id	sid	tid	token	lemma	pos
fcis-5209	1	1	frontiers	frontier	NOUN
fcis-5209	1	2	in	in	ADP
fcis-5209	1	3	computing	computing	NOUN
fcis-5209	1	4	and	and	CCONJ
fcis-5209	1	5	intelligent	intelligent	ADJ
fcis-5209	1	6	systems	system	NOUN
fcis-5209	1	7	issn	issn	VERB
fcis-5209	1	8	:	:	PUNCT
fcis-5209	1	9	2832	2832	NUM
fcis-5209	1	10	-	-	SYM
fcis-5209	1	11	6024	6024	NUM
fcis-5209	1	12	|	|	NOUN
fcis-5209	1	13	vol	vol	NOUN
fcis-5209	1	14	.	.	PROPN
fcis-5209	2	1	2	2	NUM
fcis-5209	2	2	,	,	PUNCT
fcis-5209	2	3	no	no	INTJ
fcis-5209	2	4	.	.	NOUN
fcis-5209	2	5	3	3	NUM
fcis-5209	2	6	,	,	PUNCT
fcis-5209	2	7	2022	2022	NUM
fcis-5209	2	8	40	40	NUM
fcis-5209	2	9	improved	improved	ADJ
fcis-5209	2	10	text	text	NOUN
fcis-5209	2	11	matching	matching	NOUN
fcis-5209	2	12	model	model	NOUN
fcis-5209	2	13	based	base	VERB
fcis-5209	2	14	on	on	ADP
fcis-5209	2	15	bert	bert	PROPN
fcis-5209	2	16	qingyu	qingyu	PROPN
fcis-5209	2	17	li	li	PROPN
fcis-5209	2	18	,	,	PUNCT
fcis-5209	2	19	yujun	yujun	PROPN
fcis-5209	2	20	zhang	zhang	PROPN
fcis-5209	2	21	*	*	PUNCT
fcis-5209	2	22	school	school	NOUN
fcis-5209	2	23	of	of	ADP
fcis-5209	2	24	computer	computer	NOUN
fcis-5209	2	25	and	and	CCONJ
fcis-5209	2	26	software	software	NOUN
fcis-5209	2	27	engineering	engineering	NOUN
fcis-5209	2	28	,	,	PUNCT
fcis-5209	2	29	university	university	NOUN
fcis-5209	2	30	of	of	ADP
fcis-5209	2	31	science	science	NOUN
fcis-5209	2	32	and	and	CCONJ
fcis-5209	2	33	technology	technology	NOUN
fcis-5209	2	34	liaoning	liaoning	NOUN
fcis-5209	2	35	,	,	PUNCT
fcis-5209	2	36	anshan	anshan	PROPN
fcis-5209	2	37	114051	114051	NUM
fcis-5209	2	38	,	,	PUNCT
fcis-5209	2	39	china	china	PROPN
fcis-5209	2	40	*	*	PUNCT
fcis-5209	2	41	corresponding	correspond	VERB
fcis-5209	2	42	author	author	NOUN
fcis-5209	2	43	:	:	PUNCT
fcis-5209	2	44	yujun	yujun	PROPN
fcis-5209	2	45	zhang	zhang	PROPN
fcis-5209	2	46	abstract	abstract	PROPN
fcis-5209	2	47	:	:	PUNCT
fcis-5209	3	1	text	text	NOUN
fcis-5209	3	2	matching	matching	NOUN
fcis-5209	3	3	is	be	AUX
fcis-5209	3	4	a	a	DET
fcis-5209	3	5	basic	basic	ADJ
fcis-5209	3	6	and	and	CCONJ
fcis-5209	3	7	important	important	ADJ
fcis-5209	3	8	task	task	NOUN
fcis-5209	3	9	in	in	ADP
fcis-5209	3	10	natural	natural	ADJ
fcis-5209	3	11	language	language	NOUN
fcis-5209	3	12	understanding	understanding	NOUN
fcis-5209	3	13	,	,	PUNCT
fcis-5209	3	14	this	this	DET
fcis-5209	3	15	paper	paper	NOUN
fcis-5209	3	16	proposes	propose	VERB
fcis-5209	3	17	a	a	DET
fcis-5209	3	18	new	new	ADJ
fcis-5209	3	19	model	model	NOUN
fcis-5209	3	20	bbmc	bbmc	NOUN
fcis-5209	3	21	for	for	ADP
fcis-5209	3	22	the	the	DET
fcis-5209	3	23	problem	problem	NOUN
fcis-5209	3	24	of	of	ADP
fcis-5209	3	25	insufficient	insufficient	ADJ
fcis-5209	3	26	feature	feature	NOUN
fcis-5209	3	27	extraction	extraction	NOUN
fcis-5209	3	28	ability	ability	NOUN
fcis-5209	3	29	of	of	ADP
fcis-5209	3	30	existing	exist	VERB
fcis-5209	3	31	text	text	NOUN
fcis-5209	3	32	matching	matching	NOUN
fcis-5209	3	33	models	model	NOUN
fcis-5209	3	34	,	,	PUNCT
fcis-5209	3	35	which	which	PRON
fcis-5209	3	36	integrates	integrate	VERB
fcis-5209	3	37	bilstm	bilstm	NOUN
fcis-5209	3	38	and	and	CCONJ
fcis-5209	3	39	multi	multi	ADJ
fcis-5209	3	40	-	-	ADJ
fcis-5209	3	41	scale	scale	ADJ
fcis-5209	3	42	cnn	cnn	PROPN
fcis-5209	3	43	on	on	ADP
fcis-5209	3	44	the	the	DET
fcis-5209	3	45	basis	basis	NOUN
fcis-5209	3	46	of	of	ADP
fcis-5209	3	47	bert	bert	PROPN
fcis-5209	3	48	.	.	PUNCT
fcis-5209	4	1	first	first	ADV
fcis-5209	4	2	,	,	PUNCT
fcis-5209	4	3	the	the	DET
fcis-5209	4	4	word	word	NOUN
fcis-5209	4	5	embedding	embed	VERB
fcis-5209	4	6	representation	representation	NOUN
fcis-5209	4	7	of	of	ADP
fcis-5209	4	8	the	the	DET
fcis-5209	4	9	text	text	NOUN
fcis-5209	4	10	is	be	AUX
fcis-5209	4	11	obtained	obtain	VERB
fcis-5209	4	12	by	by	ADP
fcis-5209	4	13	the	the	DET
fcis-5209	4	14	bert	bert	NOUN
fcis-5209	4	15	,	,	PUNCT
fcis-5209	4	16	and	and	CCONJ
fcis-5209	4	17	then	then	ADV
fcis-5209	4	18	the	the	DET
fcis-5209	4	19	semantic	semantic	ADJ
fcis-5209	4	20	features	feature	NOUN
fcis-5209	4	21	of	of	ADP
fcis-5209	4	22	the	the	DET
fcis-5209	4	23	text	text	NOUN
fcis-5209	4	24	are	be	AUX
fcis-5209	4	25	further	far	ADV
fcis-5209	4	26	extracted	extract	VERB
fcis-5209	4	27	by	by	ADP
fcis-5209	4	28	the	the	DET
fcis-5209	4	29	double	double	ADJ
fcis-5209	4	30	-	-	PUNCT
fcis-5209	4	31	layer	layer	NOUN
fcis-5209	4	32	bilstm	bilstm	NOUN
fcis-5209	4	33	,	,	PUNCT
fcis-5209	4	34	followed	follow	VERB
fcis-5209	4	35	by	by	ADP
fcis-5209	4	36	the	the	DET
fcis-5209	4	37	multi	multi	ADJ
fcis-5209	4	38	-	-	ADJ
fcis-5209	4	39	scale	scale	ADJ
fcis-5209	4	40	cnn	cnn	PROPN
fcis-5209	4	41	model	model	NOUN
fcis-5209	4	42	,	,	PUNCT
fcis-5209	4	43	the	the	DET
fcis-5209	4	44	key	key	ADJ
fcis-5209	4	45	local	local	ADJ
fcis-5209	4	46	features	feature	NOUN
fcis-5209	4	47	are	be	AUX
fcis-5209	4	48	extracted	extract	VERB
fcis-5209	4	49	,	,	PUNCT
fcis-5209	4	50	and	and	CCONJ
fcis-5209	4	51	finally	finally	ADV
fcis-5209	4	52	the	the	DET
fcis-5209	4	53	linear	linear	ADJ
fcis-5209	4	54	and	and	CCONJ
fcis-5209	4	55	softmax	softmax	NOUN
fcis-5209	4	56	function	function	NOUN
fcis-5209	4	57	are	be	AUX
fcis-5209	4	58	used	use	VERB
fcis-5209	4	59	to	to	PART
fcis-5209	4	60	classify	classify	VERB
fcis-5209	4	61	.	.	PUNCT
fcis-5209	5	1	experimental	experimental	ADJ
fcis-5209	5	2	results	result	NOUN
fcis-5209	5	3	on	on	ADP
fcis-5209	5	4	the	the	DET
fcis-5209	5	5	lcqmc	lcqmc	ADJ
fcis-5209	5	6	dataset	dataset	NOUN
fcis-5209	5	7	show	show	NOUN
fcis-5209	5	8	that	that	SCONJ
fcis-5209	5	9	the	the	DET
fcis-5209	5	10	bbmc	bbmc	NOUN
fcis-5209	5	11	has	have	AUX
fcis-5209	5	12	been	be	AUX
fcis-5209	5	13	improved	improve	VERB
fcis-5209	5	14	to	to	ADP
fcis-5209	5	15	a	a	DET
fcis-5209	5	16	certain	certain	ADJ
fcis-5209	5	17	extent	extent	NOUN
fcis-5209	5	18	compared	compare	VERB
fcis-5209	5	19	with	with	ADP
fcis-5209	5	20	other	other	ADJ
fcis-5209	5	21	methods	method	NOUN
fcis-5209	5	22	,	,	PUNCT
fcis-5209	5	23	and	and	CCONJ
fcis-5209	5	24	the	the	DET
fcis-5209	5	25	accuracy	accuracy	NOUN
fcis-5209	5	26	on	on	ADP
fcis-5209	5	27	the	the	DET
fcis-5209	5	28	test	test	NOUN
fcis-5209	5	29	set	set	NOUN
fcis-5209	5	30	can	can	AUX
fcis-5209	5	31	be	be	AUX
fcis-5209	5	32	best	well	ADV
fcis-5209	5	33	achieved	achieve	VERB
fcis-5209	5	34	88.01	88.01	NUM
fcis-5209	5	35	%	%	NOUN
fcis-5209	5	36	.	.	PUNCT
fcis-5209	6	1	keywords	keyword	NOUN
fcis-5209	6	2	:	:	PUNCT
fcis-5209	6	3	bert	bert	PROPN
fcis-5209	6	4	;	;	PUNCT
fcis-5209	6	5	bilstm	bilstm	NOUN
fcis-5209	6	6	;	;	PUNCT
fcis-5209	6	7	cnn	cnn	PROPN
fcis-5209	6	8	;	;	PUNCT
fcis-5209	6	9	text	text	NOUN
fcis-5209	6	10	matching	matching	NOUN
fcis-5209	6	11	;	;	PUNCT
fcis-5209	6	12	nlp	nlp	NOUN
fcis-5209	6	13	.	.	NOUN
fcis-5209	6	14	1	1	NUM
fcis-5209	6	15	.	.	X
fcis-5209	6	16	introduction	introduction	NOUN
fcis-5209	6	17	as	as	ADP
fcis-5209	6	18	a	a	DET
fcis-5209	6	19	core	core	NOUN
fcis-5209	6	20	problem	problem	NOUN
fcis-5209	6	21	in	in	ADP
fcis-5209	6	22	natural	natural	ADJ
fcis-5209	6	23	language	language	NOUN
fcis-5209	6	24	understanding	understanding	NOUN
fcis-5209	6	25	,	,	PUNCT
fcis-5209	6	26	text	text	NOUN
fcis-5209	6	27	matching	matching	NOUN
fcis-5209	6	28	is	be	AUX
fcis-5209	6	29	to	to	PART
fcis-5209	6	30	take	take	VERB
fcis-5209	6	31	two	two	NUM
fcis-5209	6	32	texts	text	NOUN
fcis-5209	6	33	as	as	ADP
fcis-5209	6	34	inputs	input	NOUN
fcis-5209	6	35	and	and	CCONJ
fcis-5209	6	36	predict	predict	VERB
fcis-5209	6	37	their	their	PRON
fcis-5209	6	38	relationship	relationship	NOUN
fcis-5209	6	39	categories	category	NOUN
fcis-5209	6	40	or	or	CCONJ
fcis-5209	6	41	correlation	correlation	NOUN
fcis-5209	6	42	scores	score	NOUN
fcis-5209	6	43	by	by	ADP
fcis-5209	6	44	understanding	understand	VERB
fcis-5209	6	45	their	their	PRON
fcis-5209	6	46	respective	respective	ADJ
fcis-5209	6	47	semantics	semantic	NOUN
fcis-5209	6	48	.	.	PUNCT
fcis-5209	7	1	usually	usually	ADV
fcis-5209	7	2	,	,	PUNCT
fcis-5209	7	3	this	this	DET
fcis-5209	7	4	task	task	NOUN
fcis-5209	7	5	is	be	AUX
fcis-5209	7	6	treated	treat	VERB
fcis-5209	7	7	as	as	ADP
fcis-5209	7	8	a	a	DET
fcis-5209	7	9	binary	binary	ADJ
fcis-5209	7	10	task	task	NOUN
fcis-5209	7	11	.	.	PUNCT
fcis-5209	8	1	first	first	ADJ
fcis-5209	8	2	,	,	PUNCT
fcis-5209	8	3	rich	rich	ADJ
fcis-5209	8	4	text	text	NOUN
fcis-5209	8	5	feature	feature	NOUN
fcis-5209	8	6	information	information	NOUN
fcis-5209	8	7	is	be	AUX
fcis-5209	8	8	obtained	obtain	VERB
fcis-5209	8	9	through	through	ADP
fcis-5209	8	10	various	various	ADJ
fcis-5209	8	11	methods	method	NOUN
fcis-5209	8	12	,	,	PUNCT
fcis-5209	8	13	and	and	CCONJ
fcis-5209	8	14	then	then	ADV
fcis-5209	8	15	the	the	DET
fcis-5209	8	16	extracted	extract	VERB
fcis-5209	8	17	features	feature	NOUN
fcis-5209	8	18	are	be	AUX
fcis-5209	8	19	classified	classified	ADJ
fcis-5209	8	20	.	.	PUNCT
fcis-5209	9	1	many	many	ADJ
fcis-5209	9	2	tasks	task	NOUN
fcis-5209	9	3	in	in	ADP
fcis-5209	9	4	natural	natural	ADJ
fcis-5209	9	5	language	language	NOUN
fcis-5209	9	6	processing	processing	NOUN
fcis-5209	9	7	can	can	AUX
fcis-5209	9	8	be	be	AUX
fcis-5209	9	9	regarded	regard	VERB
fcis-5209	9	10	as	as	ADP
fcis-5209	9	11	text	text	NOUN
fcis-5209	9	12	matching	matching	NOUN
fcis-5209	9	13	tasks	task	NOUN
fcis-5209	9	14	,	,	PUNCT
fcis-5209	9	15	so	so	SCONJ
fcis-5209	9	16	it	it	PRON
fcis-5209	9	17	is	be	AUX
fcis-5209	9	18	particularly	particularly	ADV
fcis-5209	9	19	important	important	ADJ
fcis-5209	9	20	to	to	PART
fcis-5209	9	21	study	study	VERB
fcis-5209	9	22	efficient	efficient	ADJ
fcis-5209	9	23	text	text	NOUN
fcis-5209	9	24	matching	matching	NOUN
fcis-5209	9	25	methods	method	NOUN
fcis-5209	9	26	.	.	PUNCT
fcis-5209	10	1	traditional	traditional	ADJ
fcis-5209	10	2	machine	machine	NOUN
fcis-5209	10	3	learning	learning	NOUN
fcis-5209	10	4	methods	method	NOUN
fcis-5209	10	5	need	need	VERB
fcis-5209	10	6	to	to	PART
fcis-5209	10	7	rely	rely	VERB
fcis-5209	10	8	on	on	ADP
fcis-5209	10	9	manual	manual	ADJ
fcis-5209	10	10	access	access	NOUN
fcis-5209	10	11	to	to	ADP
fcis-5209	10	12	the	the	DET
fcis-5209	10	13	shallow	shallow	ADJ
fcis-5209	10	14	features	feature	NOUN
fcis-5209	10	15	of	of	ADP
fcis-5209	10	16	the	the	DET
fcis-5209	10	17	text	text	NOUN
fcis-5209	10	18	,	,	PUNCT
fcis-5209	10	19	which	which	PRON
fcis-5209	10	20	can	can	AUX
fcis-5209	10	21	not	not	PART
fcis-5209	10	22	well	well	ADV
fcis-5209	10	23	represent	represent	VERB
fcis-5209	10	24	the	the	DET
fcis-5209	10	25	semantic	semantic	ADJ
fcis-5209	10	26	information	information	NOUN
fcis-5209	10	27	of	of	ADP
fcis-5209	10	28	the	the	DET
fcis-5209	10	29	text	text	NOUN
fcis-5209	10	30	.	.	PUNCT
fcis-5209	11	1	with	with	ADP
fcis-5209	11	2	the	the	DET
fcis-5209	11	3	rapid	rapid	ADJ
fcis-5209	11	4	development	development	NOUN
fcis-5209	11	5	of	of	ADP
fcis-5209	11	6	deep	deep	ADJ
fcis-5209	11	7	learning	learning	NOUN
fcis-5209	11	8	,	,	PUNCT
fcis-5209	11	9	a	a	DET
fcis-5209	11	10	large	large	ADJ
fcis-5209	11	11	number	number	NOUN
fcis-5209	11	12	of	of	ADP
fcis-5209	11	13	text	text	NOUN
fcis-5209	11	14	matching	matching	NOUN
fcis-5209	11	15	models	model	NOUN
fcis-5209	11	16	based	base	VERB
fcis-5209	11	17	on	on	ADP
fcis-5209	11	18	deep	deep	ADJ
fcis-5209	11	19	learning	learning	NOUN
fcis-5209	11	20	have	have	AUX
fcis-5209	11	21	emerged	emerge	VERB
fcis-5209	11	22	,	,	PUNCT
fcis-5209	11	23	which	which	PRON
fcis-5209	11	24	can	can	AUX
fcis-5209	11	25	be	be	AUX
fcis-5209	11	26	roughly	roughly	ADV
fcis-5209	11	27	divided	divide	VERB
fcis-5209	11	28	into	into	ADP
fcis-5209	11	29	three	three	NUM
fcis-5209	11	30	types	type	NOUN
fcis-5209	11	31	:	:	PUNCT
fcis-5209	11	32	representation	representation	NOUN
fcis-5209	11	33	based	base	VERB
fcis-5209	11	34	method	method	NOUN
fcis-5209	11	35	,	,	PUNCT
fcis-5209	11	36	interaction	interaction	NOUN
fcis-5209	11	37	based	base	VERB
fcis-5209	11	38	method	method	NOUN
fcis-5209	11	39	and	and	CCONJ
fcis-5209	11	40	pretraining	pretraine	VERB
fcis-5209	11	41	language	language	NOUN
fcis-5209	11	42	model	model	NOUN
fcis-5209	11	43	based	base	VERB
fcis-5209	11	44	method	method	NOUN
fcis-5209	11	45	.	.	PUNCT
fcis-5209	12	1	the	the	DET
fcis-5209	12	2	representation	representation	NOUN
fcis-5209	12	3	based	base	VERB
fcis-5209	12	4	method	method	NOUN
fcis-5209	12	5	obtains	obtain	VERB
fcis-5209	12	6	the	the	DET
fcis-5209	12	7	semantic	semantic	ADJ
fcis-5209	12	8	representation	representation	NOUN
fcis-5209	12	9	of	of	ADP
fcis-5209	12	10	text	text	NOUN
fcis-5209	12	11	in	in	ADP
fcis-5209	12	12	various	various	ADJ
fcis-5209	12	13	ways	way	NOUN
fcis-5209	12	14	for	for	ADP
fcis-5209	12	15	matching	match	VERB
fcis-5209	12	16	calculation	calculation	NOUN
fcis-5209	12	17	.	.	PUNCT
fcis-5209	13	1	in	in	ADP
fcis-5209	13	2	2013	2013	NUM
fcis-5209	13	3	,	,	PUNCT
fcis-5209	13	4	huang	huang	PROPN
fcis-5209	13	5	et	et	PROPN
fcis-5209	13	6	al	al	PROPN
fcis-5209	13	7	.	.	PROPN
fcis-5209	13	8	proposed	propose	VERB
fcis-5209	13	9	the	the	DET
fcis-5209	13	10	dssm	dssm	PROPN
fcis-5209	13	11	model	model	NOUN
fcis-5209	13	12	,	,	PUNCT
fcis-5209	13	13	which	which	PRON
fcis-5209	13	14	is	be	AUX
fcis-5209	13	15	the	the	DET
fcis-5209	13	16	pioneering	pioneering	ADJ
fcis-5209	13	17	work	work	NOUN
fcis-5209	13	18	of	of	ADP
fcis-5209	13	19	deep	deep	ADJ
fcis-5209	13	20	learning	learning	NOUN
fcis-5209	13	21	in	in	ADP
fcis-5209	13	22	text	text	NOUN
fcis-5209	13	23	matching	matching	NOUN
