id	sid	tid	token	lemma	pos
ajst-9864	1	1	academic	academic	ADJ
ajst-9864	1	2	journal	journal	NOUN
ajst-9864	1	3	of	of	ADP
ajst-9864	1	4	science	science	NOUN
ajst-9864	1	5	and	and	CCONJ
ajst-9864	1	6	technology	technology	NOUN
ajst-9864	1	7	issn	issn	NOUN
ajst-9864	1	8	:	:	PUNCT
ajst-9864	1	9	2771	2771	NUM
ajst-9864	1	10	-	-	SYM
ajst-9864	1	11	3032	3032	NUM
ajst-9864	1	12	|	|	NOUN
ajst-9864	1	13	vol	vol	NOUN
ajst-9864	1	14	.	.	PROPN
ajst-9864	2	1	6	6	NUM
ajst-9864	2	2	,	,	PUNCT
ajst-9864	2	3	no	no	INTJ
ajst-9864	2	4	.	.	NOUN
ajst-9864	2	5	2	2	NUM
ajst-9864	2	6	,	,	PUNCT
ajst-9864	2	7	2023	2023	NUM
ajst-9864	2	8	128	128	NUM
ajst-9864	2	9	an	an	DET
ajst-9864	2	10	entity	entity	NOUN
ajst-9864	2	11	based	base	VERB
ajst-9864	2	12	approach	approach	NOUN
ajst-9864	2	13	for	for	ADP
ajst-9864	2	14	extracting	extract	VERB
ajst-9864	2	15	relationship	relationship	NOUN
ajst-9864	2	16	between	between	ADP
ajst-9864	2	17	chinese	chinese	ADJ
ajst-9864	2	18	characters	character	NOUN
ajst-9864	2	19	yinggang	yinggang	PROPN
ajst-9864	2	20	jie1	jie1	PROPN
ajst-9864	2	21	,	,	PUNCT
ajst-9864	2	22	jiarong	jiarong	PROPN
ajst-9864	2	23	huang2	huang2	PROPN
ajst-9864	2	24	,	,	PUNCT
ajst-9864	2	25	yingyan	yingyan	NOUN
ajst-9864	2	26	wu3	wu3	VERB
ajst-9864	2	27	1	1	NUM
ajst-9864	2	28	faculty	faculty	NOUN
ajst-9864	2	29	of	of	ADP
ajst-9864	2	30	computer	computer	NOUN
ajst-9864	2	31	science	science	NOUN
ajst-9864	2	32	and	and	CCONJ
ajst-9864	2	33	information	information	NOUN
ajst-9864	2	34	technology	technology	NOUN
ajst-9864	2	35	,	,	PUNCT
ajst-9864	2	36	university	university	PROPN
ajst-9864	2	37	of	of	ADP
ajst-9864	2	38	malaya	malaya	PROPN
ajst-9864	2	39	,	,	PUNCT
ajst-9864	2	40	kuala	kuala	PROPN
ajst-9864	2	41	lumpur	lumpur	PROPN
ajst-9864	2	42	,	,	PUNCT
ajst-9864	2	43	50603	50603	NUM
ajst-9864	2	44	,	,	PUNCT
ajst-9864	2	45	malaysia	malaysia	NOUN
ajst-9864	2	46	2	2	NUM
ajst-9864	2	47	faculty	faculty	NOUN
ajst-9864	2	48	of	of	ADP
ajst-9864	2	49	business	business	NOUN
ajst-9864	2	50	and	and	CCONJ
ajst-9864	2	51	economics	economic	NOUN
ajst-9864	2	52	,	,	PUNCT
ajst-9864	2	53	university	university	PROPN
ajst-9864	2	54	of	of	ADP
ajst-9864	2	55	malaya	malaya	PROPN
ajst-9864	2	56	,	,	PUNCT
ajst-9864	2	57	kuala	kuala	PROPN
ajst-9864	2	58	lumpur	lumpur	PROPN
ajst-9864	2	59	,	,	PUNCT
ajst-9864	2	60	50603	50603	NUM
ajst-9864	2	61	,	,	PUNCT
ajst-9864	2	62	malaysia	malaysia	NOUN
ajst-9864	2	63	3	3	NUM
ajst-9864	2	64	faculty	faculty	NOUN
ajst-9864	2	65	of	of	ADP
ajst-9864	2	66	computer	computer	NOUN
ajst-9864	2	67	science	science	NOUN
ajst-9864	2	68	,	,	PUNCT
ajst-9864	2	69	inner	inner	PROPN
ajst-9864	2	70	mongolia	mongolia	PROPN
ajst-9864	2	71	university	university	PROPN
ajst-9864	2	72	,	,	PUNCT
ajst-9864	2	73	inner	inner	PROPN
ajst-9864	2	74	mongolia	mongolia	PROPN
ajst-9864	2	75	,	,	PUNCT
ajst-9864	2	76	010000	010000	NUM
ajst-9864	2	77	,	,	PUNCT
ajst-9864	2	78	china	china	PROPN
ajst-9864	2	79	abstract	abstract	NOUN
ajst-9864	2	80	:	:	PUNCT
ajst-9864	2	81	knowledge	knowledge	NOUN
ajst-9864	2	82	extraction	extraction	NOUN
ajst-9864	2	83	includes	include	VERB
ajst-9864	2	84	the	the	DET
ajst-9864	2	85	extraction	extraction	NOUN
ajst-9864	2	86	of	of	ADP
ajst-9864	2	87	entity	entity	NOUN
ajst-9864	2	88	relationships	relationship	NOUN
ajst-9864	2	89	,	,	PUNCT
ajst-9864	2	90	which	which	PRON
ajst-9864	2	91	is	be	AUX
ajst-9864	2	92	crucial	crucial	ADJ
ajst-9864	2	93	.	.	PUNCT
ajst-9864	3	1	deep	deep	ADJ
ajst-9864	3	2	learning	learn	VERB
ajst-9864	3	3	algorithms	algorithm	NOUN
ajst-9864	3	4	are	be	AUX
ajst-9864	3	5	increasingly	increasingly	ADV
ajst-9864	3	6	prevalent	prevalent	ADJ
ajst-9864	3	7	in	in	ADP
ajst-9864	3	8	relation	relation	NOUN
ajst-9864	3	9	extraction	extraction	NOUN
ajst-9864	3	10	tasks	task	NOUN
ajst-9864	3	11	when	when	SCONJ
ajst-9864	3	12	compared	compare	VERB
ajst-9864	3	13	to	to	ADP
ajst-9864	3	14	conventional	conventional	ADJ
ajst-9864	3	15	pattern	pattern	NOUN
ajst-9864	3	16	recognition	recognition	NOUN
ajst-9864	3	17	techniques	technique	NOUN
ajst-9864	3	18	.	.	PUNCT
ajst-9864	4	1	the	the	DET
ajst-9864	4	2	majority	majority	NOUN
ajst-9864	4	3	of	of	ADP
ajst-9864	4	4	current	current	ADJ
ajst-9864	4	5	research	research	NOUN
ajst-9864	4	6	on	on	ADP
ajst-9864	4	7	remotely	remotely	ADV
ajst-9864	4	8	supervised	supervise	VERB
ajst-9864	4	9	methods	method	NOUN
ajst-9864	4	10	and	and	CCONJ
ajst-9864	4	11	kernel	kernel	PROPN
ajst-9864	4	12	functions	function	NOUN
ajst-9864	4	13	in	in	ADP
ajst-9864	4	14	relationship	relationship	NOUN
ajst-9864	4	15	extraction	extraction	NOUN
ajst-9864	4	16	strategies	strategy	NOUN
ajst-9864	4	17	for	for	ADP
ajst-9864	4	18	chinese	chinese	ADJ
ajst-9864	4	19	language	language	NOUN
ajst-9864	4	20	can	can	AUX
ajst-9864	4	21	not	not	PART
ajst-9864	4	22	disregard	disregard	VERB
ajst-9864	4	23	the	the	DET
ajst-9864	4	24	detrimental	detrimental	ADJ
ajst-9864	4	25	effects	effect	NOUN
ajst-9864	4	26	of	of	ADP
ajst-9864	4	27	noisy	noisy	ADJ
ajst-9864	4	28	data	datum	NOUN
ajst-9864	4	29	in	in	ADP
ajst-9864	4	30	the	the	DET
ajst-9864	4	31	dataset	dataset	NOUN
ajst-9864	4	32	on	on	ADP
ajst-9864	4	33	the	the	DET
ajst-9864	4	34	experimental	experimental	ADJ
ajst-9864	4	35	findings	finding	NOUN
ajst-9864	4	36	.	.	PUNCT
ajst-9864	5	1	a	a	DET
ajst-9864	5	2	bidirectional	bidirectional	PROPN
ajst-9864	5	3	gru	gru	PROPN
ajst-9864	5	4	neural	neural	ADJ
ajst-9864	5	5	network	network	NOUN
ajst-9864	5	6	,	,	PUNCT
ajst-9864	5	7	a	a	DET
ajst-9864	5	8	two	two	NUM
ajst-9864	5	9	-	-	PUNCT
ajst-9864	5	10	layer	layer	NOUN
ajst-9864	5	11	attention	attention	NOUN
ajst-9864	5	12	mechanism	mechanism	NOUN
ajst-9864	5	13	,	,	PUNCT
ajst-9864	5	14	and	and	CCONJ
ajst-9864	5	15	a	a	DET
ajst-9864	5	16	chinese	chinese	ADJ
ajst-9864	5	17	relationship	relationship	NOUN
ajst-9864	5	18	extraction	extraction	NOUN
ajst-9864	5	19	model	model	NOUN
ajst-9864	5	20	are	be	AUX
ajst-9864	5	21	proposed	propose	VERB
ajst-9864	5	22	.	.	PUNCT
ajst-9864	6	1	for	for	ADP
ajst-9864	6	2	the	the	DET
ajst-9864	6	3	forgetfulness	forgetfulness	NOUN
ajst-9864	6	4	problem	problem	NOUN
ajst-9864	6	5	,	,	PUNCT
ajst-9864	6	6	a	a	DET
ajst-9864	6	7	bidirectional	bidirectional	PROPN
ajst-9864	6	8	gru	gru	PROPN
ajst-9864	6	9	neural	neural	ADJ
ajst-9864	6	10	network	network	NOUN
ajst-9864	6	11	is	be	AUX
ajst-9864	6	12	utilised	utilise	VERB
ajst-9864	6	13	to	to	PART
ajst-9864	6	14	fuse	fuse	VERB
ajst-9864	6	15	the	the	DET
ajst-9864	6	16	input	input	NOUN
ajst-9864	6	17	vectors	vector	NOUN
ajst-9864	6	18	.	.	PUNCT
ajst-9864	7	1	integrating	integrate	VERB
ajst-9864	7	2	the	the	DET
ajst-9864	7	3	structural	structural	ADJ
ajst-9864	7	4	properties	property	NOUN
ajst-9864	7	5	of	of	ADP
ajst-9864	7	6	the	the	DET
ajst-9864	7	7	mandarin	mandarin	NOUN
ajst-9864	7	8	language	language	NOUN
ajst-9864	7	9	,	,	PUNCT
ajst-9864	7	10	word	word	NOUN
ajst-9864	7	11	vectors	vector	NOUN
ajst-9864	7	12	are	be	AUX
ajst-9864	7	13	employed	employ	VERB
ajst-9864	7	14	as	as	ADP
ajst-9864	7	15	the	the	DET
ajst-9864	7	16	input	input	NOUN
ajst-9864	7	17	.	.	PUNCT
ajst-9864	8	1	sentence	sentence	NOUN
ajst-9864	8	2	-	-	PUNCT
ajst-9864	8	3	level	level	NOUN
ajst-9864	8	4	attention	attention	NOUN
ajst-9864	8	5	mechanisms	mechanism	NOUN
ajst-9864	8	6	are	be	AUX
ajst-9864	8	7	utilised	utilise	VERB
ajst-9864	8	8	to	to	PART
ajst-9864	8	9	extract	extract	VERB
ajst-9864	8	10	sentence	sentence	NOUN
ajst-9864	8	11	features	feature	NOUN
ajst-9864	8	12	,	,	PUNCT
ajst-9864	8	13	and	and	CCONJ
ajst-9864	8	14	word	word	NOUN
ajst-9864	8	15	-	-	PUNCT
ajst-9864	8	16	level	level	NOUN
ajst-9864	8	17	feature	feature	NOUN
ajst-9864	8	18	information	information	NOUN
ajst-9864	8	19	is	be	AUX
ajst-9864	8	20	retrieved	retrieve	VERB
ajst-9864	8	21	from	from	ADP
ajst-9864	8	22	a	a	DET
ajst-9864	8	23	sentence	sentence	NOUN
ajst-9864	8	24	.	.	PUNCT
ajst-9864	9	1	about	about	ADP
ajst-9864	9	2	1300	1300	NUM
ajst-9864	9	3	bits	bit	NOUN
ajst-9864	9	4	of	of	ADP
ajst-9864	9	5	data	datum	NOUN
ajst-9864	9	6	are	be	AUX
ajst-9864	9	7	extracted	extract	VERB
ajst-9864	9	8	from	from	ADP
ajst-9864	9	9	news	news	NOUN
ajst-9864	9	10	websites	website	NOUN
ajst-9864	9	11	using	use	VERB
ajst-9864	9	12	a	a	DET
ajst-9864	9	13	remotely	remotely	ADV
ajst-9864	9	14	supervised	supervise	VERB
ajst-9864	9	15	methodology	methodology	NOUN
ajst-9864	9	16	for	for	ADP
ajst-9864	9	17	validation	validation	NOUN
ajst-9864	9	18	.	.	PUNCT
ajst-9864	10	1	according	accord	VERB
ajst-9864	10	2	to	to	ADP
ajst-9864	10	3	the	the	DET
ajst-9864	10	4	experimental	experimental	ADJ
ajst-9864	10	5	findings	finding	NOUN
ajst-9864	10	6	,	,	PUNCT
ajst-9864	10	7	the	the	DET
ajst-9864	10	8	neural	neural	ADJ
ajst-9864	10	9	network	network	NOUN
ajst-9864	10	10	model	model	NOUN
ajst-9864	10	11	with	with	ADP
ajst-9864	10	12	a	a	DET
ajst-9864	10	13	two	two	NUM
ajst-9864	10	14	-	-	PUNCT
ajst-9864	10	15	layer	layer	NOUN
ajst-9864	10	16	attention	attention	NOUN
ajst-9864	10	17	mechanism	mechanism	NOUN
ajst-9864	10	18	is	be	AUX
ajst-9864	10	19	able	able	ADJ
ajst-9864	10	20	to	to	PART
ajst-9864	10	21	fully	fully	ADV
ajst-9864	10	22	utilise	utilise	VERB
ajst-9864	10	23	all	all	PRON
ajst-9864	10	24	of	of	ADP
ajst-9864	10	25	the	the	DET
ajst-9864	10	26	feature	feature	NOUN
ajst-9864	10	27	information	information	NOUN
ajst-9864	10	28	in	in	ADP
ajst-9864	10	29	a	a	DET
ajst-9864	10	30	sentence	sentence	NOUN
ajst-9864	10	31	,	,	PUNCT
ajst-9864	10	32	and	and	CCONJ
ajst-9864	10	33	it	it	PRON
ajst-9864	10	34	performs	perform	VERB
ajst-9864	10	35	significantly	significantly	ADV
ajst-9864	10	36	better	well	ADJ
ajst-9864	10	37	than	than	ADP
ajst-9864	10	38	the	the	DET
ajst-9864	10	39	neural	neural	ADJ
ajst-9864	10	40	network	network	NOUN
ajst-9864	10	41	model	model	NOUN
ajst-9864	10	42	without	without	ADP
ajst-9864	10	43	an	an	DET
ajst-9864	10	44	attention	attention	NOUN
ajst-9864	10	45	mechanism	mechanism	NOUN
ajst-9864	10	46	in	in	ADP
ajst-9864	10	47	terms	term	NOUN
ajst-9864	10	48	of	of	ADP
ajst-9864	10	49	accuracy	accuracy	NOUN
ajst-9864	10	50	and	and	CCONJ
ajst-9864	10	51	recall	recall	NOUN
ajst-9864	10	52	rates	rate	NOUN
ajst-9864	10	53	.	.	PUNCT
ajst-9864	11	1	compared	compare	VERB
ajst-9864	11	2	to	to	ADP
ajst-9864	11	3	the	the	DET
ajst-9864	11	4	model	model	NOUN
ajst-9864	11	5	without	without	ADP
ajst-9864	11	6	the	the	DET
ajst-9864	11	7	attention	attention	NOUN
ajst-9864	11	8	mechanism	mechanism	NOUN
ajst-9864	11	9	,	,	PUNCT
ajst-9864	11	10	the	the	DET
ajst-9864	11	11	accuracy	accuracy	NOUN
ajst-9864	11	12	and	and	CCONJ
ajst-9864	11	13	recall	recall	NOUN
ajst-9864	11	14	rates	rate	NOUN
ajst-9864	11	15	are	be	AUX
ajst-9864	11	16	noticeably	noticeably	ADV
ajst-9864	11	17	greater	great	ADJ
ajst-9864	11	18	.	.	PUNCT
ajst-9864	12	1	keywords	keyword	NOUN
ajst-9864	12	2	:	:	PUNCT
ajst-9864	12	3	chinese	chinese	ADJ
ajst-9864	12	4	relation	relation	NOUN
ajst-9864	12	5	extraction	extraction	NOUN
ajst-9864	12	6	,	,	PUNCT
ajst-9864	12	7	bidirectional	bidirectional	PROPN
ajst-9864	12	8	gru	gru	PROPN
ajst-9864	12	9	,	,	PUNCT
ajst-9864	12	10	neural	neural	ADJ
ajst-9864	12	11	network	network	NOUN
ajst-9864	12	12	,	,	PUNCT
ajst-9864	12	13	attention	attention	NOUN
ajst-9864	12	14	mechanism	mechanism	NOUN
ajst-9864	12	15	,	,	PUNCT
ajst-9864	12	16	word	word	NOUN
ajst-9864	12	17	vector	vector	NOUN
ajst-9864	12	18	.	.	PUNCT
ajst-9864	13	1	1	1	X
ajst-9864	13	2	.	.	X
ajst-9864	13	3	introduction	introduction	NOUN
ajst-9864	13	4	1.1	1.1	NUM
ajst-9864	13	5	.	.	PUNCT
ajst-9864	14	1	research	research	NOUN
ajst-9864	14	2	problem	problem	NOUN
ajst-9864	14	3	statement	statement	NOUN
ajst-9864	14	4	the	the	DET
ajst-9864	14	5	pervasive	pervasive	ADJ
ajst-9864	14	6	use	use	NOUN
ajst-9864	14	7	of	of	ADP
ajst-9864	14	8	the	the	DET
ajst-9864	14	9	internet	internet	NOUN
ajst-9864	14	10	has	have	AUX
ajst-9864	14	11	transitioned	transition	VERB
ajst-9864	14	12	a	a	DET
ajst-9864	14	13	multitude	multitude	NOUN
ajst-9864	14	14	of	of	ADP
ajst-9864	14	15	offline	offline	ADJ
ajst-9864	14	16	applications	application	NOUN
ajst-9864	14	17	into	into	ADP
ajst-9864	14	18	the	the	DET
ajst-9864	14	19	online	online	ADJ
ajst-9864	14	20	sphere	sphere	NOUN
ajst-9864	14	21	,	,	PUNCT
ajst-9864	14	22	resulting	result	VERB
ajst-9864	14	23	in	in	ADP
ajst-9864	14	24	an	an	DET
ajst-9864	14	25	influx	influx	NOUN
ajst-9864	14	26	of	of	ADP
ajst-9864	14	27	massive	massive	ADJ
ajst-9864	14	28	quantities	quantity	NOUN
ajst-9864	14	29	of	of	ADP
ajst-9864	14	30	social	social	ADJ
ajst-9864	14	31	data	datum	NOUN
ajst-9864	14	32	.	.	PUNCT
ajst-9864	15	1	a	a	DET
ajst-9864	15	2	considerable	considerable	ADJ
ajst-9864	15	3	portion	portion	NOUN
ajst-9864	15	4	of	of	ADP
ajst-9864	15	5	this	this	DET
ajst-9864	15	6	data	datum	NOUN
ajst-9864	15	7	,	,	PUNCT
ajst-9864	15	8	although	although	SCONJ
ajst-9864	15	9	potentially	potentially	ADV
ajst-9864	15	10	valuable	valuable	ADJ
ajst-9864	15	11	,	,	PUNCT
ajst-9864	15	12	is	be	AUX
ajst-9864	15	13	predominantly	predominantly	ADV
ajst-9864	15	14	unstructured	unstructured	ADJ
ajst-9864	15	15	natural	natural	ADJ
ajst-9864	15	16	language	language	NOUN
ajst-9864	15	17	data	datum	NOUN
ajst-9864	15	18	and	and	CCONJ
ajst-9864	15	19	,	,	PUNCT
ajst-9864	15	20	thus	thus	ADV
ajst-9864	15	21	,	,	PUNCT
ajst-9864	15	22	remains	remain	VERB
ajst-9864	15	23	underexploited	underexploite	VERB
ajst-9864	15	24	due	due	ADP
ajst-9864	15	25	to	to	ADP
ajst-9864	15	26	the	the	DET
ajst-9864	15	27	high	high	ADJ
ajst-9864	15	28	incidence	incidence	NOUN
ajst-9864	15	29	of	of	ADP
ajst-9864	15	30	noise	noise	NOUN
ajst-9864	15	31	.	.	PUNCT
ajst-9864	16	1	the	the	DET
ajst-9864	16	2	extraction	extraction	NOUN
ajst-9864	16	3	of	of	ADP
ajst-9864	16	4	meaningful	meaningful	ADJ
ajst-9864	16	5	insights	insight	NOUN
ajst-9864	16	6	from	from	ADP
ajst-9864	16	7	this	this	DET
ajst-9864	16	8	data	datum	NOUN
ajst-9864	16	9	can	can	AUX
ajst-9864	16	10	be	be	AUX
ajst-9864	16	11	accomplished	accomplish	VERB
ajst-9864	16	12	through	through	ADP
ajst-9864	16	13	text	text	NOUN
ajst-9864	16	14	mining	mining	NOUN
ajst-9864	16	15	,	,	PUNCT
ajst-9864	16	16	with	with	ADP
ajst-9864	16	17	a	a	DET
ajst-9864	16	18	particular	particular	ADJ
ajst-9864	16	19	emphasis	emphasis	NOUN
ajst-9864	16	20	on	on	ADP
ajst-9864	16	21	character	character	NOUN
ajst-9864	16	22	relationship	relationship	NOUN
ajst-9864	16	23	extraction	extraction	NOUN
ajst-9864	16	24	.	.	PUNCT
ajst-9864	17	1	this	this	DET
ajst-9864	17	2	paper	paper	NOUN
ajst-9864	17	3	confronts	confront	VERB
ajst-9864	17	4	the	the	DET
ajst-9864	17	5	challenge	challenge	NOUN
ajst-9864	17	6	of	of	ADP
ajst-9864	17	7	discerning	discern	VERB
ajst-9864	17	8	character	character	NOUN
ajst-9864	17	9	entities	entity	NOUN
ajst-9864	17	10	and	and	CCONJ
ajst-9864	17	11	their	their	PRON
ajst-9864	17	12	interconnectedness	interconnectedness	NOUN
ajst-9864	17	13	from	from	ADP
ajst-9864	17	14	extensive	extensive	ADJ
ajst-9864	17	15	amounts	amount	NOUN
ajst-9864	17	16	of	of	ADP
ajst-9864	17	17	unstructured	unstructured	ADJ
ajst-9864	17	18	chinese	chinese	ADJ
ajst-9864	17	19	text	text	NOUN
ajst-9864	17	20	data	datum	NOUN
ajst-9864	17	21	via	via	ADP
ajst-9864	17	22	a	a	DET
ajst-9864	17	23	text	text	NOUN
ajst-9864	17	24	mining	mining	NOUN
ajst-9864	17	25	-	-	PUNCT
ajst-9864	17	26	based	base	VERB
ajst-9864	17	27	methodology	methodology	NOUN
ajst-9864	17	28	.	.	PUNCT
ajst-9864	18	1	the	the	DET
ajst-9864	18	2	extraction	extraction	NOUN
ajst-9864	18	3	procedure	procedure	NOUN
ajst-9864	18	4	generates	generate	VERB
ajst-9864	18	5	triples	triple	NOUN
ajst-9864	18	6	,	,	PUNCT
ajst-9864	18	7	such	such	ADJ
ajst-9864	18	8	as	as	ADP
ajst-9864	18	9	"	"	PUNCT
ajst-9864	18	10	andy	andy	PROPN
ajst-9864	18	11	lau	lau	PROPN
ajst-9864	18	12	,	,	PUNCT
ajst-9864	18	13	his	his	PRON
ajst-9864	18	14	wife	wife	NOUN
ajst-9864	18	15	,	,	PUNCT
ajst-9864	18	16	and	and	CCONJ
ajst-9864	18	17	lilian	lilian	PROPN
ajst-9864	18	18	chu	chu	PROPN
ajst-9864	18	19	"	"	PUNCT
ajst-9864	18	20	,	,	PUNCT
ajst-9864	18	21	encapsulating	encapsulate	VERB
ajst-9864	18	22	relationships	relationship	NOUN
ajst-9864	18	23	that	that	PRON
ajst-9864	18	24	constitute	constitute	VERB
ajst-9864	18	25	the	the	DET
ajst-9864	18	26	foundation	foundation	NOUN
ajst-9864	18	27	of	of	ADP
ajst-9864	18	28	knowledge	knowledge	NOUN
ajst-9864	18	29	graphs	graph	NOUN
ajst-9864	18	30	with	with	ADP
ajst-9864	18	31	wide	wide	ADV
ajst-9864	18	32	-	-	PUNCT
ajst-9864	18	33	ranging	range	VERB
ajst-9864	18	34	applications[1	applications[1	NOUN
ajst-9864	18	35	]	]	PUNCT
ajst-9864	18	36	.	.	PUNCT
ajst-9864	19	1	deep	deep	ADJ
ajst-9864	19	2	neural	neural	ADJ
ajst-9864	19	3	networks	network	NOUN
ajst-9864	19	4	have	have	AUX
ajst-9864	19	5	been	be	AUX
ajst-9864	19	6	utilized	utilize	VERB
ajst-9864	19	7	in	in	ADP
ajst-9864	19	8	tasks	task	NOUN
ajst-9864	19	9	of	of	ADP
ajst-9864	19	10	relational	relational	ADJ
ajst-9864	19	11	extraction	extraction	NOUN
ajst-9864	19	12	to	to	ADP
ajst-9864	19	13	decipher	decipher	ADJ
ajst-9864	19	14	underlying	underlie	VERB
ajst-9864	19	15	information	information	NOUN
ajst-9864	19	16	and	and	CCONJ
ajst-9864	19	17	minimize	minimize	VERB
ajst-9864	19	18	errors	error	NOUN
ajst-9864	19	19	from	from	ADP
ajst-9864	19	20	manual	manual	ADJ
ajst-9864	19	21	labelling	labelling	NOUN
ajst-9864	19	22	.	.	PUNCT
ajst-9864	20	1	however	however	ADV
ajst-9864	20	2	,	,	PUNCT
ajst-9864	20	3	these	these	DET
ajst-9864	20	4	models	model	NOUN
ajst-9864	20	5	often	often	ADV
ajst-9864	20	6	overlook	overlook	VERB
ajst-9864	20	7	the	the	DET
ajst-9864	20	8	presence	presence	NOUN
ajst-9864	20	9	of	of	ADP
ajst-9864	20	10	multiple	multiple	ADJ
ajst-9864	20	11	sentences	sentence	NOUN
ajst-9864	20	12	illustrating	illustrate	VERB
ajst-9864	20	13	the	the	DET
ajst-9864	20	14	same	same	ADJ
ajst-9864	20	15	relationship	relationship	NOUN
ajst-9864	20	16	,	,	PUNCT
ajst-9864	20	17	thereby	thereby	ADV
ajst-9864	20	18	potentially	potentially	ADV
ajst-9864	20	19	forfeiting	forfeit	VERB
ajst-9864	20	20	valuable	valuable	ADJ
ajst-9864	20	21	information	information	NOUN
ajst-9864	20	22	.	.	PUNCT
ajst-9864	21	1	while	while	SCONJ
ajst-9864	21	2	neural	neural	ADJ
ajst-9864	21	3	network	network	NOUN
ajst-9864	21	4	models	model	NOUN
ajst-9864	21	5	employing	employ	VERB
ajst-9864	21	6	attention	attention	NOUN
ajst-9864	21	7	mechanisms	mechanism	NOUN
ajst-9864	21	8	have	have	AUX
ajst-9864	21	9	yielded	yield	VERB
ajst-9864	21	10	encouraging	encouraging	ADJ
ajst-9864	21	11	results	result	NOUN
ajst-9864	21	12	,	,	PUNCT
ajst-9864	21	13	current	current	ADJ
ajst-9864	21	14	models	model	NOUN
ajst-9864	21	15	integrate	integrate	VERB
ajst-9864	21	16	only	only	ADV
ajst-9864	21	17	a	a	DET
ajst-9864	21	18	single	single	ADJ
ajst-9864	21	19	layer	layer	NOUN
ajst-9864	21	20	of	of	ADP
ajst-9864	21	21	attention	attention	NOUN
ajst-9864	21	22	and	and	CCONJ
ajst-9864	21	23	do	do	AUX
ajst-9864	21	24	not	not	PART
ajst-9864	21	25	thoroughly	thoroughly	ADV
ajst-9864	21	26	account	account	VERB
ajst-9864	21	27	for	for	ADP
ajst-9864	21	28	the	the	DET
ajst-9864	21	29	causal	causal	ADJ
ajst-9864	21	30	relationships	relationship	NOUN
ajst-9864	21	31	within	within	ADP
ajst-9864	21	32	the	the	DET
ajst-9864	21	33	data[2	data[2	NOUN
ajst-9864	21	34	]	]	PUNCT
ajst-9864	21	35	.	.	PUNCT
ajst-9864	22	1	this	this	DET
ajst-9864	22	2	paper	paper	NOUN
ajst-9864	22	3	introduces	introduce	VERB
ajst-9864	22	4	a	a	DET
ajst-9864	22	5	bidirectional	bidirectional	PROPN
ajst-9864	22	6	gru	gru	PROPN
ajst-9864	22	7	neural	neural	PROPN
ajst-9864	22	8	network	network	NOUN
ajst-9864	22	9	model	model	NOUN
ajst-9864	22	10	endowed	endow	VERB
ajst-9864	22	11	with	with	ADP
ajst-9864	22	12	a	a	DET
ajst-9864	22	13	dual	dual	ADJ
ajst-9864	22	14	-	-	PUNCT
ajst-9864	22	15	layer	layer	NOUN
ajst-9864	22	16	attention	attention	NOUN
ajst-9864	22	17	mechanism	mechanism	NOUN
ajst-9864	22	18	operating	operate	VERB
ajst-9864	22	19	at	at	ADP
ajst-9864	22	20	both	both	CCONJ
ajst-9864	22	21	the	the	DET
ajst-9864	22	22	word	word	NOUN
ajst-9864	22	23	and	and	CCONJ
ajst-9864	22	24	sentence	sentence	NOUN
ajst-9864	22	25	levels	level	NOUN
ajst-9864	22	26	.	.	PUNCT
ajst-9864	23	1	this	this	DET
ajst-9864	23	2	model	model	NOUN
ajst-9864	23	3	effectively	effectively	ADV
ajst-9864	23	4	segregates	segregate	VERB
ajst-9864	23	5	noise	noise	NOUN
ajst-9864	23	6	data	datum	NOUN
ajst-9864	23	7	and	and	CCONJ
ajst-9864	23	8	fully	fully	ADV
ajst-9864	23	9	capitalizes	capitalize	VERB
ajst-9864	23	10	on	on	ADP
ajst-9864	23	11	valuable	valuable	ADJ
ajst-9864	23	12	sentences	sentence	NOUN
ajst-9864	23	13	for	for	ADP
ajst-9864	23	14	relationship	relationship	NOUN
ajst-9864	23	15	extraction	extraction	NOUN
ajst-9864	23	16	tasks	task	NOUN
ajst-9864	23	17	,	,	PUNCT
ajst-9864	23	18	even	even	ADV
ajst-9864	23	19	in	in	ADP
ajst-9864	23	20	the	the	DET
ajst-9864	23	21	absence	absence	NOUN
ajst-9864	23	22	of	of	ADP
ajst-9864	23	23	a	a	DET
ajst-9864	23	24	natural	natural	ADJ
ajst-9864	23	25	language	language	NOUN
ajst-9864	23	26	processing	processing	NOUN
ajst-9864	23	27	system	system	NOUN
ajst-9864	23	28	.	.	PUNCT
ajst-9864	24	1	the	the	DET
ajst-9864	24	2	salient	salient	ADJ
ajst-9864	24	3	contributions	contribution	NOUN
ajst-9864	24	4	of	of	ADP
ajst-9864	24	5	this	this	DET
ajst-9864	24	6	paper	paper	NOUN
ajst-9864	24	7	include	include	VERB
ajst-9864	24	8	a	a	DET
ajst-9864	24	9	two	two	NUM
ajst-9864	24	10	-	-	PUNCT
ajst-9864	24	11	layer	layer	NOUN
ajst-9864	24	12	attention	attention	NOUN
ajst-9864	24	13	mechanism	mechanism	NOUN
ajst-9864	24	14	capable	capable	ADJ
ajst-9864	24	15	of	of	ADP
ajst-9864	24	16	encapsulating	encapsulate	VERB
ajst-9864	24	17	the	the	DET
ajst-9864	24	18	feature	feature	NOUN
ajst-9864	24	19	information	information	NOUN
ajst-9864	24	20	of	of	ADP
ajst-9864	24	21	chinese	chinese	ADJ
ajst-9864	24	22	sentences	sentence	NOUN
ajst-9864	24	23	and	and	CCONJ
ajst-9864	24	24	a	a	DET
ajst-9864	24	25	more	more	ADV
ajst-9864	24	26	straightforward	straightforward	ADJ
ajst-9864	24	27	,	,	PUNCT
ajst-9864	24	28	less	less	ADV
ajst-9864	24	29	susceptible	susceptible	ADJ
ajst-9864	24	30	to	to	ADP
ajst-9864	24	31	overfitting	overfitte	VERB
ajst-9864	24	32	,	,	PUNCT
ajst-9864	24	33	bidirectional	bidirectional	PROPN
ajst-9864	24	34	gru	gru	PROPN
ajst-9864	24	35	neural	neural	PROPN
ajst-9864	24	36	network	network	NOUN
ajst-9864	24	37	model[3	model[3	PROPN
ajst-9864	24	38	]	]	NOUN
ajst-9864	24	39	.	.	PUNCT
ajst-9864	25	1	1.2	1.2	NUM
ajst-9864	25	2	.	.	PUNCT
ajst-9864	25	3	research	research	NOUN
ajst-9864	25	4	objectives	objective	NOUN
ajst-9864	25	5	and	and	CCONJ
ajst-9864	25	6	questions	question	NOUN
ajst-9864	25	7	this	this	DET
ajst-9864	25	8	study	study	NOUN
ajst-9864	25	9	seeks	seek	VERB
ajst-9864	25	10	to	to	PART
ajst-9864	25	11	perform	perform	VERB
ajst-9864	25	12	a	a	DET
ajst-9864	25	13	comparative	comparative	ADJ
ajst-9864	25	14	analysis	analysis	NOUN
ajst-9864	25	15	to	to	PART
ajst-9864	25	16	identify	identify	VERB
ajst-9864	25	17	a	a	DET
ajst-9864	25	18	suitable	suitable	ADJ
ajst-9864	25	19	natural	natural	ADJ
ajst-9864	25	20	language	language	NOUN
ajst-9864	25	21	processing	processing	NOUN
ajst-9864	25	22	(	(	PUNCT
ajst-9864	25	23	nlp	nlp	NOUN
ajst-9864	25	24	)	)	PUNCT
ajst-9864	25	25	model	model	NOUN
ajst-9864	25	26	for	for	ADP
ajst-9864	25	27	chinese	chinese	ADJ
ajst-9864	25	28	text	text	NOUN
ajst-9864	25	29	relationship	relationship	NOUN
ajst-9864	25	30	analysis	analysis	NOUN
ajst-9864	25	31	and	and	CCONJ
ajst-9864	25	32	to	to	PART
ajst-9864	25	33	construct	construct	VERB
ajst-9864	25	34	a	a	DET
ajst-9864	25	35	bidirectional	bidirectional	PROPN
ajst-9864	25	36	gru	gru	PROPN
ajst-9864	25	37	neural	neural	ADJ
ajst-9864	25	38	network	network	NOUN
ajst-9864	25	39	model	model	NOUN
ajst-9864	25	40	for	for	ADP
ajst-9864	25	41	inter	inter	ADJ
ajst-9864	25	42	-	-	ADJ
ajst-9864	25	43	character	character	ADJ
ajst-9864	25	44	relationship	relationship	NOUN
ajst-9864	25	45	detection	detection	NOUN
ajst-9864	25	46	.	.	PUNCT
ajst-9864	26	1	it	it	PRON
ajst-9864	26	2	further	far	ADV
ajst-9864	26	3	aims	aim	VERB
ajst-9864	26	4	to	to	PART
ajst-9864	26	5	incorporate	incorporate	VERB
ajst-9864	26	6	a	a	DET
ajst-9864	26	7	doublelayer	doublelayer	NOUN
ajst-9864	26	8	attention	attention	NOUN
ajst-9864	26	9	mechanism	mechanism	NOUN
ajst-9864	26	10	in	in	ADP
ajst-9864	26	11	the	the	DET
ajst-9864	26	12	bi	bi	PROPN
ajst-9864	26	13	-	-	PROPN
ajst-9864	26	14	gru	gru	NOUN
ajst-9864	26	15	model	model	NOUN
ajst-9864	26	16	to	to	PART
ajst-9864	26	17	better	well	ADV
ajst-9864	26	18	capture	capture	VERB
ajst-9864	26	19	chinese	chinese	ADJ
ajst-9864	26	20	sentence	sentence	NOUN
ajst-9864	26	21	features	feature	NOUN
ajst-9864	26	22	and	and	CCONJ
ajst-9864	26	23	mitigate	mitigate	VERB
ajst-9864	26	24	the	the	DET
ajst-9864	26	25	impact	impact	NOUN
ajst-9864	26	26	of	of	ADP
ajst-9864	26	27	noisy	noisy	ADJ
ajst-9864	26	28	data	datum	NOUN
ajst-9864	26	29	.	.	PUNCT
ajst-9864	27	1	the	the	DET
ajst-9864	27	2	primary	primary	ADJ
ajst-9864	27	3	research	research	NOUN
ajst-9864	27	4	questions	question	NOUN
ajst-9864	27	5	include	include	VERB
ajst-9864	27	6	understanding	understand	VERB
ajst-9864	27	7	the	the	DET
ajst-9864	27	8	significance	significance	NOUN
ajst-9864	27	9	of	of	ADP
ajst-9864	27	10	chinese	chinese	ADJ
ajst-9864	27	11	text	text	NOUN
ajst-9864	27	12	relationship	relationship	NOUN
ajst-9864	27	13	mining	mining	NOUN
ajst-9864	27	14	,	,	PUNCT
ajst-9864	27	15	methods	method	NOUN
ajst-9864	27	16	to	to	PART
ajst-9864	27	17	translate	translate	VERB
ajst-9864	27	18	chinese	chinese	ADJ
ajst-9864	27	19	words	word	NOUN
ajst-9864	27	20	into	into	ADP
ajst-9864	27	21	word	word	NOUN
ajst-9864	27	22	vectors	vector	NOUN
ajst-9864	27	23	for	for	ADP
ajst-9864	27	24	nlp	nlp	ADJ
ajst-9864	27	25	classification	classification	NOUN
ajst-9864	27	26	,	,	PUNCT
ajst-9864	27	27	techniques	technique	NOUN
ajst-9864	27	28	to	to	PART
ajst-9864	27	29	improve	improve	VERB
ajst-9864	27	30	machine	machine	NOUN
ajst-9864	27	31	learning	learning	NOUN
ajst-9864	27	32	models	model	NOUN
ajst-9864	27	33	for	for	ADP
ajst-9864	27	34	entity	entity	NOUN
ajst-9864	27	35	relationship	relationship	NOUN
ajst-9864	27	36	extraction	extraction	NOUN
ajst-9864	27	37	in	in	ADP
ajst-9864	27	38	complex	complex	ADJ
ajst-9864	27	39	chinese	chinese	ADJ
ajst-9864	27	40	sentences	sentence	NOUN
ajst-9864	27	41	,	,	PUNCT
ajst-9864	27	42	and	and	CCONJ
ajst-9864	27	43	strategies	strategy	NOUN
ajst-9864	27	44	to	to	PART
ajst-9864	27	45	enhance	enhance	VERB
ajst-9864	27	46	the	the	DET
ajst-9864	27	47	accuracy	accuracy	NOUN
ajst-9864	27	48	of	of	ADP
ajst-9864	27	49	chinese	chinese	ADJ
ajst-9864	27	50	text	text	NOUN
ajst-9864	27	51	mining	mining	NOUN
ajst-9864	27	52	models	model	NOUN
ajst-9864	27	53	.	.	PUNCT
ajst-9864	28	1	1.3	1.3	NUM
ajst-9864	28	2	.	.	PUNCT
ajst-9864	29	1	research	research	NOUN
ajst-9864	29	2	motivation	motivation	NOUN
ajst-9864	29	3	the	the	DET
ajst-9864	29	4	growth	growth	NOUN
ajst-9864	29	5	of	of	ADP
ajst-9864	29	6	internet	internet	NOUN
ajst-9864	29	7	use	use	NOUN
ajst-9864	29	8	and	and	CCONJ
ajst-9864	29	9	big	big	ADJ
ajst-9864	29	10	data	datum	NOUN
ajst-9864	29	11	has	have	AUX
ajst-9864	29	12	led	lead	VERB
ajst-9864	29	13	to	to	ADP
ajst-9864	29	14	a	a	DET
ajst-9864	29	15	surge	surge	NOUN
ajst-9864	29	16	in	in	ADP
ajst-9864	29	17	text	text	NOUN
ajst-9864	29	18	data	datum	NOUN
ajst-9864	29	19	,	,	PUNCT
ajst-9864	29	20	rich	rich	ADJ
ajst-9864	29	21	with	with	ADP
ajst-9864	29	22	potential	potential	ADJ
ajst-9864	29	23	information	information	NOUN
ajst-9864	29	24	.	.	PUNCT
ajst-9864	30	1	however	however	ADV
ajst-9864	30	2	,	,	PUNCT
ajst-9864	30	3	the	the	DET
ajst-9864	30	4	natural	natural	ADJ
ajst-9864	30	5	language	language	NOUN
ajst-9864	30	6	format	format	NOUN
ajst-9864	30	7	poses	pose	VERB
ajst-9864	30	8	a	a	DET
ajst-9864	30	9	challenge	challenge	NOUN
ajst-9864	30	10	for	for	ADP
ajst-9864	30	11	direct	direct	ADJ
ajst-9864	30	12	computer	computer	NOUN
ajst-9864	30	13	utilization	utilization	NOUN
ajst-9864	30	14	,	,	PUNCT
ajst-9864	30	15	necessitating	necessitate	VERB
ajst-9864	30	16	natural	natural	ADJ
ajst-9864	30	17	language	language	NOUN
ajst-9864	30	18	processing	processing	NOUN
ajst-9864	30	19	technology	technology	NOUN
ajst-9864	30	20	to	to	PART
ajst-9864	30	21	convert	convert	VERB
ajst-9864	30	22	data	datum	NOUN
ajst-9864	30	23	into	into	ADP
ajst-9864	30	24	a	a	DET
ajst-9864	30	25	computer	computer	NOUN
ajst-9864	30	26	-	-	PUNCT
ajst-9864	30	27	readable	readable	ADJ
ajst-9864	30	28	format	format	NOUN
ajst-9864	30	29	.	.	PUNCT
ajst-9864	31	1	entity	entity	NOUN
ajst-9864	31	2	relationship	relationship	NOUN
ajst-9864	31	3	extraction	extraction	NOUN
ajst-9864	31	4	,	,	PUNCT
ajst-9864	31	5	a	a	DET
ajst-9864	31	6	vital	vital	ADJ
ajst-9864	31	7	component	component	NOUN
ajst-9864	31	8	of	of	ADP
ajst-9864	31	9	nlp	nlp	NOUN
ajst-9864	31	10	,	,	PUNCT
ajst-9864	31	11	faces	face	VERB
ajst-9864	31	12	unique	unique	ADJ
ajst-9864	31	13	challenges	challenge	NOUN
ajst-9864	31	14	in	in	ADP
ajst-9864	31	15	chinese	chinese	ADJ
ajst-9864	31	16	due	due	ADP
ajst-9864	31	17	to	to	ADP
ajst-9864	31	18	its	its	PRON
ajst-9864	31	19	complex	complex	ADJ
ajst-9864	31	20	language	language	NOUN
ajst-9864	31	21	features	feature	NOUN
ajst-9864	31	22	and	and	CCONJ
ajst-9864	31	23	shorter	short	ADJ
ajst-9864	31	24	texts	text	NOUN
ajst-9864	31	25	that	that	PRON
ajst-9864	31	26	provide	provide	VERB
ajst-9864	31	27	insufficient	insufficient	ADJ
ajst-9864	31	28	contextual	contextual	ADJ
ajst-9864	31	29	information	information	NOUN
ajst-9864	31	30	.	.	PUNCT
ajst-9864	32	1	the	the	DET
ajst-9864	32	2	dearth	dearth	NOUN
ajst-9864	32	3	of	of	ADP
ajst-9864	32	4	research	research	NOUN
ajst-9864	32	5	on	on	ADP
ajst-9864	32	6	chinese	chinese	ADJ
ajst-9864	32	7	entity	entity	NOUN
ajst-9864	32	8	relation	relation	NOUN
ajst-9864	32	9	extraction	extraction	NOUN
ajst-9864	32	10	further	far	ADV
ajst-9864	32	11	compounds	compound	VERB
ajst-9864	32	12	these	these	DET
ajst-9864	32	13	issues	issue	NOUN
ajst-9864	32	14	.	.	PUNCT
ajst-9864	33	1	this	this	DET
ajst-9864	33	2	study	study	NOUN
ajst-9864	33	3	aims	aim	VERB
ajst-9864	33	4	to	to	PART
ajst-9864	33	5	address	address	VERB
ajst-9864	33	6	these	these	DET
ajst-9864	33	7	challenges	challenge	NOUN
ajst-9864	33	8	through	through	ADP
ajst-9864	33	9	deep	deep	ADJ
ajst-9864	33	10	learningbased	learningbase	VERB
ajst-9864	33	11	chinese	chinese	ADJ
ajst-9864	33	12	text	text	NOUN
ajst-9864	33	13	entity	entity	NOUN
ajst-9864	33	14	relation	relation	NOUN
ajst-9864	33	15	extraction	extraction	NOUN
ajst-9864	33	16	models	model	NOUN
ajst-9864	33	17	,	,	PUNCT
ajst-9864	33	18	129	129	NUM
ajst-9864	33	19	enhancing	enhance	VERB
ajst-9864	33	20	accuracy	accuracy	NOUN
ajst-9864	33	21	and	and	CCONJ
ajst-9864	33	22	efficiency	efficiency	NOUN
ajst-9864	33	23	while	while	SCONJ
ajst-9864	33	24	supporting	support	VERB
ajst-9864	33	25	the	the	DET
ajst-9864	33	26	advancement	advancement	NOUN
ajst-9864	33	27	of	of	ADP
ajst-9864	33	28	chinese	chinese	ADJ
ajst-9864	33	29	nlp	nlp	NOUN
ajst-9864	33	30	technology	technology	NOUN
ajst-9864	33	31	and	and	CCONJ
ajst-9864	33	32	the	the	DET
ajst-9864	33	33	application	application	NOUN
ajst-9864	33	34	of	of	ADP
ajst-9864	33	35	deep	deep	ADJ
ajst-9864	33	36	learning	learning	NOUN
ajst-9864	33	37	in	in	ADP
ajst-9864	33	38	nlp	nlp	NOUN
ajst-9864	33	39	.	.	PROPN
ajst-9864	34	1	1.4	1.4	NUM
ajst-9864	34	2	.	.	PUNCT
ajst-9864	35	1	research	research	NOUN
ajst-9864	35	2	significance	significance	NOUN
ajst-9864	35	3	there	there	PRON
ajst-9864	35	4	is	be	VERB
ajst-9864	35	5	a	a	DET
ajst-9864	35	6	wealth	wealth	NOUN
ajst-9864	35	7	of	of	ADP
ajst-9864	35	8	chinese	chinese	ADJ
ajst-9864	35	9	social	social	ADJ
ajst-9864	35	10	data	datum	NOUN
ajst-9864	35	11	available	available	ADJ
ajst-9864	35	12	on	on	ADP
ajst-9864	35	13	the	the	DET
ajst-9864	35	14	internet	internet	NOUN
ajst-9864	35	15	,	,	PUNCT
ajst-9864	35	16	yet	yet	CCONJ
ajst-9864	35	17	many	many	ADJ
ajst-9864	35	18	advanced	advanced	ADJ
ajst-9864	35	19	entity	entity	NOUN
ajst-9864	35	20	relationship	relationship	NOUN
ajst-9864	35	21	extraction	extraction	NOUN
ajst-9864	35	22	methods	method	NOUN
