	namely	namely	ADV
bracis-28433	157	13	conll03	conll03	VERB
bracis-28433	158	1	[	[	X
bracis-28433	158	2	17	17	NUM
bracis-28433	158	3	]	]	PUNCT
bracis-28433	158	4	and	and	CCONJ
bracis-28433	158	5	ontonotes5.0	ontonotes5.0	X
bracis-28433	159	1	[	[	X
bracis-28433	159	2	15	15	NUM
bracis-28433	159	3	]	]	PUNCT
bracis-28433	159	4	,	,	PUNCT
bracis-28433	159	5	and	and	CCONJ
bracis-28433	159	6	one	one	NUM
bracis-28433	159	7	legal	legal	ADJ
bracis-28433	159	8	domain	domain	NOUN
bracis-28433	159	9	portuguese	portuguese	ADJ
bracis-28433	159	10	ner	ner	NOUN
bracis-28433	159	11	dataset	dataset	VERB
bracis-28433	159	12	,	,	PUNCT
bracis-28433	159	13	named	name	VERB
bracis-28433	159	14	aposentadoria	aposentadoria	NOUN
bracis-28433	159	15	[	[	X
bracis-28433	159	16	2	2	NUM
bracis-28433	159	17	]	]	PUNCT
bracis-28433	159	18	.	.	PUNCT
bracis-28433	160	1	aposentadoria	aposentadoria	PROPN
bracis-28433	160	2	is	be	AUX
bracis-28433	160	3	a	a	DET
bracis-28433	160	4	new	new	ADJ
bracis-28433	160	5	legal	legal	ADJ
bracis-28433	160	6	domain	domain	NOUN
bracis-28433	160	7	ner	ner	NOUN
bracis-28433	160	8	dataset	dataset	VERB
bracis-28433	160	9	.	.	PUNCT
bracis-28433	161	1	it	it	PRON
bracis-28433	161	2	contains	contain	VERB
bracis-28433	161	3	named	name	VERB
bracis-28433	161	4	entities	entity	NOUN
bracis-28433	161	5	from	from	ADP
bracis-28433	161	6	10	10	NUM
bracis-28433	161	7	classes	class	NOUN
bracis-28433	161	8	associated	associate	VERB
bracis-28433	161	9	with	with	ADP
bracis-28433	161	10	retirement	retirement	NOUN
bracis-28433	161	11	acts	act	NOUN
bracis-28433	161	12	of	of	ADP
bracis-28433	161	13	public	public	ADJ
bracis-28433	161	14	employees	employee	NOUN
bracis-28433	161	15	from	from	ADP
bracis-28433	161	16	the	the	DET
bracis-28433	161	17	diário	diário	PROPN
bracis-28433	161	18	oficial	oficial	NOUN
bracis-28433	161	19	do	do	VERB
bracis-28433	161	20	distrito	distrito	PROPN
bracis-28433	161	21	federal	federal	PROPN
bracis-28433	161	22	(	(	PUNCT
bracis-28433	161	23	brazilian	brazilian	ADJ
bracis-28433	161	24	federal	federal	ADJ
bracis-28433	161	25	district	district	PROPN
bracis-28433	161	26	official	official	ADJ
bracis-28433	161	27	gazette	gazette	PROPN
bracis-28433	161	28	,	,	PUNCT
bracis-28433	161	29	in	in	ADP
bracis-28433	161	30	direct	direct	ADJ
bracis-28433	161	31	translation	translation	NOUN
bracis-28433	161	32	)	)	PUNCT
bracis-28433	161	33	.	.	PUNCT
bracis-28433	162	1	the	the	DET
bracis-28433	162	2	datasets	dataset	NOUN
bracis-28433	162	3	chosen	choose	VERB
bracis-28433	162	4	for	for	ADP
bracis-28433	162	5	use	use	NOUN
bracis-28433	162	6	are	be	AUX
bracis-28433	162	7	listed	list	VERB
bracis-28433	162	8	in	in	ADP
bracis-28433	162	9	table	table	NOUN
bracis-28433	162	10	 	 	SPACE
bracis-28433	162	11	1	1	NUM
bracis-28433	162	12	,	,	PUNCT
bracis-28433	162	13	along	along	ADP
bracis-28433	162	14	with	with	ADP
bracis-28433	162	15	relevant	relevant	ADJ
bracis-28433	162	16	information	information	NOUN
bracis-28433	162	17	such	such	ADJ
bracis-28433	162	18	as	as	ADP
bracis-28433	162	19	their	their	PRON
bracis-28433	162	20	language	language	NOUN
bracis-28433	162	21	,	,	PUNCT
bracis-28433	162	22	subject	subject	ADJ
bracis-28433	162	23	domain	domain	NOUN
bracis-28433	162	24	.	.	PUNCT
bracis-28433	163	1	table	table	NOUN
bracis-28433	163	2	1	1	NUM
bracis-28433	163	3	.	.	PUNCT
bracis-28433	164	1	datasets	dataset	NOUN
bracis-28433	164	2	description.full	description.full	PROPN
bracis-28433	164	3	size	size	NOUN
bracis-28433	164	4	table	table	NOUN
bracis-28433	164	5	we	we	PRON
bracis-28433	164	6	use	use	VERB
bracis-28433	164	7	two	two	NUM
bracis-28433	164	8	neural	neural	ADJ
bracis-28433	164	9	models	model	NOUN
bracis-28433	164	10	for	for	ADP
bracis-28433	164	11	sentence	sentence	NOUN
bracis-28433	164	12	-	-	PUNCT
bracis-28433	164	13	level	level	NOUN
bracis-28433	164	14	and	and	CCONJ
bracis-28433	164	15	one	one	NUM
bracis-28433	164	16	for	for	ADP
bracis-28433	164	17	token	token	VERB
bracis-28433	164	18	-	-	PUNCT
bracis-28433	164	19	level	level	NOUN
bracis-28433	164	20	active	active	ADJ
bracis-28433	164	21	self	self	NOUN
bracis-28433	164	22	-	-	PUNCT
bracis-28433	164	23	learning	learning	NOUN
bracis-28433	164	24	.	.	PUNCT
bracis-28433	165	1	for	for	ADP
bracis-28433	165	2	sentence	sentence	NOUN
bracis-28433	165	3	level	level	NOUN
bracis-28433	165	4	,	,	PUNCT
bracis-28433	165	5	we	we	PRON
bracis-28433	165	6	use	use	VERB
bracis-28433	165	7	cnn	cnn	PROPN
bracis-28433	165	8	-	-	PUNCT
bracis-28433	165	9	cnn	cnn	PROPN
bracis-28433	165	10	-	-	PUNCT
bracis-28433	165	11	lstm	lstm	NOUN
bracis-28433	165	12	proposed	propose	VERB
bracis-28433	165	13	by	by	ADP
bracis-28433	165	14	shen	shen	PROPN
bracis-28433	165	15	et	et	PROPN
bracis-28433	165	16	al	al	PROPN
bracis-28433	165	17	.	.	PUNCT
bracis-28433	166	1	[	[	X
bracis-28433	166	2	18	18	NUM
bracis-28433	166	3	]	]	PUNCT
bracis-28433	166	4	and	and	CCONJ
bracis-28433	166	5	the	the	DET
bracis-28433	166	6	cnn	cnn	PROPN
bracis-28433	166	7	-	-	PUNCT
bracis-28433	166	8	bilstm	bilstm	NOUN
bracis-28433	166	9	-	-	PUNCT
bracis-28433	166	10	crf	crf	NOUN
bracis-28433	166	11	proposed	propose	VERB
bracis-28433	166	12	by	by	ADP
bracis-28433	166	13	ma	ma	PROPN
bracis-28433	166	14	and	and	CCONJ
bracis-28433	166	15	hovy	hovy	VERB
bracis-28433	167	1	[	[	X
bracis-28433	167	2	9	9	NUM
bracis-28433	167	3	]	]	PUNCT
bracis-28433	167	4	.	.	PUNCT
bracis-28433	168	1	we	we	PRON
bracis-28433	168	2	could	could	AUX
bracis-28433	168	3	not	not	PART
bracis-28433	168	4	use	use	VERB
bracis-28433	168	5	cnn	cnn	PROPN
bracis-28433	168	6	-	-	PUNCT
bracis-28433	168	7	bilstm	bilstm	NOUN
bracis-28433	168	8	-	-	PUNCT
bracis-28433	168	9	crf	crf	NOUN
bracis-28433	168	10	in	in	ADP
bracis-28433	168	11	the	the	DET
bracis-28433	168	12	token	token	ADJ
bracis-28433	168	13	-	-	PUNCT
bracis-28433	168	14	level	level	NOUN
bracis-28433	168	15	case	case	NOUN
bracis-28433	168	16	because	because	SCONJ
bracis-28433	168	17	crf	crf	NOUN
bracis-28433	168	18	classification	classification	NOUN
bracis-28433	168	19	layers	layer	NOUN
bracis-28433	168	20	output	output	VERB
bracis-28433	168	21	a	a	DET
bracis-28433	168	22	distribution	distribution	NOUN
bracis-28433	168	23	over	over	ADP
bracis-28433	168	24	the	the	DET
bracis-28433	168	25	entire	entire	ADJ
bracis-28433	168	26	sentence	sentence	NOUN
bracis-28433	168	27	instead	instead	ADV
bracis-28433	168	28	of	of	ADP
bracis-28433	168	29	a	a	DET
bracis-28433	168	30	distribution	distribution	NOUN
bracis-28433	168	31	of	of	ADP
bracis-28433	168	32	classes	class	NOUN
bracis-28433	168	33	for	for	ADP
bracis-28433	168	34	each	each	DET
bracis-28433	168	35	token	token	VERB
bracis-28433	168	36	.	.	PUNCT
bracis-28433	169	1	more	more	ADJ
bracis-28433	169	2	details	detail	NOUN
bracis-28433	169	3	about	about	ADP
bracis-28433	169	4	the	the	DET
bracis-28433	169	5	model	model	NOUN
bracis-28433	169	6	’s	’s	PART
bracis-28433	169	7	and	and	CCONJ
bracis-28433	169	8	training	training	NOUN
bracis-28433	169	9	algorithm	algorithm	NOUN
bracis-28433	169	10	’s	’s	PART
bracis-28433	169	11	hyperparameters	hyperparameter	NOUN
bracis-28433	169	12	are	be	AUX
bracis-28433	169	13	presented	present	VERB
bracis-28433	169	14	in	in	ADP
bracis-28433	169	15	the	the	DET
bracis-28433	169	16	following	follow	VERB
bracis-28433	169	17	subsections	subsection	NOUN
bracis-28433	169	18	.	.	PUNCT
bracis-28433	170	1	we	we	PRON
bracis-28433	170	2	evaluate	evaluate	VERB
bracis-28433	170	3	four	four	NUM
bracis-28433	170	4	versions	version	NOUN
bracis-28433	170	5	of	of	ADP
bracis-28433	170	6	our	our	PRON
bracis-28433	170	7	proposed	propose	VERB
bracis-28433	170	8	active	active	ADJ
bracis-28433	170	9	self	self	NOUN
bracis-28433	170	10	-	-	PUNCT
bracis-28433	170	11	learning	learn	VERB
bracis-28433	170	12	algorithm	algorithm	NOUN
bracis-28433	170	13	using	use	VERB
bracis-28433	170	14	self	self	NOUN
bracis-28433	170	15	-	-	PUNCT
bracis-28433	170	16	training	training	NOUN
bracis-28433	170	17	techniques	technique	NOUN
bracis-28433	170	18	at	at	ADP
bracis-28433	170	19	the	the	DET
bracis-28433	170	20	sentence	sentence	NOUN
bracis-28433	170	21	level	level	NOUN
bracis-28433	170	22	(	(	PUNCT
bracis-28433	170	23	sect	sect	NOUN
bracis-28433	170	24	.	.	PUNCT
bracis-28433	170	25	 	 	SPACE
bracis-28433	170	26	4	4	NUM
bracis-28433	170	27	)	)	PUNCT
bracis-28433	170	28	.	.	PUNCT
bracis-28433	171	1	we	we	PRON
bracis-28433	171	2	also	also	ADV
bracis-28433	171	3	compare	compare	VERB
bracis-28433	171	4	the	the	DET
bracis-28433	171	5	results	result	NOUN
bracis-28433	171	6	with	with	ADP
bracis-28433	171	7	the	the	DET
bracis-28433	171	8	deep	deep	ADJ
bracis-28433	171	9	active	active	ADJ
bracis-28433	171	10	learning	learning	NOUN
bracis-28433	171	11	algorithm	algorithm	NOUN
bracis-28433	171	12	proposed	propose	VERB
bracis-28433	171	13	by	by	ADP
bracis-28433	171	14	shen	shen	PROPN
bracis-28433	171	15	et	et	PROPN
bracis-28433	171	16	al	al	PROPN
bracis-28433	171	17	.	.	PUNCT
bracis-28433	172	1	[	[	X
bracis-28433	172	2	18	18	NUM
bracis-28433	172	3	]	]	PUNCT
bracis-28433	172	4	.	.	PUNCT
bracis-28433	173	1	moreover	moreover	ADV
bracis-28433	173	2	,	,	PUNCT
bracis-28433	173	3	we	we	PRON
bracis-28433	173	4	compare	compare	VERB
bracis-28433	173	5	the	the	DET
bracis-28433	173	6	token	token	ADJ
bracis-28433	173	7	-	-	PUNCT
bracis-28433	173	8	level	level	NOUN
bracis-28433	173	9	active	active	ADJ
bracis-28433	173	10	self	self	NOUN
bracis-28433	173	11	-	-	PUNCT
bracis-28433	173	12	learning	learn	VERB
bracis-28433	173	13	algorithm	algorithm	NOUN
bracis-28433	173	14	with	with	ADP
bracis-28433	173	15	the	the	DET
bracis-28433	173	16	subsequence	subsequence	NOUN
bracis-28433	173	17	-	-	PUNCT
bracis-28433	173	18	based	base	VERB
bracis-28433	173	19	active	active	ADJ
bracis-28433	173	20	learning	learn	VERB
bracis-28433	173	21	algorithm	algorithm	NOUN
bracis-28433	173	22	by	by	ADP
bracis-28433	173	23	radmard	radmard	PROPN
bracis-28433	173	24	et	et	PROPN
bracis-28433	173	25	al	al	PROPN
bracis-28433	173	26	.	.	PUNCT
bracis-28433	174	1	[	[	X
bracis-28433	174	2	16	16	NUM
bracis-28433	174	3	]	]	PUNCT
bracis-28433	174	4	.	.	PUNCT
bracis-28433	175	1	this	this	DET
bracis-28433	175	2	algorithm	algorithm	NOUN
bracis-28433	175	3	was	be	AUX
bracis-28433	175	4	chosen	choose	VERB
bracis-28433	175	5	as	as	ADP
bracis-28433	175	6	a	a	DET
bracis-28433	175	7	baseline	baseline	NOUN
bracis-28433	175	8	because	because	SCONJ
bracis-28433	175	9	its	its	PRON
bracis-28433	175	10	queries	query	NOUN
bracis-28433	175	11	use	use	VERB
bracis-28433	175	12	sub	sub	NOUN
bracis-28433	175	13	-	-	NOUN
bracis-28433	175	14	sentences	sentence	NOUN
bracis-28433	175	15	instead	instead	ADV
bracis-28433	175	16	of	of	ADP
bracis-28433	175	17	whole	whole	ADJ
bracis-28433	175	18	sentences	sentence	NOUN
bracis-28433	175	19	.	.	PUNCT
bracis-28433	176	1	for	for	ADP
bracis-28433	176	2	all	all	DET
bracis-28433	176	3	proposed	propose	VERB
bracis-28433	176	4	algorithms	algorithm	NOUN
bracis-28433	176	5	,	,	PUNCT
bracis-28433	176	6	we	we	PRON
bracis-28433	176	7	compute	compute	VERB
bracis-28433	176	8	the	the	DET
bracis-28433	176	9	maximum	maximum	NOUN
bracis-28433	176	10	normalized	normalize	VERB
bracis-28433	176	11	log	log	NOUN
bracis-28433	176	12	-	-	PUNCT
bracis-28433	176	13	probability	probability	NOUN
bracis-28433	176	14	(	(	PUNCT
bracis-28433	176	15	mnlp	mnlp	ADJ
bracis-28433	176	16	)	)	PUNCT
bracis-28433	176	17	measure	measure	NOUN
bracis-28433	176	18	proposed	propose	VERB
bracis-28433	176	19	by	by	ADP
bracis-28433	176	20	shen	shen	PROPN
bracis-28433	176	21	et	et	PROPN
bracis-28433	176	22	al	al	PROPN
bracis-28433	176	23	.	.	PUNCT
bracis-28433	177	1	[	[	X
bracis-28433	177	2	18	18	NUM
bracis-28433	177	3	]	]	PUNCT
bracis-28433	177	4	to	to	PART
bracis-28433	177	5	generate	generate	VERB
bracis-28433	177	6	queries	query	NOUN
bracis-28433	177	7	for	for	ADP
bracis-28433	177	8	new	new	ADJ
bracis-28433	177	9	samples	sample	NOUN
bracis-28433	177	10	.	.	PUNCT
bracis-28433	178	1	the	the	DET
bracis-28433	178	2	mnlp	mnlp	ADJ
bracis-28433	178	3	value	value	NOUN
bracis-28433	178	4	for	for	ADP
bracis-28433	178	5	a	a	DET
bracis-28433	178	6	sentence	sentence	NOUN
bracis-28433	178	7	x	x	PUNCT
bracis-28433	178	8	of	of	ADP
bracis-28433	178	9	length	length	NOUN
bracis-28433	178	10	n	n	NUM
bracis-28433	178	11	can	can	AUX
bracis-28433	178	12	be	be	AUX
bracis-28433	178	13	calculated	calculate	VERB
bracis-28433	178	14	as	as	ADP
bracis-28433	178	15	$	$	SYM
bracis-28433	178	16	$	$	SYM
bracis-28433	178	17	\begin{aligned	\begin{aligne	VERB
bracis-28433	178	18	}	}	PUNCT
bracis-28433	178	19	mnlp(x	mnlp(x	PROPN
bracis-28433	178	20	)	)	PUNCT
bracis-28433	178	21	=	=	NOUN
bracis-28433	179	1	\max	\max	PROPN
bracis-28433	179	2	_	_	PUNCT
bracis-28433	179	3	{	{	PUNCT
bracis-28433	179	4	y_1	y_1	PROPN
bracis-28433	179	5	,	,	PUNCT
bracis-28433	179	6	...	...	PUNCT
bracis-28433	179	7	,	,	PUNCT
bracis-28433	179	8	y_{n-1}}\frac{1}{n}\sum	y_{n-1}}\frac{1}{n}\sum	PROPN
bracis-28433	179	9	^n_{i=0}log\	^n_{i=0}log\	NOUN
bracis-28433	179	10	p(y_i\vert	p(y_i\vert	PROPN
bracis-28433	179	11	x_i	x_i	PROPN
bracis-28433	179	12	,	,	PUNCT
bracis-28433	179	13	y_0	y_0	PROPN
bracis-28433	179	14	,	,	PUNCT
bracis-28433	179	15	y_1	y_1	PROPN
bracis-28433	179	16	,	,	PUNCT
bracis-28433	179	17	...	...	PUNCT
bracis-28433	179	18	,	,	PUNCT
bracis-28433	179	19	y_{i-1	y_{i-1	NUM
bracis-28433	179	20	}	}	PUNCT
bracis-28433	179	21	)	)	PUNCT
bracis-28433	179	22	.	.	PUNCT
bracis-28433	180	1	\end{aligned}$$	\end{aligned}$$	PROPN
bracis-28433	180	2	(	(	PUNCT
bracis-28433	180	3	1	1	NUM
bracis-28433	180	4	)	)	PUNCT
bracis-28433	180	5	where	where	SCONJ
bracis-28433	180	6	\(x_i\	\(x_i\	NOUN
bracis-28433	180	7	)	)	PUNCT
bracis-28433	180	8	is	be	AUX
bracis-28433	180	9	the	the	DET
bracis-28433	180	10	i	i	PROPN
bracis-28433	180	11	-	-	PUNCT
bracis-28433	180	12	th	th	X
bracis-28433	180	13	token	token	NOUN
bracis-28433	180	14	of	of	ADP
bracis-28433	180	15	a	a	DET
bracis-28433	180	16	sentence	sentence	NOUN
bracis-28433	180	17	x	x	NOUN
bracis-28433	180	18	,	,	PUNCT
bracis-28433	180	19	and	and	CCONJ
bracis-28433	180	20	\(y_i\	\(y_i\	NOUN
bracis-28433	180	21	)	)	PUNCT
bracis-28433	180	22	is	be	AUX
bracis-28433	180	23	the	the	DET
bracis-28433	180	24	class	class	NOUN
bracis-28433	180	25	predicted	predict	VERB
bracis-28433	180	26	for	for	ADP
bracis-28433	180	27	the	the	DET
bracis-28433	180	28	i	i	PROPN
bracis-28433	180	29	-	-	PUNCT
bracis-28433	180	30	th	th	X
bracis-28433	180	31	token	token	VERB
bracis-28433	180	32	\(x_i\	\(x_i\	NOUN
bracis-28433	180	33	)	)	PUNCT
bracis-28433	180	34	.	.	PUNCT
bracis-28433	181	1	the	the	DET
bracis-28433	181	2	querying	query	VERB
bracis-28433	181	3	for	for	ADP
bracis-28433	181	4	the	the	DET
bracis-28433	181	5	oracle	oracle	NOUN
bracis-28433	181	6	and	and	CCONJ
bracis-28433	181	7	the	the	DET
bracis-28433	181	8	trained	train	VERB
bracis-28433	181	9	model	model	NOUN
bracis-28433	181	10	is	be	AUX
bracis-28433	181	11	based	base	VERB
bracis-28433	181	12	on	on	ADP
bracis-28433	181	13	the	the	DET
bracis-28433	181	14	mnlp	mnlp	ADJ
bracis-28433	181	15	measure	measure	NOUN
bracis-28433	181	16	.	.	PUNCT
bracis-28433	182	1	we	we	PRON
bracis-28433	182	2	use	use	VERB
bracis-28433	182	3	the	the	DET
bracis-28433	182	4	least	least	ADJ
bracis-28433	182	5	confidence	confidence	NOUN
bracis-28433	182	6	for	for	ADP
bracis-28433	182	7	the	the	DET
bracis-28433	182	8	oracle	oracle	NOUN
bracis-28433	182	9	annotation	annotation	NOUN
bracis-28433	182	10	,	,	PUNCT
bracis-28433	182	11	where	where	SCONJ
bracis-28433	182	12	unlabeled	unlabele	VERB
bracis-28433	182	13	samples	sample	NOUN
bracis-28433	182	14	with	with	ADP
bracis-28433	182	15	the	the	DET
bracis-28433	182	16	lowest	low	ADJ
bracis-28433	182	17	mnlp	mnlp	ADJ
bracis-28433	182	18	scores	score	NOUN
bracis-28433	182	19	in	in	ADP
bracis-28433	182	20	a	a	DET
bracis-28433	182	21	given	give	VERB
bracis-28433	182	22	algorithm	algorithm	NOUN
bracis-28433	182	23	iteration	iteration	NOUN
bracis-28433	182	24	will	will	AUX
bracis-28433	182	25	be	be	AUX
bracis-28433	182	26	queried	query	VERB
bracis-28433	182	27	to	to	ADP
bracis-28433	182	28	the	the	DET
bracis-28433	182	29	oracle	oracle	NOUN
bracis-28433	182	30	.	.	PUNCT
bracis-28433	183	1	for	for	ADP
bracis-28433	183	2	automatic	automatic	ADJ
bracis-28433	183	3	annotation	annotation	NOUN
bracis-28433	183	4	,	,	PUNCT
bracis-28433	183	5	we	we	PRON
bracis-28433	183	6	use	use	VERB
bracis-28433	183	7	the	the	DET
bracis-28433	183	8	exponentiated	exponentiate	VERB
bracis-28433	183	9	mnlp	mnlp	ADJ
bracis-28433	183	10	measure	measure	NOUN
bracis-28433	183	11	.	.	PUNCT
bracis-28433	184	1	the	the	DET
bracis-28433	184	2	unlabeled	unlabele	VERB
bracis-28433	184	3	samples	sample	NOUN
bracis-28433	184	4	with	with	ADP
bracis-28433	184	5	exponentiated	exponentiated	ADJ
bracis-28433	184	6	mnlp	mnlp	ADJ
bracis-28433	184	7	measure	measure	NOUN
bracis-28433	184	8	higher	high	ADJ
bracis-28433	184	9	than	than	ADP
bracis-28433	184	10	a	a	DET
bracis-28433	184	11	predefined	predefine	VERB
bracis-28433	184	12	threshold	threshold	NOUN
bracis-28433	184	13	will	will	AUX
bracis-28433	184	14	be	be	AUX
bracis-28433	184	15	used	use	VERB
bracis-28433	184	16	for	for	ADP
bracis-28433	184	17	self	self	NOUN
bracis-28433	184	18	-	-	PUNCT
bracis-28433	184	19	training	training	NOUN
bracis-28433	184	20	in	in	ADP
bracis-28433	184	21	the	the	DET
bracis-28433	184	22	next	next	ADJ
bracis-28433	184	23	iteration	iteration	NOUN
bracis-28433	184	24	of	of	ADP
bracis-28433	184	25	the	the	DET
bracis-28433	184	26	active	active	ADJ
bracis-28433	184	27	self	self	NOUN
bracis-28433	184	28	-	-	PUNCT
bracis-28433	184	29	learning	learn	VERB
bracis-28433	184	30	algorithm	algorithm	NOUN
bracis-28433	184	31	.	.	PUNCT
bracis-28433	185	1	for	for	ADP
bracis-28433	185	2	all	all	DET
bracis-28433	185	3	experiments	experiment	NOUN
bracis-28433	185	4	,	,	PUNCT
bracis-28433	185	5	the	the	DET
bracis-28433	185	6	threshold	threshold	NOUN
bracis-28433	185	7	selected	select	VERB
bracis-28433	185	8	empirically	empirically	ADV