fcis-5209	13	24	tasks	task	NOUN
fcis-5209	13	25	.	.	PUNCT
fcis-5209	14	1	later	later	ADV
fcis-5209	14	2	,	,	PUNCT
fcis-5209	14	3	with	with	ADP
fcis-5209	14	4	the	the	DET
fcis-5209	14	5	popularity	popularity	NOUN
fcis-5209	14	6	of	of	ADP
fcis-5209	14	7	cnn	cnn	PROPN
fcis-5209	14	8	(	(	PUNCT
fcis-5209	14	9	convolutional	convolutional	ADJ
fcis-5209	14	10	neural	neural	ADJ
fcis-5209	14	11	network	network	NOUN
fcis-5209	14	12	,	,	PUNCT
fcis-5209	14	13	cnn	cnn	PROPN
fcis-5209	14	14	)	)	PUNCT
fcis-5209	14	15	and	and	CCONJ
fcis-5209	14	16	lstm	lstm	NOUN
fcis-5209	14	17	(	(	PUNCT
fcis-5209	14	18	long	long	ADJ
fcis-5209	14	19	short	short	ADJ
fcis-5209	14	20	term	term	NOUN
fcis-5209	14	21	memory	memory	NOUN
fcis-5209	14	22	,	,	PUNCT
fcis-5209	14	23	lstm	lstm	NOUN
fcis-5209	14	24	)	)	PUNCT
fcis-5209	14	25	,	,	PUNCT
fcis-5209	14	26	etc	etc	X
fcis-5209	14	27	.	.	X
fcis-5209	14	28	,	,	PUNCT
fcis-5209	14	29	shen	shen	PROPN
fcis-5209	14	30	et	et	PROPN
fcis-5209	14	31	al	al	PROPN
fcis-5209	14	32	.	.	PROPN
fcis-5209	14	33	introduced	introduce	VERB
fcis-5209	14	34	cnn	cnn	PROPN
fcis-5209	14	35	into	into	ADP
fcis-5209	14	36	the	the	DET
fcis-5209	14	37	dssm	dssm	PROPN
fcis-5209	14	38	model	model	NOUN
fcis-5209	14	39	to	to	PART
fcis-5209	14	40	solve	solve	VERB
fcis-5209	14	41	the	the	DET
fcis-5209	14	42	problem	problem	NOUN
fcis-5209	14	43	of	of	ADP
fcis-5209	14	44	losing	lose	VERB
fcis-5209	14	45	context	context	NOUN
fcis-5209	14	46	information	information	NOUN
fcis-5209	14	47	in	in	ADP
fcis-5209	14	48	dssm	dssm	PROPN
fcis-5209	14	49	.	.	PUNCT
fcis-5209	15	1	palangi	palangi	PROPN
fcis-5209	15	2	et	et	PROPN
fcis-5209	15	3	al	al	PROPN
fcis-5209	15	4	.	.	PROPN
fcis-5209	15	5	introduced	introduce	VERB
fcis-5209	15	6	lstm	lstm	NOUN
fcis-5209	15	7	into	into	ADP
fcis-5209	15	8	dssm	dssm	NOUN
fcis-5209	15	9	to	to	PART
fcis-5209	15	10	capture	capture	VERB
fcis-5209	15	11	long	long	ADJ
fcis-5209	15	12	-	-	PUNCT
fcis-5209	15	13	term	term	NOUN
fcis-5209	15	14	context	context	NOUN
fcis-5209	15	15	and	and	CCONJ
fcis-5209	15	16	proposed	propose	VERB
fcis-5209	15	17	lstm	lstm	ADJ
fcis-5209	15	18	-	-	PUNCT
fcis-5209	15	19	dssm	dssm	ADJ
fcis-5209	15	20	model	model	NOUN
fcis-5209	15	21	.	.	PUNCT
fcis-5209	16	1	this	this	DET
fcis-5209	16	2	kind	kind	NOUN
fcis-5209	16	3	of	of	ADP
fcis-5209	16	4	model	model	NOUN
fcis-5209	16	5	only	only	ADV
fcis-5209	16	6	obtains	obtain	VERB
fcis-5209	16	7	the	the	DET
fcis-5209	16	8	semantic	semantic	ADJ
fcis-5209	16	9	representation	representation	NOUN
fcis-5209	16	10	of	of	ADP
fcis-5209	16	11	the	the	DET
fcis-5209	16	12	text	text	NOUN
fcis-5209	16	13	,	,	PUNCT
fcis-5209	16	14	but	but	CCONJ
fcis-5209	16	15	does	do	AUX
fcis-5209	16	16	not	not	PART
fcis-5209	16	17	take	take	VERB
fcis-5209	16	18	into	into	ADP
fcis-5209	16	19	account	account	NOUN
fcis-5209	16	20	the	the	DET
fcis-5209	16	21	interaction	interaction	NOUN
fcis-5209	16	22	information	information	NOUN
fcis-5209	16	23	between	between	ADP
fcis-5209	16	24	texts	text	NOUN
fcis-5209	16	25	.	.	PUNCT
fcis-5209	17	1	the	the	DET
fcis-5209	17	2	interaction	interaction	NOUN
fcis-5209	17	3	based	base	VERB
fcis-5209	17	4	method	method	NOUN
fcis-5209	17	5	realizes	realize	VERB
fcis-5209	17	6	text	text	NOUN
fcis-5209	17	7	interaction	interaction	NOUN
fcis-5209	17	8	through	through	ADP
fcis-5209	17	9	various	various	ADJ
fcis-5209	17	10	attention	attention	NOUN
fcis-5209	17	11	mechanisms	mechanism	NOUN
fcis-5209	17	12	to	to	PART
fcis-5209	17	13	obtain	obtain	VERB
fcis-5209	17	14	context	context	NOUN
fcis-5209	17	15	information	information	NOUN
fcis-5209	17	16	.	.	PUNCT
fcis-5209	18	1	hu	hu	PROPN
fcis-5209	18	2	et	et	PROPN
fcis-5209	18	3	al	al	PROPN
fcis-5209	18	4	.	.	PUNCT
fcis-5209	18	5	use	use	VERB
fcis-5209	18	6	one	one	NUM
fcis-5209	18	7	-	-	PUNCT
fcis-5209	18	8	dimensional	dimensional	ADJ
fcis-5209	18	9	convolution	convolution	NOUN
fcis-5209	18	10	to	to	PART
fcis-5209	18	11	focus	focus	VERB
fcis-5209	18	12	on	on	ADP
fcis-5209	18	13	adjacent	adjacent	ADJ
fcis-5209	18	14	word	word	NOUN
fcis-5209	18	15	vectors	vector	NOUN
fcis-5209	18	16	for	for	ADP
fcis-5209	18	17	two	two	NUM
fcis-5209	18	18	pieces	piece	NOUN
fcis-5209	18	19	of	of	ADP
fcis-5209	18	20	text	text	NOUN
fcis-5209	18	21	respectively	respectively	ADV
fcis-5209	18	22	,	,	PUNCT
fcis-5209	18	23	then	then	ADV
fcis-5209	18	24	combine	combine	VERB
fcis-5209	18	25	the	the	DET
fcis-5209	18	26	two	two	NUM
fcis-5209	18	27	tensors	tensor	NOUN
fcis-5209	18	28	obtained	obtain	VERB
fcis-5209	18	29	after	after	ADP
fcis-5209	18	30	convolution	convolution	NOUN
fcis-5209	18	31	,	,	PUNCT
fcis-5209	18	32	and	and	CCONJ
fcis-5209	18	33	finally	finally	ADV
fcis-5209	18	34	propose	propose	VERB
fcis-5209	18	35	the	the	DET
fcis-5209	18	36	arc	arc	NOUN
fcis-5209	18	37	-	-	PUNCT
fcis-5209	18	38	ii	ii	NOUN
fcis-5209	18	39	model	model	NOUN
fcis-5209	18	40	using	use	VERB
fcis-5209	18	41	the	the	DET
fcis-5209	18	42	classification	classification	NOUN
fcis-5209	18	43	method	method	NOUN
fcis-5209	18	44	of	of	ADP
fcis-5209	18	45	multilayer	multilayer	PROPN
fcis-5209	18	46	perceptron	perceptron	PROPN
fcis-5209	18	47	.	.	PUNCT
fcis-5209	19	1	yin	yin	PROPN
fcis-5209	19	2	et	et	PROPN
fcis-5209	19	3	al	al	PROPN
fcis-5209	19	4	.	.	PROPN
fcis-5209	19	5	introduced	introduce	VERB
fcis-5209	19	6	the	the	DET
fcis-5209	19	7	attention	attention	NOUN
fcis-5209	19	8	mechanism	mechanism	NOUN
fcis-5209	19	9	on	on	ADP
fcis-5209	19	10	the	the	DET
fcis-5209	19	11	basis	basis	NOUN
fcis-5209	19	12	of	of	ADP
fcis-5209	19	13	cnn	cnn	PROPN
fcis-5209	19	14	,	,	PUNCT
fcis-5209	19	15	used	use	VERB
fcis-5209	19	16	the	the	DET
fcis-5209	19	17	attention	attention	NOUN
fcis-5209	19	18	weight	weight	NOUN
fcis-5209	19	19	to	to	PART
fcis-5209	19	20	fuse	fuse	VERB
fcis-5209	19	21	the	the	DET
fcis-5209	19	22	output	output	NOUN
fcis-5209	19	23	of	of	ADP
fcis-5209	19	24	the	the	DET
fcis-5209	19	25	convolution	convolution	NOUN
fcis-5209	19	26	layer	layer	NOUN
fcis-5209	19	27	,	,	PUNCT
fcis-5209	19	28	and	and	CCONJ
fcis-5209	19	29	finally	finally	ADV
fcis-5209	19	30	proposed	propose	VERB
fcis-5209	19	31	the	the	DET
fcis-5209	19	32	abcnn	abcnn	PROPN
fcis-5209	19	33	model	model	NOUN
fcis-5209	19	34	.	.	PUNCT
fcis-5209	20	1	pang	pang	NOUN
fcis-5209	20	2	et	et	PROPN
fcis-5209	20	3	al	al	PROPN
fcis-5209	20	4	.	.	PROPN
fcis-5209	21	1	used	use	VERB
fcis-5209	21	2	dot	dot	NOUN
fcis-5209	21	3	product	product	NOUN
fcis-5209	21	4	operation	operation	NOUN
fcis-5209	21	5	on	on	ADP
fcis-5209	21	6	word	word	NOUN
fcis-5209	21	7	vectors	vector	NOUN
fcis-5209	21	8	of	of	ADP
fcis-5209	21	9	two	two	NUM
fcis-5209	21	10	texts	text	NOUN
fcis-5209	21	11	to	to	PART
fcis-5209	21	12	interact	interact	VERB
fcis-5209	21	13	,	,	PUNCT
fcis-5209	21	14	and	and	CCONJ
fcis-5209	21	15	then	then	ADV
fcis-5209	21	16	proposed	propose	VERB
fcis-5209	21	17	matchpyramid	matchpyramid	NOUN
fcis-5209	21	18	model	model	NOUN
fcis-5209	21	19	by	by	ADP
fcis-5209	21	20	using	use	VERB
fcis-5209	21	21	convolution	convolution	NOUN
fcis-5209	21	22	and	and	CCONJ
fcis-5209	21	23	pooling	pool	VERB
fcis-5209	21	24	to	to	PART
fcis-5209	21	25	extract	extract	VERB
fcis-5209	21	26	features	feature	NOUN
fcis-5209	21	27	.	.	PUNCT
fcis-5209	22	1	chen	chen	PROPN
fcis-5209	22	2	et	et	PROPN
fcis-5209	22	3	al	al	PROPN
fcis-5209	22	4	.	.	PROPN
fcis-5209	22	5	proposed	propose	VERB
fcis-5209	22	6	a	a	DET
fcis-5209	22	7	simple	simple	ADJ
fcis-5209	22	8	and	and	CCONJ
fcis-5209	22	9	efficient	efficient	ADJ
fcis-5209	22	10	semantic	semantic	ADJ
fcis-5209	22	11	matching	matching	NOUN
fcis-5209	22	12	network	network	NOUN
fcis-5209	22	13	esim	esim	ADJ
fcis-5209	22	14	by	by	ADP
fcis-5209	22	15	using	use	VERB
fcis-5209	22	16	bi	bi	ADJ
fcis-5209	22	17	-	-	ADJ
fcis-5209	22	18	directional	directional	ADJ
fcis-5209	22	19	recurrent	recurrent	ADJ
fcis-5209	22	20	neural	neural	ADJ
fcis-5209	22	21	network	network	NOUN
fcis-5209	22	22	bilstm	bilstm	NOUN
fcis-5209	22	23	to	to	PART
fcis-5209	22	24	obtain	obtain	VERB
fcis-5209	22	25	context	context	NOUN
fcis-5209	22	26	representation	representation	NOUN
fcis-5209	22	27	and	and	CCONJ
fcis-5209	22	28	introducing	introduce	VERB
fcis-5209	22	29	attention	attention	NOUN
fcis-5209	22	30	mechanism	mechanism	NOUN
fcis-5209	22	31	to	to	PART
fcis-5209	22	32	strengthen	strengthen	VERB
fcis-5209	22	33	the	the	DET
fcis-5209	22	34	interaction	interaction	NOUN
fcis-5209	22	35	between	between	ADP
fcis-5209	22	36	texts	text	NOUN
fcis-5209	22	37	.	.	PUNCT
fcis-5209	23	1	in	in	ADP
fcis-5209	23	2	order	order	NOUN
fcis-5209	23	3	to	to	PART
fcis-5209	23	4	solve	solve	VERB
fcis-5209	23	5	the	the	DET
fcis-5209	23	6	problem	problem	NOUN
fcis-5209	23	7	of	of	ADP
fcis-5209	23	8	insufficient	insufficient	ADJ
fcis-5209	23	9	interaction	interaction	NOUN
fcis-5209	23	10	between	between	ADP
fcis-5209	23	11	texts	text	NOUN
fcis-5209	23	12	,	,	PUNCT
fcis-5209	23	13	wang	wang	PROPN
fcis-5209	23	14	et	et	PROPN
fcis-5209	23	15	al	al	PROPN
fcis-5209	23	16	.	.	PROPN
fcis-5209	23	17	proposed	propose	VERB
fcis-5209	23	18	a	a	DET
fcis-5209	23	19	multi	multi	ADJ
fcis-5209	23	20	angle	angle	NOUN
fcis-5209	23	21	matching	matching	NOUN
fcis-5209	23	22	model	model	NOUN
fcis-5209	23	23	bimpm	bimpm	NOUN
fcis-5209	23	24	,	,	PUNCT
fcis-5209	23	25	which	which	PRON
fcis-5209	23	26	uses	use	VERB
fcis-5209	23	27	a	a	DET
fcis-5209	23	28	variety	variety	NOUN
fcis-5209	23	29	of	of	ADP
fcis-5209	23	30	information	information	NOUN
fcis-5209	23	31	interaction	interaction	NOUN
fcis-5209	23	32	methods	method	NOUN
fcis-5209	23	33	to	to	PART
fcis-5209	23	34	improve	improve	VERB
fcis-5209	23	35	the	the	DET
fcis-5209	23	36	degree	degree	NOUN
fcis-5209	23	37	of	of	ADP
fcis-5209	23	38	text	text	NOUN
fcis-5209	23	39	interaction	interaction	NOUN
fcis-5209	23	40	.	.	PUNCT
fcis-5209	24	1	although	although	SCONJ
fcis-5209	24	2	the	the	DET
fcis-5209	24	3	above	above	ADJ
fcis-5209	24	4	model	model	NOUN
fcis-5209	24	5	can	can	AUX
fcis-5209	24	6	make	make	VERB
fcis-5209	24	7	use	use	NOUN
fcis-5209	24	8	of	of	ADP
fcis-5209	24	9	the	the	DET
fcis-5209	24	10	semantic	semantic	ADJ
fcis-5209	24	11	information	information	NOUN
fcis-5209	24	12	and	and	CCONJ
fcis-5209	24	13	interactive	interactive	ADJ
fcis-5209	24	14	information	information	NOUN
fcis-5209	24	15	between	between	ADP
fcis-5209	24	16	texts	text	NOUN
fcis-5209	24	17	,	,	PUNCT
fcis-5209	24	18	the	the	DET
fcis-5209	24	19	traditional	traditional	ADJ
fcis-5209	24	20	static	static	ADJ
fcis-5209	24	21	text	text	NOUN
fcis-5209	24	22	semantic	semantic	ADJ
fcis-5209	24	23	acquisition	acquisition	NOUN
fcis-5209	24	24	methods	method	NOUN
fcis-5209	24	25	such	such	ADJ
fcis-5209	24	26	as	as	ADP
fcis-5209	24	27	word2vec	word2vec	PRON
fcis-5209	24	28	and	and	CCONJ
fcis-5209	24	29	glove	glove	NOUN
fcis-5209	24	30	are	be	AUX
fcis-5209	24	31	still	still	ADV
fcis-5209	24	32	used	use	VERB
fcis-5209	24	33	in	in	ADP
fcis-5209	24	34	the	the	DET
fcis-5209	24	35	semantic	semantic	ADJ
fcis-5209	24	36	coding	code	VERB
fcis-5209	24	37	stage	stage	NOUN
fcis-5209	24	38	,	,	PUNCT
fcis-5209	24	39	which	which	PRON
fcis-5209	24	40	can	can	AUX
fcis-5209	24	41	not	not	PART
fcis-5209	24	42	express	express	VERB
fcis-5209	24	43	the	the	DET
fcis-5209	24	44	semantic	semantic	ADJ
fcis-5209	24	45	information	information	NOUN
fcis-5209	24	46	of	of	ADP
fcis-5209	24	47	text	text	NOUN
fcis-5209	24	48	well	well	ADV
fcis-5209	24	49	.	.	PUNCT
fcis-5209	25	1	the	the	DET
fcis-5209	25	2	method	method	NOUN
fcis-5209	25	3	based	base	VERB
fcis-5209	25	4	on	on	ADP
fcis-5209	25	5	pretraining	pretraine	VERB
fcis-5209	25	6	language	language	NOUN
fcis-5209	25	7	model	model	NOUN
fcis-5209	25	8	refers	refer	VERB
fcis-5209	25	9	to	to	ADP
fcis-5209	25	10	the	the	DET
fcis-5209	25	11	unsupervised	unsupervised	ADJ
fcis-5209	25	12	way	way	NOUN
fcis-5209	25	13	to	to	PART
fcis-5209	25	14	train	train	VERB
fcis-5209	25	15	the	the	DET
fcis-5209	25	16	language	language	NOUN
fcis-5209	25	17	model	model	NOUN
fcis-5209	25	18	in	in	ADP
fcis-5209	25	19	large	large	ADJ
fcis-5209	25	20	corpus	corpu	NOUN
fcis-5209	25	21	in	in	ADP
fcis-5209	25	22	advance	advance	NOUN
fcis-5209	25	23	,	,	PUNCT
fcis-5209	25	24	and	and	CCONJ
fcis-5209	25	25	then	then	ADV
fcis-5209	25	26	load	load	VERB
fcis-5209	25	27	the	the	DET
fcis-5209	25	28	pretrained	pretraine	VERB
fcis-5209	25	29	model	model	NOUN
fcis-5209	25	30	weights	weight	VERB
fcis-5209	25	31	to	to	PART
fcis-5209	25	32	finetune	finetune	VERB
fcis-5209	25	33	the	the	DET
fcis-5209	25	34	specific	specific	ADJ
fcis-5209	25	35	downstream	downstream	ADJ
fcis-5209	25	36	tasks	task	NOUN
fcis-5209	25	37	.	.	PUNCT
fcis-5209	26	1	in	in	ADP
fcis-5209	26	2	2018	2018	NUM
fcis-5209	26	3	,	,	PUNCT
fcis-5209	26	4	the	the	DET
fcis-5209	26	5	bert	bert	PROPN
fcis-5209	26	6	(	(	PUNCT
fcis-5209	26	7	bidirectional	bidirectional	ADJ
fcis-5209	26	8	encoder	encoder	NOUN
fcis-5209	26	9	representations	representation	VERB
fcis-5209	26	10	from	from	ADP
fcis-5209	26	11	transformers	transformer	NOUN
fcis-5209	26	12	)	)	PUNCT
fcis-5209	26	13	model	model	NOUN
fcis-5209	26	14	based	base	VERB
fcis-5209	26	15	on	on	ADP
fcis-5209	26	16	bidirectional	bidirectional	ADJ
fcis-5209	26	17	transformer	transformer	NOUN
fcis-5209	26	18	came	come	VERB
fcis-5209	26	19	out	out	ADP
fcis-5209	26	20	,	,	PUNCT
fcis-5209	26	21	refreshing	refresh	VERB
fcis-5209	26	22	the	the	DET
fcis-5209	26	23	list	list	NOUN