ajst-9864	35	23	cater	cater	VERB
ajst-9864	35	24	primarily	primarily	ADV
ajst-9864	35	25	to	to	ADP
ajst-9864	35	26	english	english	ADJ
ajst-9864	35	27	texts	text	NOUN
ajst-9864	35	28	.	.	PUNCT
ajst-9864	36	1	with	with	ADP
ajst-9864	36	2	chinese	chinese	ADJ
ajst-9864	36	3	free	free	ADJ
ajst-9864	36	4	text	text	NOUN
ajst-9864	36	5	exhibiting	exhibit	VERB
ajst-9864	36	6	complex	complex	ADJ
ajst-9864	36	7	sentence	sentence	NOUN
ajst-9864	36	8	structures	structure	NOUN
ajst-9864	36	9	and	and	CCONJ
ajst-9864	36	10	contextspecific	contextspecific	ADJ
ajst-9864	36	11	word	word	NOUN
ajst-9864	36	12	meanings	meaning	NOUN
ajst-9864	36	13	,	,	PUNCT
ajst-9864	36	14	traditional	traditional	ADJ
ajst-9864	36	15	character	character	NOUN
ajst-9864	36	16	relationship	relationship	NOUN
ajst-9864	36	17	extraction	extraction	NOUN
ajst-9864	36	18	methods	method	NOUN
ajst-9864	36	19	often	often	ADV
ajst-9864	36	20	fall	fall	VERB
ajst-9864	36	21	short	short	ADJ
ajst-9864	36	22	.	.	PUNCT
ajst-9864	37	1	in	in	ADP
ajst-9864	37	2	the	the	DET
ajst-9864	37	3	era	era	NOUN
ajst-9864	37	4	of	of	ADP
ajst-9864	37	5	artificial	artificial	ADJ
ajst-9864	37	6	intelligence	intelligence	NOUN
ajst-9864	37	7	,	,	PUNCT
ajst-9864	37	8	companies	company	NOUN
ajst-9864	37	9	are	be	AUX
ajst-9864	37	10	investing	invest	VERB
ajst-9864	37	11	heavily	heavily	ADV
ajst-9864	37	12	in	in	ADP
ajst-9864	37	13	advanced	advanced	ADJ
ajst-9864	37	14	algorithms	algorithm	NOUN
ajst-9864	37	15	for	for	ADP
ajst-9864	37	16	chinese	chinese	ADJ
ajst-9864	37	17	character	character	NOUN
ajst-9864	37	18	relationship	relationship	NOUN
ajst-9864	37	19	mining	mining	NOUN
ajst-9864	37	20	,	,	PUNCT
ajst-9864	37	21	enhancing	enhance	VERB
ajst-9864	37	22	both	both	DET
ajst-9864	37	23	efficiency	efficiency	NOUN
ajst-9864	37	24	and	and	CCONJ
ajst-9864	37	25	accuracy	accuracy	NOUN
ajst-9864	37	26	.	.	PUNCT
ajst-9864	38	1	this	this	DET
ajst-9864	38	2	paper	paper	NOUN
ajst-9864	38	3	proposes	propose	VERB
ajst-9864	38	4	a	a	DET
ajst-9864	38	5	new	new	ADJ
ajst-9864	38	6	method	method	NOUN
ajst-9864	38	7	for	for	ADP
ajst-9864	38	8	this	this	DET
ajst-9864	38	9	task	task	NOUN
ajst-9864	38	10	,	,	PUNCT
ajst-9864	38	11	thus	thus	ADV
ajst-9864	38	12	bearing	bear	VERB
ajst-9864	38	13	significant	significant	ADJ
ajst-9864	38	14	relevance[4	relevance[4	NOUN
ajst-9864	38	15	]	]	PUNCT
ajst-9864	38	16	.	.	PUNCT
ajst-9864	39	1	knowledge	knowledge	NOUN
ajst-9864	39	2	bases	basis	NOUN
ajst-9864	39	3	containing	contain	VERB
ajst-9864	39	4	personal	personal	ADJ
ajst-9864	39	5	relationships	relationship	NOUN
ajst-9864	39	6	are	be	AUX
ajst-9864	39	7	integral	integral	ADJ
ajst-9864	39	8	to	to	ADP
ajst-9864	39	9	numerous	numerous	ADJ
ajst-9864	39	10	daily	daily	ADJ
ajst-9864	39	11	life	life	NOUN
ajst-9864	39	12	scenarios	scenario	NOUN
ajst-9864	39	13	and	and	CCONJ
ajst-9864	39	14	enterprise	enterprise	NOUN
ajst-9864	39	15	business	business	NOUN
ajst-9864	39	16	segments	segment	NOUN
ajst-9864	39	17	,	,	PUNCT
ajst-9864	39	18	especially	especially	ADV
ajst-9864	39	19	in	in	ADP
ajst-9864	39	20	search	search	NOUN
ajst-9864	39	21	,	,	PUNCT
ajst-9864	39	22	recommendation	recommendation	NOUN
ajst-9864	39	23	,	,	PUNCT
ajst-9864	39	24	and	and	CCONJ
ajst-9864	39	25	advertising	advertising	NOUN
ajst-9864	39	26	.	.	PUNCT
ajst-9864	40	1	as	as	ADP
ajst-9864	40	2	such	such	ADJ
ajst-9864	40	3	,	,	PUNCT
ajst-9864	40	4	refining	refine	VERB
ajst-9864	40	5	the	the	DET
ajst-9864	40	6	extraction	extraction	NOUN
ajst-9864	40	7	method	method	NOUN
ajst-9864	40	8	for	for	ADP
ajst-9864	40	9	chinese	chinese	ADJ
ajst-9864	40	10	personal	personal	ADJ
ajst-9864	40	11	relationships	relationship	NOUN
ajst-9864	40	12	and	and	CCONJ
ajst-9864	40	13	improving	improve	VERB
ajst-9864	40	14	accuracy	accuracy	NOUN
ajst-9864	40	15	is	be	AUX
ajst-9864	40	16	crucial	crucial	ADJ
ajst-9864	40	17	.	.	PUNCT
ajst-9864	41	1	deep	deep	ADJ
ajst-9864	41	2	learning	learning	NOUN
ajst-9864	41	3	-	-	PUNCT
ajst-9864	41	4	based	base	VERB
ajst-9864	41	5	models	model	NOUN
ajst-9864	41	6	hold	hold	VERB
ajst-9864	41	7	the	the	DET
ajst-9864	41	8	key	key	NOUN
ajst-9864	41	9	to	to	ADP
ajst-9864	41	10	enhancing	enhance	VERB
ajst-9864	41	11	entity	entity	NOUN
ajst-9864	41	12	relationship	relationship	NOUN
ajst-9864	41	13	extraction	extraction	NOUN
ajst-9864	41	14	from	from	ADP
ajst-9864	41	15	chinese	chinese	ADJ
ajst-9864	41	16	texts	text	NOUN
ajst-9864	41	17	,	,	PUNCT
ajst-9864	41	18	encompassing	encompass	VERB
ajst-9864	41	19	entity	entity	NOUN
ajst-9864	41	20	recognition	recognition	NOUN
ajst-9864	41	21	,	,	PUNCT
ajst-9864	41	22	relationship	relationship	NOUN
ajst-9864	41	23	extraction	extraction	NOUN
ajst-9864	41	24	,	,	PUNCT
ajst-9864	41	25	and	and	CCONJ
ajst-9864	41	26	text	text	NOUN
ajst-9864	41	27	comprehension	comprehension	NOUN
ajst-9864	41	28	.	.	PUNCT
ajst-9864	42	1	such	such	ADJ
ajst-9864	42	2	research	research	NOUN
ajst-9864	42	3	bears	bear	VERB
ajst-9864	42	4	multiple	multiple	ADJ
ajst-9864	42	5	levels	level	NOUN
ajst-9864	42	6	of	of	ADP
ajst-9864	42	7	significance	significance	NOUN
ajst-9864	42	8	:	:	PUNCT
ajst-9864	42	9	it	it	PRON
ajst-9864	42	10	can	can	AUX
ajst-9864	42	11	enhance	enhance	VERB
ajst-9864	42	12	the	the	DET
ajst-9864	42	13	efficiency	efficiency	NOUN
ajst-9864	42	14	and	and	CCONJ
ajst-9864	42	15	accuracy	accuracy	NOUN
ajst-9864	42	16	of	of	ADP
ajst-9864	42	17	natural	natural	ADJ
ajst-9864	42	18	language	language	NOUN
ajst-9864	42	19	processing	processing	NOUN
ajst-9864	42	20	,	,	PUNCT
ajst-9864	42	21	foster	foster	VERB
ajst-9864	42	22	the	the	DET
ajst-9864	42	23	development	development	NOUN
ajst-9864	42	24	of	of	ADP
ajst-9864	42	25	information	information	NOUN
ajst-9864	42	26	extraction	extraction	NOUN
ajst-9864	42	27	and	and	CCONJ
ajst-9864	42	28	knowledge	knowledge	NOUN
ajst-9864	42	29	graphs	graph	NOUN
ajst-9864	42	30	,	,	PUNCT
ajst-9864	42	31	further	far	ADV
ajst-9864	42	32	the	the	DET
ajst-9864	42	33	application	application	NOUN
ajst-9864	42	34	of	of	ADP
ajst-9864	42	35	deep	deep	ADJ
ajst-9864	42	36	learning	learning	NOUN
ajst-9864	42	37	in	in	ADP
ajst-9864	42	38	natural	natural	ADJ
ajst-9864	42	39	language	language	NOUN
ajst-9864	42	40	processing	processing	NOUN
ajst-9864	42	41	,	,	PUNCT
ajst-9864	42	42	and	and	CCONJ
ajst-9864	42	43	promote	promote	VERB
ajst-9864	42	44	the	the	DET
ajst-9864	42	45	advancement	advancement	NOUN
ajst-9864	42	46	of	of	ADP
ajst-9864	42	47	chinese	chinese	ADJ
ajst-9864	42	48	natural	natural	ADJ
ajst-9864	42	49	language	language	NOUN
ajst-9864	42	50	processing	processing	NOUN
ajst-9864	42	51	.	.	PUNCT
ajst-9864	43	1	in	in	ADP
ajst-9864	43	2	essence	essence	NOUN
ajst-9864	43	3	,	,	PUNCT
ajst-9864	43	4	this	this	DET
ajst-9864	43	5	research	research	NOUN
ajst-9864	43	6	can	can	AUX
ajst-9864	43	7	contribute	contribute	VERB
ajst-9864	43	8	to	to	ADP
ajst-9864	43	9	a	a	DET
ajst-9864	43	10	broader	broad	ADJ
ajst-9864	43	11	adoption	adoption	NOUN
ajst-9864	43	12	and	and	CCONJ
ajst-9864	43	13	development	development	NOUN
ajst-9864	43	14	of	of	ADP
ajst-9864	43	15	artificial	artificial	ADJ
ajst-9864	43	16	intelligence	intelligence	NOUN
ajst-9864	43	17	.	.	PUNCT
ajst-9864	44	1	2	2	X
ajst-9864	44	2	.	.	X
ajst-9864	44	3	literature	literature	NOUN
ajst-9864	44	4	references	reference	NOUN
ajst-9864	44	5	entity	entity	NOUN
ajst-9864	44	6	relation	relation	NOUN
ajst-9864	44	7	extraction	extraction	NOUN
ajst-9864	44	8	is	be	AUX
ajst-9864	44	9	a	a	DET
ajst-9864	44	10	key	key	ADJ
ajst-9864	44	11	task	task	NOUN
ajst-9864	44	12	in	in	ADP
ajst-9864	44	13	natural	natural	ADJ
ajst-9864	44	14	language	language	NOUN
ajst-9864	44	15	processing	processing	NOUN
ajst-9864	44	16	,	,	PUNCT
ajst-9864	44	17	identifying	identify	VERB
ajst-9864	44	18	relationships	relationship	NOUN
ajst-9864	44	19	between	between	ADP
ajst-9864	44	20	entities	entity	NOUN
ajst-9864	44	21	within	within	ADP
ajst-9864	44	22	text	text	NOUN
ajst-9864	44	23	,	,	PUNCT
ajst-9864	44	24	which	which	PRON
ajst-9864	44	25	supports	support	VERB
ajst-9864	44	26	knowledge	knowledge	NOUN
ajst-9864	44	27	graph	graph	NOUN
ajst-9864	44	28	creation	creation	NOUN
ajst-9864	44	29	and	and	CCONJ
ajst-9864	44	30	information	information	NOUN
ajst-9864	44	31	retrieval	retrieval	NOUN
ajst-9864	44	32	applications	application	NOUN
ajst-9864	44	33	.	.	PUNCT
ajst-9864	45	1	techniques	technique	NOUN
ajst-9864	45	2	used	use	VERB
ajst-9864	45	3	for	for	ADP
ajst-9864	45	4	extraction	extraction	NOUN
ajst-9864	45	5	can	can	AUX
ajst-9864	45	6	be	be	AUX
ajst-9864	45	7	categorized	categorize	VERB
ajst-9864	45	8	as	as	ADP
ajst-9864	45	9	rule	rule	NOUN
ajst-9864	45	10	-	-	PUNCT
ajst-9864	45	11	based	base	VERB
ajst-9864	45	12	,	,	PUNCT
ajst-9864	45	13	statistical	statistical	ADJ
ajst-9864	45	14	,	,	PUNCT
ajst-9864	45	15	or	or	CCONJ
ajst-9864	45	16	deep	deep	ADJ
ajst-9864	45	17	learning	learning	NOUN
ajst-9864	45	18	-	-	PUNCT
ajst-9864	45	19	based[5	based[5	PROPN
ajst-9864	45	20	]	]	PUNCT
ajst-9864	45	21	.	.	PUNCT
ajst-9864	46	1	rule	rule	NOUN
ajst-9864	46	2	-	-	PUNCT
ajst-9864	46	3	based	base	VERB
ajst-9864	46	4	techniques	technique	NOUN
ajst-9864	46	5	utilize	utilize	VERB
ajst-9864	46	6	manually	manually	ADV
ajst-9864	46	7	crafted	craft	VERB
ajst-9864	46	8	rules	rule	NOUN
ajst-9864	46	9	based	base	VERB
ajst-9864	46	10	on	on	ADP
ajst-9864	46	11	linguistic	linguistic	ADJ
ajst-9864	46	12	knowledge	knowledge	NOUN
ajst-9864	46	13	and	and	CCONJ
ajst-9864	46	14	existing	exist	VERB
ajst-9864	46	15	information	information	NOUN
ajst-9864	46	16	.	.	PUNCT
ajst-9864	47	1	despite	despite	SCONJ
ajst-9864	47	2	their	their	PRON
ajst-9864	47	3	ability	ability	NOUN
ajst-9864	47	4	to	to	PART
ajst-9864	47	5	deliver	deliver	VERB
ajst-9864	47	6	high	high	ADJ
ajst-9864	47	7	-	-	PUNCT
ajst-9864	47	8	precision	precision	NOUN
ajst-9864	47	9	results	result	NOUN
ajst-9864	47	10	,	,	PUNCT
ajst-9864	47	11	the	the	DET
ajst-9864	47	12	intricacies	intricacy	NOUN
ajst-9864	47	13	of	of	ADP
ajst-9864	47	14	language	language	NOUN
ajst-9864	47	15	and	and	CCONJ
ajst-9864	47	16	a	a	DET
ajst-9864	47	17	need	need	NOUN
ajst-9864	47	18	to	to	PART
ajst-9864	47	19	cover	cover	VERB
ajst-9864	47	20	every	every	DET
ajst-9864	47	21	possible	possible	ADJ
ajst-9864	47	22	case	case	NOUN
ajst-9864	47	23	can	can	AUX
ajst-9864	47	24	lead	lead	VERB
ajst-9864	47	25	to	to	ADP
ajst-9864	47	26	errors	error	NOUN
ajst-9864	47	27	.	.	PUNCT
ajst-9864	48	1	statistical	statistical	ADJ
ajst-9864	48	2	methods	method	NOUN
ajst-9864	48	3	employ	employ	VERB
ajst-9864	48	4	statistical	statistical	ADJ
ajst-9864	48	5	learning	learning	NOUN
ajst-9864	48	6	algorithms	algorithm	NOUN
ajst-9864	48	7	,	,	PUNCT
ajst-9864	48	8	requiring	require	VERB
ajst-9864	48	9	vast	vast	ADJ
ajst-9864	48	10	quantities	quantity	NOUN
ajst-9864	48	11	of	of	ADP
ajst-9864	48	12	annotated	annotated	ADJ
ajst-9864	48	13	data	datum	NOUN
ajst-9864	48	14	,	,	PUNCT
ajst-9864	48	15	and	and	CCONJ
ajst-9864	48	16	often	often	ADV
ajst-9864	48	17	struggle	struggle	VERB
ajst-9864	48	18	with	with	ADP
ajst-9864	48	19	the	the	DET
ajst-9864	48	20	complexity	complexity	NOUN
ajst-9864	48	21	of	of	ADP
ajst-9864	48	22	semantic	semantic	ADJ
ajst-9864	48	23	relationships	relationship	NOUN
ajst-9864	48	24	.	.	PUNCT
ajst-9864	49	1	deep	deep	ADJ
ajst-9864	49	2	learning	learning	NOUN
ajst-9864	49	3	-	-	PUNCT
ajst-9864	49	4	based	base	VERB
ajst-9864	49	5	methods	method	NOUN
ajst-9864	49	6	leverage	leverage	VERB
ajst-9864	49	7	neural	neural	ADJ
ajst-9864	49	8	networks	network	NOUN
ajst-9864	49	9	to	to	PART
ajst-9864	49	10	automatically	automatically	ADV
ajst-9864	49	11	learn	learn	VERB
ajst-9864	49	12	entity	entity	NOUN
ajst-9864	49	13	relations	relation	NOUN
ajst-9864	49	14	,	,	PUNCT
ajst-9864	49	15	bypassing	bypass	VERB
ajst-9864	49	16	the	the	DET
ajst-9864	49	17	need	need	NOUN
ajst-9864	49	18	for	for	ADP
ajst-9864	49	19	manual	manual	ADJ
ajst-9864	49	20	rule	rule	NOUN
ajst-9864	49	21	or	or	CCONJ
ajst-9864	49	22	feature	feature	NOUN
ajst-9864	49	23	definition	definition	NOUN
ajst-9864	49	24	.	.	PUNCT
ajst-9864	50	1	these	these	DET
ajst-9864	50	2	techniques	technique	NOUN
ajst-9864	50	3	,	,	PUNCT
ajst-9864	50	4	including	include	VERB
ajst-9864	50	5	convolutional	convolutional	ADJ
ajst-9864	50	6	neural	neural	ADJ
ajst-9864	50	7	networks	network	NOUN
ajst-9864	50	8	(	(	PUNCT
ajst-9864	50	9	cnn	cnn	PROPN
ajst-9864	50	10	)	)	PUNCT
ajst-9864	50	11	,	,	PUNCT
ajst-9864	50	12	recurrent	recurrent	ADJ
ajst-9864	50	13	neural	neural	ADJ
ajst-9864	50	14	networks	network	NOUN
ajst-9864	50	15	(	(	PUNCT
ajst-9864	50	16	rnn	rnn	PROPN
ajst-9864	50	17	)	)	PUNCT
ajst-9864	50	18	,	,	PUNCT
ajst-9864	50	19	long	long	ADJ
ajst-9864	50	20	short	short	ADJ
ajst-9864	50	21	-	-	PUNCT
ajst-9864	50	22	term	term	NOUN
ajst-9864	50	23	memory	memory	NOUN
ajst-9864	50	24	networks	network	NOUN
ajst-9864	50	25	(	(	PUNCT
ajst-9864	50	26	lstm	lstm	PROPN
ajst-9864	50	27	)	)	PUNCT
ajst-9864	50	28	,	,	PUNCT
ajst-9864	50	29	gated	gate	VERB
ajst-9864	50	30	recurrent	recurrent	ADJ
ajst-9864	50	31	unit	unit	NOUN
ajst-9864	50	32	networks	network	NOUN
ajst-9864	50	33	(	(	PUNCT
ajst-9864	50	34	gru	gru	NOUN
ajst-9864	50	35	)	)	PUNCT
ajst-9864	50	36	,	,	PUNCT
ajst-9864	50	37	and	and	CCONJ
ajst-9864	50	38	attention	attention	NOUN
ajst-9864	50	39	mechanisms	mechanism	NOUN
ajst-9864	50	40	,	,	PUNCT
ajst-9864	50	41	can	can	AUX
ajst-9864	50	42	capture	capture	VERB
ajst-9864	50	43	intricate	intricate	ADJ
ajst-9864	50	44	semantic	semantic	ADJ
ajst-9864	50	45	relationships	relationship	NOUN
ajst-9864	50	46	,	,	PUNCT
ajst-9864	50	47	but	but	CCONJ
ajst-9864	50	48	require	require	VERB
ajst-9864	50	49	significant	significant	ADJ
ajst-9864	50	50	data	datum	NOUN
ajst-9864	50	51	and	and	CCONJ
ajst-9864	50	52	computational	computational	ADJ
ajst-9864	50	53	resources[6	resources[6	ADP
ajst-9864	50	54	]	]	X
ajst-9864	50	55	..	..	PUNCT
ajst-9864	50	56	most	most	ADJ
ajst-9864	50	57	current	current	ADJ
ajst-9864	50	58	entity	entity	NOUN
ajst-9864	50	59	relationship	relationship	NOUN
ajst-9864	50	60	extraction	extraction	NOUN
ajst-9864	50	61	research	research	NOUN
ajst-9864	50	62	is	be	AUX
ajst-9864	50	63	centered	center	VERB
ajst-9864	50	64	around	around	ADP
ajst-9864	50	65	english	english	PROPN
ajst-9864	50	66	,	,	PUNCT
ajst-9864	50	67	with	with	ADP
ajst-9864	50	68	fewer	few	ADJ
ajst-9864	50	69	studies	study	NOUN
ajst-9864	50	70	focusing	focus	VERB
ajst-9864	50	71	on	on	ADP
ajst-9864	50	72	chinese	chinese	PROPN
ajst-9864	50	73	,	,	PUNCT
ajst-9864	50	74	leading	lead	VERB
ajst-9864	50	75	to	to	ADP
ajst-9864	50	76	language	language	NOUN
ajst-9864	50	77	-	-	PUNCT
ajst-9864	50	78	specific	specific	ADJ
ajst-9864	50	79	challenges	challenge	NOUN
ajst-9864	50	80	.	.	PUNCT
ajst-9864	51	1	several	several	ADJ
ajst-9864	51	2	researchers	researcher	NOUN
ajst-9864	51	3	have	have	AUX
ajst-9864	51	4	proposed	propose	VERB
ajst-9864	51	5	solutions	solution	NOUN
ajst-9864	51	6	such	such	ADJ
ajst-9864	51	7	as	as	ADP
ajst-9864	51	8	feature	feature	NOUN
ajst-9864	51	9	-	-	PUNCT
ajst-9864	51	10	based	base	VERB
ajst-9864	51	11	,	,	PUNCT
ajst-9864	51	12	kernel	kernel	PROPN
ajst-9864	51	13	function	function	NOUN
ajst-9864	51	14	-	-	PUNCT
ajst-9864	51	15	based	base	VERB
ajst-9864	51	16	,	,	PUNCT
ajst-9864	51	17	or	or	CCONJ
ajst-9864	51	18	deep	deep	ADJ
ajst-9864	51	19	belief	belief	NOUN
ajst-9864	51	20	network	network	NOUN
ajst-9864	51	21	-	-	PUNCT
ajst-9864	51	22	based	base	VERB
ajst-9864	51	23	extraction	extraction	NOUN
ajst-9864	51	24	methods	method	NOUN
ajst-9864	51	25	for	for	ADP
ajst-9864	51	26	chinese	chinese	ADJ
ajst-9864	51	27	entities	entity	NOUN
ajst-9864	51	28	.	.	PUNCT
ajst-9864	52	1	deep	deep	ADJ
ajst-9864	52	2	learning	learning	NOUN
ajst-9864	52	3	has	have	AUX
ajst-9864	52	4	made	make	VERB
ajst-9864	52	5	substantial	substantial	ADJ
ajst-9864	52	6	strides	stride	NOUN
ajst-9864	52	7	in	in	ADP
ajst-9864	52	8	recent	recent	ADJ
ajst-9864	52	9	years	year	NOUN
ajst-9864	52	10	in	in	ADP
ajst-9864	52	11	the	the	DET
ajst-9864	52	12	domain	domain	NOUN
ajst-9864	52	13	of	of	ADP
ajst-9864	52	14	text	text	NOUN
ajst-9864	52	15	entity	entity	NOUN
ajst-9864	52	16	relationship	relationship	NOUN
ajst-9864	52	17	extraction	extraction	NOUN
ajst-9864	52	18	.	.	PUNCT
ajst-9864	53	1	studies	study	NOUN
ajst-9864	53	2	are	be	AUX
ajst-9864	53	3	now	now	ADV
ajst-9864	53	4	focusing	focus	VERB
ajst-9864	53	5	on	on	ADP
ajst-9864	53	6	integrating	integrate	VERB
ajst-9864	53	7	deep	deep	ADJ
ajst-9864	53	8	learning	learning	NOUN
ajst-9864	53	9	with	with	ADP
ajst-9864	53	10	external	external	ADJ
ajst-9864	53	11	knowledge	knowledge	NOUN
ajst-9864	53	12	bases	basis	NOUN
ajst-9864	53	13	or	or	CCONJ
ajst-9864	53	14	utilizing	utilize	VERB
ajst-9864	53	15	advanced	advanced	ADJ
ajst-9864	53	16	techniques	technique	NOUN
ajst-9864	53	17	like	like	ADP
ajst-9864	53	18	migration	migration	NOUN
ajst-9864	53	19	learning	learning	NOUN
ajst-9864	53	20	and	and	CCONJ
ajst-9864	53	21	reinforcement	reinforcement	NOUN
ajst-9864	53	22	learning	learning	NOUN
ajst-9864	53	23	to	to	PART
ajst-9864	53	24	enhance	enhance	VERB
ajst-9864	53	25	model	model	NOUN
ajst-9864	53	26	performance	performance	NOUN
ajst-9864	53	27	.	.	PUNCT
ajst-9864	54	1	graph	graph	VERB
ajst-9864	54	2	neural	neural	ADJ
ajst-9864	54	3	network	network	NOUN
ajst-9864	54	4	-	-	PUNCT
ajst-9864	54	5	based	base	VERB
ajst-9864	54	6	methods	method	NOUN
ajst-9864	54	7	have	have	AUX
ajst-9864	54	8	also	also	ADV
ajst-9864	54	9	shown	show	VERB
ajst-9864	54	10	promising	promising	ADJ
ajst-9864	54	11	results	result	NOUN
ajst-9864	54	12	,	,	PUNCT
ajst-9864	54	13	using	use	VERB
ajst-9864	54	14	graph	graph	NOUN
ajst-9864	54	15	structures	structure	NOUN
ajst-9864	54	16	to	to	PART
ajst-9864	54	17	represent	represent	VERB
ajst-9864	54	18	entities	entity	NOUN
ajst-9864	54	19	and	and	CCONJ
ajst-9864	54	20	relationships[7	relationships[7	NOUN
ajst-9864	54	21	]	]	X
ajst-9864	54	22	.	.	PUNCT
ajst-9864	55	1	the	the	DET
ajst-9864	55	2	gru	gru	PROPN
ajst-9864	55	3	dual	dual	ADV
ajst-9864	55	4	-	-	PUNCT
ajst-9864	55	5	supervised	supervise	VERB
ajst-9864	55	6	model	model	NOUN
ajst-9864	55	7	,	,	PUNCT
ajst-9864	55	8	utilizing	utilize	VERB
ajst-9864	55	9	both	both	DET
ajst-9864	55	10	labelled	label	VERB
ajst-9864	55	11	and	and	CCONJ
ajst-9864	55	12	unlabeled	unlabeled	ADJ
ajst-9864	55	13	data	datum	NOUN
ajst-9864	55	14	,	,	PUNCT
ajst-9864	55	15	is	be	AUX
ajst-9864	55	16	a	a	DET
ajst-9864	55	17	valuable	valuable	ADJ
ajst-9864	55	18	text	text	NOUN
ajst-9864	55	19	classification	classification	NOUN
ajst-9864	55	20	method	method	NOUN
ajst-9864	55	21	,	,	PUNCT
ajst-9864	55	22	expected	expect	VERB
ajst-9864	55	23	to	to	PART
ajst-9864	55	24	see	see	VERB
ajst-9864	55	25	wider	wide	ADJ
ajst-9864	55	26	application	application	NOUN
ajst-9864	55	27	and	and	CCONJ
ajst-9864	55	28	further	further	ADJ
ajst-9864	55	29	improvement	improvement	NOUN
ajst-9864	55	30	in	in	ADP
ajst-9864	55	31	future	future	ADJ
ajst-9864	55	32	research	research	NOUN
ajst-9864	55	33	.	.	PUNCT
ajst-9864	56	1	with	with	ADP
ajst-9864	56	2	continued	continue	VERB
ajst-9864	56	3	evolution	evolution	NOUN
ajst-9864	56	4	in	in	ADP
ajst-9864	56	5	technology	technology	NOUN
ajst-9864	56	6	and	and	CCONJ
ajst-9864	56	7	expanding	expand	VERB
ajst-9864	56	8	datasets	dataset	NOUN
ajst-9864	56	9	,	,	PUNCT
ajst-9864	56	10	deep	deep	ADJ
ajst-9864	56	11	learning	learning	NOUN
ajst-9864	56	12	models	model	NOUN
ajst-9864	56	13	are	be	AUX
ajst-9864	56	14	poised	poise	VERB
ajst-9864	56	15	to	to	PART
ajst-9864	56	16	become	become	VERB
ajst-9864	56	17	an	an	DET
ajst-9864	56	18	increasingly	increasingly	ADV
ajst-9864	56	19	important	important	ADJ
ajst-9864	56	20	player	player	NOUN
ajst-9864	56	21	in	in	ADP
ajst-9864	56	22	the	the	DET
ajst-9864	56	23	field	field	NOUN
ajst-9864	56	24	of	of	ADP
ajst-9864	56	25	text	text	NOUN
ajst-9864	56	26	entity	entity	NOUN
ajst-9864	56	27	relationship	relationship	NOUN
ajst-9864	56	28	extraction	extraction	NOUN
ajst-9864	56	29	.	.	PUNCT
ajst-9864	57	1	3	3	X
ajst-9864	57	2	.	.	X
ajst-9864	57	3	research	research	NOUN
ajst-9864	57	4	methods	method	NOUN
ajst-9864	57	5	for	for	ADP
ajst-9864	57	6	the	the	DET
ajst-9864	57	7	extraction	extraction	NOUN
ajst-9864	57	8	of	of	ADP
ajst-9864	57	9	character	character	NOUN
ajst-9864	57	10	associations	association	NOUN
ajst-9864	57	11	in	in	ADP
ajst-9864	57	12	chinese	chinese	ADJ
ajst-9864	57	13	text	text	NOUN
ajst-9864	57	14	,	,	PUNCT
ajst-9864	57	15	we	we	PRON
ajst-9864	57	16	suggest	suggest	VERB
ajst-9864	57	17	a	a	DET
ajst-9864	57	18	two	two	NUM
ajst-9864	57	19	-	-	PUNCT
ajst-9864	57	20	way	way	NOUN
ajst-9864	57	21	gru	gru	NOUN
ajst-9864	57	22	neural	neural	ADJ
ajst-9864	57	23	network	network	NOUN
ajst-9864	57	24	in	in	ADP
ajst-9864	57	25	this	this	DET
ajst-9864	57	26	research	research	NOUN
ajst-9864	57	27	using	use	VERB
ajst-9864	57	28	a	a	DET
ajst-9864	57	29	two	two	NUM
ajst-9864	57	30	-	-	PUNCT
ajst-9864	57	31	level	level	NOUN
ajst-9864	57	32	attention	attention	NOUN
ajst-9864	57	33	mechanism	mechanism	NOUN
ajst-9864	57	34	.	.	PUNCT
ajst-9864	58	1	the	the	DET
ajst-9864	58	2	model	model	NOUN
ajst-9864	58	3	's	's	PART
ajst-9864	58	4	architecture	architecture	NOUN
ajst-9864	58	5	is	be	AUX
ajst-9864	58	6	shown	show	VERB
ajst-9864	58	7	in	in	ADP
ajst-9864	58	8	figure	figure	NOUN
ajst-9864	58	9	1	1	NUM
ajst-9864	58	10	,	,	PUNCT
ajst-9864	58	11	with	with	ADP
ajst-9864	58	12	word	word	NOUN
ajst-9864	58	13	-	-	PUNCT
ajst-9864	58	14	level	level	NOUN
ajst-9864	58	15	attention	attention	NOUN
ajst-9864	58	16	representing	represent	VERB
ajst-9864	58	17	the	the	DET
ajst-9864	58	18	word	word	NOUN
ajst-9864	58	19	-	-	PUNCT
ajst-9864	58	20	level	level	NOUN
ajst-9864	58	21	attention	attention	NOUN
ajst-9864	58	22	mechanism	mechanism	NOUN
ajst-9864	58	23	and	and	CCONJ
ajst-9864	58	24	sentence	sentence	NOUN
ajst-9864	58	25	-	-	PUNCT
ajst-9864	58	26	level	level	NOUN
ajst-9864	58	27	attention	attention	NOUN
ajst-9864	58	28	representing	represent	VERB
ajst-9864	58	29	the	the	DET
ajst-9864	58	30	sentence	sentence	NOUN
ajst-9864	58	31	-	-	PUNCT
ajst-9864	58	32	level	level	NOUN
ajst-9864	58	33	attention	attention	NOUN
ajst-9864	58	34	mechanism	mechanism	NOUN
ajst-9864	58	35	.	.	PUNCT
ajst-9864	59	1	the	the	DET
ajst-9864	59	2	feature	feature	NOUN
ajst-9864	59	3	vectors	vector	NOUN
ajst-9864	59	4	are	be	AUX
ajst-9864	59	5	then	then	ADV
ajst-9864	59	6	used	use	VERB
ajst-9864	59	7	to	to	PART
ajst-9864	59	8	construct	construct	VERB
ajst-9864	59	9	word	word	NOUN
ajst-9864	59	10	-	-	PUNCT
ajst-9864	59	11	level	level	NOUN
ajst-9864	59	12	and	and	CCONJ
ajst-9864	59	13	sentence	sentence	NOUN
ajst-9864	59	14	-	-	PUNCT
ajst-9864	59	15	level	level	NOUN
ajst-9864	59	16	attention	attention	NOUN
ajst-9864	59	17	approaches	approach	NOUN
ajst-9864	59	18	,	,	PUNCT
ajst-9864	59	19	both	both	PRON
ajst-9864	59	20	of	of	ADP
ajst-9864	59	21	which	which	PRON
ajst-9864	59	22	are	be	AUX
ajst-9864	59	23	used	use	VERB
ajst-9864	59	24	to	to	PART
ajst-9864	59	25	mitigate	mitigate	VERB
ajst-9864	59	26	the	the	DET
ajst-9864	59	27	effects	effect	NOUN
ajst-9864	59	28	of	of	ADP
ajst-9864	59	29	noisy	noisy	ADJ
ajst-9864	59	30	data	datum	NOUN
ajst-9864	59	31	.	.	PUNCT
ajst-9864	60	1	experiment	experiment	NOUN
ajst-9864	60	2	findings	finding	NOUN
ajst-9864	60	3	show	show	VERB
ajst-9864	60	4	that	that	SCONJ
ajst-9864	60	5	the	the	DET
ajst-9864	60	6	bidirectional	bidirectional	PROPN
ajst-9864	60	7	gru	gru	PROPN
ajst-9864	60	8	neural	neural	PROPN
ajst-9864	60	9	network	network	NOUN
ajst-9864	60	10	outperforms	outperform	VERB
ajst-9864	60	11	the	the	DET
ajst-9864	60	12	more	more	ADV
ajst-9864	60	13	common	common	ADJ
ajst-9864	60	14	neural	neural	ADJ
ajst-9864	60	15	network	network	NOUN
ajst-9864	60	16	model	model	NOUN
ajst-9864	60	17	in	in	ADP
ajst-9864	60	18	terms	term	NOUN
ajst-9864	60	19	of	of	ADP
ajst-9864	60	20	accuracy[8	accuracy[8	NOUN
ajst-9864	60	21	]	]	PUNCT
ajst-9864	60	22	.	.	PUNCT
ajst-9864	61	1	the	the	DET
ajst-9864	61	2	proposed	propose	VERB
ajst-9864	61	3	bidirectional	bidirectional	PROPN
ajst-9864	61	4	gru	gru	PROPN
ajst-9864	61	5	+	+	CCONJ
ajst-9864	61	6	word	word	NOUN
ajst-9864	61	7	-	-	PUNCT
ajst-9864	61	8	level	level	NOUN
ajst-9864	61	9	attention	attention	NOUN
ajst-9864	61	10	+	+	CCONJ
ajst-9864	61	11	sentence	sentence	NOUN
ajst-9864	61	12	-	-	PUNCT
ajst-9864	61	13	level	level	NOUN
ajst-9864	61	14	attention	attention	NOUN
ajst-9864	61	15	model	model	NOUN
ajst-9864	61	16	consists	consist	VERB
ajst-9864	61	17	of	of	ADP
ajst-9864	61	18	6	6	NUM
ajst-9864	61	19	parts	part	NOUN
ajst-9864	61	20	:	:	PUNCT
ajst-9864	61	21	1	1	NUM
ajst-9864	61	22	)	)	PUNCT
ajst-9864	61	23	input	input	NOUN
ajst-9864	61	24	layer	layer	NOUN
ajst-9864	61	25	:	:	PUNCT
ajst-9864	61	26	the	the	DET
ajst-9864	61	27	sentences	sentence	NOUN
ajst-9864	61	28	are	be	AUX
ajst-9864	61	29	input	input	VERB
ajst-9864	61	30	into	into	ADP
ajst-9864	61	31	the	the	DET
ajst-9864	61	32	model	model	NOUN
ajst-9864	61	33	.	.	PUNCT
ajst-9864	62	1	2	2	X
ajst-9864	62	2	)	)	PUNCT
ajst-9864	62	3	embedding	embed	VERB
ajst-9864	62	4	layer	layer	NOUN
ajst-9864	62	5	:	:	PUNCT
ajst-9864	62	6	mapping	map	VERB
ajst-9864	62	7	each	each	DET
ajst-9864	62	8	word	word	NOUN
ajst-9864	62	9	to	to	ADP
ajst-9864	62	10	a	a	DET
ajst-9864	62	11	lowdimensional	lowdimensional	ADJ
ajst-9864	62	12	vector	vector	NOUN
ajst-9864	62	13	using	use	VERB
ajst-9864	62	14	word2vec	word2vec	PROPN
ajst-9864	62	15	tool	tool	NOUN
ajst-9864	62	16	.	.	PUNCT
ajst-9864	63	1	3	3	X
ajst-9864	63	2	)	)	PUNCT
ajst-9864	63	3	gru	gru	NOUN
ajst-9864	63	4	layer	layer	NOUN
ajst-9864	63	5	:	:	PUNCT
ajst-9864	63	6	obtains	obtain	VERB
ajst-9864	63	7	the	the	DET
ajst-9864	63	8	feature	feature	NOUN
ajst-9864	63	9	vector	vector	NOUN
ajst-9864	63	10	of	of	ADP
ajst-9864	63	11	the	the	DET
ajst-9864	63	12	original	original	ADJ
ajst-9864	63	13	sentence	sentence	NOUN
ajst-9864	63	14	through	through	ADP
ajst-9864	63	15	a	a	DET
ajst-9864	63	16	bidirectional	bidirectional	PROPN
ajst-9864	63	17	gru	gru	PROPN
ajst-9864	63	18	neural	neural	ADJ
ajst-9864	63	19	network	network	NOUN
ajst-9864	63	20	;	;	PUNCT
ajst-9864	63	21	4	4	X
ajst-9864	63	22	)	)	PUNCT
ajst-9864	63	23	word	word	NOUN
ajst-9864	63	24	-	-	PUNCT
ajst-9864	63	25	level	level	NOUN
ajst-9864	63	26	attention	attention	NOUN
ajst-9864	63	27	layer	layer	NOUN
ajst-9864	63	28	:	:	PUNCT
ajst-9864	63	29	generate	generate	VERB
ajst-9864	63	30	the	the	DET
ajst-9864	63	31	weight	weight	NOUN
ajst-9864	63	32	vector	vector	NOUN
ajst-9864	63	33	w	w	NOUN
ajst-9864	63	34	and	and	CCONJ
ajst-9864	63	35	multiply	multiply	VERB
ajst-9864	63	36	the	the	DET
ajst-9864	63	37	feature	feature	NOUN
ajst-9864	63	38	vector	vector	NOUN
ajst-9864	63	39	obtained	obtain	VERB
ajst-9864	63	40	from	from	ADP
ajst-9864	63	41	the	the	DET
ajst-9864	63	42	gru	gru	NOUN
ajst-9864	63	43	layer	layer	NOUN
ajst-9864	63	44	by	by	ADP
ajst-9864	63	45	the	the	DET
ajst-9864	63	46	weight	weight	NOUN
ajst-9864	63	47	vector	vector	NOUN
ajst-9864	63	48	w	w	ADP
ajst-9864	63	49	to	to	PART
ajst-9864	63	50	obtain	obtain	VERB
ajst-9864	63	51	the	the	DET
ajst-9864	63	52	word	word	NOUN
ajst-9864	63	53	-	-	PUNCT
ajst-9864	63	54	level	level	NOUN
ajst-9864	63	55	feature	feature	NOUN
ajst-9864	63	56	vector	vector	NOUN
ajst-9864	63	57	.	.	NOUN
ajst-9864	64	1	5	5	NUM
ajst-9864	64	2	)	)	PUNCT
ajst-9864	64	3	sentence	sentence	NOUN
ajst-9864	64	4	level	level	NOUN
ajst-9864	64	5	attention	attention	NOUN
ajst-9864	64	6	layer	layer	NOUN
ajst-9864	64	7	:	:	PUNCT
ajst-9864	64	8	generate	generate	VERB
ajst-9864	64	9	the	the	DET
ajst-9864	64	10	weight	weight	NOUN
ajst-9864	64	11	vector	vector	NOUN
ajst-9864	64	12	and	and	CCONJ
ajst-9864	64	13	multiply	multiply	VERB
ajst-9864	64	14	the	the	DET
ajst-9864	64	15	word	word	NOUN
ajst-9864	64	16	level	level	NOUN
ajst-9864	64	17	feature	feature	NOUN
ajst-9864	64	18	vector	vector	NOUN
ajst-9864	64	19	in	in	ADP
ajst-9864	64	20	4	4	NUM
ajst-9864	64	21	)	)	PUNCT
ajst-9864	64	22	by	by	ADP
ajst-9864	64	23	the	the	DET
ajst-9864	64	24	weight	weight	NOUN
ajst-9864	64	25	vector	vector	NOUN
ajst-9864	64	26	a	a	PRON
ajst-9864	64	27	to	to	PART
ajst-9864	64	28	get	get	VERB
ajst-9864	64	29	the	the	DET
ajst-9864	64	30	sentence	sentence	NOUN
ajst-9864	64	31	level	level	NOUN
ajst-9864	64	32	feature	feature	NOUN
ajst-9864	64	33	vector	vector	NOUN
ajst-9864	64	34	.	.	NOUN
ajst-9864	65	1	6	6	NUM
ajst-9864	65	2	)	)	PUNCT
ajst-9864	65	3	output	output	NOUN
ajst-9864	65	4	layer	layer	NOUN
ajst-9864	65	5	:	:	PUNCT
ajst-9864	65	6	the	the	DET
ajst-9864	65	7	feature	feature	NOUN
ajst-9864	65	8	vectors	vector	NOUN
ajst-9864	65	9	obtained	obtain	VERB
ajst-9864	65	10	by	by	ADP
ajst-9864	65	11	the	the	DET
ajst-9864	65	12	twolayer	twolayer	NOUN
ajst-9864	65	13	attention	attention	NOUN
ajst-9864	65	14	mechanism	mechanism	NOUN
ajst-9864	65	15	are	be	AUX
ajst-9864	65	16	passed	pass	VERB
ajst-9864	65	17	through	through	ADP
ajst-9864	65	18	the	the	DET
ajst-9864	65	19	softmax	softmax	NOUN
ajst-9864	65	20	classifier	classifier	NOUN
ajst-9864	65	21	to	to	PART
ajst-9864	65	22	obtain	obtain	VERB
ajst-9864	65	23	the	the	DET
ajst-9864	65	24	final	final	ADJ
ajst-9864	65	25	relationship	relationship	NOUN
ajst-9864	65	26	classification	classification	NOUN
ajst-9864	65	27	results	result	NOUN
ajst-9864	65	28	.	.	PUNCT
ajst-9864	66	1	130	130	NUM
ajst-9864	66	2	figure	figure	NOUN
ajst-9864	66	3	1	1	NUM
ajst-9864	66	4	.	.	PUNCT
ajst-9864	66	5	bidirectional	bidirectional	PROPN
ajst-9864	66	6	gru	gru	PROPN
ajst-9864	66	7	+	+	CCONJ
ajst-9864	66	8	word	word	NOUN
ajst-9864	66	9	-	-	PUNCT
ajst-9864	66	10	level	level	NOUN
ajst-9864	66	11	attention	attention	NOUN
ajst-9864	66	12	+	+	CCONJ
ajst-9864	66	13	sentence	sentence	NOUN
ajst-9864	66	14	-	-	PUNCT
ajst-9864	66	15	level	level	NOUN
ajst-9864	66	16	attention	attention	NOUN
ajst-9864	66	17	3.1	3.1	NUM
ajst-9864	66	18	.	.	PUNCT
ajst-9864	67	1	research	research	NOUN
ajst-9864	67	2	design	design	NOUN
ajst-9864	67	3	3.1.1	3.1.1	NUM
ajst-9864	67	4	.	.	PUNCT
ajst-9864	68	1	word	word	NOUN
ajst-9864	68	2	sequence	sequence	NOUN
ajst-9864	68	3	processing	process	VERB
ajst-9864	68	4	deep	deep	ADJ
ajst-9864	68	5	learning	learning	NOUN
ajst-9864	68	6	requires	require	VERB
ajst-9864	68	7	the	the	DET
ajst-9864	68	8	mathematical	mathematical	ADJ
ajst-9864	68	9	representation	representation	NOUN
ajst-9864	68	10	of	of	ADP
ajst-9864	68	11	language	language	NOUN
ajst-9864	68	12	,	,	PUNCT
ajst-9864	68	13	converting	convert	VERB
ajst-9864	68	14	text	text	NOUN
ajst-9864	68	15	to	to	ADP
ajst-9864	68	16	vector	vector	NOUN
ajst-9864	68	17	form	form	NOUN
ajst-9864	68	18	for	for	ADP
ajst-9864	68	19	machine	machine	NOUN
ajst-9864	68	20	processing	processing	NOUN
ajst-9864	68	21	.	.	PUNCT
ajst-9864	69	1	traditional	traditional	ADJ
ajst-9864	69	2	word	word	NOUN
ajst-9864	69	3	embedding	embed	VERB
ajst-9864	69	4	approaches	approach	NOUN
ajst-9864	69	5	,	,	PUNCT
ajst-9864	69	6	such	such	ADJ
ajst-9864	69	7	as	as	ADP
ajst-9864	69	8	word2vec	word2vec	X
ajst-9864	69	9	,	,	PUNCT
ajst-9864	69	10	often	often	ADV
ajst-9864	69	11	ignore	ignore	VERB
ajst-9864	69	12	the	the	DET
ajst-9864	69	13	internal	internal	ADJ
ajst-9864	69	14	structure	structure	NOUN
ajst-9864	69	15	of	of	ADP
ajst-9864	69	16	words	word	NOUN
ajst-9864	69	17	,	,	PUNCT
ajst-9864	69	18	especially	especially	ADV
ajst-9864	69	19	relevant	relevant	ADJ
ajst-9864	69	20	in	in	ADP
ajst-9864	69	21	languages	language	NOUN
ajst-9864	69	22	like	like	ADP
ajst-9864	69	23	chinese	chinese	PROPN
ajst-9864	69	24	.	.	PUNCT
ajst-9864	70	1	in	in	ADP
ajst-9864	70	2	this	this	DET
ajst-9864	70	3	study	study	NOUN
ajst-9864	70	4	,	,	PUNCT
ajst-9864	70	5	we	we	PRON
ajst-9864	70	6	employ	employ	VERB
ajst-9864	70	7	word2vec	word2vec	X
ajst-9864	70	8	for	for	ADP
ajst-9864	70	9	creating	create	VERB
ajst-9864	70	10	chinese	chinese	ADJ
ajst-9864	70	11	word	word	NOUN
ajst-9864	70	12	vectors	vector	NOUN
ajst-9864	70	13	.	.	PUNCT
ajst-9864	71	1	word2vec	word2vec	X
ajst-9864	71	2	,	,	PUNCT
ajst-9864	71	3	using	use	VERB
ajst-9864	71	4	methods	method	NOUN
ajst-9864	71	5	like	like	ADP
ajst-9864	71	6	continuous	continuous	ADJ
ajst-9864	71	7	bag	bag	NOUN