bracis-28433	185	9	was	be	AUX
bracis-28433	185	10	0.99	0.99	NUM
bracis-28433	185	11	.	.	PUNCT
bracis-28433	186	1	this	this	DET
bracis-28433	186	2	threshold	threshold	NOUN
bracis-28433	186	3	means	mean	VERB
bracis-28433	186	4	that	that	SCONJ
bracis-28433	186	5	the	the	DET
bracis-28433	186	6	model	model	NOUN
bracis-28433	186	7	’s	’s	PART
bracis-28433	186	8	prediction	prediction	NOUN
bracis-28433	186	9	confidence	confidence	NOUN
bracis-28433	186	10	must	must	AUX
bracis-28433	186	11	be	be	AUX
bracis-28433	186	12	equal	equal	ADJ
bracis-28433	186	13	to	to	ADP
bracis-28433	186	14	or	or	CCONJ
bracis-28433	186	15	higher	high	ADJ
bracis-28433	186	16	than	than	ADP
bracis-28433	186	17	0.99	0.99	NUM
bracis-28433	186	18	for	for	SCONJ
bracis-28433	186	19	the	the	DET
bracis-28433	186	20	sample	sample	NOUN
bracis-28433	186	21	to	to	PART
bracis-28433	186	22	be	be	AUX
bracis-28433	186	23	used	use	VERB
bracis-28433	186	24	for	for	ADP
bracis-28433	186	25	self	self	NOUN
bracis-28433	186	26	-	-	PUNCT
bracis-28433	186	27	training	training	NOUN
bracis-28433	186	28	.	.	PUNCT
bracis-28433	187	1	similar	similar	ADJ
bracis-28433	187	2	to	to	ADP
bracis-28433	187	3	previous	previous	ADJ
bracis-28433	187	4	works	work	NOUN
bracis-28433	187	5	in	in	ADP
bracis-28433	187	6	the	the	DET
bracis-28433	187	7	literature	literature	NOUN
bracis-28433	187	8	[	[	X
bracis-28433	187	9	18	18	NUM
bracis-28433	187	10	,	,	PUNCT
bracis-28433	187	11	19	19	NUM
bracis-28433	187	12	]	]	PUNCT
bracis-28433	187	13	,	,	PUNCT
bracis-28433	187	14	we	we	PRON
bracis-28433	187	15	apply	apply	VERB
bracis-28433	187	16	early	early	ADJ
bracis-28433	187	17	stopping	stopping	NOUN
bracis-28433	187	18	of	of	ADP
bracis-28433	187	19	the	the	DET
bracis-28433	187	20	model	model	NOUN
bracis-28433	187	21	training	training	NOUN
bracis-28433	187	22	.	.	PUNCT
bracis-28433	188	1	the	the	DET
bracis-28433	188	2	early	early	ADJ
bracis-28433	188	3	stopping	stopping	NOUN
bracis-28433	188	4	is	be	AUX
bracis-28433	188	5	based	base	VERB
bracis-28433	188	6	on	on	ADP
bracis-28433	188	7	the	the	DET
bracis-28433	188	8	model	model	NOUN
bracis-28433	188	9	’s	’s	PART
bracis-28433	188	10	performance	performance	NOUN
bracis-28433	188	11	on	on	ADP
bracis-28433	188	12	the	the	DET
bracis-28433	188	13	validation	validation	NOUN
bracis-28433	188	14	set	set	NOUN
bracis-28433	188	15	.	.	PUNCT
bracis-28433	189	1	we	we	PRON
bracis-28433	189	2	empirically	empirically	ADV
bracis-28433	189	3	chose	choose	VERB
bracis-28433	189	4	a	a	DET
bracis-28433	189	5	patience	patience	NOUN
bracis-28433	189	6	of	of	ADP
bracis-28433	189	7	15	15	NUM
bracis-28433	189	8	epochs	epoch	NOUN
bracis-28433	189	9	,	,	PUNCT
bracis-28433	189	10	meaning	mean	VERB
bracis-28433	189	11	that	that	SCONJ
bracis-28433	189	12	if	if	SCONJ
bracis-28433	189	13	the	the	DET
bracis-28433	189	14	model	model	NOUN
bracis-28433	189	15	’s	’s	PART
bracis-28433	189	16	performance	performance	NOUN
bracis-28433	189	17	does	do	AUX
bracis-28433	189	18	not	not	PART
bracis-28433	189	19	improve	improve	VERB
bracis-28433	189	20	in	in	ADP
bracis-28433	189	21	15	15	NUM
bracis-28433	189	22	consecutive	consecutive	ADJ
bracis-28433	189	23	epochs	epoch	NOUN
bracis-28433	189	24	,	,	PUNCT
bracis-28433	189	25	the	the	DET
bracis-28433	189	26	model	model	NOUN
bracis-28433	189	27	’s	’s	PART
bracis-28433	189	28	training	training	NOUN
bracis-28433	189	29	is	be	AUX
bracis-28433	189	30	interrupted	interrupt	VERB
bracis-28433	189	31	.	.	PUNCT
bracis-28433	190	1	6.1	6.1	NUM
bracis-28433	190	2	cnn	cnn	PROPN
bracis-28433	190	3	-	-	PUNCT
bracis-28433	190	4	cnn	cnn	PROPN
bracis-28433	190	5	-	-	PUNCT
bracis-28433	190	6	lstm	lstm	PROPN
bracis-28433	190	7	model	model	NOUN
bracis-28433	190	8	the	the	DET
bracis-28433	190	9	cnn	cnn	PROPN
bracis-28433	190	10	-	-	PUNCT
bracis-28433	190	11	cnn	cnn	PROPN
bracis-28433	190	12	-	-	PUNCT
bracis-28433	190	13	lstm	lstm	ADJ
bracis-28433	190	14	model	model	NOUN
bracis-28433	190	15	consists	consist	VERB
bracis-28433	190	16	of	of	ADP
bracis-28433	190	17	a	a	DET
bracis-28433	190	18	character	character	NOUN
bracis-28433	190	19	-	-	PUNCT
bracis-28433	190	20	level	level	NOUN
bracis-28433	190	21	convolutional	convolutional	ADJ
bracis-28433	190	22	encoder	encoder	NOUN
bracis-28433	190	23	,	,	PUNCT
bracis-28433	190	24	a	a	DET
bracis-28433	190	25	word	word	NOUN
bracis-28433	190	26	-	-	PUNCT
bracis-28433	190	27	level	level	NOUN
bracis-28433	190	28	convolutional	convolutional	ADJ
bracis-28433	190	29	encoder	encoder	NOUN
bracis-28433	190	30	,	,	PUNCT
bracis-28433	190	31	and	and	CCONJ
bracis-28433	190	32	an	an	DET
bracis-28433	190	33	lstm	lstm	ADJ
bracis-28433	190	34	tag	tag	NOUN
bracis-28433	190	35	decoder	decoder	NOUN
bracis-28433	190	36	.	.	PUNCT
bracis-28433	191	1	the	the	DET
bracis-28433	191	2	character	character	NOUN
bracis-28433	191	3	-	-	PUNCT
bracis-28433	191	4	level	level	NOUN
bracis-28433	191	5	cnn	cnn	PROPN
bracis-28433	191	6	is	be	AUX
bracis-28433	191	7	used	use	VERB
bracis-28433	191	8	to	to	PART
bracis-28433	191	9	generate	generate	VERB
bracis-28433	191	10	character	character	NOUN
bracis-28433	191	11	-	-	PUNCT
bracis-28433	191	12	level	level	NOUN
bracis-28433	191	13	vector	vector	NOUN
bracis-28433	191	14	representations	representation	NOUN
bracis-28433	191	15	of	of	ADP
bracis-28433	191	16	a	a	DET
bracis-28433	191	17	word	word	NOUN
bracis-28433	191	18	.	.	PUNCT
bracis-28433	192	1	this	this	DET
bracis-28433	192	2	cnn	cnn	PROPN
bracis-28433	192	3	works	work	VERB
bracis-28433	192	4	by	by	ADP
bracis-28433	192	5	first	first	ADV
bracis-28433	192	6	transforming	transform	VERB
bracis-28433	192	7	each	each	DET
bracis-28433	192	8	character	character	NOUN
bracis-28433	192	9	of	of	ADP
bracis-28433	192	10	a	a	DET
bracis-28433	192	11	word	word	NOUN
bracis-28433	192	12	into	into	ADP
bracis-28433	192	13	a	a	DET
bracis-28433	192	14	vector	vector	NOUN
bracis-28433	192	15	representation	representation	NOUN
bracis-28433	192	16	,	,	PUNCT
bracis-28433	192	17	with	with	ADP
bracis-28433	192	18	embeddings	embedding	NOUN
bracis-28433	192	19	being	be	AUX
bracis-28433	192	20	initialized	initialize	VERB
bracis-28433	192	21	with	with	ADP
bracis-28433	192	22	uniform	uniform	ADJ
bracis-28433	192	23	samples	sample	NOUN
bracis-28433	192	24	from	from	ADP
bracis-28433	192	25	\(\left	\(\left	PROPN
bracis-28433	192	26	[	[	PUNCT
bracis-28433	192	27	-\sqrt{\frac{3}{dim	-\sqrt{\frac{3}{dim	PROPN
bracis-28433	192	28	}	}	PUNCT
bracis-28433	192	29	}	}	PUNCT
bracis-28433	192	30	,	,	PUNCT
bracis-28433	192	31	+	+	ADJ
bracis-28433	192	32	\sqrt{\frac{3}{dim	\sqrt{\frac{3}{dim	NOUN
bracis-28433	192	33	}	}	PUNCT
bracis-28433	192	34	}	}	PUNCT
bracis-28433	192	35	\right	\right	PROPN
bracis-28433	192	36	]	]	PUNCT
bracis-28433	192	37	\	\	NOUN
bracis-28433	192	38	)	)	PUNCT
bracis-28433	192	39	as	as	SCONJ
bracis-28433	192	40	proposed	propose	VERB
bracis-28433	192	41	by	by	ADP
bracis-28433	192	42	ma	ma	PROPN
bracis-28433	192	43	and	and	CCONJ
bracis-28433	192	44	hovy	hovy	VERB
bracis-28433	193	1	[	[	X
bracis-28433	193	2	9	9	NUM
bracis-28433	193	3	]	]	PUNCT
bracis-28433	193	4	.	.	PUNCT
bracis-28433	194	1	dropout	dropout	NOUN
bracis-28433	195	1	[	[	X
bracis-28433	195	2	20	20	NUM
bracis-28433	195	3	]	]	PUNCT
bracis-28433	195	4	is	be	AUX
bracis-28433	195	5	applied	apply	VERB
bracis-28433	195	6	to	to	ADP
bracis-28433	195	7	the	the	DET
bracis-28433	195	8	generated	generate	VERB
bracis-28433	195	9	embeddings	embedding	NOUN
bracis-28433	195	10	as	as	SCONJ
bracis-28433	195	11	presented	present	VERB
bracis-28433	195	12	by	by	ADP
bracis-28433	195	13	ma	ma	PROPN
bracis-28433	195	14	and	and	CCONJ
bracis-28433	195	15	hovy	hovy	VERB
bracis-28433	196	1	[	[	X
bracis-28433	196	2	9	9	NUM
bracis-28433	196	3	]	]	PUNCT
bracis-28433	196	4	.	.	PUNCT
bracis-28433	197	1	then	then	ADV
bracis-28433	197	2	,	,	PUNCT
bracis-28433	197	3	a	a	DET
bracis-28433	197	4	one	one	NUM
bracis-28433	197	5	-	-	PUNCT
bracis-28433	197	6	dimensional	dimensional	ADJ
bracis-28433	197	7	convolutional	convolutional	ADJ
bracis-28433	197	8	layer	layer	NOUN
bracis-28433	197	9	is	be	AUX
bracis-28433	197	10	used	use	VERB
bracis-28433	197	11	to	to	PART
bracis-28433	197	12	extract	extract	VERB
bracis-28433	197	13	information	information	NOUN
bracis-28433	197	14	of	of	ADP
bracis-28433	197	15	neighboring	neighboring	NOUN
bracis-28433	197	16	characters	character	NOUN
bracis-28433	197	17	,	,	PUNCT
bracis-28433	197	18	followed	follow	VERB
bracis-28433	197	19	by	by	ADP
bracis-28433	197	20	relu	relu	NOUN
bracis-28433	197	21	activation	activation	NOUN
bracis-28433	197	22	[	[	X
bracis-28433	197	23	11	11	NUM
bracis-28433	197	24	]	]	PUNCT
bracis-28433	197	25	and	and	CCONJ
bracis-28433	197	26	max	max	PROPN
bracis-28433	197	27	-	-	PUNCT
bracis-28433	197	28	pooling	pooling	NOUN
bracis-28433	197	29	to	to	PART
bracis-28433	197	30	generate	generate	VERB
bracis-28433	197	31	fixed	fix	VERB
bracis-28433	197	32	-	-	PUNCT
bracis-28433	197	33	size	size	NOUN
bracis-28433	197	34	character	character	NOUN
bracis-28433	197	35	-	-	PUNCT
bracis-28433	197	36	level	level	NOUN
bracis-28433	197	37	word	word	NOUN
bracis-28433	197	38	embeddings	embedding	NOUN
bracis-28433	197	39	.	.	PUNCT
bracis-28433	198	1	we	we	PRON
bracis-28433	198	2	generate	generate	VERB
bracis-28433	198	3	a	a	DET
bracis-28433	198	4	word	word	NOUN
bracis-28433	198	5	representation	representation	NOUN
bracis-28433	198	6	by	by	ADP
bracis-28433	198	7	concatenating	concatenate	VERB
bracis-28433	198	8	a	a	DET
bracis-28433	198	9	character	character	NOUN
bracis-28433	198	10	-	-	PUNCT
bracis-28433	198	11	level	level	NOUN
bracis-28433	198	12	vector	vector	NOUN
bracis-28433	198	13	with	with	ADP
bracis-28433	198	14	a	a	DET
bracis-28433	198	15	vector	vector	NOUN
bracis-28433	198	16	representation	representation	NOUN
bracis-28433	198	17	from	from	ADP
bracis-28433	198	18	a	a	DET
bracis-28433	198	19	lookup	lookup	NOUN
bracis-28433	198	20	table	table	NOUN
bracis-28433	198	21	,	,	PUNCT
bracis-28433	198	22	which	which	PRON
bracis-28433	198	23	we	we	PRON
bracis-28433	198	24	initialize	initialize	VERB
bracis-28433	198	25	with	with	ADP
bracis-28433	198	26	pre	pre	ADJ
bracis-28433	198	27	-	-	ADJ
bracis-28433	198	28	trained	train	VERB
bracis-28433	198	29	word	word	NOUN
bracis-28433	198	30	embeddings	embedding	NOUN
bracis-28433	198	31	that	that	PRON
bracis-28433	198	32	we	we	PRON
bracis-28433	198	33	update	update	VERB
bracis-28433	198	34	throughout	throughout	ADP
bracis-28433	198	35	the	the	DET
bracis-28433	198	36	training	training	NOUN
bracis-28433	198	37	process	process	NOUN
bracis-28433	198	38	.	.	PUNCT
bracis-28433	199	1	we	we	PRON
bracis-28433	199	2	used	use	VERB
bracis-28433	199	3	glove	glove	NOUN
bracis-28433	199	4	embeddings	embedding	NOUN
bracis-28433	199	5	of	of	ADP
bracis-28433	199	6	100	100	NUM
bracis-28433	199	7	dimensions	dimension	NOUN
bracis-28433	199	8	pre	pre	VERB
bracis-28433	199	9	-	-	VERB
bracis-28433	199	10	trained	train	VERB
bracis-28433	199	11	on	on	ADP
bracis-28433	199	12	english	english	PROPN
bracis-28433	199	13	newswire	newswire	PROPN
bracis-28433	199	14	corpus	corpus	NOUN
bracis-28433	200	1	[	[	X
bracis-28433	200	2	14	14	NUM
bracis-28433	200	3	]	]	PUNCT
bracis-28433	200	4	,	,	PUNCT
bracis-28433	200	5	and	and	CCONJ
bracis-28433	200	6	glove	glove	NOUN
bracis-28433	200	7	embeddings	embedding	NOUN
bracis-28433	200	8	of	of	ADP
bracis-28433	200	9	300	300	NUM
bracis-28433	200	10	dimensions	dimension	NOUN
bracis-28433	200	11	pre	pre	VERB
bracis-28433	200	12	-	-	VERB
bracis-28433	200	13	trained	train	VERB
bracis-28433	200	14	on	on	ADP
bracis-28433	200	15	multi	multi	ADJ
bracis-28433	200	16	-	-	ADJ
bracis-28433	200	17	genre	genre	ADJ
bracis-28433	200	18	portuguese	portuguese	ADJ
bracis-28433	200	19	corpus	corpus	NOUN
bracis-28433	200	20	[	[	X
bracis-28433	200	21	5	5	NUM
bracis-28433	200	22	]	]	PUNCT
bracis-28433	200	23	.	.	PUNCT
bracis-28433	201	1	the	the	DET
bracis-28433	201	2	lstm	lstm	PROPN
bracis-28433	201	3	tag	tag	NOUN
bracis-28433	201	4	decoder	decoder	NOUN
bracis-28433	201	5	is	be	AUX
bracis-28433	201	6	responsible	responsible	ADJ
bracis-28433	201	7	for	for	ADP
bracis-28433	201	8	the	the	DET
bracis-28433	201	9	tagging	tagging	NOUN
bracis-28433	201	10	of	of	ADP
bracis-28433	201	11	each	each	DET
bracis-28433	201	12	word	word	NOUN
bracis-28433	201	13	.	.	PUNCT
bracis-28433	202	1	the	the	DET
bracis-28433	202	2	lstm	lstm	PROPN
bracis-28433	202	3	tag	tag	NOUN
bracis-28433	202	4	decoder	decoder	NOUN
bracis-28433	202	5	uses	use	VERB
bracis-28433	202	6	a	a	DET
bracis-28433	202	7	one	one	NUM
bracis-28433	202	8	-	-	PUNCT
bracis-28433	202	9	hot	hot	ADJ
bracis-28433	202	10	encoded	encode	VERB
bracis-28433	202	11	vector	vector	NOUN
bracis-28433	202	12	of	of	ADP
bracis-28433	202	13	the	the	DET
bracis-28433	202	14	previous	previous	ADJ
bracis-28433	202	15	tag	tag	NOUN
bracis-28433	202	16	,	,	PUNCT
bracis-28433	202	17	concatenates	concatenate	VERB
bracis-28433	202	18	it	it	PRON
bracis-28433	202	19	to	to	ADP
bracis-28433	202	20	the	the	DET
bracis-28433	202	21	encoded	encode	VERB
bracis-28433	202	22	vector	vector	NOUN
bracis-28433	202	23	representation	representation	NOUN
bracis-28433	202	24	of	of	ADP
bracis-28433	202	25	the	the	DET
bracis-28433	202	26	current	current	ADJ
bracis-28433	202	27	word	word	NOUN
bracis-28433	202	28	,	,	PUNCT
bracis-28433	202	29	and	and	CCONJ
bracis-28433	202	30	uses	use	VERB
bracis-28433	202	31	it	it	PRON
bracis-28433	202	32	as	as	ADP
bracis-28433	202	33	input	input	NOUN
bracis-28433	202	34	.	.	PUNCT
bracis-28433	203	1	a	a	DET
bracis-28433	203	2	thorough	thorough	ADJ
bracis-28433	203	3	explanation	explanation	NOUN
bracis-28433	203	4	of	of	ADP
bracis-28433	203	5	the	the	DET
bracis-28433	203	6	hyperparameter	hyperparameter	NOUN
bracis-28433	203	7	tuning	tune	VERB
bracis-28433	203	8	routine	routine	ADJ
bracis-28433	203	9	and	and	CCONJ
bracis-28433	203	10	hyperparameter	hyperparameter	NOUN
bracis-28433	203	11	values	value	NOUN
bracis-28433	203	12	used	use	VERB
bracis-28433	203	13	in	in	ADP
bracis-28433	203	14	our	our	PRON
bracis-28433	203	15	experiments	experiment	NOUN
bracis-28433	203	16	is	be	AUX
bracis-28433	203	17	presented	present	VERB
bracis-28433	203	18	in	in	ADP
bracis-28433	203	19	[	[	X
bracis-28433	203	20	1	1	NUM
bracis-28433	203	21	]	]	PUNCT
bracis-28433	203	22	.	.	PUNCT
bracis-28433	204	1	6.2	6.2	NUM
bracis-28433	204	2	cnn	cnn	PROPN
bracis-28433	204	3	-	-	PUNCT
bracis-28433	204	4	bilstm	bilstm	NOUN
bracis-28433	204	5	-	-	PUNCT
bracis-28433	204	6	crf	crf	NOUN
bracis-28433	204	7	model	model	NOUN
bracis-28433	204	8	similarly	similarly	ADV
bracis-28433	204	9	to	to	ADP
bracis-28433	204	10	the	the	DET
bracis-28433	204	11	cnn	cnn	PROPN
bracis-28433	204	12	-	-	PUNCT
bracis-28433	204	13	cnn	cnn	PROPN
bracis-28433	204	14	-	-	PUNCT
bracis-28433	204	15	lstm	lstm	PROPN
bracis-28433	204	16	model	model	NOUN
bracis-28433	204	17	,	,	PUNCT
bracis-28433	204	18	the	the	DET
bracis-28433	204	19	cnn	cnn	PROPN
bracis-28433	204	20	-	-	PUNCT
bracis-28433	204	21	bilstm	bilstm	NOUN
bracis-28433	204	22	-	-	PUNCT
bracis-28433	204	23	crf	crf	NOUN
bracis-28433	204	24	model	model	NOUN
bracis-28433	204	25	uses	use	VERB
bracis-28433	204	26	a	a	DET
bracis-28433	204	27	cnn	cnn	NOUN
bracis-28433	204	28	to	to	PART
bracis-28433	204	29	generate	generate	VERB
bracis-28433	204	30	a	a	DET
bracis-28433	204	31	character	character	NOUN
bracis-28433	204	32	-	-	PUNCT
bracis-28433	204	33	level	level	NOUN
bracis-28433	204	34	representation	representation	NOUN
bracis-28433	204	35	for	for	ADP
bracis-28433	204	36	each	each	DET
bracis-28433	204	37	word	word	NOUN
bracis-28433	204	38	.	.	PUNCT
bracis-28433	205	1	the	the	DET
bracis-28433	205	2	full	full	ADJ
bracis-28433	205	3	embedding	embedding	NOUN
bracis-28433	205	4	,	,	PUNCT
bracis-28433	205	5	formed	form	VERB
bracis-28433	205	6	by	by	ADP
bracis-28433	205	7	the	the	DET
bracis-28433	205	8	concatenation	concatenation	NOUN
bracis-28433	205	9	of	of	ADP
bracis-28433	205	10	character	character	NOUN
bracis-28433	205	11	-	-	PUNCT
bracis-28433	205	12	level	level	NOUN
bracis-28433	205	13	and	and	CCONJ
bracis-28433	205	14	word	word	NOUN
bracis-28433	205	15	-	-	PUNCT
bracis-28433	205	16	level	level	NOUN
bracis-28433	205	17	embeddings	embedding	NOUN
bracis-28433	205	18	,	,	PUNCT
bracis-28433	205	19	is	be	AUX
bracis-28433	205	20	fed	feed	VERB
bracis-28433	205	21	to	to	ADP
bracis-28433	205	22	a	a	DET
bracis-28433	205	23	bilstm	bilstm	NOUN
bracis-28433	205	24	layer	layer	NOUN
bracis-28433	205	25	that	that	PRON
bracis-28433	205	26	generates	generate	VERB
bracis-28433	205	27	vector	vector	NOUN
bracis-28433	205	28	representations	representation	NOUN
bracis-28433	205	29	for	for	ADP
bracis-28433	205	30	each	each	DET
bracis-28433	205	31	word	word	NOUN
bracis-28433	205	32	in	in	ADP
bracis-28433	205	33	a	a	DET
bracis-28433	205	34	sentence	sentence	NOUN
bracis-28433	205	35	.	.	PUNCT
bracis-28433	206	1	a	a	DET
bracis-28433	206	2	fully	fully	ADV
bracis-28433	206	3	-	-	PUNCT
bracis-28433	206	4	connected	connect	VERB
bracis-28433	206	5	layer	layer	NOUN
bracis-28433	206	6	is	be	AUX
bracis-28433	206	7	then	then	ADV
bracis-28433	206	8	used	use	VERB
bracis-28433	206	9	to	to	PART
bracis-28433	206	10	reduce	reduce	VERB
bracis-28433	206	11	the	the	DET
bracis-28433	206	12	encoded	encode	VERB
bracis-28433	206	13	vector	vector	NOUN
bracis-28433	206	14	’s	’s	PART
bracis-28433	206	15	dimension	dimension	NOUN
bracis-28433	206	16	to	to	ADP
bracis-28433	206	17	the	the	DET
bracis-28433	206	18	number	number	NOUN
bracis-28433	206	19	of	of	ADP
bracis-28433	206	20	possible	possible	ADJ
bracis-28433	206	21	tags	tag	NOUN
bracis-28433	206	22	.	.	PUNCT
bracis-28433	207	1	the	the	DET
bracis-28433	207	2	reduced	reduced	ADJ
bracis-28433	207	3	dimension	dimension	NOUN
bracis-28433	207	4	vector	vector	NOUN
bracis-28433	207	5	is	be	AUX
bracis-28433	207	6	then	then	ADV
bracis-28433	207	7	fed	feed	VERB
bracis-28433	207	8	to	to	ADP
bracis-28433	207	9	a	a	DET
bracis-28433	207	10	crf	crf	NOUN
bracis-28433	207	11	layer	layer	NOUN
bracis-28433	207	12	.	.	PUNCT