fcis-5209	26	24	of	of	ADP
fcis-5209	26	25	many	many	ADJ
fcis-5209	26	26	tasks	task	NOUN
fcis-5209	26	27	in	in	ADP
fcis-5209	26	28	nlp	nlp	ADJ
fcis-5209	26	29	field	field	NOUN
fcis-5209	26	30	.	.	PUNCT
fcis-5209	27	1	therefore	therefore	ADV
fcis-5209	27	2	,	,	PUNCT
fcis-5209	27	3	the	the	DET
fcis-5209	27	4	method	method	NOUN
fcis-5209	27	5	based	base	VERB
fcis-5209	27	6	on	on	ADP
fcis-5209	27	7	the	the	DET
fcis-5209	27	8	pretraining	pretraine	VERB
fcis-5209	27	9	language	language	NOUN
fcis-5209	27	10	model	model	NOUN
fcis-5209	27	11	has	have	AUX
fcis-5209	27	12	become	become	VERB
fcis-5209	27	13	a	a	DET
fcis-5209	27	14	research	research	NOUN
fcis-5209	27	15	hotspot	hotspot	NOUN
fcis-5209	27	16	in	in	ADP
fcis-5209	27	17	recent	recent	ADJ
fcis-5209	27	18	years	year	NOUN
fcis-5209	27	19	.	.	PUNCT
fcis-5209	28	1	so	so	ADV
fcis-5209	28	2	since	since	SCONJ
fcis-5209	28	3	2018	2018	NUM
fcis-5209	28	4	,	,	PUNCT
fcis-5209	28	5	many	many	ADJ
fcis-5209	28	6	researchers	researcher	NOUN
fcis-5209	28	7	have	have	AUX
fcis-5209	28	8	improved	improve	VERB
fcis-5209	28	9	the	the	DET
fcis-5209	28	10	bert	bert	PROPN
fcis-5209	28	11	model	model	NOUN
fcis-5209	28	12	[	[	X
fcis-5209	28	13	16	16	NUM
fcis-5209	28	14	,	,	PUNCT
fcis-5209	28	15	17	17	NUM
fcis-5209	28	16	,	,	PUNCT
fcis-5209	28	17	18	18	NUM
fcis-5209	28	18	]	]	PUNCT
fcis-5209	28	19	.	.	PUNCT
fcis-5209	29	1	cui	cui	NOUN
fcis-5209	29	2	et	et	PROPN
fcis-5209	29	3	al	al	PROPN
fcis-5209	29	4	.	.	PROPN
fcis-5209	29	5	changed	change	VERB
fcis-5209	29	6	the	the	DET
fcis-5209	29	7	pretraining	pretraine	VERB
fcis-5209	29	8	method	method	NOUN
fcis-5209	29	9	on	on	ADP
fcis-5209	29	10	the	the	DET
fcis-5209	29	11	basis	basis	NOUN
fcis-5209	29	12	of	of	ADP
fcis-5209	29	13	bert	bert	PROPN
fcis-5209	29	14	model	model	NOUN
fcis-5209	29	15	and	and	CCONJ
fcis-5209	29	16	trained	train	VERB
fcis-5209	29	17	the	the	DET
fcis-5209	29	18	chinese	chinese	PROPN
fcis-5209	29	19	-	-	PUNCT
fcis-5209	29	20	wwm	wwm	PROPN
fcis-5209	29	21	-	-	PUNCT
fcis-5209	29	22	bert	bert	PROPN
fcis-5209	29	23	model	model	NOUN
fcis-5209	29	24	for	for	ADP
fcis-5209	29	25	chinese	chinese	ADJ
fcis-5209	29	26	tasks	task	NOUN
fcis-5209	29	27	on	on	ADP
fcis-5209	29	28	a	a	DET
fcis-5209	29	29	large	large	ADJ
fcis-5209	29	30	chinese	chinese	ADJ
fcis-5209	29	31	corpus	corpus	NOUN
fcis-5209	30	1	[	[	X
fcis-5209	30	2	19	19	NUM
fcis-5209	30	3	,	,	PUNCT
fcis-5209	30	4	20	20	NUM
fcis-5209	30	5	]	]	PUNCT
fcis-5209	30	6	.	.	PUNCT
fcis-5209	31	1	this	this	DET
fcis-5209	31	2	model	model	NOUN
fcis-5209	31	3	has	have	AUX
fcis-5209	31	4	achieved	achieve	VERB
fcis-5209	31	5	good	good	ADJ
fcis-5209	31	6	results	result	NOUN
fcis-5209	31	7	in	in	ADP
fcis-5209	31	8	chinese	chinese	ADJ
fcis-5209	31	9	natural	natural	ADJ
fcis-5209	31	10	language	language	NOUN
fcis-5209	31	11	processing	processing	NOUN
fcis-5209	31	12	tasks	task	NOUN
fcis-5209	31	13	,	,	PUNCT
fcis-5209	31	14	including	include	VERB
fcis-5209	31	15	text	text	NOUN
fcis-5209	31	16	matching	matching	NOUN
fcis-5209	31	17	tasks	task	NOUN
fcis-5209	31	18	.	.	PUNCT
fcis-5209	32	1	however	however	ADV
fcis-5209	32	2	,	,	PUNCT
fcis-5209	32	3	the	the	DET
fcis-5209	32	4	text	text	NOUN
fcis-5209	32	5	semantic	semantic	ADJ
fcis-5209	32	6	information	information	NOUN
fcis-5209	32	7	obtained	obtain	VERB
fcis-5209	32	8	by	by	ADP
fcis-5209	32	9	fine	fine	ADV
fcis-5209	32	10	-	-	PUNCT
fcis-5209	32	11	tuning	tune	VERB
fcis-5209	32	12	specific	specific	ADJ
fcis-5209	32	13	text	text	NOUN
fcis-5209	32	14	matching	matching	NOUN
fcis-5209	32	15	tasks	task	NOUN
fcis-5209	32	16	using	use	VERB
fcis-5209	32	17	bert	bert	PROPN
fcis-5209	32	18	model	model	NOUN
fcis-5209	32	19	alone	alone	ADV
fcis-5209	32	20	is	be	AUX
fcis-5209	32	21	insufficient	insufficient	ADJ
fcis-5209	32	22	,	,	PUNCT
fcis-5209	32	23	therefore	therefore	ADV
fcis-5209	32	24	,	,	PUNCT
fcis-5209	32	25	based	base	VERB
fcis-5209	32	26	on	on	ADP
fcis-5209	32	27	this	this	DET
fcis-5209	32	28	problem	problem	NOUN
fcis-5209	32	29	,	,	PUNCT
fcis-5209	32	30	this	this	DET
fcis-5209	32	31	paper	paper	NOUN
fcis-5209	32	32	proposes	propose	VERB
fcis-5209	32	33	a	a	DET
fcis-5209	32	34	method	method	NOUN
fcis-5209	32	35	based	base	VERB
fcis-5209	32	36	on	on	ADP
fcis-5209	32	37	pretraining	pretraine	VERB
fcis-5209	32	38	bert	bert	PROPN
fcis-5209	32	39	model	model	NOUN
fcis-5209	32	40	fusion	fusion	NOUN
fcis-5209	32	41	of	of	ADP
fcis-5209	32	42	bilstm	bilstm	NOUN
fcis-5209	32	43	and	and	CCONJ
fcis-5209	32	44	multi	multi	ADJ
fcis-5209	32	45	-	-	ADJ
fcis-5209	32	46	scale	scale	ADJ
fcis-5209	32	47	cnn	cnn	PROPN
fcis-5209	32	48	to	to	PART
fcis-5209	32	49	remodel	remodel	VERB
fcis-5209	32	50	the	the	DET
fcis-5209	32	51	text	text	NOUN
fcis-5209	32	52	matching	matching	NOUN
fcis-5209	32	53	task	task	NOUN
fcis-5209	32	54	,	,	PUNCT
fcis-5209	32	55	further	far	ADV
fcis-5209	32	56	extract	extract	VERB
fcis-5209	32	57	global	global	ADJ
fcis-5209	32	58	semantic	semantic	ADJ
fcis-5209	32	59	information	information	NOUN
fcis-5209	32	60	and	and	CCONJ
fcis-5209	32	61	local	local	ADJ
fcis-5209	32	62	semantic	semantic	ADJ
fcis-5209	32	63	information	information	NOUN
fcis-5209	32	64	to	to	PART
fcis-5209	32	65	improve	improve	VERB
fcis-5209	32	66	the	the	DET
fcis-5209	32	67	accuracy	accuracy	NOUN
fcis-5209	32	68	of	of	ADP
fcis-5209	32	69	text	text	NOUN
fcis-5209	32	70	matching	matching	NOUN
fcis-5209	32	71	.	.	PUNCT
fcis-5209	33	1	2	2	X
fcis-5209	33	2	.	.	X
fcis-5209	33	3	our	our	PRON
fcis-5209	33	4	method	method	NOUN
fcis-5209	33	5	the	the	DET
fcis-5209	33	6	model	model	NOUN
fcis-5209	33	7	proposed	propose	VERB
fcis-5209	33	8	in	in	ADP
fcis-5209	33	9	this	this	DET
fcis-5209	33	10	paper	paper	NOUN
fcis-5209	33	11	mainly	mainly	ADV
fcis-5209	33	12	obtains	obtain	VERB
fcis-5209	33	13	the	the	DET
fcis-5209	33	14	word	word	NOUN
fcis-5209	33	15	vector	vector	NOUN
fcis-5209	33	16	representation	representation	NOUN
fcis-5209	33	17	of	of	ADP
fcis-5209	33	18	the	the	DET
fcis-5209	33	19	text	text	NOUN
fcis-5209	33	20	with	with	ADP
fcis-5209	33	21	context	context	PROPN
fcis-5209	33	22	semantic	semantic	ADJ
fcis-5209	33	23	41	41	NUM
fcis-5209	33	24	information	information	NOUN
fcis-5209	33	25	through	through	ADP
fcis-5209	33	26	bert	bert	PROPN
fcis-5209	33	27	,	,	PUNCT
fcis-5209	33	28	then	then	ADV
fcis-5209	33	29	further	far	ADV
fcis-5209	33	30	obtains	obtain	VERB
fcis-5209	33	31	the	the	DET
fcis-5209	33	32	semantic	semantic	ADJ
fcis-5209	33	33	information	information	NOUN
fcis-5209	33	34	of	of	ADP
fcis-5209	33	35	the	the	DET
fcis-5209	33	36	text	text	NOUN
fcis-5209	33	37	through	through	ADP
fcis-5209	33	38	the	the	DET
fcis-5209	33	39	bilstm	bilstm	NOUN
fcis-5209	33	40	model	model	NOUN
fcis-5209	33	41	,	,	PUNCT
fcis-5209	33	42	and	and	CCONJ
fcis-5209	33	43	then	then	ADV
fcis-5209	33	44	the	the	DET
fcis-5209	33	45	key	key	ADJ
fcis-5209	33	46	information	information	NOUN
fcis-5209	33	47	of	of	ADP
fcis-5209	33	48	different	different	ADJ
fcis-5209	33	49	scales	scale	NOUN
fcis-5209	33	50	obtained	obtain	VERB
fcis-5209	33	51	through	through	ADP
fcis-5209	33	52	cnn	cnn	PROPN
fcis-5209	33	53	with	with	ADP
fcis-5209	33	54	different	different	ADJ
fcis-5209	33	55	size	size	NOUN
fcis-5209	33	56	convolution	convolution	NOUN
fcis-5209	33	57	cores	core	NOUN
fcis-5209	33	58	,	,	PUNCT
fcis-5209	33	59	and	and	CCONJ
fcis-5209	33	60	finally	finally	ADV
fcis-5209	33	61	classifies	classify	VERB
fcis-5209	33	62	through	through	ADP
fcis-5209	33	63	the	the	DET
fcis-5209	33	64	classification	classification	NOUN
fcis-5209	33	65	layer	layer	NOUN
fcis-5209	33	66	.	.	PUNCT
fcis-5209	34	1	the	the	DET
fcis-5209	34	2	overall	overall	ADJ
fcis-5209	34	3	architecture	architecture	NOUN
fcis-5209	34	4	is	be	AUX
fcis-5209	34	5	shown	show	VERB
fcis-5209	34	6	in	in	ADP
fcis-5209	34	7	the	the	DET
fcis-5209	34	8	following	following	NOUN
fcis-5209	34	9	,	,	PUNCT
fcis-5209	34	10	see	see	VERB
fcis-5209	34	11	figure	figure	NOUN
fcis-5209	34	12	1	1	NUM
fcis-5209	34	13	.	.	PUNCT
fcis-5209	34	14	figure	figure	NOUN
fcis-5209	34	15	1	1	NUM
fcis-5209	34	16	.	.	PUNCT
fcis-5209	34	17	overall	overall	ADJ
fcis-5209	34	18	architecture	architecture	NOUN
fcis-5209	34	19	diagram	diagram	NOUN
fcis-5209	34	20	(	(	PUNCT
fcis-5209	34	21	1	1	X
fcis-5209	34	22	)	)	PUNCT
fcis-5209	34	23	bert	bert	NOUN
fcis-5209	34	24	according	accord	VERB
fcis-5209	34	25	to	to	ADP
fcis-5209	34	26	the	the	DET
fcis-5209	34	27	characteristics	characteristic	NOUN
fcis-5209	34	28	of	of	ADP
fcis-5209	34	29	the	the	DET
fcis-5209	34	30	input	input	NOUN
fcis-5209	34	31	layer	layer	NOUN
fcis-5209	34	32	of	of	ADP
fcis-5209	34	33	the	the	DET
fcis-5209	34	34	bert	bert	PROPN
fcis-5209	34	35	model	model	NOUN
fcis-5209	34	36	,	,	PUNCT
fcis-5209	34	37	this	this	DET
fcis-5209	34	38	paper	paper	NOUN
fcis-5209	34	39	first	first	ADV
fcis-5209	34	40	separates	separate	VERB
fcis-5209	34	41	the	the	DET
fcis-5209	34	42	two	two	NUM
fcis-5209	34	43	sentences	sentence	NOUN
fcis-5209	34	44	with	with	ADP
fcis-5209	34	45	[	[	X
fcis-5209	34	46	cls	cls	X
fcis-5209	34	47	]	]	PUNCT
fcis-5209	34	48	and	and	CCONJ
fcis-5209	34	49	[	[	X
fcis-5209	34	50	sep	sep	X
fcis-5209	34	51	]	]	X
fcis-5209	34	52	,	,	PUNCT
fcis-5209	34	53	and	and	CCONJ
fcis-5209	34	54	then	then	ADV
fcis-5209	34	55	gets	get	VERB
fcis-5209	34	56	the	the	DET
fcis-5209	34	57	word	word	NOUN
fcis-5209	34	58	embedding	embed	VERB
fcis-5209	34	59	of	of	ADP
fcis-5209	34	60	the	the	DET
fcis-5209	34	61	text	text	NOUN
fcis-5209	34	62	through	through	ADP
fcis-5209	34	63	the	the	DET
fcis-5209	34	64	bert	bert	NOUN
fcis-5209	34	65	embedding	embed	VERB
fcis-5209	34	66	layer	layer	NOUN
fcis-5209	34	67	.	.	PUNCT
fcis-5209	35	1	it	it	PRON
fcis-5209	35	2	consists	consist	VERB
fcis-5209	35	3	of	of	ADP
fcis-5209	35	4	three	three	NUM
fcis-5209	35	5	parts	part	NOUN
fcis-5209	35	6	,	,	PUNCT
fcis-5209	35	7	namely	namely	ADV
fcis-5209	35	8	token	token	ADJ
fcis-5209	35	9	embedding	embed	VERB
fcis-5209	35	10	,	,	PUNCT
fcis-5209	35	11	segment	segment	NOUN
fcis-5209	35	12	embedding	embed	VERB
fcis-5209	35	13	,	,	PUNCT
fcis-5209	35	14	and	and	CCONJ
fcis-5209	35	15	position	position	NOUN
fcis-5209	35	16	embedding	embed	VERB
fcis-5209	35	17	.	.	PUNCT
fcis-5209	36	1	the	the	DET
fcis-5209	36	2	calculation	calculation	NOUN
fcis-5209	36	3	formula	formula	NOUN
fcis-5209	36	4	is	be	AUX
fcis-5209	36	5	as	as	SCONJ
fcis-5209	36	6	follows	follow	VERB
fcis-5209	36	7	:	:	PUNCT
fcis-5209	36	8	token	token	PROPN
fcis-5209	36	9	seg	seg	PROPN
fcis-5209	36	10	pose	pose	VERB
fcis-5209	36	11	e	e	NOUN
fcis-5209	36	12	e	e	NOUN
fcis-5209	36	13	e=	e=	X
fcis-5209	37	1	+	+	X
fcis-5209	38	1	+	+	CCONJ
fcis-5209	38	2	(	(	PUNCT
fcis-5209	38	3	1	1	X
fcis-5209	38	4	)	)	PUNCT
fcis-5209	38	5	input	input	NOUN
fcis-5209	38	6	the	the	DET
fcis-5209	38	7	word	word	NOUN
fcis-5209	38	8	embedding	embed	VERB
fcis-5209	38	9	e	e	NOUN
fcis-5209	38	10	obtained	obtain	VERB
fcis-5209	38	11	from	from	ADP
fcis-5209	38	12	bert	bert	NOUN
fcis-5209	38	13	embedding	embed	VERB
fcis-5209	38	14	layer	layer	NOUN
fcis-5209	38	15	into	into	ADP
fcis-5209	38	16	the	the	DET
fcis-5209	38	17	bert	bert	NOUN
fcis-5209	38	18	coding	code	VERB
fcis-5209	38	19	layer	layer	NOUN
fcis-5209	38	20	,	,	PUNCT
fcis-5209	38	21	and	and	CCONJ
fcis-5209	38	22	then	then	ADV
fcis-5209	38	23	the	the	DET
fcis-5209	38	24	sum	sum	NOUN
fcis-5209	38	25	of	of	ADP
fcis-5209	38	26	hidden	hide	VERB
fcis-5209	38	27	layer	layer	NOUN
fcis-5209	38	28	vectors	vector	NOUN
fcis-5209	38	29	output	output	VERB
fcis-5209	38	30	by	by	ADP
fcis-5209	38	31	the	the	DET
fcis-5209	38	32	last	last	ADJ
fcis-5209	38	33	four	four	NUM
fcis-5209	38	34	layers	layer	NOUN
fcis-5209	38	35	of	of	ADP
fcis-5209	38	36	transformer	transformer	NOUN
fcis-5209	38	37	in	in	ADP
fcis-5209	38	38	bert	bert	NOUN
fcis-5209	38	39	coding	code	VERB
fcis-5209	38	40	layer	layer	NOUN
fcis-5209	38	41	is	be	AUX
fcis-5209	38	42	selected	select	VERB
fcis-5209	38	43	as	as	SCONJ
fcis-5209	38	44	the	the	DET
fcis-5209	38	45	semantic	semantic	ADJ
fcis-5209	38	46	feature	feature	NOUN
fcis-5209	38	47	h	h	NOUN
fcis-5209	38	48	extracted	extract	VERB
fcis-5209	38	49	from	from	ADP
fcis-5209	38	50	bert	bert	PROPN
fcis-5209	38	51	model	model	NOUN
fcis-5209	38	52	.	.	PUNCT
fcis-5209	39	1	the	the	DET
fcis-5209	39	2	calculation	calculation	NOUN
fcis-5209	39	3	formula	formula	NOUN
fcis-5209	39	4	is	be	AUX
fcis-5209	39	5	as	as	SCONJ
fcis-5209	39	6	follows	follow	VERB
fcis-5209	39	7	:	:	PUNCT
fcis-5209	39	8	12	12	NUM
fcis-5209	39	9	9	9	NUM
fcis-5209	39	10	_	_	PUNCT
fcis-5209	40	1	i	i	PRON
fcis-5209	40	2	i	i	PRON
fcis-5209	40	3	h	h	VERB
fcis-5209	40	4	hidden	hide	VERB
fcis-5209	40	5	state	state	NOUN
fcis-5209	40	6	=	=	SYM
fcis-5209	40	7	=	=	NOUN
fcis-5209	40	8			X
fcis-5209	40	9	(	(	PUNCT
fcis-5209	40	10	2	2	NUM
fcis-5209	40	11	)	)	PUNCT
fcis-5209	40	12	where	where	SCONJ
fcis-5209	40	13	i	i	PRON
fcis-5209	40	14	represents	represent	VERB
fcis-5209	40	15	the	the	DET
fcis-5209	40	16	number	number	NOUN
fcis-5209	40	17	of	of	ADP
fcis-5209	40	18	layers	layer	NOUN
fcis-5209	40	19	of	of	ADP
fcis-5209	40	20	the	the	DET
fcis-5209	40	21	transformer	transformer	NOUN
fcis-5209	40	22	,	,	PUNCT
fcis-5209	40	23	hidden_states	hidden_state	NOUN
fcis-5209	40	24	represents	represent	VERB
fcis-5209	40	25	the	the	DET
fcis-5209	40	26	hidden	hide	VERB
fcis-5209	40	27	layer	layer	NOUN
fcis-5209	40	28	vector	vector	NOUN