ajst-9864	71	8	-	-	PUNCT
ajst-9864	71	9	of	of	ADP
ajst-9864	71	10	-	-	PUNCT
ajst-9864	71	11	words	word	NOUN
ajst-9864	71	12	(	(	PUNCT
ajst-9864	71	13	cbow	cbow	VERB
ajst-9864	71	14	)	)	PUNCT
ajst-9864	71	15	and	and	CCONJ
ajst-9864	71	16	skip	skip	VERB
ajst-9864	71	17	-	-	PUNCT
ajst-9864	71	18	gram	gram	NOUN
ajst-9864	71	19	,	,	PUNCT
ajst-9864	71	20	predicts	predict	VERB
ajst-9864	71	21	either	either	CCONJ
ajst-9864	71	22	the	the	DET
ajst-9864	71	23	centre	centre	ADJ
ajst-9864	71	24	word	word	NOUN
ajst-9864	71	25	or	or	CCONJ
ajst-9864	71	26	the	the	DET
ajst-9864	71	27	surrounding	surround	VERB
ajst-9864	71	28	words	word	NOUN
ajst-9864	71	29	in	in	ADP
ajst-9864	71	30	a	a	DET
ajst-9864	71	31	phrase	phrase	NOUN
ajst-9864	71	32	,	,	PUNCT
ajst-9864	71	33	aiming	aim	VERB
ajst-9864	71	34	to	to	ADP
ajst-9864	71	35	group	group	NOUN
ajst-9864	71	36	similar	similar	ADJ
ajst-9864	71	37	words	word	NOUN
ajst-9864	71	38	in	in	ADP
ajst-9864	71	39	the	the	DET
ajst-9864	71	40	vector	vector	NOUN
ajst-9864	71	41	space[9	space[9	PROPN
ajst-9864	71	42	]	]	PUNCT
ajst-9864	71	43	.	.	PUNCT
ajst-9864	72	1	key	key	ADJ
ajst-9864	72	2	properties	property	NOUN
ajst-9864	72	3	of	of	ADP
ajst-9864	72	4	word	word	NOUN
ajst-9864	72	5	vectors	vector	NOUN
ajst-9864	72	6	include	include	VERB
ajst-9864	72	7	:	:	PUNCT
ajst-9864	72	8	(	(	PUNCT
ajst-9864	72	9	1	1	X
ajst-9864	72	10	)	)	PUNCT
ajst-9864	72	11	similarity	similarity	NOUN
ajst-9864	72	12	between	between	ADP
ajst-9864	72	13	vectors	vector	NOUN
ajst-9864	72	14	represents	represent	VERB
ajst-9864	72	15	similarity	similarity	NOUN
ajst-9864	72	16	between	between	ADP
ajst-9864	72	17	words	word	NOUN
ajst-9864	72	18	,	,	PUNCT
ajst-9864	72	19	useful	useful	ADJ
ajst-9864	72	20	for	for	ADP
ajst-9864	72	21	tasks	task	NOUN
ajst-9864	72	22	like	like	ADP
ajst-9864	72	23	sentence	sentence	NOUN
ajst-9864	72	24	similarity	similarity	NOUN
ajst-9864	72	25	and	and	CCONJ
ajst-9864	72	26	text	text	NOUN
ajst-9864	72	27	classification	classification	NOUN
ajst-9864	72	28	,	,	PUNCT
ajst-9864	72	29	and	and	CCONJ
ajst-9864	72	30	(	(	PUNCT
ajst-9864	72	31	2	2	X
ajst-9864	72	32	)	)	PUNCT
ajst-9864	72	33	vector	vector	NOUN
ajst-9864	72	34	operations	operation	NOUN
ajst-9864	72	35	can	can	AUX
ajst-9864	72	36	depict	depict	VERB
ajst-9864	72	37	semantic	semantic	ADJ
ajst-9864	72	38	relationships	relationship	NOUN
ajst-9864	72	39	between	between	ADP
ajst-9864	72	40	words	word	NOUN
ajst-9864	72	41	,	,	PUNCT
ajst-9864	72	42	for	for	ADP
ajst-9864	72	43	instance	instance	NOUN
ajst-9864	72	44	,	,	PUNCT
ajst-9864	72	45	solving	solve	VERB
ajst-9864	72	46	analogy	analogy	NOUN
ajst-9864	72	47	problems	problem	NOUN
ajst-9864	72	48	.	.	PUNCT
ajst-9864	73	1	word2vec	word2vec	PROPN
ajst-9864	73	2	vectorizes	vectorize	VERB
ajst-9864	73	3	the	the	DET
ajst-9864	73	4	contextual	contextual	ADJ
ajst-9864	73	5	information	information	NOUN
ajst-9864	73	6	of	of	ADP
ajst-9864	73	7	words	word	NOUN
ajst-9864	73	8	,	,	PUNCT
ajst-9864	73	9	initially	initially	ADV
ajst-9864	73	10	mapping	map	VERB
ajst-9864	73	11	each	each	DET
ajst-9864	73	12	word	word	NOUN
ajst-9864	73	13	to	to	ADP
ajst-9864	73	14	a	a	DET
ajst-9864	73	15	high	high	ADV
ajst-9864	73	16	-	-	PUNCT
ajst-9864	73	17	dimensional	dimensional	ADJ
ajst-9864	73	18	vector	vector	NOUN
ajst-9864	73	19	,	,	PUNCT
ajst-9864	73	20	later	later	ADV
ajst-9864	73	21	reduced	reduce	VERB
ajst-9864	73	22	to	to	ADP
ajst-9864	73	23	a	a	DET
ajst-9864	73	24	lower	lower	ADV
ajst-9864	73	25	-	-	PUNCT
ajst-9864	73	26	dimensional	dimensional	ADJ
ajst-9864	73	27	space	space	NOUN
ajst-9864	73	28	.	.	PUNCT
ajst-9864	74	1	it	it	PRON
ajst-9864	74	2	updates	update	VERB
ajst-9864	74	3	word	word	NOUN
ajst-9864	74	4	vectors	vector	NOUN
ajst-9864	74	5	based	base	VERB
ajst-9864	74	6	on	on	ADP
ajst-9864	74	7	contextual	contextual	ADJ
ajst-9864	74	8	information	information	NOUN
ajst-9864	74	9	and	and	CCONJ
ajst-9864	74	10	has	have	VERB
ajst-9864	74	11	widespread	widespread	ADJ
ajst-9864	74	12	application	application	NOUN
ajst-9864	74	13	due	due	ADP
ajst-9864	74	14	to	to	ADP
ajst-9864	74	15	its	its	PRON
ajst-9864	74	16	desirable	desirable	ADJ
ajst-9864	74	17	properties	property	NOUN
ajst-9864	74	18	.	.	PUNCT
ajst-9864	75	1	the	the	DET
ajst-9864	75	2	conversion	conversion	NOUN
ajst-9864	75	3	of	of	ADP
ajst-9864	75	4	a	a	DET
ajst-9864	75	5	chinese	chinese	ADJ
ajst-9864	75	6	statement	statement	NOUN
ajst-9864	75	7	into	into	ADP
ajst-9864	75	8	a	a	DET
ajst-9864	75	9	vector	vector	NOUN
ajst-9864	75	10	form	form	NOUN
ajst-9864	75	11	follows	follow	VERB
ajst-9864	75	12	the	the	DET
ajst-9864	75	13	equation	equation	NOUN
ajst-9864	75	14	ei	ei	NOUN
ajst-9864	75	15	=	=	PUNCT
ajst-9864	75	16	wword	wword	PROPN
ajst-9864	75	17	vi	vi	PROPN
ajst-9864	75	18	(	(	PUNCT
ajst-9864	75	19	1	1	NUM
ajst-9864	75	20	)	)	PUNCT
ajst-9864	75	21	where	where	SCONJ
ajst-9864	75	22	:	:	PUNCT
ajst-9864	75	23	vi	vi	PROPN
ajst-9864	75	24	is	be	AUX
ajst-9864	75	25	a	a	DET
ajst-9864	75	26	one	one	NUM
ajst-9864	75	27	-	-	PUNCT
ajst-9864	75	28	dimensional	dimensional	ADJ
ajst-9864	75	29	vector	vector	NOUN
ajst-9864	75	30	of	of	ADP
ajst-9864	75	31	size	size	NOUN
ajst-9864	75	32	|v|	|v|	INTJ
ajst-9864	75	33	with	with	ADP
ajst-9864	75	34	a	a	DET
ajst-9864	75	35	value	value	NOUN
ajst-9864	75	36	of	of	ADP
ajst-9864	75	37	1	1	NUM
ajst-9864	75	38	at	at	ADP
ajst-9864	75	39	the	the	DET
ajst-9864	75	40	subscript	subscript	NOUN
ajst-9864	75	41	ei	ei	NOUN
ajst-9864	75	42	and	and	CCONJ
ajst-9864	75	43	a	a	DET
ajst-9864	75	44	value	value	NOUN
ajst-9864	75	45	of	of	ADP
ajst-9864	75	46	0	0	NUM
ajst-9864	75	47	elsewhere	elsewhere	ADV
ajst-9864	75	48	.	.	PUNCT
ajst-9864	76	1	thus	thus	ADV
ajst-9864	76	2	,	,	PUNCT
ajst-9864	76	3	the	the	DET
ajst-9864	76	4	chinese	chinese	ADJ
ajst-9864	76	5	statement	statement	NOUN
ajst-9864	76	6	is	be	AUX
ajst-9864	76	7	converted	convert	VERB
ajst-9864	76	8	into	into	ADP
ajst-9864	76	9	a	a	DET
ajst-9864	76	10	vector	vector	NOUN
ajst-9864	76	11	embs	embs	ADJ
ajst-9864	76	12	=	=	SYM
ajst-9864	76	13	{	{	PUNCT
ajst-9864	76	14	e1	e1	PROPN
ajst-9864	76	15	,	,	PUNCT
ajst-9864	76	16	e2	e2	PROPN
ajst-9864	76	17	,	,	PUNCT
ajst-9864	76	18	...	...	PUNCT
ajst-9864	76	19	,	,	PUNCT
ajst-9864	76	20	et	et	NOUN
ajst-9864	76	21	}	}	PUNCT
ajst-9864	76	22	.	.	PUNCT
ajst-9864	77	1	3.1.2	3.1.2	X
ajst-9864	77	2	.	.	NOUN
ajst-9864	77	3	gru	gru	PROPN
ajst-9864	77	4	network	network	NOUN
ajst-9864	77	5	model	model	NOUN
ajst-9864	77	6	in	in	ADP
ajst-9864	77	7	natural	natural	ADJ
ajst-9864	77	8	language	language	NOUN
ajst-9864	77	9	processing	processing	NOUN
ajst-9864	77	10	(	(	PUNCT
ajst-9864	77	11	nlp	nlp	NOUN
ajst-9864	77	12	)	)	PUNCT
ajst-9864	77	13	,	,	PUNCT
ajst-9864	77	14	text	text	NOUN
ajst-9864	77	15	data	datum	NOUN
ajst-9864	77	16	is	be	AUX
ajst-9864	77	17	processed	process	VERB
ajst-9864	77	18	as	as	ADP
ajst-9864	77	19	sequence	sequence	NOUN
ajst-9864	77	20	data	datum	NOUN
ajst-9864	77	21	.	.	PUNCT
ajst-9864	78	1	traditional	traditional	ADJ
ajst-9864	78	2	models	model	NOUN
ajst-9864	78	3	based	base	VERB
ajst-9864	78	4	on	on	ADP
ajst-9864	78	5	ngram	ngram	NOUN
ajst-9864	78	6	struggle	struggle	NOUN
ajst-9864	78	7	with	with	ADP
ajst-9864	78	8	long	long	ADJ
ajst-9864	78	9	-	-	PUNCT
ajst-9864	78	10	term	term	NOUN
ajst-9864	78	11	dependencies	dependency	NOUN
ajst-9864	78	12	and	and	CCONJ
ajst-9864	78	13	lack	lack	NOUN
ajst-9864	78	14	flexibility	flexibility	NOUN
ajst-9864	78	15	.	.	PUNCT
ajst-9864	79	1	to	to	PART
ajst-9864	79	2	combat	combat	VERB
ajst-9864	79	3	this	this	PRON
ajst-9864	79	4	,	,	PUNCT
ajst-9864	79	5	recurrent	recurrent	ADJ
ajst-9864	79	6	neural	neural	ADJ
ajst-9864	79	7	networks	network	NOUN
ajst-9864	79	8	(	(	PUNCT
ajst-9864	79	9	rnns	rnns	PROPN
ajst-9864	79	10	)	)	PUNCT
ajst-9864	79	11	were	be	AUX
ajst-9864	79	12	introduced	introduce	VERB
ajst-9864	79	13	,	,	PUNCT
ajst-9864	79	14	offering	offer	VERB
ajst-9864	79	15	memory	memory	NOUN
ajst-9864	79	16	of	of	ADP
ajst-9864	79	17	past	past	ADJ
ajst-9864	79	18	inputs	input	NOUN
ajst-9864	79	19	but	but	CCONJ
ajst-9864	79	20	suffering	suffer	VERB
ajst-9864	79	21	from	from	ADP
ajst-9864	79	22	vanishing	vanish	VERB
ajst-9864	79	23	and	and	CCONJ
ajst-9864	79	24	exploding	explode	VERB
ajst-9864	79	25	gradients	gradient	NOUN
ajst-9864	79	26	during	during	ADP
ajst-9864	79	27	training[10	training[10	PRON
ajst-9864	79	28	]	]	PUNCT
ajst-9864	79	29	..	..	PUNCT
ajst-9864	80	1	the	the	DET
ajst-9864	80	2	long	long	ADJ
ajst-9864	80	3	short	short	ADJ
ajst-9864	80	4	-	-	PUNCT
ajst-9864	80	5	term	term	NOUN
ajst-9864	80	6	memory	memory	NOUN
ajst-9864	80	7	(	(	PUNCT
ajst-9864	80	8	lstm	lstm	NOUN
ajst-9864	80	9	)	)	PUNCT
ajst-9864	80	10	model	model	NOUN
ajst-9864	80	11	was	be	AUX
ajst-9864	80	12	developed	develop	VERB
ajst-9864	80	13	to	to	PART
ajst-9864	80	14	mitigate	mitigate	VERB
ajst-9864	80	15	these	these	DET
ajst-9864	80	16	problems	problem	NOUN
ajst-9864	80	17	by	by	ADP
ajst-9864	80	18	regulating	regulate	VERB
ajst-9864	80	19	hidden	hide	VERB
ajst-9864	80	20	state	state	NOUN
ajst-9864	80	21	updates	update	NOUN
ajst-9864	80	22	using	use	VERB
ajst-9864	80	23	gate	gate	NOUN
ajst-9864	80	24	mechanisms	mechanism	NOUN
ajst-9864	80	25	.	.	PUNCT
ajst-9864	81	1	despite	despite	SCONJ
ajst-9864	81	2	addressing	address	VERB
ajst-9864	81	3	these	these	DET
ajst-9864	81	4	issues	issue	NOUN
ajst-9864	81	5	,	,	PUNCT
ajst-9864	81	6	lstms	lstms	NOUN
ajst-9864	81	7	are	be	AUX
ajst-9864	81	8	computationally	computationally	ADV
ajst-9864	81	9	costly	costly	ADJ
ajst-9864	81	10	and	and	CCONJ
ajst-9864	81	11	prone	prone	ADJ
ajst-9864	81	12	to	to	ADP
ajst-9864	81	13	overfitting	overfitte	VERB
ajst-9864	81	14	due	due	ADP
ajst-9864	81	15	to	to	ADP
ajst-9864	81	16	their	their	PRON
ajst-9864	81	17	complex	complex	ADJ
ajst-9864	81	18	structure	structure	NOUN
ajst-9864	81	19	and	and	CCONJ
ajst-9864	81	20	wide	wide	ADJ
ajst-9864	81	21	parameter	parameter	NOUN
ajst-9864	81	22	space	space	NOUN
ajst-9864	81	23	.	.	PUNCT
ajst-9864	82	1	gated	gate	VERB
ajst-9864	82	2	recurrent	recurrent	ADJ
ajst-9864	82	3	units	unit	NOUN
ajst-9864	82	4	(	(	PUNCT
ajst-9864	82	5	gru	gru	NOUN
ajst-9864	82	6	)	)	PUNCT
ajst-9864	82	7	,	,	PUNCT
ajst-9864	82	8	designed	design	VERB
ajst-9864	82	9	to	to	PART
ajst-9864	82	10	address	address	VERB
ajst-9864	82	11	these	these	DET
ajst-9864	82	12	limitations	limitation	NOUN
ajst-9864	82	13	,	,	PUNCT
ajst-9864	82	14	have	have	VERB
ajst-9864	82	15	only	only	ADV
ajst-9864	82	16	two	two	NUM
ajst-9864	82	17	gates	gate	NOUN
ajst-9864	82	18	and	and	CCONJ
ajst-9864	82	19	no	no	DET
ajst-9864	82	20	cell	cell	NOUN
ajst-9864	82	21	states	state	NOUN
ajst-9864	82	22	,	,	PUNCT
ajst-9864	82	23	making	make	VERB
ajst-9864	82	24	them	they	PRON
ajst-9864	82	25	simpler	simple	ADJ
ajst-9864	82	26	and	and	CCONJ
ajst-9864	82	27	more	more	ADV
ajst-9864	82	28	efficient	efficient	ADJ
ajst-9864	82	29	to	to	PART
ajst-9864	82	30	train	train	VERB
ajst-9864	82	31	.	.	PUNCT
ajst-9864	83	1	they	they	PRON
ajst-9864	83	2	also	also	ADV
ajst-9864	83	3	require	require	VERB
ajst-9864	83	4	less	less	ADJ
ajst-9864	83	5	data	datum	NOUN
ajst-9864	83	6	and	and	CCONJ
ajst-9864	83	7	offer	offer	VERB
ajst-9864	83	8	better	well	ADJ
ajst-9864	83	9	generalisation	generalisation	NOUN
ajst-9864	83	10	,	,	PUNCT
ajst-9864	83	11	as	as	SCONJ
ajst-9864	83	12	they	they	PRON
ajst-9864	83	13	can	can	AUX
ajst-9864	83	14	effectively	effectively	ADV
ajst-9864	83	15	handle	handle	VERB
ajst-9864	83	16	long	long	ADJ
ajst-9864	83	17	-	-	PUNCT
ajst-9864	83	18	term	term	NOUN
ajst-9864	83	19	dependencies	dependency	NOUN
ajst-9864	83	20	.	.	PUNCT
ajst-9864	84	1	the	the	DET
ajst-9864	84	2	bidirectional	bidirectional	PROPN
ajst-9864	84	3	gru	gru	PROPN
ajst-9864	84	4	model	model	NOUN
ajst-9864	84	5	,	,	PUNCT
ajst-9864	84	6	composed	compose	VERB
ajst-9864	84	7	of	of	ADP
ajst-9864	84	8	two	two	NUM
ajst-9864	84	9	independent	independent	ADJ
ajst-9864	84	10	gru	gru	NOUN
ajst-9864	84	11	operating	operate	VERB
ajst-9864	84	12	in	in	ADP
ajst-9864	84	13	opposite	opposite	ADJ
ajst-9864	84	14	directions	direction	NOUN
ajst-9864	84	15	,	,	PUNCT
ajst-9864	84	16	enhances	enhance	VERB
ajst-9864	84	17	the	the	DET
ajst-9864	84	18	model	model	NOUN
ajst-9864	84	19	's	's	PART
ajst-9864	84	20	expressiveness	expressiveness	NOUN
ajst-9864	84	21	by	by	ADP
ajst-9864	84	22	considering	consider	VERB
ajst-9864	84	23	both	both	CCONJ
ajst-9864	84	24	past	past	ADJ
ajst-9864	84	25	and	and	CCONJ
ajst-9864	84	26	future	future	ADJ
ajst-9864	84	27	information	information	NOUN
ajst-9864	84	28	in	in	ADP
ajst-9864	84	29	the	the	DET
ajst-9864	84	30	input	input	NOUN
ajst-9864	84	31	sequence	sequence	NOUN
ajst-9864	84	32	,	,	PUNCT
ajst-9864	84	33	minimising	minimise	VERB
ajst-9864	84	34	the	the	DET
ajst-9864	84	35	prediction	prediction	NOUN
ajst-9864	84	36	gap	gap	NOUN
ajst-9864	84	37	during	during	ADP
ajst-9864	84	38	training	training	NOUN
ajst-9864	84	39	.	.	PUNCT
ajst-9864	85	1	compared	compare	VERB
ajst-9864	85	2	to	to	ADP
ajst-9864	85	3	traditional	traditional	ADJ
ajst-9864	85	4	clm	clm	NOUN
ajst-9864	85	5	models	model	NOUN
ajst-9864	85	6	,	,	PUNCT
ajst-9864	85	7	gru	gru	PROPN
ajst-9864	85	8	has	have	VERB
ajst-9864	85	9	better	well	ADJ
ajst-9864	85	10	long	long	ADJ
ajst-9864	85	11	-	-	PUNCT
ajst-9864	85	12	term	term	NOUN
ajst-9864	85	13	dependencies	dependency	NOUN
ajst-9864	85	14	and	and	CCONJ
ajst-9864	85	15	flexibility	flexibility	NOUN
ajst-9864	85	16	and	and	CCONJ
ajst-9864	85	17	can	can	AUX
ajst-9864	85	18	handle	handle	VERB
ajst-9864	85	19	sequence	sequence	NOUN
ajst-9864	85	20	data	datum	NOUN
ajst-9864	85	21	of	of	ADP
ajst-9864	85	22	any	any	DET
ajst-9864	85	23	length	length	NOUN
ajst-9864	85	24	,	,	PUNCT
ajst-9864	85	25	making	make	VERB
ajst-9864	85	26	it	it	PRON
ajst-9864	85	27	perform	perform	VERB
ajst-9864	85	28	better	well	ADV
ajst-9864	85	29	in	in	ADP
ajst-9864	85	30	nlp	nlp	ADJ
ajst-9864	85	31	tasks	task	NOUN
ajst-9864	85	32	.	.	PUNCT
ajst-9864	86	1	compared	compare	VERB
ajst-9864	86	2	to	to	ADP
ajst-9864	86	3	lstm	lstm	PROPN
ajst-9864	86	4	,	,	PUNCT
ajst-9864	86	5	gru	gru	PROPN
ajst-9864	86	6	is	be	AUX
ajst-9864	86	7	simpler	simple	ADJ
ajst-9864	86	8	and	and	CCONJ
ajst-9864	86	9	more	more	ADV
ajst-9864	86	10	computationally	computationally	ADV
ajst-9864	86	11	efficient	efficient	ADJ
ajst-9864	86	12	,	,	PUNCT
ajst-9864	86	13	requiring	require	VERB
ajst-9864	86	14	less	less	ADJ
ajst-9864	86	15	data	datum	NOUN
ajst-9864	86	16	for	for	ADP
ajst-9864	86	17	training	training	NOUN
ajst-9864	86	18	,	,	PUNCT
ajst-9864	86	19	making	make	VERB
ajst-9864	86	20	it	it	PRON
ajst-9864	86	21	more	more	ADV
ajst-9864	86	22	advantageous	advantageous	ADJ
ajst-9864	86	23	for	for	ADP
ajst-9864	86	24	handling	handle	VERB
ajst-9864	86	25	small	small	ADJ
ajst-9864	86	26	datasets	dataset	NOUN
ajst-9864	86	27	.	.	PUNCT
ajst-9864	87	1	table	table	NOUN
ajst-9864	87	2	1	1	NUM
ajst-9864	87	3	compares	compare	VERB
ajst-9864	87	4	clms	clm	NOUN
ajst-9864	87	5	,	,	PUNCT
ajst-9864	87	6	lstm	lstm	ADJ
ajst-9864	87	7	,	,	PUNCT
ajst-9864	87	8	and	and	CCONJ
ajst-9864	87	9	gru	gru	NOUN
ajst-9864	87	10	models	model	NOUN
ajst-9864	87	11	.	.	PUNCT
ajst-9864	88	1	model	model	NOUN
ajst-9864	88	2	flexibility	flexibility	NOUN
ajst-9864	88	3	computational	computational	ADJ
ajst-9864	88	4	cost	cost	NOUN
ajst-9864	88	5	effectiveness	effectiveness	NOUN
ajst-9864	88	6	in	in	ADP
ajst-9864	88	7	processing	process	VERB
ajst-9864	88	8	chinese	chinese	ADJ
ajst-9864	88	9	text	text	NOUN
ajst-9864	88	10	number	number	NOUN
ajst-9864	88	11	of	of	ADP
ajst-9864	88	12	parameters	parameter	NOUN
ajst-9864	88	13	computational	computational	ADJ
ajst-9864	88	14	efficiency	efficiency	NOUN
ajst-9864	88	15	clms	clm	NOUN
ajst-9864	88	16	low	low	ADJ
ajst-9864	88	17	high	high	ADJ
ajst-9864	88	18	low	low	ADJ
ajst-9864	88	19	few	few	ADJ
ajst-9864	88	20	high	high	ADJ
ajst-9864	88	21	lstm	lstm	ADJ
ajst-9864	88	22	high	high	ADJ
ajst-9864	88	23	high	high	ADJ
ajst-9864	88	24	moderate	moderate	ADJ
ajst-9864	88	25	to	to	PART
ajst-9864	88	26	high	high	VERB
ajst-9864	88	27	many	many	ADJ
ajst-9864	88	28	medium	medium	PROPN
ajst-9864	88	29	gru	gru	NOUN
ajst-9864	88	30	medium	medium	ADJ
ajst-9864	88	31	low	low	ADJ
ajst-9864	88	32	high	high	ADJ
ajst-9864	88	33	few	few	ADJ
ajst-9864	88	34	high	high	ADJ
ajst-9864	88	35	131	131	NUM
ajst-9864	88	36	the	the	DET
ajst-9864	88	37	above	above	ADJ
ajst-9864	88	38	table	table	NOUN
ajst-9864	89	1	1	1	NUM
ajst-9864	89	2	compares	compare	VERB
ajst-9864	89	3	clms	clm	NOUN
ajst-9864	89	4	,	,	PUNCT
ajst-9864	89	5	lstm	lstm	ADJ
ajst-9864	89	6	,	,	PUNCT
ajst-9864	89	7	and	and	CCONJ
ajst-9864	89	8	gru	gru	NOUN
ajst-9864	89	9	models	model	NOUN
ajst-9864	89	10	across	across	ADP
ajst-9864	89	11	different	different	ADJ
ajst-9864	89	12	dimensions	dimension	NOUN
ajst-9864	89	13	.	.	PUNCT
ajst-9864	90	1	flexibility	flexibility	NOUN
ajst-9864	90	2	:	:	PUNCT
ajst-9864	90	3	flexibility	flexibility	NOUN
ajst-9864	90	4	refers	refer	VERB
ajst-9864	90	5	to	to	ADP
ajst-9864	90	6	the	the	DET
ajst-9864	90	7	model	model	NOUN
ajst-9864	90	8	's	's	PART
ajst-9864	90	9	adaptability	adaptability	NOUN
ajst-9864	90	10	in	in	ADP
ajst-9864	90	11	handling	handle	VERB
ajst-9864	90	12	different	different	ADJ
ajst-9864	90	13	types	type	NOUN
ajst-9864	90	14	of	of	ADP
ajst-9864	90	15	tasks	task	NOUN
ajst-9864	90	16	.	.	PUNCT
ajst-9864	91	1	clms	clm	NOUN
ajst-9864	91	2	has	have	VERB
ajst-9864	91	3	low	low	ADJ
ajst-9864	91	4	flexibility	flexibility	NOUN
ajst-9864	91	5	and	and	CCONJ
ajst-9864	91	6	is	be	AUX
ajst-9864	91	7	mainly	mainly	ADV
ajst-9864	91	8	used	use	VERB
ajst-9864	91	9	for	for	ADP
ajst-9864	91	10	speech	speech	NOUN
ajst-9864	91	11	recognition	recognition	NOUN
ajst-9864	91	12	tasks	task	NOUN
ajst-9864	91	13	.	.	PUNCT
ajst-9864	92	1	lstm	lstm	PROPN
ajst-9864	92	2	and	and	CCONJ
ajst-9864	92	3	gru	gru	PROPN
ajst-9864	92	4	have	have	VERB
ajst-9864	92	5	higher	high	ADJ
ajst-9864	92	6	flexibility	flexibility	NOUN
ajst-9864	92	7	and	and	CCONJ
ajst-9864	92	8	can	can	AUX
ajst-9864	92	9	be	be	AUX
ajst-9864	92	10	used	use	VERB
ajst-9864	92	11	for	for	ADP
ajst-9864	92	12	multiple	multiple	ADJ
ajst-9864	92	13	tasks	task	NOUN
ajst-9864	92	14	such	such	ADJ
ajst-9864	92	15	as	as	ADP
ajst-9864	92	16	text	text	NOUN
ajst-9864	92	17	classification	classification	NOUN
ajst-9864	92	18	,	,	PUNCT
ajst-9864	92	19	machine	machine	NOUN
ajst-9864	92	20	translation	translation	NOUN
ajst-9864	92	21	,	,	PUNCT
ajst-9864	92	22	speech	speech	NOUN
ajst-9864	92	23	recognition	recognition	NOUN
ajst-9864	92	24	,	,	PUNCT
ajst-9864	92	25	etc	etc	X
ajst-9864	92	26	.	.	X
ajst-9864	92	27	computational	computational	ADJ
ajst-9864	92	28	cost	cost	NOUN
ajst-9864	92	29	:	:	PUNCT
ajst-9864	92	30	the	the	DET
ajst-9864	92	31	amount	amount	NOUN
ajst-9864	92	32	of	of	ADP
ajst-9864	92	33	computing	compute	VERB
ajst-9864	92	34	resources	resource	NOUN
ajst-9864	92	35	required	require	VERB
ajst-9864	92	36	during	during	ADP
ajst-9864	92	37	training	training	NOUN
ajst-9864	92	38	and	and	CCONJ
ajst-9864	92	39	inference	inference	NOUN
ajst-9864	92	40	is	be	AUX
ajst-9864	92	41	referred	refer	VERB
ajst-9864	92	42	to	to	ADP
ajst-9864	92	43	as	as	SCONJ
ajst-9864	92	44	the	the	DET
ajst-9864	92	45	computational	computational	ADJ
ajst-9864	92	46	cost.clms	cost.clm	NOUN
ajst-9864	92	47	has	have	VERB
ajst-9864	92	48	a	a	DET
ajst-9864	92	49	high	high	ADJ
ajst-9864	92	50	computational	computational	ADJ
ajst-9864	92	51	cost	cost	NOUN
ajst-9864	92	52	and	and	CCONJ
ajst-9864	92	53	requires	require	VERB
ajst-9864	92	54	a	a	DET
ajst-9864	92	55	lot	lot	NOUN
ajst-9864	92	56	of	of	ADP
ajst-9864	92	57	computing	compute	VERB
ajst-9864	92	58	resources	resource	NOUN
ajst-9864	92	59	for	for	ADP
ajst-9864	92	60	training	training	NOUN
ajst-9864	92	61	and	and	CCONJ
ajst-9864	92	62	inference	inference	NOUN
ajst-9864	92	63	.	.	PUNCT
ajst-9864	93	1	lstm	lstm	PROPN
ajst-9864	93	2	and	and	CCONJ
ajst-9864	93	3	gru	gru	PROPN
ajst-9864	93	4	have	have	VERB
ajst-9864	93	5	relatively	relatively	ADV
ajst-9864	93	6	high	high	ADJ
ajst-9864	93	7	computational	computational	ADJ
ajst-9864	93	8	costs	cost	NOUN
ajst-9864	93	9	,	,	PUNCT
ajst-9864	93	10	but	but	CCONJ
ajst-9864	93	11	are	be	AUX
ajst-9864	93	12	lower	low	ADJ
ajst-9864	93	13	than	than	ADP
ajst-9864	93	14	clms	clm	NOUN
ajst-9864	93	15	.	.	PUNCT
ajst-9864	94	1	effectiveness	effectiveness	NOUN
ajst-9864	94	2	in	in	ADP
ajst-9864	94	3	processing	process	VERB
ajst-9864	94	4	chinese	chinese	ADJ
ajst-9864	94	5	text	text	NOUN
ajst-9864	94	6	:	:	PUNCT
ajst-9864	94	7	this	this	PRON
ajst-9864	94	8	refers	refer	VERB
ajst-9864	94	9	to	to	ADP
ajst-9864	94	10	the	the	DET
ajst-9864	94	11	model	model	NOUN
ajst-9864	94	12	's	's	PART
ajst-9864	94	13	performance	performance	NOUN
ajst-9864	94	14	in	in	ADP
ajst-9864	94	15	processing	process	VERB
ajst-9864	94	16	chinese	chinese	ADJ
ajst-9864	94	17	text	text	NOUN
ajst-9864	94	18	.	.	PUNCT
ajst-9864	95	1	clms	clm	NOUN
ajst-9864	95	2	has	have	VERB
ajst-9864	95	3	low	low	ADJ
ajst-9864	95	4	effectiveness	effectiveness	NOUN
ajst-9864	95	5	in	in	ADP
ajst-9864	95	6	processing	process	VERB
ajst-9864	95	7	chinese	chinese	ADJ
ajst-9864	95	8	text	text	NOUN
ajst-9864	95	9	because	because	SCONJ
ajst-9864	95	10	it	it	PRON
ajst-9864	95	11	is	be	AUX
ajst-9864	95	12	not	not	PART
ajst-9864	95	13	good	good	ADJ
ajst-9864	95	14	at	at	ADP
ajst-9864	95	15	handling	handle	VERB
ajst-9864	95	16	the	the	DET
ajst-9864	95	17	prosody	prosody	NOUN
ajst-9864	95	18	and	and	CCONJ
ajst-9864	95	19	tone	tone	NOUN
ajst-9864	95	20	of	of	ADP
ajst-9864	95	21	chinese	chinese	ADJ
ajst-9864	95	22	text	text	NOUN
ajst-9864	95	23	.	.	PUNCT
ajst-9864	96	1	lstm	lstm	PROPN
ajst-9864	96	2	and	and	CCONJ
ajst-9864	96	3	gru	gru	PROPN
ajst-9864	96	4	have	have	VERB
ajst-9864	96	5	good	good	ADJ
ajst-9864	96	6	effectiveness	effectiveness	NOUN
ajst-9864	96	7	in	in	ADP
ajst-9864	96	8	processing	process	VERB
ajst-9864	96	9	chinese	chinese	ADJ
ajst-9864	96	10	text	text	NOUN
ajst-9864	96	11	and	and	CCONJ
ajst-9864	96	12	can	can	AUX
ajst-9864	96	13	handle	handle	VERB
ajst-9864	96	14	the	the	DET
ajst-9864	96	15	semantics	semantic	NOUN
ajst-9864	96	16	and	and	CCONJ
ajst-9864	96	17	syntax	syntax	NOUN
ajst-9864	96	18	of	of	ADP
ajst-9864	96	19	chinese	chinese	ADJ
ajst-9864	96	20	text	text	NOUN
ajst-9864	96	21	well	well	ADV
ajst-9864	96	22	.	.	PUNCT
ajst-9864	97	1	number	number	NOUN
ajst-9864	97	2	of	of	ADP
ajst-9864	97	3	parameters	parameter	NOUN
ajst-9864	97	4	:	:	PUNCT
ajst-9864	97	5	the	the	DET
ajst-9864	97	6	"	"	PUNCT
ajst-9864	97	7	number	number	NOUN
ajst-9864	97	8	of	of	ADP
ajst-9864	97	9	parameters	parameter	NOUN
ajst-9864	97	10	"	"	PUNCT
ajst-9864	97	11	refers	refer	VERB
ajst-9864	97	12	to	to	ADP
ajst-9864	97	13	the	the	DET
ajst-9864	97	14	number	number	NOUN
ajst-9864	97	15	of	of	ADP
ajst-9864	97	16	trainable	trainable	ADJ
ajst-9864	97	17	parameters	parameter	NOUN
ajst-9864	97	18	in	in	ADP
ajst-9864	97	19	the	the	DET
ajst-9864	97	20	model	model	NOUN
ajst-9864	97	21	.	.	PUNCT
ajst-9864	98	1	clms	clm	NOUN
ajst-9864	98	2	has	have	VERB
ajst-9864	98	3	fewer	few	ADJ
ajst-9864	98	4	parameters	parameter	NOUN
ajst-9864	98	5	,	,	PUNCT
ajst-9864	98	6	lstm	lstm	NOUN
ajst-9864	98	7	has	have	VERB
ajst-9864	98	8	more	more	ADJ
ajst-9864	98	9	parameters	parameter	NOUN
ajst-9864	98	10	,	,	PUNCT
ajst-9864	98	11	and	and	CCONJ
ajst-9864	98	12	gru	gru	NOUN
ajst-9864	98	13	has	have	VERB
ajst-9864	98	14	relatively	relatively	ADV
ajst-9864	98	15	fewer	few	ADJ
ajst-9864	98	16	parameters	parameter	NOUN
ajst-9864	98	17	.	.	PUNCT
ajst-9864	99	1	computational	computational	ADJ
ajst-9864	99	2	efficiency	efficiency	NOUN
ajst-9864	99	3	:	:	PUNCT
ajst-9864	99	4	computational	computational	ADJ
ajst-9864	99	5	efficiency	efficiency	NOUN
ajst-9864	99	6	refers	refer	VERB
ajst-9864	99	7	to	to	ADP
ajst-9864	99	8	the	the	DET
ajst-9864	99	9	speed	speed	NOUN
ajst-9864	99	10	of	of	ADP
ajst-9864	99	11	the	the	DET
ajst-9864	99	12	model	model	NOUN
ajst-9864	99	13	's	's	PART
ajst-9864	99	14	computation	computation	NOUN
ajst-9864	99	15	during	during	ADP
ajst-9864	99	16	inference	inference	NOUN
ajst-9864	99	17	.	.	PUNCT
ajst-9864	100	1	clms	clm	NOUN
ajst-9864	100	2	has	have	VERB
ajst-9864	100	3	high	high	ADJ
ajst-9864	100	4	computational	computational	ADJ
ajst-9864	100	5	efficiency	efficiency	NOUN
ajst-9864	100	6	,	,	PUNCT
ajst-9864	100	7	lstm	lstm	NOUN
ajst-9864	100	8	has	have	VERB
ajst-9864	100	9	medium	medium	ADJ
ajst-9864	100	10	computational	computational	ADJ
ajst-9864	100	11	efficiency	efficiency	NOUN
ajst-9864	100	12	,	,	PUNCT
ajst-9864	100	13	and	and	CCONJ
ajst-9864	100	14	gru	gru	PROPN
ajst-9864	100	15	has	have	VERB
ajst-9864	100	16	good	good	ADJ
ajst-9864	100	17	computational	computational	ADJ
ajst-9864	100	18	efficiency	efficiency	NOUN
ajst-9864	100	19	.	.	PUNCT
ajst-9864	101	1	this	this	PRON
ajst-9864	101	2	is	be	AUX
ajst-9864	101	3	because	because	SCONJ
ajst-9864	101	4	the	the	DET
ajst-9864	101	5	gating	gate	VERB
ajst-9864	101	6	structure	structure	NOUN
ajst-9864	101	7	in	in	ADP
ajst-9864	101	8	gru	gru	NOUN
ajst-9864	101	9	is	be	AUX
ajst-9864	101	10	simpler	simple	ADJ
ajst-9864	101	11	than	than	ADP
ajst-9864	101	12	lstm	lstm	ADJ
ajst-9864	101	13	,	,	PUNCT
ajst-9864	101	14	with	with	ADP
ajst-9864	101	15	lower	low	ADJ
ajst-9864	101	16	computational	computational	ADJ
ajst-9864	101	17	complexity	complexity	NOUN
ajst-9864	101	18	,	,	PUNCT
ajst-9864	101	19	thereby	thereby	ADV
ajst-9864	101	20	improving	improve	VERB
ajst-9864	101	21	computational	computational	ADJ
ajst-9864	101	22	efficiency	efficiency	NOUN
ajst-9864	101	23	.	.	PUNCT
ajst-9864	102	1	figure	figure	NOUN
ajst-9864	102	2	2	2	NUM
ajst-9864	102	3	.	.	NOUN
ajst-9864	102	4	gru	gru	NOUN
ajst-9864	102	5	model	model	PROPN
ajst-9864	102	6	workflow	workflow	NOUN
ajst-9864	102	7	the	the	DET
ajst-9864	102	8	workflow	workflow	NOUN
ajst-9864	102	9	of	of	ADP
ajst-9864	102	10	the	the	DET
ajst-9864	102	11	gru	gru	NOUN
ajst-9864	102	12	network	network	NOUN
ajst-9864	102	13	is	be	AUX
ajst-9864	102	14	as	as	SCONJ
ajst-9864	102	15	follows	follow	VERB
ajst-9864	102	16	:	:	PUNCT
ajst-9864	102	17	1	1	X
ajst-9864	102	18	)	)	PUNCT
ajst-9864	102	19	calculate	calculate	VERB
ajst-9864	102	20	the	the	DET
ajst-9864	102	21	reset	reset	NOUN
ajst-9864	102	22	gate	gate	NOUN
ajst-9864	102	23	rt	rt	PROPN
ajst-9864	102	24	and	and	CCONJ
ajst-9864	102	25	the	the	DET
ajst-9864	102	26	candidate	candidate	NOUN
ajst-9864	102	27	state	state	NOUN
ajst-9864	102	28	reset	reset	NOUN
ajst-9864	102	29	gate	gate	NOUN
ajst-9864	102	30	𝐡𝐭	𝐡𝐭	VERB
ajst-9864	102	31	.	.	PUNCT
ajst-9864	103	1	used	use	VERB
ajst-9864	103	2	rt	rt	PROPN
ajst-9864	103	3	to	to	PART
ajst-9864	103	4	control	control	VERB
ajst-9864	103	5	how	how	SCONJ
ajst-9864	103	6	much	much	ADJ
ajst-9864	103	7	information	information	NOUN
ajst-9864	103	8	in	in	ADP
ajst-9864	103	9	the	the	DET
ajst-9864	103	10	candidate	candidate	NOUN
ajst-9864	103	11	state	state	NOUN
ajst-9864	103	12	ht	ht	PROPN
ajst-9864	103	13	comes	come	VERB
ajst-9864	103	14	from	from	ADP
ajst-9864	103	15	the	the	DET
ajst-9864	103	16	state	state	NOUN
ajst-9864	103	17	ht-1	ht-1	VERB
ajst-9864	103	18	at	at	ADP
ajst-9864	103	19	the	the	DET
ajst-9864	103	20	previous	previous	ADJ
ajst-9864	103	21	moment	moment	NOUN
ajst-9864	103	22	,	,	PUNCT
ajst-9864	103	23	mathematically	mathematically	ADV
ajst-9864	103	24	represented	represent	VERB
ajst-9864	103	25	as	as	ADP
ajst-9864	103	26	:	:	PUNCT
ajst-9864	103	27	rt	rt	PROPN
ajst-9864	103	28	=	=	PROPN
ajst-9864	103	29	σ	σ	PROPN
ajst-9864	103	30	(	(	PUNCT
ajst-9864	103	31	wrxt	wrxt	NOUN
ajst-9864	103	32	+	+	CCONJ
ajst-9864	103	33	urht	urht	ADJ
ajst-9864	103	34	－	－	ADJ
ajst-9864	103	35	1	1	NUM
ajst-9864	103	36	+	+	NUM
ajst-9864	103	37	br	br	NOUN
ajst-9864	103	38	)	)	PUNCT
ajst-9864	103	39	(	(	PUNCT
ajst-9864	103	40	2	2	X
ajst-9864	103	41	)	)	PUNCT
ajst-9864	103	42	where	where	SCONJ
ajst-9864	103	43	σ	σ	PROPN
ajst-9864	103	44	denotes	denote	VERB
ajst-9864	103	45	the	the	DET
ajst-9864	103	46	activation	activation	NOUN
ajst-9864	103	47	function	function	NOUN
ajst-9864	103	48	,	,	PUNCT
ajst-9864	103	49	wr	wr	PROPN
ajst-9864	103	50	,	,	PUNCT
ajst-9864	103	51	ur	ur	INTJ
ajst-9864	103	52	,	,	PUNCT
ajst-9864	103	53	and	and	CCONJ
ajst-9864	103	54	br	br	NOUN
ajst-9864	103	55	are	be	AUX
ajst-9864	103	56	the	the	DET
ajst-9864	103	57	parameters	parameter	NOUN
ajst-9864	103	58	of	of	ADP
ajst-9864	103	59	the	the	DET
ajst-9864	103	60	reset	reset	NOUN
ajst-9864	103	61	gate	gate	NOUN
ajst-9864	103	62	,	,	PUNCT
ajst-9864	103	63	xt	xt	PROPN
ajst-9864	103	64	is	be	AUX
ajst-9864	103	65	the	the	DET
ajst-9864	103	66	input	input	NOUN
ajst-9864	103	67	at	at	ADP
ajst-9864	103	68	the	the	DET
ajst-9864	103	69	t	t	NOUN
ajst-9864	103	70	level	level	NOUN
ajst-9864	103	71	,	,	PUNCT
ajst-9864	103	72	and	and	CCONJ
ajst-9864	103	73	ht-1	ht-1	NUM
ajst-9864	103	74	is	be	AUX
ajst-9864	103	75	the	the	DET
ajst-9864	103	76	state	state	NOUN
ajst-9864	103	77	at	at	ADP
ajst-9864	103	78	the	the	DET
ajst-9864	103	79	previous	previous	ADJ
ajst-9864	103	80	moment	moment	NOUN
ajst-9864	103	81	.	.	PUNCT
ajst-9864	104	1	the	the	DET
ajst-9864	104	2	candidate	candidate	NOUN
ajst-9864	104	3	states	state	VERB
ajst-9864	104	4	ht	ht	X
ajst-9864	104	5	at	at	ADP
ajst-9864	104	6	the	the	DET
ajst-9864	104	7	current	current	ADJ
ajst-9864	104	8	moment	moment	NOUN
ajst-9864	104	9	are	be	AUX
ajst-9864	104	10	mathematically	mathematically	ADV
ajst-9864	104	11	represented	represent	VERB
ajst-9864	104	12	as	as	ADP
ajst-9864	104	13	:	:	PUNCT
ajst-9864	104	14	ht	ht	PROPN
ajst-9864	104	15	=	=	SYM
ajst-9864	104	16	tanh	tanh	PROPN
ajst-9864	104	17	(	(	PUNCT
ajst-9864	104	18	wh	wh	NOUN
ajst-9864	104	19	xt	xt	PROPN
ajst-9864	105	1	+	+	CCONJ
ajst-9864	105	2	uh	uh	INTJ
ajst-9864	105	3	(	(	PUNCT
ajst-9864	105	4	rt⊙ht	rt⊙ht	X
ajst-9864	105	5	－	－	PROPN
ajst-9864	105	6	1	1	NUM
ajst-9864	105	7	+	+	NUM
ajst-9864	105	8	br	br	NOUN
ajst-9864	105	9	)	)	PUNCT
ajst-9864	105	10	)	)	PUNCT
ajst-9864	106	1	(	(	PUNCT
ajst-9864	106	2	3	3	X
ajst-9864	106	3	)	)	PUNCT
ajst-9864	106	4	where	where	SCONJ
ajst-9864	106	5	⊙	⊙	PROPN
ajst-9864	106	6	denotes	denote	VERB
ajst-9864	106	7	the	the	DET
ajst-9864	106	8	matrix	matrix	NOUN
ajst-9864	106	9	element	element	NOUN
ajst-9864	106	10	multiplication	multiplication	NOUN
ajst-9864	106	11	,	,	PUNCT
ajst-9864	106	12	wh	wh	PROPN
ajst-9864	106	13	,	,	PUNCT
ajst-9864	106	14	uh	uh	INTJ
ajst-9864	106	15	are	be	AUX
ajst-9864	106	16	the	the	DET
ajst-9864	106	17	parameters	parameter	NOUN
ajst-9864	106	18	of	of	ADP
ajst-9864	106	19	the	the	DET
ajst-9864	106	20	candidate	candidate	NOUN
ajst-9864	106	21	states	state	NOUN
ajst-9864	106	22	,	,	PUNCT
ajst-9864	106	23	and	and	CCONJ
ajst-9864	106	24	others	other	NOUN
ajst-9864	106	25	as	as	ADP
ajst-9864	106	26	above	above	ADV
ajst-9864	106	27	.	.	PUNCT
ajst-9864	107	1	2	2	X
ajst-9864	107	2	)	)	PUNCT
ajst-9864	107	3	compute	compute	VERB
ajst-9864	107	4	the	the	DET
ajst-9864	107	5	update	update	NOUN
ajst-9864	107	6	gate	gate	NOUN
ajst-9864	107	7	zt	zt	PROPN
ajst-9864	107	8	and	and	CCONJ
ajst-9864	107	9	the	the	DET
ajst-9864	107	10	current	current	ADJ
ajst-9864	107	11	state	state	NOUN
ajst-9864	107	12	ht	ht	PROPN
ajst-9864	107	13	.	.	PUNCT
ajst-9864	108	1	the	the	DET
ajst-9864	108	2	update	update	NOUN
ajst-9864	108	3	gate	gate	NOUN
ajst-9864	108	4	zt	zt	PROPN
ajst-9864	108	5	is	be	AUX
ajst-9864	108	6	used	use	VERB
ajst-9864	108	7	to	to	PART
ajst-9864	108	8	control	control	VERB
ajst-9864	108	9	how	how	SCONJ
ajst-9864	108	10	much	much	ADJ
ajst-9864	108	11	information	information	NOUN
ajst-9864	108	12	from	from	ADP
ajst-9864	108	13	the	the	DET
ajst-9864	108	14	historical	historical	ADJ
ajst-9864	108	15	state	state	NOUN
ajst-9864	108	16	ht	ht	PROPN
ajst-9864	108	17	1	1	NUM
ajst-9864	108	18	is	be	AUX
ajst-9864	108	19	retained	retain	VERB
ajst-9864	108	20	in	in	ADP
ajst-9864	108	21	the	the	DET
ajst-9864	108	22	current	current	ADJ
ajst-9864	108	23	state	state	NOUN
ajst-9864	108	24	ht	ht	PROPN
ajst-9864	108	25	and	and	CCONJ
ajst-9864	108	26	how	how	SCONJ
ajst-9864	108	27	much	much	ADJ
ajst-9864	108	28	information	information	NOUN
ajst-9864	108	29	from	from	ADP