bracis-28433	208	1	dropout	dropout	NOUN
bracis-28433	208	2	layers	layer	NOUN
bracis-28433	208	3	are	be	AUX
bracis-28433	208	4	used	use	VERB
bracis-28433	208	5	before	before	ADP
bracis-28433	208	6	and	and	CCONJ
bracis-28433	208	7	after	after	ADP
bracis-28433	208	8	the	the	DET
bracis-28433	208	9	bilstm	bilstm	NOUN
bracis-28433	208	10	layer	layer	NOUN
bracis-28433	208	11	.	.	PUNCT
bracis-28433	209	1	all	all	DET
bracis-28433	209	2	weight	weight	NOUN
bracis-28433	209	3	matrices	matrix	NOUN
bracis-28433	209	4	for	for	ADP
bracis-28433	209	5	the	the	DET
bracis-28433	209	6	bilstm	bilstm	NOUN
bracis-28433	209	7	and	and	CCONJ
bracis-28433	209	8	fully	fully	ADV
bracis-28433	209	9	-	-	PUNCT
bracis-28433	209	10	connected	connect	VERB
bracis-28433	209	11	layers	layer	NOUN
bracis-28433	209	12	are	be	AUX
bracis-28433	209	13	randomly	randomly	ADV
bracis-28433	209	14	initialized	initialize	VERB
bracis-28433	209	15	using	use	VERB
bracis-28433	209	16	a	a	DET
bracis-28433	209	17	uniform	uniform	ADJ
bracis-28433	209	18	distribution	distribution	NOUN
bracis-28433	209	19	to	to	PART
bracis-28433	209	20	select	select	VERB
bracis-28433	209	21	samples	sample	NOUN
bracis-28433	209	22	from	from	ADP
bracis-28433	209	23	\([-\sqrt{\frac{6}{r+c	\([-\sqrt{\frac{6}{r+c	ADJ
bracis-28433	209	24	}	}	PUNCT
bracis-28433	209	25	}	}	PUNCT
bracis-28433	209	26	,	,	PUNCT
bracis-28433	209	27	+	+	NOUN
bracis-28433	209	28	\sqrt{\frac{6}{r+c}}\	\sqrt{\frac{6}{r+c}}\	NOUN
bracis-28433	209	29	)	)	PUNCT
bracis-28433	209	30	as	as	SCONJ
bracis-28433	209	31	proposed	propose	VERB
bracis-28433	209	32	by	by	ADP
bracis-28433	209	33	ma	ma	PROPN
bracis-28433	209	34	and	and	CCONJ
bracis-28433	209	35	hovy	hovy	VERB
bracis-28433	210	1	[	[	X
bracis-28433	210	2	9	9	NUM
bracis-28433	210	3	]	]	PUNCT
bracis-28433	210	4	,	,	PUNCT
bracis-28433	210	5	where	where	SCONJ
bracis-28433	210	6	r	r	NOUN
bracis-28433	210	7	and	and	CCONJ
bracis-28433	210	8	c	c	NOUN
bracis-28433	210	9	are	be	AUX
bracis-28433	210	10	the	the	DET
bracis-28433	210	11	numbers	number	NOUN
bracis-28433	210	12	of	of	ADP
bracis-28433	210	13	rows	row	NOUN
bracis-28433	210	14	and	and	CCONJ
bracis-28433	210	15	columns	column	NOUN
bracis-28433	210	16	in	in	ADP
bracis-28433	210	17	the	the	DET
bracis-28433	210	18	weight	weight	NOUN
bracis-28433	210	19	matrix	matrix	NOUN
bracis-28433	210	20	.	.	PUNCT
bracis-28433	211	1	we	we	PRON
bracis-28433	211	2	initialize	initialize	VERB
bracis-28433	211	3	the	the	DET
bracis-28433	211	4	bias	bias	NOUN
bracis-28433	211	5	parameters	parameter	NOUN
bracis-28433	211	6	from	from	ADP
bracis-28433	211	7	the	the	DET
bracis-28433	211	8	bilstm	bilstm	NOUN
bracis-28433	211	9	layer	layer	NOUN
bracis-28433	211	10	to	to	ADP
bracis-28433	211	11	0.0	0.0	NUM
bracis-28433	211	12	and	and	CCONJ
bracis-28433	211	13	the	the	DET
bracis-28433	211	14	forget	forget	PROPN
bracis-28433	211	15	gate	gate	PROPN
bracis-28433	211	16	bias	bias	NOUN
bracis-28433	211	17	to	to	ADP
bracis-28433	211	18	1.0	1.0	NUM
bracis-28433	211	19	.	.	PUNCT
bracis-28433	212	1	the	the	DET
bracis-28433	212	2	model	model	NOUN
bracis-28433	212	3	’s	’s	PART
bracis-28433	212	4	hyperparameters	hyperparameter	NOUN
bracis-28433	212	5	for	for	ADP
bracis-28433	212	6	the	the	DET
bracis-28433	212	7	english	english	ADJ
bracis-28433	212	8	ner	ner	NOUN
bracis-28433	212	9	datasets	dataset	NOUN
bracis-28433	212	10	are	be	AUX
bracis-28433	212	11	similar	similar	ADJ
bracis-28433	212	12	to	to	ADP
bracis-28433	212	13	those	those	PRON
bracis-28433	212	14	presented	present	VERB
bracis-28433	212	15	in	in	ADP
bracis-28433	212	16	the	the	DET
bracis-28433	212	17	experiments	experiment	NOUN
bracis-28433	212	18	of	of	ADP
bracis-28433	212	19	siddhant	siddhant	NOUN
bracis-28433	212	20	and	and	CCONJ
bracis-28433	212	21	lipton	lipton	PROPN
bracis-28433	213	1	[	[	X
bracis-28433	213	2	19	19	NUM
bracis-28433	213	3	]	]	PUNCT
bracis-28433	213	4	.	.	PUNCT
bracis-28433	214	1	a	a	DET
bracis-28433	214	2	grid	grid	NOUN
bracis-28433	214	3	search	search	NOUN
bracis-28433	214	4	was	be	AUX
bracis-28433	214	5	employed	employ	VERB
bracis-28433	214	6	for	for	ADP
bracis-28433	214	7	the	the	DET
bracis-28433	214	8	portuguese	portuguese	PROPN
bracis-28433	214	9	dataset	dataset	VERB
bracis-28433	214	10	to	to	PART
bracis-28433	214	11	define	define	VERB
bracis-28433	214	12	the	the	DET
bracis-28433	214	13	model	model	NOUN
bracis-28433	214	14	’s	’s	PART
bracis-28433	214	15	hyperparameters	hyperparameter	NOUN
bracis-28433	214	16	.	.	PUNCT
bracis-28433	215	1	a	a	DET
bracis-28433	215	2	thorough	thorough	ADJ
bracis-28433	215	3	explanation	explanation	NOUN
bracis-28433	215	4	of	of	ADP
bracis-28433	215	5	the	the	DET
bracis-28433	215	6	hyperparameter	hyperparameter	NOUN
bracis-28433	215	7	tuning	tune	VERB
bracis-28433	215	8	routine	routine	ADJ
bracis-28433	215	9	and	and	CCONJ
bracis-28433	215	10	hyperparameter	hyperparameter	NOUN
bracis-28433	215	11	values	value	NOUN
bracis-28433	215	12	used	use	VERB
bracis-28433	215	13	in	in	ADP
bracis-28433	215	14	our	our	PRON
bracis-28433	215	15	experiments	experiment	NOUN
bracis-28433	215	16	is	be	AUX
bracis-28433	215	17	presented	present	VERB
bracis-28433	215	18	in	in	ADP
bracis-28433	215	19	[	[	X
bracis-28433	215	20	1	1	NUM
bracis-28433	215	21	]	]	PUNCT
bracis-28433	215	22	.	.	PUNCT
bracis-28433	216	1	7	7	NUM
bracis-28433	216	2	results	result	NOUN
bracis-28433	216	3	we	we	PRON
bracis-28433	216	4	present	present	VERB
bracis-28433	216	5	and	and	CCONJ
bracis-28433	216	6	discuss	discuss	VERB
bracis-28433	216	7	the	the	DET
bracis-28433	216	8	experimental	experimental	ADJ
bracis-28433	216	9	results	result	NOUN
bracis-28433	216	10	focusing	focus	VERB
bracis-28433	216	11	on	on	ADP
bracis-28433	216	12	the	the	DET
bracis-28433	216	13	performance	performance	NOUN
bracis-28433	216	14	overview	overview	NOUN
bracis-28433	216	15	(	(	PUNCT
bracis-28433	216	16	f1	f1	NOUN
bracis-28433	216	17	-	-	PUNCT
bracis-28433	216	18	score	score	NOUN
bracis-28433	216	19	)	)	PUNCT
bracis-28433	216	20	of	of	ADP
bracis-28433	216	21	the	the	DET
bracis-28433	216	22	models	model	NOUN
bracis-28433	216	23	trained	train	VERB
bracis-28433	216	24	on	on	ADP
bracis-28433	216	25	each	each	PRON
bracis-28433	216	26	of	of	ADP
bracis-28433	216	27	the	the	DET
bracis-28433	216	28	three	three	NUM
bracis-28433	216	29	datasets	dataset	NOUN
bracis-28433	216	30	(	(	PUNCT
bracis-28433	216	31	see	see	VERB
bracis-28433	216	32	table	table	NOUN
bracis-28433	216	33	 	 	SPACE
bracis-28433	216	34	1	1	NUM
bracis-28433	216	35	)	)	PUNCT
bracis-28433	216	36	.	.	PUNCT
bracis-28433	217	1	additionally	additionally	ADV
bracis-28433	217	2	,	,	PUNCT
bracis-28433	217	3	we	we	PRON
bracis-28433	217	4	compare	compare	VERB
bracis-28433	217	5	the	the	DET
bracis-28433	217	6	performance	performance	NOUN
bracis-28433	217	7	of	of	ADP
bracis-28433	217	8	neural	neural	ADJ
bracis-28433	217	9	models	model	NOUN
bracis-28433	217	10	trained	train	VERB
bracis-28433	217	11	using	use	VERB
bracis-28433	217	12	various	various	ADJ
bracis-28433	217	13	active	active	ADJ
bracis-28433	217	14	and	and	CCONJ
bracis-28433	217	15	active	active	ADJ
bracis-28433	217	16	self	self	NOUN
bracis-28433	217	17	-	-	PUNCT
bracis-28433	217	18	learning	learn	VERB
bracis-28433	217	19	algorithms	algorithm	NOUN
bracis-28433	217	20	.	.	PUNCT
bracis-28433	218	1	in	in	ADP
bracis-28433	218	2	separate	separate	ADJ
bracis-28433	218	3	sections	section	NOUN
bracis-28433	218	4	,	,	PUNCT
bracis-28433	218	5	we	we	PRON
bracis-28433	218	6	present	present	VERB
bracis-28433	218	7	the	the	DET
bracis-28433	218	8	results	result	NOUN
bracis-28433	218	9	of	of	ADP
bracis-28433	218	10	sentence	sentence	NOUN
bracis-28433	218	11	-	-	PUNCT
bracis-28433	218	12	level	level	NOUN
bracis-28433	218	13	and	and	CCONJ
bracis-28433	218	14	token	token	ADJ
bracis-28433	218	15	-	-	PUNCT
bracis-28433	218	16	level	level	NOUN
bracis-28433	218	17	querying	query	VERB
bracis-28433	218	18	strategies	strategy	NOUN
bracis-28433	218	19	.	.	PUNCT
bracis-28433	219	1	7.1	7.1	NUM
bracis-28433	219	2	results	result	NOUN
bracis-28433	219	3	of	of	ADP
bracis-28433	219	4	 	 	SPACE
bracis-28433	219	5	sentence	sentence	NOUN
bracis-28433	219	6	-	-	PUNCT
bracis-28433	219	7	level	level	NOUN
bracis-28433	219	8	strategy	strategy	NOUN
bracis-28433	219	9	the	the	DET
bracis-28433	219	10	asl	asl	PROPN
bracis-28433	219	11	algorithm	algorithm	NOUN
bracis-28433	219	12	that	that	PRON
bracis-28433	219	13	uses	use	VERB
bracis-28433	219	14	pseudo	pseudo	NOUN
bracis-28433	219	15	-	-	NOUN
bracis-28433	219	16	labeling	labeling	NOUN
bracis-28433	219	17	as	as	ADP
bracis-28433	219	18	the	the	DET
bracis-28433	219	19	default	default	NOUN
bracis-28433	219	20	self	self	NOUN
bracis-28433	219	21	-	-	PUNCT
bracis-28433	219	22	learning	learn	VERB
bracis-28433	219	23	strategy	strategy	NOUN
bracis-28433	219	24	as	as	SCONJ
bracis-28433	219	25	proposed	propose	VERB
bracis-28433	219	26	by	by	ADP
bracis-28433	219	27	[	[	X
bracis-28433	219	28	2	2	NUM
bracis-28433	219	29	]	]	PUNCT
bracis-28433	219	30	,	,	PUNCT
bracis-28433	219	31	is	be	AUX
bracis-28433	219	32	referred	refer	VERB
bracis-28433	219	33	to	to	ADP
bracis-28433	219	34	as	as	ADV
bracis-28433	219	35	dasl	dasl	PROPN
bracis-28433	219	36	(	(	PUNCT
bracis-28433	219	37	deep	deep	ADJ
bracis-28433	219	38	active	active	ADJ
bracis-28433	219	39	self	self	NOUN
bracis-28433	219	40	-	-	PUNCT
bracis-28433	219	41	learning	learning	NOUN
bracis-28433	219	42	)	)	PUNCT
bracis-28433	219	43	.	.	PUNCT
bracis-28433	220	1	to	to	PART
bracis-28433	220	2	conduct	conduct	VERB
bracis-28433	220	3	an	an	DET
bracis-28433	220	4	almost	almost	ADV
bracis-28433	220	5	comprehensive	comprehensive	ADJ
bracis-28433	220	6	experiment	experiment	NOUN
bracis-28433	220	7	with	with	ADP
bracis-28433	220	8	various	various	ADJ
bracis-28433	220	9	self	self	NOUN
bracis-28433	220	10	-	-	PUNCT
bracis-28433	220	11	learning	learn	VERB
bracis-28433	220	12	methods	method	NOUN
bracis-28433	220	13	,	,	PUNCT
bracis-28433	220	14	we	we	PRON
bracis-28433	220	15	replace	replace	VERB
bracis-28433	220	16	pseudo	pseudo	NOUN
bracis-28433	220	17	-	-	NOUN
bracis-28433	220	18	labeling	labeling	NOUN
bracis-28433	220	19	with	with	ADP
bracis-28433	220	20	three	three	NUM
bracis-28433	220	21	other	other	ADJ
bracis-28433	220	22	techniques	technique	NOUN
bracis-28433	220	23	:	:	PUNCT
bracis-28433	220	24	dasl_wd	dasl_wd	PROPN
bracis-28433	220	25	(	(	PUNCT
bracis-28433	220	26	deep	deep	ADV
bracis-28433	220	27	active	active	ADJ
bracis-28433	220	28	self	self	NOUN
bracis-28433	220	29	-	-	PUNCT
bracis-28433	220	30	learning	learning	NOUN
bracis-28433	220	31	with	with	ADP
bracis-28433	220	32	word	word	NOUN
bracis-28433	220	33	dropout	dropout	NOUN
bracis-28433	220	34	)	)	PUNCT
bracis-28433	220	35	,	,	PUNCT
bracis-28433	220	36	dasl_vat	dasl_vat	NOUN
bracis-28433	220	37	(	(	PUNCT
bracis-28433	220	38	deep	deep	ADJ
bracis-28433	220	39	active	active	ADJ
bracis-28433	220	40	self	self	NOUN
bracis-28433	220	41	-	-	PUNCT
bracis-28433	220	42	learning	learning	NOUN
bracis-28433	220	43	with	with	ADP
bracis-28433	220	44	virtual	virtual	ADJ
bracis-28433	220	45	adversarial	adversarial	ADJ
bracis-28433	220	46	technique	technique	NOUN
bracis-28433	220	47	)	)	PUNCT
bracis-28433	220	48	,	,	PUNCT
bracis-28433	220	49	and	and	CCONJ
bracis-28433	220	50	dasl_cvt	dasl_cvt	NUM
bracis-28433	220	51	(	(	PUNCT
bracis-28433	220	52	deep	deep	ADJ
bracis-28433	220	53	active	active	ADJ
bracis-28433	220	54	self	self	NOUN
bracis-28433	220	55	-	-	PUNCT
bracis-28433	220	56	learning	learning	NOUN
bracis-28433	220	57	with	with	ADP
bracis-28433	220	58	cross	cross	ADJ
bracis-28433	220	59	-	-	ADJ
bracis-28433	220	60	view	view	ADJ
bracis-28433	220	61	training	training	NOUN
bracis-28433	220	62	)	)	PUNCT
bracis-28433	220	63	.	.	PUNCT
bracis-28433	221	1	we	we	PRON
bracis-28433	221	2	also	also	ADV
bracis-28433	221	3	compare	compare	VERB
bracis-28433	221	4	the	the	DET
bracis-28433	221	5	results	result	NOUN
bracis-28433	221	6	of	of	ADP
bracis-28433	221	7	active	active	ADJ
bracis-28433	221	8	self	self	NOUN
bracis-28433	221	9	-	-	PUNCT
bracis-28433	221	10	learning	learn	VERB
bracis-28433	221	11	strategies	strategy	NOUN
bracis-28433	221	12	with	with	ADP
bracis-28433	221	13	deep	deep	ADJ
bracis-28433	221	14	active	active	ADJ
bracis-28433	221	15	learning	learning	NOUN
bracis-28433	221	16	(	(	PUNCT
bracis-28433	221	17	dal	dal	NOUN
bracis-28433	221	18	)	)	PUNCT
bracis-28433	221	19	.	.	PUNCT
bracis-28433	222	1	in	in	ADP
bracis-28433	222	2	fig	fig	NOUN
bracis-28433	222	3	.	.	PUNCT
bracis-28433	222	4	 	 	SPACE
bracis-28433	222	5	1	1	NUM
bracis-28433	222	6	,	,	PUNCT
bracis-28433	222	7	supervised	supervise	VERB
bracis-28433	222	8	(	(	PUNCT
bracis-28433	222	9	i.e.	i.e.	X
bracis-28433	222	10	dashed	dash	VERB
bracis-28433	222	11	line	line	NOUN
bracis-28433	222	12	)	)	PUNCT
bracis-28433	222	13	represents	represent	VERB
bracis-28433	222	14	the	the	DET
bracis-28433	222	15	best	good	ADJ
bracis-28433	222	16	performance	performance	NOUN
bracis-28433	222	17	of	of	ADP
bracis-28433	222	18	the	the	DET
bracis-28433	222	19	neural	neural	ADJ
bracis-28433	222	20	models	model	NOUN
bracis-28433	222	21	when	when	SCONJ
bracis-28433	222	22	trained	train	VERB
bracis-28433	222	23	using	use	VERB
bracis-28433	222	24	the	the	DET
bracis-28433	222	25	entire	entire	ADJ
bracis-28433	222	26	training	training	NOUN
bracis-28433	222	27	set	set	VERB
bracis-28433	222	28	with	with	ADP
bracis-28433	222	29	true	true	ADJ
bracis-28433	222	30	labels	label	NOUN
bracis-28433	222	31	in	in	ADP
bracis-28433	222	32	a	a	DET
bracis-28433	222	33	supervised	supervised	ADJ
bracis-28433	222	34	fashion	fashion	NOUN
bracis-28433	222	35	.	.	PUNCT
bracis-28433	223	1	dal	dal	NOUN
bracis-28433	223	2	represents	represent	VERB
bracis-28433	223	3	the	the	DET
bracis-28433	223	4	baseline	baseline	ADJ
bracis-28433	223	5	active	active	ADJ
bracis-28433	223	6	learning	learning	NOUN
bracis-28433	223	7	algorithm	algorithm	NOUN
bracis-28433	223	8	by	by	ADP
bracis-28433	223	9	shen	shen	PROPN
bracis-28433	223	10	et	et	PROPN
bracis-28433	223	11	al	al	PROPN
bracis-28433	223	12	.	.	PUNCT
bracis-28433	224	1	[	[	X
bracis-28433	224	2	18	18	NUM
bracis-28433	224	3	]	]	PUNCT
bracis-28433	224	4	.	.	PUNCT
bracis-28433	225	1	figure	figure	NOUN
bracis-28433	225	2	 	 	SPACE
bracis-28433	225	3	3	3	NUM
bracis-28433	225	4	presents	present	VERB
bracis-28433	225	5	the	the	DET
bracis-28433	225	6	performance	performance	NOUN
bracis-28433	225	7	(	(	PUNCT
bracis-28433	225	8	i.e.	i.e.	X
bracis-28433	225	9	,	,	PUNCT
bracis-28433	225	10	f1	f1	ADJ
bracis-28433	225	11	-	-	PUNCT
bracis-28433	225	12	score	score	NOUN
bracis-28433	225	13	)	)	PUNCT
bracis-28433	225	14	achieved	achieve	VERB
bracis-28433	225	15	by	by	ADP
bracis-28433	225	16	the	the	DET
bracis-28433	225	17	models	model	NOUN
bracis-28433	225	18	and	and	CCONJ
bracis-28433	225	19	the	the	DET
bracis-28433	225	20	percentage	percentage	NOUN
bracis-28433	225	21	of	of	ADP
bracis-28433	225	22	oracle	oracle	NOUN
bracis-28433	225	23	annotated	annotate	VERB
bracis-28433	225	24	tokens	token	NOUN
bracis-28433	225	25	(	(	PUNCT
bracis-28433	225	26	human	human	ADJ
bracis-28433	225	27	labeled	label	VERB
bracis-28433	225	28	tokens	token	NOUN
bracis-28433	225	29	)	)	PUNCT
bracis-28433	225	30	.	.	PUNCT
bracis-28433	226	1	it	it	PRON
bracis-28433	226	2	should	should	AUX
bracis-28433	226	3	be	be	AUX
bracis-28433	226	4	noted	note	VERB
bracis-28433	226	5	that	that	SCONJ
bracis-28433	226	6	active	active	ADJ
bracis-28433	226	7	self	self	NOUN
bracis-28433	226	8	-	-	PUNCT
bracis-28433	226	9	learning	learn	VERB
bracis-28433	226	10	algorithms	algorithm	NOUN
bracis-28433	226	11	use	use	VERB
bracis-28433	226	12	additional	additional	ADJ
bracis-28433	226	13	data	datum	NOUN
bracis-28433	226	14	for	for	ADP
bracis-28433	226	15	training	training	NOUN
bracis-28433	226	16	in	in	ADP
bracis-28433	226	17	addition	addition	NOUN
bracis-28433	226	18	to	to	ADP
bracis-28433	226	19	the	the	DET
bracis-28433	226	20	hand	hand	NOUN
bracis-28433	226	21	-	-	PUNCT
bracis-28433	226	22	annotated	annotate	VERB
bracis-28433	226	23	data	datum	NOUN
bracis-28433	226	24	.	.	PUNCT
bracis-28433	227	1	note	note	VERB
bracis-28433	227	2	that	that	SCONJ
bracis-28433	227	3	the	the	DET
bracis-28433	227	4	dasl_cvt	dasl_cvt	NUM
bracis-28433	227	5	technique	technique	NOUN
bracis-28433	227	6	is	be	AUX
bracis-28433	227	7	only	only	ADV
bracis-28433	227	8	implemented	implement	VERB
bracis-28433	227	9	with	with	ADP
bracis-28433	227	10	the	the	DET
bracis-28433	227	11	cnn	cnn	PROPN
bracis-28433	227	12	-	-	PUNCT
bracis-28433	227	13	bilstm	bilstm	NOUN
bracis-28433	227	14	-	-	PUNCT
bracis-28433	227	15	crf	crf	NOUN
bracis-28433	227	16	model	model	NOUN
bracis-28433	227	17	.	.	PUNCT
bracis-28433	228	1	this	this	DET
bracis-28433	228	2	technique	technique	NOUN
bracis-28433	228	3	required	require	VERB
bracis-28433	228	4	multiple	multiple	ADJ
bracis-28433	228	5	views	view	NOUN
bracis-28433	228	6	of	of	ADP
bracis-28433	228	7	the	the	DET
bracis-28433	228	8	input	input	NOUN
bracis-28433	228	9	,	,	PUNCT
bracis-28433	228	10	which	which	PRON
bracis-28433	228	11	is	be	AUX
bracis-28433	228	12	more	more	ADV
bracis-28433	228	13	computationally	computationally	ADV
bracis-28433	228	14	costly	costly	ADJ
bracis-28433	228	15	to	to	PART
bracis-28433	228	16	implement	implement	VERB
bracis-28433	228	17	in	in	ADP
bracis-28433	228	18	neural	neural	ADJ
bracis-28433	228	19	models	model	NOUN
bracis-28433	228	20	with	with	ADP
bracis-28433	228	21	cnn	cnn	PROPN
bracis-28433	228	22	-	-	PUNCT
bracis-28433	228	23	only	only	ADV
bracis-28433	228	24	encoders	encoder	NOUN
bracis-28433	228	25	such	such	ADJ
bracis-28433	228	26	as	as	ADP
bracis-28433	228	27	the	the	DET
bracis-28433	228	28	cnn	cnn	PROPN
bracis-28433	228	29	-	-	PUNCT
bracis-28433	228	30	cnn	cnn	PROPN
bracis-28433	228	31	-	-	PUNCT
bracis-28433	228	32	lstm	lstm	PROPN
bracis-28433	228	33	.	.	PUNCT
bracis-28433	229	1	from	from	ADP
bracis-28433	229	2	the	the	DET
bracis-28433	229	3	experiments	experiment	NOUN