fcis-5209	40	29	obtained	obtain	VERB
fcis-5209	40	30	through	through	ADP
fcis-5209	40	31	the	the	DET
fcis-5209	40	32	ith	ith	PROPN
fcis-5209	40	33	transformer	transformer	NOUN
fcis-5209	40	34	.	.	PUNCT
fcis-5209	41	1	after	after	ADP
fcis-5209	41	2	calculating	calculate	VERB
fcis-5209	41	3	h	h	NOUN
fcis-5209	41	4	,	,	PUNCT
fcis-5209	41	5	input	input	NOUN
fcis-5209	41	6	e	e	NOUN
fcis-5209	41	7	and	and	CCONJ
fcis-5209	41	8	h	h	NOUN
fcis-5209	41	9	into	into	ADP
fcis-5209	41	10	the	the	DET
fcis-5209	41	11	subsequent	subsequent	ADJ
fcis-5209	41	12	layers	layer	NOUN
fcis-5209	41	13	to	to	PART
fcis-5209	41	14	further	far	ADV
fcis-5209	41	15	extract	extract	VERB
fcis-5209	41	16	the	the	DET
fcis-5209	41	17	text	text	NOUN
fcis-5209	41	18	feature	feature	NOUN
fcis-5209	41	19	information	information	NOUN
fcis-5209	41	20	.	.	PUNCT
fcis-5209	42	1	(	(	PUNCT
fcis-5209	42	2	2	2	X
fcis-5209	42	3	)	)	PUNCT
fcis-5209	42	4	bilstm	bilstm	NOUN
fcis-5209	42	5	blocks	block	NOUN
fcis-5209	42	6	this	this	DET
fcis-5209	42	7	module	module	NOUN
fcis-5209	42	8	consists	consist	VERB
fcis-5209	42	9	of	of	ADP
fcis-5209	42	10	two	two	NUM
fcis-5209	42	11	layers	layer	NOUN
fcis-5209	42	12	of	of	ADP
fcis-5209	42	13	bilstm	bilstm	NOUN
fcis-5209	42	14	.	.	PUNCT
fcis-5209	43	1	the	the	DET
fcis-5209	43	2	architecture	architecture	NOUN
fcis-5209	43	3	of	of	ADP
fcis-5209	43	4	bilstm	bilstm	NOUN
fcis-5209	43	5	is	be	AUX
fcis-5209	43	6	shown	show	VERB
fcis-5209	43	7	in	in	ADP
fcis-5209	43	8	the	the	DET
fcis-5209	43	9	following	following	NOUN
fcis-5209	43	10	,	,	PUNCT
fcis-5209	43	11	see	see	VERB
fcis-5209	43	12	figure	figure	NOUN
fcis-5209	43	13	2	2	NUM
fcis-5209	43	14	.	.	PUNCT
fcis-5209	43	15	figure	figure	NOUN
fcis-5209	43	16	2	2	NUM
fcis-5209	43	17	.	.	PUNCT
fcis-5209	43	18	bilstm	bilstm	NOUN
fcis-5209	43	19	where	where	SCONJ
fcis-5209	43	20	f	f	PROPN
fcis-5209	43	21	represents	represent	VERB
fcis-5209	43	22	forward	forward	ADV
fcis-5209	43	23	lstm	lstm	NOUN
fcis-5209	43	24	and	and	CCONJ
fcis-5209	43	25	b	b	NOUN
fcis-5209	43	26	represents	represent	VERB
fcis-5209	43	27	backward	backward	ADJ
fcis-5209	43	28	lstm	lstm	NOUN
fcis-5209	43	29	.	.	PUNCT
fcis-5209	44	1	the	the	DET
fcis-5209	44	2	calculation	calculation	NOUN
fcis-5209	44	3	formula	formula	NOUN
fcis-5209	44	4	is	be	AUX
fcis-5209	44	5	as	as	SCONJ
fcis-5209	44	6	follows	follow	VERB
fcis-5209	44	7	:	:	PUNCT
fcis-5209	44	8	1	1	NUM
fcis-5209	44	9	1	1	NUM
fcis-5209	44	10	1	1	NUM
fcis-5209	44	11	lstm	lstm	NOUN
fcis-5209	44	12	(	(	PUNCT
fcis-5209	44	13	,	,	PUNCT
fcis-5209	44	14	)	)	PUNCT
fcis-5209	44	15	(	(	PUNCT
fcis-5209	44	16	,	,	PUNCT
fcis-5209	44	17	)	)	PUNCT
fcis-5209	44	18	(	(	PUNCT
fcis-5209	44	19	)	)	PUNCT
fcis-5209	45	1	i	i	PRON
fcis-5209	45	2	i	i	PRON
fcis-5209	46	1	i	i	PRON
fcis-5209	46	2	i	i	PRON
fcis-5209	47	1	i	i	PRON
fcis-5209	47	2	i	i	PRON
fcis-5209	48	1	i	i	PRON
fcis-5209	48	2	i	i	VERB
fcis-5209	49	1	i	i	PRON
fcis-5209	49	2	n	n	VERB
fcis-5209	50	1	i	i	PRON
fcis-5209	51	1	i	i	PRON
fcis-5209	52	1	f	f	PROPN
fcis-5209	52	2	f	f	PROPN
fcis-5209	52	3	w	w	PROPN
fcis-5209	52	4	b	b	PROPN
fcis-5209	52	5	lstm	lstm	PROPN
fcis-5209	52	6	b	b	PROPN
fcis-5209	52	7	w	w	PROPN
fcis-5209	52	8	t	t	PROPN
fcis-5209	52	9	f	f	PROPN
fcis-5209	52	10	b	b	PROPN
fcis-5209	52	11	t	t	PROPN
fcis-5209	52	12	bilstm	bilstm	NOUN
fcis-5209	52	13	w	w	PROPN
fcis-5209	52	14	t	t	PROPN
fcis-5209	52	15	−	−	PROPN
fcis-5209	53	1	+	+	CCONJ
fcis-5209	53	2	=	=	SYM
fcis-5209	53	3	=	=	PUNCT
fcis-5209	53	4	=	=	PUNCT
fcis-5209	53	5	=	=	PUNCT
fcis-5209	53	6			ADJ
fcis-5209	53	7	=	=	SYM
fcis-5209	54	1	=	=	NOUN
fcis-5209	54	2			X
fcis-5209	54	3	(	(	PUNCT
fcis-5209	54	4	3	3	X
fcis-5209	54	5	)	)	PUNCT
fcis-5209	54	6	where	where	SCONJ
fcis-5209	54	7	represents	represent	VERB
fcis-5209	54	8	forward	forward	ADV
fcis-5209	54	9	lstm	lstm	NOUN
fcis-5209	54	10	unit	unit	NOUN
fcis-5209	54	11	,	,	PUNCT
fcis-5209	54	12	represents	represent	VERB
fcis-5209	54	13	backward	backward	ADJ
fcis-5209	54	14	lstm	lstm	ADJ
fcis-5209	54	15	unit	unit	NOUN
fcis-5209	54	16	,	,	PUNCT
fcis-5209	54	17			PROPN
fcis-5209	54	18	represents	represent	VERB
fcis-5209	54	19	splicing	splicing	NOUN
fcis-5209	54	20	operation	operation	NOUN
fcis-5209	54	21	,	,	PUNCT
fcis-5209	54	22	wi	wi	PROPN
fcis-5209	54	23	represents	represent	VERB
fcis-5209	54	24	the	the	DET
fcis-5209	54	25	ith	ith	PROPN
fcis-5209	54	26	word	word	NOUN
fcis-5209	54	27	,	,	PUNCT
fcis-5209	54	28	if	if	SCONJ
fcis-5209	54	29	represents	represent	VERB
fcis-5209	54	30	the	the	DET
fcis-5209	54	31	output	output	NOUN
fcis-5209	54	32	of	of	ADP
fcis-5209	54	33	wi	wi	PROPN
fcis-5209	54	34	through	through	ADP
fcis-5209	54	35	forward	forward	ADV
fcis-5209	54	36	lstm	lstm	PROPN
fcis-5209	54	37	,	,	PUNCT
fcis-5209	54	38	ib	ib	NOUN
fcis-5209	54	39	represents	represent	VERB
fcis-5209	54	40	the	the	DET
fcis-5209	54	41	output	output	NOUN
fcis-5209	54	42	of	of	ADP
fcis-5209	54	43	wi	wi	PROPN
fcis-5209	54	44	through	through	ADP
fcis-5209	54	45	backward	backward	ADJ
fcis-5209	54	46	lstm	lstm	PROPN
fcis-5209	54	47	,	,	PUNCT
fcis-5209	54	48	it	it	PRON
fcis-5209	54	49	represents	represent	VERB
fcis-5209	54	50	the	the	DET
fcis-5209	54	51	output	output	NOUN
fcis-5209	54	52	of	of	ADP
fcis-5209	54	53	wi	wi	PROPN
fcis-5209	54	54	through	through	ADP
fcis-5209	54	55	bilstm	bilstm	NOUN
fcis-5209	54	56	,	,	PUNCT
fcis-5209	54	57	t	t	PROPN
fcis-5209	54	58	represents	represent	VERB
fcis-5209	54	59	the	the	DET
fcis-5209	54	60	output	output	NOUN
fcis-5209	54	61	of	of	ADP
fcis-5209	54	62	the	the	DET
fcis-5209	54	63	whole	whole	ADJ
fcis-5209	54	64	sentence	sentence	NOUN
fcis-5209	54	65	w	w	NOUN
fcis-5209	54	66	after	after	ADP
fcis-5209	54	67	passing	pass	VERB
fcis-5209	54	68	through	through	ADP
fcis-5209	54	69	bilstm	bilstm	NOUN
fcis-5209	54	70	.	.	PUNCT
fcis-5209	55	1	in	in	ADP
fcis-5209	55	2	this	this	DET
fcis-5209	55	3	paper	paper	NOUN
fcis-5209	55	4	,	,	PUNCT
fcis-5209	55	5	the	the	DET
fcis-5209	55	6	word	word	NOUN
fcis-5209	55	7	embedding	embed	VERB
fcis-5209	55	8	e	e	NOUN
fcis-5209	55	9	and	and	CCONJ
fcis-5209	55	10	semantic	semantic	ADJ
fcis-5209	55	11	feature	feature	NOUN
fcis-5209	55	12	h	h	NOUN
fcis-5209	55	13	obtained	obtain	VERB
fcis-5209	55	14	from	from	ADP
fcis-5209	55	15	the	the	DET
fcis-5209	55	16	bert	bert	NOUN
fcis-5209	55	17	layer	layer	NOUN
fcis-5209	55	18	are	be	AUX
fcis-5209	55	19	input	input	VERB
fcis-5209	55	20	into	into	ADP
fcis-5209	55	21	this	this	DET
fcis-5209	55	22	layer	layer	NOUN
fcis-5209	55	23	respectively	respectively	ADV
fcis-5209	55	24	to	to	PART
fcis-5209	55	25	obtain	obtain	VERB
fcis-5209	55	26	the	the	DET
fcis-5209	55	27	overall	overall	ADJ
fcis-5209	55	28	semantic	semantic	ADJ
fcis-5209	55	29	information	information	NOUN
fcis-5209	55	30	te	te	PROPN
fcis-5209	55	31	and	and	CCONJ
fcis-5209	55	32	th	th	X
fcis-5209	55	33	,	,	PUNCT
fcis-5209	55	34	according	accord	VERB
fcis-5209	55	35	to	to	ADP
fcis-5209	55	36	the	the	DET
fcis-5209	55	37	above	above	ADJ
fcis-5209	55	38	bilstm	bilstm	NOUN
fcis-5209	55	39	calculation	calculation	NOUN
fcis-5209	55	40	formula	formula	NOUN
fcis-5209	55	41	,	,	PUNCT
fcis-5209	55	42	te	te	PROPN
fcis-5209	55	43	and	and	CCONJ
fcis-5209	55	44	th	th	NUM
fcis-5209	55	45	formulas	formula	NOUN
fcis-5209	55	46	are	be	AUX
fcis-5209	55	47	obtained	obtain	VERB
fcis-5209	55	48	as	as	SCONJ
fcis-5209	55	49	follows	follow	VERB
fcis-5209	55	50	:	:	PUNCT
fcis-5209	55	51	1	1	NUM
fcis-5209	55	52	1	1	NUM
fcis-5209	55	53	1	1	NUM
fcis-5209	55	54	1	1	NUM
fcis-5209	55	55	(	(	PUNCT
fcis-5209	55	56	)	)	PUNCT
fcis-5209	55	57	(	(	PUNCT
fcis-5209	55	58	)	)	PUNCT
fcis-5209	55	59	(	(	PUNCT
fcis-5209	55	60	)	)	PUNCT
fcis-5209	55	61	(	(	PUNCT
fcis-5209	55	62	)	)	PUNCT
fcis-5209	55	63	e	e	X
fcis-5209	55	64	e	e	X
fcis-5209	55	65	e	e	X
fcis-5209	55	66	h	h	NOUN
fcis-5209	55	67	h	h	NOUN
fcis-5209	55	68	h	h	NOUN
fcis-5209	55	69	t	t	PROPN
fcis-5209	55	70	bilstm	bilstm	NOUN
fcis-5209	55	71	e	e	PROPN
fcis-5209	55	72	t	t	PROPN
fcis-5209	55	73	bilstm	bilstm	NOUN
fcis-5209	55	74	t	t	PROPN
fcis-5209	55	75	t	t	PROPN
fcis-5209	55	76	bilstm	bilstm	NOUN
fcis-5209	55	77	h	h	PROPN
fcis-5209	55	78	t	t	PROPN
fcis-5209	55	79	bilstm	bilstm	NOUN
fcis-5209	55	80	t	t	NOUN
fcis-5209	56	1	=	=	PUNCT
fcis-5209	56	2	=	=	PUNCT
fcis-5209	57	1	=	=	PUNCT
fcis-5209	57	2	=	=	SYM
fcis-5209	57	3	(	(	PUNCT
fcis-5209	57	4	4	4	NUM
fcis-5209	57	5	)	)	PUNCT
fcis-5209	57	6	where	where	SCONJ
fcis-5209	57	7	represents	represent	VERB
fcis-5209	57	8	the	the	DET
fcis-5209	57	9	feature	feature	NOUN
fcis-5209	57	10	obtained	obtain	VERB
fcis-5209	57	11	after	after	ADP
fcis-5209	57	12	passing	pass	VERB
fcis-5209	57	13	through	through	ADP
fcis-5209	57	14	the	the	DET
fcis-5209	57	15	first	first	ADJ
fcis-5209	57	16	layer	layer	NOUN
fcis-5209	57	17	of	of	ADP
fcis-5209	57	18	bilstm	bilstm	NOUN
fcis-5209	57	19	,	,	PUNCT
fcis-5209	57	20	and	and	CCONJ
fcis-5209	57	21	then	then	ADV
fcis-5209	57	22	input	input	VERB
fcis-5209	57	23	it	it	PRON
fcis-5209	57	24	into	into	ADP
fcis-5209	57	25	the	the	DET
fcis-5209	57	26	second	second	ADJ
fcis-5209	57	27	layer	layer	NOUN
fcis-5209	57	28	of	of	ADP
fcis-5209	57	29	bilstm	bilstm	NOUN
fcis-5209	57	30	to	to	PART
fcis-5209	57	31	obtain	obtain	VERB
fcis-5209	57	32	et	et	NOUN
fcis-5209	57	33	,	,	PUNCT
fcis-5209	57	34	and	and	CCONJ
fcis-5209	57	35	the	the	DET
fcis-5209	57	36	calculation	calculation	NOUN
fcis-5209	57	37	of	of	ADP
fcis-5209	57	38	ht	ht	PROPN
fcis-5209	57	39	is	be	AUX
fcis-5209	57	40	the	the	DET
fcis-5209	57	41	same	same	ADJ
fcis-5209	57	42	as	as	ADP
fcis-5209	57	43	above	above	ADV
fcis-5209	57	44	.	.	PUNCT
fcis-5209	58	1	(	(	PUNCT
fcis-5209	58	2	3	3	X
fcis-5209	58	3	)	)	PUNCT
fcis-5209	58	4	conv	conv	ADJ
fcis-5209	58	5	block	block	NOUN
fcis-5209	58	6	this	this	DET
fcis-5209	58	7	layer	layer	NOUN
fcis-5209	58	8	is	be	AUX
fcis-5209	58	9	a	a	DET
fcis-5209	58	10	multi	multi	ADJ
fcis-5209	58	11	-	-	ADJ
fcis-5209	58	12	scale	scale	ADJ
fcis-5209	58	13	convolution	convolution	NOUN
fcis-5209	58	14	block	block	NOUN
fcis-5209	58	15	.	.	PUNCT
fcis-5209	59	1	cnn	cnn	PROPN
fcis-5209	59	2	with	with	ADP
fcis-5209	59	3	different	different	ADJ
fcis-5209	59	4	convolution	convolution	NOUN
fcis-5209	59	5	kernel	kernel	NOUN
fcis-5209	59	6	sizes	size	NOUN
fcis-5209	59	7	is	be	AUX
fcis-5209	59	8	used	use	VERB
fcis-5209	59	9	to	to	PART
fcis-5209	59	10	further	far	ADV
fcis-5209	59	11	process	process	VERB
fcis-5209	59	12	the	the	DET
fcis-5209	59	13	semantic	semantic	ADJ
fcis-5209	59	14	features	feature	NOUN
fcis-5209	59	15	extracted	extract	VERB
fcis-5209	59	16	by	by	ADP
fcis-5209	59	17	bilstm	bilstm	NOUN
fcis-5209	59	18	to	to	PART
fcis-5209	59	19	obtain	obtain	VERB
fcis-5209	59	20	the	the	DET
fcis-5209	59	21	local	local	ADJ
fcis-5209	59	22	key	key	ADJ
fcis-5209	59	23	information	information	NOUN
fcis-5209	59	24	of	of	ADP
fcis-5209	59	25	the	the	DET
fcis-5209	59	26	text	text	NOUN
fcis-5209	59	27	.	.	PUNCT
fcis-5209	60	1	since	since	SCONJ
fcis-5209	60	2	the	the	DET
fcis-5209	60	3	dimensions	dimension	NOUN
fcis-5209	60	4	of	of	ADP
fcis-5209	60	5	the	the	DET
fcis-5209	60	6	semantic	semantic	ADJ
fcis-5209	60	7	features	feature	NOUN
fcis-5209	60	8	te	te	PROPN
fcis-5209	60	9	and	and	CCONJ
fcis-5209	60	10	th	th	X
fcis-5209	60	11	extracted	extract	VERB
fcis-5209	60	12	from	from	ADP
fcis-5209	60	13	the	the	DET
fcis-5209	60	14	bilstm	bilstm	NOUN
fcis-5209	60	15	layer	layer	NOUN
fcis-5209	60	16	are	be	AUX
fcis-5209	60	17	all	all	PRON
fcis-5209	60	18	[	[	X
fcis-5209	60	19	batch_size	batch_size	VERB
fcis-5209	60	20	,	,	PUNCT
fcis-5209	60	21	seq_len	seq_len	ADJ
fcis-5209	60	22	,	,	PUNCT
fcis-5209	60	23	768	768	NUM
fcis-5209	60	24	]	]	PUNCT
fcis-5209	60	25	,	,	PUNCT
fcis-5209	60	26	and	and	CCONJ
fcis-5209	60	27	the	the	DET
fcis-5209	60	28	length	length	NOUN
fcis-5209	60	29	of	of	ADP
fcis-5209	60	30	the	the	DET
fcis-5209	60	31	phase	phase	NOUN
fcis-5209	60	32	mostly	mostly	ADV
fcis-5209	60	33	2	2	NUM
fcis-5209	60	34	to	to	PART
fcis-5209	60	35	4	4	NUM
fcis-5209	60	36	characters	character	NOUN
fcis-5209	60	37	,	,	PUNCT
fcis-5209	60	38	this	this	DET
fcis-5209	60	39	layer	layer	NOUN
fcis-5209	60	40	selects	select	VERB
fcis-5209	60	41	twodimensional	twodimensional	ADJ
fcis-5209	60	42	convolutions	convolution	NOUN
fcis-5209	60	43	with	with	ADP
fcis-5209	60	44	convolution	convolution	NOUN
fcis-5209	60	45	kernels	kernel	NOUN
fcis-5209	60	46	of	of	ADP
fcis-5209	60	47	2	2	NUM
fcis-5209	60	48	*	*	SYM
fcis-5209	60	49	768	768	NUM
fcis-5209	60	50	,	,	PUNCT
fcis-5209	60	51	3	3	NUM
fcis-5209	60	52	*	*	SYM
fcis-5209	60	53	768	768	NUM
fcis-5209	60	54	,	,	PUNCT
fcis-5209	60	55	and	and	CCONJ
fcis-5209	60	56	4	4	NUM
fcis-5209	60	57	*	*	NOUN
fcis-5209	60	58	768	768	NUM
fcis-5209	60	59	respectively	respectively	ADV
fcis-5209	60	60	to	to	PART
fcis-5209	60	61	extract	extract	VERB
fcis-5209	60	62	local	local	ADJ
fcis-5209	60	63	key	key	ADJ
fcis-5209	60	64	features	feature	NOUN
fcis-5209	60	65	,	,	PUNCT
fcis-5209	60	66	and	and	CCONJ
fcis-5209	60	67	then	then	ADV
fcis-5209	60	68	uses	use	VERB
fcis-5209	60	69	the	the	DET
fcis-5209	60	70	relu	relu	NOUN
fcis-5209	60	71	activation	activation	NOUN
fcis-5209	60	72	function	function	VERB
fcis-5209	60	73	to	to	PART
fcis-5209	60	74	modify	modify	VERB
fcis-5209	60	75	them	they	PRON
fcis-5209	60	76	,	,	PUNCT
fcis-5209	60	77	and	and	CCONJ
fcis-5209	60	78	further	far	ADV
fcis-5209	60	79	compresses	compress	VERB