ajst-9864	108	30	the	the	DET
ajst-9864	108	31	candidate	candidate	NOUN
ajst-9864	108	32	state	state	NOUN
ajst-9864	108	33	ht	ht	PROPN
ajst-9864	108	34	is	be	AUX
ajst-9864	108	35	received	receive	VERB
ajst-9864	108	36	in	in	ADP
ajst-9864	108	37	the	the	DET
ajst-9864	108	38	current	current	ADJ
ajst-9864	108	39	state	state	NOUN
ajst-9864	108	40	ht	ht	PROPN
ajst-9864	108	41	,	,	PUNCT
ajst-9864	108	42	mathematically	mathematically	ADV
ajst-9864	108	43	expressed	express	VERB
ajst-9864	108	44	as	as	ADP
ajst-9864	108	45	:	:	PUNCT
ajst-9864	108	46	zt	zt	PROPN
ajst-9864	108	47	=	=	PROPN
ajst-9864	108	48	σ	σ	PROPN
ajst-9864	108	49	(	(	PUNCT
ajst-9864	108	50	wzxt	wzxt	PROPN
ajst-9864	108	51	+	+	CCONJ
ajst-9864	108	52	uzht	uzht	PROPN
ajst-9864	108	53	－	－	PROPN
ajst-9864	108	54	1	1	NUM
ajst-9864	108	55	+	+	CCONJ
ajst-9864	108	56	bz	bz	X
ajst-9864	108	57	)	)	PUNCT
ajst-9864	108	58	(	(	PUNCT
ajst-9864	108	59	4	4	X
ajst-9864	108	60	)	)	PUNCT
ajst-9864	108	61	where	where	SCONJ
ajst-9864	108	62	σ	σ	PROPN
ajst-9864	108	63	denotes	denote	VERB
ajst-9864	108	64	the	the	DET
ajst-9864	108	65	activation	activation	NOUN
ajst-9864	108	66	function	function	NOUN
ajst-9864	108	67	,	,	PUNCT
ajst-9864	108	68	wz	wz	PROPN
ajst-9864	108	69	,	,	PUNCT
ajst-9864	108	70	uz	uz	PROPN
ajst-9864	108	71	,	,	PUNCT
ajst-9864	108	72	bz	bz	PROPN
ajst-9864	108	73	are	be	AUX
ajst-9864	108	74	the	the	DET
ajst-9864	108	75	parameters	parameter	NOUN
ajst-9864	108	76	of	of	ADP
ajst-9864	108	77	the	the	DET
ajst-9864	108	78	update	update	NOUN
ajst-9864	108	79	gate	gate	NOUN
ajst-9864	108	80	,	,	PUNCT
ajst-9864	108	81	xt	xt	PROPN
ajst-9864	108	82	is	be	AUX
ajst-9864	108	83	the	the	DET
ajst-9864	108	84	input	input	NOUN
ajst-9864	108	85	at	at	ADP
ajst-9864	108	86	layer	layer	NOUN
ajst-9864	108	87	t	t	PROPN
ajst-9864	108	88	,	,	PUNCT
ajst-9864	108	89	and	and	CCONJ
ajst-9864	108	90	ht1	ht1	X
ajst-9864	108	91	is	be	AUX
ajst-9864	108	92	the	the	DET
ajst-9864	108	93	state	state	NOUN
ajst-9864	108	94	at	at	ADP
ajst-9864	108	95	the	the	DET
ajst-9864	108	96	previous	previous	ADJ
ajst-9864	108	97	moment	moment	NOUN
ajst-9864	108	98	.	.	PUNCT
ajst-9864	109	1	the	the	DET
ajst-9864	109	2	current	current	ADJ
ajst-9864	109	3	moment	moment	NOUN
ajst-9864	109	4	hidden	hide	VERB
ajst-9864	109	5	state	state	NOUN
ajst-9864	109	6	,	,	PUNCT
ajst-9864	109	7	ht	ht	PROPN
ajst-9864	109	8	,	,	PUNCT
ajst-9864	109	9	is	be	AUX
ajst-9864	109	10	mathematically	mathematically	ADV
ajst-9864	109	11	represented	represent	VERB
ajst-9864	109	12	as	as	ADP
ajst-9864	109	13	ht	ht	PROPN
ajst-9864	109	14	=	=	SYM
ajst-9864	109	15	zt⊙ht	zt⊙ht	X
ajst-9864	109	16	－	－	NOUN
ajst-9864	109	17	1	1	NUM
ajst-9864	110	1	+	+	CCONJ
ajst-9864	110	2	(	(	PUNCT
ajst-9864	110	3	1	1	NUM
ajst-9864	110	4	－	－	PROPN
ajst-9864	110	5	zt	zt	PROPN
ajst-9864	110	6	)	)	PUNCT
ajst-9864	110	7	⊙ht	⊙ht	NOUN
ajst-9864	110	8	.	.	PUNCT
ajst-9864	111	1	(	(	PUNCT
ajst-9864	111	2	5	5	NUM
ajst-9864	111	3	)	)	PUNCT
ajst-9864	111	4	where	where	SCONJ
ajst-9864	111	5	zt	zt	PROPN
ajst-9864	111	6	is	be	AUX
ajst-9864	111	7	the	the	DET
ajst-9864	111	8	information	information	NOUN
ajst-9864	111	9	captured	capture	VERB
ajst-9864	111	10	by	by	ADP
ajst-9864	111	11	the	the	DET
ajst-9864	111	12	update	update	NOUN
ajst-9864	111	13	gate	gate	NOUN
ajst-9864	111	14	,	,	PUNCT
ajst-9864	111	15	ht1	ht1	X
ajst-9864	111	16	is	be	AUX
ajst-9864	111	17	the	the	DET
ajst-9864	111	18	state	state	NOUN
ajst-9864	111	19	at	at	ADP
ajst-9864	111	20	the	the	DET
ajst-9864	111	21	previous	previous	ADJ
ajst-9864	111	22	moment	moment	NOUN
ajst-9864	111	23	,	,	PUNCT
ajst-9864	111	24	⊙	⊙	PROPN
ajst-9864	111	25	denotes	denote	VERB
ajst-9864	111	26	the	the	DET
ajst-9864	111	27	matrix	matrix	NOUN
ajst-9864	111	28	element	element	NOUN
ajst-9864	111	29	multiplication	multiplication	NOUN
ajst-9864	111	30	,	,	PUNCT
ajst-9864	111	31	and	and	CCONJ
ajst-9864	111	32	ht	ht	PROPN
ajst-9864	111	33	is	be	AUX
ajst-9864	111	34	the	the	DET
ajst-9864	111	35	current	current	ADJ
ajst-9864	111	36	candidate	candidate	NOUN
ajst-9864	111	37	state	state	NOUN
ajst-9864	111	38	.	.	PUNCT
ajst-9864	112	1	3.1.3	3.1.3	NUM
ajst-9864	112	2	.	.	PUNCT
ajst-9864	112	3	attention	attention	NOUN
ajst-9864	112	4	mechanism	mechanism	NOUN
ajst-9864	112	5	the	the	DET
ajst-9864	112	6	attention	attention	NOUN
ajst-9864	112	7	mechanism	mechanism	NOUN
ajst-9864	112	8	,	,	PUNCT
ajst-9864	112	9	inspired	inspire	VERB
ajst-9864	112	10	by	by	ADP
ajst-9864	112	11	human	human	ADJ
ajst-9864	112	12	visual	visual	ADJ
ajst-9864	112	13	attention	attention	NOUN
ajst-9864	112	14	,	,	PUNCT
ajst-9864	112	15	emphasizes	emphasize	VERB
ajst-9864	112	16	relevant	relevant	ADJ
ajst-9864	112	17	information	information	NOUN
ajst-9864	112	18	for	for	ADP
ajst-9864	112	19	tasks	task	NOUN
ajst-9864	112	20	at	at	ADP
ajst-9864	112	21	hand	hand	NOUN
ajst-9864	112	22	.	.	PUNCT
ajst-9864	113	1	initially	initially	ADV
ajst-9864	113	2	applied	apply	VERB
ajst-9864	113	3	to	to	ADP
ajst-9864	113	4	text	text	NOUN
ajst-9864	113	5	representation	representation	NOUN
ajst-9864	113	6	by	by	ADP
ajst-9864	113	7	google	google	PROPN
ajst-9864	113	8	's	's	PART
ajst-9864	113	9	machine	machine	NOUN
ajst-9864	113	10	translation	translation	NOUN
ajst-9864	113	11	team	team	NOUN
ajst-9864	113	12	,	,	PUNCT
ajst-9864	113	13	self	self	NOUN
ajst-9864	113	14	-	-	PUNCT
ajst-9864	113	15	attention	attention	NOUN
ajst-9864	113	16	identified	identify	VERB
ajst-9864	113	17	syntactic	syntactic	ADJ
ajst-9864	113	18	or	or	CCONJ
ajst-9864	113	19	semantic	semantic	ADJ
ajst-9864	113	20	differences	difference	NOUN
ajst-9864	113	21	in	in	ADP
ajst-9864	113	22	phrases	phrase	NOUN
ajst-9864	113	23	,	,	PUNCT
ajst-9864	113	24	aiding	aid	VERB
ajst-9864	113	25	the	the	DET
ajst-9864	113	26	recognition	recognition	NOUN
ajst-9864	113	27	of	of	ADP
ajst-9864	113	28	long	long	ADJ
ajst-9864	113	29	-	-	PUNCT
ajst-9864	113	30	distance	distance	NOUN
ajst-9864	113	31	dependencies	dependency	NOUN
ajst-9864	113	32	.	.	PUNCT
ajst-9864	114	1	its	its	PRON
ajst-9864	114	2	value	value	NOUN
ajst-9864	114	3	becomes	become	VERB
ajst-9864	114	4	more	more	ADV
ajst-9864	114	5	evident	evident	ADJ
ajst-9864	114	6	with	with	ADP
ajst-9864	114	7	longer	long	ADJ
ajst-9864	114	8	inputs	input	NOUN
ajst-9864	114	9	,	,	PUNCT
ajst-9864	114	10	where	where	SCONJ
ajst-9864	114	11	,	,	PUNCT
ajst-9864	114	12	without	without	ADP
ajst-9864	114	13	attention	attention	NOUN
ajst-9864	114	14	,	,	PUNCT
ajst-9864	114	15	a	a	DET
ajst-9864	114	16	single	single	ADJ
ajst-9864	114	17	intermediate	intermediate	ADJ
ajst-9864	114	18	semantic	semantic	ADJ
ajst-9864	114	19	vector	vector	NOUN
ajst-9864	114	20	can	can	AUX
ajst-9864	114	21	lose	lose	VERB
ajst-9864	114	22	significant	significant	ADJ
ajst-9864	114	23	detail[11	detail[11	NOUN
ajst-9864	114	24	]	]	X
ajst-9864	114	25	..	..	PUNCT
ajst-9864	114	26	attention	attention	NOUN
ajst-9864	114	27	can	can	AUX
ajst-9864	114	28	be	be	AUX
ajst-9864	114	29	word	word	NOUN
ajst-9864	114	30	-	-	PUNCT
ajst-9864	114	31	level	level	NOUN
ajst-9864	114	32	,	,	PUNCT
ajst-9864	114	33	used	use	VERB
ajst-9864	114	34	to	to	PART
ajst-9864	114	35	discern	discern	VERB
ajst-9864	114	36	relationships	relationship	NOUN
ajst-9864	114	37	and	and	CCONJ
ajst-9864	114	38	significance	significance	NOUN
ajst-9864	114	39	between	between	ADP
ajst-9864	114	40	words	word	NOUN
ajst-9864	114	41	.	.	PUNCT
ajst-9864	115	1	recurrent	recurrent	ADJ
ajst-9864	115	2	neural	neural	ADJ
ajst-9864	115	3	networks	network	NOUN
ajst-9864	115	4	(	(	PUNCT
ajst-9864	115	5	rnn	rnn	PROPN
ajst-9864	115	6	)	)	PUNCT
ajst-9864	115	7	assign	assign	VERB
ajst-9864	115	8	different	different	ADJ
ajst-9864	115	9	weights	weight	NOUN
ajst-9864	115	10	to	to	ADP
ajst-9864	115	11	each	each	DET
ajst-9864	115	12	time	time	NOUN
ajst-9864	115	13	step	step	NOUN
ajst-9864	115	14	for	for	ADP
ajst-9864	115	15	focus	focus	NOUN
ajst-9864	115	16	on	on	ADP
ajst-9864	115	17	crucial	crucial	ADJ
ajst-9864	115	18	words	word	NOUN
ajst-9864	115	19	.	.	PUNCT
ajst-9864	116	1	convolutional	convolutional	ADJ
ajst-9864	116	2	neural	neural	ADJ
ajst-9864	116	3	networks	network	NOUN
ajst-9864	116	4	(	(	PUNCT
ajst-9864	116	5	cnn	cnn	PROPN
ajst-9864	116	6	)	)	PUNCT
ajst-9864	116	7	,	,	PUNCT
ajst-9864	116	8	typically	typically	ADV
ajst-9864	116	9	used	use	VERB
ajst-9864	116	10	in	in	ADP
ajst-9864	116	11	image	image	NOUN
ajst-9864	116	12	processing	processing	NOUN
ajst-9864	116	13	,	,	PUNCT
ajst-9864	116	14	can	can	AUX
ajst-9864	116	15	also	also	ADV
ajst-9864	116	16	apply	apply	VERB
ajst-9864	116	17	this	this	DET
ajst-9864	116	18	approach	approach	NOUN
ajst-9864	116	19	to	to	ADP
ajst-9864	116	20	nlp	nlp	NOUN
ajst-9864	116	21	tasks	task	NOUN
ajst-9864	116	22	.	.	PUNCT
ajst-9864	117	1	sentence	sentence	NOUN
ajst-9864	117	2	-	-	PUNCT
ajst-9864	117	3	level	level	NOUN
ajst-9864	117	4	attention	attention	NOUN
ajst-9864	117	5	helps	help	VERB
ajst-9864	117	6	in	in	ADP
ajst-9864	117	7	understanding	understand	VERB
ajst-9864	117	8	entire	entire	ADJ
ajst-9864	117	9	sentences	sentence	NOUN
ajst-9864	117	10	by	by	ADP
ajst-9864	117	11	allocating	allocate	VERB
ajst-9864	117	12	varying	vary	VERB
ajst-9864	117	13	weights	weight	NOUN
ajst-9864	117	14	to	to	ADP
ajst-9864	117	15	different	different	ADJ
ajst-9864	117	16	sentence	sentence	NOUN
ajst-9864	117	17	parts	part	NOUN
ajst-9864	117	18	.	.	PUNCT
ajst-9864	118	1	hierarchical	hierarchical	ADJ
ajst-9864	118	2	attention	attention	NOUN
ajst-9864	118	3	networks	network	NOUN
ajst-9864	118	4	(	(	PUNCT
ajst-9864	118	5	han	han	PROPN
ajst-9864	118	6	)	)	PUNCT
ajst-9864	118	7	model	model	NOUN
ajst-9864	118	8	long	long	ADJ
ajst-9864	118	9	text	text	NOUN
ajst-9864	118	10	at	at	ADP
ajst-9864	118	11	both	both	CCONJ
ajst-9864	118	12	the	the	DET
ajst-9864	118	13	word	word	NOUN
ajst-9864	118	14	and	and	CCONJ
ajst-9864	118	15	sentence	sentence	NOUN
ajst-9864	118	16	levels	level	NOUN
ajst-9864	118	17	,	,	PUNCT
ajst-9864	118	18	while	while	SCONJ
ajst-9864	118	19	transformers	transformer	NOUN
ajst-9864	118	20	use	use	VERB
ajst-9864	118	21	self	self	NOUN
ajst-9864	118	22	-	-	PUNCT
ajst-9864	118	23	attention	attention	NOUN
ajst-9864	118	24	to	to	PART
ajst-9864	118	25	encode	encode	VERB
ajst-9864	118	26	the	the	DET
ajst-9864	118	27	input	input	NOUN
ajst-9864	118	28	sequence	sequence	NOUN
ajst-9864	118	29	and	and	CCONJ
ajst-9864	118	30	predict	predict	VERB
ajst-9864	118	31	downstream	downstream	ADJ
ajst-9864	118	32	tasks	task	NOUN
ajst-9864	118	33	,	,	PUNCT
ajst-9864	118	34	emphasizing	emphasize	VERB
ajst-9864	118	35	important	important	ADJ
ajst-9864	118	36	sentence	sentence	NOUN
ajst-9864	118	37	parts	part	NOUN
ajst-9864	118	38	.	.	PUNCT
ajst-9864	119	1	this	this	DET
ajst-9864	119	2	study	study	NOUN
ajst-9864	119	3	uses	use	VERB
ajst-9864	119	4	transformers	transformer	NOUN
ajst-9864	119	5	for	for	ADP
ajst-9864	119	6	sentence	sentence	NOUN
ajst-9864	119	7	-	-	PUNCT
ajst-9864	119	8	level	level	NOUN
ajst-9864	119	9	attention[12	attention[12	NOUN
ajst-9864	119	10	]	]	X
ajst-9864	119	11	..	..	PUNCT
ajst-9864	119	12	to	to	PART
ajst-9864	119	13	boost	boost	VERB
ajst-9864	119	14	its	its	PRON
ajst-9864	119	15	effectiveness	effectiveness	NOUN
ajst-9864	119	16	in	in	ADP
ajst-9864	119	17	connection	connection	NOUN
ajst-9864	119	18	extraction	extraction	NOUN
ajst-9864	119	19	tasks	task	NOUN
ajst-9864	119	20	,	,	PUNCT
ajst-9864	119	21	the	the	DET
ajst-9864	119	22	article	article	NOUN
ajst-9864	119	23	suggests	suggest	VERB
ajst-9864	119	24	training	train	VERB
ajst-9864	119	25	a	a	DET
ajst-9864	119	26	bidirectional	bidirectional	ADJ
ajst-9864	119	27	gru	gru	PROPN
ajst-9864	119	28	neural	neural	ADJ
ajst-9864	119	29	network	network	NOUN
ajst-9864	119	30	to	to	PART
ajst-9864	119	31	simultaneously	simultaneously	ADV
ajst-9864	119	32	pay	pay	VERB
ajst-9864	119	33	attention	attention	NOUN
ajst-9864	119	34	to	to	ADP
ajst-9864	119	35	both	both	DET
ajst-9864	119	36	words	word	NOUN
ajst-9864	119	37	and	and	CCONJ
ajst-9864	119	38	sentences	sentence	NOUN
ajst-9864	119	39	.	.	PUNCT
ajst-9864	120	1	let	let	VERB
ajst-9864	120	2	the	the	DET
ajst-9864	120	3	bidirectional	bidirectional	PROPN
ajst-9864	120	4	gru	gru	PROPN
ajst-9864	120	5	neural	neural	ADJ
ajst-9864	120	6	network	network	NOUN
ajst-9864	120	7	layer	layer	NOUN
ajst-9864	120	8	's	's	PART
ajst-9864	120	9	output	output	NOUN
ajst-9864	120	10	vectors	vector	NOUN
ajst-9864	120	11	h	h	NOUN
ajst-9864	120	12	form	form	VERB
ajst-9864	120	13	a	a	DET
ajst-9864	120	14	matrix	matrix	NOUN
ajst-9864	120	15	,	,	PUNCT
ajst-9864	120	16	denoted	denote	VERB
ajst-9864	120	17	as	as	ADP
ajst-9864	120	18	h1	h1	PROPN
ajst-9864	120	19	,	,	PUNCT
ajst-9864	120	20	h2	h2	PROPN
ajst-9864	120	21	,	,	PUNCT
ajst-9864	120	22	...	...	PUNCT
ajst-9864	120	23	,	,	PUNCT
ajst-9864	120	24	ht	ht	INTJ
ajst-9864	120	25	,	,	PUNCT
ajst-9864	120	26	where	where	SCONJ
ajst-9864	120	27	t	t	PROPN
ajst-9864	120	28	is	be	AUX
ajst-9864	120	29	the	the	DET
ajst-9864	120	30	total	total	ADJ
ajst-9864	120	31	number	number	NOUN
ajst-9864	120	32	of	of	ADP
ajst-9864	120	33	words	word	NOUN
ajst-9864	120	34	in	in	ADP
ajst-9864	120	35	the	the	DET
ajst-9864	120	36	sentence	sentence	NOUN
ajst-9864	120	37	.	.	PUNCT
ajst-9864	121	1	the	the	DET
ajst-9864	121	2	following	follow	VERB
ajst-9864	121	3	output	output	NOUN
ajst-9864	121	4	vectors	vector	NOUN
ajst-9864	121	5	are	be	AUX
ajst-9864	121	6	weighted	weight	VERB
ajst-9864	121	7	and	and	CCONJ
ajst-9864	121	8	added	add	VERB
ajst-9864	121	9	together	together	ADV
ajst-9864	121	10	to	to	PART
ajst-9864	121	11	generate	generate	VERB
ajst-9864	121	12	the	the	DET
ajst-9864	121	13	representation	representation	NOUN
ajst-9864	121	14	of	of	ADP
ajst-9864	121	15	phrase	phrase	NOUN
ajst-9864	121	16	r	r	NOUN
ajst-9864	121	17	:	:	PUNCT
ajst-9864	121	18	m	m	PROPN
ajst-9864	121	19	tanh	tanh	PROPN
ajst-9864	121	20	h	h	PROPN
ajst-9864	121	21	(	(	PUNCT
ajst-9864	121	22	6	6	NUM
ajst-9864	121	23	)	)	PUNCT
ajst-9864	121	24	α	α	PRON
ajst-9864	121	25	software	software	NOUN
ajst-9864	121	26	ωtm	ωtm	NOUN
ajst-9864	121	27	(	(	PUNCT
ajst-9864	121	28	7	7	X
ajst-9864	121	29	)	)	PUNCT
ajst-9864	121	30	r	r	NOUN
ajst-9864	121	31	hαt	hαt	NOUN
ajst-9864	121	32	(	(	PUNCT
ajst-9864	121	33	8)	8)	NUM
ajst-9864	121	34	where	where	SCONJ
ajst-9864	121	35	ｈ∈ｒｄｗ×ｔ	ｈ∈ｒｄｗ×ｔ	PROPN
ajst-9864	121	36	,	,	PUNCT
ajst-9864	121	37	  	  	SPACE
ajst-9864	121	38	ｍ∈ｒｄｗ×ｔ	ｍ∈ｒｄｗ×ｔ	NOUN
ajst-9864	121	39	.	.	PUNCT
ajst-9864	122	1	the	the	DET
ajst-9864	122	2	tanh	tanh	PROPN
ajst-9864	122	3	function	function	NOUN
ajst-9864	122	4	is	be	AUX
ajst-9864	122	5	used	use	VERB
ajst-9864	122	6	to	to	PART
ajst-9864	122	7	transform	transform	VERB
ajst-9864	122	8	the	the	DET
ajst-9864	122	9	original	original	ADJ
ajst-9864	122	10	vector	vector	NOUN
ajst-9864	122	11	to	to	ADP
ajst-9864	122	12	between	between	ADP
ajst-9864	122	13	[	[	X
ajst-9864	122	14	-1	-1	X
ajst-9864	122	15	,	,	PUNCT
ajst-9864	122	16	1	1	NUM
ajst-9864	122	17	]	]	PUNCT
ajst-9864	122	18	.	.	PUNCT
ajst-9864	123	1	the	the	DET
ajst-9864	123	2	size	size	NOUN
ajst-9864	123	3	of	of	ADP
ajst-9864	123	4	is	be	AUX
ajst-9864	123	5	t	t	PROPN
ajst-9864	123	6	,	,	PUNCT
ajst-9864	123	7	w	w	PROPN
ajst-9864	123	8	is	be	AUX
ajst-9864	123	9	a	a	DET
ajst-9864	123	10	vector	vector	NOUN
ajst-9864	123	11	of	of	ADP
ajst-9864	123	12	training	training	NOUN
ajst-9864	123	13	parameters	parameter	NOUN
ajst-9864	123	14	,	,	PUNCT
ajst-9864	123	15	wt	wt	PROPN
ajst-9864	123	16	is	be	AUX
ajst-9864	123	17	the	the	DET
ajst-9864	123	18	transpose	transpose	NOUN
ajst-9864	123	19	of	of	ADP
ajst-9864	123	20	w	w	PROPN
ajst-9864	123	21	,	,	PUNCT
ajst-9864	123	22	and	and	CCONJ
ajst-9864	123	23	the	the	DET
ajst-9864	123	24	dimension	dimension	NOUN
ajst-9864	123	25	of	of	ADP
ajst-9864	123	26	the	the	DET
ajst-9864	123	27	word	word	NOUN
ajst-9864	123	28	vector	vector	NOUN
ajst-9864	123	29	is	be	AUX
ajst-9864	123	30	.	.	PUNCT
ajst-9864	124	1	using	use	VERB
ajst-9864	124	2	this	this	DET
ajst-9864	124	3	equation	equation	NOUN
ajst-9864	124	4	,	,	PUNCT
ajst-9864	124	5	we	we	PRON
ajst-9864	124	6	may	may	AUX
ajst-9864	124	7	derive	derive	VERB
ajst-9864	124	8	a	a	DET
ajst-9864	124	9	representation	representation	NOUN
ajst-9864	124	10	of	of	ADP
ajst-9864	124	11	the	the	DET
ajst-9864	124	12	phrase	phrase	NOUN
ajst-9864	124	13	132	132	NUM
ajst-9864	124	14	that	that	PRON
ajst-9864	124	15	is	be	AUX
ajst-9864	124	16	based	base	VERB
ajst-9864	124	17	on	on	ADP
ajst-9864	124	18	a	a	DET
ajst-9864	124	19	system	system	NOUN
ajst-9864	124	20	for	for	ADP
ajst-9864	124	21	paying	pay	VERB
ajst-9864	124	22	attention	attention	NOUN
ajst-9864	124	23	to	to	ADP
ajst-9864	124	24	individual	individual	ADJ
ajst-9864	124	25	words.(1	words.(1	PROPN
ajst-9864	124	26	)	)	PUNCT
ajst-9864	124	27	.	.	PUNCT
ajst-9864	125	1	h∗	h∗	PROPN
ajst-9864	125	2	tanh	tanh	PROPN
ajst-9864	125	3	r	r	PROPN
ajst-9864	125	4	(	(	PUNCT
ajst-9864	125	5	9	9	NUM
ajst-9864	125	6	)	)	PUNCT
ajst-9864	125	7	introduction	introduction	NOUN
ajst-9864	125	8	of	of	ADP
ajst-9864	125	9	a	a	DET
ajst-9864	125	10	sentence	sentence	NOUN
ajst-9864	125	11	-	-	PUNCT
ajst-9864	125	12	level	level	NOUN
ajst-9864	125	13	attention	attention	NOUN
ajst-9864	125	14	-	-	PUNCT
ajst-9864	125	15	based	base	VERB
ajst-9864	125	16	mechanism	mechanism	NOUN
ajst-9864	125	17	on	on	ADP
ajst-9864	125	18	a	a	DET
ajst-9864	125	19	sentence	sentence	NOUN
ajst-9864	125	20	representation	representation	NOUN
ajst-9864	125	21	of	of	ADP
ajst-9864	125	22	a	a	DET
ajst-9864	125	23	word	word	NOUN
ajst-9864	125	24	-	-	PUNCT
ajst-9864	125	25	level	level	NOUN
ajst-9864	125	26	attention	attention	NOUN
ajst-9864	125	27	-	-	PUNCT
ajst-9864	125	28	based	base	VERB
ajst-9864	125	29	mechanism	mechanism	NOUN
ajst-9864	125	30	.	.	PUNCT
ajst-9864	126	1	β	β	X
ajst-9864	126	2	ℓ	ℓ	PROPN
ajst-9864	126	3	∑	∑	PROPN
ajst-9864	126	4	ℓ	ℓ	PROPN
ajst-9864	126	5	(	(	PUNCT
ajst-9864	126	6	10	10	NUM
ajst-9864	126	7	)	)	PUNCT
ajst-9864	126	8	h∗	h∗	PROPN
ajst-9864	126	9	∑	∑	PROPN
ajst-9864	126	10	h∗	h∗	PROPN
ajst-9864	126	11	⋅	⋅	PROPN
ajst-9864	126	12	β	β	X
ajst-9864	126	13	(	(	PUNCT
ajst-9864	126	14	11	11	NUM
ajst-9864	126	15	)	)	PUNCT
ajst-9864	126	16	where	where	SCONJ
ajst-9864	126	17	:	:	PUNCT
ajst-9864	126	18	βｉ	βｉ	PRON
ajst-9864	126	19	is	be	AUX
ajst-9864	126	20	the	the	DET
ajst-9864	126	21	weight	weight	NOUN
ajst-9864	126	22	vector	vector	NOUN
ajst-9864	126	23	based	base	VERB
ajst-9864	126	24	on	on	ADP
ajst-9864	126	25	the	the	DET
ajst-9864	126	26	sentence	sentence	NOUN
ajst-9864	126	27	-	-	PUNCT
ajst-9864	126	28	level	level	NOUN
ajst-9864	126	29	attention	attention	NOUN
ajst-9864	126	30	mechanism	mechanism	NOUN
ajst-9864	126	31	.	.	PUNCT
ajst-9864	127	1	the	the	DET
ajst-9864	127	2	combination	combination	NOUN
ajst-9864	127	3	of	of	ADP
ajst-9864	127	4	word	word	NOUN
ajst-9864	127	5	-	-	PUNCT
ajst-9864	127	6	level	level	NOUN
ajst-9864	127	7	supervision	supervision	NOUN
ajst-9864	127	8	and	and	CCONJ
ajst-9864	127	9	sentencelevel	sentencelevel	NOUN
ajst-9864	127	10	supervision	supervision	NOUN
ajst-9864	127	11	can	can	AUX
ajst-9864	127	12	improve	improve	VERB
ajst-9864	127	13	the	the	DET
ajst-9864	127	14	accuracy	accuracy	NOUN
ajst-9864	127	15	and	and	CCONJ
ajst-9864	127	16	recall	recall	NOUN
ajst-9864	127	17	of	of	ADP
ajst-9864	127	18	chinese	chinese	ADJ
ajst-9864	127	19	text	text	NOUN
ajst-9864	127	20	relationship	relationship	NOUN
ajst-9864	127	21	extraction	extraction	NOUN
ajst-9864	127	22	and	and	CCONJ
ajst-9864	127	23	can	can	AUX
ajst-9864	127	24	solve	solve	VERB
ajst-9864	127	25	some	some	PRON
ajst-9864	127	26	of	of	ADP
ajst-9864	127	27	the	the	DET
ajst-9864	127	28	problems	problem	NOUN
ajst-9864	127	29	that	that	PRON
ajst-9864	127	30	exist	exist	VERB
ajst-9864	127	31	when	when	SCONJ
ajst-9864	127	32	using	use	VERB
ajst-9864	127	33	word	word	NOUN
ajst-9864	127	34	-	-	PUNCT
ajst-9864	127	35	level	level	NOUN
ajst-9864	127	36	supervision	supervision	NOUN
ajst-9864	127	37	or	or	CCONJ
ajst-9864	127	38	sentence	sentence	NOUN
ajst-9864	127	39	-	-	PUNCT
ajst-9864	127	40	level	level	NOUN
ajst-9864	127	41	supervision	supervision	NOUN
ajst-9864	127	42	alone	alone	ADV
ajst-9864	127	43	.	.	PUNCT
ajst-9864	128	1	3.1.4	3.1.4	X
ajst-9864	128	2	.	.	PUNCT
ajst-9864	128	3	softmax	softmax	PROPN
ajst-9864	128	4	classifier	classifier	NOUN
ajst-9864	128	5	using	use	VERB
ajst-9864	128	6	the	the	DET
ajst-9864	128	7	softmax	softmax	NOUN
ajst-9864	128	8	function	function	NOUN
ajst-9864	128	9	,	,	PUNCT
ajst-9864	128	10	a	a	DET
ajst-9864	128	11	set	set	NOUN
ajst-9864	128	12	of	of	ADP
ajst-9864	128	13	k	k	PROPN
ajst-9864	128	14	real	real	ADJ
ajst-9864	128	15	values	value	NOUN
ajst-9864	128	16	may	may	AUX
ajst-9864	128	17	be	be	AUX
ajst-9864	128	18	converted	convert	VERB
ajst-9864	128	19	into	into	ADP
ajst-9864	128	20	a	a	DET
ajst-9864	128	21	set	set	NOUN
ajst-9864	128	22	of	of	ADP
ajst-9864	128	23	k	k	PROPN
ajst-9864	128	24	real	real	ADJ
ajst-9864	128	25	values	value	NOUN
ajst-9864	128	26	that	that	PRON
ajst-9864	128	27	sum	sum	VERB
ajst-9864	128	28	to	to	ADP
ajst-9864	128	29	1	1	NUM
ajst-9864	128	30	.	.	PUNCT
ajst-9864	129	1	in	in	ADP
ajst-9864	129	2	order	order	NOUN
ajst-9864	129	3	to	to	PART
ajst-9864	129	4	be	be	AUX
ajst-9864	129	5	comprehended	comprehend	VERB
ajst-9864	129	6	as	as	ADP
ajst-9864	129	7	probabilities	probability	NOUN
ajst-9864	129	8	,	,	PUNCT
ajst-9864	129	9	softmax	softmax	NOUN
ajst-9864	129	10	converts	convert	VERB
ajst-9864	129	11	the	the	DET
ajst-9864	129	12	inputs	input	NOUN
ajst-9864	129	13	,	,	PUNCT
ajst-9864	129	14	which	which	PRON
ajst-9864	129	15	might	might	AUX
ajst-9864	129	16	be	be	AUX
ajst-9864	129	17	positive	positive	ADJ
ajst-9864	129	18	,	,	PUNCT
ajst-9864	129	19	negative	negative	ADJ
ajst-9864	129	20	,	,	PUNCT
ajst-9864	129	21	zero	zero	NUM
ajst-9864	129	22	,	,	PUNCT
ajst-9864	129	23	or	or	CCONJ
ajst-9864	129	24	more	more	ADJ
ajst-9864	129	25	than	than	ADP
ajst-9864	129	26	1	1	NUM
ajst-9864	129	27	,	,	PUNCT
ajst-9864	129	28	into	into	ADP
ajst-9864	129	29	values	value	NOUN
ajst-9864	129	30	between	between	ADP
ajst-9864	129	31	0	0	NUM
ajst-9864	129	32	and	and	CCONJ
ajst-9864	129	33	1	1	NUM
ajst-9864	129	34	.	.	X
ajst-9864	130	1	if	if	SCONJ
ajst-9864	130	2	the	the	DET
ajst-9864	130	3	input	input	NOUN
ajst-9864	130	4	is	be	AUX
ajst-9864	130	5	tiny	tiny	ADJ
ajst-9864	130	6	or	or	CCONJ
ajst-9864	130	7	negative	negative	ADJ
ajst-9864	130	8	,	,	PUNCT
ajst-9864	130	9	the	the	DET
ajst-9864	130	10	resulting	result	VERB
ajst-9864	130	11	probability	probability	NOUN
ajst-9864	130	12	is	be	AUX
ajst-9864	130	13	also	also	ADV
ajst-9864	130	14	small	small	ADJ
ajst-9864	130	15	,	,	PUNCT
ajst-9864	130	16	and	and	CCONJ
ajst-9864	130	17	if	if	SCONJ
ajst-9864	130	18	the	the	DET
ajst-9864	130	19	input	input	NOUN
ajst-9864	130	20	is	be	AUX
ajst-9864	130	21	high	high	ADJ
ajst-9864	130	22	,	,	PUNCT
ajst-9864	130	23	the	the	DET
ajst-9864	130	24	resulting	result	VERB
ajst-9864	130	25	probability	probability	NOUN
ajst-9864	130	26	is	be	AUX
ajst-9864	130	27	huge	huge	ADJ
ajst-9864	130	28	,	,	PUNCT
ajst-9864	130	29	but	but	CCONJ
ajst-9864	130	30	the	the	DET
ajst-9864	130	31	range	range	NOUN
ajst-9864	130	32	of	of	ADP
ajst-9864	130	33	values	value	NOUN
ajst-9864	130	34	never	never	ADV
ajst-9864	130	35	goes	go	VERB
ajst-9864	130	36	outside	outside	ADV
ajst-9864	130	37	of	of	ADP
ajst-9864	130	38	0	0	NUM
ajst-9864	130	39	and	and	CCONJ
ajst-9864	130	40	1[13	1[13	NUM
ajst-9864	130	41	]	]	X
ajst-9864	130	42	.	.	PUNCT
ajst-9864	131	1	a	a	DET
ajst-9864	131	2	real	real	ADV
ajst-9864	131	3	-	-	PUNCT
ajst-9864	131	4	valued	value	VERB
ajst-9864	131	5	score	score	NOUN
ajst-9864	131	6	is	be	AUX
ajst-9864	131	7	generated	generate	VERB
ajst-9864	131	8	at	at	ADP
ajst-9864	131	9	the	the	DET
ajst-9864	131	10	penultimate	penultimate	NOUN
ajst-9864	131	11	layer	layer	NOUN
ajst-9864	131	12	of	of	ADP
ajst-9864	131	13	many	many	ADJ
ajst-9864	131	14	multilayer	multilayer	ADJ
ajst-9864	131	15	neural	neural	ADJ
ajst-9864	131	16	networks	network	NOUN
ajst-9864	131	17	,	,	PUNCT
ajst-9864	131	18	which	which	PRON
ajst-9864	131	19	may	may	AUX
ajst-9864	131	20	be	be	AUX
ajst-9864	131	21	hard	hard	ADJ
ajst-9864	131	22	to	to	PART
ajst-9864	131	23	scale	scale	VERB
ajst-9864	131	24	and	and	CCONJ
ajst-9864	131	25	may	may	AUX
ajst-9864	131	26	be	be	AUX
ajst-9864	131	27	tricky	tricky	ADJ
ajst-9864	131	28	to	to	PART
ajst-9864	131	29	employ	employ	VERB
ajst-9864	131	30	.	.	PUNCT
ajst-9864	132	1	softmax	softmax	PROPN
ajst-9864	132	2	is	be	AUX
ajst-9864	132	3	useful	useful	ADJ
ajst-9864	132	4	here	here	ADV
ajst-9864	132	5	since	since	SCONJ
ajst-9864	132	6	it	it	PRON
ajst-9864	132	7	converts	convert	VERB
ajst-9864	132	8	the	the	DET
ajst-9864	132	9	findings	finding	NOUN
ajst-9864	132	10	into	into	ADP
ajst-9864	132	11	a	a	DET
ajst-9864	132	12	normalised	normalise	VERB
ajst-9864	132	13	probability	probability	NOUN
ajst-9864	132	14	distribution	distribution	NOUN
ajst-9864	132	15	that	that	PRON
ajst-9864	132	16	may	may	AUX
ajst-9864	132	17	be	be	AUX
ajst-9864	132	18	shown	show	VERB
ajst-9864	132	19	to	to	ADP
ajst-9864	132	20	the	the	DET
ajst-9864	132	21	user	user	NOUN
ajst-9864	132	22	or	or	CCONJ
ajst-9864	132	23	used	use	VERB
ajst-9864	132	24	as	as	ADP
ajst-9864	132	25	input	input	NOUN
ajst-9864	132	26	to	to	ADP
ajst-9864	132	27	other	other	ADJ
ajst-9864	132	28	systems	system	NOUN
ajst-9864	132	29	.	.	PUNCT
ajst-9864	133	1	for	for	ADP
ajst-9864	133	2	this	this	DET
ajst-9864	133	3	reason	reason	NOUN
ajst-9864	133	4	,	,	PUNCT
ajst-9864	133	5	the	the	DET
ajst-9864	133	6	neural	neural	ADJ
ajst-9864	133	7	network	network	NOUN
ajst-9864	133	8	will	will	AUX
ajst-9864	133	9	adopt	adopt	VERB
ajst-9864	133	10	a	a	DET
ajst-9864	133	11	softmax	softmax	NOUN
ajst-9864	133	12	function	function	NOUN
ajst-9864	133	13	as	as	ADP
ajst-9864	133	14	its	its	PRON
ajst-9864	133	15	last	last	ADJ
ajst-9864	133	16	hidden	hide	VERB
ajst-9864	133	17	layer	layer	NOUN
ajst-9864	133	18	.	.	PUNCT
ajst-9864	134	1	the	the	DET
ajst-9864	134	2	formula	formula	NOUN
ajst-9864	134	3	gives	give	VERB
ajst-9864	134	4	the	the	DET
ajst-9864	134	5	definition	definition	NOUN
ajst-9864	134	6	of	of	ADP
ajst-9864	134	7	the	the	DET
ajst-9864	134	8	common	common	ADJ
ajst-9864	134	9	(	(	PUNCT
ajst-9864	134	10	unit	unit	NOUN
ajst-9864	134	11	)	)	PUNCT
ajst-9864	134	12	softmax	softmax	NOUN
ajst-9864	134	13	function	function	NOUN
ajst-9864	134	14	.	.	PUNCT
ajst-9864	135	1	σ	σ	X
ajst-9864	135	2	z	z	PROPN
ajst-9864	135	3	∑	∑	PUNCT
ajst-9864	135	4	fori=1,	fori=1,	PROPN
ajst-9864	135	5	…	…	SYM
ajst-9864	135	6	,kandz=(z	,kandz=(z	X
ajst-9864	135	7	,	,	PUNCT
ajst-9864	135	8	…	…	PUNCT
ajst-9864	135	9	,	,	PUNCT
ajst-9864	135	10	z	z	NOUN
ajst-9864	135	11	∈	∈	PROPN
ajst-9864	135	12	ℝ	ℝ	PROPN
ajst-9864	135	13	(	(	PUNCT
ajst-9864	135	14	12	12	NUM
ajst-9864	135	15	)	)	PUNCT
ajst-9864	135	16	the	the	DET
ajst-9864	135	17	classifier	classifier	NOUN
ajst-9864	135	18	takes	take	VERB
ajst-9864	135	19	the	the	DET
ajst-9864	135	20	hidden	hide	VERB
ajst-9864	135	21	state	state	NOUN
ajst-9864	135	22	hs	hs	INTJ
ajst-9864	135	23	*	*	PUNCT
ajst-9864	135	24	as	as	ADP
ajst-9864	135	25	input	input	NOUN
ajst-9864	135	26	:	:	PUNCT
ajst-9864	135	27	p	p	X
ajst-9864	135	28	(	(	PUNCT
ajst-9864	135	29	y	y	PROPN
ajst-9864	135	30	|	|	NOUN
ajst-9864	135	31	s	s	PART
ajst-9864	135	32	)	)	PUNCT
ajst-9864	135	33	=	=	SYM
ajst-9864	135	34	softmax	softmax	NOUN
ajst-9864	135	35	(	(	PUNCT
ajst-9864	135	36	w(s)h*s	w(s)h*s	PROPN
ajst-9864	135	37	+	+	CCONJ
ajst-9864	135	38	b	b	PROPN
ajst-9864	135	39	(	(	PUNCT
ajst-9864	135	40	s	s	NOUN
ajst-9864	135	41	)	)	PUNCT
ajst-9864	135	42	)	)	PUNCT
ajst-9864	135	43	(	(	PUNCT
ajst-9864	135	44	13	13	X
ajst-9864	135	45	)	)	PUNCT
ajst-9864	135	46	y	y	NOUN
ajst-9864	135	47	=	=	PUNCT
ajst-9864	135	48	argmax	argmax	PROPN
ajst-9864	135	49	(	(	PUNCT
ajst-9864	135	50	y	y	PROPN
ajst-9864	135	51	|	|	ADV
ajst-9864	135	52	s	s	PART
ajst-9864	135	53	)	)	PUNCT
ajst-9864	135	54	(	(	PUNCT
ajst-9864	135	55	14	14	NUM
ajst-9864	135	56	)	)	PUNCT
ajst-9864	135	57	the	the	DET
ajst-9864	135	58	loss	loss	NOUN
ajst-9864	135	59	function	function	NOUN
ajst-9864	135	60	is	be	AUX
ajst-9864	135	61	the	the	DET
ajst-9864	135	62	negative	negative	ADJ
ajst-9864	135	63	log	log	NOUN
ajst-9864	135	64	-	-	PUNCT
ajst-9864	135	65	likelihood	likelihood	NOUN
ajst-9864	135	66	value	value	NOUN
ajst-9864	135	67	of	of	ADP
ajst-9864	135	68	the	the	DET
ajst-9864	135	69	true	true	ADJ
ajst-9864	135	70	category	category	NOUN
ajst-9864	135	71	label	label	NOUN
ajst-9864	135	72	y	y	NOUN
ajst-9864	135	73	:	:	PUNCT
ajst-9864	135	74	j(θ	j(θ	PROPN
ajst-9864	135	75	)	)	PUNCT
ajst-9864	136	1	=	=	SYM
ajst-9864	136	2	－	－	ADJ
ajst-9864	136	3	∑	∑	PUNCT
ajst-9864	136	4	ti	ti	NOUN
ajst-9864	136	5	log(yi	log(yi	ADV
ajst-9864	136	6	)	)	PUNCT
ajst-9864	137	1	+	+	CCONJ
ajst-9864	137	2	λǁθǁ2f	λǁθǁ2f	X
ajst-9864	137	3	(	(	PUNCT
ajst-9864	137	4	15	15	NUM
ajst-9864	137	5	)	)	PUNCT
ajst-9864	137	6	where	where	SCONJ
ajst-9864	137	7	:	:	PUNCT
ajst-9864	137	8	t	t	PROPN
ajst-9864	137	9	is	be	AUX
ajst-9864	137	10	the	the	DET
ajst-9864	137	11	true	true	ADJ
ajst-9864	137	12	value	value	NOUN
ajst-9864	137	13	using	use	VERB
ajst-9864	137	14	one	one	NUM
ajst-9864	137	15	-	-	PUNCT
ajst-9864	137	16	hot	hot	ADJ
ajst-9864	137	17	representation	representation	NOUN
ajst-9864	137	18	;	;	PUNCT
ajst-9864	137	19	y	y	PROPN
ajst-9864	137	20	is	be	AUX
ajst-9864	137	21	the	the	DET
ajst-9864	137	22	probability	probability	NOUN
ajst-9864	137	23	of	of	ADP
ajst-9864	137	24	each	each	DET
ajst-9864	137	25	category	category	NOUN
ajst-9864	137	26	estimated	estimate	VERB
ajst-9864	137	27	using	use	VERB
ajst-9864	137	28	the	the	DET
ajst-9864	137	29	softmax	softmax	NOUN
ajst-9864	137	30	function	function	NOUN
ajst-9864	137	31	;	;	PUNCT
ajst-9864	137	32	and	and	CCONJ
ajst-9864	137	33	λ	λ	NOUN
ajst-9864	137	34	is	be	AUX
ajst-9864	137	35	the	the	DET
ajst-9864	137	36	l2	l2	NOUN
ajst-9864	137	37	regularization	regularization	NOUN
ajst-9864	137	38	parameter	parameter	NOUN
ajst-9864	137	39	.	.	PUNCT
ajst-9864	138	1	furthermore	furthermore	ADV
ajst-9864	138	2	,	,	PUNCT
ajst-9864	138	3	hinton	hinton	PROPN
ajst-9864	138	4	has	have	AUX
ajst-9864	138	5	recommended	recommend	VERB
ajst-9864	138	6	using	use	VERB
ajst-9864	138	7	dropout	dropout	NOUN
ajst-9864	138	8	(	(	PUNCT
ajst-9864	138	9	ignoring	ignore	VERB
ajst-9864	138	10	certain	certain	ADJ
ajst-9864	138	11	units	unit	NOUN
ajst-9864	138	12	at	at	ADP
ajst-9864	138	13	random	random	ADJ
ajst-9864	138	14	in	in	ADP
ajst-9864	138	15	the	the	DET
ajst-9864	138	16	neural	neural	ADJ
ajst-9864	138	17	network	network	NOUN
ajst-9864	138	18	)	)	PUNCT
ajst-9864	138	19	to	to	PART
ajst-9864	138	20	avoid	avoid	VERB
ajst-9864	138	21	overloading	overload	VERB
ajst-9864	138	22	the	the	DET
ajst-9864	138	23	hidden	hidden	ADJ
ajst-9864	138	24	layer	layer	NOUN
ajst-9864	138	25	.	.	PUNCT
ajst-9864	139	1	to	to	PART
ajst-9864	139	2	combat	combat	VERB
ajst-9864	139	3	the	the	DET
ajst-9864	139	4	overfitting	overfitte	VERB
ajst-9864	139	5	issue	issue	NOUN
ajst-9864	139	6	of	of	ADP
ajst-9864	139	7	hidden	hide	VERB
ajst-9864	139	8	layer	layer	NOUN
ajst-9864	139	9	units	unit	NOUN
ajst-9864	139	10	,	,	PUNCT
ajst-9864	139	11	hinton	hinton	PROPN
ajst-9864	139	12	advocated	advocate	VERB
ajst-9864	139	13	using	use	VERB
ajst-9864	139	14	dropout	dropout	NOUN
ajst-9864	139	15	(	(	PUNCT
ajst-9864	139	16	randomly	randomly	ADV
ajst-9864	139	17	disregarding	disregard	VERB
ajst-9864	139	18	certain	certain	ADJ
ajst-9864	139	19	units	unit	NOUN
ajst-9864	139	20	in	in	ADP
ajst-9864	139	21	the	the	DET
ajst-9864	139	22	neural	neural	ADJ
ajst-9864	139	23	network	network	NOUN
ajst-9864	139	24	)	)	PUNCT
ajst-9864	139	25	.	.	PUNCT
ajst-9864	140	1	because	because	SCONJ
ajst-9864	140	2	of	of	ADP
ajst-9864	140	3	this	this	PRON
ajst-9864	140	4	,	,	PUNCT
ajst-9864	140	5	we	we	PRON
ajst-9864	140	6	use	use	VERB
ajst-9864	140	7	drop	drop	NOUN
ajst-9864	140	8	-	-	PUNCT
ajst-9864	140	9	out	out	NOUN
ajst-9864	140	10	on	on	ADP
ajst-9864	140	11	the	the	DET
ajst-9864	140	12	gru	gru	NOUN
ajst-9864	140	13	layer	layer	NOUN
ajst-9864	140	14	,	,	PUNCT
ajst-9864	140	15	the	the	DET
ajst-9864	140	16	penultimate	penultimate	NOUN