bracis-28433	229	4	performed	perform	VERB
bracis-28433	229	5	in	in	ADP
bracis-28433	229	6	fig	fig	NOUN
bracis-28433	229	7	.	.	PUNCT
bracis-28433	229	8	 	 	SPACE
bracis-28433	229	9	3	3	NUM
bracis-28433	229	10	,	,	PUNCT
bracis-28433	229	11	we	we	PRON
bracis-28433	229	12	noticed	notice	VERB
bracis-28433	229	13	that	that	SCONJ
bracis-28433	229	14	no	no	DET
bracis-28433	229	15	technique	technique	NOUN
bracis-28433	229	16	performed	perform	VERB
bracis-28433	229	17	consistently	consistently	ADV
bracis-28433	229	18	better	well	ADV
bracis-28433	229	19	than	than	ADP
bracis-28433	229	20	the	the	DET
bracis-28433	229	21	others	other	NOUN
bracis-28433	229	22	.	.	PUNCT
bracis-28433	230	1	however	however	ADV
bracis-28433	230	2	,	,	PUNCT
bracis-28433	230	3	for	for	ADP
bracis-28433	230	4	the	the	DET
bracis-28433	230	5	aposentadoria	aposentadoria	PROPN
bracis-28433	230	6	dataset	dataset	NOUN
bracis-28433	230	7	,	,	PUNCT
bracis-28433	230	8	the	the	DET
bracis-28433	230	9	dasl_vat	dasl_vat	NOUN
bracis-28433	230	10	and	and	CCONJ
bracis-28433	230	11	dasv_cvt	dasv_cvt	VERB
bracis-28433	230	12	techniques	technique	NOUN
bracis-28433	230	13	help	help	VERB
bracis-28433	230	14	the	the	DET
bracis-28433	230	15	model	model	NOUN
bracis-28433	230	16	achieve	achieve	VERB
bracis-28433	230	17	its	its	PRON
bracis-28433	230	18	best	good	ADJ
bracis-28433	230	19	performance	performance	NOUN
bracis-28433	230	20	faster	fast	ADV
bracis-28433	230	21	than	than	ADP
bracis-28433	230	22	the	the	DET
bracis-28433	230	23	baseline	baseline	NOUN
bracis-28433	230	24	algorithms	algorithm	NOUN
bracis-28433	230	25	.	.	PUNCT
bracis-28433	231	1	we	we	PRON
bracis-28433	231	2	believe	believe	VERB
bracis-28433	231	3	this	this	DET
bracis-28433	231	4	behavior	behavior	NOUN
bracis-28433	231	5	is	be	AUX
bracis-28433	231	6	because	because	SCONJ
bracis-28433	231	7	this	this	DET
bracis-28433	231	8	dataset	dataset	NOUN
bracis-28433	231	9	is	be	AUX
bracis-28433	231	10	less	less	ADV
bracis-28433	231	11	imbalanced	imbalanced	ADJ
bracis-28433	231	12	than	than	ADP
bracis-28433	231	13	the	the	DET
bracis-28433	231	14	other	other	ADJ
bracis-28433	231	15	datasets	dataset	NOUN
bracis-28433	231	16	,	,	PUNCT
bracis-28433	231	17	meaning	mean	VERB
bracis-28433	231	18	that	that	SCONJ
bracis-28433	231	19	most	most	ADJ
bracis-28433	231	20	ner	ner	NOUN
bracis-28433	231	21	entities	entity	NOUN
bracis-28433	231	22	are	be	AUX
bracis-28433	231	23	present	present	ADJ
bracis-28433	231	24	in	in	ADP
bracis-28433	231	25	most	most	ADJ
bracis-28433	231	26	sentences	sentence	NOUN
bracis-28433	231	27	.	.	PUNCT
bracis-28433	232	1	as	as	ADP
bracis-28433	232	2	a	a	DET
bracis-28433	232	3	result	result	NOUN
bracis-28433	232	4	,	,	PUNCT
bracis-28433	232	5	the	the	DET
bracis-28433	232	6	model	model	NOUN
bracis-28433	232	7	generalizes	generalize	VERB
bracis-28433	232	8	faster	fast	ADV
bracis-28433	232	9	with	with	ADP
bracis-28433	232	10	less	less	ADJ
bracis-28433	232	11	hand	hand	NOUN
bracis-28433	232	12	-	-	PUNCT
bracis-28433	232	13	annotated	annotate	VERB
bracis-28433	232	14	data	datum	NOUN
bracis-28433	232	15	and	and	CCONJ
bracis-28433	232	16	more	more	ADV
bracis-28433	232	17	effectively	effectively	ADV
bracis-28433	232	18	utilizes	utilize	VERB
bracis-28433	232	19	unlabeled	unlabeled	ADJ
bracis-28433	232	20	data	datum	NOUN
bracis-28433	232	21	through	through	ADP
bracis-28433	232	22	self	self	NOUN
bracis-28433	232	23	-	-	PUNCT
bracis-28433	232	24	training	training	NOUN
bracis-28433	232	25	in	in	ADP
bracis-28433	232	26	the	the	DET
bracis-28433	232	27	early	early	ADJ
bracis-28433	232	28	iterations	iteration	NOUN
bracis-28433	232	29	of	of	ADP
bracis-28433	232	30	the	the	DET
bracis-28433	232	31	active	active	ADJ
bracis-28433	232	32	self	self	NOUN
bracis-28433	232	33	-	-	PUNCT
bracis-28433	232	34	learning	learn	VERB
bracis-28433	232	35	algorithms	algorithm	NOUN
bracis-28433	232	36	.	.	PUNCT
bracis-28433	233	1	fig	fig	NOUN
bracis-28433	233	2	.	.	PUNCT
bracis-28433	234	1	3	3	X
bracis-28433	234	2	.	.	X
bracis-28433	234	3	the	the	DET
bracis-28433	234	4	graph	graph	NOUN
bracis-28433	234	5	shows	show	VERB
bracis-28433	234	6	the	the	DET
bracis-28433	234	7	performance	performance	NOUN
bracis-28433	234	8	(	(	PUNCT
bracis-28433	234	9	f1	f1	NOUN
bracis-28433	234	10	-	-	PUNCT
bracis-28433	234	11	score	score	NOUN
bracis-28433	234	12	)	)	PUNCT
bracis-28433	234	13	of	of	ADP
bracis-28433	234	14	the	the	DET
bracis-28433	234	15	neural	neural	ADJ
bracis-28433	234	16	models	model	NOUN
bracis-28433	234	17	as	as	ADP
bracis-28433	234	18	a	a	DET
bracis-28433	234	19	function	function	NOUN
bracis-28433	234	20	of	of	ADP
bracis-28433	234	21	the	the	DET
bracis-28433	234	22	percentage	percentage	NOUN
bracis-28433	234	23	of	of	ADP
bracis-28433	234	24	hand	hand	NOUN
bracis-28433	234	25	-	-	PUNCT
bracis-28433	234	26	annotated	annotate	VERB
bracis-28433	234	27	tokens	token	NOUN
bracis-28433	234	28	in	in	ADP
bracis-28433	234	29	each	each	DET
bracis-28433	234	30	iteration	iteration	NOUN
bracis-28433	234	31	of	of	ADP
bracis-28433	234	32	the	the	DET
bracis-28433	234	33	active	active	ADJ
bracis-28433	234	34	and	and	CCONJ
bracis-28433	234	35	active	active	ADJ
bracis-28433	234	36	self	self	NOUN
bracis-28433	234	37	-	-	PUNCT
bracis-28433	234	38	learning	learn	VERB
bracis-28433	234	39	algorithms	algorithm	NOUN
bracis-28433	234	40	.	.	PUNCT
bracis-28433	235	1	full	full	ADJ
bracis-28433	235	2	size	size	NOUN
bracis-28433	235	3	image	image	NOUN
bracis-28433	235	4	7.2	7.2	NUM
bracis-28433	235	5	results	result	NOUN
bracis-28433	235	6	of	of	ADP
bracis-28433	235	7	 	 	SPACE
bracis-28433	235	8	token	token	ADJ
bracis-28433	235	9	-	-	PUNCT
bracis-28433	235	10	level	level	NOUN
bracis-28433	235	11	strategy	strategy	NOUN
bracis-28433	235	12	in	in	ADP
bracis-28433	235	13	this	this	DET
bracis-28433	235	14	section	section	NOUN
bracis-28433	235	15	,	,	PUNCT
bracis-28433	235	16	we	we	PRON
bracis-28433	235	17	present	present	VERB
bracis-28433	235	18	the	the	DET
bracis-28433	235	19	results	result	NOUN
bracis-28433	235	20	of	of	ADP
bracis-28433	235	21	the	the	DET
bracis-28433	235	22	token	token	ADJ
bracis-28433	235	23	-	-	PUNCT
bracis-28433	235	24	level	level	NOUN
bracis-28433	235	25	active	active	ADJ
bracis-28433	235	26	self	self	NOUN
bracis-28433	235	27	-	-	PUNCT
bracis-28433	235	28	learning	learn	VERB
bracis-28433	235	29	algorithm	algorithm	NOUN
bracis-28433	235	30	proposed	propose	VERB
bracis-28433	235	31	.	.	PUNCT
bracis-28433	236	1	our	our	PRON
bracis-28433	236	2	algorithm	algorithm	NOUN
bracis-28433	236	3	,	,	PUNCT
bracis-28433	236	4	namely	namely	ADV
bracis-28433	236	5	mdal	mdal	NOUN
bracis-28433	236	6	,	,	PUNCT
bracis-28433	236	7	is	be	AUX
bracis-28433	236	8	a	a	DET
bracis-28433	236	9	modified	modify	VERB
bracis-28433	236	10	version	version	NOUN
bracis-28433	236	11	of	of	ADP
bracis-28433	236	12	deep	deep	ADJ
bracis-28433	236	13	active	active	ADJ
bracis-28433	236	14	learning	learning	NOUN
bracis-28433	236	15	(	(	PUNCT
bracis-28433	236	16	dal	dal	NOUN
bracis-28433	236	17	)	)	PUNCT
bracis-28433	236	18	.	.	PUNCT
bracis-28433	237	1	figure	figure	NOUN
bracis-28433	237	2	 	 	SPACE
bracis-28433	237	3	4	4	NUM
bracis-28433	237	4	compares	compare	VERB
bracis-28433	237	5	the	the	DET
bracis-28433	237	6	original	original	ADJ
bracis-28433	237	7	and	and	CCONJ
bracis-28433	237	8	the	the	DET
bracis-28433	237	9	modified	modified	ADJ
bracis-28433	237	10	dal	dal	NOUN
bracis-28433	237	11	algorithms	algorithm	NOUN
bracis-28433	237	12	,	,	PUNCT
bracis-28433	237	13	both	both	PRON
bracis-28433	237	14	using	use	VERB
bracis-28433	237	15	the	the	DET
bracis-28433	237	16	validation	validation	NOUN
bracis-28433	237	17	f1	f1	NOUN
bracis-28433	237	18	-	-	PUNCT
bracis-28433	237	19	score	score	NOUN
bracis-28433	237	20	for	for	ADP
bracis-28433	237	21	early	early	ADJ
bracis-28433	237	22	stopping	stopping	NOUN
bracis-28433	237	23	.	.	PUNCT
bracis-28433	238	1	the	the	DET
bracis-28433	238	2	graphs	graph	NOUN
bracis-28433	238	3	show	show	VERB
bracis-28433	238	4	that	that	SCONJ
bracis-28433	238	5	both	both	CCONJ
bracis-28433	238	6	the	the	DET
bracis-28433	238	7	original	original	ADJ
bracis-28433	238	8	and	and	CCONJ
bracis-28433	238	9	the	the	DET
bracis-28433	238	10	modified	modify	VERB
bracis-28433	238	11	dal	dal	NOUN
bracis-28433	238	12	algorithms	algorithm	NOUN
bracis-28433	238	13	achieve	achieve	VERB
bracis-28433	238	14	similar	similar	ADJ
bracis-28433	238	15	performance	performance	NOUN
bracis-28433	238	16	with	with	ADP
bracis-28433	238	17	the	the	DET
bracis-28433	238	18	same	same	ADJ
bracis-28433	238	19	amount	amount	NOUN
bracis-28433	238	20	of	of	ADP
bracis-28433	238	21	labeled	label	VERB
bracis-28433	238	22	data	datum	NOUN
bracis-28433	238	23	but	but	CCONJ
bracis-28433	238	24	with	with	ADP
bracis-28433	238	25	our	our	PRON
bracis-28433	238	26	modified	modify	VERB
bracis-28433	238	27	algorithm	algorithm	NOUN
bracis-28433	238	28	allowing	allow	VERB
bracis-28433	238	29	for	for	ADP
bracis-28433	238	30	tokens	token	NOUN
bracis-28433	238	31	to	to	PART
bracis-28433	238	32	be	be	AUX
bracis-28433	238	33	self	self	NOUN
bracis-28433	238	34	-	-	PUNCT
bracis-28433	238	35	labeled	label	VERB
bracis-28433	238	36	.	.	PUNCT
bracis-28433	239	1	from	from	ADP
bracis-28433	239	2	the	the	DET
bracis-28433	239	3	graphs	graph	NOUN
bracis-28433	239	4	in	in	ADP
bracis-28433	239	5	the	the	DET
bracis-28433	239	6	center	center	NOUN
bracis-28433	239	7	,	,	PUNCT
bracis-28433	239	8	we	we	PRON
bracis-28433	239	9	observe	observe	VERB
bracis-28433	239	10	that	that	SCONJ
bracis-28433	239	11	the	the	DET
bracis-28433	239	12	models	model	NOUN
bracis-28433	239	13	trained	train	VERB
bracis-28433	239	14	with	with	ADP
bracis-28433	239	15	the	the	DET
bracis-28433	239	16	modified	modify	VERB
bracis-28433	239	17	dal	dal	NOUN
bracis-28433	239	18	algorithm	algorithm	NOUN
bracis-28433	239	19	reach	reach	VERB
bracis-28433	239	20	peak	peak	NOUN
bracis-28433	239	21	performance	performance	NOUN
bracis-28433	239	22	with	with	ADP
bracis-28433	239	23	significantly	significantly	ADV
bracis-28433	239	24	less	less	ADJ
bracis-28433	239	25	hand	hand	NOUN
bracis-28433	239	26	annotated	annotated	ADJ
bracis-28433	239	27	tokens	token	NOUN
bracis-28433	239	28	.	.	PUNCT
bracis-28433	240	1	this	this	PRON
bracis-28433	240	2	is	be	AUX
bracis-28433	240	3	justified	justify	VERB
bracis-28433	240	4	by	by	ADP
bracis-28433	240	5	the	the	DET
bracis-28433	240	6	fact	fact	NOUN
bracis-28433	240	7	that	that	SCONJ
bracis-28433	240	8	for	for	ADP
bracis-28433	240	9	the	the	DET
bracis-28433	240	10	modified	modify	VERB
bracis-28433	240	11	algorithm	algorithm	NOUN
bracis-28433	240	12	,	,	PUNCT
bracis-28433	240	13	most	most	ADJ
bracis-28433	240	14	tokens	token	NOUN
bracis-28433	240	15	are	be	AUX
bracis-28433	240	16	self	self	NOUN
bracis-28433	240	17	-	-	PUNCT
bracis-28433	240	18	annotated	annotate	VERB
bracis-28433	240	19	by	by	ADP
bracis-28433	240	20	the	the	DET
bracis-28433	240	21	trained	train	VERB
bracis-28433	240	22	model	model	NOUN
bracis-28433	240	23	reliably	reliably	ADV
bracis-28433	240	24	,	,	PUNCT
bracis-28433	240	25	as	as	SCONJ
bracis-28433	240	26	shown	show	VERB
bracis-28433	240	27	in	in	ADP
bracis-28433	240	28	the	the	DET
bracis-28433	240	29	graphs	graph	NOUN
bracis-28433	240	30	to	to	ADP
bracis-28433	240	31	the	the	DET
bracis-28433	240	32	right	right	NOUN
bracis-28433	240	33	in	in	ADP
bracis-28433	240	34	fig	fig	NOUN
bracis-28433	240	35	.	.	PUNCT
bracis-28433	240	36	 	 	SPACE
bracis-28433	241	1	4	4	NUM
bracis-28433	241	2	.	.	PUNCT
bracis-28433	241	3	table	table	NOUN
bracis-28433	241	4	 	 	SPACE
bracis-28433	241	5	2	2	NUM
bracis-28433	241	6	presents	present	VERB
bracis-28433	241	7	the	the	DET
bracis-28433	241	8	percentage	percentage	NOUN
bracis-28433	241	9	of	of	ADP
bracis-28433	241	10	hand	hand	NOUN
bracis-28433	241	11	-	-	PUNCT
bracis-28433	241	12	annotated	annotate	VERB
bracis-28433	241	13	tokens	token	NOUN
bracis-28433	241	14	required	require	VERB
bracis-28433	241	15	for	for	ADP
bracis-28433	241	16	the	the	DET
bracis-28433	241	17	trained	train	VERB
bracis-28433	241	18	model	model	NOUN
bracis-28433	241	19	to	to	PART
bracis-28433	241	20	reach	reach	VERB
bracis-28433	241	21	99	99	NUM
bracis-28433	241	22	%	%	NOUN
bracis-28433	241	23	of	of	ADP
bracis-28433	241	24	its	its	PRON
bracis-28433	241	25	peak	peak	NOUN
bracis-28433	241	26	f1	f1	NOUN
bracis-28433	241	27	-	-	PUNCT
bracis-28433	241	28	score	score	NOUN
bracis-28433	241	29	performance	performance	NOUN
bracis-28433	241	30	.	.	PUNCT
bracis-28433	242	1	we	we	PRON
bracis-28433	242	2	observe	observe	VERB
bracis-28433	242	3	that	that	SCONJ
bracis-28433	242	4	the	the	DET
bracis-28433	242	5	proposed	propose	VERB
bracis-28433	242	6	method	method	NOUN
bracis-28433	242	7	requires	require	VERB
bracis-28433	242	8	significantly	significantly	ADV
bracis-28433	242	9	fewer	few	ADJ
bracis-28433	242	10	hand	hand	NOUN
bracis-28433	242	11	-	-	PUNCT
bracis-28433	242	12	annotated	annotate	VERB
bracis-28433	242	13	tokens	token	NOUN
bracis-28433	242	14	.	.	PUNCT
bracis-28433	243	1	however	however	ADV
bracis-28433	243	2	,	,	PUNCT
bracis-28433	243	3	we	we	PRON
bracis-28433	243	4	do	do	AUX
bracis-28433	243	5	note	note	VERB
bracis-28433	243	6	that	that	SCONJ
bracis-28433	243	7	these	these	DET
bracis-28433	243	8	results	result	NOUN
bracis-28433	243	9	do	do	AUX
bracis-28433	243	10	not	not	PART
bracis-28433	243	11	consider	consider	VERB
bracis-28433	243	12	that	that	SCONJ
bracis-28433	243	13	many	many	ADJ
bracis-28433	243	14	of	of	ADP
bracis-28433	243	15	the	the	DET
bracis-28433	243	16	tokens	token	NOUN
bracis-28433	243	17	in	in	ADP
bracis-28433	243	18	the	the	DET
bracis-28433	243	19	queried	query	VERB
bracis-28433	243	20	sentences	sentence	NOUN
bracis-28433	243	21	were	be	AUX
bracis-28433	243	22	not	not	PART
bracis-28433	243	23	named	name	VERB
bracis-28433	243	24	entities	entity	NOUN
bracis-28433	243	25	and	and	CCONJ
bracis-28433	243	26	therefore	therefore	ADV
bracis-28433	243	27	were	be	AUX
bracis-28433	243	28	not	not	PART
bracis-28433	243	29	required	require	VERB
bracis-28433	243	30	to	to	PART
bracis-28433	243	31	be	be	AUX
bracis-28433	243	32	annotated	annotate	VERB
bracis-28433	243	33	.	.	PUNCT
bracis-28433	244	1	this	this	PRON
bracis-28433	244	2	is	be	AUX
bracis-28433	244	3	the	the	DET
bracis-28433	244	4	reason	reason	NOUN
bracis-28433	244	5	why	why	SCONJ
bracis-28433	244	6	our	our	PRON
bracis-28433	244	7	algorithm	algorithm	NOUN
bracis-28433	244	8	is	be	AUX
bracis-28433	244	9	capable	capable	ADJ
bracis-28433	244	10	of	of	ADP
bracis-28433	244	11	training	train	VERB
bracis-28433	244	12	a	a	DET
bracis-28433	244	13	model	model	NOUN
bracis-28433	244	14	to	to	PART
bracis-28433	244	15	peak	peak	VERB
bracis-28433	244	16	performance	performance	NOUN
bracis-28433	244	17	with	with	ADP
bracis-28433	244	18	only	only	ADV
bracis-28433	244	19	6.96	6.96	NUM
bracis-28433	244	20	%	%	NOUN
bracis-28433	244	21	of	of	ADP
bracis-28433	244	22	human	human	NOUN
bracis-28433	244	23	-	-	PUNCT
bracis-28433	244	24	labeled	label	VERB
bracis-28433	244	25	tokens	token	NOUN
bracis-28433	244	26	in	in	ADP
bracis-28433	244	27	the	the	DET
bracis-28433	244	28	conll2003	conll2003	PROPN
bracis-28433	244	29	dataset	dataset	VERB
bracis-28433	244	30	when	when	SCONJ
bracis-28433	244	31	compared	compare	VERB
bracis-28433	244	32	to	to	ADP
bracis-28433	244	33	the	the	DET
bracis-28433	244	34	36.20	36.20	NUM
bracis-28433	244	35	%	%	NOUN
bracis-28433	244	36	and	and	CCONJ
bracis-28433	244	37	27	27	NUM
bracis-28433	244	38	%	%	NOUN
bracis-28433	244	39	of	of	ADP
bracis-28433	244	40	the	the	DET
bracis-28433	244	41	baseline	baseline	ADJ
bracis-28433	244	42	algorithm	algorithm	NOUN
bracis-28433	244	43	.	.	PUNCT
bracis-28433	245	1	however	however	ADV
bracis-28433	245	2	,	,	PUNCT
bracis-28433	245	3	for	for	ADP
bracis-28433	245	4	fine	fine	ADV
bracis-28433	245	5	-	-	PUNCT
bracis-28433	245	6	grained	grain	VERB
bracis-28433	245	7	ner	ner	NOUN
bracis-28433	245	8	datasets	dataset	NOUN
bracis-28433	245	9	,	,	PUNCT
bracis-28433	245	10	such	such	ADJ
bracis-28433	245	11	as	as	ADP
bracis-28433	245	12	the	the	DET
bracis-28433	245	13	ontonotes5.0	ontonotes5.0	NOUN
bracis-28433	245	14	,	,	PUNCT
bracis-28433	245	15	and	and	CCONJ
bracis-28433	245	16	datasets	dataset	NOUN
bracis-28433	245	17	with	with	ADP
bracis-28433	245	18	few	few	ADJ
bracis-28433	245	19	tokens	token	NOUN
bracis-28433	245	20	that	that	PRON
bracis-28433	245	21	do	do	AUX
bracis-28433	245	22	not	not	PART
bracis-28433	245	23	have	have	VERB
bracis-28433	245	24	a	a	DET
bracis-28433	245	25	named	name	VERB
bracis-28433	245	26	entity	entity	NOUN
bracis-28433	245	27	class	class	NOUN
bracis-28433	245	28	,	,	PUNCT
bracis-28433	245	29	such	such	ADJ
bracis-28433	245	30	as	as	ADP
bracis-28433	245	31	the	the	DET
bracis-28433	245	32	aposentadoria	aposentadoria	PROPN
bracis-28433	245	33	dataset	dataset	NOUN
bracis-28433	245	34	,	,	PUNCT
bracis-28433	245	35	the	the	DET
bracis-28433	245	36	presented	present	VERB
bracis-28433	245	37	results	result	NOUN
bracis-28433	245	38	are	be	AUX
bracis-28433	245	39	convincing	convince	VERB
bracis-28433	245	40	evidence	evidence	NOUN
bracis-28433	245	41	that	that	SCONJ
bracis-28433	245	42	our	our	PRON
bracis-28433	245	43	method	method	NOUN
bracis-28433	245	44	is	be	AUX
bracis-28433	245	45	capable	capable	ADJ
bracis-28433	245	46	of	of	ADP
bracis-28433	245	47	reducing	reduce	VERB
bracis-28433	245	48	the	the	DET
bracis-28433	245	49	human	human	ADJ
bracis-28433	245	50	annotation	annotation	NOUN
bracis-28433	245	51	costs	cost	NOUN
bracis-28433	245	52	in	in	ADP
bracis-28433	245	53	the	the	DET
bracis-28433	245	54	active	active	ADJ
bracis-28433	245	55	learning	learning	NOUN
bracis-28433	245	56	process	process	NOUN
bracis-28433	245	57	when	when	SCONJ
bracis-28433	245	58	compared	compare	VERB
bracis-28433	245	59	to	to	ADP
bracis-28433	245	60	the	the	DET
bracis-28433	245	61	baselines	baseline	NOUN
bracis-28433	245	62	.	.	PUNCT
bracis-28433	246	1	table	table	NOUN
bracis-28433	246	2	2	2	NUM
bracis-28433	246	3	.	.	X