fcis-5209	60	80	the	the	DET
fcis-5209	60	81	features	feature	NOUN
fcis-5209	60	82	by	by	ADP
fcis-5209	60	83	using	use	VERB
fcis-5209	60	84	the	the	DET
fcis-5209	60	85	maximum	maximum	ADJ
fcis-5209	60	86	pooling	pooling	NOUN
fcis-5209	60	87	method	method	NOUN
fcis-5209	60	88	to	to	PART
fcis-5209	60	89	remove	remove	VERB
fcis-5209	60	90	redundant	redundant	ADJ
fcis-5209	60	91	information	information	NOUN
fcis-5209	60	92	.	.	PUNCT
fcis-5209	61	1	finally	finally	ADV
fcis-5209	61	2	,	,	PUNCT
fcis-5209	61	3	the	the	DET
fcis-5209	61	4	three	three	NUM
fcis-5209	61	5	convolution	convolution	NOUN
fcis-5209	61	6	features	feature	NOUN
fcis-5209	61	7	are	be	AUX
fcis-5209	61	8	spliced	splice	VERB
fcis-5209	61	9	to	to	PART
fcis-5209	61	10	obtain	obtain	VERB
fcis-5209	61	11	m	m	PRON
fcis-5209	61	12	,	,	PUNCT
fcis-5209	61	13	and	and	CCONJ
fcis-5209	61	14	the	the	DET
fcis-5209	61	15	calculation	calculation	NOUN
fcis-5209	61	16	formula	formula	NOUN
fcis-5209	61	17	is	be	AUX
fcis-5209	61	18	as	as	SCONJ
fcis-5209	61	19	follows	follow	VERB
fcis-5209	61	20	:	:	PUNCT
fcis-5209	61	21	1	1	NUM
fcis-5209	61	22	1	1	NUM
fcis-5209	61	23	2	2	NUM
fcis-5209	61	24	2	2	NUM
fcis-5209	61	25	3	3	NUM
fcis-5209	61	26	3	3	NUM
fcis-5209	61	27	1	1	NUM
fcis-5209	61	28	2	2	NUM
fcis-5209	61	29	3	3	NUM
fcis-5209	61	30	(	(	PUNCT
fcis-5209	61	31	(	(	PUNCT
fcis-5209	61	32	(	(	PUNCT
fcis-5209	61	33	,	,	PUNCT
fcis-5209	61	34	2	2	NUM
fcis-5209	61	35	*	*	NUM
fcis-5209	61	36	768	768	NUM
fcis-5209	61	37	)	)	PUNCT
fcis-5209	61	38	)	)	PUNCT
fcis-5209	61	39	)	)	PUNCT
fcis-5209	62	1	(	(	PUNCT
fcis-5209	62	2	(	(	PUNCT
fcis-5209	62	3	(	(	PUNCT
fcis-5209	62	4	,	,	PUNCT
fcis-5209	62	5	2	2	NUM
fcis-5209	62	6	*	*	NUM
fcis-5209	62	7	768	768	NUM
fcis-5209	62	8	)	)	PUNCT
fcis-5209	62	9	)	)	PUNCT
fcis-5209	62	10	)	)	PUNCT
fcis-5209	62	11	(	(	PUNCT
fcis-5209	62	12	(	(	PUNCT
fcis-5209	62	13	(	(	PUNCT
fcis-5209	62	14	,	,	PUNCT
fcis-5209	62	15	3	3	NUM
fcis-5209	62	16	*	*	NOUN
fcis-5209	62	17	768	768	NUM
fcis-5209	62	18	)	)	PUNCT
fcis-5209	62	19	)	)	PUNCT
fcis-5209	62	20	)	)	PUNCT
fcis-5209	62	21	(	(	PUNCT
fcis-5209	62	22	(	(	PUNCT
fcis-5209	62	23	(	(	PUNCT
fcis-5209	62	24	,	,	PUNCT
fcis-5209	62	25	3	3	NUM
fcis-5209	62	26	*	*	NOUN
fcis-5209	62	27	768	768	NUM
fcis-5209	62	28	)	)	PUNCT
fcis-5209	62	29	)	)	PUNCT
fcis-5209	62	30	)	)	PUNCT
fcis-5209	62	31	(	(	PUNCT
fcis-5209	62	32	(	(	PUNCT
fcis-5209	62	33	(	(	PUNCT
fcis-5209	62	34	,	,	PUNCT
fcis-5209	62	35	4	4	NUM
fcis-5209	62	36	*	*	NOUN
fcis-5209	62	37	768	768	NUM
fcis-5209	62	38	)	)	PUNCT
fcis-5209	62	39	)	)	PUNCT
fcis-5209	62	40	)	)	PUNCT
fcis-5209	62	41	(	(	PUNCT
fcis-5209	62	42	(	(	PUNCT
fcis-5209	62	43	(	(	PUNCT
fcis-5209	62	44	,	,	PUNCT
fcis-5209	62	45	4	4	NUM
fcis-5209	62	46	*	*	NOUN
fcis-5209	62	47	768	768	NUM
fcis-5209	62	48	)	)	PUNCT
fcis-5209	62	49	)	)	PUNCT
fcis-5209	62	50	)	)	PUNCT
fcis-5209	62	51	(	(	PUNCT
fcis-5209	62	52	,	,	PUNCT
fcis-5209	62	53	,	,	PUNCT
fcis-5209	62	54	)	)	PUNCT
fcis-5209	62	55	(	(	PUNCT
fcis-5209	62	56	e	e	X
fcis-5209	62	57	e	e	NOUN
fcis-5209	62	58	h	h	NOUN
fcis-5209	62	59	h	h	NOUN
fcis-5209	62	60	e	e	NOUN
fcis-5209	62	61	e	e	NOUN
fcis-5209	62	62	h	h	NOUN
fcis-5209	62	63	h	h	NOUN
fcis-5209	62	64	e	e	NOUN
fcis-5209	62	65	e	e	NOUN
fcis-5209	62	66	h	h	NOUN
fcis-5209	62	67	h	h	NOUN
fcis-5209	62	68	e	e	NOUN
fcis-5209	62	69	e	e	X
fcis-5209	62	70	e	e	X
fcis-5209	62	71	e	e	PROPN
fcis-5209	62	72	h	h	PROPN
fcis-5209	62	73	m	m	PROPN
fcis-5209	62	74	mp	mp	PROPN
fcis-5209	62	75	relu	relu	PROPN
fcis-5209	62	76	conv	conv	PROPN
fcis-5209	62	77	t	t	PROPN
fcis-5209	62	78	m	m	PROPN
fcis-5209	62	79	mp	mp	PROPN
fcis-5209	62	80	relu	relu	PROPN
fcis-5209	62	81	conv	conv	PROPN
fcis-5209	62	82	t	t	PROPN
fcis-5209	62	83	m	m	PROPN
fcis-5209	62	84	mp	mp	PROPN
fcis-5209	62	85	relu	relu	PROPN
fcis-5209	62	86	conv	conv	PROPN
fcis-5209	62	87	t	t	PROPN
fcis-5209	62	88	m	m	PROPN
fcis-5209	62	89	mp	mp	PROPN
fcis-5209	62	90	relu	relu	PROPN
fcis-5209	62	91	conv	conv	PROPN
fcis-5209	62	92	t	t	PROPN
fcis-5209	62	93	m	m	PROPN
fcis-5209	62	94	mp	mp	PROPN
fcis-5209	62	95	relu	relu	PROPN
fcis-5209	62	96	conv	conv	PROPN
fcis-5209	62	97	t	t	PROPN
fcis-5209	62	98	m	m	PROPN
fcis-5209	62	99	mp	mp	PROPN
fcis-5209	62	100	relu	relu	PROPN
fcis-5209	62	101	conv	conv	PROPN
fcis-5209	62	102	t	t	PROPN
fcis-5209	62	103	m	m	PROPN
fcis-5209	62	104	concat	concat	PROPN
fcis-5209	62	105	m	m	VERB
fcis-5209	62	106	m	m	VERB
fcis-5209	62	107	m	m	VERB
fcis-5209	62	108	m	m	NOUN
fcis-5209	62	109	concat	concat	NOUN
fcis-5209	62	110	=	=	PUNCT
fcis-5209	63	1	=	=	PUNCT
fcis-5209	63	2	=	=	PUNCT
fcis-5209	63	3	=	=	PUNCT
fcis-5209	63	4	=	=	PUNCT
fcis-5209	63	5	=	=	PUNCT
fcis-5209	63	6	=	=	SYM
fcis-5209	63	7	=	=	SYM
fcis-5209	63	8	1	1	NUM
fcis-5209	63	9	2	2	NUM
fcis-5209	63	10	3	3	NUM
fcis-5209	63	11	,	,	PUNCT
fcis-5209	63	12	,	,	PUNCT
fcis-5209	63	13	)	)	PUNCT
fcis-5209	63	14	(	(	PUNCT
fcis-5209	63	15	,	,	PUNCT
fcis-5209	63	16	)	)	PUNCT
fcis-5209	63	17	h	h	NOUN
fcis-5209	63	18	h	h	NOUN
fcis-5209	63	19	h	h	NOUN
fcis-5209	64	1	e	e	PROPN
fcis-5209	64	2	h	h	NOUN
fcis-5209	64	3	m	m	VERB
fcis-5209	64	4	m	m	VERB
fcis-5209	64	5	m	m	VERB
fcis-5209	64	6	m	m	NOUN
fcis-5209	64	7	concat	concat	NOUN
fcis-5209	64	8	m	m	VERB
fcis-5209	64	9	m=	m=	X
fcis-5209	64	10	(	(	PUNCT
fcis-5209	64	11	5	5	NUM
fcis-5209	64	12	)	)	PUNCT
fcis-5209	64	13	where	where	SCONJ
fcis-5209	64	14	conv	conv	NOUN
fcis-5209	64	15	represents	represent	VERB
fcis-5209	64	16	the	the	DET
fcis-5209	64	17	convolution	convolution	NOUN
fcis-5209	64	18	operation	operation	NOUN
fcis-5209	64	19	,	,	PUNCT
fcis-5209	64	20	and	and	CCONJ
fcis-5209	64	21	the	the	DET
fcis-5209	64	22	two	two	NUM
fcis-5209	64	23	parameters	parameter	NOUN
fcis-5209	64	24	represent	represent	VERB
fcis-5209	64	25	the	the	DET
fcis-5209	64	26	input	input	NOUN
fcis-5209	64	27	and	and	CCONJ
fcis-5209	64	28	convolution	convolution	NOUN
fcis-5209	64	29	kernel	kernel	NOUN
fcis-5209	64	30	size	size	NOUN
fcis-5209	64	31	respectively	respectively	ADV
fcis-5209	64	32	,	,	PUNCT
fcis-5209	64	33	relu	relu	NOUN
fcis-5209	64	34	represents	represent	VERB
fcis-5209	64	35	the	the	DET
fcis-5209	64	36	activation	activation	NOUN
fcis-5209	64	37	function	function	NOUN
fcis-5209	64	38	,	,	PUNCT
fcis-5209	64	39	mp	mp	PROPN
fcis-5209	64	40	represents	represent	VERB
fcis-5209	64	41	the	the	DET
fcis-5209	64	42	max	max	PROPN
fcis-5209	64	43	pooling	pooling	NOUN
fcis-5209	64	44	,	,	PUNCT
fcis-5209	64	45	and	and	CCONJ
fcis-5209	64	46	concat	concat	NOUN
fcis-5209	64	47	represents	represent	VERB
fcis-5209	64	48	the	the	DET
fcis-5209	64	49	splicing	splicing	NOUN
fcis-5209	64	50	operation	operation	NOUN
fcis-5209	64	51	.	.	PUNCT
fcis-5209	65	1	(	(	PUNCT
fcis-5209	65	2	4	4	X
fcis-5209	65	3	)	)	PUNCT
fcis-5209	65	4	classification	classification	NOUN
fcis-5209	65	5	layer	layer	NOUN
fcis-5209	65	6	the	the	DET
fcis-5209	65	7	classification	classification	NOUN
fcis-5209	65	8	layer	layer	NOUN
fcis-5209	65	9	classifies	classify	VERB
fcis-5209	65	10	the	the	DET
fcis-5209	65	11	local	local	ADJ
fcis-5209	65	12	semantic	semantic	ADJ
fcis-5209	65	13	information	information	NOUN
fcis-5209	65	14	m	m	AUX
fcis-5209	65	15	obtained	obtain	VERB
fcis-5209	65	16	from	from	ADP
fcis-5209	65	17	the	the	DET
fcis-5209	65	18	conv	conv	ADJ
fcis-5209	65	19	block	block	NOUN
fcis-5209	65	20	on	on	ADP
fcis-5209	65	21	the	the	DET
fcis-5209	65	22	upper	upper	ADJ
fcis-5209	65	23	layer	layer	NOUN
fcis-5209	65	24	through	through	ADP
fcis-5209	65	25	a	a	DET
fcis-5209	65	26	layer	layer	NOUN
fcis-5209	65	27	of	of	ADP
fcis-5209	65	28	fully	fully	ADV
fcis-5209	65	29	connected	connect	VERB
fcis-5209	65	30	neural	neural	ADJ
fcis-5209	65	31	network	network	NOUN
fcis-5209	65	32	,	,	PUNCT
fcis-5209	65	33	and	and	CCONJ
fcis-5209	65	34	then	then	ADV
fcis-5209	65	35	gets	get	VERB
fcis-5209	65	36	the	the	DET
fcis-5209	65	37	matching	matching	NOUN
fcis-5209	65	38	or	or	CCONJ
fcis-5209	65	39	mismatching	mismatch	VERB
fcis-5209	65	40	results	result	NOUN
fcis-5209	65	41	through	through	ADP
fcis-5209	65	42	softmax	softmax	NOUN
fcis-5209	65	43	function	function	NOUN
fcis-5209	65	44	.	.	PUNCT
fcis-5209	66	1	the	the	DET
fcis-5209	66	2	calculation	calculation	NOUN
fcis-5209	66	3	formula	formula	NOUN
fcis-5209	66	4	is	be	AUX
fcis-5209	66	5	as	as	SCONJ
fcis-5209	66	6	follows	follow	VERB
fcis-5209	66	7	:	:	PUNCT
fcis-5209	66	8	max	max	PROPN
fcis-5209	66	9	(	(	PUNCT
fcis-5209	66	10	(	(	PUNCT
fcis-5209	66	11	)	)	PUNCT
fcis-5209	66	12	)	)	PUNCT
fcis-5209	66	13	y	y	PROPN
fcis-5209	66	14	soft	soft	ADJ
fcis-5209	66	15	linear	linear	NOUN
fcis-5209	66	16	m=	m=	X
fcis-5209	66	17	(	(	PUNCT
fcis-5209	66	18	6	6	NUM
fcis-5209	66	19	)	)	SYM
fcis-5209	66	20	3	3	NUM
fcis-5209	66	21	.	.	X
fcis-5209	66	22	experiment	experiment	NOUN
fcis-5209	66	23	in	in	ADP
fcis-5209	66	24	this	this	DET
fcis-5209	66	25	paper	paper	NOUN
fcis-5209	66	26	,	,	PUNCT
fcis-5209	66	27	the	the	DET
fcis-5209	66	28	proposed	propose	VERB
fcis-5209	66	29	model	model	NOUN
fcis-5209	66	30	is	be	AUX
fcis-5209	66	31	tested	test	VERB
fcis-5209	66	32	in	in	ADP
fcis-5209	66	33	the	the	DET
fcis-5209	66	34	text	text	NOUN
fcis-5209	66	35	lstm	lstm	NOUN
fcis-5209	66	36	lstm	lstm	PROPN
fcis-5209	66	37	1et	1et	ADJ
fcis-5209	66	38	42	42	NUM
fcis-5209	66	39	matching	matching	NOUN
fcis-5209	66	40	dataset	dataset	ADJ
fcis-5209	66	41	lcqmc	lcqmc	NOUN
fcis-5209	66	42	,	,	PUNCT
fcis-5209	66	43	and	and	CCONJ
fcis-5209	66	44	compared	compare	VERB
fcis-5209	66	45	with	with	ADP
fcis-5209	66	46	other	other	ADJ
fcis-5209	66	47	model	model	NOUN
fcis-5209	66	48	methods	method	NOUN
fcis-5209	66	49	.	.	PUNCT
fcis-5209	67	1	acc	acc	PROPN
fcis-5209	67	2	and	and	CCONJ
fcis-5209	67	3	f1	f1	NOUN
fcis-5209	67	4	-	-	PUNCT
fcis-5209	67	5	score	score	NOUN
fcis-5209	67	6	are	be	AUX
fcis-5209	67	7	used	use	VERB
fcis-5209	67	8	to	to	PART
fcis-5209	67	9	evaluate	evaluate	VERB
fcis-5209	67	10	the	the	DET
fcis-5209	67	11	effectiveness	effectiveness	NOUN
fcis-5209	67	12	of	of	ADP
fcis-5209	67	13	the	the	DET
fcis-5209	67	14	model	model	NOUN
fcis-5209	67	15	.	.	PUNCT
fcis-5209	68	1	3.1	3.1	NUM
fcis-5209	68	2	.	.	PUNCT
fcis-5209	68	3	dataset	dataset	ADJ
fcis-5209	68	4	lcqmc	lcqmc	PROPN
fcis-5209	68	5	is	be	AUX
fcis-5209	68	6	a	a	DET
fcis-5209	68	7	large	large	ADJ
fcis-5209	68	8	-	-	PUNCT
fcis-5209	68	9	scale	scale	NOUN
fcis-5209	68	10	open	open	ADJ
fcis-5209	68	11	corpus	corpus	NOUN
fcis-5209	68	12	for	for	ADP
fcis-5209	68	13	text	text	NOUN
fcis-5209	68	14	matching	matching	NOUN
fcis-5209	68	15	.	.	PUNCT
fcis-5209	69	1	the	the	DET
fcis-5209	69	2	dataset	dataset	NOUN
fcis-5209	69	3	contains	contain	VERB
fcis-5209	69	4	260068	260068	NUM
fcis-5209	69	5	pieces	piece	NOUN
fcis-5209	69	6	of	of	ADP
fcis-5209	69	7	data	datum	NOUN
fcis-5209	69	8	in	in	ADP
fcis-5209	69	9	total	total	NOUN
fcis-5209	69	10	,	,	PUNCT
fcis-5209	69	11	including	include	VERB
fcis-5209	69	12	238766	238766	NUM
fcis-5209	69	13	training	training	NOUN
fcis-5209	69	14	sets	set	NOUN
fcis-5209	69	15	,	,	PUNCT
fcis-5209	69	16	8802	8802	NUM
fcis-5209	69	17	validation	validation	NOUN
fcis-5209	69	18	sets	set	NOUN
fcis-5209	69	19	,	,	PUNCT
fcis-5209	69	20	and	and	CCONJ
fcis-5209	69	21	12500	12500	NUM
fcis-5209	69	22	test	test	NOUN
fcis-5209	69	23	sets	set	NOUN
fcis-5209	69	24	.	.	PUNCT
fcis-5209	70	1	each	each	DET
fcis-5209	70	2	data	datum	NOUN
fcis-5209	70	3	is	be	AUX
fcis-5209	70	4	composed	compose	VERB
fcis-5209	70	5	of	of	ADP
fcis-5209	70	6	two	two	NUM
fcis-5209	70	7	sentences	sentence	NOUN
fcis-5209	70	8	and	and	CCONJ
fcis-5209	70	9	a	a	DET
fcis-5209	70	10	label	label	NOUN
fcis-5209	70	11	.	.	PUNCT
fcis-5209	71	1	the	the	DET
fcis-5209	71	2	label	label	NOUN
fcis-5209	71	3	is	be	AUX
fcis-5209	71	4	divided	divide	VERB
fcis-5209	71	5	into	into	ADP
fcis-5209	71	6	0	0	NUM
fcis-5209	71	7	and	and	CCONJ
fcis-5209	71	8	1	1	NUM
fcis-5209	71	9	,	,	PUNCT
fcis-5209	71	10	where	where	SCONJ
fcis-5209	71	11	0	0	NUM
fcis-5209	71	12	represents	represent	VERB
fcis-5209	71	13	the	the	DET
fcis-5209	71	14	semantic	semantic	ADJ
fcis-5209	71	15	mismatch	mismatch	NOUN
fcis-5209	71	16	of	of	ADP
fcis-5209	71	17	two	two	NUM
fcis-5209	71	18	sentences	sentence	NOUN
fcis-5209	71	19	,	,	PUNCT
fcis-5209	71	20	and	and	CCONJ
fcis-5209	71	21	1	1	NUM
fcis-5209	71	22	represents	represent	VERB
fcis-5209	71	23	the	the	DET
fcis-5209	71	24	semantic	semantic	ADJ
fcis-5209	71	25	match	match	NOUN
fcis-5209	71	26	of	of	ADP
fcis-5209	71	27	two	two	NUM
fcis-5209	71	28	sentences	sentence	NOUN
fcis-5209	71	29	.	.	PUNCT
fcis-5209	72	1	data	datum	NOUN
fcis-5209	72	2	examples	example	NOUN
fcis-5209	72	3	are	be	AUX
fcis-5209	72	4	shown	show	VERB
fcis-5209	72	5	in	in	ADP
fcis-5209	72	6	the	the	DET
fcis-5209	72	7	following	following	NOUN
fcis-5209	72	8	,	,	PUNCT
fcis-5209	72	9	see	see	VERB
fcis-5209	72	10	table	table	NOUN
fcis-5209	72	11	1	1	NUM
fcis-5209	72	12	.	.	PUNCT
fcis-5209	72	13	table	table	NOUN
fcis-5209	72	14	1	1	NUM
fcis-5209	72	15	.	.	PUNCT
fcis-5209	72	16	example	example	NOUN
fcis-5209	72	17	of	of	ADP
fcis-5209	72	18	lcqmc	lcqmc	PROPN
fcis-5209	72	19	sentence1	sentence1	PROPN
fcis-5209	72	20	sentence2	sentence2	PROPN
fcis-5209	72	21	labe	labe	PROPN
fcis-5209	72	22	l	l	PROPN
fcis-5209	72	23	这腰带是什么牌子	这腰带是什么牌子	NOUN
fcis-5209	72	24	护腰带是什么牌子	护腰带是什么牌子	NOUN
fcis-5209	72	25	0	0	NUM
fcis-5209	72	26	货到付款的网站是哪	货到付款的网站是哪	PROPN
fcis-5209	72	27	个	个	PROPN
fcis-5209	72	28	什么购物网站是货到付款	什么购物网站是货到付款	NOUN