ajst-9864	140	17	layer	layer	NOUN
ajst-9864	140	18	,	,	PUNCT
ajst-9864	140	19	and	and	CCONJ
ajst-9864	140	20	the	the	DET
ajst-9864	140	21	embedding	embed	VERB
ajst-9864	140	22	layer	layer	NOUN
ajst-9864	140	23	.	.	PUNCT
ajst-9864	141	1	using	use	VERB
ajst-9864	141	2	eq	eq	X
ajst-9864	141	3	.	.	PUNCT
ajst-9864	142	1	(	(	PUNCT
ajst-9864	142	2	15	15	NUM
ajst-9864	142	3	)	)	PUNCT
ajst-9864	142	4	,	,	PUNCT
ajst-9864	142	5	we	we	PRON
ajst-9864	142	6	can	can	AUX
ajst-9864	142	7	see	see	VERB
ajst-9864	142	8	that	that	SCONJ
ajst-9864	142	9	after	after	SCONJ
ajst-9864	142	10	the	the	DET
ajst-9864	142	11	gradient	gradient	ADJ
ajst-9864	142	12	descent	descent	NOUN
ajst-9864	142	13	phase	phase	NOUN
ajst-9864	142	14	is	be	AUX
ajst-9864	142	15	complete	complete	ADJ
ajst-9864	142	16	,	,	PUNCT
ajst-9864	142	17	the	the	DET
ajst-9864	142	18	l2	l2	NOUN
ajst-9864	142	19	parametrization	parametrization	NOUN
ajst-9864	142	20	is	be	AUX
ajst-9864	142	21	constrained	constrain	VERB
ajst-9864	142	22	to	to	ADP
ajst-9864	142	23	values	value	NOUN
ajst-9864	142	24	of	of	ADP
ajst-9864	142	25	=	=	PUNCT
ajst-9864	142	26	s	s	PROPN
ajst-9864	142	27	for	for	ADP
ajst-9864	142	28	all	all	PRON
ajst-9864	142	29	>	>	X
ajst-9864	142	30	0	0	X
ajst-9864	142	31	.	.	PUNCT
ajst-9864	143	1	by	by	ADP
ajst-9864	143	2	combining	combine	VERB
ajst-9864	143	3	the	the	DET
ajst-9864	143	4	loss	loss	NOUN
ajst-9864	143	5	value	value	NOUN
ajst-9864	143	6	with	with	ADP
ajst-9864	143	7	l2	l2	NOUN
ajst-9864	143	8	regularisation	regularisation	NOUN
ajst-9864	143	9	,	,	PUNCT
ajst-9864	143	10	this	this	DET
ajst-9864	143	11	paper	paper	NOUN
ajst-9864	143	12	's	's	PART
ajst-9864	143	13	studies	study	NOUN
ajst-9864	143	14	successfully	successfully	ADV
ajst-9864	143	15	reduce	reduce	VERB
ajst-9864	143	16	the	the	DET
ajst-9864	143	17	risk	risk	NOUN
ajst-9864	143	18	of	of	ADP
ajst-9864	143	19	overfitting	overfitte	VERB
ajst-9864	143	20	.	.	PUNCT
ajst-9864	144	1	4	4	X
ajst-9864	144	2	.	.	X
ajst-9864	144	3	experiment	experiment	NOUN
ajst-9864	144	4	4.1	4.1	NUM
ajst-9864	144	5	.	.	PUNCT
ajst-9864	145	1	data	datum	NOUN
ajst-9864	145	2	set	set	VERB
ajst-9864	145	3	most	most	ADJ
ajst-9864	145	4	relationship	relationship	NOUN
ajst-9864	145	5	extraction	extraction	NOUN
ajst-9864	145	6	experiments	experiment	NOUN
ajst-9864	145	7	use	use	VERB
ajst-9864	145	8	english	english	PROPN
ajst-9864	145	9	corpora	corpora	PROPN
ajst-9864	145	10	due	due	ADP
ajst-9864	145	11	to	to	ADP
ajst-9864	145	12	limited	limited	ADJ
ajst-9864	145	13	public	public	ADJ
ajst-9864	145	14	chinese	chinese	ADJ
ajst-9864	145	15	character	character	NOUN
ajst-9864	145	16	relationship	relationship	NOUN
ajst-9864	145	17	corpora	corpora	PROPN
ajst-9864	145	18	.	.	PUNCT
ajst-9864	146	1	this	this	DET
ajst-9864	146	2	study	study	NOUN
ajst-9864	146	3	uses	use	VERB
ajst-9864	146	4	a	a	DET
ajst-9864	146	5	remotely	remotely	ADV
ajst-9864	146	6	supervised	supervise	VERB
ajst-9864	146	7	approach	approach	NOUN
ajst-9864	146	8	to	to	PART
ajst-9864	146	9	acquire	acquire	VERB
ajst-9864	146	10	training	training	NOUN
ajst-9864	146	11	samples	sample	NOUN
ajst-9864	146	12	from	from	ADP
ajst-9864	146	13	online	online	ADJ
ajst-9864	146	14	knowledge	knowledge	NOUN
ajst-9864	146	15	bases	basis	NOUN
ajst-9864	146	16	like	like	ADP
ajst-9864	146	17	cndb	cndb	PROPN
ajst-9864	146	18	pedia	pedia	PROPN
ajst-9864	146	19	and	and	CCONJ
ajst-9864	146	20	fudan	fudan	PROPN
ajst-9864	146	21	university	university	NOUN
ajst-9864	146	22	's	's	PART
ajst-9864	146	23	knowledge	knowledge	NOUN
ajst-9864	146	24	factory	factory	NOUN
ajst-9864	146	25	laboratory	laboratory	NOUN
ajst-9864	146	26	.	.	PUNCT
ajst-9864	147	1	using	use	VERB
ajst-9864	147	2	crawlers	crawler	NOUN
ajst-9864	147	3	,	,	PUNCT
ajst-9864	147	4	686	686	NUM
ajst-9864	147	5	entity	entity	NOUN
ajst-9864	147	6	pairs	pair	NOUN
ajst-9864	147	7	with	with	ADP
ajst-9864	147	8	defined	define	VERB
ajst-9864	147	9	relationships	relationship	NOUN
ajst-9864	147	10	and	and	CCONJ
ajst-9864	147	11	653	653	NUM
ajst-9864	147	12	pairs	pair	NOUN
ajst-9864	147	13	without	without	ADP
ajst-9864	147	14	definitive	definitive	ADJ
ajst-9864	147	15	relationships	relationship	NOUN
ajst-9864	147	16	were	be	AUX
ajst-9864	147	17	extracted	extract	VERB
ajst-9864	147	18	from	from	ADP
ajst-9864	147	19	news	news	NOUN
ajst-9864	147	20	websites	website	NOUN
ajst-9864	147	21	.	.	PUNCT
ajst-9864	148	1	these	these	PRON
ajst-9864	148	2	,	,	PUNCT
ajst-9864	148	3	classified	classify	VERB
ajst-9864	148	4	into	into	ADP
ajst-9864	148	5	11	11	NUM
ajst-9864	148	6	categories	category	NOUN
ajst-9864	148	7	,	,	PUNCT
ajst-9864	148	8	form	form	VERB
ajst-9864	148	9	the	the	DET
ajst-9864	148	10	dataset	dataset	NOUN
ajst-9864	148	11	of	of	ADP
ajst-9864	148	12	around	around	ADP
ajst-9864	148	13	1140	1140	NUM
ajst-9864	148	14	sentences	sentence	NOUN
ajst-9864	148	15	.	.	PUNCT
ajst-9864	149	1	adhering	adhere	VERB
ajst-9864	149	2	to	to	ADP
ajst-9864	149	3	the	the	DET
ajst-9864	149	4	"	"	PUNCT
ajst-9864	149	5	two	two	NUM
ajst-9864	149	6	and	and	CCONJ
ajst-9864	149	7	eight	eight	NUM
ajst-9864	149	8	rule	rule	NOUN
ajst-9864	149	9	"	"	PUNCT
ajst-9864	149	10	,	,	PUNCT
ajst-9864	149	11	80	80	NUM
ajst-9864	149	12	%	%	NOUN
ajst-9864	149	13	of	of	ADP
ajst-9864	149	14	this	this	DET
ajst-9864	149	15	data	data	NOUN
ajst-9864	149	16	is	be	AUX
ajst-9864	149	17	arbitrarily	arbitrarily	ADV
ajst-9864	149	18	split	split	VERB
ajst-9864	149	19	for	for	ADP
ajst-9864	149	20	training	training	NOUN
ajst-9864	149	21	and	and	CCONJ
ajst-9864	149	22	testing	testing	NOUN
ajst-9864	149	23	.	.	PUNCT
ajst-9864	150	1	table	table	NOUN
ajst-9864	150	2	2	2	NUM
ajst-9864	150	3	.	.	PUNCT
ajst-9864	151	1	various	various	ADJ
ajst-9864	151	2	types	type	NOUN
ajst-9864	151	3	of	of	ADP
ajst-9864	151	4	relationships	relationship	NOUN
ajst-9864	151	5	between	between	ADP
ajst-9864	151	6	entities	entity	NOUN
ajst-9864	151	7	relationship	relationship	NOUN
ajst-9864	151	8	number	number	NOUN
ajst-9864	151	9	category	category	NOUN
ajst-9864	151	10	unknown	unknown	ADJ
ajst-9864	151	11	455	455	NUM
ajst-9864	151	12	0	0	NUM
ajst-9864	151	13	parent	parent	NOUN
ajst-9864	151	14	155	155	NUM
ajst-9864	151	15	1	1	NUM
ajst-9864	151	16	marital	marital	ADJ
ajst-9864	151	17	196	196	NUM
ajst-9864	151	18	2	2	NUM
ajst-9864	151	19	teacher	teacher	NOUN
ajst-9864	151	20	-	-	PUNCT
ajst-9864	151	21	student	student	NOUN
ajst-9864	151	22	52	52	NUM
ajst-9864	151	23	3	3	NUM
ajst-9864	151	24	sibling	sible	VERB
ajst-9864	151	25	72	72	NUM
ajst-9864	151	26	4	4	NUM
ajst-9864	151	27	cooperative	cooperative	ADJ
ajst-9864	151	28	85	85	NUM
ajst-9864	151	29	5	5	NUM
ajst-9864	151	30	romantic	romantic	ADJ
ajst-9864	151	31	40	40	NUM
ajst-9864	151	32	6	6	NUM
ajst-9864	151	33	grandparent	grandparent	NOUN
ajst-9864	151	34	-	-	PUNCT
ajst-9864	151	35	grandchild	grandchild	ADJ
ajst-9864	151	36	10	10	NUM
ajst-9864	151	37	7	7	NUM
ajst-9864	151	38	friendship	friendship	NOUN
ajst-9864	151	39	28	28	NUM
ajst-9864	151	40	8	8	NUM
ajst-9864	151	41	kinship	kinship	NOUN
ajst-9864	151	42	20	20	NUM
ajst-9864	151	43	9	9	NUM
ajst-9864	151	44	same	same	ADJ
ajst-9864	151	45	teacher	teacher	NOUN
ajst-9864	151	46	10	10	NUM
ajst-9864	151	47	10	10	NUM
ajst-9864	151	48	superior	superior	ADJ
ajst-9864	151	49	subordinate	subordinate	ADJ
ajst-9864	151	50	15	15	NUM
ajst-9864	151	51	11	11	NUM
ajst-9864	151	52	133	133	NUM
ajst-9864	151	53	table	table	NOUN
ajst-9864	151	54	3	3	NUM
ajst-9864	151	55	.	.	PUNCT
ajst-9864	151	56	data	datum	NOUN
ajst-9864	151	57	set	set	VERB
ajst-9864	151	58	sample	sample	NOUN
ajst-9864	151	59	entity	entity	NOUN
ajst-9864	151	60	1	1	NUM
ajst-9864	151	61	entity	entity	NOUN
ajst-9864	151	62	2	2	NUM
ajst-9864	151	63	relation	relation	NOUN
ajst-9864	151	64	ship	ship	NOUN
ajst-9864	151	65	statement	statement	NOUN
ajst-9864	151	66	gong	gong	PROPN
ajst-9864	151	67	li	li	PROPN
ajst-9864	152	1	huang	huang	PROPN
ajst-9864	152	2	he	he	PROPN
ajst-9864	152	3	xiang	xiang	PROPN
ajst-9864	152	4	marital	marital	PROPN
ajst-9864	152	5	after	after	ADP
ajst-9864	152	6	their	their	PRON
ajst-9864	152	7	marriage	marriage	NOUN
ajst-9864	152	8	,	,	PUNCT
ajst-9864	152	9	gong	gong	PROPN
ajst-9864	152	10	li	li	PROPN
ajst-9864	152	11	and	and	CCONJ
ajst-9864	152	12	huang	huang	PROPN
ajst-9864	152	13	he	he	PRON
ajst-9864	152	14	xiang	xiang	PROPN
ajst-9864	152	15	made	make	VERB
ajst-9864	152	16	their	their	PRON
ajst-9864	152	17	home	home	NOUN
ajst-9864	152	18	in	in	ADP
ajst-9864	152	19	hong	hong	PROPN
ajst-9864	152	20	kong	kong	PROPN
ajst-9864	152	21	and	and	CCONJ
ajst-9864	152	22	lived	live	VERB
ajst-9864	152	23	an	an	DET
ajst-9864	152	24	ordinary	ordinary	ADJ
ajst-9864	152	25	life	life	NOUN
ajst-9864	152	26	as	as	ADP
ajst-9864	152	27	ordinary	ordinary	ADJ
ajst-9864	152	28	people	people	NOUN
ajst-9864	152	29	.	.	PUNCT
ajst-9864	153	1	in	in	ADP
ajst-9864	153	2	the	the	DET
ajst-9864	153	3	experiment	experiment	NOUN
ajst-9864	153	4	,	,	PUNCT
ajst-9864	153	5	google	google	PROPN
ajst-9864	153	6	's	's	PART
ajst-9864	153	7	word2vec	word2vec	NOUN
ajst-9864	153	8	tool	tool	NOUN
ajst-9864	153	9	creates	create	VERB
ajst-9864	153	10	word	word	NOUN
ajst-9864	153	11	vectors	vector	NOUN
ajst-9864	153	12	that	that	PRON
ajst-9864	153	13	become	become	VERB
ajst-9864	153	14	the	the	DET
ajst-9864	153	15	model	model	NOUN
ajst-9864	153	16	's	's	PART
ajst-9864	153	17	input	input	NOUN
ajst-9864	153	18	.	.	PUNCT
ajst-9864	154	1	pre	pre	VERB
ajst-9864	154	2	-	-	ADJ
ajst-9864	154	3	trained	train	VERB
ajst-9864	154	4	word	word	NOUN
ajst-9864	154	5	vectors	vector	NOUN
ajst-9864	154	6	are	be	AUX
ajst-9864	154	7	transformed	transform	VERB
ajst-9864	154	8	into	into	ADP
ajst-9864	154	9	unique	unique	ADJ
ajst-9864	154	10	word	word	NOUN
ajst-9864	154	11	ids	id	NOUN
ajst-9864	154	12	,	,	PUNCT
ajst-9864	154	13	and	and	CCONJ
ajst-9864	154	14	each	each	DET
ajst-9864	154	15	sentence	sentence	NOUN
ajst-9864	154	16	word	word	NOUN
ajst-9864	154	17	is	be	AUX
ajst-9864	154	18	expressed	express	VERB
ajst-9864	154	19	as	as	ADP
ajst-9864	154	20	a	a	DET
ajst-9864	154	21	3x1	3x1	NUM
ajst-9864	154	22	matrix	matrix	NOUN
ajst-9864	154	23	with	with	ADP
ajst-9864	154	24	format	format	NOUN
ajst-9864	155	1	[	[	X
ajst-9864	155	2	word	word	NOUN
ajst-9864	155	3	i	i	PROPN
ajst-9864	155	4	d	d	PROPN
ajst-9864	155	5	,	,	PUNCT
ajst-9864	155	6	distance	distance	NOUN
ajst-9864	155	7	to	to	ADP
ajst-9864	155	8	entity	entity	NOUN
ajst-9864	155	9	1	1	NUM
ajst-9864	155	10	,	,	PUNCT
ajst-9864	155	11	entity	entity	NOUN
ajst-9864	155	12	2	2	NUM
ajst-9864	155	13	]	]	PUNCT
ajst-9864	155	14	.	.	PUNCT
ajst-9864	156	1	these	these	DET
ajst-9864	156	2	matrices	matrix	NOUN
ajst-9864	156	3	form	form	VERB
ajst-9864	156	4	the	the	DET
ajst-9864	156	5	sentence	sentence	NOUN
ajst-9864	156	6	representation	representation	NOUN
ajst-9864	156	7	.	.	PUNCT
ajst-9864	157	1	entity	entity	NOUN
ajst-9864	157	2	pair	pair	NOUN
ajst-9864	157	3	relationships	relationship	NOUN
ajst-9864	157	4	are	be	AUX
ajst-9864	157	5	denoted	denote	VERB
ajst-9864	157	6	by	by	ADP
ajst-9864	157	7	a	a	DET
ajst-9864	157	8	one	one	NUM
ajst-9864	157	9	-	-	PUNCT
ajst-9864	157	10	dimensional	dimensional	ADJ
ajst-9864	157	11	binary	binary	NOUN
ajst-9864	157	12	array	array	NOUN
ajst-9864	157	13	,	,	PUNCT
ajst-9864	157	14	with	with	ADP
ajst-9864	157	15	'	'	NUM
ajst-9864	157	16	1	1	NUM
ajst-9864	157	17	'	'	PUNCT
ajst-9864	157	18	at	at	ADP
ajst-9864	157	19	the	the	DET
ajst-9864	157	20	index	index	NOUN
ajst-9864	157	21	corresponding	correspond	VERB
ajst-9864	157	22	to	to	ADP
ajst-9864	157	23	the	the	DET
ajst-9864	157	24	relationship	relationship	NOUN
ajst-9864	157	25	.	.	PUNCT
ajst-9864	158	1	this	this	DET
ajst-9864	158	2	efficient	efficient	ADJ
ajst-9864	158	3	representation	representation	NOUN
ajst-9864	158	4	aids	aid	VERB
ajst-9864	158	5	the	the	DET
ajst-9864	158	6	model	model	NOUN
ajst-9864	158	7	in	in	ADP
ajst-9864	158	8	identifying	identify	VERB
ajst-9864	158	9	and	and	CCONJ
ajst-9864	158	10	predicting	predict	VERB
ajst-9864	158	11	entity	entity	NOUN
ajst-9864	158	12	relationships	relationship	NOUN
ajst-9864	158	13	.	.	PUNCT
ajst-9864	159	1	4.2	4.2	NUM
ajst-9864	159	2	.	.	PUNCT
ajst-9864	159	3	related	relate	VERB
ajst-9864	159	4	experimental	experimental	ADJ
ajst-9864	159	5	configurations	configuration	NOUN
ajst-9864	159	6	4.2.1	4.2.1	NUM
ajst-9864	159	7	.	.	PUNCT
ajst-9864	160	1	software	software	NOUN
ajst-9864	160	2	experimental	experimental	ADJ
ajst-9864	160	3	environment	environment	NOUN
ajst-9864	160	4	1	1	NUM
ajst-9864	160	5	)	)	PUNCT
ajst-9864	160	6	development	development	NOUN
ajst-9864	160	7	platform	platform	NOUN
ajst-9864	160	8	:	:	PUNCT
ajst-9864	160	9	this	this	DET
ajst-9864	160	10	thesis	thesis	NOUN
ajst-9864	160	11	was	be	AUX
ajst-9864	160	12	conducted	conduct	VERB
ajst-9864	160	13	under	under	ADP
ajst-9864	160	14	an	an	DET
ajst-9864	160	15	ubuntu	ubuntu	ADJ
ajst-9864	160	16	system	system	NOUN
ajst-9864	160	17	;	;	PUNCT
ajst-9864	160	18	the	the	DET
ajst-9864	160	19	main	main	ADJ
ajst-9864	160	20	programming	programming	NOUN
ajst-9864	160	21	language	language	NOUN
ajst-9864	160	22	used	use	VERB
ajst-9864	160	23	for	for	ADP
ajst-9864	160	24	this	this	DET
ajst-9864	160	25	model	model	NOUN
ajst-9864	160	26	is	be	AUX
ajst-9864	160	27	python	python	NOUN
ajst-9864	160	28	3	3	NUM
ajst-9864	160	29	.	.	PUNCT
ajst-9864	160	30	version	version	NOUN
ajst-9864	160	31	6.12	6.12	NUM
ajst-9864	160	32	;	;	PUNCT
ajst-9864	160	33	all	all	DET
ajst-9864	160	34	development	development	NOUN
ajst-9864	160	35	tools	tool	NOUN
ajst-9864	160	36	are	be	AUX
ajst-9864	160	37	pycharm	pycharm	ADJ
ajst-9864	160	38	2020.1	2020.1	NUM
ajst-9864	160	39	;	;	PUNCT
ajst-9864	160	40	the	the	DET
ajst-9864	160	41	deep	deep	ADJ
ajst-9864	160	42	learning	learning	NOUN
ajst-9864	160	43	frameworks	framework	NOUN
ajst-9864	160	44	used	use	VERB
ajst-9864	160	45	are	be	AUX
ajst-9864	160	46	tensorflow	tensorflow	NOUN
ajst-9864	160	47	1.15	1.15	NUM
ajst-9864	160	48	and	and	CCONJ
ajst-9864	160	49	keras	keras	PROPN
ajst-9864	160	50	2.3.1	2.3.1	NUM
ajst-9864	160	51	;	;	PUNCT
ajst-9864	160	52	and	and	CCONJ
ajst-9864	160	53	the	the	DET
ajst-9864	160	54	crawler	crawler	NOUN
ajst-9864	160	55	framework	framework	NOUN
ajst-9864	160	56	used	use	VERB
ajst-9864	160	57	is	be	AUX
ajst-9864	160	58	scrapy	scrapy	NOUN
ajst-9864	160	59	2.1	2.1	NUM
ajst-9864	160	60	.	.	PUNCT
ajst-9864	161	1	the	the	DET
ajst-9864	161	2	data	data	NOUN
ajst-9864	161	3	analysis	analysis	NOUN
ajst-9864	161	4	frameworks	framework	NOUN
ajst-9864	161	5	used	use	VERB
ajst-9864	161	6	were	be	AUX
ajst-9864	161	7	numpy	numpy	NOUN
ajst-9864	161	8	1.16.2	1.16.2	NUM
ajst-9864	161	9	,	,	PUNCT
ajst-9864	161	10	pandas	panda	NOUN
ajst-9864	161	11	0.23.4	0.23.4	NUM
ajst-9864	161	12	,	,	PUNCT
ajst-9864	161	13	and	and	CCONJ
ajst-9864	161	14	matplotlib	matplotlib	PROPN
ajst-9864	161	15	2.2.4	2.2.4	NUM
ajst-9864	161	16	.	.	NOUN
ajst-9864	161	17	2	2	NUM
ajst-9864	161	18	)	)	PUNCT
ajst-9864	161	19	database	database	NOUN
ajst-9864	161	20	platform	platform	NOUN
ajst-9864	161	21	:	:	PUNCT
ajst-9864	161	22	the	the	DET
ajst-9864	161	23	databases	database	NOUN
ajst-9864	161	24	used	use	VERB
ajst-9864	161	25	in	in	ADP
ajst-9864	161	26	the	the	DET
ajst-9864	161	27	model	model	NOUN
ajst-9864	161	28	are	be	AUX
ajst-9864	161	29	redis	redis	NOUN
ajst-9864	161	30	and	and	CCONJ
ajst-9864	161	31	mongodb	mongodb	NOUN
ajst-9864	161	32	,	,	PUNCT
ajst-9864	161	33	where	where	SCONJ
ajst-9864	161	34	redis	redis	NOUN
ajst-9864	161	35	is	be	AUX
ajst-9864	161	36	used	use	VERB
ajst-9864	161	37	to	to	PART
ajst-9864	161	38	store	store	VERB
ajst-9864	161	39	the	the	DET
ajst-9864	161	40	character	character	NOUN
ajst-9864	161	41	relationship	relationship	NOUN
ajst-9864	161	42	pairs	pair	NOUN
ajst-9864	161	43	downloaded	download	VERB
ajst-9864	161	44	from	from	ADP
ajst-9864	161	45	the	the	DET
ajst-9864	161	46	interactive	interactive	ADJ
ajst-9864	161	47	encyclopaedia	encyclopaedia	NOUN
ajst-9864	161	48	and	and	CCONJ
ajst-9864	161	49	used	use	VERB
ajst-9864	161	50	when	when	SCONJ
ajst-9864	161	51	crawling	crawl	VERB
ajst-9864	161	52	text	text	NOUN
ajst-9864	161	53	.	.	PUNCT
ajst-9864	162	1	redis	redi	NOUN
ajst-9864	162	2	has	have	VERB
ajst-9864	162	3	certain	certain	ADJ
ajst-9864	162	4	persistence	persistence	NOUN
ajst-9864	162	5	layer	layer	NOUN
ajst-9864	162	6	functions	function	NOUN
ajst-9864	162	7	and	and	CCONJ
ajst-9864	162	8	can	can	AUX
ajst-9864	162	9	be	be	AUX
ajst-9864	162	10	used	use	VERB
ajst-9864	162	11	as	as	ADP
ajst-9864	162	12	a	a	DET
ajst-9864	162	13	caching	cache	VERB
ajst-9864	162	14	tool	tool	NOUN
ajst-9864	162	15	.	.	PUNCT
ajst-9864	163	1	the	the	DET
ajst-9864	163	2	mongodb	mongodb	NOUN
ajst-9864	163	3	database	database	NOUN
ajst-9864	163	4	is	be	AUX
ajst-9864	163	5	used	use	VERB
ajst-9864	163	6	to	to	PART
ajst-9864	163	7	store	store	VERB
ajst-9864	163	8	the	the	DET
ajst-9864	163	9	unprocessed	unprocessed	ADJ
ajst-9864	163	10	,	,	PUNCT
ajst-9864	163	11	unstructured	unstructured	ADJ
ajst-9864	163	12	text	text	NOUN
ajst-9864	163	13	crawled	crawl	VERB
ajst-9864	163	14	by	by	ADP
ajst-9864	163	15	the	the	DET
ajst-9864	163	16	crawler	crawler	NOUN
ajst-9864	163	17	and	and	CCONJ
ajst-9864	163	18	the	the	DET
ajst-9864	163	19	corpus	corpus	NOUN
ajst-9864	163	20	of	of	ADP
ajst-9864	163	21	relationships	relationship	NOUN
ajst-9864	163	22	constructed	construct	VERB
ajst-9864	163	23	later	later	ADV
ajst-9864	163	24	by	by	ADP
ajst-9864	163	25	manual	manual	ADJ
ajst-9864	163	26	cleaning	cleaning	NOUN
ajst-9864	163	27	and	and	CCONJ
ajst-9864	163	28	annotation	annotation	NOUN
ajst-9864	163	29	.	.	PUNCT
ajst-9864	164	1	4.2.2	4.2.2	X
ajst-9864	164	2	.	.	PUNCT
ajst-9864	164	3	hardware	hardware	NOUN
ajst-9864	164	4	support	support	NOUN
ajst-9864	164	5	environment	environment	NOUN
ajst-9864	164	6	a	a	DET
ajst-9864	164	7	model	model	NOUN
ajst-9864	164	8	writing	write	VERB
ajst-9864	164	9	environment	environment	NOUN
ajst-9864	164	10	:	:	PUNCT
ajst-9864	164	11	macbook	macbook	NOUN
ajst-9864	164	12	pro	pro	X
ajst-9864	164	13	(	(	PUNCT
ajst-9864	164	14	13	13	NUM
ajst-9864	164	15	-	-	PUNCT
ajst-9864	164	16	inch	inch	NOUN
ajst-9864	164	17	,	,	PUNCT
ajst-9864	164	18	2018	2018	NUM
ajst-9864	164	19	;	;	PUNCT
ajst-9864	164	20	four	four	NUM
ajst-9864	164	21	thunderbolt	thunderbolt	NOUN
ajst-9864	164	22	3	3	NUM
ajst-9864	164	23	ports	port	NOUN
ajst-9864	164	24	)	)	PUNCT
ajst-9864	164	25	operating	operating	NOUN
ajst-9864	164	26	system	system	NOUN
ajst-9864	164	27	:	:	PUNCT
ajst-9864	164	28	maccatalina	maccatalina	NOUN
ajst-9864	164	29	10.15	10.15	NUM
ajst-9864	164	30	processor	processor	NOUN
ajst-9864	164	31	:	:	PUNCT
ajst-9864	164	32	2.3	2.3	NUM
ajst-9864	164	33	ghz	ghz	NOUN
ajst-9864	164	34	p4	p4	ADJ
ajst-9864	164	35	t	t	PROPN
ajst-9864	164	36	intel	intel	PROPN
ajst-9864	164	37	core	core	PROPN
ajst-9864	164	38	i5	i5	PROPN
ajst-9864	164	39	memory	memory	NOUN
ajst-9864	164	40	:	:	PUNCT
ajst-9864	164	41	8	8	NUM
ajst-9864	164	42	gb	gb	NOUN
ajst-9864	164	43	at	at	ADP
ajst-9864	164	44	2133	2133	NUM
ajst-9864	164	45	mhz	mhz	NOUN
ajst-9864	164	46	lpddr3	lpddr3	NOUN
ajst-9864	164	47	boot	boot	NOUN
ajst-9864	164	48	disk	disk	NOUN
ajst-9864	164	49	:	:	PUNCT
ajst-9864	164	50	macintosh	macintosh	PROPN
ajst-9864	164	51	hd	hd	PROPN
ajst-9864	164	52	256	256	NUM
ajst-9864	164	53	g	g	PROPN
ajst-9864	164	54	graphics	graphic	NOUN
ajst-9864	164	55	card	card	NOUN
ajst-9864	164	56	:	:	PUNCT
ajst-9864	164	57	intel	intel	PROPN
ajst-9864	164	58	iris	iris	NOUN
ajst-9864	164	59	plus	plus	CCONJ
ajst-9864	164	60	graphics	graphic	NOUN
ajst-9864	164	61	655/1536	655/1536	NUM
ajst-9864	164	62	mb	mb	ADP
ajst-9864	164	63	model	model	NOUN
ajst-9864	164	64	training	training	NOUN
ajst-9864	164	65	and	and	CCONJ
ajst-9864	164	66	testing	testing	NOUN
ajst-9864	164	67	environment	environment	NOUN
ajst-9864	164	68	:	:	PUNCT
ajst-9864	164	69	operating	operate	VERB
ajst-9864	164	70	system	system	NOUN
ajst-9864	164	71	:	:	PUNCT
ajst-9864	164	72	ubuntu	ubuntu	NOUN
ajst-9864	164	73	18.04.2	18.04.2	NUM
ajst-9864	164	74	cpu	cpu	NOUN
ajst-9864	164	75	:	:	PUNCT
ajst-9864	164	76	intel(r	intel(r	PROPN
ajst-9864	164	77	)	)	PUNCT
ajst-9864	164	78	core(tm	core(tm	NOUN
ajst-9864	164	79	)	)	PUNCT
ajst-9864	164	80	i5	i5	NOUN
ajst-9864	164	81	-	-	PUNCT
ajst-9864	164	82	10400f	10400f	NUM
ajst-9864	164	83	@	@	ADP
ajst-9864	164	84	2.90	2.90	NUM
ajst-9864	164	85	ghz	ghz	NOUN
ajst-9864	164	86	memory	memory	NOUN
ajst-9864	164	87	:	:	PUNCT
ajst-9864	164	88	kingston	kingston	PROPN
ajst-9864	164	89	ddr4	ddr4	PROPN
ajst-9864	164	90	-	-	PUNCT
ajst-9864	164	91	2666	2666	NUM
ajst-9864	164	92	8gbx4	8gbx4	NUM
ajst-9864	164	93	graphics	graphic	NOUN
ajst-9864	164	94	:	:	PUNCT
ajst-9864	164	95	nvidia	nvidia	PROPN
ajst-9864	164	96	geforce	geforce	NOUN
ajst-9864	164	97	rtx	rtx	PROPN
ajst-9864	164	98	2080	2080	NUM
ajst-9864	164	99	ti	ti	NOUN
ajst-9864	164	100	11	11	NUM
ajst-9864	164	101	g	g	PROPN
ajst-9864	164	102	4.2.3	4.2.3	NUM
ajst-9864	164	103	.	.	PUNCT
ajst-9864	165	1	experimental	experimental	ADJ
ajst-9864	165	2	flow	flow	NOUN
ajst-9864	165	3	design	design	NOUN
ajst-9864	165	4	this	this	DET
ajst-9864	165	5	study	study	NOUN
ajst-9864	165	6	verifies	verify	VERB
ajst-9864	165	7	the	the	DET
ajst-9864	165	8	effectiveness	effectiveness	NOUN
ajst-9864	165	9	of	of	ADP
ajst-9864	165	10	our	our	PRON
ajst-9864	165	11	proposed	propose	VERB
ajst-9864	165	12	gru	gru	NOUN
ajst-9864	165	13	dual	dual	ADV
ajst-9864	165	14	-	-	PUNCT
ajst-9864	165	15	supervised	supervise	VERB
ajst-9864	165	16	model	model	NOUN
ajst-9864	165	17	through	through	ADP
ajst-9864	165	18	a	a	DET
ajst-9864	165	19	specific	specific	ADJ
ajst-9864	165	20	experiment	experiment	NOUN
ajst-9864	165	21	.	.	PUNCT
ajst-9864	166	1	the	the	DET
ajst-9864	166	2	specific	specific	ADJ
ajst-9864	166	3	experimental	experimental	ADJ
ajst-9864	166	4	steps	step	NOUN
ajst-9864	166	5	are	be	AUX
ajst-9864	166	6	as	as	SCONJ
ajst-9864	166	7	follows	follow	VERB
ajst-9864	166	8	:	:	PUNCT
ajst-9864	166	9	s1	s1	NOUN
ajst-9864	166	10	:	:	PUNCT
ajst-9864	166	11	manually	manually	ADV
ajst-9864	166	12	annotate	annotate	VERB
ajst-9864	166	13	entity	entity	NOUN
ajst-9864	166	14	and	and	CCONJ
ajst-9864	166	15	relationship	relationship	NOUN
ajst-9864	166	16	data	datum	NOUN
ajst-9864	166	17	.	.	PUNCT
ajst-9864	167	1	s2	s2	PROPN
ajst-9864	167	2	:	:	PUNCT
ajst-9864	167	3	preprocess	preprocess	NOUN
ajst-9864	167	4	the	the	DET
ajst-9864	167	5	annotated	annotate	VERB
ajst-9864	167	6	data	datum	NOUN
ajst-9864	167	7	to	to	PART
ajst-9864	167	8	generate	generate	VERB
ajst-9864	167	9	training	training	NOUN
ajst-9864	167	10	and	and	CCONJ
ajst-9864	167	11	testing	testing	NOUN
ajst-9864	167	12	sets	set	NOUN
ajst-9864	167	13	for	for	ADP
ajst-9864	167	14	the	the	DET
ajst-9864	167	15	entity	entity	NOUN
ajst-9864	167	16	extraction	extraction	NOUN
ajst-9864	167	17	model	model	NOUN
ajst-9864	167	18	and	and	CCONJ
ajst-9864	167	19	the	the	DET
ajst-9864	167	20	relationship	relationship	NOUN
ajst-9864	167	21	extraction	extraction	NOUN
ajst-9864	167	22	model	model	NOUN
ajst-9864	167	23	:	:	PUNCT
ajst-9864	167	24	convert	convert	VERB
ajst-9864	167	25	entity	entity	NOUN
ajst-9864	167	26	annotation	annotation	NOUN
ajst-9864	167	27	data	datum	NOUN
ajst-9864	167	28	to	to	ADP
ajst-9864	167	29	the	the	DET
ajst-9864	167	30	bmes	bme	NOUN
ajst-9864	167	31	entity	entity	NOUN
ajst-9864	167	32	annotation	annotation	NOUN
ajst-9864	167	33	system	system	NOUN
ajst-9864	167	34	,	,	PUNCT
ajst-9864	167	35	where	where	SCONJ
ajst-9864	167	36	b	b	NOUN
ajst-9864	167	37	stands	stand	VERB
ajst-9864	167	38	for	for	ADP
ajst-9864	167	39	the	the	DET
ajst-9864	167	40	entity	entity	NOUN
ajst-9864	167	41	's	's	PART
ajst-9864	167	42	beginning	beginning	NOUN
ajst-9864	167	43	position	position	NOUN
ajst-9864	167	44	,	,	PUNCT
ajst-9864	167	45	m	m	VERB
ajst-9864	167	46	for	for	ADP
ajst-9864	167	47	its	its	PRON
ajst-9864	167	48	middle	middle	ADJ
ajst-9864	167	49	section	section	NOUN
ajst-9864	167	50	,	,	PUNCT
ajst-9864	167	51	e	e	NOUN
ajst-9864	167	52	for	for	ADP
ajst-9864	167	53	its	its	PRON
ajst-9864	167	54	conclusion	conclusion	NOUN
ajst-9864	167	55	,	,	PUNCT
ajst-9864	167	56	and	and	CCONJ
ajst-9864	167	57	s	s	VERB
ajst-9864	167	58	for	for	ADP
ajst-9864	167	59	a	a	DET
ajst-9864	167	60	single	single	ADJ
ajst-9864	167	61	-	-	PUNCT
ajst-9864	167	62	character	character	NOUN
ajst-9864	167	63	entity	entity	NOUN
ajst-9864	167	64	.	.	PUNCT
ajst-9864	168	1	convert	convert	VERB
ajst-9864	168	2	relationship	relationship	NOUN
ajst-9864	168	3	extraction	extraction	NOUN
ajst-9864	168	4	data	datum	NOUN
ajst-9864	168	5	to	to	ADP
ajst-9864	168	6	the	the	DET
ajst-9864	168	7	format	format	NOUN
ajst-9864	168	8	of	of	ADP
ajst-9864	168	9	"	"	PUNCT
ajst-9864	168	10	entity1	entity1	ADV
ajst-9864	168	11	,	,	PUNCT
ajst-9864	168	12	entity2	entity2	PROPN
ajst-9864	168	13	,	,	PUNCT
ajst-9864	168	14	"	"	PUNCT
ajst-9864	168	15	"	"	PUNCT
ajst-9864	168	16	entity1	entity1	ADV
ajst-9864	168	17	start	start	VERB
ajst-9864	168	18	position	position	NOUN
ajst-9864	168	19	,	,	PUNCT
ajst-9864	168	20	entity1	entity1	ADV
ajst-9864	168	21	end	end	NOUN
ajst-9864	168	22	position	position	NOUN
ajst-9864	168	23	,	,	PUNCT
ajst-9864	168	24	entity1	entity1	ADV
ajst-9864	168	25	label	label	NOUN
ajst-9864	168	26	,	,	PUNCT
ajst-9864	168	27	"	"	PUNCT
ajst-9864	168	28	"	"	PUNCT
ajst-9864	168	29	entity2	entity2	NOUN
ajst-9864	168	30	start	start	VERB
ajst-9864	168	31	position	position	NOUN
ajst-9864	168	32	,	,	PUNCT
ajst-9864	168	33	entity2	entity2	NOUN
ajst-9864	168	34	end	end	NOUN
ajst-9864	168	35	position	position	NOUN
ajst-9864	168	36	,	,	PUNCT
ajst-9864	168	37	entity2	entity2	NOUN
ajst-9864	168	38	label	label	NOUN
ajst-9864	168	39	,	,	PUNCT
ajst-9864	168	40	text	text	NOUN
ajst-9864	168	41	paragraphd	paragraphd	NOUN
ajst-9864	168	42	s	s	PART
ajst-9864	168	43	represents	represent	VERB
ajst-9864	168	44	a	a	DET
ajst-9864	168	45	single	single	ADJ
ajst-9864	168	46	-	-	PUNCT
ajst-9864	168	47	character	character	NOUN
ajst-9864	168	48	entity	entity	NOUN
ajst-9864	168	49	.	.	PUNCT
ajst-9864	169	1	convert	convert	VERB
ajst-9864	169	2	relationship	relationship	NOUN
ajst-9864	169	3	extraction	extraction	NOUN
ajst-9864	169	4	data	datum	NOUN
ajst-9864	169	5	to	to	ADP
ajst-9864	169	6	the	the	DET
ajst-9864	169	7	format	format	NOUN
ajst-9864	169	8	of	of	ADP
ajst-9864	169	9	"	"	PUNCT
ajst-9864	169	10	entity1	entity1	ADV
ajst-9864	169	11	,	,	PUNCT
ajst-9864	169	12	entity2	entity2	PROPN
ajst-9864	169	13	,	,	PUNCT
ajst-9864	169	14	"	"	PUNCT
ajst-9864	169	15	"	"	PUNCT
ajst-9864	169	16	entity1	entity1	ADV
ajst-9864	169	17	start	start	VERB
ajst-9864	169	18	position	position	NOUN
ajst-9864	169	19	,	,	PUNCT
ajst-9864	169	20	entity1	entity1	ADV
ajst-9864	169	21	end	end	NOUN
ajst-9864	169	22	position	position	NOUN
ajst-9864	169	23	,	,	PUNCT
ajst-9864	169	24	entity1	entity1	ADV
ajst-9864	169	25	label	label	NOUN
ajst-9864	169	26	,	,	PUNCT
ajst-9864	169	27	"	"	PUNCT
ajst-9864	169	28	"	"	PUNCT
ajst-9864	169	29	entity2	entity2	NOUN
ajst-9864	169	30	start	start	VERB
ajst-9864	169	31	position	position	NOUN
ajst-9864	169	32	,	,	PUNCT
ajst-9864	169	33	entity2	entity2	NOUN
ajst-9864	169	34	end	end	NOUN
ajst-9864	169	35	position	position	NOUN
ajst-9864	169	36	,	,	PUNCT
ajst-9864	169	37	entity2	entity2	NOUN
ajst-9864	169	38	label	label	NOUN
ajst-9864	169	39	,	,	PUNCT
ajst-9864	169	40	text	text	NOUN
ajst-9864	169	41	paragraph	paragraph	NOUN
ajst-9864	169	42	.	.	PUNCT
ajst-9864	169	43	"	"	PUNCT
ajst-9864	170	1	s3	s3	PROPN
ajst-9864	170	2	:	:	PUNCT
ajst-9864	170	3	construct	construct	VERB
ajst-9864	170	4	the	the	DET
ajst-9864	170	5	bigru-2att	bigru-2att	PROPN
ajst-9864	170	6	relationship	relationship	NOUN
ajst-9864	170	7	extraction	extraction	NOUN
ajst-9864	170	8	network	network	NOUN
ajst-9864	170	9	,	,	PUNCT
ajst-9864	170	10	with	with	ADP
ajst-9864	170	11	specific	specific	ADJ
ajst-9864	170	12	steps	step	NOUN
ajst-9864	170	13	as	as	SCONJ
ajst-9864	170	14	follows	follow	VERB
ajst-9864	170	15	:	:	PUNCT
ajst-9864	170	16	s3	s3	PROPN
ajst-9864	170	17	-	-	PUNCT
ajst-9864	170	18	1	1	NUM
ajst-9864	170	19	:	:	PUNCT
ajst-9864	170	20	expand	expand	VERB
ajst-9864	170	21	the	the	DET
ajst-9864	170	22	vector	vector	NOUN
ajst-9864	170	23	feature	feature	NOUN
ajst-9864	170	24	of	of	ADP
ajst-9864	170	25	entity	entity	NOUN
ajst-9864	170	26	position	position	NOUN
ajst-9864	170	27	information	information	NOUN
ajst-9864	170	28	(	(	PUNCT
ajst-9864	170	29	including	include	VERB
ajst-9864	170	30	start	start	NOUN
ajst-9864	170	31	position	position	NOUN
ajst-9864	170	32	and	and	CCONJ
ajst-9864	170	33	end	end	NOUN
ajst-9864	170	34	position	position	NOUN
ajst-9864	170	35	)	)	PUNCT
ajst-9864	170	36	and	and	CCONJ
ajst-9864	170	37	entity	entity	NOUN
ajst-9864	170	38	label	label	NOUN
ajst-9864	170	39	information	information	NOUN
ajst-9864	170	40	to	to	PART
ajst-9864	170	41	vectorize	vectorize	VERB
ajst-9864	170	42	the	the	DET
ajst-9864	170	43	text	text	NOUN
ajst-9864	170	44	information	information	NOUN
ajst-9864	170	45	as	as	ADP
ajst-9864	170	46	model	model	NOUN
ajst-9864	170	47	input	input	NOUN
ajst-9864	170	48	.	.	PUNCT
ajst-9864	171	1	s3	s3	PROPN
ajst-9864	171	2	-	-	PUNCT
ajst-9864	171	3	2	2	NUM
ajst-9864	171	4	:	:	PUNCT
ajst-9864	171	5	the	the	DET
ajst-9864	171	6	first	first	ADJ
ajst-9864	171	7	layer	layer	NOUN
ajst-9864	171	8	of	of	ADP
ajst-9864	171	9	the	the	DET
ajst-9864	171	10	model	model	NOUN
ajst-9864	171	11	network	network	NOUN
ajst-9864	171	12	is	be	AUX
ajst-9864	171	13	a	a	DET
ajst-9864	171	14	bidirectional	bidirectional	ADJ
ajst-9864	171	15	gru	gru	PROPN
ajst-9864	171	16	.	.	PROPN
ajst-9864	171	17	figure	figure	NOUN
ajst-9864	171	18	3	3	NUM
ajst-9864	171	19	gru	gru	NOUN
ajst-9864	171	20	structure	structure	NOUN
ajst-9864	171	21	as	as	SCONJ
ajst-9864	171	22	can	can	AUX
ajst-9864	171	23	be	be	AUX
ajst-9864	171	24	seen	see	VERB
ajst-9864	171	25	in	in	ADP
ajst-9864	171	26	figure	figure	NOUN
ajst-9864	171	27	3	3	NUM
ajst-9864	171	28	,	,	PUNCT
ajst-9864	171	29	each	each	DET
ajst-9864	171	30	gru	gru	NOUN
ajst-9864	171	31	unit	unit	NOUN
ajst-9864	171	32	has	have	VERB
ajst-9864	171	33	both	both	CCONJ
ajst-9864	171	34	a	a	DET
ajst-9864	171	35	reset	reset	NOUN
ajst-9864	171	36	gate	gate	NOUN
ajst-9864	171	37	and	and	CCONJ
ajst-9864	171	38	an	an	DET
ajst-9864	171	39	update	update	NOUN
ajst-9864	171	40	gate	gate	NOUN
ajst-9864	171	41	.	.	PUNCT
ajst-9864	172	1	the	the	DET
ajst-9864	172	2	output	output	NOUN
ajst-9864	172	3	ht	ht	PROPN
ajst-9864	172	4	of	of	ADP
ajst-9864	172	5	the	the	DET
ajst-9864	172	6	gating	gate	VERB
ajst-9864	172	7	unit	unit	NOUN
ajst-9864	172	8	at	at	ADP
ajst-9864	172	9	time	time	NOUN
ajst-9864	172	10	t	t	PROPN
ajst-9864	172	11	is	be	AUX
ajst-9864	172	12	determined	determine	VERB
ajst-9864	172	13	by	by	ADP
ajst-9864	172	14	combining	combine	VERB
ajst-9864	172	15	the	the	DET
ajst-9864	172	16	information	information	NOUN
ajst-9864	172	17	from	from	ADP
ajst-9864	172	18	the	the	DET
ajst-9864	172	19	output	output	NOUN
ajst-9864	172	20	ht-1	ht-1	PUNCT
ajst-9864	172	21	of	of	ADP
ajst-9864	172	22	the	the	DET
ajst-9864	172	23	previous	previous	ADJ
ajst-9864	172	24	time	time	NOUN
ajst-9864	172	25	step	step	NOUN
ajst-9864	172	26	with	with	ADP
ajst-9864	172	27	the	the	DET
ajst-9864	172	28	information	information	NOUN
ajst-9864	172	29	from	from	ADP
ajst-9864	172	30	the	the	DET
ajst-9864	172	31	input	input	NOUN
ajst-9864	172	32	xt	xt	ADP
ajst-9864	172	33	of	of	ADP
ajst-9864	172	34	the	the	DET
ajst-9864	172	35	current	current	ADJ
ajst-9864	172	36	time	time	NOUN
ajst-9864	172	37	step	step	NOUN
ajst-9864	172	38	.	.	PUNCT
ajst-9864	173	1	there	there	PRON
ajst-9864	173	2	will	will	AUX
ajst-9864	173	3	be	be	AUX
ajst-9864	173	4	a	a	DET
ajst-9864	173	5	greater	great	ADJ
ajst-9864	173	6	degree	degree	NOUN
ajst-9864	173	7	of	of	ADP
ajst-9864	173	8	retention	retention	NOUN
ajst-9864	173	9	the	the	PRON
ajst-9864	173	10	higher	high	ADJ
ajst-9864	173	11	the	the	DET
ajst-9864	173	12	value.the	value.the	X
ajst-9864	173	13	reset	reset	NOUN
ajst-9864	173	14	gate	gate	NOUN
ajst-9864	173	15	rt	rt	PROPN
ajst-9864	173	16	determines	determine	VERB
ajst-9864	173	17	the	the	DET