bracis-28433	246	4	percentage	percentage	NOUN
bracis-28433	246	5	of	of	ADP
bracis-28433	246	6	the	the	DET
bracis-28433	246	7	training	training	NOUN
bracis-28433	246	8	set	set	NOUN
bracis-28433	246	9	that	that	PRON
bracis-28433	246	10	was	be	AUX
bracis-28433	246	11	annotated	annotate	VERB
bracis-28433	246	12	by	by	ADP
bracis-28433	246	13	the	the	DET
bracis-28433	246	14	oracle	oracle	NOUN
bracis-28433	246	15	in	in	ADP
bracis-28433	246	16	order	order	NOUN
bracis-28433	246	17	to	to	PART
bracis-28433	246	18	train	train	VERB
bracis-28433	246	19	a	a	DET
bracis-28433	246	20	model	model	NOUN
bracis-28433	246	21	that	that	PRON
bracis-28433	246	22	reaches	reach	VERB
bracis-28433	246	23	99	99	NUM
bracis-28433	246	24	%	%	NOUN
bracis-28433	246	25	of	of	ADP
bracis-28433	246	26	its	its	PRON
bracis-28433	246	27	peak	peak	NOUN
bracis-28433	246	28	performance.full	performance.full	ADJ
bracis-28433	246	29	size	size	NOUN
bracis-28433	246	30	table	table	NOUN
bracis-28433	246	31	fig	fig	NOUN
bracis-28433	246	32	.	.	PUNCT
bracis-28433	247	1	4	4	X
bracis-28433	247	2	.	.	X
bracis-28433	247	3	comparison	comparison	NOUN
bracis-28433	247	4	between	between	ADP
bracis-28433	247	5	the	the	DET
bracis-28433	247	6	original	original	ADJ
bracis-28433	247	7	and	and	CCONJ
bracis-28433	247	8	modified	modify	VERB
bracis-28433	247	9	dal	dal	NOUN
bracis-28433	247	10	algorithms	algorithm	NOUN
bracis-28433	247	11	.	.	PUNCT
bracis-28433	248	1	graphs	graph	NOUN
bracis-28433	248	2	on	on	ADP
bracis-28433	248	3	the	the	DET
bracis-28433	248	4	left	left	ADJ
bracis-28433	248	5	column	column	NOUN
bracis-28433	248	6	compare	compare	VERB
bracis-28433	248	7	the	the	DET
bracis-28433	248	8	performance	performance	NOUN
bracis-28433	248	9	of	of	ADP
bracis-28433	248	10	the	the	DET
bracis-28433	248	11	trained	train	VERB
bracis-28433	248	12	model	model	NOUN
bracis-28433	248	13	(	(	PUNCT
bracis-28433	248	14	y	y	NOUN
bracis-28433	248	15	-	-	PUNCT
bracis-28433	248	16	axis	axis	NOUN
bracis-28433	248	17	)	)	PUNCT
bracis-28433	248	18	,	,	PUNCT
bracis-28433	248	19	by	by	ADP
bracis-28433	248	20	the	the	DET
bracis-28433	248	21	total	total	ADJ
bracis-28433	248	22	amount	amount	NOUN
bracis-28433	248	23	of	of	ADP
bracis-28433	248	24	labeled	label	VERB
bracis-28433	248	25	data	datum	NOUN
bracis-28433	248	26	(	(	PUNCT
bracis-28433	248	27	x	x	NOUN
bracis-28433	248	28	-	-	NOUN
bracis-28433	248	29	axis	axis	NOUN
bracis-28433	248	30	)	)	PUNCT
bracis-28433	248	31	,	,	PUNCT
bracis-28433	248	32	including	include	VERB
bracis-28433	248	33	tokens	token	NOUN
bracis-28433	248	34	annotated	annotate	VERB
bracis-28433	248	35	by	by	ADP
bracis-28433	248	36	the	the	DET
bracis-28433	248	37	oracle	oracle	NOUN
bracis-28433	248	38	and	and	CCONJ
bracis-28433	248	39	self	self	NOUN
bracis-28433	248	40	-	-	PUNCT
bracis-28433	248	41	annotated	annotate	VERB
bracis-28433	248	42	by	by	ADP
bracis-28433	248	43	the	the	DET
bracis-28433	248	44	trained	train	VERB
bracis-28433	248	45	model	model	NOUN
bracis-28433	248	46	in	in	ADP
bracis-28433	248	47	the	the	DET
bracis-28433	248	48	case	case	NOUN
bracis-28433	248	49	of	of	ADP
bracis-28433	248	50	our	our	PRON
bracis-28433	248	51	modified	modified	ADJ
bracis-28433	248	52	algorithm	algorithm	NOUN
bracis-28433	248	53	.	.	PUNCT
bracis-28433	249	1	graphs	graph	NOUN
bracis-28433	249	2	on	on	ADP
bracis-28433	249	3	the	the	DET
bracis-28433	249	4	center	center	NOUN
bracis-28433	249	5	column	column	NOUN
bracis-28433	249	6	compare	compare	VERB
bracis-28433	249	7	the	the	DET
bracis-28433	249	8	f1	f1	NOUN
bracis-28433	249	9	-	-	PUNCT
bracis-28433	249	10	score	score	NOUN
bracis-28433	249	11	of	of	ADP
bracis-28433	249	12	the	the	DET
bracis-28433	249	13	trained	train	VERB
bracis-28433	249	14	model	model	NOUN
bracis-28433	249	15	(	(	PUNCT
bracis-28433	249	16	y	y	NOUN
bracis-28433	249	17	-	-	PUNCT
bracis-28433	249	18	axis	axis	NOUN
bracis-28433	249	19	)	)	PUNCT
bracis-28433	249	20	by	by	ADP
bracis-28433	249	21	the	the	DET
bracis-28433	249	22	amount	amount	NOUN
bracis-28433	249	23	of	of	ADP
bracis-28433	249	24	data	datum	NOUN
bracis-28433	249	25	annotated	annotate	VERB
bracis-28433	249	26	by	by	ADP
bracis-28433	249	27	the	the	DET
bracis-28433	249	28	oracle	oracle	NOUN
bracis-28433	249	29	(	(	PUNCT
bracis-28433	249	30	x	x	NOUN
bracis-28433	249	31	-	-	NOUN
bracis-28433	249	32	axis	axis	ADJ
bracis-28433	249	33	)	)	PUNCT
bracis-28433	249	34	.	.	PUNCT
bracis-28433	250	1	graphs	graph	NOUN
bracis-28433	250	2	on	on	ADP
bracis-28433	250	3	the	the	DET
bracis-28433	250	4	right	right	ADJ
bracis-28433	250	5	column	column	NOUN
bracis-28433	250	6	compare	compare	VERB
bracis-28433	250	7	the	the	DET
bracis-28433	250	8	percentage	percentage	NOUN
bracis-28433	250	9	of	of	ADP
bracis-28433	250	10	the	the	DET
bracis-28433	250	11	training	training	NOUN
bracis-28433	250	12	set	set	NOUN
bracis-28433	250	13	that	that	PRON
bracis-28433	250	14	was	be	AUX
bracis-28433	250	15	annotated	annotate	VERB
bracis-28433	250	16	by	by	ADP
bracis-28433	250	17	a	a	DET
bracis-28433	250	18	human	human	NOUN
bracis-28433	250	19	,	,	PUNCT
bracis-28433	250	20	by	by	ADP
bracis-28433	250	21	the	the	DET
bracis-28433	250	22	model	model	NOUN
bracis-28433	250	23	through	through	ADP
bracis-28433	250	24	self	self	NOUN
bracis-28433	250	25	-	-	PUNCT
bracis-28433	250	26	labeling	labeling	NOUN
bracis-28433	250	27	,	,	PUNCT
bracis-28433	250	28	and	and	CCONJ
bracis-28433	250	29	the	the	DET
bracis-28433	250	30	samples	sample	NOUN
bracis-28433	250	31	that	that	PRON
bracis-28433	250	32	were	be	AUX
bracis-28433	250	33	mislabeled	mislabele	VERB
bracis-28433	250	34	.	.	PUNCT
bracis-28433	251	1	in	in	ADP
bracis-28433	251	2	all	all	DET
bracis-28433	251	3	graphs	graph	NOUN
bracis-28433	251	4	,	,	PUNCT
bracis-28433	251	5	dal	dal	NOUN
bracis-28433	251	6	represents	represent	VERB
bracis-28433	251	7	the	the	DET
bracis-28433	251	8	original	original	ADJ
bracis-28433	251	9	deep	deep	ADJ
bracis-28433	251	10	active	active	ADJ
bracis-28433	251	11	learning	learn	VERB
bracis-28433	251	12	algorithm	algorithm	NOUN
bracis-28433	251	13	from	from	ADP
bracis-28433	251	14	the	the	DET
bracis-28433	251	15	literature	literature	NOUN
bracis-28433	251	16	,	,	PUNCT
bracis-28433	251	17	while	while	SCONJ
bracis-28433	251	18	mdal	mdal	NOUN
bracis-28433	251	19	indicates	indicate	VERB
bracis-28433	251	20	our	our	PRON
bracis-28433	251	21	modified	modified	ADJ
bracis-28433	251	22	algorithm	algorithm	NOUN
bracis-28433	251	23	with	with	ADP
bracis-28433	251	24	token	token	VERB
bracis-28433	251	25	-	-	PUNCT
bracis-28433	251	26	level	level	NOUN
bracis-28433	251	27	self	self	NOUN
bracis-28433	251	28	-	-	PUNCT
bracis-28433	251	29	labeling	labeling	NOUN
bracis-28433	251	30	.	.	PUNCT
bracis-28433	252	1	full	full	ADJ
bracis-28433	252	2	size	size	NOUN
bracis-28433	252	3	image	image	NOUN
bracis-28433	252	4	8	8	NUM
bracis-28433	252	5	conclusion	conclusion	NOUN
bracis-28433	252	6	in	in	ADP
bracis-28433	252	7	this	this	DET
bracis-28433	252	8	work	work	NOUN
bracis-28433	252	9	,	,	PUNCT
bracis-28433	252	10	we	we	PRON
bracis-28433	252	11	presented	present	VERB
bracis-28433	252	12	an	an	DET
bracis-28433	252	13	investigation	investigation	NOUN
bracis-28433	252	14	into	into	ADP
bracis-28433	252	15	different	different	ADJ
bracis-28433	252	16	types	type	NOUN
bracis-28433	252	17	of	of	ADP
bracis-28433	252	18	active	active	ADJ
bracis-28433	252	19	-	-	PUNCT
bracis-28433	252	20	self	self	NOUN
bracis-28433	252	21	learning	learn	VERB
bracis-28433	252	22	algorithms	algorithm	NOUN
bracis-28433	252	23	for	for	ADP
bracis-28433	252	24	named	name	VERB
bracis-28433	252	25	entity	entity	NOUN
bracis-28433	252	26	recognition	recognition	NOUN
bracis-28433	252	27	tasks	task	NOUN
bracis-28433	252	28	.	.	PUNCT
bracis-28433	253	1	from	from	ADP
bracis-28433	253	2	the	the	DET
bracis-28433	253	3	many	many	ADJ
bracis-28433	253	4	experiments	experiment	NOUN
bracis-28433	253	5	performed	perform	VERB
bracis-28433	253	6	,	,	PUNCT
bracis-28433	253	7	the	the	DET
bracis-28433	253	8	sentence	sentence	NOUN
bracis-28433	253	9	-	-	PUNCT
bracis-28433	253	10	level	level	NOUN
bracis-28433	253	11	active	active	ADJ
bracis-28433	253	12	self	self	NOUN
bracis-28433	253	13	-	-	PUNCT
bracis-28433	253	14	learning	learn	VERB
bracis-28433	253	15	algorithm	algorithm	NOUN
bracis-28433	253	16	could	could	AUX
bracis-28433	253	17	not	not	PART
bracis-28433	253	18	consistently	consistently	ADV
bracis-28433	253	19	achieve	achieve	VERB
bracis-28433	253	20	significant	significant	ADJ
bracis-28433	253	21	results	result	NOUN
bracis-28433	253	22	compared	compare	VERB
bracis-28433	253	23	to	to	ADP
bracis-28433	253	24	pure	pure	ADJ
bracis-28433	253	25	active	active	ADJ
bracis-28433	253	26	learning	learning	NOUN
bracis-28433	253	27	.	.	PUNCT
bracis-28433	254	1	however	however	ADV
bracis-28433	254	2	,	,	PUNCT
bracis-28433	254	3	the	the	DET
bracis-28433	254	4	proposed	propose	VERB
bracis-28433	254	5	token	token	VERB
bracis-28433	254	6	-	-	PUNCT
bracis-28433	254	7	level	level	NOUN
bracis-28433	254	8	active	active	ADJ
bracis-28433	254	9	self	self	NOUN
bracis-28433	254	10	-	-	PUNCT
bracis-28433	254	11	learning	learning	NOUN
bracis-28433	254	12	could	could	AUX
bracis-28433	254	13	train	train	VERB
bracis-28433	254	14	a	a	DET
bracis-28433	254	15	neural	neural	ADJ
bracis-28433	254	16	model	model	NOUN
bracis-28433	254	17	to	to	ADP
bracis-28433	254	18	near	near	ADJ
bracis-28433	254	19	-	-	PUNCT
bracis-28433	254	20	peak	peak	NOUN
bracis-28433	254	21	performance	performance	NOUN
bracis-28433	254	22	using	use	VERB
bracis-28433	254	23	fewer	few	ADJ
bracis-28433	254	24	human	human	ADJ
bracis-28433	254	25	-	-	PUNCT
bracis-28433	254	26	annotated	annotate	VERB
bracis-28433	254	27	tokens	token	NOUN
bracis-28433	254	28	compared	compare	VERB
bracis-28433	254	29	to	to	ADP
bracis-28433	254	30	the	the	DET
bracis-28433	254	31	state	state	NOUN
bracis-28433	254	32	-	-	PUNCT
bracis-28433	254	33	of	of	ADP
bracis-28433	254	34	-	-	PUNCT
bracis-28433	254	35	the	the	DET
bracis-28433	254	36	-	-	PUNCT
bracis-28433	254	37	art	art	NOUN
bracis-28433	254	38	algorithms	algorithm	NOUN
bracis-28433	254	39	used	use	VERB
bracis-28433	254	40	as	as	ADP
bracis-28433	254	41	baselines	baseline	NOUN
bracis-28433	254	42	.	.	PUNCT
bracis-28433	255	1	our	our	PRON
bracis-28433	255	2	proposed	propose	VERB
bracis-28433	255	3	token	token	VERB
bracis-28433	255	4	-	-	PUNCT
bracis-28433	255	5	level	level	NOUN
bracis-28433	255	6	algorithm	algorithm	NOUN
bracis-28433	255	7	is	be	AUX
bracis-28433	255	8	particularly	particularly	ADV
bracis-28433	255	9	effective	effective	ADJ
bracis-28433	255	10	for	for	ADP
bracis-28433	255	11	fine	fine	ADV
bracis-28433	255	12	-	-	PUNCT
bracis-28433	255	13	grained	grain	VERB
bracis-28433	255	14	datasets	dataset	NOUN
bracis-28433	255	15	where	where	SCONJ
bracis-28433	255	16	most	most	ADJ
bracis-28433	255	17	tokens	token	NOUN
bracis-28433	255	18	are	be	AUX
bracis-28433	255	19	assigned	assign	VERB
bracis-28433	255	20	a	a	DET
bracis-28433	255	21	named	name	VERB
bracis-28433	255	22	entity	entity	NOUN
bracis-28433	255	23	class	class	NOUN
bracis-28433	255	24	.	.	PUNCT
bracis-28433	256	1	while	while	SCONJ
bracis-28433	256	2	we	we	PRON
bracis-28433	256	3	failed	fail	VERB
bracis-28433	256	4	at	at	ADP
bracis-28433	256	5	creating	create	VERB
bracis-28433	256	6	a	a	DET
bracis-28433	256	7	sentence	sentence	NOUN
bracis-28433	256	8	-	-	PUNCT
bracis-28433	256	9	level	level	NOUN
bracis-28433	256	10	active	active	ADJ
bracis-28433	256	11	self	self	NOUN
bracis-28433	256	12	-	-	PUNCT
bracis-28433	256	13	learning	learning	NOUN
bracis-28433	256	14	that	that	PRON
bracis-28433	256	15	overcomes	overcome	VERB
bracis-28433	256	16	the	the	DET
bracis-28433	256	17	current	current	ADJ
bracis-28433	256	18	state	state	NOUN
bracis-28433	256	19	-	-	PUNCT
bracis-28433	256	20	of	of	ADP
bracis-28433	256	21	-	-	PUNCT
bracis-28433	256	22	the	the	DET
bracis-28433	256	23	-	-	PUNCT
bracis-28433	256	24	art	art	NOUN
bracis-28433	256	25	,	,	PUNCT
bracis-28433	256	26	future	future	ADJ
bracis-28433	256	27	research	research	NOUN
bracis-28433	256	28	may	may	AUX
bracis-28433	256	29	investigate	investigate	VERB
bracis-28433	256	30	how	how	SCONJ
bracis-28433	256	31	pretrained	pretraine	VERB
bracis-28433	256	32	transformer	transformer	NOUN
bracis-28433	256	33	models	model	NOUN
bracis-28433	256	34	may	may	AUX
bracis-28433	256	35	impact	impact	VERB
bracis-28433	256	36	these	these	DET
bracis-28433	256	37	algorithms	algorithm	NOUN
bracis-28433	256	38	.	.	PUNCT
bracis-28433	257	1	bridging	bridge	VERB
bracis-28433	257	2	the	the	DET
bracis-28433	257	3	current	current	ADJ
bracis-28433	257	4	gaps	gap	NOUN
bracis-28433	257	5	in	in	ADP
bracis-28433	257	6	sentence	sentence	NOUN
bracis-28433	257	7	-	-	PUNCT
bracis-28433	257	8	level	level	NOUN
bracis-28433	257	9	active	active	ADJ
bracis-28433	257	10	self	self	NOUN
bracis-28433	257	11	-	-	PUNCT
bracis-28433	257	12	learning	learn	VERB
bracis-28433	257	13	algorithms	algorithm	NOUN
bracis-28433	257	14	research	research	NOUN
bracis-28433	257	15	using	use	VERB
bracis-28433	257	16	models	model	NOUN
bracis-28433	257	17	with	with	ADP
bracis-28433	257	18	substantial	substantial	ADJ
bracis-28433	257	19	a	a	DET
bracis-28433	257	20	priori	priori	ADJ
bracis-28433	257	21	information	information	NOUN
bracis-28433	257	22	may	may	AUX
bracis-28433	257	23	be	be	AUX
bracis-28433	257	24	possible	possible	ADJ
bracis-28433	257	25	.	.	PUNCT
bracis-28433	258	1	references	reference	NOUN
bracis-28433	258	2	neto	neto	PROPN
bracis-28433	258	3	,	,	PUNCT
bracis-28433	258	4	j.r.c.s.a.v.s	j.r.c.s.a.v.s	PROPN
bracis-28433	258	5	.	.	PUNCT
bracis-28433	258	6	:	:	PUNCT
bracis-28433	259	1	deep	deep	ADJ
bracis-28433	259	2	active	active	ADJ
bracis-28433	259	3	learning	learning	NOUN
bracis-28433	259	4	approaches	approach	NOUN
bracis-28433	259	5	to	to	ADP
bracis-28433	259	6	the	the	DET
bracis-28433	259	7	task	task	NOUN
bracis-28433	259	8	of	of	ADP
bracis-28433	259	9	named	name	VERB
bracis-28433	259	10	entity	entity	NOUN
bracis-28433	259	11	recognition	recognition	NOUN
bracis-28433	259	12	.	.	PUNCT
bracis-28433	260	1	masters	master	NOUN
bracis-28433	260	2	dissertation	dissertation	NOUN
bracis-28433	261	1	[	[	X
bracis-28433	261	2	university	university	NOUN
bracis-28433	261	3	of	of	ADP
bracis-28433	261	4	brasilia	brasilia	PROPN
bracis-28433	261	5	]	]	PUNCT
bracis-28433	261	6	(	(	PUNCT
bracis-28433	261	7	2021	2021	NUM
bracis-28433	261	8	)	)	PUNCT
bracis-28433	261	9	google	google	PROPN
bracis-28433	261	10	scholar	scholar	NOUN
bracis-28433	261	11	  	  	SPACE
bracis-28433	261	12	neto	neto	NOUN
bracis-28433	261	13	,	,	PUNCT
bracis-28433	261	14	j.r.c.s.a.v.s	j.r.c.s.a.v.s	PROPN
bracis-28433	261	15	.	.	PROPN
bracis-28433	261	16	,	,	PUNCT
bracis-28433	261	17	faleiros	faleiros	PROPN
bracis-28433	261	18	,	,	PUNCT
bracis-28433	261	19	t.p	t.p	PROPN
bracis-28433	261	20	.	.	PROPN
bracis-28433	261	21	:	:	PUNCT
bracis-28433	261	22	deep	deep	ADJ
bracis-28433	261	23	active	active	ADJ
bracis-28433	261	24	-	-	PUNCT
bracis-28433	261	25	self	self	NOUN
bracis-28433	261	26	learning	learning	NOUN
bracis-28433	261	27	applied	apply	VERB
bracis-28433	261	28	to	to	ADP
bracis-28433	261	29	named	name	VERB
bracis-28433	261	30	entity	entity	NOUN
bracis-28433	261	31	recognition	recognition	NOUN
bracis-28433	261	32	.	.	PUNCT
bracis-28433	262	1	in	in	ADP
bracis-28433	262	2	:	:	PUNCT
bracis-28433	262	3	britto	britto	PROPN
bracis-28433	262	4	,	,	PUNCT
bracis-28433	262	5	a.	a.	PROPN
bracis-28433	262	6	,	,	PUNCT
bracis-28433	262	7	valdivia	valdivia	PROPN
bracis-28433	262	8	delgado	delgado	PROPN
bracis-28433	262	9	,	,	PUNCT
bracis-28433	262	10	k.	k.	PROPN
bracis-28433	262	11	(	(	PUNCT
bracis-28433	262	12	eds	eds	PROPN
bracis-28433	262	13	.	.	PUNCT
bracis-28433	262	14	)	)	PUNCT
bracis-28433	262	15	bracis	bracis	PROPN
bracis-28433	262	16	2021	2021	NUM
bracis-28433	262	17	.	.	PUNCT
bracis-28433	263	1	lncs	lncs	PROPN
bracis-28433	263	2	(	(	PUNCT
bracis-28433	263	3	lnai	lnai	ADJ
bracis-28433	263	4	)	)	PUNCT
bracis-28433	263	5	,	,	PUNCT
bracis-28433	263	6	vol	vol	NOUN
bracis-28433	263	7	.	.	PROPN
bracis-28433	263	8	13074	13074	NUM
bracis-28433	263	9	,	,	PUNCT
bracis-28433	263	10	pp	pp	ADJ
bracis-28433	263	11	.	.	PUNCT
bracis-28433	264	1	405–418	405–418	NUM
bracis-28433	264	2	.	.	PUNCT
bracis-28433	265	1	springer	springer	NOUN
bracis-28433	265	2	,	,	PUNCT
bracis-28433	265	3	cham	cham	PROPN
bracis-28433	265	4	(	(	PUNCT
bracis-28433	265	5	2021	2021	NUM
bracis-28433	265	6	)	)	PUNCT
bracis-28433	265	7	.	.	PUNCT
bracis-28433	266	1	https://doi.org/10.1007/978-3-030-91699-2_28	https://doi.org/10.1007/978-3-030-91699-2_28	NOUN
bracis-28433	266	2	chapter	chapter	NOUN
bracis-28433	266	3	  	  	SPACE
bracis-28433	266	4	google	google	PROPN
bracis-28433	266	5	scholar	scholar	NOUN
bracis-28433	266	6	  	  	SPACE
bracis-28433	266	7	clark	clark	NOUN
bracis-28433	266	8	,	,	PUNCT
bracis-28433	266	9	k.	k.	PROPN
bracis-28433	266	10	,	,	PUNCT
bracis-28433	266	11	luong	luong	PROPN
bracis-28433	266	12	,	,	PUNCT
bracis-28433	266	13	m.t	m.t	PROPN
bracis-28433	266	14	.	.	PROPN
bracis-28433	266	15	,	,	PUNCT
bracis-28433	266	16	manning	manning	NOUN
bracis-28433	266	17	,	,	PUNCT
bracis-28433	266	18	c.d	c.d	PROPN
bracis-28433	266	19	.	.	PROPN
bracis-28433	266	20	,	,	PUNCT
bracis-28433	266	21	le	le	PROPN
bracis-28433	266	22	,	,	PUNCT
bracis-28433	266	23	q.	q.	PROPN
bracis-28433	266	24	:	:	PUNCT
bracis-28433	266	25	semi	semi	ADJ
bracis-28433	266	26	-	-	ADJ
bracis-28433	266	27	supervised	supervised	ADJ
bracis-28433	266	28	sequence	sequence	NOUN
bracis-28433	266	29	modeling	modeling	NOUN
bracis-28433	266	30	with	with	ADP
bracis-28433	266	31	cross	cross	ADJ
bracis-28433	266	32	-	-	ADJ
bracis-28433	266	33	view	view	ADJ
bracis-28433	266	34	training	training	NOUN
bracis-28433	266	35	.	.	PUNCT
bracis-28433	267	1	in	in	ADP
bracis-28433	267	2	:	:	PUNCT
bracis-28433	267	3	proceedings	proceeding	NOUN
bracis-28433	267	4	of	of	ADP
bracis-28433	267	5	the	the	DET
bracis-28433	267	6	2018	2018	NUM
bracis-28433	267	7	conference	conference	NOUN