fcis-5209	72	29	的	的	ADP
fcis-5209	72	30	1	1	NUM
fcis-5209	72	31	3.2	3.2	NUM
fcis-5209	72	32	.	.	PUNCT
fcis-5209	73	1	experimental	experimental	ADJ
fcis-5209	73	2	environment	environment	NOUN
fcis-5209	73	3	the	the	DET
fcis-5209	73	4	training	training	NOUN
fcis-5209	73	5	,	,	PUNCT
fcis-5209	73	6	validating	validate	VERB
fcis-5209	73	7	and	and	CCONJ
fcis-5209	73	8	testing	testing	NOUN
fcis-5209	73	9	of	of	ADP
fcis-5209	73	10	all	all	DET
fcis-5209	73	11	models	model	NOUN
fcis-5209	73	12	in	in	ADP
fcis-5209	73	13	this	this	DET
fcis-5209	73	14	paper	paper	NOUN
fcis-5209	73	15	are	be	AUX
fcis-5209	73	16	based	base	VERB
fcis-5209	73	17	on	on	ADP
fcis-5209	73	18	the	the	DET
fcis-5209	73	19	deep	deep	ADJ
fcis-5209	73	20	learning	learning	NOUN
fcis-5209	73	21	framework	framework	NOUN
fcis-5209	73	22	pytorch	pytorch	NOUN
fcis-5209	73	23	.	.	PUNCT
fcis-5209	74	1	the	the	DET
fcis-5209	74	2	specific	specific	ADJ
fcis-5209	74	3	experimental	experimental	ADJ
fcis-5209	74	4	environment	environment	NOUN
fcis-5209	74	5	configuration	configuration	NOUN
fcis-5209	74	6	is	be	AUX
fcis-5209	74	7	shown	show	VERB
fcis-5209	74	8	in	in	ADP
fcis-5209	74	9	the	the	DET
fcis-5209	74	10	following	following	NOUN
fcis-5209	74	11	,	,	PUNCT
fcis-5209	74	12	see	see	VERB
fcis-5209	74	13	table	table	NOUN
fcis-5209	74	14	2	2	NUM
fcis-5209	74	15	.	.	PUNCT
fcis-5209	74	16	table	table	NOUN
fcis-5209	74	17	2	2	NUM
fcis-5209	74	18	.	.	PUNCT
fcis-5209	74	19	experimental	experimental	ADJ
fcis-5209	74	20	environment	environment	NOUN
fcis-5209	74	21	configuration	configuration	NOUN
fcis-5209	74	22	software	software	NOUN
fcis-5209	74	23	and	and	CCONJ
fcis-5209	74	24	hardware	hardware	NOUN
fcis-5209	74	25	configuration	configuration	NOUN
fcis-5209	74	26	operating	operating	NOUN
fcis-5209	74	27	system	system	NOUN
fcis-5209	74	28	ubuntu	ubuntu	ADJ
fcis-5209	74	29	20.04.4	20.04.4	NUM
fcis-5209	74	30	lts	lts	NOUN
fcis-5209	74	31	development	development	NOUN
fcis-5209	74	32	tools	tool	NOUN
fcis-5209	74	33	and	and	CCONJ
fcis-5209	74	34	languages	language	NOUN
fcis-5209	74	35	pycharm	pycharm	VERB
fcis-5209	74	36	and	and	CCONJ
fcis-5209	74	37	python	python	NOUN
fcis-5209	74	38	deep	deep	ADJ
fcis-5209	74	39	learning	learning	NOUN
fcis-5209	74	40	framework	framework	NOUN
fcis-5209	74	41	pytorch	pytorch	NOUN
fcis-5209	74	42	1.11.0	1.11.0	NUM
fcis-5209	74	43	gpu	gpu	PROPN
fcis-5209	74	44	nvidia	nvidia	PROPN
fcis-5209	74	45	geforce	geforce	NOUN
fcis-5209	74	46	rtx	rtx	PROPN
fcis-5209	74	47	3090	3090	NUM
fcis-5209	75	1	24	24	NUM
fcis-5209	75	2	g	g	PROPN
fcis-5209	75	3	cpu	cpu	NOUN
fcis-5209	75	4	intel	intel	PROPN
fcis-5209	75	5	®	®	NOUN
fcis-5209	75	6	core	core	NOUN
fcis-5209	75	7	™	™	VERB
fcis-5209	75	8	i9	i9	NOUN
fcis-5209	75	9	-	-	PUNCT
fcis-5209	75	10	9900k	9900k	NOUN
fcis-5209	75	11	cpu	cpu	NOUN
fcis-5209	75	12	@	@	ADP
fcis-5209	75	13	3.60ghz	3.60ghz	NUM
fcis-5209	75	14	×	×	NOUN
fcis-5209	75	15	16	16	NUM
fcis-5209	75	16	memory	memory	NOUN
fcis-5209	75	17	64	64	NUM
fcis-5209	75	18	g	g	NOUN
fcis-5209	75	19	3.3	3.3	NUM
fcis-5209	75	20	.	.	PUNCT
fcis-5209	76	1	experimental	experimental	ADJ
fcis-5209	76	2	parameters	parameter	NOUN
fcis-5209	76	3	the	the	DET
fcis-5209	76	4	bert	bert	PROPN
fcis-5209	76	5	model	model	NOUN
fcis-5209	76	6	selected	select	VERB
fcis-5209	76	7	in	in	ADP
fcis-5209	76	8	this	this	DET
fcis-5209	76	9	paper	paper	NOUN
fcis-5209	76	10	is	be	AUX
fcis-5209	76	11	the	the	DET
fcis-5209	76	12	chinesewwm	chinesewwm	NOUN
fcis-5209	76	13	-	-	PUNCT
fcis-5209	76	14	ext	ext	PROPN
fcis-5209	76	15	model	model	NOUN
fcis-5209	76	16	pretrained	pretraine	VERB
fcis-5209	76	17	by	by	ADP
fcis-5209	76	18	cui	cui	PROPN
fcis-5209	76	19	et	et	PROPN
fcis-5209	76	20	al	al	PROPN
fcis-5209	76	21	.	.	PROPN
fcis-5209	77	1	on	on	ADP
fcis-5209	77	2	a	a	DET
fcis-5209	77	3	large	large	ADJ
fcis-5209	77	4	chinese	chinese	ADJ
fcis-5209	77	5	corpus	corpus	NOUN
fcis-5209	77	6	.	.	PUNCT
fcis-5209	78	1	the	the	DET
fcis-5209	78	2	max	max	PROPN
fcis-5209	78	3	length	length	NOUN
fcis-5209	78	4	of	of	ADP
fcis-5209	78	5	input	input	NOUN
fcis-5209	78	6	data	datum	NOUN
fcis-5209	78	7	is	be	AUX
fcis-5209	78	8	128	128	NUM
fcis-5209	78	9	and	and	CCONJ
fcis-5209	78	10	the	the	DET
fcis-5209	78	11	hidden	hidden	ADJ
fcis-5209	78	12	size	size	NOUN
fcis-5209	78	13	of	of	ADP
fcis-5209	78	14	bert	bert	PROPN
fcis-5209	78	15	is	be	AUX
fcis-5209	78	16	768	768	NUM
fcis-5209	78	17	,	,	PUNCT
fcis-5209	78	18	the	the	DET
fcis-5209	78	19	hidden	hidden	ADJ
fcis-5209	78	20	size	size	NOUN
fcis-5209	78	21	of	of	ADP
fcis-5209	78	22	bilstm	bilstm	NOUN
fcis-5209	78	23	is	be	AUX
fcis-5209	78	24	334	334	NUM
fcis-5209	78	25	,	,	PUNCT
fcis-5209	78	26	the	the	DET
fcis-5209	78	27	convolution	convolution	NOUN
fcis-5209	78	28	kernel	kernel	NOUN
fcis-5209	78	29	of	of	ADP
fcis-5209	78	30	multi	multi	ADJ
fcis-5209	78	31	-	-	ADJ
fcis-5209	78	32	scale	scale	ADJ
fcis-5209	78	33	convolution	convolution	NOUN
fcis-5209	78	34	is	be	AUX
fcis-5209	78	35	2	2	NUM
fcis-5209	78	36	*	*	SYM
fcis-5209	78	37	768	768	NUM
fcis-5209	78	38	,	,	PUNCT
fcis-5209	78	39	3	3	NUM
fcis-5209	78	40	*	*	SYM
fcis-5209	78	41	768	768	NUM
fcis-5209	78	42	and	and	CCONJ
fcis-5209	78	43	4	4	NUM
fcis-5209	78	44	*	*	NOUN
fcis-5209	78	45	768	768	NUM
fcis-5209	78	46	respectively	respectively	ADV
fcis-5209	78	47	,	,	PUNCT
fcis-5209	78	48	the	the	DET
fcis-5209	78	49	random	random	ADJ
fcis-5209	78	50	seed	seed	NOUN
fcis-5209	78	51	is	be	AUX
fcis-5209	78	52	47	47	NUM
fcis-5209	78	53	,	,	PUNCT
fcis-5209	78	54	batch	batch	NOUN
fcis-5209	78	55	size	size	NOUN
fcis-5209	78	56	is	be	AUX
fcis-5209	78	57	64	64	NUM
fcis-5209	78	58	,	,	PUNCT
fcis-5209	78	59	learning	learn	VERB
fcis-5209	78	60	rate	rate	NOUN
fcis-5209	78	61	is	be	AUX
fcis-5209	78	62	2e-5	2e-5	PROPN
fcis-5209	78	63	,	,	PUNCT
fcis-5209	78	64	dropout	dropout	NOUN
fcis-5209	78	65	is	be	AUX
fcis-5209	78	66	0.2	0.2	NUM
fcis-5209	78	67	and	and	CCONJ
fcis-5209	78	68	the	the	DET
fcis-5209	78	69	epoch	epoch	NOUN
fcis-5209	78	70	is	be	AUX
fcis-5209	78	71	3	3	NUM
fcis-5209	78	72	.	.	ADP
fcis-5209	78	73	3.4	3.4	NUM
fcis-5209	78	74	.	.	PUNCT
fcis-5209	79	1	evaluation	evaluation	NOUN
fcis-5209	79	2	criteria	criterion	NOUN
fcis-5209	79	3	in	in	ADP
fcis-5209	79	4	order	order	NOUN
fcis-5209	79	5	to	to	PART
fcis-5209	79	6	verify	verify	VERB
fcis-5209	79	7	the	the	DET
fcis-5209	79	8	effectiveness	effectiveness	NOUN
fcis-5209	79	9	of	of	ADP
fcis-5209	79	10	the	the	DET
fcis-5209	79	11	proposed	propose	VERB
fcis-5209	79	12	model	model	NOUN
fcis-5209	79	13	,	,	PUNCT
fcis-5209	79	14	the	the	DET
fcis-5209	79	15	accuracy	accuracy	NOUN
fcis-5209	79	16	acc	acc	NOUN
fcis-5209	79	17	and	and	CCONJ
fcis-5209	79	18	f1	f1	NOUN
fcis-5209	79	19	-	-	PUNCT
fcis-5209	79	20	score	score	NOUN
fcis-5209	79	21	are	be	AUX
fcis-5209	79	22	used	use	VERB
fcis-5209	79	23	to	to	PART
fcis-5209	79	24	evaluate	evaluate	VERB
fcis-5209	79	25	the	the	DET
fcis-5209	79	26	model	model	NOUN
fcis-5209	79	27	.	.	PUNCT
fcis-5209	80	1	the	the	DET
fcis-5209	80	2	calculation	calculation	NOUN
fcis-5209	80	3	formula	formula	NOUN
fcis-5209	80	4	of	of	ADP
fcis-5209	80	5	acc	acc	PROPN
fcis-5209	80	6	and	and	CCONJ
fcis-5209	80	7	f1	f1	NOUN
fcis-5209	80	8	-	-	PUNCT
fcis-5209	80	9	score	score	NOUN
fcis-5209	80	10	is	be	AUX
fcis-5209	80	11	as	as	SCONJ
fcis-5209	80	12	follows	follow	VERB
fcis-5209	80	13	:	:	PUNCT
fcis-5209	80	14	(	(	PUNCT
fcis-5209	80	15	)	)	PUNCT
fcis-5209	80	16	(	(	PUNCT
fcis-5209	80	17	)	)	PUNCT
fcis-5209	81	1	p	p	NOUN
fcis-5209	81	2	n	n	NUM
fcis-5209	81	3	p	p	NOUN
fcis-5209	81	4	n	n	PROPN
fcis-5209	81	5	p	p	NOUN
fcis-5209	81	6	n	n	ADP
fcis-5209	81	7	t	t	PROPN
fcis-5209	81	8	t	t	PROPN
fcis-5209	81	9	acc	acc	PROPN
fcis-5209	82	1	t	t	PROPN
fcis-5209	82	2	t	t	PROPN
fcis-5209	82	3	f	f	PROPN
fcis-5209	82	4	f	f	PROPN
fcis-5209	83	1	+	+	CCONJ
fcis-5209	83	2	=	=	PUNCT
fcis-5209	84	1	+	+	PUNCT
fcis-5209	84	2	+	+	PUNCT
fcis-5209	84	3	+	+	CCONJ
fcis-5209	84	4	(	(	PUNCT
fcis-5209	84	5	7	7	X
fcis-5209	84	6	)	)	PUNCT
fcis-5209	84	7	𝑃	𝑃	NOUN
fcis-5209	84	8	=	=	PUNCT
fcis-5209	84	9	𝑇𝑝	𝑇𝑝	PROPN
fcis-5209	84	10	𝑇𝑝+𝐹𝑝	𝑇𝑝+𝐹𝑝	NOUN
fcis-5209	84	11	,	,	PUNCT
fcis-5209	84	12	𝑅	𝑅	PROPN
fcis-5209	84	13	=	=	SYM
fcis-5209	84	14	𝑇𝑝	𝑇𝑝	PROPN
fcis-5209	84	15	𝑇𝑝+𝐹𝑁	𝑇𝑝+𝐹𝑁	PROPN
fcis-5209	84	16	,	,	PUNCT
fcis-5209	84	17	𝐹1	𝐹1	PROPN
fcis-5209	84	18	−	−	PROPN
fcis-5209	84	19	𝑠𝑐𝑜𝑟𝑒	𝑠𝑐𝑜𝑟𝑒	NOUN
fcis-5209	84	20	=	=	SYM
fcis-5209	84	21	2*𝑃*𝑅	2*𝑃*𝑅	PROPN
fcis-5209	84	22	𝑃+𝑅	𝑃+𝑅	NOUN
fcis-5209	84	23	(	(	PUNCT
fcis-5209	84	24	8)	8)	NUM
fcis-5209	84	25	tp	tp	NOUN
fcis-5209	84	26	represents	represent	VERB
fcis-5209	84	27	the	the	DET
fcis-5209	84	28	positive	positive	ADJ
fcis-5209	84	29	sample	sample	NOUN
fcis-5209	84	30	predicted	predict	VERB
fcis-5209	84	31	by	by	ADP
fcis-5209	84	32	the	the	DET
fcis-5209	84	33	model	model	NOUN
fcis-5209	84	34	as	as	ADP
fcis-5209	84	35	a	a	DET
fcis-5209	84	36	positive	positive	ADJ
fcis-5209	84	37	class	class	NOUN
fcis-5209	84	38	,	,	PUNCT
fcis-5209	84	39	tn	tn	PROPN
fcis-5209	84	40	represents	represent	VERB
fcis-5209	84	41	the	the	DET
fcis-5209	84	42	negative	negative	ADJ
fcis-5209	84	43	sample	sample	NOUN
fcis-5209	84	44	predicted	predict	VERB
fcis-5209	84	45	by	by	ADP
fcis-5209	84	46	the	the	DET
fcis-5209	84	47	model	model	NOUN
fcis-5209	84	48	as	as	ADP
fcis-5209	84	49	a	a	DET
fcis-5209	84	50	negative	negative	ADJ
fcis-5209	84	51	class	class	NOUN
fcis-5209	84	52	,	,	PUNCT
fcis-5209	84	53	fp	fp	X
fcis-5209	84	54	represents	represent	VERB
fcis-5209	84	55	the	the	DET
fcis-5209	84	56	negative	negative	ADJ
fcis-5209	84	57	sample	sample	NOUN
fcis-5209	84	58	predicted	predict	VERB
fcis-5209	84	59	by	by	ADP
fcis-5209	84	60	the	the	DET
fcis-5209	84	61	model	model	NOUN
fcis-5209	84	62	as	as	ADP
fcis-5209	84	63	a	a	DET
fcis-5209	84	64	positive	positive	ADJ
fcis-5209	84	65	class	class	NOUN
fcis-5209	84	66	,	,	PUNCT
fcis-5209	84	67	and	and	CCONJ
fcis-5209	84	68	fn	fn	NOUN
fcis-5209	84	69	represents	represent	VERB
fcis-5209	84	70	the	the	DET
fcis-5209	84	71	positive	positive	ADJ
fcis-5209	84	72	sample	sample	NOUN
fcis-5209	84	73	predicted	predict	VERB
fcis-5209	84	74	by	by	ADP
fcis-5209	84	75	the	the	DET
fcis-5209	84	76	model	model	NOUN
fcis-5209	84	77	as	as	ADP
fcis-5209	84	78	a	a	DET
fcis-5209	84	79	negative	negative	ADJ
fcis-5209	84	80	class	class	NOUN
fcis-5209	84	81	.	.	PUNCT
fcis-5209	85	1	3.5	3.5	NUM
fcis-5209	85	2	.	.	PUNCT
fcis-5209	85	3	evaluation	evaluation	NOUN
fcis-5209	85	4	criteria	criterion	NOUN
fcis-5209	85	5	in	in	ADP
fcis-5209	85	6	this	this	DET
fcis-5209	85	7	paper	paper	NOUN
fcis-5209	85	8	,	,	PUNCT
fcis-5209	85	9	all	all	DET
fcis-5209	85	10	experiments	experiment	NOUN
fcis-5209	85	11	are	be	AUX
fcis-5209	85	12	carried	carry	VERB
fcis-5209	85	13	out	out	ADP
fcis-5209	85	14	under	under	ADP
fcis-5209	85	15	the	the	DET
fcis-5209	85	16	same	same	ADJ
fcis-5209	85	17	experimental	experimental	ADJ
fcis-5209	85	18	environment	environment	NOUN
fcis-5209	85	19	as	as	ADV
fcis-5209	85	20	far	far	ADV
fcis-5209	85	21	as	as	ADP
fcis-5209	85	22	possible	possible	ADJ
fcis-5209	85	23	.	.	PUNCT
fcis-5209	86	1	first	first	ADV
fcis-5209	86	2	,	,	PUNCT
fcis-5209	86	3	the	the	DET
fcis-5209	86	4	baseline	baseline	PROPN
fcis-5209	86	5	model	model	PROPN
fcis-5209	86	6	bert	bert	PROPN
fcis-5209	86	7	is	be	AUX
fcis-5209	86	8	implemented	implement	VERB
fcis-5209	86	9	,	,	PUNCT
fcis-5209	86	10	and	and	CCONJ
fcis-5209	86	11	the	the	DET
fcis-5209	86	12	output	output	NOUN
fcis-5209	86	13	of	of	ADP
fcis-5209	86	14	its	its	PRON
fcis-5209	86	15	different	different	ADJ
fcis-5209	86	16	layers	layer	NOUN
fcis-5209	86	17	are	be	AUX
fcis-5209	86	18	used	use	VERB
fcis-5209	86	19	as	as	ADP
fcis-5209	86	20	the	the	DET
fcis-5209	86	21	semantic	semantic	ADJ
fcis-5209	86	22	features	feature	NOUN
fcis-5209	86	23	of	of	ADP
fcis-5209	86	24	the	the	DET
fcis-5209	86	25	text	text	NOUN
fcis-5209	86	26	,	,	PUNCT
fcis-5209	86	27	and	and	CCONJ
fcis-5209	86	28	experiment	experiment	NOUN
fcis-5209	86	29	on	on	ADP
fcis-5209	86	30	the	the	DET
fcis-5209	86	31	lcqmc	lcqmc	PROPN
fcis-5209	86	32	dataset	dataset	NOUN
fcis-5209	86	33	.	.	PUNCT
fcis-5209	87	1	then	then	ADV
fcis-5209	87	2	,	,	PUNCT
fcis-5209	87	3	the	the	DET
fcis-5209	87	4	bert	bert	PROPN
fcis-5209	87	5	model	model	NOUN
fcis-5209	87	6	is	be	AUX
fcis-5209	87	7	spliced	splice	VERB
fcis-5209	87	8	into	into	ADP
fcis-5209	87	9	bilstm	bilstm	NOUN
fcis-5209	87	10	and	and	CCONJ
fcis-5209	87	11	conv	conv	ADJ
fcis-5209	87	12	block	block	NOUN
fcis-5209	87	13	respectively	respectively	ADV
fcis-5209	87	14	to	to	PART
fcis-5209	87	15	process	process	VERB
fcis-5209	87	16	the	the	DET
fcis-5209	87	17	features	feature	NOUN
fcis-5209	87	18	.	.	PUNCT
fcis-5209	88	1	finally	finally	ADV
fcis-5209	88	2	,	,	PUNCT
fcis-5209	88	3	our	our	PRON
fcis-5209	88	4	method	method	NOUN
fcis-5209	88	5	is	be	AUX
fcis-5209	88	6	tested	test	VERB
fcis-5209	88	7	.	.	PUNCT
fcis-5209	89	1	the	the	DET
fcis-5209	89	2	experimental	experimental	ADJ
fcis-5209	89	3	results	result	NOUN
fcis-5209	89	4	are	be	AUX
fcis-5209	89	5	shown	show	VERB
fcis-5209	89	6	in	in	ADP
fcis-5209	89	7	following	follow	VERB
fcis-5209	89	8	,	,	PUNCT
fcis-5209	89	9	see	see	VERB
fcis-5209	89	10	table3	table3	PROPN
fcis-5209	89	11	.	.	PUNCT
fcis-5209	90	1	table	table	NOUN
fcis-5209	90	2	3	3	NUM
fcis-5209	90	3	.	.	PUNCT
fcis-5209	90	4	experiment	experiment	NOUN
fcis-5209	90	5	result	result	PROPN