ajst-9864	173	18	degree	degree	NOUN
ajst-9864	173	19	of	of	ADP
ajst-9864	173	20	forgetting	forget	VERB
ajst-9864	173	21	of	of	ADP
ajst-9864	173	22	information	information	NOUN
ajst-9864	173	23	in	in	ADP
ajst-9864	173	24	ht-1	ht-1	PUNCT
ajst-9864	173	25	based	base	VERB
ajst-9864	173	26	on	on	ADP
ajst-9864	173	27	xt	xt	PROPN
ajst-9864	173	28	,	,	PUNCT
ajst-9864	173	29	with	with	ADP
ajst-9864	173	30	smaller	small	ADJ
ajst-9864	173	31	values	value	NOUN
ajst-9864	173	32	of	of	ADP
ajst-9864	173	33	rt	rt	PROPN
ajst-9864	173	34	indicating	indicate	VERB
ajst-9864	173	35	a	a	DET
ajst-9864	173	36	higher	high	ADJ
ajst-9864	173	37	degree	degree	NOUN
ajst-9864	173	38	of	of	ADP
ajst-9864	173	39	ignoring	ignore	VERB
ajst-9864	173	40	.	.	PUNCT
ajst-9864	174	1	the	the	DET
ajst-9864	174	2	current	current	ADJ
ajst-9864	174	3	time	time	NOUN
ajst-9864	174	4	step	step	NOUN
ajst-9864	174	5	's	's	PART
ajst-9864	174	6	memory	memory	NOUN
ajst-9864	174	7	and	and	CCONJ
ajst-9864	174	8	hidden	hidden	ADJ
ajst-9864	174	9	state	state	NOUN
ajst-9864	174	10	ht	ht	PROPN
ajst-9864	174	11	after	after	ADP
ajst-9864	174	12	passing	pass	VERB
ajst-9864	174	13	through	through	ADP
ajst-9864	174	14	the	the	DET
ajst-9864	174	15	reset	reset	NOUN
ajst-9864	174	16	and	and	CCONJ
ajst-9864	174	17	update	update	NOUN
ajst-9864	174	18	gates	gate	NOUN
ajst-9864	174	19	are	be	AUX
ajst-9864	174	20	calculated	calculate	VERB
ajst-9864	174	21	.	.	PUNCT
ajst-9864	175	1	the	the	DET
ajst-9864	175	2	update	update	PROPN
ajst-9864	175	3	gate	gate	PROPN
ajst-9864	175	4	zt	zt	PROPN
ajst-9864	175	5	,	,	PUNCT
ajst-9864	175	6	reset	reset	NOUN
ajst-9864	175	7	gate	gate	PROPN
ajst-9864	175	8	rt	rt	PROPN
ajst-9864	175	9	,	,	PUNCT
ajst-9864	175	10	new	new	ADJ
ajst-9864	175	11	memory	memory	NOUN
ajst-9864	175	12	𝐡	𝐡	NOUN
ajst-9864	175	13	and	and	CCONJ
ajst-9864	175	14	final	final	ADJ
ajst-9864	175	15	hidden	hidden	ADJ
ajst-9864	175	16	state	state	NOUN
ajst-9864	175	17	htfor	htfor	ADP
ajst-9864	175	18	the	the	DET
ajst-9864	175	19	t	t	PROPN
ajst-9864	175	20	-	-	PUNCT
ajst-9864	175	21	th	th	PROPN
ajst-9864	175	22	gru	gru	NOUN
ajst-9864	175	23	unit	unit	NOUN
ajst-9864	175	24	are	be	AUX
ajst-9864	175	25	calculated	calculate	VERB
ajst-9864	175	26	as	as	SCONJ
ajst-9864	175	27	follows	follow	VERB
ajst-9864	175	28	:	:	PUNCT
ajst-9864	175	29	zt＝σ(wz·[ht-1	zt＝σ(wz·[ht-1	NUM
ajst-9864	175	30	,	,	PUNCT
ajst-9864	175	31	xt	xt	ADP
ajst-9864	175	32	]	]	X
ajst-9864	175	33	)	)	PUNCT
ajst-9864	175	34	(	(	PUNCT
ajst-9864	175	35	16	16	NUM
ajst-9864	175	36	)	)	PUNCT
ajst-9864	175	37	rt＝σ(wr·[ht-1	rt＝σ(wr·[ht-1	PUNCT
ajst-9864	175	38	,	,	PUNCT
ajst-9864	175	39	xt	xt	X
ajst-9864	175	40	]	]	X
ajst-9864	175	41	)	)	PUNCT
ajst-9864	175	42	(	(	PUNCT
ajst-9864	175	43	17	17	NUM
ajst-9864	175	44	)	)	PUNCT
ajst-9864	175	45	h̃	h̃	PROPN
ajst-9864	175	46	tanh	tanh	PROPN
ajst-9864	175	47	w	w	PROPN
ajst-9864	175	48	⋅	⋅	PROPN
ajst-9864	175	49	h	h	PROPN
ajst-9864	175	50	∗	∗	NOUN
ajst-9864	175	51	r	r	NOUN
ajst-9864	175	52	,	,	PUNCT
ajst-9864	175	53	x	x	X
ajst-9864	175	54	(	(	PUNCT
ajst-9864	175	55	18	18	NUM
ajst-9864	175	56	)	)	PUNCT
ajst-9864	175	57	h	h	NOUN
ajst-9864	175	58	1	1	NUM
ajst-9864	175	59	z	z	NOUN
ajst-9864	175	60	∗	∗	NOUN
ajst-9864	175	61	h̃	h̃	PROPN
ajst-9864	175	62	z	z	PROPN
ajst-9864	175	63	∗	∗	NOUN
ajst-9864	175	64	h	h	NOUN
ajst-9864	175	65	(	(	PUNCT
ajst-9864	175	66	19	19	NUM
ajst-9864	175	67	)	)	PUNCT
ajst-9864	175	68	where	where	SCONJ
ajst-9864	175	69	σ	σ	PROPN
ajst-9864	175	70	(	(	PUNCT
ajst-9864	175	71	)	)	PUNCT
ajst-9864	175	72	is	be	AUX
ajst-9864	175	73	a	a	DET
ajst-9864	175	74	sigmoid	sigmoid	NOUN
ajst-9864	175	75	nonlinear	nonlinear	ADJ
ajst-9864	175	76	activation	activation	NOUN
ajst-9864	175	77	function	function	VERB
ajst-9864	175	78	to	to	PART
ajst-9864	175	79	enhance	enhance	VERB
ajst-9864	175	80	the	the	DET
ajst-9864	175	81	model	model	NOUN
ajst-9864	175	82	's	's	PART
ajst-9864	175	83	ability	ability	NOUN
ajst-9864	175	84	to	to	PART
ajst-9864	175	85	handle	handle	VERB
ajst-9864	175	86	nonlinear	nonlinear	ADJ
ajst-9864	175	87	data	datum	NOUN
ajst-9864	175	88	.	.	PUNCT
ajst-9864	176	1	x	x	X
ajst-9864	176	2	1/	1/	NUM
ajst-9864	176	3	1	1	NUM
ajst-9864	176	4	e	e	NOUN
ajst-9864	176	5	x	x	PROPN
ajst-9864	176	6	。	。	NOUN
ajst-9864	176	7	*	*	AUX
ajst-9864	176	8	denotes	denote	NOUN
ajst-9864	176	9	dot	dot	VERB
ajst-9864	176	10	product.tanh	product.tanh	NOUN
ajst-9864	176	11	x	x	PUNCT
ajst-9864	176	12	e	e	NOUN
ajst-9864	176	13	e	e	X
ajst-9864	176	14	/	/	SYM
ajst-9864	176	15	e	e	X
ajst-9864	176	16	e	e	X
ajst-9864	176	17	。	。	PROPN
ajst-9864	176	18	w	w	PROPN
ajst-9864	176	19	,	,	PUNCT
ajst-9864	176	20	wr	wr	PROPN
ajst-9864	176	21	,	,	PUNCT
ajst-9864	176	22	wz	wz	PROPN
ajst-9864	176	23	are	be	AUX
ajst-9864	176	24	the	the	DET
ajst-9864	176	25	weight	weight	NOUN
ajst-9864	176	26	matrices	matrix	NOUN
ajst-9864	176	27	of	of	ADP
ajst-9864	176	28	the	the	DET
ajst-9864	176	29	model	model	NOUN
ajst-9864	176	30	.	.	PUNCT
ajst-9864	177	1	[	[	PUNCT
ajst-9864	177	2	]	]	X
ajst-9864	177	3	denotes	denote	NOUN
ajst-9864	177	4	joining	join	VERB
ajst-9864	177	5	the	the	DET
ajst-9864	177	6	two	two	NUM
ajst-9864	177	7	vectors	vector	NOUN
ajst-9864	177	8	.	.	PUNCT
ajst-9864	178	1	s3	s3	PROPN
ajst-9864	178	2	-	-	PUNCT
ajst-9864	178	3	3	3	NUM
ajst-9864	178	4	:	:	PUNCT
ajst-9864	178	5	the	the	DET
ajst-9864	178	6	second	second	ADJ
ajst-9864	178	7	layer	layer	NOUN
ajst-9864	178	8	of	of	ADP
ajst-9864	178	9	the	the	DET
ajst-9864	178	10	model	model	NOUN
ajst-9864	178	11	network	network	NOUN
ajst-9864	178	12	is	be	AUX
ajst-9864	178	13	the	the	DET
ajst-9864	178	14	wordlevel	wordlevel	NOUN
ajst-9864	178	15	attention	attention	NOUN
ajst-9864	178	16	layer	layer	NOUN
ajst-9864	178	17	.	.	PUNCT
ajst-9864	179	1	for	for	ADP
ajst-9864	179	2	a	a	DET
ajst-9864	179	3	sentence	sentence	NOUN
ajst-9864	179	4	vector	vector	NOUN
ajst-9864	179	5	w	w	NOUN
ajst-9864	179	6	=	=	SYM
ajst-9864	179	7	{	{	PUNCT
ajst-9864	179	8	w1	w1	NOUN
ajst-9864	179	9	,	,	PUNCT
ajst-9864	179	10	w2	w2	NOUN
ajst-9864	179	11	,	,	PUNCT
ajst-9864	179	12	...	...	PUNCT
ajst-9864	179	13	,	,	PUNCT
ajst-9864	179	14	wt	wt	ADP
ajst-9864	179	15	}	}	PUNCT
ajst-9864	179	16	take	take	VERB
ajst-9864	179	17	the	the	DET
ajst-9864	179	18	result	result	NOUN
ajst-9864	179	19	ht	ht	PROPN
ajst-9864	179	20	obtained	obtain	VERB
ajst-9864	179	21	in	in	ADP
ajst-9864	179	22	step	step	NOUN
ajst-9864	179	23	s3	s3	PROPN
ajst-9864	179	24	-	-	PUNCT
ajst-9864	179	25	2	2	NUM
ajst-9864	179	26	and	and	CCONJ
ajst-9864	179	27	process	process	VERB
ajst-9864	179	28	it	it	PRON
ajst-9864	179	29	by	by	ADP
ajst-9864	179	30	the	the	DET
ajst-9864	179	31	following	follow	VERB
ajst-9864	179	32	equation	equation	NOUN
ajst-9864	179	33	to	to	PART
ajst-9864	179	34	obtain	obtain	VERB
ajst-9864	179	35	ut	ut	PROPN
ajst-9864	179	36	.	.	PROPN
ajst-9864	179	37	134	134	NUM
ajst-9864	179	38	ut＝tanh(ww·ht+bw	ut＝tanh(ww·ht+bw	NUM
ajst-9864	179	39	)	)	PUNCT
ajst-9864	179	40	(	(	PUNCT
ajst-9864	179	41	20	20	NUM
ajst-9864	179	42	)	)	PUNCT
ajst-9864	179	43	in	in	ADP
ajst-9864	179	44	fact	fact	NOUN
ajst-9864	179	45	,	,	PUNCT
ajst-9864	179	46	each	each	DET
ajst-9864	179	47	word	word	NOUN
ajst-9864	179	48	in	in	ADP
ajst-9864	179	49	the	the	DET
ajst-9864	179	50	sentence	sentence	NOUN
ajst-9864	179	51	produces	produce	VERB
ajst-9864	179	52	an	an	DET
ajst-9864	179	53	unequal	unequal	ADJ
ajst-9864	179	54	effect	effect	NOUN
ajst-9864	179	55	on	on	ADP
ajst-9864	179	56	the	the	DET
ajst-9864	179	57	expression	expression	NOUN
ajst-9864	179	58	of	of	ADP
ajst-9864	179	59	the	the	DET
ajst-9864	179	60	sentence	sentence	NOUN
ajst-9864	179	61	's	's	PART
ajst-9864	179	62	meaning	meaning	NOUN
ajst-9864	179	63	.	.	PUNCT
ajst-9864	180	1	a	a	DET
ajst-9864	180	2	randomly	randomly	ADV
ajst-9864	180	3	initialised	initialise	VERB
ajst-9864	180	4	word	word	NOUN
ajst-9864	180	5	context	context	PROPN
ajst-9864	180	6	vector	vector	NOUN
ajst-9864	180	7	uw	uw	PROPN
ajst-9864	180	8	is	be	AUX
ajst-9864	180	9	added	add	VERB
ajst-9864	180	10	to	to	ADP
ajst-9864	180	11	the	the	DET
ajst-9864	180	12	word	word	NOUN
ajst-9864	180	13	-	-	PUNCT
ajst-9864	180	14	level	level	NOUN
ajst-9864	180	15	attention	attention	NOUN
ajst-9864	180	16	layer	layer	NOUN
ajst-9864	180	17	training	training	NOUN
ajst-9864	180	18	process	process	NOUN
ajst-9864	180	19	for	for	ADP
ajst-9864	180	20	co	co	NOUN
ajst-9864	180	21	-	-	NOUN
ajst-9864	180	22	training	training	NOUN
ajst-9864	180	23	.	.	PUNCT
ajst-9864	181	1	the	the	DET
ajst-9864	181	2	degree	degree	NOUN
ajst-9864	181	3	of	of	ADP
ajst-9864	181	4	connection	connection	NOUN
ajst-9864	181	5	between	between	ADP
ajst-9864	181	6	the	the	DET
ajst-9864	181	7	word	word	NOUN
ajst-9864	181	8	and	and	CCONJ
ajst-9864	181	9	the	the	DET
ajst-9864	181	10	relation	relation	NOUN
ajst-9864	181	11	is	be	AUX
ajst-9864	181	12	computed	compute	VERB
ajst-9864	181	13	by	by	ADP
ajst-9864	181	14	adding	add	VERB
ajst-9864	181	15	a	a	DET
ajst-9864	181	16	word	word	NOUN
ajst-9864	181	17	-	-	PUNCT
ajst-9864	181	18	level	level	NOUN
ajst-9864	181	19	attention	attention	NOUN
ajst-9864	181	20	layer	layer	NOUN
ajst-9864	181	21	,	,	PUNCT
ajst-9864	181	22	resulting	result	VERB
ajst-9864	181	23	in	in	ADP
ajst-9864	181	24	the	the	DET
ajst-9864	181	25	word	word	NOUN
ajst-9864	181	26	-	-	PUNCT
ajst-9864	181	27	level	level	NOUN
ajst-9864	181	28	attention	attention	NOUN
ajst-9864	181	29	layer	layer	NOUN
ajst-9864	181	30	sentence	sentence	NOUN
ajst-9864	181	31	vector	vector	NOUN
ajst-9864	181	32	.	.	PUNCT
ajst-9864	182	1	the	the	DET
ajst-9864	182	2	wordlevel	wordlevel	NOUN
ajst-9864	182	3	attention	attention	NOUN
ajst-9864	182	4	layer	layer	NOUN
ajst-9864	182	5	is	be	AUX
ajst-9864	182	6	computed	compute	VERB
ajst-9864	182	7	as	as	SCONJ
ajst-9864	182	8	follows	follow	VERB
ajst-9864	182	9	.	.	PUNCT
ajst-9864	183	1	α	α	X
ajst-9864	183	2	∑	∑	PROPN
ajst-9864	183	3	(	(	PUNCT
ajst-9864	183	4	21	21	NUM
ajst-9864	183	5	)	)	PUNCT
ajst-9864	183	6	s	s	PART
ajst-9864	183	7	∑	∑	PUNCT
ajst-9864	183	8	α	α	PRON
ajst-9864	183	9	u	u	X
ajst-9864	183	10	(	(	PUNCT
ajst-9864	183	11	22	22	NUM
ajst-9864	183	12	)	)	PUNCT
ajst-9864	183	13	s3	s3	PROPN
ajst-9864	183	14	-	-	PUNCT
ajst-9864	183	15	4	4	NUM
ajst-9864	183	16	:	:	PUNCT
ajst-9864	183	17	the	the	DET
ajst-9864	183	18	third	third	ADJ
ajst-9864	183	19	layer	layer	NOUN
ajst-9864	183	20	is	be	AUX
ajst-9864	183	21	the	the	DET
ajst-9864	183	22	sentence	sentence	NOUN
ajst-9864	183	23	-	-	PUNCT
ajst-9864	183	24	level	level	NOUN
ajst-9864	183	25	attention	attention	NOUN
ajst-9864	183	26	layer	layer	NOUN
ajst-9864	183	27	.	.	PUNCT
ajst-9864	184	1	sentence	sentence	NOUN
ajst-9864	184	2	feature	feature	NOUN
ajst-9864	184	3	values	value	NOUN
ajst-9864	184	4	consisting	consist	VERB
ajst-9864	184	5	of	of	ADP
ajst-9864	184	6	the	the	DET
ajst-9864	184	7	outputs	output	NOUN
ajst-9864	184	8	of	of	ADP
ajst-9864	184	9	the	the	DET
ajst-9864	184	10	wordlevel	wordlevel	NOUN
ajst-9864	184	11	attention	attention	NOUN
ajst-9864	184	12	layer	layer	NOUN
ajst-9864	184	13	are	be	AUX
ajst-9864	184	14	used	use	VERB
ajst-9864	184	15	as	as	ADP
ajst-9864	184	16	inputs	input	NOUN
ajst-9864	184	17	to	to	ADP
ajst-9864	184	18	the	the	DET
ajst-9864	184	19	sentence	sentence	NOUN
ajst-9864	184	20	-	-	PUNCT
ajst-9864	184	21	level	level	NOUN
ajst-9864	184	22	attention	attention	NOUN
ajst-9864	184	23	layer	layer	NOUN
ajst-9864	184	24	.	.	PUNCT
ajst-9864	185	1	similar	similar	ADJ
ajst-9864	185	2	to	to	ADP
ajst-9864	185	3	the	the	DET
ajst-9864	185	4	word	word	NOUN
ajst-9864	185	5	-	-	PUNCT
ajst-9864	185	6	level	level	NOUN
ajst-9864	185	7	attention	attention	NOUN
ajst-9864	185	8	layer	layer	NOUN
ajst-9864	185	9	,	,	PUNCT
ajst-9864	185	10	a	a	DET
ajst-9864	185	11	randomly	randomly	ADV
ajst-9864	185	12	initialised	initialise	VERB
ajst-9864	185	13	word	word	NOUN
ajst-9864	185	14	context	context	NOUN
ajst-9864	185	15	vector	vector	NOUN
ajst-9864	185	16	us	we	PRON
ajst-9864	185	17	is	be	AUX
ajst-9864	185	18	added	add	VERB
ajst-9864	185	19	for	for	ADP
ajst-9864	185	20	cotraining	cotraining	NOUN
ajst-9864	185	21	,	,	PUNCT
ajst-9864	185	22	and	and	CCONJ
ajst-9864	185	23	v	v	NOUN
ajst-9864	185	24	is	be	AUX
ajst-9864	185	25	the	the	DET
ajst-9864	185	26	vector	vector	NOUN
ajst-9864	185	27	sum	sum	NOUN
ajst-9864	185	28	of	of	ADP
ajst-9864	185	29	all	all	DET
ajst-9864	185	30	sentences	sentence	NOUN
ajst-9864	185	31	,	,	PUNCT
ajst-9864	185	32	as	as	SCONJ
ajst-9864	185	33	follows	follow	VERB
ajst-9864	185	34	:	:	PUNCT
ajst-9864	185	35	ui＝tanh(ws·si+bs	ui＝tanh(ws·si+bs	NUM
ajst-9864	185	36	)	)	PUNCT
ajst-9864	185	37	(	(	PUNCT
ajst-9864	185	38	23	23	X
ajst-9864	185	39	)	)	PUNCT
ajst-9864	185	40	α	α	NOUN
ajst-9864	185	41	∑	∑	PROPN
ajst-9864	185	42	(	(	PUNCT
ajst-9864	185	43	24	24	NUM
ajst-9864	185	44	)	)	PUNCT
ajst-9864	185	45	v	v	NOUN
ajst-9864	185	46	∑	∑	PROPN
ajst-9864	185	47	α	α	PROPN
ajst-9864	185	48	u	u	NOUN
ajst-9864	185	49	(	(	PUNCT
ajst-9864	185	50	25	25	NUM
ajst-9864	185	51	)	)	PUNCT
ajst-9864	185	52	s3	s3	PROPN
ajst-9864	185	53	-	-	PUNCT
ajst-9864	185	54	5	5	NUM
ajst-9864	185	55	:	:	PUNCT
ajst-9864	185	56	the	the	DET
ajst-9864	185	57	fourth	fourth	ADJ
ajst-9864	185	58	layer	layer	NOUN
ajst-9864	185	59	is	be	AUX
ajst-9864	185	60	the	the	DET
ajst-9864	185	61	softmax	softmax	NOUN
ajst-9864	185	62	classifier	classifier	NOUN
ajst-9864	185	63	.	.	PUNCT
ajst-9864	186	1	the	the	DET
ajst-9864	186	2	softmax	softmax	NOUN
ajst-9864	186	3	classifier	classifier	NOUN
ajst-9864	186	4	maps	map	VERB
ajst-9864	186	5	v	v	NOUN
ajst-9864	186	6	to	to	ADP
ajst-9864	186	7	a	a	DET
ajst-9864	186	8	vector	vector	NOUN
ajst-9864	186	9	of	of	ADP
ajst-9864	186	10	elements	element	NOUN
ajst-9864	186	11	in	in	ADP
ajst-9864	186	12	the	the	DET
ajst-9864	186	13	range	range	NOUN
ajst-9864	186	14	[	[	X
ajst-9864	186	15	0	0	NUM
ajst-9864	186	16	,	,	PUNCT
ajst-9864	186	17	1	1	NUM
ajst-9864	186	18	]	]	PUNCT
ajst-9864	186	19	whose	whose	DET
ajst-9864	186	20	sum	sum	NOUN
ajst-9864	186	21	is	be	AUX
ajst-9864	186	22	1	1	NUM
ajst-9864	186	23	,	,	PUNCT
ajst-9864	186	24	as	as	SCONJ
ajst-9864	186	25	shown	show	VERB
ajst-9864	186	26	below	below	ADP
ajst-9864	186	27	:	:	PUNCT
ajst-9864	186	28	y	y	PROPN
ajst-9864	186	29	=	=	PUNCT
ajst-9864	186	30	softmax(v	softmax(v	PROPN
ajst-9864	186	31	)	)	PUNCT
ajst-9864	186	32	,	,	PUNCT
ajst-9864	186	33	where	where	SCONJ
ajst-9864	186	34	y＝[y1	y＝[y1	ADJ
ajst-9864	186	35	,	,	PUNCT
ajst-9864	186	36	y2	y2	INTJ
ajst-9864	186	37	,	,	PUNCT
ajst-9864	186	38	…	…	PUNCT
ajst-9864	186	39	,	,	PUNCT
ajst-9864	186	40	yn	yn	X
ajst-9864	186	41	]	]	PUNCT
ajst-9864	186	42	and	and	CCONJ
ajst-9864	186	43	yi∈[0	yi∈[0	NOUN
ajst-9864	186	44	,	,	PUNCT
ajst-9864	186	45	1]and	1]and	NUM
ajst-9864	186	46	∑yi＝1	∑yi＝1	PROPN
ajst-9864	186	47	.	.	PUNCT
ajst-9864	187	1	n	n	PRON
ajst-9864	187	2	is	be	AUX
ajst-9864	187	3	the	the	DET
ajst-9864	187	4	number	number	NOUN
ajst-9864	187	5	of	of	ADP
ajst-9864	187	6	relation	relation	NOUN
ajst-9864	187	7	labels	label	NOUN
ajst-9864	187	8	,	,	PUNCT
ajst-9864	187	9	that	that	ADV
ajst-9864	187	10	is	is	ADV
ajst-9864	187	11	,	,	PUNCT
ajst-9864	187	12	the	the	DET
ajst-9864	187	13	number	number	NOUN
ajst-9864	187	14	of	of	ADP
ajst-9864	187	15	categories	category	NOUN
ajst-9864	187	16	for	for	ADP
ajst-9864	187	17	relation	relation	NOUN
ajst-9864	187	18	extraction	extraction	NOUN
ajst-9864	187	19	.	.	PUNCT
ajst-9864	188	1	s3	s3	PROPN
ajst-9864	188	2	-	-	PUNCT
ajst-9864	188	3	6	6	NUM
ajst-9864	188	4	:	:	PUNCT
ajst-9864	188	5	the	the	DET
ajst-9864	188	6	classification	classification	NOUN
ajst-9864	188	7	results	result	NOUN
ajst-9864	188	8	are	be	AUX
ajst-9864	188	9	generated	generate	VERB
ajst-9864	188	10	after	after	ADP
ajst-9864	188	11	passing	pass	VERB
ajst-9864	188	12	through	through	ADP
ajst-9864	188	13	the	the	DET
ajst-9864	188	14	four	four	NUM
ajst-9864	188	15	layers	layer	NOUN
ajst-9864	188	16	of	of	ADP
ajst-9864	188	17	the	the	DET
ajst-9864	188	18	network	network	NOUN
ajst-9864	188	19	.	.	PUNCT
ajst-9864	189	1	s4	s4	PROPN
ajst-9864	189	2	:	:	PUNCT
ajst-9864	189	3	bigru-2att	bigru-2att	PROPN
ajst-9864	189	4	relation	relation	NOUN
ajst-9864	189	5	extraction	extraction	NOUN
ajst-9864	189	6	model	model	NOUN
ajst-9864	189	7	training	training	NOUN
ajst-9864	189	8	.	.	PUNCT
ajst-9864	190	1	the	the	DET
ajst-9864	190	2	bigru-2att	bigru-2att	PROPN
ajst-9864	190	3	relation	relation	NOUN
ajst-9864	190	4	extraction	extraction	NOUN
ajst-9864	190	5	network	network	NOUN
ajst-9864	190	6	uses	use	VERB
ajst-9864	190	7	the	the	DET
ajst-9864	190	8	sigmoid	sigmoid	NOUN
ajst-9864	190	9	function	function	NOUN
ajst-9864	190	10	as	as	ADP
ajst-9864	190	11	the	the	DET
ajst-9864	190	12	activation	activation	NOUN
ajst-9864	190	13	function	function	NOUN
ajst-9864	190	14	and	and	CCONJ
ajst-9864	190	15	employs	employ	VERB
ajst-9864	190	16	softmax	softmax	NOUN
ajst-9864	190	17	as	as	ADP
ajst-9864	190	18	the	the	DET
ajst-9864	190	19	classifier	classifier	NOUN
ajst-9864	190	20	.	.	PUNCT
ajst-9864	191	1	to	to	PART
ajst-9864	191	2	prevent	prevent	VERB
ajst-9864	191	3	overfitting	overfitting	NOUN
ajst-9864	191	4	during	during	ADP
ajst-9864	191	5	the	the	DET
ajst-9864	191	6	training	training	NOUN
ajst-9864	191	7	process	process	NOUN
ajst-9864	191	8	,	,	PUNCT
ajst-9864	191	9	l2	l2	NOUN
ajst-9864	191	10	regularisation	regularisation	NOUN
ajst-9864	191	11	is	be	AUX
ajst-9864	191	12	added	add	VERB
ajst-9864	191	13	to	to	PART
ajst-9864	191	14	constrain	constrain	VERB
ajst-9864	191	15	the	the	DET
ajst-9864	191	16	bigru-2att	bigru-2att	PROPN
ajst-9864	191	17	network	network	NOUN
ajst-9864	191	18	.	.	PUNCT
ajst-9864	192	1	the	the	DET
ajst-9864	192	2	dropout	dropout	NOUN
ajst-9864	192	3	strategy	strategy	NOUN
ajst-9864	192	4	is	be	AUX
ajst-9864	192	5	introduced	introduce	VERB
ajst-9864	192	6	during	during	ADP
ajst-9864	192	7	training	training	NOUN
ajst-9864	192	8	with	with	ADP
ajst-9864	192	9	a	a	DET
ajst-9864	192	10	suppression	suppression	NOUN
ajst-9864	192	11	probability	probability	NOUN
ajst-9864	192	12	of	of	ADP
ajst-9864	192	13	0.5	0.5	NUM
ajst-9864	192	14	.	.	PUNCT
ajst-9864	193	1	the	the	DET
ajst-9864	193	2	model	model	NOUN
ajst-9864	193	3	parameters	parameter	NOUN
ajst-9864	193	4	are	be	AUX
ajst-9864	193	5	trained	train	VERB
ajst-9864	193	6	using	use	VERB
ajst-9864	193	7	the	the	DET
ajst-9864	193	8	batch	batch	NOUN
ajst-9864	193	9	adam	adam	PROPN
ajst-9864	193	10	optimization	optimization	NOUN
ajst-9864	193	11	method	method	NOUN
ajst-9864	193	12	.	.	PUNCT
ajst-9864	194	1	the	the	DET
ajst-9864	194	2	detailed	detailed	ADJ
ajst-9864	194	3	model	model	NOUN
ajst-9864	194	4	parameters	parameter	NOUN
ajst-9864	194	5	are	be	AUX
ajst-9864	194	6	shown	show	VERB
ajst-9864	194	7	in	in	ADP
ajst-9864	194	8	table	table	NOUN
ajst-9864	194	9	4	4	NUM
ajst-9864	194	10	.	.	PUNCT
ajst-9864	194	11	table	table	NOUN
ajst-9864	194	12	4	4	NUM
ajst-9864	194	13	model	model	NOUN
ajst-9864	194	14	parameters	parameter	NOUN
ajst-9864	194	15	parameters	parameter	NOUN
ajst-9864	194	16	meaning	mean	VERB
ajst-9864	194	17	value	value	NOUN
ajst-9864	194	18	batch	batch	NOUN
ajst-9864	194	19	the	the	DET
ajst-9864	194	20	number	number	NOUN
ajst-9864	194	21	of	of	ADP
ajst-9864	194	22	samples	sample	NOUN
ajst-9864	194	23	used	use	VERB
ajst-9864	194	24	per	per	ADP
ajst-9864	194	25	batch	batch	NOUN
ajst-9864	194	26	70	70	NUM
ajst-9864	194	27	epoch	epoch	NOUN
ajst-9864	194	28	the	the	DET
ajst-9864	194	29	number	number	NOUN
ajst-9864	194	30	of	of	ADP
ajst-9864	194	31	iterations	iteration	NOUN
ajst-9864	194	32	50	50	NUM
ajst-9864	194	33	gru_size	gru_size	NOUN
ajst-9864	194	34	the	the	DET
ajst-9864	194	35	number	number	NOUN
ajst-9864	194	36	of	of	ADP
ajst-9864	194	37	gru	gru	NOUN
ajst-9864	194	38	units	unit	NOUN
ajst-9864	194	39	,	,	PUNCT
ajst-9864	194	40	i.e.	i.e.	X
ajst-9864	194	41	the	the	DET
ajst-9864	194	42	size	size	NOUN
ajst-9864	194	43	of	of	ADP
ajst-9864	194	44	sentence	sentence	NOUN
ajst-9864	194	45	embeddings	embedding	NOUN
ajst-9864	194	46	200	200	NUM
ajst-9864	194	47	big_num	big_num	ADJ
ajst-9864	194	48	the	the	DET
ajst-9864	194	49	number	number	NOUN
ajst-9864	194	50	of	of	ADP
ajst-9864	194	51	entity	entity	NOUN
ajst-9864	194	52	pairs	pair	NOUN
ajst-9864	194	53	per	per	ADP
ajst-9864	194	54	batch	batch	NOUN
ajst-9864	194	55	during	during	ADP
ajst-9864	194	56	training	training	NOUN
ajst-9864	194	57	or	or	CCONJ
ajst-9864	194	58	testing	test	VERB
ajst-9864	194	59	30	30	NUM
ajst-9864	194	60	dropout	dropout	NOUN
ajst-9864	194	61	dropout	dropout	NOUN
ajst-9864	194	62	rate	rate	NOUN
ajst-9864	194	63	0.5	0.5	NUM
ajst-9864	194	64	learning_r	learning_r	NOUN
ajst-9864	194	65	ate	eat	VERB
ajst-9864	194	66	learning	learn	VERB
ajst-9864	194	67	rate	rate	NOUN
ajst-9864	194	68	1	1	NUM
ajst-9864	194	69	regularizat	regularizat	NOUN
ajst-9864	194	70	ion_rate	ion_rate	ADP
ajst-9864	194	71	l2	l2	NOUN
ajst-9864	194	72	regularization	regularization	NOUN
ajst-9864	194	73	coefficient	coefficient	NOUN
ajst-9864	194	74	0.0001	0.0001	NUM
ajst-9864	194	75	s5	s5	NOUN
ajst-9864	194	76	:	:	PUNCT
ajst-9864	194	77	input	input	VERB
ajst-9864	194	78	the	the	DET
ajst-9864	194	79	entity	entity	NOUN
ajst-9864	194	80	identification	identification	NOUN
ajst-9864	194	81	results	result	NOUN
ajst-9864	194	82	and	and	CCONJ
ajst-9864	194	83	test	test	NOUN
ajst-9864	194	84	set	set	VERB
ajst-9864	194	85	data	datum	NOUN
ajst-9864	194	86	into	into	ADP
ajst-9864	194	87	the	the	DET
ajst-9864	194	88	bigru-2att	bigru-2att	PROPN
ajst-9864	194	89	relationship	relationship	NOUN
ajst-9864	194	90	extraction	extraction	NOUN
ajst-9864	194	91	model	model	NOUN
ajst-9864	194	92	to	to	PART
ajst-9864	194	93	obtain	obtain	VERB
ajst-9864	194	94	the	the	DET
ajst-9864	194	95	relationship	relationship	NOUN
ajst-9864	194	96	extraction	extraction	NOUN
ajst-9864	194	97	results	result	NOUN
ajst-9864	194	98	.	.	PUNCT
ajst-9864	195	1	4.3	4.3	NUM
ajst-9864	195	2	.	.	PUNCT
ajst-9864	196	1	experimental	experimental	ADJ
ajst-9864	196	2	results	result	VERB
ajst-9864	196	3	the	the	DET
ajst-9864	196	4	results	result	NOUN
ajst-9864	196	5	of	of	ADP
ajst-9864	196	6	the	the	DET
ajst-9864	196	7	relationship	relationship	NOUN
ajst-9864	196	8	extraction	extraction	NOUN
ajst-9864	196	9	obtained	obtain	VERB
ajst-9864	196	10	in	in	ADP
ajst-9864	196	11	this	this	DET
ajst-9864	196	12	implementation	implementation	NOUN
ajst-9864	196	13	are	be	AUX
ajst-9864	196	14	evaluated	evaluate	VERB
ajst-9864	196	15	for	for	ADP
ajst-9864	196	16	performance	performance	NOUN
ajst-9864	196	17	,	,	PUNCT
ajst-9864	196	18	and	and	CCONJ
ajst-9864	196	19	the	the	DET
ajst-9864	196	20	performance	performance	NOUN
ajst-9864	196	21	evaluation	evaluation	NOUN
ajst-9864	196	22	metric	metric	NOUN
ajst-9864	196	23	uses	use	VERB
ajst-9864	196	24	the	the	DET
ajst-9864	196	25	accuracy	accuracy	NOUN
ajst-9864	196	26	rate	rate	NOUN
ajst-9864	196	27	,	,	PUNCT
ajst-9864	196	28	calculated	calculate	VERB
ajst-9864	196	29	as	as	SCONJ
ajst-9864	196	30	follows.the	follows.the	DET
ajst-9864	196	31	results	result	NOUN
ajst-9864	196	32	are	be	AUX
ajst-9864	196	33	shown	show	VERB
ajst-9864	196	34	in	in	ADP
ajst-9864	196	35	figure	figure	NOUN
ajst-9864	196	36	4	4	NUM
ajst-9864	196	37	precision	precision	NOUN
ajst-9864	196	38	(	(	PUNCT
ajst-9864	196	39	26	26	NUM
ajst-9864	196	40	)	)	PUNCT
ajst-9864	196	41	recall	recall	NOUN
ajst-9864	196	42	(	(	PUNCT
ajst-9864	196	43	27	27	NUM
ajst-9864	196	44	)	)	PUNCT
ajst-9864	196	45	(	(	PUNCT
ajst-9864	196	46	28	28	NUM
ajst-9864	196	47	)	)	PUNCT
ajst-9864	196	48	the	the	DET
ajst-9864	196	49	terms	term	NOUN
ajst-9864	196	50	used	use	VERB
ajst-9864	196	51	are	be	AUX
ajst-9864	196	52	as	as	SCONJ
ajst-9864	196	53	follows	follow	VERB
ajst-9864	196	54	:	:	PUNCT
ajst-9864	196	55	true	true	ADJ
ajst-9864	196	56	positives	positive	NOUN
ajst-9864	196	57	(	(	PUNCT
ajst-9864	196	58	tp	tp	NOUN
ajst-9864	196	59	)	)	PUNCT
ajst-9864	196	60	equal	equal	VERB
ajst-9864	196	61	the	the	DET
ajst-9864	196	62	number	number	NOUN
ajst-9864	196	63	of	of	ADP
ajst-9864	196	64	occurrences	occurrence	NOUN
ajst-9864	196	65	that	that	PRON
ajst-9864	196	66	were	be	AUX
ajst-9864	196	67	properly	properly	ADV
ajst-9864	196	68	categorised	categorise	VERB
ajst-9864	196	69	,	,	PUNCT
ajst-9864	196	70	false	false	ADJ
ajst-9864	196	71	positives	positive	NOUN
ajst-9864	196	72	(	(	PUNCT
ajst-9864	196	73	fp	fp	X
ajst-9864	196	74	)	)	PUNCT
ajst-9864	196	75	equal	equal	VERB
ajst-9864	196	76	the	the	DET
ajst-9864	196	77	number	number	NOUN
ajst-9864	196	78	of	of	ADP
ajst-9864	196	79	instances	instance	NOUN
ajst-9864	196	80	that	that	PRON
ajst-9864	196	81	were	be	AUX
ajst-9864	196	82	incorrectly	incorrectly	ADV
ajst-9864	196	83	forecasted	forecast	VERB
ajst-9864	196	84	as	as	ADP
ajst-9864	196	85	positive	positive	ADJ
ajst-9864	196	86	,	,	PUNCT
ajst-9864	196	87	and	and	CCONJ
ajst-9864	196	88	false	false	ADJ
ajst-9864	196	89	negatives	negative	NOUN
ajst-9864	196	90	(	(	PUNCT
ajst-9864	196	91	fn	fn	NOUN
ajst-9864	196	92	)	)	PUNCT
ajst-9864	196	93	equal	equal	ADJ
ajst-9864	196	94	the	the	DET
ajst-9864	196	95	number	number	NOUN
ajst-9864	196	96	of	of	ADP
ajst-9864	196	97	instances	instance	NOUN
ajst-9864	196	98	that	that	PRON
ajst-9864	196	99	were	be	AUX
ajst-9864	196	100	incorrectly	incorrectly	ADV
ajst-9864	196	101	projected	project	VERB
ajst-9864	196	102	as	as	ADP
ajst-9864	196	103	negative	negative	ADJ
ajst-9864	196	104	.	.	PUNCT
ajst-9864	197	1	figure	figure	NOUN
ajst-9864	197	2	4	4	NUM
ajst-9864	197	3	.	.	PUNCT
ajst-9864	197	4	results	result	VERB
ajst-9864	197	5	4.4	4.4	NUM
ajst-9864	197	6	.	.	PUNCT
ajst-9864	198	1	comparison	comparison	NOUN
ajst-9864	198	2	of	of	ADP
ajst-9864	198	3	experimental	experimental	ADJ
ajst-9864	198	4	results	result	NOUN
ajst-9864	198	5	after	after	ADP
ajst-9864	198	6	identifying	identify	VERB
ajst-9864	198	7	the	the	DET
ajst-9864	198	8	test	test	NOUN
ajst-9864	198	9	samples	sample	NOUN
ajst-9864	198	10	,	,	PUNCT
ajst-9864	198	11	the	the	DET
ajst-9864	198	12	relationship	relationship	NOUN
ajst-9864	198	13	extraction	extraction	NOUN
ajst-9864	198	14	's	's	PART
ajst-9864	198	15	precision	precision	NOUN
ajst-9864	198	16	,	,	PUNCT
ajst-9864	198	17	recall	recall	NOUN
ajst-9864	198	18	,	,	PUNCT
ajst-9864	198	19	and	and	CCONJ
ajst-9864	198	20	f1	f1	NOUN
ajst-9864	198	21	score	score	NOUN
ajst-9864	198	22	were	be	AUX
ajst-9864	198	23	determined	determine	VERB
ajst-9864	198	24	to	to	PART
ajst-9864	198	25	be	be	AUX
ajst-9864	198	26	85.22	85.22	NUM
ajst-9864	198	27	%	%	NOUN
ajst-9864	198	28	,	,	PUNCT
ajst-9864	198	29	87.57	87.57	NUM
ajst-9864	198	30	%	%	NOUN
ajst-9864	198	31	,	,	PUNCT
ajst-9864	198	32	and	and	CCONJ
ajst-9864	198	33	86.40	86.40	NUM
ajst-9864	198	34	%	%	NOUN
ajst-9864	198	35	,	,	PUNCT
ajst-9864	198	36	respectively	respectively	ADV
ajst-9864	198	37	.	.	PUNCT
ajst-9864	199	1	as	as	ADP
ajst-9864	199	2	a	a	DET
ajst-9864	199	3	result	result	NOUN
ajst-9864	199	4	,	,	PUNCT
ajst-9864	199	5	it	it	PRON
ajst-9864	199	6	is	be	AUX
ajst-9864	199	7	clear	clear	ADJ
ajst-9864	199	8	that	that	SCONJ
ajst-9864	199	9	the	the	DET
ajst-9864	199	10	suggested	suggest	VERB
ajst-9864	199	11	strategy	strategy	NOUN
ajst-9864	199	12	significantly	significantly	ADV
ajst-9864	199	13	raises	raise	VERB
ajst-9864	199	14	relationship	relationship	NOUN
ajst-9864	199	15	extraction	extraction	NOUN
ajst-9864	199	16	's	's	PART
ajst-9864	199	17	precision	precision	NOUN
ajst-9864	199	18	,	,	PUNCT
ajst-9864	199	19	recall	recall	NOUN
ajst-9864	199	20	,	,	PUNCT
ajst-9864	199	21	and	and	CCONJ
ajst-9864	199	22	f1	f1	PROPN
ajst-9864	199	23	score	score	NOUN
ajst-9864	199	24	.	.	PUNCT
ajst-9864	200	1	experimental	experimental	ADJ
ajst-9864	200	2	results	result	NOUN
ajst-9864	200	3	using	use	VERB
ajst-9864	200	4	the	the	DET
ajst-9864	200	5	data	datum	NOUN
ajst-9864	200	6	and	and	CCONJ
ajst-9864	200	7	parameter	parameter	NOUN
ajst-9864	200	8	settings	setting	NOUN
ajst-9864	200	9	acquired	acquire	VERB
ajst-9864	200	10	in	in	ADP
ajst-9864	200	11	section	section	NOUN
ajst-9864	200	12	3.1	3.1	NUM
ajst-9864	200	13	verify	verify	VERB
ajst-9864	200	14	the	the	DET
ajst-9864	200	15	extraction	extraction	NOUN
ajst-9864	200	16	effectiveness	effectiveness	NOUN
ajst-9864	200	17	of	of	ADP
ajst-9864	200	18	the	the	DET
ajst-9864	200	19	bidirectional	bidirectional	PROPN
ajst-9864	200	20	gru	gru	PROPN
ajst-9864	200	21	neural	neural	PROPN
ajst-9864	200	22	network	network	NOUN
ajst-9864	200	23	model	model	NOUN
ajst-9864	200	24	with	with	ADP
ajst-9864	200	25	an	an	DET
ajst-9864	200	26	attention	attention	NOUN
ajst-9864	200	27	mechanism	mechanism	NOUN
ajst-9864	200	28	integrated	integrate	VERB
ajst-9864	200	29	for	for	ADP
ajst-9864	200	30	chinese	chinese	ADJ
ajst-9864	200	31	character	character	NOUN
ajst-9864	200	32	connection	connection	NOUN
ajst-9864	200	33	extraction	extraction	NOUN
ajst-9864	200	34	.	.	PUNCT
ajst-9864	201	1	table	table	NOUN
ajst-9864	201	2	5	5	NUM
ajst-9864	201	3	summarised	summarise	VERB
ajst-9864	201	4	the	the	DET
ajst-9864	201	5	accuracy	accuracy	NOUN
ajst-9864	201	6	,	,	PUNCT
ajst-9864	201	7	recall	recall	NOUN
ajst-9864	201	8	,	,	PUNCT
ajst-9864	201	9	and	and	CCONJ
ajst-9864	201	10	fvalue	fvalue	NOUN
ajst-9864	201	11	of	of	ADP
ajst-9864	201	12	each	each	DET
ajst-9864	201	13	model	model	NOUN
ajst-9864	201	14	for	for	ADP
ajst-9864	201	15	the	the	DET
ajst-9864	201	16	test	test	NOUN
ajst-9864	201	17	set	set	VERB
ajst-9864	201	18	.	.	PUNCT
ajst-9864	202	1	table	table	NOUN
ajst-9864	202	2	5	5	NUM
ajst-9864	202	3	.	.	PUNCT
ajst-9864	202	4	model	model	NOUN
ajst-9864	202	5	parameters	parameters	PROPN
ajst-9864	202	6	model	model	PROPN
ajst-9864	202	7	precision	precision	PROPN
ajst-9864	202	8	recall	recall	PROPN
ajst-9864	202	9	f1	f1	PROPN
ajst-9864	202	10	score	score	NOUN
ajst-9864	202	11	gru	gru	PROPN
ajst-9864	202	12	51.25	51.25	NUM
ajst-9864	202	13	30.12	30.12	NUM
ajst-9864	202	14	40.15	40.15	NUM
ajst-9864	202	15	gru+catt	gru+catt	NOUN
ajst-9864	202	16	58.30	58.30	NUM
ajst-9864	202	17	36.21	36.21	NUM
ajst-9864	202	18	45.78	45.78	NUM
ajst-9864	202	19	gru+satt	gru+satt	NOUN
ajst-9864	202	20	62.40	62.40	NUM
ajst-9864	202	21	45.90	45.90	NUM
ajst-9864	202	22	50.12	50.12	NUM
ajst-9864	202	23	gru+catt+satt	gru+catt+satt	PROPN
ajst-9864	202	24	68.90	68.90	NUM
ajst-9864	202	25	50.12	50.12	NUM
ajst-9864	202	26	57.21	57.21	NUM
ajst-9864	202	27	according	accord	VERB
ajst-9864	202	28	to	to	ADP
ajst-9864	202	29	table	table	NOUN
ajst-9864	202	30	5	5	NUM
ajst-9864	202	31	,	,	PUNCT
ajst-9864	202	32	the	the	DET
ajst-9864	202	33	accuracy	accuracy	NOUN
ajst-9864	202	34	and	and	CCONJ
ajst-9864	202	35	recall	recall	NOUN
ajst-9864	202	36	of	of	ADP
ajst-9864	202	37	the	the	DET
ajst-9864	202	38	regular	regular	ADJ
ajst-9864	202	39	bidirectional	bidirectional	PROPN
ajst-9864	202	40	gru	gru	PROPN
ajst-9864	202	41	neural	neural	PROPN
ajst-9864	202	42	network	network	NOUN
ajst-9864	202	43	model	model	NOUN
ajst-9864	202	44	are	be	AUX
ajst-9864	202	45	51.25	51.25	NUM
ajst-9864	202	46	%	%	NOUN
ajst-9864	202	47	and	and	CCONJ
ajst-9864	202	48	30.18	30.18	NUM
ajst-9864	202	49	%	%	NOUN
ajst-9864	202	50	,	,	PUNCT
ajst-9864	202	51	respectively	respectively	ADV
ajst-9864	202	52	.	.	PUNCT
ajst-9864	203	1	the	the	DET
ajst-9864	203	2	accuracy	accuracy	NOUN
ajst-9864	203	3	and	and	CCONJ
ajst-9864	203	4	recall	recall	NOUN
ajst-9864	203	5	of	of	ADP
ajst-9864	203	6	the	the	DET
ajst-9864	203	7	model	model	NOUN
ajst-9864	203	8	with	with	ADP
ajst-9864	203	9	a	a	DET
ajst-9864	203	10	character	character	NOUN
ajst-9864	203	11	-	-	PUNCT
ajst-9864	203	12	level	level	NOUN
ajst-9864	203	13	attention	attention	NOUN
ajst-9864	203	14	mechanism	mechanism	NOUN
ajst-9864	203	15	added	add	VERB
ajst-9864	203	16	are	be	AUX
ajst-9864	203	17	58.31	58.31	NUM
ajst-9864	203	18	%	%	NOUN
ajst-9864	203	19	and	and	CCONJ
ajst-9864	203	20	36.21	36.21	NUM
ajst-9864	203	21	%	%	NOUN
ajst-9864	203	22	,	,	PUNCT
ajst-9864	203	23	respectively	respectively	ADV
ajst-9864	203	24	.	.	PUNCT
ajst-9864	204	1	the	the	DET
ajst-9864	204	2	accuracy	accuracy	NOUN
ajst-9864	204	3	and	and	CCONJ
ajst-9864	204	4	recall	recall	NOUN
ajst-9864	204	5	of	of	ADP
ajst-9864	204	6	the	the	DET
ajst-9864	204	7	model	model	NOUN