bracis-28433	267	8	on	on	ADP
bracis-28433	267	9	empirical	empirical	ADJ
bracis-28433	267	10	methods	method	NOUN
bracis-28433	267	11	in	in	ADP
bracis-28433	267	12	natural	natural	ADJ
bracis-28433	267	13	language	language	NOUN
bracis-28433	267	14	processing	processing	NOUN
bracis-28433	267	15	,	,	PUNCT
bracis-28433	267	16	pp	pp	ADJ
bracis-28433	267	17	.	.	PUNCT
bracis-28433	267	18	1914–1925	1914–1925	NUM
bracis-28433	267	19	.	.	PUNCT
bracis-28433	267	20	association	association	NOUN
bracis-28433	267	21	for	for	ADP
bracis-28433	267	22	computational	computational	ADJ
bracis-28433	267	23	linguistics	linguistic	NOUN
bracis-28433	267	24	,	,	PUNCT
bracis-28433	267	25	brussels	brussels	PROPN
bracis-28433	267	26	,	,	PUNCT
bracis-28433	267	27	belgium	belgium	PROPN
bracis-28433	267	28	,	,	PUNCT
bracis-28433	267	29	october	october	PROPN
bracis-28433	267	30	–	–	PUNCT
bracis-28433	267	31	november	november	PROPN
bracis-28433	267	32	2018	2018	NUM
bracis-28433	267	33	google	google	PROPN
bracis-28433	267	34	scholar	scholar	NOUN
bracis-28433	267	35	  	  	SPACE
bracis-28433	267	36	goodfellow	goodfellow	PROPN
bracis-28433	267	37	,	,	PUNCT
bracis-28433	267	38	i.j	i.j	PROPN
bracis-28433	267	39	.	.	PROPN
bracis-28433	267	40	,	,	PUNCT
bracis-28433	267	41	shlens	shlens	PROPN
bracis-28433	267	42	,	,	PUNCT
bracis-28433	267	43	j.	j.	PROPN
bracis-28433	267	44	,	,	PUNCT
bracis-28433	267	45	szegedy	szegedy	PROPN
bracis-28433	267	46	,	,	PUNCT
bracis-28433	267	47	c.	c.	NOUN
bracis-28433	267	48	:	:	PUNCT
bracis-28433	267	49	explaining	explain	VERB
bracis-28433	267	50	and	and	CCONJ
bracis-28433	267	51	harnessing	harness	VERB
bracis-28433	267	52	adversarial	adversarial	ADJ
bracis-28433	267	53	examples	example	NOUN
bracis-28433	267	54	(	(	PUNCT
bracis-28433	267	55	2015	2015	NUM
bracis-28433	267	56	)	)	PUNCT
bracis-28433	267	57	google	google	PROPN
bracis-28433	267	58	scholar	scholar	NOUN
bracis-28433	267	59	  	  	SPACE
bracis-28433	267	60	hartmann	hartmann	PROPN
bracis-28433	267	61	,	,	PUNCT
bracis-28433	267	62	n.s	n.s	PROPN
bracis-28433	267	63	.	.	PROPN
bracis-28433	267	64	,	,	PUNCT
bracis-28433	267	65	fonseca	fonseca	PROPN
bracis-28433	267	66	,	,	PUNCT
bracis-28433	267	67	e.r	e.r	PROPN
bracis-28433	267	68	.	.	PROPN
bracis-28433	267	69	,	,	PUNCT
bracis-28433	267	70	shulby	shulby	PROPN
bracis-28433	267	71	,	,	PUNCT
bracis-28433	267	72	c.d	c.d	PROPN
bracis-28433	267	73	.	.	PROPN
bracis-28433	267	74	,	,	PUNCT
bracis-28433	267	75	treviso	treviso	PROPN
bracis-28433	267	76	,	,	PUNCT
bracis-28433	267	77	m.v	m.v	PROPN
bracis-28433	267	78	.	.	PROPN
bracis-28433	267	79	,	,	PUNCT
bracis-28433	267	80	rodrigues	rodrigues	PROPN
bracis-28433	267	81	,	,	PUNCT
bracis-28433	267	82	j.s	j.s	PROPN
bracis-28433	267	83	.	.	PROPN
bracis-28433	267	84	,	,	PUNCT
bracis-28433	267	85	aluísio	aluísio	PROPN
bracis-28433	267	86	,	,	PUNCT
bracis-28433	267	87	s.m	s.m	PROPN
bracis-28433	267	88	.	.	PROPN
bracis-28433	267	89	:	:	PUNCT
bracis-28433	268	1	portuguese	portuguese	ADJ
bracis-28433	268	2	word	word	NOUN
bracis-28433	268	3	embeddings	embedding	NOUN
bracis-28433	268	4	:	:	PUNCT
bracis-28433	268	5	evaluating	evaluate	VERB
bracis-28433	268	6	on	on	ADP
bracis-28433	268	7	word	word	NOUN
bracis-28433	268	8	analogies	analogy	NOUN
bracis-28433	268	9	and	and	CCONJ
bracis-28433	268	10	natural	natural	ADJ
bracis-28433	268	11	language	language	NOUN
bracis-28433	268	12	tasks	task	NOUN
bracis-28433	268	13	.	.	PUNCT
bracis-28433	269	1	in	in	ADP
bracis-28433	269	2	:	:	PUNCT
bracis-28433	269	3	anais	anais	PROPN
bracis-28433	269	4	do	do	AUX
bracis-28433	269	5	xi	xi	PROPN
bracis-28433	269	6	simpósio	simpósio	VERB
bracis-28433	269	7	brasileiro	brasileiro	PROPN
bracis-28433	269	8	de	de	PROPN
bracis-28433	269	9	tecnologia	tecnologia	PROPN
bracis-28433	269	10	da	da	PROPN
bracis-28433	269	11	informação	informação	PROPN
bracis-28433	269	12	e	e	PROPN
bracis-28433	269	13	da	da	PROPN
bracis-28433	269	14	linguagem	linguagem	PROPN
bracis-28433	269	15	humana	humana	PROPN
bracis-28433	269	16	,	,	PUNCT
bracis-28433	269	17	pp	pp	ADJ
bracis-28433	269	18	.	.	PUNCT
bracis-28433	270	1	122–131	122–131	NUM
bracis-28433	270	2	.	.	PUNCT
bracis-28433	271	1	sbc	sbc	PROPN
bracis-28433	271	2	,	,	PUNCT
bracis-28433	271	3	porto	porto	PROPN
bracis-28433	271	4	alegre	alegre	PROPN
bracis-28433	271	5	,	,	PUNCT
bracis-28433	271	6	rs	rs	NOUN
bracis-28433	271	7	,	,	PUNCT
bracis-28433	271	8	brasil	brasil	PROPN
bracis-28433	271	9	(	(	PUNCT
bracis-28433	271	10	2017	2017	NUM
bracis-28433	271	11	)	)	PUNCT
bracis-28433	271	12	google	google	PROPN
bracis-28433	271	13	scholar	scholar	NOUN
bracis-28433	271	14	  	  	SPACE
bracis-28433	271	15	houlsby	houlsby	ADJ
bracis-28433	271	16	,	,	PUNCT
bracis-28433	271	17	n.	n.	NOUN
bracis-28433	271	18	,	,	PUNCT
bracis-28433	271	19	huszár	huszár	NOUN
bracis-28433	271	20	,	,	PUNCT
bracis-28433	271	21	f.	f.	PROPN
bracis-28433	271	22	,	,	PUNCT
bracis-28433	271	23	ghahramani	ghahramani	PROPN
bracis-28433	271	24	,	,	PUNCT
bracis-28433	271	25	z.	z.	PROPN
bracis-28433	271	26	,	,	PUNCT
bracis-28433	271	27	lengyel	lengyel	PROPN
bracis-28433	271	28	,	,	PUNCT
bracis-28433	271	29	m.	m.	NOUN
bracis-28433	271	30	:	:	PUNCT
bracis-28433	271	31	bayesian	bayesian	NOUN
bracis-28433	271	32	active	active	ADJ
bracis-28433	271	33	learning	learning	NOUN
bracis-28433	271	34	for	for	ADP
bracis-28433	271	35	classification	classification	NOUN
bracis-28433	271	36	and	and	CCONJ
bracis-28433	271	37	preference	preference	NOUN
bracis-28433	271	38	learning	learning	NOUN
bracis-28433	271	39	(	(	PUNCT
bracis-28433	271	40	2011	2011	NUM
bracis-28433	271	41	)	)	PUNCT
bracis-28433	271	42	google	google	PROPN
bracis-28433	271	43	scholar	scholar	NOUN
bracis-28433	271	44	  	  	SPACE
bracis-28433	271	45	kobayashi	kobayashi	PROPN
bracis-28433	271	46	,	,	PUNCT
bracis-28433	271	47	k.	k.	PROPN
bracis-28433	271	48	,	,	PUNCT
bracis-28433	271	49	wakabayashi	wakabayashi	PROPN
bracis-28433	271	50	,	,	PUNCT
bracis-28433	271	51	k.	k.	PROPN
bracis-28433	271	52	:	:	PUNCT
bracis-28433	271	53	named	name	VERB
bracis-28433	271	54	entity	entity	NOUN
bracis-28433	271	55	recognition	recognition	NOUN
bracis-28433	271	56	using	use	VERB
bracis-28433	271	57	point	point	NOUN
bracis-28433	271	58	prediction	prediction	NOUN
bracis-28433	271	59	and	and	CCONJ
bracis-28433	271	60	active	active	ADJ
bracis-28433	271	61	learning	learning	NOUN
bracis-28433	271	62	.	.	PUNCT
bracis-28433	272	1	in	in	ADP
bracis-28433	272	2	:	:	PUNCT
bracis-28433	272	3	proceedings	proceeding	NOUN
bracis-28433	272	4	of	of	ADP
bracis-28433	272	5	the	the	DET
bracis-28433	272	6	21st	21st	ADJ
bracis-28433	272	7	international	international	ADJ
bracis-28433	272	8	conference	conference	NOUN
bracis-28433	272	9	on	on	ADP
bracis-28433	272	10	information	information	NOUN
bracis-28433	272	11	integration	integration	NOUN
bracis-28433	272	12	and	and	CCONJ
bracis-28433	272	13	web	web	NOUN
bracis-28433	272	14	-	-	PUNCT
bracis-28433	272	15	based	base	VERB
bracis-28433	272	16	applications	application	NOUN
bracis-28433	272	17	and	and	CCONJ
bracis-28433	272	18	services	service	NOUN
bracis-28433	272	19	,	,	PUNCT
bracis-28433	272	20	iiwas2019	iiwas2019	PROPN
bracis-28433	272	21	,	,	PUNCT
bracis-28433	272	22	pp	pp	X
bracis-28433	272	23	.	.	PUNCT
bracis-28433	273	1	287–293	287–293	NUM
bracis-28433	273	2	.	.	PUNCT
bracis-28433	273	3	association	association	NOUN
bracis-28433	273	4	for	for	ADP
bracis-28433	273	5	computing	compute	VERB
bracis-28433	273	6	machinery	machinery	NOUN
bracis-28433	273	7	,	,	PUNCT
bracis-28433	273	8	new	new	PROPN
bracis-28433	273	9	york	york	PROPN
bracis-28433	273	10	,	,	PUNCT
bracis-28433	273	11	ny	ny	PROPN
bracis-28433	273	12	,	,	PUNCT
bracis-28433	273	13	usa	usa	PROPN
bracis-28433	273	14	(	(	PUNCT
bracis-28433	273	15	2019	2019	NUM
bracis-28433	273	16	)	)	PUNCT
bracis-28433	273	17	google	google	NOUN
bracis-28433	273	18	scholar	scholar	NOUN
bracis-28433	273	19	  	  	SPACE
bracis-28433	273	20	lakshmi	lakshmi	NOUN
bracis-28433	273	21	narayan	narayan	NOUN
bracis-28433	273	22	,	,	PUNCT
bracis-28433	273	23	p.	p.	PROPN
bracis-28433	273	24	,	,	PUNCT
bracis-28433	273	25	nagesh	nagesh	PROPN
bracis-28433	273	26	,	,	PUNCT
bracis-28433	273	27	a.	a.	PROPN
bracis-28433	273	28	,	,	PUNCT
bracis-28433	273	29	surdeanu	surdeanu	NOUN
bracis-28433	273	30	,	,	PUNCT
bracis-28433	273	31	m.	m.	NOUN
bracis-28433	273	32	:	:	PUNCT
bracis-28433	273	33	exploration	exploration	NOUN
bracis-28433	273	34	of	of	ADP
bracis-28433	273	35	noise	noise	NOUN
bracis-28433	273	36	strategies	strategy	NOUN
bracis-28433	273	37	in	in	ADP
bracis-28433	273	38	semi	semi	ADJ
bracis-28433	273	39	-	-	ADJ
bracis-28433	273	40	supervised	supervised	ADJ
bracis-28433	273	41	named	name	VERB
bracis-28433	273	42	entity	entity	NOUN
bracis-28433	273	43	classification	classification	NOUN
bracis-28433	273	44	.	.	PUNCT
bracis-28433	274	1	in	in	ADP
bracis-28433	274	2	:	:	PUNCT
bracis-28433	274	3	proceedings	proceeding	NOUN
bracis-28433	274	4	of	of	ADP
bracis-28433	274	5	the	the	DET
bracis-28433	274	6	eighth	eighth	ADJ
bracis-28433	274	7	joint	joint	ADJ
bracis-28433	274	8	conference	conference	NOUN
bracis-28433	274	9	on	on	ADP
bracis-28433	274	10	lexical	lexical	ADJ
bracis-28433	274	11	and	and	CCONJ
bracis-28433	274	12	computational	computational	ADJ
bracis-28433	274	13	semantics	semantic	NOUN
bracis-28433	274	14	(	(	PUNCT
bracis-28433	274	15	*	*	PUNCT
bracis-28433	274	16	sem	sem	NOUN
bracis-28433	274	17	2019	2019	NUM
bracis-28433	274	18	)	)	PUNCT
bracis-28433	274	19	,	,	PUNCT
bracis-28433	274	20	pp	pp	ADP
bracis-28433	274	21	.	.	PUNCT
bracis-28433	275	1	186–191	186–191	NUM
bracis-28433	275	2	.	.	PUNCT
bracis-28433	275	3	association	association	NOUN
bracis-28433	275	4	for	for	ADP
bracis-28433	275	5	computational	computational	ADJ
bracis-28433	275	6	linguistics	linguistic	NOUN
bracis-28433	275	7	,	,	PUNCT
bracis-28433	275	8	minneapolis	minneapolis	PROPN
bracis-28433	275	9	,	,	PUNCT
bracis-28433	275	10	minnesota	minnesota	PROPN
bracis-28433	275	11	,	,	PUNCT
bracis-28433	275	12	june	june	PROPN
bracis-28433	275	13	2019	2019	NUM
bracis-28433	275	14	google	google	PROPN
bracis-28433	275	15	scholar	scholar	NOUN
bracis-28433	275	16	  	  	SPACE
bracis-28433	275	17	ma	ma	PROPN
bracis-28433	275	18	,	,	PUNCT
bracis-28433	275	19	x.	x.	NOUN
bracis-28433	275	20	,	,	PUNCT
bracis-28433	275	21	hovy	hovy	PROPN
bracis-28433	275	22	,	,	PUNCT
bracis-28433	275	23	e.	e.	PROPN
bracis-28433	275	24	:	:	PUNCT
bracis-28433	275	25	end	end	VERB
bracis-28433	275	26	-	-	PUNCT
bracis-28433	275	27	to	to	ADP
bracis-28433	275	28	-	-	PUNCT
bracis-28433	275	29	end	end	NOUN
bracis-28433	275	30	sequence	sequence	NOUN
bracis-28433	275	31	labeling	labeling	NOUN
bracis-28433	275	32	via	via	ADP
bracis-28433	275	33	bi	bi	ADJ
bracis-28433	275	34	-	-	ADJ
bracis-28433	275	35	directional	directional	ADJ
bracis-28433	275	36	lstm	lstm	ADJ
bracis-28433	275	37	-	-	PUNCT
bracis-28433	275	38	cnns	cnns	PROPN
bracis-28433	275	39	-	-	PUNCT
bracis-28433	275	40	crf	crf	NOUN
bracis-28433	275	41	.	.	PUNCT
bracis-28433	276	1	in	in	ADP
bracis-28433	276	2	:	:	PUNCT
bracis-28433	276	3	proceedings	proceeding	NOUN
bracis-28433	276	4	of	of	ADP
bracis-28433	276	5	the	the	DET
bracis-28433	276	6	54th	54th	ADJ
bracis-28433	276	7	annual	annual	ADJ
bracis-28433	276	8	meeting	meeting	NOUN
bracis-28433	276	9	of	of	ADP
bracis-28433	276	10	the	the	DET
bracis-28433	276	11	association	association	NOUN
bracis-28433	276	12	for	for	ADP
bracis-28433	276	13	computational	computational	ADJ
bracis-28433	276	14	linguistics	linguistic	NOUN
bracis-28433	276	15	(	(	PUNCT
bracis-28433	276	16	volume	volume	NOUN
bracis-28433	276	17	1	1	NUM
bracis-28433	276	18	:	:	PUNCT
bracis-28433	276	19	long	long	ADJ
bracis-28433	276	20	papers	paper	NOUN
bracis-28433	276	21	)	)	PUNCT
bracis-28433	276	22	,	,	PUNCT
bracis-28433	276	23	pp	pp	ADP
bracis-28433	276	24	.	.	PUNCT
bracis-28433	277	1	1064–1074	1064–1074	NUM
bracis-28433	277	2	.	.	PUNCT
bracis-28433	278	1	association	association	NOUN
bracis-28433	278	2	for	for	ADP
bracis-28433	278	3	computational	computational	ADJ
bracis-28433	278	4	linguistics	linguistic	NOUN
bracis-28433	278	5	,	,	PUNCT
bracis-28433	278	6	berlin	berlin	PROPN
bracis-28433	278	7	,	,	PUNCT
bracis-28433	278	8	germany	germany	PROPN
bracis-28433	278	9	,	,	PUNCT
bracis-28433	278	10	august	august	PROPN
bracis-28433	278	11	2016	2016	NUM
bracis-28433	278	12	google	google	PROPN
bracis-28433	278	13	scholar	scholar	NOUN
bracis-28433	278	14	  	  	SPACE
bracis-28433	278	15	miyato	miyato	NOUN
bracis-28433	278	16	,	,	PUNCT
bracis-28433	278	17	t.	t.	PROPN
bracis-28433	278	18	,	,	PUNCT
bracis-28433	278	19	dai	dai	PROPN
bracis-28433	278	20	,	,	PUNCT
bracis-28433	278	21	a.m.	a.m.	ADV
bracis-28433	278	22	,	,	PUNCT
bracis-28433	278	23	goodfellow	goodfellow	PROPN
bracis-28433	278	24	,	,	PUNCT
bracis-28433	278	25	i.	i.	NOUN
bracis-28433	278	26	:	:	PUNCT
bracis-28433	278	27	adversarial	adversarial	ADJ
bracis-28433	278	28	training	training	NOUN
bracis-28433	278	29	methods	method	NOUN
bracis-28433	278	30	for	for	ADP
bracis-28433	278	31	semi	semi	ADJ
bracis-28433	278	32	-	-	ADJ
bracis-28433	278	33	supervised	supervised	ADJ
bracis-28433	278	34	text	text	NOUN
bracis-28433	278	35	classification	classification	NOUN
bracis-28433	278	36	.	.	PUNCT
bracis-28433	279	1	in	in	ADP
bracis-28433	279	2	:	:	PUNCT
bracis-28433	279	3	international	international	ADJ
bracis-28433	279	4	conference	conference	NOUN
bracis-28433	279	5	on	on	ADP
bracis-28433	279	6	learning	learn	VERB
bracis-28433	279	7	representations	representation	NOUN
bracis-28433	279	8	(	(	PUNCT
bracis-28433	279	9	iclr	iclr	NOUN
bracis-28433	279	10	)	)	PUNCT
bracis-28433	279	11	(	(	PUNCT
bracis-28433	279	12	2017	2017	NUM
bracis-28433	279	13	)	)	PUNCT
bracis-28433	279	14	google	google	NOUN
bracis-28433	279	15	scholar	scholar	NOUN
bracis-28433	279	16	  	  	SPACE
bracis-28433	279	17	nair	nair	NOUN
bracis-28433	279	18	,	,	PUNCT
bracis-28433	279	19	v.	v.	PROPN
bracis-28433	279	20	,	,	PUNCT
bracis-28433	279	21	hinton	hinton	PROPN
bracis-28433	279	22	,	,	PUNCT
bracis-28433	279	23	g.e	g.e	PROPN
bracis-28433	279	24	.	.	PUNCT
bracis-28433	279	25	:	:	PUNCT
bracis-28433	279	26	rectified	rectify	VERB
bracis-28433	279	27	linear	linear	ADJ
bracis-28433	279	28	units	unit	NOUN
bracis-28433	279	29	improve	improve	VERB
bracis-28433	279	30	restricted	restrict	VERB
bracis-28433	279	31	boltzmann	boltzmann	PROPN
bracis-28433	279	32	machines	machine	NOUN
bracis-28433	279	33	.	.	PUNCT
bracis-28433	280	1	in	in	ADP
bracis-28433	280	2	:	:	PUNCT
bracis-28433	280	3	proceedings	proceeding	NOUN
bracis-28433	280	4	of	of	ADP
bracis-28433	280	5	the	the	DET
bracis-28433	280	6	27th	27th	ADJ
bracis-28433	280	7	international	international	ADJ
bracis-28433	280	8	conference	conference	NOUN
bracis-28433	280	9	on	on	ADP
bracis-28433	280	10	international	international	ADJ
bracis-28433	280	11	conference	conference	NOUN
bracis-28433	280	12	on	on	ADP
bracis-28433	280	13	machine	machine	NOUN
bracis-28433	280	14	learning	learning	NOUN
bracis-28433	280	15	,	,	PUNCT
bracis-28433	280	16	icml	icml	VERB
bracis-28433	280	17	2010	2010	NUM
bracis-28433	280	18	,	,	PUNCT
bracis-28433	280	19	pp	pp	ADP
bracis-28433	280	20	.	.	PUNCT
bracis-28433	281	1	807–814	807–814	NUM
bracis-28433	281	2	.	.	PUNCT
bracis-28433	282	1	omnipress	omnipress	ADJ
bracis-28433	282	2	,	,	PUNCT
bracis-28433	282	3	madison	madison	PROPN
bracis-28433	282	4	,	,	PUNCT
bracis-28433	282	5	wi	wi	PROPN
bracis-28433	282	6	,	,	PUNCT
bracis-28433	282	7	usa	usa	PROPN
bracis-28433	282	8	(	(	PUNCT
bracis-28433	282	9	2010	2010	NUM
bracis-28433	282	10	)	)	PUNCT
bracis-28433	282	11	google	google	PROPN
bracis-28433	282	12	scholar	scholar	NOUN
bracis-28433	282	13	  	  	SPACE
bracis-28433	282	14	neubig	neubig	NOUN
bracis-28433	282	15	,	,	PUNCT
bracis-28433	282	16	g.	g.	NOUN
bracis-28433	282	17	,	,	PUNCT
bracis-28433	282	18	nakata	nakata	NOUN
bracis-28433	282	19	,	,	PUNCT
bracis-28433	282	20	y.	y.	PROPN
bracis-28433	282	21	,	,	PUNCT
bracis-28433	282	22	mori	mori	PROPN
bracis-28433	282	23	,	,	PUNCT
bracis-28433	282	24	s.	s.	PROPN
bracis-28433	282	25	:	:	PUNCT
bracis-28433	282	26	pointwise	pointwise	VERB
bracis-28433	282	27	prediction	prediction	NOUN
bracis-28433	282	28	for	for	ADP
bracis-28433	282	29	robust	robust	ADJ
bracis-28433	282	30	,	,	PUNCT
bracis-28433	282	31	adaptable	adaptable	ADJ
bracis-28433	282	32	japanese	japanese	ADJ
bracis-28433	282	33	morphological	morphological	ADJ
bracis-28433	282	34	analysis	analysis	NOUN
bracis-28433	282	35	.	.	PUNCT
bracis-28433	283	1	in	in	ADP
bracis-28433	283	2	:	:	PUNCT
bracis-28433	283	3	proceedings	proceeding	NOUN
bracis-28433	283	4	of	of	ADP
bracis-28433	283	5	the	the	DET
bracis-28433	283	6	49th	49th	ADJ
bracis-28433	283	7	annual	annual	ADJ
bracis-28433	283	8	meeting	meeting	NOUN
bracis-28433	283	9	of	of	ADP
bracis-28433	283	10	the	the	DET
bracis-28433	283	11	association	association	NOUN
bracis-28433	283	12	for	for	ADP
bracis-28433	283	13	computational	computational	ADJ
bracis-28433	283	14	linguistics	linguistic	NOUN
bracis-28433	283	15	:	:	PUNCT
bracis-28433	283	16	human	human	ADJ
bracis-28433	283	17	language	language	NOUN
bracis-28433	283	18	technologies	technology	NOUN
bracis-28433	283	19	,	,	PUNCT
bracis-28433	283	20	pp	pp	ADJ
bracis-28433	283	21	.	.	PUNCT
bracis-28433	284	1	529–533	529–533	NUM
bracis-28433	284	2	.	.	PUNCT
bracis-28433	284	3	association	association	NOUN
bracis-28433	284	4	for	for	ADP
bracis-28433	284	5	computational	computational	ADJ
bracis-28433	284	6	linguistics	linguistic	NOUN
bracis-28433	284	7	,	,	PUNCT
bracis-28433	284	8	portland	portland	PROPN
bracis-28433	284	9	,	,	PUNCT
bracis-28433	284	10	oregon	oregon	PROPN
bracis-28433	284	11	,	,	PUNCT
bracis-28433	284	12	usa	usa	PROPN
bracis-28433	284	13	,	,	PUNCT
bracis-28433	284	14	june	june	PROPN
bracis-28433	284	15	2011	2011	NUM
bracis-28433	284	16	google	google	PROPN
bracis-28433	284	17	scholar	scholar	NOUN
bracis-28433	284	18	  	  	SPACE
bracis-28433	284	19	park	park	NOUN