fcis-5209	90	6	model	model	PROPN
fcis-5209	90	7	acc	acc	PROPN
fcis-5209	90	8	(	(	PUNCT
fcis-5209	90	9	%	%	INTJ
fcis-5209	90	10	)	)	PUNCT
fcis-5209	90	11	f1	f1	NOUN
fcis-5209	90	12	-	-	PUNCT
fcis-5209	90	13	score	score	NOUN
fcis-5209	90	14	(	(	PUNCT
fcis-5209	90	15	%	%	INTJ
fcis-5209	90	16	)	)	PUNCT
fcis-5209	90	17	bert	bert	PROPN
fcis-5209	90	18	-	-	PUNCT
fcis-5209	90	19	cls	cls	NOUN
fcis-5209	90	20	86.84	86.84	NUM
fcis-5209	90	21	87.82	87.82	NUM
fcis-5209	90	22	bert	bert	NOUN
fcis-5209	90	23	-	-	PUNCT
fcis-5209	90	24	pooler	pooler	PROPN
fcis-5209	90	25	86.83	86.83	NUM
fcis-5209	90	26	87.76	87.76	NUM
fcis-5209	90	27	bert	bert	NOUN
fcis-5209	90	28	-	-	PUNCT
fcis-5209	90	29	last	last	ADJ
fcis-5209	90	30	-	-	PUNCT
fcis-5209	90	31	avg	avg	NOUN
fcis-5209	90	32	86.59	86.59	NUM
fcis-5209	90	33	87.65	87.65	NUM
fcis-5209	90	34	bert	bert	NOUN
fcis-5209	90	35	-	-	PUNCT
fcis-5209	90	36	first	first	ADJ
fcis-5209	90	37	-	-	PUNCT
fcis-5209	90	38	last	last	ADJ
fcis-5209	90	39	-	-	PUNCT
fcis-5209	90	40	avg	avg	NOUN
fcis-5209	90	41	87.87	87.87	NUM
fcis-5209	90	42	88.54	88.54	NUM
fcis-5209	90	43	bert	bert	NOUN
fcis-5209	90	44	-	-	PUNCT
fcis-5209	90	45	last	last	ADJ
fcis-5209	90	46	-	-	PUNCT
fcis-5209	90	47	four	four	NUM
fcis-5209	90	48	-	-	PUNCT
fcis-5209	90	49	avg	avg	NOUN
fcis-5209	90	50	86.47	86.47	NUM
fcis-5209	90	51	87.50	87.50	NUM
fcis-5209	90	52	bert	bert	NOUN
fcis-5209	91	1	+	+	NOUN
fcis-5209	91	2	bilstm	bilstm	NOUN
fcis-5209	91	3	86.56	86.56	NUM
fcis-5209	91	4	87.63	87.63	NUM
fcis-5209	91	5	bert	bert	PROPN
fcis-5209	91	6	+	+	CCONJ
fcis-5209	91	7	conv	conv	ADJ
fcis-5209	91	8	block	block	NOUN
fcis-5209	91	9	88.17	88.17	NUM
fcis-5209	91	10	88.66	88.66	NUM
fcis-5209	91	11	bert	bert	NOUN
fcis-5209	91	12	+	+	CCONJ
fcis-5209	91	13	bilstm	bilstm	NOUN
fcis-5209	91	14	+	+	CCONJ
fcis-5209	91	15	conv	conv	ADJ
fcis-5209	91	16	block	block	NOUN
fcis-5209	91	17	87.87	87.87	NUM
fcis-5209	91	18	88.55	88.55	NUM
fcis-5209	91	19	diff(bert	diff(bert	PROPN
fcis-5209	91	20	-	-	PUNCT
fcis-5209	91	21	wwm	wwm	NOUN
fcis-5209	91	22	)	)	PUNCT
fcis-5209	91	23	87.80	87.80	NUM
fcis-5209	91	24	-	-	PUNCT
fcis-5209	91	25	our	our	PRON
fcis-5209	91	26	method	method	NOUN
fcis-5209	91	27	bbmc	bbmc	NOUN
fcis-5209	91	28	88.01	88.01	NUM
fcis-5209	91	29	88.72	88.72	NUM
fcis-5209	91	30	from	from	ADP
fcis-5209	91	31	the	the	DET
fcis-5209	91	32	experimental	experimental	ADJ
fcis-5209	91	33	results	result	NOUN
fcis-5209	91	34	,	,	PUNCT
fcis-5209	91	35	it	it	PRON
fcis-5209	91	36	can	can	AUX
fcis-5209	91	37	be	be	AUX
fcis-5209	91	38	seen	see	VERB
fcis-5209	91	39	that	that	SCONJ
fcis-5209	91	40	compared	compare	VERB
fcis-5209	91	41	with	with	ADP
fcis-5209	91	42	the	the	DET
fcis-5209	91	43	bert	bert	PROPN
fcis-5209	91	44	model	model	NOUN
fcis-5209	91	45	alone	alone	ADV
fcis-5209	91	46	,	,	PUNCT
fcis-5209	91	47	the	the	DET
fcis-5209	91	48	acc	acc	PROPN
fcis-5209	91	49	and	and	CCONJ
fcis-5209	91	50	f1	f1	PROPN
fcis-5209	91	51	score	score	NOUN
fcis-5209	91	52	of	of	ADP
fcis-5209	91	53	the	the	DET
fcis-5209	91	54	model	model	NOUN
fcis-5209	91	55	proposed	propose	VERB
fcis-5209	91	56	in	in	ADP
fcis-5209	91	57	this	this	DET
fcis-5209	91	58	paper	paper	NOUN
fcis-5209	91	59	are	be	AUX
fcis-5209	91	60	improved	improve	VERB
fcis-5209	91	61	.	.	PUNCT
fcis-5209	92	1	and	and	CCONJ
fcis-5209	92	2	the	the	DET
fcis-5209	92	3	acc	acc	PROPN
fcis-5209	92	4	is	be	AUX
fcis-5209	92	5	increased	increase	VERB
fcis-5209	92	6	by	by	ADP
fcis-5209	92	7	1.45	1.45	NUM
fcis-5209	92	8	%	%	NOUN
fcis-5209	92	9	and	and	CCONJ
fcis-5209	92	10	0.14	0.14	NUM
fcis-5209	92	11	%	%	NOUN
fcis-5209	92	12	respectively	respectively	ADV
fcis-5209	92	13	compared	compare	VERB
fcis-5209	92	14	with	with	ADP
fcis-5209	92	15	the	the	DET
fcis-5209	92	16	method	method	NOUN
fcis-5209	92	17	of	of	ADP
fcis-5209	92	18	splicing	splice	VERB
fcis-5209	92	19	the	the	DET
fcis-5209	92	20	bilstm	bilstm	NOUN
fcis-5209	92	21	module	module	NOUN
fcis-5209	92	22	and	and	CCONJ
fcis-5209	92	23	the	the	DET
fcis-5209	92	24	bilstm	bilstm	NOUN
fcis-5209	92	25	+	+	CCONJ
fcis-5209	92	26	conv	conv	ADJ
fcis-5209	92	27	block	block	NOUN
fcis-5209	92	28	module	module	NOUN
fcis-5209	92	29	.	.	PUNCT
fcis-5209	93	1	the	the	DET
fcis-5209	93	2	f1	f1	NOUN
fcis-5209	93	3	-	-	PUNCT
fcis-5209	93	4	score	score	NOUN
fcis-5209	93	5	is	be	AUX
fcis-5209	93	6	increased	increase	VERB
fcis-5209	93	7	by	by	ADP
fcis-5209	93	8	1.09	1.09	NUM
fcis-5209	93	9	%	%	NOUN
fcis-5209	93	10	and	and	CCONJ
fcis-5209	93	11	0.17	0.17	NUM
fcis-5209	93	12	%	%	NOUN
fcis-5209	93	13	respectively	respectively	ADV
fcis-5209	93	14	.	.	PUNCT
fcis-5209	94	1	the	the	DET
fcis-5209	94	2	acc	acc	PROPN
fcis-5209	94	3	is	be	AUX
fcis-5209	94	4	0.16	0.16	NUM
fcis-5209	94	5	%	%	NOUN
fcis-5209	94	6	lower	lower	ADV
fcis-5209	94	7	and	and	CCONJ
fcis-5209	94	8	the	the	DET
fcis-5209	94	9	f1score	f1score	NOUN
fcis-5209	94	10	is	be	AUX
fcis-5209	94	11	0.06	0.06	NUM
fcis-5209	94	12	%	%	NOUN
fcis-5209	94	13	higher	high	ADJ
fcis-5209	94	14	than	than	ADP
fcis-5209	94	15	the	the	DET
fcis-5209	94	16	method	method	NOUN
fcis-5209	94	17	of	of	ADP
fcis-5209	94	18	splicing	splice	VERB
fcis-5209	94	19	the	the	DET
fcis-5209	94	20	conv	conv	ADJ
fcis-5209	94	21	block	block	NOUN
fcis-5209	94	22	module	module	NOUN
fcis-5209	94	23	.	.	PUNCT
fcis-5209	95	1	and	and	CCONJ
fcis-5209	95	2	the	the	DET
fcis-5209	95	3	acc	acc	PROPN
fcis-5209	95	4	has	have	AUX
fcis-5209	95	5	improved	improve	VERB
fcis-5209	95	6	compared	compare	VERB
fcis-5209	95	7	with	with	ADP
fcis-5209	95	8	the	the	DET
fcis-5209	95	9	diff	diff	NOUN
fcis-5209	95	10	(	(	PUNCT
fcis-5209	95	11	bert	bert	PROPN
fcis-5209	95	12	-	-	PUNCT
fcis-5209	95	13	wwm	wwm	PROPN
fcis-5209	95	14	)	)	PUNCT
fcis-5209	95	15	model	model	NOUN
fcis-5209	95	16	proposed	propose	VERB
fcis-5209	95	17	in	in	ADP
fcis-5209	95	18	paper	paper	NOUN
fcis-5209	95	19	.	.	PUNCT
fcis-5209	96	1	this	this	PRON
fcis-5209	96	2	shows	show	VERB
fcis-5209	96	3	that	that	SCONJ
fcis-5209	96	4	the	the	DET
fcis-5209	96	5	model	model	NOUN
fcis-5209	96	6	proposed	propose	VERB
fcis-5209	96	7	in	in	ADP
fcis-5209	96	8	this	this	DET
fcis-5209	96	9	paper	paper	NOUN
fcis-5209	96	10	has	have	AUX
fcis-5209	96	11	improved	improve	VERB
fcis-5209	96	12	in	in	ADP
fcis-5209	96	13	general	general	ADJ
fcis-5209	96	14	.	.	PUNCT
fcis-5209	97	1	4	4	X
fcis-5209	97	2	.	.	X
fcis-5209	97	3	conclusion	conclusion	NOUN
fcis-5209	97	4	in	in	ADP
fcis-5209	97	5	order	order	NOUN
fcis-5209	97	6	to	to	PART
fcis-5209	97	7	cover	cover	VERB
fcis-5209	97	8	the	the	DET
fcis-5209	97	9	shortage	shortage	NOUN
fcis-5209	97	10	of	of	ADP
fcis-5209	97	11	only	only	ADV
fcis-5209	97	12	using	use	VERB
fcis-5209	97	13	the	the	DET
fcis-5209	97	14	bert	bert	PROPN
fcis-5209	97	15	model	model	NOUN
fcis-5209	97	16	can	can	AUX
fcis-5209	97	17	not	not	PART
fcis-5209	97	18	well	well	ADV
fcis-5209	97	19	express	express	VERB
fcis-5209	97	20	the	the	DET
fcis-5209	97	21	text	text	NOUN
fcis-5209	97	22	semantic	semantic	ADJ
fcis-5209	97	23	features	feature	NOUN
fcis-5209	97	24	,	,	PUNCT
fcis-5209	97	25	this	this	DET
fcis-5209	97	26	paper	paper	NOUN
fcis-5209	97	27	uses	use	VERB
fcis-5209	97	28	the	the	DET
fcis-5209	97	29	hidden	hide	VERB
fcis-5209	97	30	layer	layer	NOUN
fcis-5209	97	31	vectors	vector	NOUN
fcis-5209	97	32	of	of	ADP
fcis-5209	97	33	different	different	ADJ
fcis-5209	97	34	layers	layer	NOUN
fcis-5209	97	35	output	output	NOUN
fcis-5209	97	36	by	by	ADP
fcis-5209	97	37	the	the	DET
fcis-5209	97	38	bert	bert	PROPN
fcis-5209	97	39	model	model	NOUN
fcis-5209	97	40	and	and	CCONJ
fcis-5209	97	41	uses	use	VERB
fcis-5209	97	42	bilstm	bilstm	NOUN
fcis-5209	97	43	and	and	CCONJ
fcis-5209	97	44	multi	multi	ADJ
fcis-5209	97	45	-	-	ADJ
fcis-5209	97	46	scale	scale	ADJ
fcis-5209	97	47	convolution	convolution	NOUN
fcis-5209	97	48	network	network	NOUN
fcis-5209	97	49	to	to	PART
fcis-5209	97	50	further	far	ADV
fcis-5209	97	51	extract	extract	VERB
fcis-5209	97	52	rich	rich	ADJ
fcis-5209	97	53	semantic	semantic	ADJ
fcis-5209	97	54	features	feature	NOUN
fcis-5209	97	55	.	.	PUNCT
fcis-5209	98	1	through	through	ADP
fcis-5209	98	2	training	training	NOUN
fcis-5209	98	3	,	,	PUNCT
fcis-5209	98	4	validating	validate	VERB
fcis-5209	98	5	and	and	CCONJ
fcis-5209	98	6	testing	testing	NOUN
fcis-5209	98	7	on	on	ADP
fcis-5209	98	8	lcqmc	lcqmc	PROPN
fcis-5209	98	9	dataset	dataset	NOUN
fcis-5209	98	10	,	,	PUNCT
fcis-5209	98	11	it	it	PRON
fcis-5209	98	12	is	be	AUX
fcis-5209	98	13	proved	prove	VERB
fcis-5209	98	14	that	that	SCONJ
fcis-5209	98	15	the	the	DET
fcis-5209	98	16	fusion	fusion	NOUN
fcis-5209	98	17	of	of	ADP
fcis-5209	98	18	bilstm	bilstm	NOUN
fcis-5209	98	19	and	and	CCONJ
fcis-5209	98	20	multi	multi	ADJ
fcis-5209	98	21	-	-	ADJ
fcis-5209	98	22	scale	scale	ADJ
fcis-5209	98	23	convolution	convolution	NOUN
fcis-5209	98	24	can	can	AUX
fcis-5209	98	25	improve	improve	VERB
fcis-5209	98	26	the	the	DET
fcis-5209	98	27	effect	effect	NOUN
fcis-5209	98	28	of	of	ADP
fcis-5209	98	29	the	the	DET
fcis-5209	98	30	model	model	NOUN
fcis-5209	98	31	,	,	PUNCT
fcis-5209	98	32	which	which	PRON
fcis-5209	98	33	is	be	AUX
fcis-5209	98	34	sufficient	sufficient	ADJ
fcis-5209	98	35	to	to	PART
fcis-5209	98	36	show	show	VERB
fcis-5209	98	37	that	that	SCONJ
fcis-5209	98	38	it	it	PRON
fcis-5209	98	39	is	be	AUX
fcis-5209	98	40	feasible	feasible	ADJ
fcis-5209	98	41	to	to	ADP
fcis-5209	98	42	fuse	fuse	NOUN
fcis-5209	98	43	bilstm	bilstm	NOUN
fcis-5209	98	44	and	and	CCONJ
fcis-5209	98	45	multi	multi	ADJ
fcis-5209	98	46	-	-	ADJ
fcis-5209	98	47	scale	scale	ADJ
fcis-5209	98	48	convolution	convolution	NOUN
fcis-5209	98	49	on	on	ADP
fcis-5209	98	50	the	the	DET
fcis-5209	98	51	basis	basis	NOUN
fcis-5209	98	52	of	of	ADP
fcis-5209	98	53	pretrained	pretraine	VERB
fcis-5209	98	54	bert	bert	PROPN
fcis-5209	98	55	model	model	NOUN
fcis-5209	98	56	to	to	ADP
fcis-5209	98	57	finetune	finetune	NOUN
fcis-5209	98	58	.	.	PUNCT
fcis-5209	99	1	references	reference	NOUN
fcis-5209	99	2	[	[	X
fcis-5209	99	3	1	1	X
fcis-5209	99	4	]	]	PUNCT
fcis-5209	99	5	pang	pang	PROPN
fcis-5209	99	6	liang	liang	PROPN
fcis-5209	99	7	,	,	PUNCT
fcis-5209	99	8	lan	lan	PROPN
fcis-5209	99	9	yanyan	yanyan	PROPN
fcis-5209	99	10	,	,	PUNCT
fcis-5209	99	11	xu	xu	PROPN
fcis-5209	99	12	jun	jun	PROPN
fcis-5209	99	13	,	,	PUNCT
fcis-5209	99	14	guo	guo	PROPN
fcis-5209	99	15	jiafeng	jiafeng	PROPN
fcis-5209	99	16	,	,	PUNCT
fcis-5209	99	17	wan	wan	PROPN
fcis-5209	99	18	shengxian	shengxian	PROPN
fcis-5209	99	19	,	,	PUNCT
fcis-5209	99	20	cheng	cheng	PROPN
fcis-5209	99	21	xueqi	xueqi	PROPN
fcis-5209	99	22	.	.	PUNCT
fcis-5209	100	1	overview	overview	NOUN
fcis-5209	100	2	of	of	ADP
fcis-5209	100	3	deep	deep	ADJ
fcis-5209	100	4	text	text	NOUN
fcis-5209	100	5	matching	matching	NOUN
fcis-5209	100	6	[	[	X
fcis-5209	100	7	j	j	X
fcis-5209	100	8	]	]	X
fcis-5209	100	9	.	.	PUNCT
fcis-5209	101	1	journal	journal	PROPN
fcis-5209	101	2	of	of	ADP
fcis-5209	101	3	computer	computer	NOUN
fcis-5209	101	4	science	science	NOUN
fcis-5209	101	5	,	,	PUNCT
fcis-5209	101	6	2017,40	2017,40	NUM
fcis-5209	101	7	(	(	PUNCT
fcis-5209	101	8	04	04	NUM
fcis-5209	101	9	):	):	PUNCT
fcis-5209	101	10	985	985	NUM
fcis-5209	101	11	-	-	SYM
fcis-5209	101	12	1003	1003	NUM
fcis-5209	101	13	.	.	PUNCT
fcis-5209	102	1	[	[	X
fcis-5209	102	2	2	2	NUM
fcis-5209	102	3	]	]	PUNCT
fcis-5209	102	4	yang	yang	PROPN
fcis-5209	102	5	r	r	PROPN
fcis-5209	102	6	,	,	PUNCT
fcis-5209	102	7	zhang	zhang	PROPN
fcis-5209	102	8	j	j	PROPN
fcis-5209	102	9	,	,	PUNCT
fcis-5209	102	10	gao	gao	PROPN
fcis-5209	102	11	x	x	PROPN
fcis-5209	102	12	,	,	PUNCT
fcis-5209	102	13	ji	ji	PROPN
fcis-5209	102	14	f	f	PROPN
fcis-5209	102	15	,	,	PUNCT
fcis-5209	102	16	chen	chen	PROPN
fcis-5209	102	17	h.	h.	PROPN
fcis-5209	102	18	simple	simple	ADJ
fcis-5209	102	19	and	and	CCONJ
fcis-5209	102	20	effective	effective	ADJ
fcis-5209	102	21	text	text	NOUN
fcis-5209	102	22	matching	matching	NOUN
fcis-5209	102	23	with	with	ADP
fcis-5209	102	24	richer	rich	ADJ
fcis-5209	102	25	alignment	alignment	NOUN
fcis-5209	102	26	features[j	features[j	NOUN
fcis-5209	102	27	]	]	PUNCT
fcis-5209	102	28	.	.	PUNCT
fcis-5209	103	1	arxiv	arxiv	PROPN
fcis-5209	103	2	preprint	preprint	PROPN
fcis-5209	103	3	arxiv:1908.00300	arxiv:1908.00300	NUM
fcis-5209	103	4	,	,	PUNCT
fcis-5209	103	5	2019	2019	NUM
fcis-5209	103	6	.	.	PUNCT
fcis-5209	104	1	[	[	X
fcis-5209	104	2	3	3	X
fcis-5209	104	3	]	]	X
fcis-5209	104	4	huang	huang	PROPN
fcis-5209	104	5	p	p	PROPN
fcis-5209	104	6	s	s	PROPN
fcis-5209	104	7	,	,	PUNCT
fcis-5209	104	8	he	he	PRON
fcis-5209	104	9	x	x	PROPN
fcis-5209	104	10	,	,	PUNCT
fcis-5209	104	11	gao	gao	PROPN
fcis-5209	104	12	j	j	PROPN
fcis-5209	104	13	,	,	PUNCT
fcis-5209	104	14	deng	deng	PROPN
fcis-5209	104	15	l	l	PROPN
fcis-5209	104	16	,	,	PUNCT
fcis-5209	104	17	acero	acero	PROPN
fcis-5209	104	18	a	a	PROPN
fcis-5209	104	19	,	,	PUNCT
fcis-5209	104	20	heck	heck	INTJ
fcis-5209	104	21	l.	l.	PROPN
fcis-5209	104	22	learning	learn	VERB
fcis-5209	104	23	deep	deep	ADV
fcis-5209	104	24	structured	structured	ADJ
fcis-5209	104	25	semantic	semantic	ADJ
fcis-5209	104	26	models	model	NOUN
fcis-5209	104	27	for	for	ADP
fcis-5209	104	28	web	web	NOUN
fcis-5209	104	29	search	search	NOUN
fcis-5209	104	30	using	use	VERB
fcis-5209	104	31	clickthrough	clickthrough	NOUN
fcis-5209	104	32	data[c]//proceedings	data[c]//proceeding	NOUN
fcis-5209	104	33	of	of	ADP
fcis-5209	104	34	the	the	DET
fcis-5209	104	35	22nd	22nd	PROPN