ajst-9864	204	8	with	with	ADP
ajst-9864	204	9	a	a	DET
ajst-9864	204	10	sentence	sentence	NOUN
ajst-9864	204	11	-	-	PUNCT
ajst-9864	204	12	level	level	NOUN
ajst-9864	204	13	attention	attention	NOUN
ajst-9864	204	14	mechanism	mechanism	NOUN
ajst-9864	204	15	added	add	VERB
ajst-9864	204	16	are	be	AUX
ajst-9864	204	17	62.40	62.40	NUM
ajst-9864	204	18	%	%	NOUN
ajst-9864	204	19	and	and	CCONJ
ajst-9864	204	20	45.90	45.90	NUM
ajst-9864	204	21	%	%	NOUN
ajst-9864	204	22	,	,	PUNCT
ajst-9864	204	23	respectively	respectively	ADV
ajst-9864	204	24	.	.	PUNCT
ajst-9864	205	1	the	the	DET
ajst-9864	205	2	accuracy	accuracy	NOUN
ajst-9864	205	3	and	and	CCONJ
ajst-9864	205	4	recall	recall	NOUN
ajst-9864	205	5	of	of	ADP
ajst-9864	205	6	the	the	DET
ajst-9864	205	7	bidirectional	bidirectional	PROPN
ajst-9864	205	8	gru	gru	PROPN
ajst-9864	205	9	neural	neural	PROPN
ajst-9864	205	10	network	network	NOUN
ajst-9864	205	11	model	model	NOUN
ajst-9864	205	12	with	with	ADP
ajst-9864	205	13	a	a	DET
ajst-9864	205	14	dual	dual	ADJ
ajst-9864	205	15	-	-	PUNCT
ajst-9864	205	16	layer	layer	NOUN
ajst-9864	205	17	attention	attention	NOUN
ajst-9864	205	18	mechanism	mechanism	NOUN
ajst-9864	205	19	added	add	VERB
ajst-9864	205	20	are	be	AUX
ajst-9864	205	21	68.90	68.90	NUM
ajst-9864	205	22	%	%	NOUN
ajst-9864	205	23	and	and	CCONJ
ajst-9864	205	24	50.12	50.12	NUM
ajst-9864	205	25	%	%	NOUN
ajst-9864	205	26	,	,	PUNCT
ajst-9864	205	27	respectively	respectively	ADV
ajst-9864	205	28	.	.	PUNCT
ajst-9864	206	1	regardless	regardless	ADV
ajst-9864	206	2	of	of	ADP
ajst-9864	206	3	whether	whether	SCONJ
ajst-9864	206	4	measured	measure	VERB
ajst-9864	206	5	by	by	ADP
ajst-9864	206	6	precision	precision	NOUN
ajst-9864	206	7	,	,	PUNCT
ajst-9864	206	8	recall	recall	NOUN
ajst-9864	206	9	,	,	PUNCT
ajst-9864	206	10	or	or	CCONJ
ajst-9864	206	11	fmeasure	fmeasure	NOUN
ajst-9864	206	12	,	,	PUNCT
ajst-9864	206	13	the	the	DET
ajst-9864	206	14	model	model	NOUN
ajst-9864	206	15	incorporating	incorporate	VERB
ajst-9864	206	16	attention	attention	NOUN
ajst-9864	206	17	mechanisms	mechanism	NOUN
ajst-9864	206	18	into	into	ADP
ajst-9864	206	19	a	a	DET
ajst-9864	206	20	bidirectional	bidirectional	ADJ
ajst-9864	206	21	gru	gru	PROPN
ajst-9864	206	22	neural	neural	ADJ
ajst-9864	206	23	network	network	NOUN
ajst-9864	206	24	outperforms	outperform	VERB
ajst-9864	206	25	the	the	DET
ajst-9864	206	26	regular	regular	ADJ
ajst-9864	206	27	bidirectional	bidirectional	PROPN
ajst-9864	206	28	gru	gru	PROPN
ajst-9864	206	29	neural	neural	PROPN
ajst-9864	206	30	network	network	NOUN
ajst-9864	206	31	model	model	NOUN
ajst-9864	206	32	.	.	PUNCT
ajst-9864	207	1	furthermore	furthermore	ADV
ajst-9864	207	2	,	,	PUNCT
ajst-9864	207	3	the	the	DET
ajst-9864	207	4	135	135	NUM
ajst-9864	207	5	relationship	relationship	NOUN
ajst-9864	207	6	extraction	extraction	NOUN
ajst-9864	207	7	model	model	NOUN
ajst-9864	207	8	incorporating	incorporate	VERB
ajst-9864	207	9	both	both	PRON
ajst-9864	207	10	characterlevel	characterlevel	NOUN
ajst-9864	207	11	and	and	CCONJ
ajst-9864	207	12	sentence	sentence	NOUN
ajst-9864	207	13	-	-	PUNCT
ajst-9864	207	14	level	level	NOUN
ajst-9864	207	15	attention	attention	NOUN
ajst-9864	207	16	mechanisms	mechanism	NOUN
ajst-9864	207	17	exhibits	exhibit	VERB
ajst-9864	207	18	only	only	ADV
ajst-9864	207	19	slight	slight	ADJ
ajst-9864	207	20	differences	difference	NOUN
ajst-9864	207	21	in	in	ADP
ajst-9864	207	22	precision	precision	NOUN
ajst-9864	207	23	and	and	CCONJ
ajst-9864	207	24	recall	recall	NOUN
ajst-9864	207	25	compared	compare	VERB
ajst-9864	207	26	to	to	ADP
ajst-9864	207	27	the	the	DET
ajst-9864	207	28	regular	regular	ADJ
ajst-9864	207	29	gru	gru	PROPN
ajst-9864	207	30	neural	neural	ADJ
ajst-9864	207	31	network	network	NOUN
ajst-9864	207	32	model	model	NOUN
ajst-9864	207	33	but	but	CCONJ
ajst-9864	207	34	still	still	ADV
ajst-9864	207	35	achieves	achieve	VERB
ajst-9864	207	36	higher	high	ADJ
ajst-9864	207	37	performance	performance	NOUN
ajst-9864	207	38	.	.	PUNCT
ajst-9864	208	1	the	the	DET
ajst-9864	208	2	gru	gru	PROPN
ajst-9864	208	3	neural	neural	PROPN
ajst-9864	208	4	network	network	NOUN
ajst-9864	208	5	model	model	NOUN
ajst-9864	208	6	with	with	ADP
ajst-9864	208	7	a	a	DET
ajst-9864	208	8	duallayer	duallayer	NOUN
ajst-9864	208	9	attention	attention	NOUN
ajst-9864	208	10	mechanism	mechanism	NOUN
ajst-9864	208	11	achieves	achieve	VERB
ajst-9864	208	12	the	the	DET
ajst-9864	208	13	highest	high	ADJ
ajst-9864	208	14	precision	precision	NOUN
ajst-9864	208	15	and	and	CCONJ
ajst-9864	208	16	recall	recall	NOUN
ajst-9864	208	17	.	.	PUNCT
ajst-9864	209	1	these	these	DET
ajst-9864	209	2	results	result	NOUN
ajst-9864	209	3	demonstrate	demonstrate	VERB
ajst-9864	209	4	that	that	SCONJ
ajst-9864	209	5	selectively	selectively	ADV
ajst-9864	209	6	attending	attend	VERB
ajst-9864	209	7	to	to	ADP
ajst-9864	209	8	important	important	ADJ
ajst-9864	209	9	information	information	NOUN
ajst-9864	209	10	in	in	ADP
ajst-9864	209	11	sentences	sentence	NOUN
ajst-9864	209	12	can	can	AUX
ajst-9864	209	13	improve	improve	VERB
ajst-9864	209	14	the	the	DET
ajst-9864	209	15	precision	precision	NOUN
ajst-9864	209	16	of	of	ADP
ajst-9864	209	17	relationship	relationship	NOUN
ajst-9864	209	18	extraction	extraction	NOUN
ajst-9864	209	19	.	.	PUNCT
ajst-9864	210	1	5	5	NUM
ajst-9864	210	2	.	.	X
ajst-9864	210	3	discussion	discussion	NOUN
ajst-9864	210	4	and	and	CCONJ
ajst-9864	210	5	future	future	ADJ
ajst-9864	210	6	work	work	NOUN
ajst-9864	210	7	5.1	5.1	NUM
ajst-9864	210	8	.	.	PUNCT
ajst-9864	211	1	discussion	discussion	NOUN
ajst-9864	211	2	the	the	DET
ajst-9864	211	3	experimental	experimental	ADJ
ajst-9864	211	4	results	result	NOUN
ajst-9864	211	5	show	show	VERB
ajst-9864	211	6	that	that	SCONJ
ajst-9864	211	7	the	the	DET
ajst-9864	211	8	gru	gru	NOUN
ajst-9864	211	9	word	word	NOUN
ajst-9864	211	10	-	-	PUNCT
ajst-9864	211	11	level	level	NOUN
ajst-9864	211	12	supervised	supervised	ADJ
ajst-9864	211	13	,	,	PUNCT
ajst-9864	211	14	and	and	CCONJ
ajst-9864	211	15	sentence	sentence	NOUN
ajst-9864	211	16	-	-	PUNCT
ajst-9864	211	17	level	level	NOUN
ajst-9864	211	18	supervised	supervised	ADJ
ajst-9864	211	19	models	model	NOUN
ajst-9864	211	20	have	have	AUX
ajst-9864	211	21	achieved	achieve	VERB
ajst-9864	211	22	certain	certain	ADJ
ajst-9864	211	23	achievements	achievement	NOUN
ajst-9864	211	24	in	in	ADP
ajst-9864	211	25	handling	handling	NOUN
ajst-9864	211	26	tasks	task	NOUN
ajst-9864	211	27	,	,	PUNCT
ajst-9864	211	28	but	but	CCONJ
ajst-9864	211	29	there	there	PRON
ajst-9864	211	30	is	be	VERB
ajst-9864	211	31	still	still	ADV
ajst-9864	211	32	room	room	NOUN
ajst-9864	211	33	for	for	ADP
ajst-9864	211	34	improvement	improvement	NOUN
ajst-9864	211	35	.	.	PUNCT
ajst-9864	212	1	measured	measure	VERB
ajst-9864	212	2	as	as	ADP
ajst-9864	212	3	a	a	DET
ajst-9864	212	4	percentage	percentage	NOUN
ajst-9864	212	5	of	of	ADP
ajst-9864	212	6	properly	properly	ADV
ajst-9864	212	7	predicted	predict	VERB
ajst-9864	212	8	samples	sample	NOUN
ajst-9864	212	9	relative	relative	ADJ
ajst-9864	212	10	to	to	ADP
ajst-9864	212	11	the	the	DET
ajst-9864	212	12	total	total	ADJ
ajst-9864	212	13	number	number	NOUN
ajst-9864	212	14	of	of	ADP
ajst-9864	212	15	samples	sample	NOUN
ajst-9864	212	16	,	,	PUNCT
ajst-9864	212	17	accuracy	accuracy	NOUN
ajst-9864	212	18	is	be	AUX
ajst-9864	212	19	a	a	DET
ajst-9864	212	20	key	key	ADJ
ajst-9864	212	21	performance	performance	NOUN
ajst-9864	212	22	metric	metric	NOUN
ajst-9864	212	23	for	for	ADP
ajst-9864	212	24	any	any	DET
ajst-9864	212	25	data	data	NOUN
ajst-9864	212	26	-	-	PUNCT
ajst-9864	212	27	driven	drive	VERB
ajst-9864	212	28	endeavour	endeavour	NOUN
ajst-9864	212	29	.	.	PUNCT
ajst-9864	213	1	the	the	DET
ajst-9864	213	2	model	model	NOUN
ajst-9864	213	3	's	's	PART
ajst-9864	213	4	recall	recall	NOUN
ajst-9864	213	5	is	be	AUX
ajst-9864	213	6	the	the	DET
ajst-9864	213	7	percentage	percentage	NOUN
ajst-9864	213	8	of	of	ADP
ajst-9864	213	9	training	training	NOUN
ajst-9864	213	10	samples	sample	NOUN
ajst-9864	213	11	that	that	PRON
ajst-9864	213	12	were	be	AUX
ajst-9864	213	13	properly	properly	ADV
ajst-9864	213	14	classified	classify	VERB
ajst-9864	213	15	as	as	ADP
ajst-9864	213	16	positive	positive	ADJ
ajst-9864	213	17	.	.	PUNCT
ajst-9864	214	1	the	the	DET
ajst-9864	214	2	f1	f1	PROPN
ajst-9864	214	3	score	score	NOUN
ajst-9864	214	4	is	be	AUX
ajst-9864	214	5	a	a	DET
ajst-9864	214	6	complete	complete	ADJ
ajst-9864	214	7	assessment	assessment	NOUN
ajst-9864	214	8	metric	metric	NOUN
ajst-9864	214	9	that	that	PRON
ajst-9864	214	10	is	be	AUX
ajst-9864	214	11	the	the	DET
ajst-9864	214	12	weighted	weight	VERB
ajst-9864	214	13	harmonic	harmonic	ADJ
ajst-9864	214	14	mean	mean	NOUN
ajst-9864	214	15	of	of	ADP
ajst-9864	214	16	accuracy	accuracy	NOUN
ajst-9864	214	17	and	and	CCONJ
ajst-9864	214	18	recall	recall	NOUN
ajst-9864	214	19	.	.	PUNCT
ajst-9864	215	1	in	in	ADP
ajst-9864	215	2	the	the	DET
ajst-9864	215	3	experimental	experimental	ADJ
ajst-9864	215	4	results	result	NOUN
ajst-9864	215	5	,	,	PUNCT
ajst-9864	215	6	the	the	DET
ajst-9864	215	7	accuracy	accuracy	NOUN
ajst-9864	215	8	of	of	ADP
ajst-9864	215	9	the	the	DET
ajst-9864	215	10	gru	gru	PROPN
ajst-9864	215	11	dualsupervised	dualsupervise	VERB
ajst-9864	215	12	model	model	NOUN
ajst-9864	215	13	is	be	AUX
ajst-9864	215	14	68.90	68.90	NUM
ajst-9864	215	15	%	%	NOUN
ajst-9864	215	16	,	,	PUNCT
ajst-9864	215	17	the	the	DET
ajst-9864	215	18	recall	recall	NOUN
ajst-9864	215	19	rate	rate	NOUN
ajst-9864	215	20	is	be	AUX
ajst-9864	215	21	50.12	50.12	NUM
ajst-9864	215	22	%	%	NOUN
ajst-9864	215	23	,	,	PUNCT
ajst-9864	215	24	and	and	CCONJ
ajst-9864	215	25	the	the	DET
ajst-9864	215	26	f1	f1	PROPN
ajst-9864	215	27	score	score	NOUN
ajst-9864	215	28	is	be	AUX
ajst-9864	215	29	57.21	57.21	NUM
ajst-9864	215	30	%	%	NOUN
ajst-9864	215	31	.	.	PUNCT
ajst-9864	216	1	this	this	PRON
ajst-9864	216	2	means	mean	VERB
ajst-9864	216	3	that	that	SCONJ
ajst-9864	216	4	the	the	DET
ajst-9864	216	5	model	model	NOUN
ajst-9864	216	6	predicts	predict	VERB
ajst-9864	216	7	correctly	correctly	ADV
ajst-9864	216	8	for	for	ADP
ajst-9864	216	9	some	some	DET
ajst-9864	216	10	samples	sample	NOUN
ajst-9864	216	11	but	but	CCONJ
ajst-9864	216	12	incorrectly	incorrectly	ADV
ajst-9864	216	13	for	for	ADP
ajst-9864	216	14	others	other	NOUN
ajst-9864	216	15	.	.	PUNCT
ajst-9864	217	1	to	to	PART
ajst-9864	217	2	improve	improve	VERB
ajst-9864	217	3	the	the	DET
ajst-9864	217	4	model	model	NOUN
ajst-9864	217	5	's	's	PART
ajst-9864	217	6	performance	performance	NOUN
ajst-9864	217	7	,	,	PUNCT
ajst-9864	217	8	future	future	ADJ
ajst-9864	217	9	work	work	NOUN
ajst-9864	217	10	can	can	AUX
ajst-9864	217	11	consider	consider	VERB
ajst-9864	217	12	improvements	improvement	NOUN
ajst-9864	217	13	in	in	ADP
ajst-9864	217	14	the	the	DET
ajst-9864	217	15	following	follow	VERB
ajst-9864	217	16	areas	area	NOUN
ajst-9864	217	17	:	:	PUNCT
ajst-9864	217	18	1	1	X
ajst-9864	217	19	)	)	PUNCT
ajst-9864	217	20	data	datum	NOUN
ajst-9864	217	21	augmentation	augmentation	NOUN
ajst-9864	217	22	:	:	PUNCT
ajst-9864	217	23	increasing	increase	VERB
ajst-9864	217	24	the	the	DET
ajst-9864	217	25	amount	amount	NOUN
ajst-9864	217	26	of	of	ADP
ajst-9864	217	27	training	training	NOUN
ajst-9864	217	28	data	datum	NOUN
ajst-9864	217	29	can	can	AUX
ajst-9864	217	30	help	help	VERB
ajst-9864	217	31	the	the	DET
ajst-9864	217	32	model	model	NOUN
ajst-9864	217	33	better	well	ADV
ajst-9864	217	34	learn	learn	VERB
ajst-9864	217	35	the	the	DET
ajst-9864	217	36	distribution	distribution	NOUN
ajst-9864	217	37	of	of	ADP
ajst-9864	217	38	the	the	DET
ajst-9864	217	39	data	datum	NOUN
ajst-9864	217	40	,	,	PUNCT
ajst-9864	217	41	thereby	thereby	ADV
ajst-9864	217	42	improving	improve	VERB
ajst-9864	217	43	its	its	PRON
ajst-9864	217	44	performance	performance	NOUN
ajst-9864	217	45	.	.	PUNCT
ajst-9864	218	1	text	text	NOUN
ajst-9864	218	2	data	datum	NOUN
ajst-9864	218	3	can	can	AUX
ajst-9864	218	4	be	be	AUX
ajst-9864	218	5	randomly	randomly	ADV
ajst-9864	218	6	transformed	transform	VERB
ajst-9864	218	7	(	(	PUNCT
ajst-9864	218	8	such	such	ADJ
ajst-9864	218	9	as	as	ADP
ajst-9864	218	10	by	by	ADP
ajst-9864	218	11	replacement	replacement	NOUN
ajst-9864	218	12	,	,	PUNCT
ajst-9864	218	13	insertion	insertion	NOUN
ajst-9864	218	14	,	,	PUNCT
ajst-9864	218	15	deletion	deletion	NOUN
ajst-9864	218	16	,	,	PUNCT
ajst-9864	218	17	etc	etc	X
ajst-9864	218	18	.	.	X
ajst-9864	218	19	)	)	PUNCT
ajst-9864	218	20	to	to	PART
ajst-9864	218	21	increase	increase	VERB
ajst-9864	218	22	data	datum	NOUN
ajst-9864	218	23	diversity	diversity	NOUN
ajst-9864	218	24	.	.	PUNCT
ajst-9864	219	1	2	2	X
ajst-9864	219	2	)	)	PUNCT
ajst-9864	219	3	model	model	NOUN
ajst-9864	219	4	improvement	improvement	NOUN
ajst-9864	219	5	:	:	PUNCT
ajst-9864	219	6	more	more	ADJ
ajst-9864	219	7	complex	complex	ADJ
ajst-9864	219	8	models	model	NOUN
ajst-9864	219	9	can	can	AUX
ajst-9864	219	10	be	be	AUX
ajst-9864	219	11	tried	try	VERB
ajst-9864	219	12	,	,	PUNCT
ajst-9864	219	13	such	such	ADJ
ajst-9864	219	14	as	as	ADP
ajst-9864	219	15	using	use	VERB
ajst-9864	219	16	more	more	ADJ
ajst-9864	219	17	gru	gru	NOUN
ajst-9864	219	18	layers	layer	NOUN
ajst-9864	219	19	or	or	CCONJ
ajst-9864	219	20	convolutional	convolutional	ADJ
ajst-9864	219	21	neural	neural	ADJ
ajst-9864	219	22	networks	network	NOUN
ajst-9864	219	23	.	.	PUNCT
ajst-9864	220	1	in	in	ADP
ajst-9864	220	2	addition	addition	NOUN
ajst-9864	220	3	,	,	PUNCT
ajst-9864	220	4	pre	pre	ADJ
ajst-9864	220	5	-	-	ADJ
ajst-9864	220	6	trained	train	VERB
ajst-9864	220	7	word	word	NOUN
ajst-9864	220	8	vectors	vector	NOUN
ajst-9864	220	9	can	can	AUX
ajst-9864	220	10	be	be	AUX
ajst-9864	220	11	used	use	VERB
ajst-9864	220	12	instead	instead	ADV
ajst-9864	220	13	of	of	ADP
ajst-9864	220	14	randomly	randomly	ADV
ajst-9864	220	15	initialised	initialise	VERB
ajst-9864	220	16	word	word	NOUN
ajst-9864	220	17	vectors	vector	NOUN
ajst-9864	220	18	.	.	PUNCT
ajst-9864	221	1	3	3	X
ajst-9864	221	2	)	)	PUNCT
ajst-9864	221	3	hyperparameter	hyperparameter	NOUN
ajst-9864	221	4	tuning	tuning	NOUN
ajst-9864	221	5	:	:	PUNCT
ajst-9864	221	6	cross	cross	ADJ
ajst-9864	221	7	-	-	ADJ
ajst-9864	221	8	validation	validation	ADJ
ajst-9864	221	9	and	and	CCONJ
ajst-9864	221	10	other	other	ADJ
ajst-9864	221	11	techniques	technique	NOUN
ajst-9864	221	12	can	can	AUX
ajst-9864	221	13	be	be	AUX
ajst-9864	221	14	used	use	VERB
ajst-9864	221	15	to	to	PART
ajst-9864	221	16	optimize	optimize	VERB
ajst-9864	221	17	the	the	DET
ajst-9864	221	18	model	model	NOUN
ajst-9864	221	19	's	's	PART
ajst-9864	221	20	hyperparameters	hyperparameter	NOUN
ajst-9864	221	21	,	,	PUNCT
ajst-9864	221	22	such	such	ADJ
ajst-9864	221	23	as	as	ADP
ajst-9864	221	24	learning	learn	VERB
ajst-9864	221	25	rate	rate	NOUN
ajst-9864	221	26	,	,	PUNCT
ajst-9864	221	27	batch	batch	NOUN
ajst-9864	221	28	size	size	NOUN
ajst-9864	221	29	,	,	PUNCT
ajst-9864	221	30	etc	etc	X
ajst-9864	221	31	.	.	X
ajst-9864	221	32	4	4	X
ajst-9864	221	33	)	)	PUNCT
ajst-9864	221	34	model	model	NOUN
ajst-9864	221	35	fusion	fusion	NOUN
ajst-9864	221	36	:	:	PUNCT
ajst-9864	221	37	multiple	multiple	ADJ
ajst-9864	221	38	models	model	NOUN
ajst-9864	221	39	can	can	AUX
ajst-9864	221	40	be	be	AUX
ajst-9864	221	41	integrated	integrate	VERB
ajst-9864	221	42	to	to	PART
ajst-9864	221	43	achieve	achieve	VERB
ajst-9864	221	44	better	well	ADJ
ajst-9864	221	45	performance	performance	NOUN
ajst-9864	221	46	.	.	PUNCT
ajst-9864	222	1	different	different	ADJ
ajst-9864	222	2	types	type	NOUN
ajst-9864	222	3	of	of	ADP
ajst-9864	222	4	models	model	NOUN
ajst-9864	222	5	,	,	PUNCT
ajst-9864	222	6	such	such	ADJ
ajst-9864	222	7	as	as	ADP
ajst-9864	222	8	rnn	rnn	PROPN
ajst-9864	222	9	,	,	PUNCT
ajst-9864	222	10	cnn	cnn	PROPN
ajst-9864	222	11	,	,	PUNCT
ajst-9864	222	12	and	and	CCONJ
ajst-9864	222	13	transformer	transformer	NOUN
ajst-9864	222	14	,	,	PUNCT
ajst-9864	222	15	can	can	AUX
ajst-9864	222	16	be	be	AUX
ajst-9864	222	17	fused	fuse	VERB
ajst-9864	222	18	,	,	PUNCT
ajst-9864	222	19	for	for	ADP
ajst-9864	222	20	example	example	NOUN
ajst-9864	222	21	.	.	PUNCT
ajst-9864	223	1	error	error	NOUN
ajst-9864	223	2	analysis	analysis	NOUN
ajst-9864	223	3	:	:	PUNCT
ajst-9864	223	4	carefully	carefully	ADV
ajst-9864	223	5	analyzing	analyze	VERB
ajst-9864	223	6	samples	sample	NOUN
ajst-9864	223	7	that	that	PRON
ajst-9864	223	8	the	the	DET
ajst-9864	223	9	model	model	NOUN
ajst-9864	223	10	predicts	predict	VERB
ajst-9864	223	11	incorrectly	incorrectly	ADV
ajst-9864	223	12	to	to	PART
ajst-9864	223	13	understand	understand	VERB
ajst-9864	223	14	the	the	DET
ajst-9864	223	15	model	model	NOUN
ajst-9864	223	16	's	's	PART
ajst-9864	223	17	weaknesses	weakness	NOUN
ajst-9864	223	18	and	and	CCONJ
ajst-9864	223	19	further	far	ADV
ajst-9864	223	20	improve	improve	VERB
ajst-9864	223	21	the	the	DET
ajst-9864	223	22	model	model	NOUN
ajst-9864	223	23	.	.	PUNCT
ajst-9864	224	1	5.2	5.2	NUM
ajst-9864	224	2	.	.	PUNCT
ajst-9864	225	1	future	future	ADJ
ajst-9864	225	2	work	work	NOUN
ajst-9864	225	3	in	in	ADP
ajst-9864	225	4	conclusion	conclusion	NOUN
ajst-9864	225	5	,	,	PUNCT
ajst-9864	225	6	based	base	VERB
ajst-9864	225	7	on	on	ADP
ajst-9864	225	8	the	the	DET
ajst-9864	225	9	research	research	NOUN
ajst-9864	225	10	goals	goal	NOUN
ajst-9864	225	11	and	and	CCONJ
ajst-9864	225	12	problems	problem	NOUN
ajst-9864	225	13	,	,	PUNCT
ajst-9864	225	14	this	this	DET
ajst-9864	225	15	study	study	NOUN
ajst-9864	225	16	aims	aim	VERB
ajst-9864	225	17	to	to	PART
ajst-9864	225	18	explore	explore	VERB
ajst-9864	225	19	suitable	suitable	ADJ
ajst-9864	225	20	natural	natural	ADJ
ajst-9864	225	21	language	language	NOUN
ajst-9864	225	22	processing	processing	NOUN
ajst-9864	225	23	algorithm	algorithm	NOUN
ajst-9864	225	24	models	model	NOUN
ajst-9864	225	25	for	for	ADP
ajst-9864	225	26	chinese	chinese	ADJ
ajst-9864	225	27	text	text	NOUN
ajst-9864	225	28	relationship	relationship	NOUN
ajst-9864	225	29	analysis	analysis	NOUN
ajst-9864	225	30	and	and	CCONJ
ajst-9864	225	31	establish	establish	VERB
ajst-9864	225	32	a	a	DET
ajst-9864	225	33	bi	bi	ADJ
ajst-9864	225	34	-	-	ADJ
ajst-9864	225	35	gru	gru	NOUN
ajst-9864	225	36	bidirectional	bidirectional	ADJ
ajst-9864	225	37	neural	neural	ADJ
ajst-9864	225	38	network	network	NOUN
ajst-9864	225	39	algorithm	algorithm	NOUN
ajst-9864	225	40	model	model	NOUN
ajst-9864	225	41	to	to	PART
ajst-9864	225	42	detect	detect	VERB
ajst-9864	225	43	character	character	NOUN
ajst-9864	225	44	relationships	relationship	NOUN
ajst-9864	225	45	.	.	PUNCT
ajst-9864	226	1	by	by	ADP
ajst-9864	226	2	comparing	compare	VERB
ajst-9864	226	3	and	and	CCONJ
ajst-9864	226	4	analysing	analyse	VERB
ajst-9864	226	5	the	the	DET
ajst-9864	226	6	ordinary	ordinary	ADJ
ajst-9864	226	7	gru	gru	NOUN
ajst-9864	226	8	model	model	NOUN
ajst-9864	226	9	and	and	CCONJ
ajst-9864	226	10	the	the	DET
ajst-9864	226	11	bigru	bigru	NOUN
ajst-9864	226	12	algorithm	algorithm	NOUN
ajst-9864	226	13	model	model	NOUN
ajst-9864	226	14	,	,	PUNCT
ajst-9864	226	15	it	it	PRON
ajst-9864	226	16	was	be	AUX
ajst-9864	226	17	found	find	VERB
ajst-9864	226	18	that	that	SCONJ
ajst-9864	226	19	the	the	DET
ajst-9864	226	20	bi	bi	PROPN
ajst-9864	226	21	-	-	PROPN
ajst-9864	226	22	gru	gru	NOUN
ajst-9864	226	23	bidirectional	bidirectional	ADJ
ajst-9864	226	24	neural	neural	ADJ
ajst-9864	226	25	network	network	NOUN
ajst-9864	226	26	algorithm	algorithm	NOUN
ajst-9864	226	27	model	model	NOUN
ajst-9864	226	28	performed	perform	VERB
ajst-9864	226	29	best	well	ADV
ajst-9864	226	30	in	in	ADP
ajst-9864	226	31	chinese	chinese	ADJ
ajst-9864	226	32	text	text	NOUN
ajst-9864	226	33	relationship	relationship	NOUN
ajst-9864	226	34	analysis	analysis	NOUN
ajst-9864	226	35	.	.	PUNCT
ajst-9864	227	1	it	it	PRON
ajst-9864	227	2	has	have	VERB
ajst-9864	227	3	a	a	DET
ajst-9864	227	4	high	high	ADJ
ajst-9864	227	5	accuracy	accuracy	NOUN
ajst-9864	227	6	and	and	CCONJ
ajst-9864	227	7	recall	recall	NOUN
ajst-9864	227	8	rate	rate	NOUN
ajst-9864	227	9	.	.	PUNCT
ajst-9864	228	1	when	when	SCONJ
ajst-9864	228	2	it	it	PRON
ajst-9864	228	3	comes	come	VERB
ajst-9864	228	4	to	to	ADP
ajst-9864	228	5	analysing	analyse	VERB
ajst-9864	228	6	the	the	DET
ajst-9864	228	7	relationships	relationship	NOUN
ajst-9864	228	8	between	between	ADP
ajst-9864	228	9	chinese	chinese	ADJ
ajst-9864	228	10	characters	character	NOUN
ajst-9864	228	11	,	,	PUNCT
ajst-9864	228	12	the	the	DET
ajst-9864	228	13	bi	bi	PROPN
ajst-9864	228	14	-	-	PROPN
ajst-9864	228	15	gru	gru	NOUN
ajst-9864	228	16	bidirectional	bidirectional	ADJ
ajst-9864	228	17	neural	neural	ADJ
ajst-9864	228	18	network	network	NOUN
ajst-9864	228	19	algorithm	algorithm	NOUN
ajst-9864	228	20	model	model	NOUN
ajst-9864	228	21	achieves	achieve	VERB
ajst-9864	228	22	the	the	DET
ajst-9864	228	23	highest	high	ADJ
ajst-9864	228	24	levels	level	NOUN
ajst-9864	228	25	of	of	ADP
ajst-9864	228	26	accuracy	accuracy	NOUN
ajst-9864	228	27	and	and	CCONJ
ajst-9864	228	28	recall	recall	NOUN
ajst-9864	228	29	.	.	PUNCT
ajst-9864	229	1	using	use	VERB
ajst-9864	229	2	a	a	DET
ajst-9864	229	3	combination	combination	NOUN
ajst-9864	229	4	of	of	ADP
ajst-9864	229	5	the	the	DET
ajst-9864	229	6	bi	bi	PROPN
ajst-9864	229	7	-	-	PROPN
ajst-9864	229	8	gru	gru	NOUN
ajst-9864	229	9	model	model	NOUN
ajst-9864	229	10	and	and	CCONJ
ajst-9864	229	11	the	the	DET
ajst-9864	229	12	double	double	ADJ
ajst-9864	229	13	-	-	PUNCT
ajst-9864	229	14	layer	layer	NOUN
ajst-9864	229	15	attention	attention	NOUN
ajst-9864	229	16	mechanism	mechanism	NOUN
ajst-9864	229	17	,	,	PUNCT
ajst-9864	229	18	the	the	DET
ajst-9864	229	19	model	model	NOUN
ajst-9864	229	20	can	can	AUX
ajst-9864	229	21	more	more	ADV
ajst-9864	229	22	accurately	accurately	ADV
ajst-9864	229	23	capture	capture	VERB
ajst-9864	229	24	the	the	DET
ajst-9864	229	25	feature	feature	NOUN
ajst-9864	229	26	information	information	NOUN
ajst-9864	229	27	of	of	ADP
ajst-9864	229	28	chinese	chinese	ADJ
ajst-9864	229	29	phrases	phrase	NOUN
ajst-9864	229	30	,	,	PUNCT
ajst-9864	229	31	giving	give	VERB
ajst-9864	229	32	more	more	ADJ
ajst-9864	229	33	weight	weight	NOUN
ajst-9864	229	34	to	to	ADP
ajst-9864	229	35	the	the	DET
ajst-9864	229	36	most	most	ADV
ajst-9864	229	37	successful	successful	ADJ
ajst-9864	229	38	sentences	sentence	NOUN
ajst-9864	229	39	via	via	ADP
ajst-9864	229	40	continuous	continuous	ADJ
ajst-9864	229	41	learning	learning	NOUN
ajst-9864	229	42	while	while	SCONJ
ajst-9864	229	43	dampening	dampen	VERB
ajst-9864	229	44	the	the	DET
ajst-9864	229	45	influence	influence	NOUN
ajst-9864	229	46	of	of	ADP
ajst-9864	229	47	noisy	noisy	ADJ
ajst-9864	229	48	data	datum	NOUN
ajst-9864	229	49	.	.	PUNCT
ajst-9864	230	1	our	our	PRON
ajst-9864	230	2	experimental	experimental	ADJ
ajst-9864	230	3	results	result	NOUN
ajst-9864	230	4	show	show	VERB
ajst-9864	230	5	that	that	SCONJ
ajst-9864	230	6	our	our	PRON
ajst-9864	230	7	model	model	NOUN
ajst-9864	230	8	achieved	achieve	VERB
ajst-9864	230	9	68.90	68.90	NUM
ajst-9864	230	10	%	%	NOUN
ajst-9864	230	11	accuracy	accuracy	NOUN
ajst-9864	230	12	,	,	PUNCT
ajst-9864	230	13	a	a	DET
ajst-9864	230	14	50.12	50.12	NUM
ajst-9864	230	15	%	%	NOUN
ajst-9864	230	16	recall	recall	NOUN
ajst-9864	230	17	rate	rate	NOUN
ajst-9864	230	18	,	,	PUNCT
ajst-9864	230	19	and	and	CCONJ
ajst-9864	230	20	a	a	DET
ajst-9864	230	21	57.21	57.21	NUM
ajst-9864	230	22	%	%	NOUN
ajst-9864	230	23	f1	f1	NOUN
ajst-9864	230	24	value	value	NOUN
ajst-9864	230	25	on	on	ADP
ajst-9864	230	26	the	the	DET
ajst-9864	230	27	dataset	dataset	NOUN
ajst-9864	230	28	,	,	PUNCT
ajst-9864	230	29	achieving	achieve	VERB
ajst-9864	230	30	better	well	ADJ
ajst-9864	230	31	performance	performance	NOUN
ajst-9864	230	32	compared	compare	VERB
ajst-9864	230	33	to	to	ADP
ajst-9864	230	34	other	other	ADJ
ajst-9864	230	35	nlp	nlp	ADJ
ajst-9864	230	36	algorithm	algorithm	NOUN
ajst-9864	230	37	models	model	NOUN
ajst-9864	230	38	.	.	PUNCT
ajst-9864	231	1	we	we	PRON
ajst-9864	231	2	believe	believe	VERB
ajst-9864	231	3	that	that	SCONJ
ajst-9864	231	4	chinese	chinese	ADJ
ajst-9864	231	5	text	text	NOUN
ajst-9864	231	6	relationship	relationship	NOUN
ajst-9864	231	7	analysis	analysis	NOUN
ajst-9864	231	8	is	be	AUX
ajst-9864	231	9	of	of	ADP
ajst-9864	231	10	significant	significant	ADJ
ajst-9864	231	11	importance	importance	NOUN
ajst-9864	231	12	in	in	ADP
ajst-9864	231	13	understanding	understand	VERB
ajst-9864	231	14	the	the	DET
ajst-9864	231	15	relationships	relationship	NOUN
ajst-9864	231	16	between	between	ADP
ajst-9864	231	17	chinese	chinese	ADJ
ajst-9864	231	18	texts	text	NOUN
ajst-9864	231	19	and	and	CCONJ
ajst-9864	231	20	improving	improve	VERB
ajst-9864	231	21	the	the	DET
ajst-9864	231	22	efficiency	efficiency	NOUN
ajst-9864	231	23	and	and	CCONJ
ajst-9864	231	24	accuracy	accuracy	NOUN
ajst-9864	231	25	of	of	ADP
ajst-9864	231	26	text	text	NOUN
ajst-9864	231	27	information	information	NOUN
ajst-9864	231	28	processing	processing	NOUN
ajst-9864	231	29	.	.	PUNCT
ajst-9864	232	1	in	in	ADP
ajst-9864	232	2	future	future	ADJ
ajst-9864	232	3	work	work	NOUN
ajst-9864	232	4	,	,	PUNCT
ajst-9864	232	5	we	we	PRON
ajst-9864	232	6	can	can	AUX
ajst-9864	232	7	consider	consider	VERB
ajst-9864	232	8	increasing	increase	VERB
ajst-9864	232	9	the	the	DET
ajst-9864	232	10	amount	amount	NOUN
ajst-9864	232	11	of	of	ADP
ajst-9864	232	12	training	training	NOUN
ajst-9864	232	13	data	datum	NOUN
ajst-9864	232	14	,	,	PUNCT
ajst-9864	232	15	trying	try	VERB
ajst-9864	232	16	more	more	ADJ
ajst-9864	232	17	complex	complex	ADJ
ajst-9864	232	18	models	model	NOUN
ajst-9864	232	19	,	,	PUNCT
ajst-9864	232	20	optimizing	optimize	VERB
ajst-9864	232	21	hyperparameters	hyperparameter	NOUN
ajst-9864	232	22	,	,	PUNCT
ajst-9864	232	23	and	and	CCONJ
ajst-9864	232	24	exploring	explore	VERB
ajst-9864	232	25	the	the	DET
ajst-9864	232	26	application	application	NOUN
ajst-9864	232	27	of	of	ADP
ajst-9864	232	28	other	other	ADJ
ajst-9864	232	29	nlp	nlp	NOUN
ajst-9864	232	30	technologies	technology	NOUN
ajst-9864	232	31	to	to	ADP
ajst-9864	232	32	chinese	chinese	ADJ
ajst-9864	232	33	text	text	NOUN
ajst-9864	232	34	relationship	relationship	NOUN
ajst-9864	232	35	analysis	analysis	NOUN
ajst-9864	232	36	,	,	PUNCT
ajst-9864	232	37	such	such	ADJ
ajst-9864	232	38	as	as	ADP
ajst-9864	232	39	bert	bert	PROPN
ajst-9864	232	40	pre	pre	ADJ
ajst-9864	232	41	-	-	ADJ
ajst-9864	232	42	training	training	ADJ
ajst-9864	232	43	models	model	NOUN
ajst-9864	232	44	,	,	PUNCT
ajst-9864	232	45	to	to	PART
ajst-9864	232	46	further	far	ADV
ajst-9864	232	47	improve	improve	VERB
ajst-9864	232	48	model	model	NOUN
ajst-9864	232	49	performance	performance	NOUN
ajst-9864	232	50	.	.	PUNCT
ajst-9864	233	1	6	6	X
ajst-9864	233	2	.	.	X
ajst-9864	233	3	conclusion	conclusion	NOUN
ajst-9864	233	4	in	in	ADP
ajst-9864	233	5	the	the	DET
ajst-9864	233	6	age	age	NOUN
ajst-9864	233	7	of	of	ADP
ajst-9864	233	8	big	big	ADJ
ajst-9864	233	9	data	datum	NOUN
ajst-9864	233	10	,	,	PUNCT
ajst-9864	233	11	this	this	DET
ajst-9864	233	12	article	article	NOUN
ajst-9864	233	13	explores	explore	VERB
ajst-9864	233	14	how	how	SCONJ
ajst-9864	233	15	to	to	PART
ajst-9864	233	16	consistently	consistently	ADV
ajst-9864	233	17	and	and	CCONJ
ajst-9864	233	18	accurately	accurately	ADV
ajst-9864	233	19	extract	extract	VERB
ajst-9864	233	20	relationships	relationship	NOUN
ajst-9864	233	21	between	between	ADP
ajst-9864	233	22	chinese	chinese	ADJ
ajst-9864	233	23	text	text	NOUN
ajst-9864	233	24	entities	entity	NOUN
ajst-9864	233	25	and	and	CCONJ
ajst-9864	233	26	offers	offer	VERB
ajst-9864	233	27	some	some	DET
ajst-9864	233	28	tips	tip	NOUN
ajst-9864	233	29	and	and	CCONJ
ajst-9864	233	30	suggestions	suggestion	NOUN
ajst-9864	233	31	to	to	PART
ajst-9864	233	32	increase	increase	VERB
ajst-9864	233	33	the	the	DET
ajst-9864	233	34	precision	precision	NOUN
ajst-9864	233	35	of	of	ADP
ajst-9864	233	36	relationship	relationship	NOUN
ajst-9864	233	37	extraction	extraction	NOUN
ajst-9864	233	38	models	model	NOUN
ajst-9864	233	39	.	.	PUNCT
ajst-9864	234	1	chinese	chinese	ADJ
ajst-9864	234	2	text	text	NOUN
ajst-9864	234	3	analysis	analysis	NOUN
ajst-9864	234	4	in	in	ADP
ajst-9864	234	5	the	the	DET
ajst-9864	234	6	context	context	NOUN
ajst-9864	234	7	of	of	ADP
ajst-9864	234	8	big	big	ADJ
ajst-9864	234	9	data	datum	NOUN
ajst-9864	234	10	is	be	AUX
ajst-9864	234	11	thoroughly	thoroughly	ADV
ajst-9864	234	12	examined	examine	VERB
ajst-9864	234	13	from	from	ADP
ajst-9864	234	14	the	the	DET
ajst-9864	234	15	perspectives	perspective	NOUN
ajst-9864	234	16	of	of	ADP
ajst-9864	234	17	management	management	NOUN
ajst-9864	234	18	,	,	PUNCT
ajst-9864	234	19	technical	technical	ADJ
ajst-9864	234	20	development	development	NOUN
ajst-9864	234	21	,	,	PUNCT
ajst-9864	234	22	and	and	CCONJ
ajst-9864	234	23	information	information	NOUN
ajst-9864	234	24	application	application	NOUN
ajst-9864	234	25	based	base	VERB
ajst-9864	234	26	on	on	ADP
ajst-9864	234	27	pertinent	pertinent	ADJ
ajst-9864	234	28	literature	literature	NOUN
ajst-9864	234	29	and	and	CCONJ
ajst-9864	234	30	theory	theory	NOUN
ajst-9864	234	31	.	.	PUNCT
ajst-9864	235	1	this	this	DET
ajst-9864	235	2	article	article	NOUN
ajst-9864	235	3	's	's	PART
ajst-9864	235	4	thorough	thorough	ADJ
ajst-9864	235	5	research	research	NOUN
ajst-9864	235	6	attempts	attempt	VERB
ajst-9864	235	7	to	to	PART
ajst-9864	235	8	build	build	VERB
ajst-9864	235	9	a	a	DET
ajst-9864	235	10	dual	dual	ADJ
ajst-9864	235	11	attention	attention	NOUN
ajst-9864	235	12	mechanism	mechanism	NOUN
ajst-9864	235	13	and	and	CCONJ
ajst-9864	235	14	optimize	optimize	VERB
ajst-9864	235	15	the	the	DET
ajst-9864	235	16	chinese	chinese	ADJ
ajst-9864	235	17	text	text	NOUN
ajst-9864	235	18	extraction	extraction	NOUN
ajst-9864	235	19	model	model	NOUN
ajst-9864	235	20	in	in	ADP
ajst-9864	235	21	order	order	NOUN
ajst-9864	235	22	to	to	PART
ajst-9864	235	23	increase	increase	VERB
ajst-9864	235	24	the	the	DET
ajst-9864	235	25	effectiveness	effectiveness	NOUN
ajst-9864	235	26	of	of	ADP
ajst-9864	235	27	chinese	chinese	ADJ
ajst-9864	235	28	text	text	NOUN
ajst-9864	235	29	data	datum	NOUN
ajst-9864	235	30	extraction	extraction	NOUN
ajst-9864	235	31	.	.	PUNCT
ajst-9864	236	1	in	in	ADP
ajst-9864	236	2	the	the	DET
ajst-9864	236	3	era	era	NOUN
ajst-9864	236	4	of	of	ADP
ajst-9864	236	5	big	big	ADJ
ajst-9864	236	6	data	datum	NOUN
ajst-9864	236	7	,	,	PUNCT
ajst-9864	236	8	by	by	ADP
ajst-9864	236	9	mining	mine	VERB
ajst-9864	236	10	chinese	chinese	ADJ
ajst-9864	236	11	text	text	NOUN
ajst-9864	236	12	relationships	relationship	NOUN
ajst-9864	236	13	,	,	PUNCT
ajst-9864	236	14	it	it	PRON
ajst-9864	236	15	can	can	AUX
ajst-9864	236	16	be	be	AUX
ajst-9864	236	17	applied	apply	VERB
ajst-9864	236	18	to	to	ADP
ajst-9864	236	19	building	build	VERB
ajst-9864	236	20	intelligent	intelligent	ADJ
ajst-9864	236	21	semantic	semantic	ADJ
ajst-9864	236	22	search	search	NOUN
ajst-9864	236	23	applications	application	NOUN
ajst-9864	236	24	,	,	PUNCT
ajst-9864	236	25	knowledge	knowledge	NOUN
ajst-9864	236	26	graph	graph	NOUN
ajst-9864	236	27	construction	construction	NOUN
ajst-9864	236	28	,	,	PUNCT
ajst-9864	236	29	question	question	NOUN
ajst-9864	236	30	-	-	PUNCT
ajst-9864	236	31	answering	answer	VERB
ajst-9864	236	32	systems	system	NOUN
ajst-9864	236	33	,	,	PUNCT
ajst-9864	236	34	etc	etc	X
ajst-9864	236	35	.	.	X
ajst-9864	237	1	at	at	ADP
ajst-9864	237	2	the	the	DET
ajst-9864	237	3	same	same	ADJ
ajst-9864	237	4	time	time	NOUN
ajst-9864	237	5	,	,	PUNCT
ajst-9864	237	6	the	the	DET
ajst-9864	237	7	mining	mining	NOUN
ajst-9864	237	8	and	and	CCONJ
ajst-9864	237	9	analysis	analysis	NOUN
ajst-9864	237	10	of	of	ADP
ajst-9864	237	11	character	character	NOUN
ajst-9864	237	12	relationships	relationship	NOUN
ajst-9864	237	13	in	in	ADP
ajst-9864	237	14	chinese	chinese	ADJ
ajst-9864	237	15	text	text	NOUN
ajst-9864	237	16	have	have	VERB
ajst-9864	237	17	a	a	DET
ajst-9864	237	18	positive	positive	ADJ
ajst-9864	237	19	and	and	CCONJ
ajst-9864	237	20	far	far	ADV
ajst-9864	237	21	-	-	PUNCT
ajst-9864	237	22	reaching	reach	VERB
ajst-9864	237	23	impact	impact	NOUN
ajst-9864	237	24	on	on	ADP
ajst-9864	237	25	the	the	DET
ajst-9864	237	26	accuracy	accuracy	NOUN
ajst-9864	237	27	of	of	ADP