bracis-28433	284	20	,	,	PUNCT
bracis-28433	284	21	j.	j.	PROPN
bracis-28433	284	22	,	,	PUNCT
bracis-28433	284	23	kim	kim	PROPN
bracis-28433	284	24	,	,	PUNCT
bracis-28433	284	25	g.	g.	PROPN
bracis-28433	284	26	,	,	PUNCT
bracis-28433	284	27	kang	kang	PROPN
bracis-28433	284	28	,	,	PUNCT
bracis-28433	284	29	j.	j.	PROPN
bracis-28433	284	30	:	:	PUNCT
bracis-28433	284	31	consistency	consistency	NOUN
bracis-28433	284	32	training	training	NOUN
bracis-28433	284	33	with	with	ADP
bracis-28433	284	34	virtual	virtual	ADJ
bracis-28433	284	35	adversarial	adversarial	ADJ
bracis-28433	284	36	discrete	discrete	ADJ
bracis-28433	284	37	perturbation	perturbation	NOUN
bracis-28433	284	38	(	(	PUNCT
bracis-28433	284	39	2021	2021	NUM
bracis-28433	284	40	)	)	PUNCT
bracis-28433	284	41	google	google	PROPN
bracis-28433	284	42	scholar	scholar	NOUN
bracis-28433	284	43	  	  	SPACE
bracis-28433	284	44	pennington	pennington	PROPN
bracis-28433	284	45	,	,	PUNCT
bracis-28433	284	46	j.	j.	PROPN
bracis-28433	284	47	,	,	PUNCT
bracis-28433	284	48	socher	socher	PROPN
bracis-28433	284	49	,	,	PUNCT
bracis-28433	284	50	r.	r.	PROPN
bracis-28433	284	51	,	,	PUNCT
bracis-28433	284	52	manning	manning	NOUN
bracis-28433	284	53	,	,	PUNCT
bracis-28433	284	54	c.	c.	NOUN
bracis-28433	284	55	:	:	PUNCT
bracis-28433	284	56	glove	glove	NOUN
bracis-28433	284	57	:	:	PUNCT
bracis-28433	284	58	global	global	ADJ
bracis-28433	284	59	vectors	vector	NOUN
bracis-28433	284	60	for	for	ADP
bracis-28433	284	61	word	word	NOUN
bracis-28433	284	62	representation	representation	NOUN
bracis-28433	284	63	.	.	PUNCT
bracis-28433	285	1	in	in	ADP
bracis-28433	285	2	:	:	PUNCT
bracis-28433	285	3	proceedings	proceeding	NOUN
bracis-28433	285	4	of	of	ADP
bracis-28433	285	5	the	the	DET
bracis-28433	285	6	2014	2014	NUM
bracis-28433	285	7	conference	conference	NOUN
bracis-28433	285	8	on	on	ADP
bracis-28433	285	9	empirical	empirical	ADJ
bracis-28433	285	10	methods	method	NOUN
bracis-28433	285	11	in	in	ADP
bracis-28433	285	12	natural	natural	ADJ
bracis-28433	285	13	language	language	NOUN
bracis-28433	285	14	processing	processing	NOUN
bracis-28433	285	15	(	(	PUNCT
bracis-28433	285	16	emnlp	emnlp	ADJ
bracis-28433	285	17	)	)	PUNCT
bracis-28433	285	18	,	,	PUNCT
bracis-28433	285	19	pp	pp	PROPN
bracis-28433	285	20	.	.	PUNCT
bracis-28433	285	21	1532–1543	1532–1543	NUM
bracis-28433	285	22	.	.	PUNCT
bracis-28433	285	23	association	association	NOUN
bracis-28433	285	24	for	for	ADP
bracis-28433	285	25	computational	computational	ADJ
bracis-28433	285	26	linguistics	linguistic	NOUN
bracis-28433	285	27	,	,	PUNCT
bracis-28433	285	28	doha	doha	PROPN
bracis-28433	285	29	,	,	PUNCT
bracis-28433	285	30	qatar	qatar	PROPN
bracis-28433	285	31	,	,	PUNCT
bracis-28433	285	32	october	october	PROPN
bracis-28433	285	33	2014	2014	NUM
bracis-28433	285	34	.	.	PUNCT
bracis-28433	286	1	https://doi.org/10.3115/v1/d14-1162	https://doi.org/10.3115/v1/d14-1162	NOUN
bracis-28433	286	2	pradhan	pradhan	PROPN
bracis-28433	286	3	,	,	PUNCT
bracis-28433	286	4	s.	s.	PROPN
bracis-28433	286	5	,	,	PUNCT
bracis-28433	286	6	et	et	PROPN
bracis-28433	286	7	al	al	PROPN
bracis-28433	286	8	.	.	PROPN
bracis-28433	286	9	:	:	PUNCT
bracis-28433	287	1	towards	towards	ADP
bracis-28433	287	2	robust	robust	ADJ
bracis-28433	287	3	linguistic	linguistic	ADJ
bracis-28433	287	4	analysis	analysis	NOUN
bracis-28433	287	5	using	use	VERB
bracis-28433	287	6	ontonotes	ontonote	NOUN
bracis-28433	287	7	.	.	PUNCT
bracis-28433	288	1	in	in	ADP
bracis-28433	288	2	:	:	PUNCT
bracis-28433	288	3	proceedings	proceeding	NOUN
bracis-28433	288	4	of	of	ADP
bracis-28433	288	5	the	the	DET
bracis-28433	288	6	seventeenth	seventeenth	ADJ
bracis-28433	288	7	conference	conference	NOUN
bracis-28433	288	8	on	on	ADP
bracis-28433	288	9	computational	computational	ADJ
bracis-28433	288	10	natural	natural	ADJ
bracis-28433	288	11	language	language	NOUN
bracis-28433	288	12	learning	learning	NOUN
bracis-28433	288	13	,	,	PUNCT
bracis-28433	288	14	pp	pp	ADP
bracis-28433	288	15	.	.	PUNCT
bracis-28433	289	1	143–152	143–152	NUM
bracis-28433	289	2	.	.	PUNCT
bracis-28433	289	3	association	association	NOUN
bracis-28433	289	4	for	for	ADP
bracis-28433	289	5	computational	computational	ADJ
bracis-28433	289	6	linguistics	linguistic	NOUN
bracis-28433	289	7	,	,	PUNCT
bracis-28433	289	8	sofia	sofia	PROPN
bracis-28433	289	9	,	,	PUNCT
bracis-28433	289	10	bulgaria	bulgaria	PROPN
bracis-28433	289	11	,	,	PUNCT
bracis-28433	289	12	august	august	PROPN
bracis-28433	289	13	2013	2013	NUM
bracis-28433	289	14	google	google	PROPN
bracis-28433	289	15	scholar	scholar	NOUN
bracis-28433	289	16	  	  	SPACE
bracis-28433	289	17	radmard	radmard	NOUN
bracis-28433	289	18	,	,	PUNCT
bracis-28433	289	19	p.	p.	NOUN
bracis-28433	289	20	,	,	PUNCT
bracis-28433	289	21	fathullah	fathullah	NOUN
bracis-28433	289	22	,	,	PUNCT
bracis-28433	289	23	y.	y.	PROPN
bracis-28433	289	24	,	,	PUNCT
bracis-28433	289	25	lipani	lipani	PROPN
bracis-28433	289	26	,	,	PUNCT
bracis-28433	289	27	a.	a.	NOUN
bracis-28433	289	28	:	:	PUNCT
bracis-28433	289	29	subsequence	subsequence	PROPN
bracis-28433	289	30	based	base	VERB
bracis-28433	289	31	deep	deep	ADJ
bracis-28433	289	32	active	active	ADJ
bracis-28433	289	33	learning	learning	NOUN
bracis-28433	289	34	for	for	ADP
bracis-28433	289	35	named	name	VERB
bracis-28433	289	36	entity	entity	NOUN
bracis-28433	289	37	recognition	recognition	NOUN
bracis-28433	289	38	.	.	PUNCT
bracis-28433	290	1	in	in	ADP
bracis-28433	290	2	:	:	PUNCT
bracis-28433	290	3	proceedings	proceeding	NOUN
bracis-28433	290	4	of	of	ADP
bracis-28433	290	5	the	the	DET
bracis-28433	290	6	59th	59th	ADJ
bracis-28433	290	7	annual	annual	ADJ
bracis-28433	290	8	meeting	meeting	NOUN
bracis-28433	290	9	of	of	ADP
bracis-28433	290	10	the	the	DET
bracis-28433	290	11	association	association	NOUN
bracis-28433	290	12	for	for	ADP
bracis-28433	290	13	computational	computational	ADJ
bracis-28433	290	14	linguistics	linguistic	NOUN
bracis-28433	290	15	and	and	CCONJ
bracis-28433	290	16	the	the	DET
bracis-28433	290	17	11th	11th	ADJ
bracis-28433	290	18	international	international	ADJ
bracis-28433	290	19	joint	joint	ADJ
bracis-28433	290	20	conference	conference	NOUN
bracis-28433	290	21	on	on	ADP
bracis-28433	290	22	natural	natural	ADJ
bracis-28433	290	23	language	language	NOUN
bracis-28433	290	24	processing	processing	NOUN
bracis-28433	290	25	(	(	PUNCT
bracis-28433	290	26	volume	volume	NOUN
bracis-28433	290	27	1	1	NUM
bracis-28433	290	28	:	:	PUNCT
bracis-28433	290	29	long	long	ADJ
bracis-28433	290	30	papers	paper	NOUN
bracis-28433	290	31	)	)	PUNCT
bracis-28433	290	32	,	,	PUNCT
bracis-28433	290	33	pp	pp	PROPN
bracis-28433	290	34	.	.	PUNCT
bracis-28433	290	35	4310–4321	4310–4321	NUM
bracis-28433	290	36	.	.	PUNCT
bracis-28433	290	37	association	association	NOUN
bracis-28433	290	38	for	for	ADP
bracis-28433	290	39	computational	computational	ADJ
bracis-28433	290	40	linguistics	linguistic	NOUN
bracis-28433	290	41	,	,	PUNCT
bracis-28433	290	42	online	online	ADJ
bracis-28433	290	43	,	,	PUNCT
bracis-28433	290	44	august	august	PROPN
bracis-28433	290	45	2021	2021	NUM
bracis-28433	290	46	google	google	PROPN
bracis-28433	290	47	scholar	scholar	NOUN
bracis-28433	290	48	  	  	SPACE
bracis-28433	290	49	sang	sing	VERB
bracis-28433	290	50	,	,	PUNCT
bracis-28433	290	51	e.f.t.k	e.f.t.k	PROPN
bracis-28433	290	52	.	.	PROPN
bracis-28433	290	53	,	,	PUNCT
bracis-28433	290	54	meulder	meulder	NOUN
bracis-28433	290	55	,	,	PUNCT
bracis-28433	290	56	f.d	f.d	PROPN
bracis-28433	290	57	.	.	PUNCT
bracis-28433	290	58	:	:	PUNCT
bracis-28433	290	59	introduction	introduction	NOUN
bracis-28433	290	60	to	to	ADP
bracis-28433	290	61	the	the	DET
bracis-28433	290	62	conll-2003	conll-2003	NOUN
bracis-28433	290	63	shared	share	VERB
bracis-28433	290	64	task	task	NOUN
bracis-28433	290	65	:	:	PUNCT
bracis-28433	290	66	language	language	NOUN
bracis-28433	290	67	-	-	PUNCT
bracis-28433	290	68	independent	independent	NOUN
bracis-28433	290	69	named	name	VERB
bracis-28433	290	70	entity	entity	NOUN
bracis-28433	290	71	recognition	recognition	NOUN
bracis-28433	290	72	.	.	PUNCT
bracis-28433	291	1	in	in	ADP
bracis-28433	291	2	:	:	PUNCT
bracis-28433	291	3	proceedings	proceeding	NOUN
bracis-28433	291	4	of	of	ADP
bracis-28433	291	5	the	the	DET
bracis-28433	291	6	seventh	seventh	ADJ
bracis-28433	291	7	conference	conference	NOUN
bracis-28433	291	8	on	on	ADP
bracis-28433	291	9	natural	natural	ADJ
bracis-28433	291	10	language	language	NOUN
bracis-28433	291	11	learning	learning	NOUN
bracis-28433	291	12	at	at	ADP
bracis-28433	291	13	hlt	hlt	PROPN
bracis-28433	291	14	-	-	PUNCT
bracis-28433	291	15	naacl	naacl	PROPN
bracis-28433	291	16	2003	2003	NUM
bracis-28433	291	17	,	,	PUNCT
bracis-28433	291	18	pp	pp	ADP
bracis-28433	291	19	.	.	PUNCT
bracis-28433	292	1	142–147	142–147	NUM
bracis-28433	292	2	(	(	PUNCT
bracis-28433	292	3	2003	2003	NUM
bracis-28433	292	4	)	)	PUNCT
bracis-28433	292	5	google	google	PROPN
bracis-28433	292	6	scholar	scholar	NOUN
bracis-28433	292	7	  	  	SPACE
bracis-28433	292	8	shen	shen	NOUN
bracis-28433	292	9	,	,	PUNCT
bracis-28433	292	10	y.	y.	PROPN
bracis-28433	292	11	,	,	PUNCT
bracis-28433	292	12	yun	yun	PROPN
bracis-28433	292	13	,	,	PUNCT
bracis-28433	292	14	h.	h.	PROPN
bracis-28433	292	15	,	,	PUNCT
bracis-28433	292	16	lipton	lipton	PROPN
bracis-28433	292	17	,	,	PUNCT
bracis-28433	292	18	z.	z.	PROPN
bracis-28433	292	19	,	,	PUNCT
bracis-28433	292	20	kronrod	kronrod	PROPN
bracis-28433	292	21	,	,	PUNCT
bracis-28433	292	22	y.	y.	PROPN
bracis-28433	292	23	,	,	PUNCT
bracis-28433	292	24	anandkumar	anandkumar	PROPN
bracis-28433	292	25	,	,	PUNCT
bracis-28433	292	26	a.	a.	NOUN
bracis-28433	292	27	:	:	PUNCT
bracis-28433	292	28	deep	deep	ADJ
bracis-28433	292	29	active	active	ADJ
bracis-28433	292	30	learning	learning	NOUN
bracis-28433	292	31	for	for	ADP
bracis-28433	292	32	named	name	VERB
bracis-28433	292	33	entity	entity	NOUN
bracis-28433	292	34	recognition	recognition	NOUN
bracis-28433	292	35	.	.	PUNCT
bracis-28433	293	1	in	in	ADP
bracis-28433	293	2	:	:	PUNCT
bracis-28433	293	3	proceedings	proceeding	NOUN
bracis-28433	293	4	of	of	ADP
bracis-28433	293	5	the	the	DET
bracis-28433	293	6	2nd	2nd	ADJ
bracis-28433	293	7	workshop	workshop	NOUN
bracis-28433	293	8	on	on	ADP
bracis-28433	293	9	representation	representation	NOUN
bracis-28433	293	10	learning	learning	NOUN
bracis-28433	293	11	for	for	ADP
bracis-28433	293	12	nlp	nlp	NOUN
bracis-28433	293	13	,	,	PUNCT
bracis-28433	293	14	pp	pp	ADJ
bracis-28433	293	15	.	.	PUNCT
bracis-28433	294	1	252–256	252–256	NUM
bracis-28433	294	2	.	.	PUNCT
bracis-28433	294	3	association	association	NOUN
bracis-28433	294	4	for	for	ADP
bracis-28433	294	5	computational	computational	ADJ
bracis-28433	294	6	linguistics	linguistic	NOUN
bracis-28433	294	7	,	,	PUNCT
bracis-28433	294	8	vancouver	vancouver	PROPN
bracis-28433	294	9	,	,	PUNCT
bracis-28433	294	10	canada	canada	PROPN
bracis-28433	294	11	,	,	PUNCT
bracis-28433	294	12	august	august	PROPN
bracis-28433	294	13	2017	2017	NUM
bracis-28433	294	14	.	.	PUNCT
bracis-28433	295	1	https://doi.org/10.18653/v1/w17-2630	https://doi.org/10.18653/v1/w17-2630	PROPN
bracis-28433	295	2	siddhant	siddhant	NOUN
bracis-28433	295	3	,	,	PUNCT
bracis-28433	295	4	a.	a.	PROPN
bracis-28433	295	5	,	,	PUNCT
bracis-28433	295	6	lipton	lipton	PROPN
bracis-28433	295	7	,	,	PUNCT
bracis-28433	295	8	z.c	z.c	PROPN
bracis-28433	295	9	.	.	PROPN
bracis-28433	295	10	:	:	PUNCT
bracis-28433	296	1	deep	deep	ADJ
bracis-28433	296	2	bayesian	bayesian	NOUN
bracis-28433	296	3	active	active	ADJ
bracis-28433	296	4	learning	learning	NOUN
bracis-28433	296	5	for	for	ADP
bracis-28433	296	6	natural	natural	ADJ
bracis-28433	296	7	language	language	NOUN
bracis-28433	296	8	processing	processing	NOUN
bracis-28433	296	9	:	:	PUNCT
bracis-28433	296	10	results	result	NOUN
bracis-28433	296	11	of	of	ADP
bracis-28433	296	12	a	a	DET
bracis-28433	296	13	large	large	ADJ
bracis-28433	296	14	-	-	PUNCT
bracis-28433	296	15	scale	scale	NOUN
bracis-28433	296	16	empirical	empirical	ADJ
bracis-28433	296	17	study	study	NOUN
bracis-28433	296	18	.	.	PUNCT
bracis-28433	297	1	in	in	ADP
bracis-28433	297	2	:	:	PUNCT
bracis-28433	297	3	proceedings	proceeding	NOUN
bracis-28433	297	4	of	of	ADP
bracis-28433	297	5	the	the	DET
bracis-28433	297	6	2018	2018	NUM
bracis-28433	297	7	conference	conference	NOUN
bracis-28433	297	8	on	on	ADP
bracis-28433	297	9	empirical	empirical	ADJ
bracis-28433	297	10	methods	method	NOUN
bracis-28433	297	11	in	in	ADP
bracis-28433	297	12	natural	natural	ADJ
bracis-28433	297	13	language	language	NOUN
bracis-28433	297	14	processing	processing	NOUN
bracis-28433	297	15	,	,	PUNCT
bracis-28433	297	16	pp	pp	ADJ
bracis-28433	297	17	.	.	PUNCT
bracis-28433	297	18	2904–2909	2904–2909	NUM
bracis-28433	297	19	.	.	PUNCT
bracis-28433	298	1	association	association	NOUN
bracis-28433	298	2	for	for	ADP
bracis-28433	298	3	computational	computational	ADJ
bracis-28433	298	4	linguistics	linguistic	NOUN
bracis-28433	298	5	,	,	PUNCT
bracis-28433	298	6	brussels	brussels	PROPN
bracis-28433	298	7	,	,	PUNCT
bracis-28433	298	8	belgium	belgium	PROPN
bracis-28433	298	9	,	,	PUNCT
bracis-28433	298	10	october	october	PROPN
bracis-28433	298	11	–	–	PUNCT
bracis-28433	298	12	november	november	PROPN
bracis-28433	298	13	2018	2018	NUM
bracis-28433	298	14	.	.	PUNCT
bracis-28433	299	1	https://doi.org/10.18653/v1/d18-1318	https://doi.org/10.18653/v1/d18-1318	PROPN
bracis-28433	299	2	srivastava	srivastava	PROPN
bracis-28433	299	3	,	,	PUNCT
bracis-28433	299	4	n.	n.	PROPN
bracis-28433	299	5	,	,	PUNCT
bracis-28433	299	6	hinton	hinton	PROPN
bracis-28433	299	7	,	,	PUNCT
bracis-28433	299	8	g.e	g.e	PROPN
bracis-28433	299	9	.	.	PROPN
bracis-28433	299	10	,	,	PUNCT
bracis-28433	299	11	krizhevsky	krizhevsky	PROPN
bracis-28433	299	12	,	,	PUNCT
bracis-28433	299	13	a.	a.	NOUN
bracis-28433	299	14	,	,	PUNCT
bracis-28433	299	15	sutskever	sutskever	PROPN
bracis-28433	299	16	,	,	PUNCT
bracis-28433	299	17	i.	i.	PROPN
bracis-28433	299	18	,	,	PUNCT
bracis-28433	299	19	salakhutdinov	salakhutdinov	PROPN
bracis-28433	299	20	,	,	PUNCT
bracis-28433	299	21	r.	r.	NOUN
bracis-28433	299	22	:	:	PUNCT
bracis-28433	299	23	dropout	dropout	NOUN
bracis-28433	299	24	:	:	PUNCT
bracis-28433	299	25	a	a	DET
bracis-28433	299	26	simple	simple	ADJ
bracis-28433	299	27	way	way	NOUN
bracis-28433	299	28	to	to	PART
bracis-28433	299	29	prevent	prevent	VERB
bracis-28433	299	30	neural	neural	ADJ
bracis-28433	299	31	networks	network	NOUN
bracis-28433	299	32	from	from	ADP
bracis-28433	299	33	overfitting	overfitte	VERB
bracis-28433	299	34	.	.	PUNCT
bracis-28433	300	1	j.	j.	PROPN
bracis-28433	300	2	mach	mach	PROPN
bracis-28433	300	3	.	.	PUNCT
bracis-28433	301	1	learn	learn	VERB
bracis-28433	301	2	.	.	PUNCT
bracis-28433	302	1	res	re	NOUN
bracis-28433	302	2	.	.	PROPN
bracis-28433	302	3	15(1	15(1	NUM
bracis-28433	302	4	)	)	PUNCT
bracis-28433	302	5	,	,	PUNCT
bracis-28433	302	6	1929–1958	1929–1958	NUM
bracis-28433	302	7	(	(	PUNCT
bracis-28433	302	8	2014	2014	NUM
bracis-28433	302	9	)	)	PUNCT
bracis-28433	302	10	mathscinet	mathscinet	NOUN
bracis-28433	302	11	  	  	SPACE
bracis-28433	302	12	math	math	NOUN
bracis-28433	302	13	  	  	SPACE
bracis-28433	302	14	google	google	PROPN
bracis-28433	302	15	scholar	scholar	NOUN
bracis-28433	302	16	  	  	SPACE
bracis-28433	302	17	tran	tran	NOUN
bracis-28433	302	18	,	,	PUNCT
bracis-28433	302	19	v.c	v.c	PROPN
bracis-28433	302	20	.	.	PROPN
bracis-28433	302	21	,	,	PUNCT
bracis-28433	302	22	nguyen	nguyen	PROPN
bracis-28433	302	23	,	,	PUNCT
bracis-28433	302	24	n.t	n.t	PROPN
bracis-28433	302	25	.	.	PROPN
bracis-28433	302	26	,	,	PUNCT
bracis-28433	302	27	fujita	fujita	PROPN
bracis-28433	302	28	,	,	PUNCT
bracis-28433	302	29	h.	h.	PROPN
bracis-28433	302	30	,	,	PUNCT
bracis-28433	302	31	hoang	hoang	PROPN
bracis-28433	302	32	,	,	PUNCT
bracis-28433	302	33	d.t	d.t	PROPN
bracis-28433	302	34	.	.	PROPN
bracis-28433	302	35	,	,	PUNCT
bracis-28433	302	36	hwang	hwang	PROPN
bracis-28433	302	37	,	,	PUNCT
bracis-28433	302	38	d.	d.	PROPN
bracis-28433	302	39	:	:	PUNCT
bracis-28433	302	40	a	a	DET
bracis-28433	302	41	combination	combination	NOUN
bracis-28433	302	42	of	of	ADP
bracis-28433	302	43	active	active	ADJ
bracis-28433	302	44	learning	learning	NOUN
bracis-28433	302	45	and	and	CCONJ
bracis-28433	302	46	self	self	NOUN
bracis-28433	302	47	-	-	PUNCT
bracis-28433	302	48	learning	learning	NOUN
bracis-28433	302	49	for	for	ADP
bracis-28433	302	50	named	name	VERB
bracis-28433	302	51	entity	entity	NOUN
bracis-28433	302	52	recognition	recognition	NOUN
bracis-28433	302	53	on	on	ADP
bracis-28433	302	54	twitter	twitter	NOUN
bracis-28433	302	55	using	use	VERB
bracis-28433	302	56	conditional	conditional	ADJ
bracis-28433	302	57	random	random	ADJ
bracis-28433	302	58	fields	field	NOUN
bracis-28433	302	59	.	.	PUNCT
bracis-28433	303	1	knowl.-based	knowl.-based	PROPN
bracis-28433	303	2	syst	syst	PROPN
bracis-28433	303	3	.	.	PUNCT
bracis-28433	304	1	132	132	NUM
bracis-28433	304	2	,	,	PUNCT
bracis-28433	304	3	179–187	179–187	NUM
bracis-28433	304	4	(	(	PUNCT
bracis-28433	304	5	2017	2017	NUM
bracis-28433	304	6	)	)	PUNCT
bracis-28433	304	7	article	article	NOUN
bracis-28433	304	8	  	  	SPACE
bracis-28433	304	9	google	google	PROPN
bracis-28433	304	10	scholar	scholar	NOUN
bracis-28433	304	11	  	  	SPACE
bracis-28433	304	12	download	download	NOUN
bracis-28433	304	13	references	reference	NOUN
bracis-28433	304	14	acknowledgements	acknowledgement	NOUN
bracis-28433	304	15	the	the	DET
bracis-28433	304	16	authors	author	NOUN
bracis-28433	304	17	were	be	AUX
bracis-28433	304	18	supported	support	VERB
bracis-28433	304	19	by	by	ADP
bracis-28433	304	20	the	the	DET
bracis-28433	304	21	fundação	fundação	NOUN
bracis-28433	304	22	de	de	PROPN
bracis-28433	304	23	apoio	apoio	NOUN
bracis-28433	304	24	a	a	DET
bracis-28433	304	25	pesquisa	pesquisa	NOUN
bracis-28433	304	26	do	do	AUX
bracis-28433	304	27	distritio	distritio	VERB
bracis-28433	304	28	federal	federal	ADJ
bracis-28433	304	29	(	(	PUNCT
bracis-28433	304	30	fap	fap	NOUN
bracis-28433	304	31	-	-	PUNCT