ajst-9864	237	28	recommendation	recommendation	NOUN
ajst-9864	237	29	systems	system	NOUN
ajst-9864	237	30	and	and	CCONJ
ajst-9864	237	31	advertising	advertising	NOUN
ajst-9864	237	32	placement	placement	NOUN
ajst-9864	237	33	.	.	PUNCT
ajst-9864	238	1	in	in	ADP
ajst-9864	238	2	order	order	NOUN
ajst-9864	238	3	to	to	PART
ajst-9864	238	4	provide	provide	VERB
ajst-9864	238	5	additional	additional	ADJ
ajst-9864	238	6	feature	feature	NOUN
ajst-9864	238	7	information	information	NOUN
ajst-9864	238	8	for	for	ADP
ajst-9864	238	9	relationship	relationship	NOUN
ajst-9864	238	10	recognition	recognition	NOUN
ajst-9864	238	11	,	,	PUNCT
ajst-9864	238	12	this	this	DET
ajst-9864	238	13	article	article	NOUN
ajst-9864	238	14	employs	employ	VERB
ajst-9864	238	15	entity	entity	NOUN
ajst-9864	238	16	location	location	NOUN
ajst-9864	238	17	information	information	NOUN
ajst-9864	238	18	and	and	CCONJ
ajst-9864	238	19	entity	entity	NOUN
ajst-9864	238	20	tag	tag	NOUN
ajst-9864	238	21	information	information	NOUN
ajst-9864	238	22	to	to	PART
ajst-9864	238	23	broaden	broaden	VERB
ajst-9864	238	24	the	the	DET
ajst-9864	238	25	term	term	NOUN
ajst-9864	238	26	"	"	PUNCT
ajst-9864	238	27	vector	vector	NOUN
ajst-9864	238	28	features	feature	NOUN
ajst-9864	238	29	"	"	PUNCT
ajst-9864	238	30	and	and	CCONJ
ajst-9864	238	31	vectorize	vectorize	VERB
ajst-9864	238	32	the	the	DET
ajst-9864	238	33	text	text	NOUN
ajst-9864	238	34	information	information	NOUN
ajst-9864	238	35	.	.	PUNCT
ajst-9864	239	1	the	the	DET
ajst-9864	239	2	bidirectional	bidirectional	PROPN
ajst-9864	239	3	gru	gru	PROPN
ajst-9864	239	4	network	network	NOUN
ajst-9864	239	5	can	can	AUX
ajst-9864	239	6	be	be	AUX
ajst-9864	239	7	strengthened	strengthen	VERB
ajst-9864	239	8	with	with	ADP
ajst-9864	239	9	the	the	DET
ajst-9864	239	10	addition	addition	NOUN
ajst-9864	239	11	of	of	ADP
ajst-9864	239	12	a	a	DET
ajst-9864	239	13	character	character	NOUN
ajst-9864	239	14	-	-	PUNCT
ajst-9864	239	15	level	level	NOUN
ajst-9864	239	16	attention	attention	NOUN
ajst-9864	239	17	layer	layer	NOUN
ajst-9864	239	18	and	and	CCONJ
ajst-9864	239	19	a	a	DET
ajst-9864	239	20	sentence	sentence	NOUN
ajst-9864	239	21	-	-	PUNCT
ajst-9864	239	22	level	level	NOUN
ajst-9864	239	23	attention	attention	NOUN
ajst-9864	239	24	layer	layer	NOUN
ajst-9864	239	25	.	.	PUNCT
ajst-9864	240	1	these	these	DET
ajst-9864	240	2	additions	addition	NOUN
ajst-9864	240	3	also	also	ADV
ajst-9864	240	4	strengthen	strengthen	VERB
ajst-9864	240	5	the	the	DET
ajst-9864	240	6	influence	influence	NOUN
ajst-9864	240	7	of	of	ADP
ajst-9864	240	8	key	key	ADJ
ajst-9864	240	9	words	word	NOUN
ajst-9864	240	10	on	on	ADP
ajst-9864	240	11	the	the	DET
ajst-9864	240	12	output	output	NOUN
ajst-9864	240	13	and	and	CCONJ
ajst-9864	240	14	increase	increase	VERB
ajst-9864	240	15	noise	noise	NOUN
ajst-9864	240	16	resistance	resistance	NOUN
ajst-9864	240	17	.	.	PUNCT
ajst-9864	241	1	the	the	DET
ajst-9864	241	2	f1	f1	PROPN
ajst-9864	241	3	value	value	NOUN
ajst-9864	241	4	,	,	PUNCT
ajst-9864	241	5	recall	recall	NOUN
ajst-9864	241	6	,	,	PUNCT
ajst-9864	241	7	and	and	CCONJ
ajst-9864	241	8	precision	precision	NOUN
ajst-9864	241	9	of	of	ADP
ajst-9864	241	10	chinese	chinese	ADJ
ajst-9864	241	11	text	text	NOUN
ajst-9864	241	12	relationship	relationship	NOUN
ajst-9864	241	13	extraction	extraction	NOUN
ajst-9864	241	14	can	can	AUX
ajst-9864	241	15	all	all	PRON
ajst-9864	241	16	be	be	AUX
ajst-9864	241	17	effectively	effectively	ADV
ajst-9864	241	18	improved	improve	VERB
ajst-9864	241	19	by	by	ADP
ajst-9864	241	20	this	this	DET
ajst-9864	241	21	paper	paper	NOUN
ajst-9864	241	22	.	.	PUNCT
ajst-9864	242	1	the	the	DET
ajst-9864	242	2	high	high	ADJ
ajst-9864	242	3	-	-	PUNCT
ajst-9864	242	4	precision	precision	NOUN
ajst-9864	242	5	chinese	chinese	ADJ
ajst-9864	242	6	text	text	NOUN
ajst-9864	242	7	entity	entity	NOUN
ajst-9864	242	8	extraction	extraction	NOUN
ajst-9864	242	9	model	model	NOUN
ajst-9864	242	10	proposed	propose	VERB
ajst-9864	242	11	in	in	ADP
ajst-9864	242	12	this	this	DET
ajst-9864	242	13	article	article	NOUN
ajst-9864	242	14	has	have	VERB
ajst-9864	242	15	great	great	ADJ
ajst-9864	242	16	significance	significance	NOUN
ajst-9864	242	17	and	and	CCONJ
ajst-9864	242	18	value	value	NOUN
ajst-9864	242	19	for	for	ADP
ajst-9864	242	20	both	both	PRON
ajst-9864	242	21	society	society	NOUN
ajst-9864	242	22	and	and	CCONJ
ajst-9864	242	23	enterprises	enterprise	NOUN
ajst-9864	242	24	.	.	PUNCT
ajst-9864	243	1	as	as	ADP
ajst-9864	243	2	a	a	DET
ajst-9864	243	3	society	society	NOUN
ajst-9864	243	4	,	,	PUNCT
ajst-9864	243	5	we	we	PRON
ajst-9864	243	6	may	may	AUX
ajst-9864	243	7	benefit	benefit	VERB
ajst-9864	243	8	from	from	ADP
ajst-9864	243	9	a	a	DET
ajst-9864	243	10	deeper	deep	ADJ
ajst-9864	243	11	comprehension	comprehension	NOUN
ajst-9864	243	12	of	of	ADP
ajst-9864	243	13	textual	textual	ADJ
ajst-9864	243	14	information	information	NOUN
ajst-9864	243	15	.	.	PUNCT
ajst-9864	244	1	the	the	DET
ajst-9864	244	2	exponential	exponential	ADJ
ajst-9864	244	3	expansion	expansion	NOUN
ajst-9864	244	4	of	of	ADP
ajst-9864	244	5	internet	internet	NOUN
ajst-9864	244	6	material	material	NOUN
ajst-9864	244	7	,	,	PUNCT
ajst-9864	244	8	including	include	VERB
ajst-9864	244	9	social	social	ADJ
ajst-9864	244	10	media	medium	NOUN
ajst-9864	244	11	,	,	PUNCT
ajst-9864	244	12	news	news	NOUN
ajst-9864	244	13	reporting	reporting	NOUN
ajst-9864	244	14	,	,	PUNCT
ajst-9864	244	15	and	and	CCONJ
ajst-9864	244	16	more	more	ADV
ajst-9864	244	17	,	,	PUNCT
ajst-9864	244	18	necessitates	necessitate	VERB
ajst-9864	244	19	improved	improved	ADJ
ajst-9864	244	20	methods	method	NOUN
ajst-9864	244	21	for	for	ADP
ajst-9864	244	22	processing	processing	NOUN
ajst-9864	244	23	and	and	CCONJ
ajst-9864	244	24	analysing	analyse	VERB
ajst-9864	244	25	this	this	DET
ajst-9864	244	26	information	information	NOUN
ajst-9864	244	27	.	.	PUNCT
ajst-9864	245	1	entity	entity	NOUN
ajst-9864	245	2	recognition	recognition	NOUN
ajst-9864	245	3	is	be	AUX
ajst-9864	245	4	the	the	DET
ajst-9864	245	5	foundation	foundation	NOUN
ajst-9864	245	6	for	for	ADP
ajst-9864	245	7	many	many	ADJ
ajst-9864	245	8	nlp	nlp	NOUN
ajst-9864	245	9	tasks	task	NOUN
ajst-9864	245	10	,	,	PUNCT
ajst-9864	245	11	including	include	VERB
ajst-9864	245	12	text	text	NOUN
ajst-9864	245	13	categorization	categorization	NOUN
ajst-9864	245	14	,	,	PUNCT
ajst-9864	245	15	information	information	NOUN
ajst-9864	245	16	extraction	extraction	NOUN
ajst-9864	245	17	,	,	PUNCT
ajst-9864	245	18	and	and	CCONJ
ajst-9864	245	19	machine	machine	NOUN
ajst-9864	245	20	translation	translation	NOUN
ajst-9864	245	21	.	.	PUNCT
ajst-9864	246	1	as	as	ADP
ajst-9864	246	2	a	a	DET
ajst-9864	246	3	consequence	consequence	NOUN
ajst-9864	246	4	,	,	PUNCT
ajst-9864	246	5	the	the	DET
ajst-9864	246	6	gru	gru	NOUN
ajst-9864	246	7	text	text	NOUN
ajst-9864	246	8	entity	entity	NOUN
ajst-9864	246	9	extraction	extraction	NOUN
ajst-9864	246	10	model	model	NOUN
ajst-9864	246	11	improves	improve	VERB
ajst-9864	246	12	our	our	PRON
ajst-9864	246	13	comprehension	comprehension	NOUN
ajst-9864	246	14	of	of	ADP
ajst-9864	246	15	text	text	NOUN
ajst-9864	246	16	data	datum	NOUN
ajst-9864	246	17	,	,	PUNCT
ajst-9864	246	18	allowing	allow	VERB
ajst-9864	246	19	us	we	PRON
ajst-9864	246	20	to	to	PART
ajst-9864	246	21	get	get	VERB
ajst-9864	246	22	more	more	ADV
ajst-9864	246	23	precise	precise	ADJ
ajst-9864	246	24	outcomes	outcome	NOUN
ajst-9864	246	25	in	in	ADP
ajst-9864	246	26	a	a	DET
ajst-9864	246	27	number	number	NOUN
ajst-9864	246	28	of	of	ADP
ajst-9864	246	29	contexts	contexts	NOUN
ajst-9864	246	30	.	.	PUNCT
ajst-9864	247	1	for	for	ADP
ajst-9864	247	2	enterprises	enterprise	NOUN
ajst-9864	247	3	,	,	PUNCT
ajst-9864	247	4	it	it	PRON
ajst-9864	247	5	can	can	AUX
ajst-9864	247	6	help	help	VERB
ajst-9864	247	7	them	they	PRON
ajst-9864	247	8	better	well	ADJ
ajst-9864	247	9	process	process	NOUN
ajst-9864	247	10	and	and	CCONJ
ajst-9864	247	11	analyse	analyse	VERB
ajst-9864	247	12	massive	massive	ADJ
ajst-9864	247	13	text	text	NOUN
ajst-9864	247	14	data	datum	NOUN
ajst-9864	247	15	.	.	PUNCT
ajst-9864	248	1	enterprises	enterprise	NOUN
ajst-9864	248	2	need	need	VERB
ajst-9864	248	3	to	to	PART
ajst-9864	248	4	process	process	VERB
ajst-9864	248	5	a	a	DET
ajst-9864	248	6	large	large	ADJ
ajst-9864	248	7	amount	amount	NOUN
ajst-9864	248	8	of	of	ADP
ajst-9864	248	9	text	text	NOUN
ajst-9864	248	10	data	datum	NOUN
ajst-9864	248	11	,	,	PUNCT
ajst-9864	248	12	including	include	VERB
ajst-9864	248	13	customer	customer	NOUN
ajst-9864	248	14	feedback	feedback	NOUN
ajst-9864	248	15	,	,	PUNCT
ajst-9864	248	16	market	market	NOUN
ajst-9864	248	17	reports	report	NOUN
ajst-9864	248	18	,	,	PUNCT
ajst-9864	248	19	competitive	competitive	ADJ
ajst-9864	248	20	intelligence	intelligence	NOUN
ajst-9864	248	21	,	,	PUNCT
ajst-9864	248	22	and	and	CCONJ
ajst-9864	248	23	social	social	ADJ
ajst-9864	248	24	media	medium	NOUN
ajst-9864	248	25	comments	comment	NOUN
ajst-9864	248	26	.	.	PUNCT
ajst-9864	249	1	by	by	ADP
ajst-9864	249	2	136	136	NUM
ajst-9864	249	3	using	use	VERB
ajst-9864	249	4	high	high	ADJ
ajst-9864	249	5	-	-	PUNCT
ajst-9864	249	6	precision	precision	NOUN
ajst-9864	249	7	text	text	NOUN
ajst-9864	249	8	entity	entity	NOUN
ajst-9864	249	9	extraction	extraction	NOUN
ajst-9864	249	10	models	model	NOUN
ajst-9864	249	11	,	,	PUNCT
ajst-9864	249	12	enterprises	enterprise	NOUN
ajst-9864	249	13	can	can	AUX
ajst-9864	249	14	extract	extract	VERB
ajst-9864	249	15	entity	entity	NOUN
ajst-9864	249	16	information	information	NOUN
ajst-9864	249	17	faster	fast	ADV
ajst-9864	249	18	and	and	CCONJ
ajst-9864	249	19	more	more	ADV
ajst-9864	249	20	accurately	accurately	ADV
ajst-9864	249	21	,	,	PUNCT
ajst-9864	249	22	thus	thus	ADV
ajst-9864	249	23	better	well	ADV
ajst-9864	249	24	understanding	understand	VERB
ajst-9864	249	25	their	their	PRON
ajst-9864	249	26	customers	customer	NOUN
ajst-9864	249	27	,	,	PUNCT
ajst-9864	249	28	markets	market	NOUN
ajst-9864	249	29	,	,	PUNCT
ajst-9864	249	30	and	and	CCONJ
ajst-9864	249	31	competitors	competitor	NOUN
ajst-9864	249	32	.	.	PUNCT
ajst-9864	250	1	in	in	ADP
ajst-9864	250	2	addition	addition	NOUN
ajst-9864	250	3	,	,	PUNCT
ajst-9864	250	4	entity	entity	NOUN
ajst-9864	250	5	recognition	recognition	NOUN
ajst-9864	250	6	can	can	AUX
ajst-9864	250	7	also	also	ADV
ajst-9864	250	8	help	help	VERB
ajst-9864	250	9	enterprises	enterprise	NOUN
ajst-9864	250	10	automate	automate	VERB
ajst-9864	250	11	many	many	ADJ
ajst-9864	250	12	repetitive	repetitive	ADJ
ajst-9864	250	13	tasks	task	NOUN
ajst-9864	250	14	,	,	PUNCT
ajst-9864	250	15	thereby	thereby	ADV
ajst-9864	250	16	improving	improve	VERB
ajst-9864	250	17	efficiency	efficiency	NOUN
ajst-9864	250	18	and	and	CCONJ
ajst-9864	250	19	reducing	reduce	VERB
ajst-9864	250	20	costs	cost	NOUN
ajst-9864	250	21	.	.	PUNCT
ajst-9864	251	1	the	the	DET
ajst-9864	251	2	gru	gru	PROPN
ajst-9864	251	3	bidirectional	bidirectional	NOUN
ajst-9864	251	4	supervised	supervise	VERB
ajst-9864	251	5	text	text	NOUN
ajst-9864	251	6	entity	entity	NOUN
ajst-9864	251	7	extraction	extraction	NOUN
ajst-9864	251	8	model	model	NOUN
ajst-9864	251	9	suggested	suggest	VERB
ajst-9864	251	10	in	in	ADP
ajst-9864	251	11	this	this	DET
ajst-9864	251	12	paper	paper	NOUN
ajst-9864	251	13	has	have	VERB
ajst-9864	251	14	a	a	DET
ajst-9864	251	15	very	very	ADV
ajst-9864	251	16	high	high	ADJ
ajst-9864	251	17	value	value	NOUN
ajst-9864	251	18	,	,	PUNCT
ajst-9864	251	19	to	to	PART
ajst-9864	251	20	sum	sum	VERB
ajst-9864	251	21	it	it	PRON
ajst-9864	251	22	up	up	ADP
ajst-9864	251	23	.	.	PUNCT
ajst-9864	252	1	it	it	PRON
ajst-9864	252	2	can	can	AUX
ajst-9864	252	3	assist	assist	VERB
ajst-9864	252	4	in	in	ADP
ajst-9864	252	5	improving	improve	VERB
ajst-9864	252	6	the	the	DET
ajst-9864	252	7	accuracy	accuracy	NOUN
ajst-9864	252	8	and	and	CCONJ
ajst-9864	252	9	effectiveness	effectiveness	NOUN
ajst-9864	252	10	of	of	ADP
ajst-9864	252	11	natural	natural	ADJ
ajst-9864	252	12	language	language	NOUN
ajst-9864	252	13	processing	processing	NOUN
ajst-9864	252	14	technologies	technology	NOUN
ajst-9864	252	15	by	by	ADP
ajst-9864	252	16	helping	help	VERB
ajst-9864	252	17	us	we	PRON
ajst-9864	252	18	process	process	VERB
ajst-9864	252	19	natural	natural	ADJ
ajst-9864	252	20	language	language	NOUN
ajst-9864	252	21	data	datum	NOUN
ajst-9864	252	22	more	more	ADV
ajst-9864	252	23	effectively	effectively	ADV
ajst-9864	252	24	.	.	PUNCT
ajst-9864	253	1	we	we	PRON
ajst-9864	253	2	need	need	VERB
ajst-9864	253	3	more	more	ADV
ajst-9864	253	4	effective	effective	ADJ
ajst-9864	253	5	and	and	CCONJ
ajst-9864	253	6	precise	precise	ADJ
ajst-9864	253	7	natural	natural	ADJ
ajst-9864	253	8	language	language	NOUN
ajst-9864	253	9	processing	processing	NOUN
ajst-9864	253	10	tools	tool	NOUN
ajst-9864	253	11	to	to	PART
ajst-9864	253	12	handle	handle	VERB
ajst-9864	253	13	the	the	DET
ajst-9864	253	14	expanding	expand	VERB
ajst-9864	253	15	volume	volume	NOUN
ajst-9864	253	16	of	of	ADP
ajst-9864	253	17	data	datum	NOUN
ajst-9864	253	18	,	,	PUNCT
ajst-9864	253	19	given	give	VERB
ajst-9864	253	20	how	how	SCONJ
ajst-9864	253	21	quickly	quickly	ADV
ajst-9864	253	22	artificial	artificial	ADJ
ajst-9864	253	23	intelligence	intelligence	NOUN
ajst-9864	253	24	and	and	CCONJ
ajst-9864	253	25	language	language	NOUN
ajst-9864	253	26	processing	processing	NOUN
ajst-9864	253	27	technology	technology	NOUN
ajst-9864	253	28	are	be	AUX
ajst-9864	253	29	developing	develop	VERB
ajst-9864	253	30	.	.	PUNCT
ajst-9864	254	1	the	the	DET
ajst-9864	254	2	gru	gru	PROPN
ajst-9864	254	3	bidirectional	bidirectional	NOUN
ajst-9864	254	4	supervised	supervise	VERB
ajst-9864	254	5	text	text	NOUN
ajst-9864	254	6	entity	entity	NOUN
ajst-9864	254	7	extraction	extraction	NOUN
ajst-9864	254	8	model	model	NOUN
ajst-9864	254	9	suggested	suggest	VERB
ajst-9864	254	10	in	in	ADP
ajst-9864	254	11	this	this	DET
ajst-9864	254	12	article	article	NOUN
ajst-9864	254	13	is	be	AUX
ajst-9864	254	14	so	so	ADV
ajst-9864	254	15	effective	effective	ADJ
ajst-9864	254	16	and	and	CCONJ
ajst-9864	254	17	precise	precise	ADJ
ajst-9864	254	18	that	that	SCONJ
ajst-9864	254	19	it	it	PRON
ajst-9864	254	20	can	can	AUX
ajst-9864	254	21	give	give	VERB
ajst-9864	254	22	us	we	PRON
ajst-9864	254	23	more	more	ADV
ajst-9864	254	24	precise	precise	ADJ
ajst-9864	254	25	and	and	CCONJ
ajst-9864	254	26	intelligent	intelligent	ADJ
ajst-9864	254	27	natural	natural	ADJ
ajst-9864	254	28	language	language	NOUN
ajst-9864	254	29	processing	processing	NOUN
ajst-9864	254	30	services	service	NOUN
ajst-9864	254	31	,	,	PUNCT
ajst-9864	254	32	enhancing	enhance	VERB
ajst-9864	254	33	society	society	NOUN
ajst-9864	254	34	and	and	CCONJ
ajst-9864	254	35	businesses	business	NOUN
ajst-9864	254	36	.	.	PUNCT
ajst-9864	255	1	7	7	X
ajst-9864	255	2	.	.	X
ajst-9864	255	3	appendix	appendix	ADJ
ajst-9864	255	4	code	code	NOUN
ajst-9864	255	5	address	address	NOUN
ajst-9864	255	6	:	:	PUNCT
ajst-9864	255	7	https://github.com/jieyinggang/re_bgru_2att	https://github.com/jieyinggang/re_bgru_2att	PROPN
ajst-9864	255	8	references	reference	NOUN
ajst-9864	255	9	[	[	X
ajst-9864	255	10	1	1	NUM
ajst-9864	255	11	]	]	PUNCT
ajst-9864	255	12	lin	lin	PROPN
ajst-9864	255	13	y	y	PROPN
ajst-9864	255	14	,	,	PUNCT
ajst-9864	255	15	shen	shen	PROPN
ajst-9864	255	16	s	s	PROPN
ajst-9864	255	17	,	,	PUNCT
ajst-9864	255	18	liu	liu	PROPN
ajst-9864	255	19	z	z	PROPN
ajst-9864	255	20	,	,	PUNCT
ajst-9864	255	21	et	et	PROPN
ajst-9864	255	22	al．	al．	NOUN
ajst-9864	255	23	neural	neural	PROPN
ajst-9864	255	24	relation	relation	NOUN
ajst-9864	255	25	extraction	extraction	NOUN
ajst-9864	255	26	with	with	ADP
ajst-9864	255	27	selective	selective	ADJ
ajst-9864	255	28	attention	attention	NOUN
ajst-9864	255	29	over	over	ADP
ajst-9864	255	30	instances	instance	NOUN
ajst-9864	256	1	[	[	X
ajst-9864	256	2	c	c	X
ajst-9864	256	3	]	]	X
ajst-9864	256	4	/	/	SYM
ajst-9864	256	5	/	/	SYM
ajst-9864	256	6	meeting	meeting	NOUN
ajst-9864	256	7	of	of	ADP
ajst-9864	256	8	the	the	DET
ajst-9864	256	9	association	association	NOUN
ajst-9864	256	10	for	for	ADP
ajst-9864	256	11	computational	computational	ADJ
ajst-9864	256	12	linguistics	linguistic	NOUN
ajst-9864	256	13	.	.	PUNCT
ajst-9864	257	1	2016	2016	NUM
ajst-9864	257	2	:	:	PUNCT
ajst-9864	257	3	2124－2133	2124－2133	NUM
ajst-9864	257	4	.	.	PUNCT
ajst-9864	258	1	[	[	X
ajst-9864	258	2	2	2	NUM
ajst-9864	258	3	]	]	X
ajst-9864	258	4	li	li	PROPN
ajst-9864	258	5	w	w	PROPN
ajst-9864	258	6	,	,	PUNCT
ajst-9864	258	7	zhang	zhang	PROPN
ajst-9864	258	8	p	p	PROPN
ajst-9864	258	9	,	,	PUNCT
ajst-9864	258	10	wei	wei	PROPN
ajst-9864	258	11	f	f	PROPN
ajst-9864	258	12	,	,	PUNCT
ajst-9864	258	13	et	et	PROPN
ajst-9864	258	14	al	al	PROPN
ajst-9864	258	15	.	.	PUNCT
ajst-9864	259	1	a	a	DET
ajst-9864	259	2	novel	novel	ADJ
ajst-9864	259	3	feature	feature	NOUN
ajst-9864	259	4	-	-	PUNCT
ajst-9864	259	5	based	base	VERB
ajst-9864	259	6	approach	approach	NOUN
ajst-9864	259	7	to	to	ADP
ajst-9864	259	8	chinese	chinese	ADJ
ajst-9864	259	9	entity	entity	NOUN
ajst-9864	259	10	relation	relation	NOUN
ajst-9864	259	11	extraction	extraction	NOUN
ajst-9864	259	12	[	[	X
ajst-9864	259	13	c	c	X
ajst-9864	259	14	]	]	X
ajst-9864	259	15	/	/	SYM
ajst-9864	259	16	/	/	SYM
ajst-9864	259	17	meeting	meeting	NOUN
ajst-9864	259	18	of	of	ADP
ajst-9864	259	19	the	the	DET
ajst-9864	259	20	association	association	NOUN
ajst-9864	259	21	for	for	ADP
ajst-9864	259	22	computational	computational	ADJ
ajst-9864	259	23	linguistics	linguistic	NOUN
ajst-9864	259	24	on	on	ADP
ajst-9864	259	25	human	human	ADJ
ajst-9864	259	26	lan	lan	PROPN
ajst-9864	259	27	guage	guage	NOUN
ajst-9864	259	28	technologies	technology	NOUN
ajst-9864	259	29	:	:	PUNCT
ajst-9864	259	30	short	short	ADJ
ajst-9864	259	31	papers	paper	NOUN
ajst-9864	259	32	.	.	PUNCT
ajst-9864	260	1	association	association	NOUN
ajst-9864	260	2	for	for	ADP
ajst-9864	260	3	computational	computational	ADJ
ajst-9864	260	4	linguistics	linguistic	NOUN
ajst-9864	260	5	,	,	PUNCT
ajst-9864	260	6	2010	2010	NUM
ajst-9864	260	7	:	:	PUNCT
ajst-9864	260	8	89－92	89－92	X
ajst-9864	260	9	.	.	PUNCT
ajst-9864	261	1	[	[	X
ajst-9864	261	2	3	3	X
ajst-9864	261	3	]	]	X
ajst-9864	261	4	che	che	X
ajst-9864	261	5	w	w	PROPN
ajst-9864	261	6	,	,	PUNCT
ajst-9864	261	7	jiang	jiang	PROPN
ajst-9864	261	8	j	j	PROPN
ajst-9864	261	9	,	,	PUNCT
ajst-9864	261	10	su	su	PROPN
ajst-9864	261	11	z	z	PROPN
ajst-9864	261	12	,	,	PUNCT
ajst-9864	261	13	et	et	PROPN
ajst-9864	261	14	al	al	PROPN
ajst-9864	261	15	.	.	PROPN
ajst-9864	261	16	improved	improve	VERB
ajst-9864	261	17	-	-	PUNCT
ajst-9864	261	18	edit	edit	NOUN
ajst-9864	261	19	-	-	PUNCT
ajst-9864	261	20	distance	distance	NOUN
ajst-9864	261	21	kernel	kernel	NOUN
ajst-9864	261	22	for	for	ADP
ajst-9864	261	23	chinese	chinese	ADJ
ajst-9864	261	24	relation	relation	NOUN
ajst-9864	261	25	extraction	extraction	NOUN
ajst-9864	261	26	[	[	X
ajst-9864	261	27	c	c	X
ajst-9864	261	28	]	]	X
ajst-9864	261	29	/	/	SYM
ajst-9864	261	30	/	/	SYM
ajst-9864	261	31	natural	natural	ADJ
ajst-9864	261	32	language	language	NOUN
ajst-9864	261	33	pro	pro	ADJ
ajst-9864	261	34	cessing	cessing	NOUN
ajst-9864	261	35	—	—	PUNCT
ajst-9864	261	36	ijcnlp	ijcnlp	NOUN
ajst-9864	261	37	2005	2005	NUM
ajst-9864	261	38	.	.	PUNCT
ajst-9864	262	1	second	second	ADJ
ajst-9864	262	2	international	international	ADJ
ajst-9864	262	3	joint	joint	ADJ
ajst-9864	262	4	confer	confer	NOUN
ajst-9864	262	5	ence	ence	NOUN
ajst-9864	262	6	,	,	PUNCT
ajst-9864	262	7	jeju	jeju	PROPN
ajst-9864	262	8	island	island	PROPN
ajst-9864	262	9	,	,	PUNCT
ajst-9864	262	10	korea	korea	PROPN
ajst-9864	262	11	,	,	PUNCT
ajst-9864	262	12	october	october	PROPN
ajst-9864	262	13	11	11	NUM
ajst-9864	262	14	13	13	NUM
ajst-9864	262	15	,	,	PUNCT
ajst-9864	262	16	2005	2005	NUM
ajst-9864	262	17	.	.	PUNCT
ajst-9864	263	1	proceedings	proceeding	NOUN
ajst-9864	263	2	.	.	PUNCT
ajst-9864	264	1	2010	2010	NUM
ajst-9864	264	2	:	:	PUNCT
ajst-9864	264	3	132	132	NUM
ajst-9864	264	4	-137	-137	X
ajst-9864	264	5	.	.	PUNCT
ajst-9864	265	1	[	[	X
ajst-9864	265	2	4	4	X
ajst-9864	265	3	]	]	X
ajst-9864	265	4	chen	chen	PROPN
ajst-9864	265	5	y	y	PROPN
ajst-9864	265	6	,	,	PUNCT
ajst-9864	265	7	zheng	zheng	PROPN
ajst-9864	265	8	d	d	X
ajst-9864	265	9	q	q	PROPN
ajst-9864	265	10	,	,	PUNCT
ajst-9864	265	11	zhao	zhao	PROPN
ajst-9864	265	12	t	t	PROPN
ajst-9864	265	13	j.	j.	PROPN
ajst-9864	265	14	chinese	chinese	PROPN
ajst-9864	265	15	relation	relation	PROPN
ajst-9864	265	16	extraction	extraction	NOUN
ajst-9864	265	17	based	base	VERB
ajst-9864	265	18	on	on	ADP
ajst-9864	265	19	deep	deep	ADJ
ajst-9864	265	20	belief	belief	NOUN
ajst-9864	265	21	nets	net	NOUN
ajst-9864	265	22	[	[	X
ajst-9864	265	23	j	j	X
ajst-9864	265	24	]	]	X
ajst-9864	265	25	.	.	PUNCT
ajst-9864	266	1	journal	journal	PROPN
ajst-9864	266	2	of	of	ADP
ajst-9864	266	3	software	software	NOUN
ajst-9864	266	4	,	,	PUNCT
ajst-9864	266	5	2012	2012	NUM
ajst-9864	266	6	,	,	PUNCT
ajst-9864	266	7	23	23	NUM
ajst-9864	266	8	(	(	PUNCT
ajst-9864	266	9	10	10	NUM
ajst-9864	266	10	)	)	PUNCT
ajst-9864	266	11	:	:	PUNCT
ajst-9864	267	1	2572－2585	2572－2585	NUM
ajst-9864	267	2	.	.	PUNCT
ajst-9864	268	1	[	[	X
ajst-9864	268	2	5	5	NUM
ajst-9864	268	3	]	]	SYM
ajst-9864	268	4	rong	rong	PROPN
ajst-9864	268	5	bohui	bohui	PROPN
ajst-9864	268	6	,	,	PUNCT
ajst-9864	268	7	fu	fu	PROPN
ajst-9864	268	8	kun	kun	PROPN
ajst-9864	268	9	,	,	PUNCT
ajst-9864	268	10	huang	huang	PROPN
ajst-9864	268	11	yu	yu	PROPN
ajst-9864	268	12	,	,	PUNCT
ajst-9864	268	13	et	et	PROPN
ajst-9864	268	14	al	al	PROPN
ajst-9864	268	15	.	.	PUNCT
ajst-9864	268	16	multi	multi	ADJ
ajst-9864	268	17	-	-	ADJ
ajst-9864	268	18	channel	channel	ADJ
ajst-9864	268	19	convolutional	convolutional	ADJ
ajst-9864	268	20	neural	neural	ADJ
ajst-9864	268	21	net	net	NOUN
ajst-9864	268	22	-	-	PUNCT
ajst-9864	268	23	based	base	VERB
ajst-9864	268	24	entity	entity	NOUN
ajst-9864	268	25	relationship	relationship	NOUN
ajst-9864	268	26	extraction	extraction	NOUN
ajst-9864	268	27	[	[	X
ajst-9864	268	28	j	j	X
ajst-9864	268	29	]	]	X
ajst-9864	268	30	.	.	PUNCT
ajst-9864	269	1	computer	computer	NOUN
ajst-9864	269	2	applications	application	NOUN
ajst-9864	269	3	research	research	NOUN
ajst-9864	269	4	,	,	PUNCT
ajst-9864	269	5	2017	2017	NUM
ajst-9864	269	6	,	,	PUNCT
ajst-9864	269	7	34	34	NUM
ajst-9864	269	8	(	(	PUNCT
ajst-9864	269	9	3	3	NUM
ajst-9864	269	10	):	):	PUNCT
ajst-9864	269	11	689	689	NUM
ajst-9864	270	1	[	[	X
ajst-9864	270	2	6	6	NUM
ajst-9864	270	3	]	]	PUNCT
ajst-9864	270	4	xu	xu	PROPN
ajst-9864	271	1	k	k	PROPN
ajst-9864	271	2	,	,	PUNCT
ajst-9864	271	3	feng	feng	PROPN
ajst-9864	271	4	y	y	PROPN
ajst-9864	271	5	,	,	PUNCT
ajst-9864	271	6	huang	huang	PROPN
ajst-9864	271	7	s	s	PROPN
ajst-9864	271	8	,	,	PUNCT
ajst-9864	271	9	et	et	PROPN
ajst-9864	271	10	al	al	PROPN
ajst-9864	271	11	.	.	PUNCT
ajst-9864	271	12	semantic	semantic	PROPN
ajst-9864	271	13	relation	relation	NOUN
ajst-9864	271	14	classification	classification	NOUN
ajst-9864	271	15	via	via	ADP
ajst-9864	271	16	convolutional	convolutional	ADJ
ajst-9864	271	17	neural	neural	ADJ
ajst-9864	271	18	networks	network	NOUN
ajst-9864	271	19	with	with	ADP
ajst-9864	271	20	simple	simple	ADJ
ajst-9864	271	21	negative	negative	ADJ
ajst-9864	271	22	sampling[j	sampling[j	NOUN
ajst-9864	271	23	]	]	PUNCT
ajst-9864	271	24	.	.	PUNCT
ajst-9864	272	1	computer	computer	NOUN
ajst-9864	272	2	ence	ence	NOUN
ajst-9864	272	3	,	,	PUNCT
ajst-9864	272	4	2015	2015	NUM
ajst-9864	272	5	,	,	PUNCT
ajst-9864	272	6	71(7	71(7	NUM
ajst-9864	272	7	):	):	PUNCT
ajst-9864	272	8	941	941	NUM
ajst-9864	272	9	-	-	SYM
ajst-9864	272	10	9	9	NUM
ajst-9864	272	11	.	.	PUNCT
ajst-9864	273	1	[	[	X
ajst-9864	273	2	7	7	X
ajst-9864	273	3	]	]	X
ajst-9864	273	4	li	li	PROPN
ajst-9864	273	5	z	z	PROPN
ajst-9864	273	6	,	,	PUNCT
ajst-9864	273	7	yang	yang	PROPN
ajst-9864	273	8	z	z	PROPN
ajst-9864	273	9	,	,	PUNCT
ajst-9864	273	10	shen	shen	PROPN
ajst-9864	273	11	c	c	X
ajst-9864	273	12	,	,	PUNCT
ajst-9864	273	13	et	et	PROPN
ajst-9864	273	14	al	al	PROPN
ajst-9864	273	15	.	.	PUNCT
ajst-9864	273	16	integrating	integrate	VERB
ajst-9864	273	17	shortest	short	ADJ
ajst-9864	273	18	dependency	dependency	NOUN
ajst-9864	273	19	path	path	NOUN
ajst-9864	273	20	and	and	CCONJ
ajst-9864	273	21	sentence	sentence	NOUN
ajst-9864	273	22	sequence	sequence	NOUN
ajst-9864	273	23	into	into	ADP
ajst-9864	273	24	a	a	DET
ajst-9864	273	25	deep	deep	ADJ
ajst-9864	273	26	learning	learning	NOUN
ajst-9864	273	27	framework	framework	NOUN
ajst-9864	273	28	for	for	ADP
ajst-9864	273	29	relation	relation	NOUN
ajst-9864	273	30	extraction	extraction	NOUN
ajst-9864	273	31	in	in	ADP
ajst-9864	273	32	clinical	clinical	ADJ
ajst-9864	273	33	text[j	text[j	NOUN
ajst-9864	273	34	]	]	PUNCT
ajst-9864	273	35	.	.	PUNCT
ajst-9864	274	1	bmc	bmc	PROPN
ajst-9864	274	2	medical	medical	ADJ
ajst-9864	274	3	informatics	informatic	NOUN
ajst-9864	274	4	and	and	CCONJ
ajst-9864	274	5	decision	decision	NOUN
ajst-9864	274	6	making	making	NOUN
ajst-9864	274	7	,	,	PUNCT
ajst-9864	274	8	2019	2019	NUM
ajst-9864	274	9	,	,	PUNCT
ajst-9864	274	10	19(s1	19(s1	NUM
ajst-9864	274	11	)	)	PUNCT
ajst-9864	274	12	.	.	PUNCT
ajst-9864	275	1	[	[	X
ajst-9864	275	2	8	8	NUM
ajst-9864	275	3	]	]	X
ajst-9864	275	4	sawant	sawant	PROPN
ajst-9864	275	5	u	u	PROPN
ajst-9864	275	6	,	,	PUNCT
ajst-9864	275	7	garg	garg	PROPN
ajst-9864	275	8	s	s	PROPN
ajst-9864	275	9	,	,	PUNCT
ajst-9864	275	10	chakrabarti	chakrabarti	PROPN
ajst-9864	275	11	s	s	PART
ajst-9864	275	12	,	,	PUNCT
ajst-9864	275	13	et	et	PROPN
ajst-9864	275	14	al	al	PROPN
ajst-9864	275	15	.	.	PROPN
ajst-9864	275	16	neural	neural	ADJ
ajst-9864	275	17	architecture	architecture	NOUN
ajst-9864	275	18	for	for	ADP
ajst-9864	275	19	question	question	NOUN
ajst-9864	275	20	answering	answer	VERB
ajst-9864	275	21	using	use	VERB
ajst-9864	275	22	a	a	DET
ajst-9864	275	23	knowledge	knowledge	NOUN
ajst-9864	275	24	graph	graph	NOUN
ajst-9864	275	25	and	and	CCONJ
ajst-9864	275	26	web	web	NOUN
ajst-9864	275	27	corpus[j	corpus[j	NOUN
ajst-9864	275	28	]	]	PUNCT
ajst-9864	275	29	.	.	PUNCT
ajst-9864	276	1	information	information	NOUN
ajst-9864	276	2	retrieval,2019	retrieval,2019	NOUN
ajst-9864	276	3	,	,	PUNCT
ajst-9864	276	4	22(3	22(3	NOUN
ajst-9864	276	5	-	-	SYM
ajst-9864	276	6	4	4	NUM
ajst-9864	276	7	):	):	PUNCT
ajst-9864	276	8	324	324	NUM
ajst-9864	276	9	-	-	SYM
ajst-9864	276	10	349	349	NUM
ajst-9864	276	11	.	.	PUNCT
ajst-9864	277	1	[	[	X
ajst-9864	277	2	9	9	NUM
ajst-9864	277	3	]	]	PUNCT
ajst-9864	277	4	xu	xu	PROPN
ajst-9864	277	5	k	k	PROPN
ajst-9864	277	6	,	,	PUNCT
ajst-9864	277	7	feng	feng	PROPN
ajst-9864	277	8	y	y	PROPN
ajst-9864	277	9	,	,	PUNCT
ajst-9864	277	10	huang	huang	PROPN
ajst-9864	277	11	s	s	PROPN
ajst-9864	277	12	,	,	PUNCT
ajst-9864	277	13	et	et	PROPN
ajst-9864	277	14	al	al	PROPN
ajst-9864	277	15	.	.	PUNCT
ajst-9864	277	16	semantic	semantic	PROPN
ajst-9864	277	17	relation	relation	NOUN
ajst-9864	277	18	classification	classification	NOUN
ajst-9864	277	19	via	via	ADP
ajst-9864	277	20	convolutional	convolutional	ADJ
ajst-9864	277	21	neural	neural	ADJ
ajst-9864	277	22	networks	network	NOUN
ajst-9864	277	23	with	with	ADP
ajst-9864	277	24	simple	simple	ADJ
ajst-9864	277	25	negative	negative	ADJ
ajst-9864	277	26	sampling[j	sampling[j	NOUN
ajst-9864	277	27	]	]	PUNCT
ajst-9864	277	28	.	.	PUNCT
ajst-9864	278	1	computer	computer	NOUN
ajst-9864	278	2	ence	ence	NOUN
ajst-9864	278	3	,	,	PUNCT
ajst-9864	278	4	2015	2015	NUM
ajst-9864	278	5	,	,	PUNCT
ajst-9864	278	6	71(7	71(7	NUM
ajst-9864	278	7	):	):	PUNCT
ajst-9864	278	8	941	941	NUM
ajst-9864	278	9	-	-	SYM
ajst-9864	278	10	9	9	NUM
ajst-9864	278	11	.	.	PUNCT
ajst-9864	279	1	[	[	X
ajst-9864	279	2	10	10	NUM
ajst-9864	279	3	]	]	X
ajst-9864	279	4	erickson	erickson	PROPN
ajst-9864	279	5	g	g	PROPN
ajst-9864	279	6	m.	m.	PROPN
ajst-9864	279	7	an	an	DET
ajst-9864	279	8	oligopoly	oligopoly	ADJ
ajst-9864	279	9	model	model	NOUN
ajst-9864	279	10	of	of	ADP
ajst-9864	279	11	dynamic	dynamic	ADJ
ajst-9864	279	12	advertising	advertising	NOUN
ajst-9864	279	13	competition[j	competition[j	PROPN
ajst-9864	279	14	]	]	PUNCT
ajst-9864	279	15	.	.	PUNCT
ajst-9864	280	1	european	european	PROPN
ajst-9864	280	2	journal	journal	PROPN
ajst-9864	280	3	of	of	ADP
ajst-9864	280	4	operational	operational	ADJ
ajst-9864	280	5	research	research	NOUN
ajst-9864	280	6	,	,	PUNCT
ajst-9864	280	7	2009	2009	NUM
ajst-9864	280	8	,	,	PUNCT
ajst-9864	280	9	197(1	197(1	NUM
ajst-9864	280	10	):	):	PUNCT
ajst-9864	280	11	374	374	NUM
ajst-9864	280	12	-	-	SYM
ajst-9864	280	13	388	388	NUM
ajst-9864	280	14	.	.	PUNCT
ajst-9864	281	1	[	[	X
ajst-9864	281	2	11	11	NUM
ajst-9864	281	3	]	]	X
ajst-9864	281	4	li	li	PROPN
ajst-9864	281	5	z	z	PROPN
ajst-9864	281	6	,	,	PUNCT
ajst-9864	281	7	yang	yang	PROPN
ajst-9864	281	8	z	z	PROPN
ajst-9864	281	9	,	,	PUNCT
ajst-9864	281	10	shen	shen	PROPN
ajst-9864	281	11	c	c	X
ajst-9864	281	12	,	,	PUNCT
ajst-9864	281	13	et	et	PROPN
ajst-9864	281	14	al	al	PROPN
ajst-9864	281	15	.	.	PUNCT
ajst-9864	281	16	integrating	integrate	VERB
ajst-9864	281	17	shortest	short	ADJ
ajst-9864	281	18	dependency	dependency	NOUN
ajst-9864	281	19	path	path	NOUN
ajst-9864	281	20	and	and	CCONJ
ajst-9864	281	21	sentence	sentence	NOUN
ajst-9864	281	22	sequence	sequence	NOUN
ajst-9864	281	23	into	into	ADP
ajst-9864	281	24	a	a	DET
ajst-9864	281	25	deep	deep	ADJ
ajst-9864	281	26	learning	learning	NOUN
ajst-9864	281	27	framework	framework	NOUN
ajst-9864	281	28	for	for	ADP
ajst-9864	281	29	relation	relation	NOUN
ajst-9864	281	30	extraction	extraction	NOUN
ajst-9864	281	31	in	in	ADP
ajst-9864	281	32	clinical	clinical	ADJ
ajst-9864	281	33	text[j	text[j	NOUN
ajst-9864	281	34	]	]	PUNCT
ajst-9864	281	35	.	.	PUNCT
ajst-9864	282	1	bmc	bmc	PROPN
ajst-9864	282	2	medical	medical	ADJ
ajst-9864	282	3	informatics	informatic	NOUN
ajst-9864	282	4	and	and	CCONJ
ajst-9864	282	5	decision	decision	NOUN
ajst-9864	282	6	making	making	NOUN
ajst-9864	282	7	,	,	PUNCT
ajst-9864	282	8	2019	2019	NUM
ajst-9864	282	9	,	,	PUNCT
ajst-9864	282	10	19(s1	19(s1	NUM
ajst-9864	282	11	)	)	PUNCT
ajst-9864	282	12	.	.	PUNCT
ajst-9864	283	1	[	[	X
ajst-9864	283	2	12	12	NUM
ajst-9864	283	3	]	]	X
ajst-9864	283	4	mikolov	mikolov	PROPN
ajst-9864	283	5	t	t	PROPN
ajst-9864	283	6	,	,	PUNCT
ajst-9864	283	7	sutskever	sutskever	VERB
ajst-9864	283	8	i	i	PRON
ajst-9864	283	9	,	,	PUNCT
ajst-9864	283	10	kai	kai	PROPN
ajst-9864	283	11	c	c	PROPN
ajst-9864	283	12	,	,	PUNCT
ajst-9864	283	13	et	et	PROPN
ajst-9864	283	14	al	al	PROPN
ajst-9864	283	15	.	.	PROPN
ajst-9864	284	1	distributed	distribute	VERB
ajst-9864	284	2	representations	representation	NOUN
ajst-9864	284	3	of	of	ADP
ajst-9864	284	4	words	word	NOUN
ajst-9864	284	5	and	and	CCONJ
ajst-9864	284	6	phrases	phrase	NOUN
ajst-9864	284	7	and	and	CCONJ
ajst-9864	284	8	their	their	PRON
ajst-9864	284	9	compositionality[j	compositionality[j	NOUN
ajst-9864	284	10	]	]	PUNCT
ajst-9864	284	11	.	.	PUNCT
ajst-9864	285	1	advances	advance	NOUN
ajst-9864	285	2	in	in	ADP
ajst-9864	285	3	neural	neural	ADJ
ajst-9864	285	4	information	information	NOUN
ajst-9864	285	5	processing	processing	NOUN
ajst-9864	285	6	systems	system	NOUN
ajst-9864	285	7	,	,	PUNCT
ajst-9864	285	8	2013	2013	NUM
ajst-9864	285	9	,	,	PUNCT
ajst-9864	285	10	26	26	NUM
ajst-9864	285	11	.	.	PUNCT
ajst-9864	286	1	[	[	X
ajst-9864	286	2	13	13	NUM
ajst-9864	286	3	]	]	X
ajst-9864	286	4	mikolov	mikolov	PROPN
ajst-9864	286	5	t	t	PROPN
ajst-9864	286	6	,	,	PUNCT
ajst-9864	286	7	chen	chen	PROPN
ajst-9864	286	8	k	k	PROPN
ajst-9864	286	9	,	,	PUNCT
ajst-9864	286	10	corrado	corrado	PROPN
ajst-9864	286	11	g	g	PROPN
ajst-9864	286	12	,	,	PUNCT
ajst-9864	286	13	et	et	PROPN
ajst-9864	287	1	al	al	PROPN
ajst-9864	287	2	.	.	PROPN
ajst-9864	287	3	efficient	efficient	ADJ
ajst-9864	287	4	estimation	estimation	NOUN
ajst-9864	287	5	of	of	ADP
ajst-9864	287	6	word	word	NOUN
ajst-9864	287	7	representations	representation	NOUN
ajst-9864	287	8	in	in	ADP
ajst-9864	287	9	vector	vector	NOUN
ajst-9864	287	10	space[d	space[d	NOUN
ajst-9864	287	11	]	]	PUNCT
ajst-9864	287	12	.	.	PUNCT
ajst-9864	288	1	computer	computer	NOUN
ajst-9864	288	2	science	science	NOUN
ajst-9864	288	3	,	,	PUNCT
ajst-9864	288	4	2013	2013	NUM
ajst-9864	288	5	.	.	PUNCT