bracis-28433	304	32	df	df	NOUN
bracis-28433	304	33	)	)	PUNCT
bracis-28433	304	34	as	as	ADP
bracis-28433	304	35	members	member	NOUN
bracis-28433	304	36	of	of	ADP
bracis-28433	304	37	the	the	DET
bracis-28433	304	38	knowledge	knowledge	NOUN
bracis-28433	304	39	extraction	extraction	NOUN
bracis-28433	304	40	from	from	ADP
bracis-28433	304	41	documents	document	NOUN
bracis-28433	304	42	of	of	ADP
bracis-28433	304	43	legal	legal	ADJ
bracis-28433	304	44	content	content	NOUN
bracis-28433	304	45	(	(	PUNCT
bracis-28433	304	46	knedle	knedle	NOUN
bracis-28433	304	47	)	)	PUNCT
bracis-28433	304	48	project	project	NOUN
bracis-28433	304	49	from	from	ADP
bracis-28433	304	50	the	the	DET
bracis-28433	304	51	university	university	PROPN
bracis-28433	304	52	of	of	ADP
bracis-28433	304	53	brasilia	brasilia	PROPN
bracis-28433	304	54	.	.	PUNCT
bracis-28433	305	1	author	author	NOUN
bracis-28433	305	2	information	information	NOUN
bracis-28433	305	3	authors	author	NOUN
bracis-28433	305	4	and	and	CCONJ
bracis-28433	305	5	affiliations	affiliation	NOUN
bracis-28433	305	6	osaka	osaka	PROPN
bracis-28433	305	7	university	university	PROPN
bracis-28433	305	8	,	,	PUNCT
bracis-28433	305	9	osaka	osaka	PROPN
bracis-28433	305	10	,	,	PUNCT
bracis-28433	305	11	japan	japan	PROPN
bracis-28433	305	12	josé	josé	PROPN
bracis-28433	305	13	reinaldo	reinaldo	PROPN
bracis-28433	305	14	cunha	cunha	VERB
bracis-28433	305	15	santos	santos	PROPN
bracis-28433	305	16	a.	a.	PROPN
bracis-28433	305	17	v.	v.	PROPN
bracis-28433	305	18	silva	silva	PROPN
bracis-28433	305	19	neto	neto	PROPN
bracis-28433	305	20	university	university	PROPN
bracis-28433	305	21	of	of	ADP
bracis-28433	305	22	brasilia	brasilia	PROPN
bracis-28433	305	23	,	,	PUNCT
bracis-28433	305	24	brasilia	brasilia	PROPN
bracis-28433	305	25	,	,	PUNCT
bracis-28433	305	26	df	df	PROPN
bracis-28433	305	27	,	,	PUNCT
bracis-28433	305	28	brazil	brazil	PROPN
bracis-28433	305	29	thiago	thiago	PROPN
bracis-28433	306	1	de	de	PROPN
bracis-28433	306	2	paulo	paulo	PROPN
bracis-28433	306	3	faleiros	faleiros	PROPN
bracis-28433	306	4	authors	author	NOUN
bracis-28433	306	5	josé	josé	PROPN
bracis-28433	306	6	reinaldo	reinaldo	PROPN
bracis-28433	306	7	cunha	cunha	VERB
bracis-28433	306	8	santos	santos	PROPN
bracis-28433	306	9	a.	a.	PROPN
bracis-28433	306	10	v.	v.	PROPN
bracis-28433	306	11	silva	silva	PROPN
bracis-28433	306	12	netoview	netoview	PROPN
bracis-28433	306	13	author	author	NOUN
bracis-28433	306	14	publications	publication	NOUN
bracis-28433	306	15	search	search	NOUN
bracis-28433	306	16	author	author	NOUN
bracis-28433	306	17	on	on	ADP
bracis-28433	306	18	:	:	PUNCT
bracis-28433	306	19	pubmed	pubmed	PROPN
bracis-28433	306	20	 	 	SPACE
bracis-28433	306	21	google	google	PROPN
bracis-28433	306	22	scholar	scholar	NOUN
bracis-28433	306	23	thiago	thiago	PROPN
bracis-28433	306	24	de	de	PROPN
bracis-28433	306	25	paulo	paulo	PROPN
bracis-28433	306	26	faleirosview	faleirosview	PROPN
bracis-28433	306	27	author	author	NOUN
bracis-28433	306	28	publications	publication	NOUN
bracis-28433	306	29	search	search	NOUN
bracis-28433	306	30	author	author	NOUN
bracis-28433	306	31	on	on	ADP
bracis-28433	306	32	:	:	PUNCT
bracis-28433	306	33	pubmed	pubmed	PROPN
bracis-28433	306	34	 	 	SPACE
bracis-28433	306	35	google	google	PROPN
bracis-28433	306	36	scholar	scholar	NOUN
bracis-28433	306	37	corresponding	correspond	VERB
bracis-28433	306	38	author	author	NOUN
bracis-28433	306	39	correspondence	correspondence	NOUN
bracis-28433	306	40	to	to	ADP
bracis-28433	306	41	josé	josé	PROPN
bracis-28433	306	42	reinaldo	reinaldo	PROPN
bracis-28433	307	1	cunha	cunha	PROPN
bracis-28433	307	2	santos	santos	PROPN
bracis-28433	307	3	a.	a.	PROPN
bracis-28433	307	4	v.	v.	PROPN
bracis-28433	307	5	silva	silva	PROPN
bracis-28433	307	6	neto	neto	PROPN
bracis-28433	307	7	.	.	PUNCT
bracis-28433	308	1	editor	editor	NOUN
bracis-28433	308	2	information	information	NOUN
bracis-28433	308	3	editors	editor	NOUN
bracis-28433	308	4	and	and	CCONJ
bracis-28433	308	5	affiliations	affiliation	NOUN
bracis-28433	308	6	federal	federal	PROPN
bracis-28433	308	7	university	university	PROPN
bracis-28433	308	8	of	of	ADP
bracis-28433	308	9	são	são	PROPN
bracis-28433	308	10	carlos	carlos	PROPN
bracis-28433	308	11	,	,	PUNCT
bracis-28433	308	12	são	são	PROPN
bracis-28433	308	13	carlos	carlos	PROPN
bracis-28433	308	14	,	,	PUNCT
bracis-28433	308	15	brazil	brazil	PROPN
bracis-28433	308	16	murilo	murilo	PROPN
bracis-28433	308	17	c.	c.	PROPN
bracis-28433	308	18	naldi	naldi	PROPN
bracis-28433	308	19	centro	centro	PROPN
bracis-28433	308	20	universitario	universitario	PROPN
bracis-28433	308	21	da	da	PROPN
bracis-28433	308	22	fei	fei	PROPN
bracis-28433	308	23	,	,	PUNCT
bracis-28433	308	24	são	são	PROPN
bracis-28433	308	25	bernardo	bernardo	PROPN
bracis-28433	308	26	do	do	AUX
bracis-28433	308	27	campo	campo	PROPN
bracis-28433	308	28	,	,	PUNCT
bracis-28433	308	29	brazil	brazil	PROPN
bracis-28433	308	30	reinaldo	reinaldo	PROPN
bracis-28433	308	31	a.	a.	PROPN
bracis-28433	308	32	c.	c.	PROPN
bracis-28433	308	33	bianchi	bianchi	PROPN
bracis-28433	308	34	rights	right	NOUN
bracis-28433	308	35	and	and	CCONJ
bracis-28433	308	36	permissions	permission	NOUN
bracis-28433	308	37	reprints	reprint	NOUN
bracis-28433	308	38	and	and	CCONJ
bracis-28433	308	39	permissions	permission	VERB
bracis-28433	308	40	copyright	copyright	NOUN
bracis-28433	308	41	information	information	NOUN
bracis-28433	308	42	©	©	ADP
bracis-28433	308	43	2023	2023	NUM
bracis-28433	308	44	the	the	DET
bracis-28433	308	45	author(s	author(s	NOUN
bracis-28433	308	46	)	)	PUNCT
bracis-28433	308	47	,	,	PUNCT
bracis-28433	308	48	under	under	ADP
bracis-28433	308	49	exclusive	exclusive	ADJ
bracis-28433	308	50	license	license	NOUN
bracis-28433	308	51	to	to	ADP
bracis-28433	308	52	springer	springer	NOUN
bracis-28433	308	53	nature	nature	PROPN
bracis-28433	308	54	switzerland	switzerland	PROPN
bracis-28433	308	55	ag	ag	PROPN
bracis-28433	308	56	about	about	ADP
bracis-28433	308	57	this	this	DET
bracis-28433	308	58	paper	paper	NOUN
bracis-28433	308	59	cite	cite	VERB
bracis-28433	308	60	this	this	DET
bracis-28433	308	61	paper	paper	NOUN
bracis-28433	308	62	cunha	cunha	VERB
bracis-28433	308	63	santos	santos	PROPN
bracis-28433	308	64	a.	a.	PROPN
bracis-28433	308	65	v.	v.	PROPN
bracis-28433	308	66	silva	silva	PROPN
bracis-28433	308	67	neto	neto	PROPN
bracis-28433	308	68	,	,	PUNCT
bracis-28433	308	69	j.r	j.r	PROPN
bracis-28433	308	70	.	.	PROPN
bracis-28433	308	71	,	,	PUNCT
bracis-28433	308	72	de	de	PROPN
bracis-28433	308	73	paulo	paulo	PROPN
bracis-28433	308	74	faleiros	faleiros	PROPN
bracis-28433	308	75	,	,	PUNCT
bracis-28433	308	76	t.	t.	PROPN
bracis-28433	308	77	(	(	PUNCT
bracis-28433	308	78	2023	2023	NUM
bracis-28433	308	79	)	)	PUNCT
bracis-28433	308	80	.	.	PUNCT
bracis-28433	309	1	investigation	investigation	NOUN
bracis-28433	309	2	of	of	ADP
bracis-28433	309	3	 	 	SPACE
bracis-28433	309	4	deep	deep	ADJ
bracis-28433	309	5	active	active	ADJ
bracis-28433	309	6	self	self	NOUN
bracis-28433	309	7	-	-	PUNCT
bracis-28433	309	8	learning	learn	VERB
bracis-28433	309	9	algorithms	algorithm	NOUN
bracis-28433	309	10	applied	apply	VERB
bracis-28433	309	11	to	to	ADP
bracis-28433	309	12	 	 	SPACE
bracis-28433	309	13	named	name	VERB
bracis-28433	309	14	entity	entity	NOUN
bracis-28433	309	15	recognition	recognition	NOUN
bracis-28433	309	16	.	.	PUNCT
bracis-28433	310	1	in	in	ADP
bracis-28433	310	2	:	:	PUNCT
bracis-28433	310	3	naldi	naldi	PROPN
bracis-28433	310	4	,	,	PUNCT
bracis-28433	310	5	m.c	m.c	PROPN
bracis-28433	310	6	.	.	PROPN
bracis-28433	310	7	,	,	PUNCT
bracis-28433	310	8	bianchi	bianchi	PROPN
bracis-28433	310	9	,	,	PUNCT
bracis-28433	310	10	r.a.c	r.a.c	ADP
bracis-28433	310	11	.	.	PUNCT
bracis-28433	310	12	(	(	PUNCT
bracis-28433	310	13	eds	ed	NOUN
bracis-28433	310	14	)	)	PUNCT
bracis-28433	310	15	intelligent	intelligent	ADJ
bracis-28433	310	16	systems	system	NOUN
bracis-28433	310	17	.	.	PUNCT
bracis-28433	311	1	bracis	bracis	PROPN
bracis-28433	311	2	2023	2023	NUM
bracis-28433	311	3	.	.	PUNCT
bracis-28433	312	1	lecture	lecture	NOUN
bracis-28433	312	2	notes	note	NOUN
bracis-28433	312	3	in	in	ADP
bracis-28433	312	4	computer	computer	NOUN
bracis-28433	312	5	science	science	NOUN
bracis-28433	312	6	(	(	PUNCT
bracis-28433	312	7	)	)	PUNCT
bracis-28433	312	8	,	,	PUNCT
bracis-28433	312	9	vol	vol	NOUN
bracis-28433	312	10	14197	14197	NUM
bracis-28433	312	11	.	.	PUNCT
bracis-28433	313	1	springer	springer	NOUN
bracis-28433	313	2	,	,	PUNCT
bracis-28433	313	3	cham	cham	PROPN
bracis-28433	313	4	.	.	PUNCT
bracis-28433	314	1	https://doi.org/10.1007/978-3-031-45392-2_31	https://doi.org/10.1007/978-3-031-45392-2_31	PROPN
bracis-28433	314	2	download	download	NOUN
bracis-28433	314	3	citation	citation	NOUN
bracis-28433	314	4	.ris	.ris	PUNCT
bracis-28433	315	1	.enw	.enw	PROPN
bracis-28433	315	2	.bib	.bib	PUNCT
bracis-28433	316	1	doi	doi	PROPN
bracis-28433	316	2	:	:	PUNCT
bracis-28433	316	3	https://doi.org/10.1007/978-3-031-45392-2_31	https://doi.org/10.1007/978-3-031-45392-2_31	VERB
bracis-28433	316	4	published	publish	VERB
bracis-28433	316	5	:	:	PUNCT
bracis-28433	316	6	12	12	NUM
bracis-28433	316	7	october	october	PROPN
bracis-28433	316	8	2023	2023	NUM
bracis-28433	316	9	publisher	publisher	NOUN
bracis-28433	316	10	name	name	NOUN
bracis-28433	316	11	:	:	PUNCT
bracis-28433	316	12	springer	springer	NOUN
bracis-28433	316	13	,	,	PUNCT
bracis-28433	316	14	cham	cham	PROPN
bracis-28433	316	15	print	print	PROPN
bracis-28433	316	16	isbn	isbn	PROPN
bracis-28433	316	17	:	:	PUNCT
bracis-28433	316	18	978	978	NUM
bracis-28433	316	19	-	-	SYM
bracis-28433	316	20	3	3	NUM
bracis-28433	316	21	-	-	PUNCT
bracis-28433	316	22	031	031	NUM
bracis-28433	316	23	-	-	PUNCT
bracis-28433	316	24	45391	45391	NUM
bracis-28433	316	25	-	-	SYM
bracis-28433	316	26	5	5	NUM
bracis-28433	316	27	online	online	ADJ
bracis-28433	316	28	isbn	isbn	NOUN
bracis-28433	316	29	:	:	PUNCT
bracis-28433	316	30	978	978	NUM
bracis-28433	316	31	-	-	SYM
bracis-28433	316	32	3	3	NUM
bracis-28433	316	33	-	-	PUNCT
bracis-28433	316	34	031	031	NUM
bracis-28433	316	35	-	-	PUNCT
bracis-28433	316	36	45392	45392	NUM
bracis-28433	316	37	-	-	SYM
bracis-28433	316	38	2	2	NUM
bracis-28433	316	39	ebook	ebook	NOUN
bracis-28433	316	40	packages	package	NOUN
bracis-28433	316	41	:	:	PUNCT
bracis-28433	316	42	computer	computer	NOUN
bracis-28433	316	43	sciencecomputer	sciencecomputer	NOUN
bracis-28433	316	44	science	science	NOUN
bracis-28433	316	45	(	(	PUNCT
bracis-28433	316	46	r0	r0	NOUN
bracis-28433	316	47	)	)	PUNCT
bracis-28433	316	48	share	share	VERB
bracis-28433	316	49	this	this	DET
bracis-28433	316	50	paper	paper	NOUN
bracis-28433	316	51	anyone	anyone	PRON
bracis-28433	316	52	you	you	PRON
bracis-28433	316	53	share	share	VERB
bracis-28433	316	54	the	the	DET
bracis-28433	316	55	following	follow	VERB
bracis-28433	316	56	link	link	NOUN
bracis-28433	316	57	with	with	ADP
bracis-28433	316	58	will	will	AUX
bracis-28433	316	59	be	be	AUX
bracis-28433	316	60	able	able	ADJ
bracis-28433	316	61	to	to	PART
bracis-28433	316	62	read	read	VERB
bracis-28433	316	63	this	this	DET
bracis-28433	316	64	content	content	NOUN
bracis-28433	316	65	:	:	PUNCT
bracis-28433	316	66	get	get	VERB
bracis-28433	316	67	shareable	shareable	ADJ
bracis-28433	316	68	linksorry	linksorry	NOUN
bracis-28433	316	69	,	,	PUNCT
bracis-28433	316	70	a	a	DET
bracis-28433	316	71	shareable	shareable	ADJ
bracis-28433	316	72	link	link	NOUN
bracis-28433	316	73	is	be	AUX
bracis-28433	316	74	not	not	PART
bracis-28433	316	75	currently	currently	ADV
bracis-28433	316	76	available	available	ADJ
bracis-28433	316	77	for	for	ADP
bracis-28433	316	78	this	this	DET
bracis-28433	316	79	article	article	NOUN
bracis-28433	316	80	.	.	PUNCT
bracis-28433	317	1	copy	copy	VERB
bracis-28433	317	2	shareable	shareable	ADJ
bracis-28433	317	3	link	link	NOUN
bracis-28433	317	4	to	to	PART
bracis-28433	317	5	clipboard	clipboard	NOUN
bracis-28433	317	6	provided	provide	VERB
bracis-28433	317	7	by	by	ADP
bracis-28433	317	8	the	the	DET
bracis-28433	317	9	springer	springer	NOUN
bracis-28433	317	10	nature	nature	PROPN
bracis-28433	317	11	sharedit	sharedit	PROPN
bracis-28433	317	12	content	content	NOUN
bracis-28433	317	13	-	-	PUNCT
bracis-28433	317	14	sharing	share	VERB
bracis-28433	317	15	initiative	initiative	NOUN
bracis-28433	317	16	keywords	keyword	VERB
bracis-28433	317	17	active	active	ADJ
bracis-28433	317	18	learning	learn	VERB
bracis-28433	317	19	active	active	ADJ
bracis-28433	317	20	self	self	NOUN
bracis-28433	317	21	-	-	PUNCT
bracis-28433	317	22	learning	learn	VERB
bracis-28433	317	23	named	name	VERB
bracis-28433	317	24	entity	entity	NOUN
bracis-28433	317	25	recognition	recognition	NOUN
bracis-28433	317	26	deep	deep	ADJ
bracis-28433	317	27	learning	learning	NOUN
bracis-28433	317	28	publish	publish	VERB
bracis-28433	317	29	with	with	ADP
bracis-28433	317	30	us	us	PROPN
bracis-28433	317	31	policies	policy	NOUN
bracis-28433	317	32	and	and	CCONJ
bracis-28433	317	33	ethics	ethic	NOUN
bracis-28433	317	34	profiles	profile	NOUN
bracis-28433	317	35	thiago	thiago	PROPN
bracis-28433	317	36	de	de	PROPN
bracis-28433	317	37	paulo	paulo	PROPN
bracis-28433	317	38	faleiros	faleiros	PROPN
bracis-28433	317	39	view	view	NOUN
bracis-28433	317	40	author	author	NOUN
bracis-28433	317	41	profile	profile	NOUN
bracis-28433	317	42	search	search	NOUN
bracis-28433	317	43	search	search	NOUN
bracis-28433	317	44	by	by	ADP
bracis-28433	317	45	keyword	keyword	NOUN
bracis-28433	317	46	or	or	CCONJ
bracis-28433	317	47	author	author	NOUN
bracis-28433	317	48	search	search	NOUN
bracis-28433	317	49	navigation	navigation	NOUN
bracis-28433	317	50	find	find	VERB
bracis-28433	317	51	a	a	DET
bracis-28433	317	52	journal	journal	NOUN
bracis-28433	317	53	publish	publish	VERB
bracis-28433	317	54	with	with	ADP
bracis-28433	317	55	us	we	PRON
bracis-28433	317	56	track	track	VERB
bracis-28433	317	57	your	your	PRON
bracis-28433	317	58	research	research	NOUN
bracis-28433	317	59	discover	discover	VERB
bracis-28433	317	60	content	content	NOUN
bracis-28433	317	61	journals	journal	NOUN
bracis-28433	317	62	a	a	DET
bracis-28433	317	63	-	-	PUNCT
bracis-28433	317	64	z	z	NOUN
bracis-28433	317	65	books	book	NOUN
bracis-28433	317	66	a	a	DET
bracis-28433	317	67	-	-	PUNCT
bracis-28433	317	68	z	z	NOUN
bracis-28433	317	69	publish	publish	NOUN
bracis-28433	317	70	with	with	ADP
bracis-28433	317	71	us	us	PROPN
bracis-28433	317	72	journal	journal	PROPN
bracis-28433	317	73	finder	finder	PROPN
bracis-28433	317	74	publish	publish	VERB
bracis-28433	317	75	your	your	PRON
bracis-28433	317	76	research	research	NOUN
bracis-28433	317	77	language	language	NOUN
bracis-28433	317	78	editing	edit	VERB
bracis-28433	317	79	open	open	ADJ
bracis-28433	317	80	access	access	NOUN
bracis-28433	317	81	publishing	publishing	NOUN
bracis-28433	317	82	products	product	NOUN
bracis-28433	317	83	and	and	CCONJ
bracis-28433	317	84	services	service	NOUN
bracis-28433	317	85	our	our	PRON
bracis-28433	317	86	products	product	NOUN
bracis-28433	317	87	librarians	librarian	VERB
bracis-28433	317	88	societies	society	NOUN
bracis-28433	317	89	partners	partner	NOUN
bracis-28433	317	90	and	and	CCONJ
bracis-28433	317	91	advertisers	advertiser	NOUN
bracis-28433	317	92	our	our	PRON
bracis-28433	317	93	brands	brand	NOUN
bracis-28433	317	94	springer	springer	NOUN
bracis-28433	317	95	nature	nature	PROPN
bracis-28433	317	96	portfolio	portfolio	PROPN
bracis-28433	317	97	bmc	bmc	PROPN
bracis-28433	317	98	palgrave	palgrave	PROPN
bracis-28433	317	99	macmillan	macmillan	PROPN
bracis-28433	317	100	apress	apress	PROPN
bracis-28433	317	101	discover	discover	VERB
bracis-28433	317	102	your	your	PRON
bracis-28433	317	103	privacy	privacy	NOUN
bracis-28433	317	104	choices	choice	NOUN
bracis-28433	317	105	/	/	SYM
bracis-28433	317	106	manage	manage	NOUN
bracis-28433	317	107	cookies	cookie	NOUN
bracis-28433	317	108	your	your	PRON
bracis-28433	317	109	us	us	PROPN
bracis-28433	318	1	state	state	NOUN
bracis-28433	318	2	privacy	privacy	NOUN
bracis-28433	318	3	rights	right	NOUN
bracis-28433	318	4	accessibility	accessibility	NOUN
bracis-28433	318	5	statement	statement	NOUN
bracis-28433	318	6	terms	term	NOUN
bracis-28433	318	7	and	and	CCONJ
bracis-28433	318	8	conditions	condition	NOUN
bracis-28433	318	9	privacy	privacy	NOUN
bracis-28433	318	10	policy	policy	NOUN
bracis-28433	318	11	help	help	NOUN
bracis-28433	318	12	and	and	CCONJ
bracis-28433	318	13	support	support	VERB
bracis-28433	318	14	legal	legal	ADJ
bracis-28433	318	15	notice	notice	NOUN
bracis-28433	318	16	cancel	cancel	VERB
bracis-28433	318	17	contracts	contract	NOUN
bracis-28433	318	18	here	here	ADV
bracis-28433	318	19	129.74.145.123	129.74.145.123	NUM
bracis-28433	318	20	hesburgh	hesburgh	PROPN
bracis-28433	318	21	library	library	PROPN
bracis-28433	318	22	er	er	INTJ
bracis-28433	318	23	unit	unit	NOUN
bracis-28433	318	24	(	(	PUNCT
bracis-28433	318	25	3005732405	3005732405	NUM
bracis-28433	318	26	)	)	PUNCT
bracis-28433	318	27	northeast	northeast	ADJ
bracis-28433	318	28	research	research	NOUN
bracis-28433	318	29	libraries	library	NOUN
bracis-28433	318	30	(	(	PUNCT
bracis-28433	318	31	nerl	nerl	PROPN
bracis-28433	318	32	)	)	PUNCT
bracis-28433	318	33	(	(	PUNCT
bracis-28433	318	34	8200828607	8200828607	NUM
bracis-28433	318	35	)	)	PUNCT
bracis-28433	318	36	nerl	nerl	VERB
bracis-28433	318	37	ta	ta	X
bracis-28433	318	38	account	account	NOUN
bracis-28433	318	39	(	(	PUNCT
bracis-28433	318	40	3006206169	3006206169	NUM
bracis-28433	318	41	)	)	PUNCT
bracis-28433	318	42	university	university	NOUN
bracis-28433	318	43	of	of	ADP
bracis-28433	318	44	notre	notre	PROPN
bracis-28433	318	45	dame	dame	PROPN
bracis-28433	318	46	hesburgh	hesburgh	PROPN
bracis-28433	318	47	library	library	NOUN
bracis-28433	318	48	(	(	PUNCT
bracis-28433	318	49	3000184373	3000184373	NUM
bracis-28433	318	50	)	)	PUNCT
bracis-28433	319	1	©	©	ADP
bracis-28433	319	2	2025	2025	NUM
bracis-28433	319	3	springer	springer	NOUN
bracis-28433	319	4	nature	nature	NOUN