fcis-5209	104	36	acm	acm	PROPN
fcis-5209	104	37	international	international	ADJ
fcis-5209	104	38	conference	conference	NOUN
fcis-5209	104	39	on	on	ADP
fcis-5209	104	40	information	information	NOUN
fcis-5209	104	41	&	&	CCONJ
fcis-5209	104	42	knowledge	knowledge	PROPN
fcis-5209	104	43	management	management	PROPN
fcis-5209	104	44	.	.	PUNCT
fcis-5209	105	1	2013	2013	NUM
fcis-5209	105	2	:	:	PUNCT
fcis-5209	105	3	2333	2333	NUM
fcis-5209	105	4	-	-	SYM
fcis-5209	105	5	2338	2338	NUM
fcis-5209	105	6	.	.	PUNCT
fcis-5209	106	1	43	43	NUM
fcis-5209	107	1	[	[	X
fcis-5209	107	2	4	4	NUM
fcis-5209	107	3	]	]	X
fcis-5209	107	4	shen	shen	PROPN
fcis-5209	107	5	y	y	PROPN
fcis-5209	107	6	,	,	PUNCT
fcis-5209	107	7	he	he	PRON
fcis-5209	107	8	x	x	PROPN
fcis-5209	107	9	,	,	PUNCT
fcis-5209	107	10	gao	gao	PROPN
fcis-5209	107	11	j	j	PROPN
fcis-5209	107	12	,	,	PUNCT
fcis-5209	107	13	deng	deng	PROPN
fcis-5209	107	14	l	l	PROPN
fcis-5209	107	15	,	,	PUNCT
fcis-5209	107	16	mesnil	mesnil	PROPN
fcis-5209	107	17	g.	g.	PROPN
fcis-5209	107	18	a	a	DET
fcis-5209	107	19	latent	latent	ADJ
fcis-5209	107	20	semantic	semantic	ADJ
fcis-5209	107	21	model	model	NOUN
fcis-5209	107	22	with	with	ADP
fcis-5209	107	23	convolutional	convolutional	ADJ
fcis-5209	107	24	-	-	PUNCT
fcis-5209	107	25	pooling	pool	VERB
fcis-5209	107	26	structure	structure	NOUN
fcis-5209	107	27	for	for	ADP
fcis-5209	107	28	information	information	NOUN
fcis-5209	107	29	retrieval[c]//proceedings	retrieval[c]//proceeding	NOUN
fcis-5209	107	30	of	of	ADP
fcis-5209	107	31	the	the	DET
fcis-5209	107	32	23rd	23rd	ADJ
fcis-5209	107	33	acm	acm	PROPN
fcis-5209	107	34	international	international	ADJ
fcis-5209	107	35	conference	conference	NOUN
fcis-5209	107	36	on	on	ADP
fcis-5209	107	37	conference	conference	NOUN
fcis-5209	107	38	on	on	ADP
fcis-5209	107	39	information	information	NOUN
fcis-5209	107	40	and	and	CCONJ
fcis-5209	107	41	knowledge	knowledge	NOUN
fcis-5209	107	42	management	management	NOUN
fcis-5209	107	43	.	.	PUNCT
fcis-5209	108	1	2014	2014	NUM
fcis-5209	108	2	:	:	PUNCT
fcis-5209	109	1	101	101	NUM
fcis-5209	109	2	-	-	SYM
fcis-5209	109	3	110	110	NUM
fcis-5209	109	4	.	.	PUNCT
fcis-5209	110	1	[	[	X
fcis-5209	110	2	5	5	X
fcis-5209	110	3	]	]	PUNCT
fcis-5209	110	4	palangi	palangi	NOUN
fcis-5209	110	5	h	h	PROPN
fcis-5209	110	6	,	,	PUNCT
fcis-5209	110	7	deng	deng	PROPN
fcis-5209	110	8	l	l	PROPN
fcis-5209	110	9	,	,	PUNCT
fcis-5209	110	10	shen	shen	PROPN
fcis-5209	110	11	y	y	PROPN
fcis-5209	110	12	,	,	PUNCT
fcis-5209	110	13	gao	gao	PROPN
fcis-5209	110	14	j	j	PROPN
fcis-5209	110	15	,	,	PUNCT
fcis-5209	110	16	he	he	PRON
fcis-5209	110	17	x	x	PROPN
fcis-5209	110	18	,	,	PUNCT
fcis-5209	110	19	chen	chen	PROPN
fcis-5209	110	20	j	j	PROPN
fcis-5209	110	21	,	,	PUNCT
fcis-5209	110	22	et	et	PROPN
fcis-5209	110	23	al	al	PROPN
fcis-5209	110	24	.	.	PROPN
fcis-5209	110	25	semantic	semantic	ADJ
fcis-5209	110	26	modelling	modelling	NOUN
fcis-5209	110	27	with	with	ADP
fcis-5209	110	28	long	long	ADJ
fcis-5209	110	29	-	-	PUNCT
fcis-5209	110	30	short	short	ADJ
fcis-5209	110	31	-	-	PUNCT
fcis-5209	110	32	term	term	NOUN
fcis-5209	110	33	memory	memory	NOUN
fcis-5209	110	34	for	for	ADP
fcis-5209	110	35	information	information	NOUN
fcis-5209	110	36	retrieval[j	retrieval[j	PROPN
fcis-5209	110	37	]	]	PUNCT
fcis-5209	110	38	.	.	PUNCT
fcis-5209	111	1	arxiv	arxiv	PROPN
fcis-5209	111	2	preprint	preprint	PROPN
fcis-5209	111	3	arxiv:1412.6629	arxiv:1412.6629	NOUN
fcis-5209	111	4	,	,	PUNCT
fcis-5209	111	5	2014	2014	NUM
fcis-5209	111	6	.	.	PUNCT
fcis-5209	112	1	[	[	X
fcis-5209	112	2	6	6	NUM
fcis-5209	112	3	]	]	SYM
fcis-5209	112	4	hu	hu	PROPN
fcis-5209	112	5	b	b	PROPN
fcis-5209	112	6	,	,	PUNCT
fcis-5209	112	7	lu	lu	PROPN
fcis-5209	112	8	z	z	PROPN
fcis-5209	112	9	,	,	PUNCT
fcis-5209	112	10	li	li	PROPN
fcis-5209	112	11	h	h	PROPN
fcis-5209	112	12	,	,	PUNCT
fcis-5209	112	13	chen	chen	PROPN
fcis-5209	112	14	q.	q.	PROPN
fcis-5209	112	15	convolutional	convolutional	ADJ
fcis-5209	112	16	neural	neural	ADJ
fcis-5209	112	17	network	network	NOUN
fcis-5209	112	18	architectures	architecture	NOUN
fcis-5209	112	19	for	for	ADP
fcis-5209	112	20	matching	match	VERB
fcis-5209	112	21	natural	natural	ADJ
fcis-5209	112	22	language	language	NOUN
fcis-5209	112	23	sentences[j	sentences[j	PROPN
fcis-5209	112	24	]	]	PUNCT
fcis-5209	112	25	.	.	PUNCT
fcis-5209	113	1	advances	advance	NOUN
fcis-5209	113	2	in	in	ADP
fcis-5209	113	3	neural	neural	ADJ
fcis-5209	113	4	information	information	NOUN
fcis-5209	113	5	processing	processing	NOUN
fcis-5209	113	6	systems	system	NOUN
fcis-5209	113	7	,	,	PUNCT
fcis-5209	113	8	2014	2014	NUM
fcis-5209	113	9	,	,	PUNCT
fcis-5209	113	10	27	27	NUM
fcis-5209	113	11	.	.	PUNCT
fcis-5209	114	1	[	[	X
fcis-5209	114	2	7	7	X
fcis-5209	114	3	]	]	X
fcis-5209	114	4	yin	yin	PROPN
fcis-5209	114	5	w	w	PROPN
fcis-5209	114	6	,	,	PUNCT
fcis-5209	114	7	schütze	schütze	PROPN
fcis-5209	114	8	h	h	PROPN
fcis-5209	114	9	,	,	PUNCT
fcis-5209	114	10	xiang	xiang	PROPN
fcis-5209	114	11	b	b	PROPN
fcis-5209	114	12	,	,	PUNCT
fcis-5209	114	13	zhou	zhou	PROPN
fcis-5209	114	14	b.	b.	PROPN
fcis-5209	114	15	abcnn	abcnn	PROPN
fcis-5209	114	16	:	:	PUNCT
fcis-5209	114	17	attention	attention	NOUN
fcis-5209	114	18	-	-	PUNCT
fcis-5209	114	19	based	base	VERB
fcis-5209	114	20	convolutional	convolutional	ADJ
fcis-5209	114	21	neural	neural	ADJ
fcis-5209	114	22	network	network	NOUN
fcis-5209	114	23	for	for	ADP
fcis-5209	114	24	modeling	model	VERB
fcis-5209	114	25	sentence	sentence	NOUN
fcis-5209	114	26	pairs[j	pairs[j	NOUN
fcis-5209	114	27	]	]	PUNCT
fcis-5209	114	28	.	.	PUNCT
fcis-5209	115	1	transactions	transaction	NOUN
fcis-5209	115	2	of	of	ADP
fcis-5209	115	3	the	the	DET
fcis-5209	115	4	association	association	NOUN
fcis-5209	115	5	for	for	ADP
fcis-5209	115	6	computational	computational	ADJ
fcis-5209	115	7	linguistics	linguistic	NOUN
fcis-5209	115	8	,	,	PUNCT
fcis-5209	115	9	2016	2016	NUM
fcis-5209	115	10	,	,	PUNCT
fcis-5209	115	11	4	4	NUM
fcis-5209	115	12	:	:	SYM
fcis-5209	115	13	259	259	NUM
fcis-5209	115	14	-	-	SYM
fcis-5209	115	15	272	272	NUM
fcis-5209	115	16	.	.	PUNCT
fcis-5209	116	1	[	[	X
fcis-5209	116	2	8	8	NUM
fcis-5209	116	3	]	]	X
fcis-5209	116	4	pang	pang	NOUN
fcis-5209	116	5	l	l	NOUN
fcis-5209	116	6	,	,	PUNCT
fcis-5209	116	7	lan	lan	PROPN
fcis-5209	116	8	y	y	PROPN
fcis-5209	116	9	,	,	PUNCT
fcis-5209	116	10	guo	guo	PROPN
fcis-5209	116	11	j	j	PROPN
fcis-5209	116	12	,	,	PUNCT
fcis-5209	117	1	xu	xu	PROPN
fcis-5209	117	2	j	j	PROPN
fcis-5209	117	3	,	,	PUNCT
fcis-5209	117	4	wan	wan	PROPN
fcis-5209	117	5	s	s	PROPN
fcis-5209	117	6	,	,	PUNCT
fcis-5209	117	7	cheng	cheng	PROPN
fcis-5209	117	8	x.	x.	PROPN
fcis-5209	117	9	text	text	PROPN
fcis-5209	117	10	matching	matching	NOUN
fcis-5209	117	11	as	as	ADP
fcis-5209	117	12	image	image	NOUN
fcis-5209	117	13	recognition[c]//proceedings	recognition[c]//proceeding	NOUN
fcis-5209	117	14	of	of	ADP
fcis-5209	117	15	the	the	DET
fcis-5209	117	16	aaai	aaai	PROPN
fcis-5209	117	17	conference	conference	NOUN
fcis-5209	117	18	on	on	ADP
fcis-5209	117	19	artificial	artificial	ADJ
fcis-5209	117	20	intelligence	intelligence	NOUN
fcis-5209	117	21	.	.	PUNCT
fcis-5209	118	1	2016	2016	NUM
fcis-5209	118	2	,	,	PUNCT
fcis-5209	118	3	30(1	30(1	NUM
fcis-5209	118	4	)	)	PUNCT
fcis-5209	118	5	.	.	PUNCT
fcis-5209	119	1	[	[	X
fcis-5209	119	2	9	9	NUM
fcis-5209	119	3	]	]	X
fcis-5209	119	4	chen	chen	PROPN
fcis-5209	119	5	q	q	X
fcis-5209	119	6	,	,	PUNCT
fcis-5209	119	7	zhu	zhu	PROPN
fcis-5209	119	8	x	x	X
fcis-5209	119	9	,	,	PUNCT
fcis-5209	119	10	ling	ling	PROPN
fcis-5209	119	11	z	z	PROPN
fcis-5209	119	12	,	,	PUNCT
fcis-5209	119	13	wei	wei	PROPN
fcis-5209	119	14	s	s	PROPN
fcis-5209	119	15	,	,	PUNCT
fcis-5209	119	16	jiang	jiang	PROPN
fcis-5209	119	17	h	h	PROPN
fcis-5209	119	18	,	,	PUNCT
fcis-5209	119	19	inkpen	inkpen	PROPN
fcis-5209	119	20	d.	d.	PROPN
fcis-5209	119	21	enhanced	enhance	VERB
fcis-5209	119	22	lstm	lstm	NOUN
fcis-5209	119	23	for	for	ADP
fcis-5209	119	24	natural	natural	ADJ
fcis-5209	119	25	language	language	NOUN
fcis-5209	119	26	inference[j	inference[j	NOUN
fcis-5209	119	27	]	]	PUNCT
fcis-5209	119	28	.	.	PUNCT
fcis-5209	120	1	arxiv	arxiv	PROPN
fcis-5209	120	2	preprint	preprint	PROPN
fcis-5209	120	3	arxiv:1609.06038	arxiv:1609.06038	PROPN
fcis-5209	120	4	,	,	PUNCT
fcis-5209	120	5	2016	2016	NUM
fcis-5209	120	6	.	.	PUNCT
fcis-5209	121	1	[	[	X
fcis-5209	121	2	10	10	NUM
fcis-5209	121	3	]	]	AUX
fcis-5209	121	4	graves	grave	VERB
fcis-5209	121	5	a	a	DET
fcis-5209	121	6	,	,	PUNCT
fcis-5209	121	7	schmidhuber	schmidhuber	NOUN
fcis-5209	121	8	j.	j.	PROPN
fcis-5209	121	9	framewise	framewise	PROPN
fcis-5209	121	10	phoneme	phoneme	NOUN
fcis-5209	121	11	classification	classification	NOUN
fcis-5209	121	12	with	with	ADP
fcis-5209	121	13	bidirectional	bidirectional	ADJ
fcis-5209	121	14	lstm	lstm	ADJ
fcis-5209	121	15	networks[c]//proceedings	networks[c]//proceedings	PROPN
fcis-5209	121	16	.	.	PUNCT
fcis-5209	121	17	2005	2005	NUM
fcis-5209	121	18	ieee	ieee	PROPN
fcis-5209	121	19	international	international	ADJ
fcis-5209	121	20	joint	joint	ADJ
fcis-5209	121	21	conference	conference	NOUN
fcis-5209	121	22	on	on	ADP
fcis-5209	121	23	neural	neural	ADJ
fcis-5209	121	24	networks	network	NOUN
fcis-5209	121	25	,	,	PUNCT
fcis-5209	121	26	2005	2005	NUM
fcis-5209	121	27	.	.	PUNCT
fcis-5209	122	1	ieee	ieee	NOUN
fcis-5209	122	2	,	,	PUNCT
fcis-5209	122	3	4	4	NUM
fcis-5209	122	4	:	:	SYM
fcis-5209	122	5	2047	2047	NUM
fcis-5209	122	6	-	-	SYM
fcis-5209	122	7	2052	2052	NUM
fcis-5209	122	8	.	.	PUNCT
fcis-5209	123	1	[	[	X
fcis-5209	123	2	11	11	NUM
fcis-5209	123	3	]	]	X
fcis-5209	123	4	wang	wang	PROPN
fcis-5209	123	5	z	z	PROPN
fcis-5209	123	6	,	,	PUNCT
fcis-5209	123	7	hamza	hamza	PROPN
fcis-5209	123	8	w	w	PROPN
fcis-5209	123	9	,	,	PUNCT
fcis-5209	123	10	florian	florian	PROPN
fcis-5209	123	11	r.	r.	PROPN
fcis-5209	123	12	bilateral	bilateral	PROPN
fcis-5209	123	13	multi	multi	ADJ
fcis-5209	123	14	-	-	NOUN
fcis-5209	123	15	perspective	perspective	ADJ
fcis-5209	123	16	matching	matching	NOUN
fcis-5209	123	17	for	for	ADP
fcis-5209	123	18	natural	natural	ADJ
fcis-5209	123	19	language	language	NOUN
fcis-5209	123	20	sentences[j	sentences[j	NOUN
fcis-5209	123	21	]	]	PUNCT
fcis-5209	123	22	.	.	PUNCT
fcis-5209	124	1	arxiv	arxiv	PROPN
fcis-5209	124	2	preprint	preprint	NOUN
fcis-5209	124	3	arxiv:1702.03814	arxiv:1702.03814	NOUN
fcis-5209	124	4	,	,	PUNCT
fcis-5209	124	5	2017	2017	NUM
fcis-5209	124	6	.	.	PUNCT
fcis-5209	125	1	[	[	X
fcis-5209	125	2	12	12	NUM
fcis-5209	125	3	]	]	X
fcis-5209	125	4	mikolov	mikolov	PROPN
fcis-5209	125	5	t	t	PROPN
fcis-5209	125	6	,	,	PUNCT
fcis-5209	125	7	sutskever	sutskever	VERB
fcis-5209	125	8	i	i	PRON
fcis-5209	125	9	,	,	PUNCT
fcis-5209	125	10	chen	chen	PROPN
fcis-5209	125	11	k	k	PROPN
fcis-5209	125	12	,	,	PUNCT
fcis-5209	125	13	corrado	corrado	PROPN
fcis-5209	125	14	g	g	PROPN
fcis-5209	125	15	,	,	PUNCT
fcis-5209	125	16	dean	dean	PROPN
fcis-5209	125	17	j.	j.	PROPN
fcis-5209	125	18	distributed	distribute	VERB
fcis-5209	125	19	representations	representation	NOUN
fcis-5209	125	20	of	of	ADP
fcis-5209	125	21	words	word	NOUN
fcis-5209	125	22	and	and	CCONJ
fcis-5209	125	23	phrases	phrase	NOUN
fcis-5209	125	24	and	and	CCONJ
fcis-5209	125	25	their	their	PRON
fcis-5209	125	26	compositionality[j	compositionality[j	NOUN
fcis-5209	125	27	]	]	PUNCT
fcis-5209	125	28	.	.	PUNCT
fcis-5209	126	1	advances	advance	NOUN
fcis-5209	126	2	in	in	ADP
fcis-5209	126	3	neural	neural	ADJ
fcis-5209	126	4	information	information	NOUN
fcis-5209	126	5	processing	processing	NOUN
fcis-5209	126	6	systems	system	NOUN
fcis-5209	126	7	,	,	PUNCT
fcis-5209	126	8	2013	2013	NUM
fcis-5209	126	9	,	,	PUNCT
fcis-5209	126	10	26	26	NUM
fcis-5209	126	11	.	.	PUNCT
fcis-5209	127	1	[	[	X
fcis-5209	127	2	13	13	NUM
fcis-5209	127	3	]	]	X
fcis-5209	127	4	pennington	pennington	PROPN
fcis-5209	127	5	j	j	PROPN
fcis-5209	127	6	,	,	PUNCT
fcis-5209	127	7	socher	socher	PROPN
fcis-5209	127	8	r	r	NOUN
fcis-5209	127	9	,	,	PUNCT
fcis-5209	127	10	manning	man	VERB
fcis-5209	127	11	c	c	PROPN
fcis-5209	127	12	d.	d.	PROPN
fcis-5209	127	13	glove	glove	PROPN
fcis-5209	127	14	:	:	PUNCT
fcis-5209	127	15	global	global	ADJ
fcis-5209	127	16	vectors	vector	NOUN
fcis-5209	127	17	for	for	ADP
fcis-5209	127	18	word	word	NOUN
fcis-5209	127	19	representation[c]//proceedings	representation[c]//proceedings	PROPN
fcis-5209	127	20	of	of	ADP
fcis-5209	127	21	the	the	DET
fcis-5209	127	22	2014	2014	NUM
fcis-5209	127	23	conference	conference	NOUN
fcis-5209	127	24	on	on	ADP
fcis-5209	127	25	empirical	empirical	ADJ
fcis-5209	127	26	methods	method	NOUN
fcis-5209	127	27	in	in	ADP
fcis-5209	127	28	natural	natural	ADJ
fcis-5209	127	29	language	language	NOUN
fcis-5209	127	30	processing	processing	NOUN
fcis-5209	127	31	(	(	PUNCT
fcis-5209	127	32	emnlp	emnlp	ADJ
fcis-5209	127	33	)	)	PUNCT
fcis-5209	127	34	.	.	PUNCT
fcis-5209	128	1	2014	2014	NUM
fcis-5209	128	2	:	:	PUNCT
fcis-5209	128	3	1532	1532	NUM
fcis-5209	128	4	-	-	SYM
fcis-5209	128	5	1543	1543	NUM
fcis-5209	128	6	.	.	PUNCT
fcis-5209	129	1	[	[	X
fcis-5209	129	2	14	14	NUM
fcis-5209	129	3	]	]	X
fcis-5209	129	4	devlin	devlin	PROPN
fcis-5209	129	5	j	j	PROPN
fcis-5209	129	6	,	,	PUNCT
fcis-5209	129	7	chang	chang	PROPN
fcis-5209	129	8	m	m	PROPN
fcis-5209	129	9	w	w	PROPN
fcis-5209	129	10	,	,	PUNCT
fcis-5209	129	11	lee	lee	PROPN
fcis-5209	129	12	k	k	PROPN
fcis-5209	129	13	,	,	PUNCT
fcis-5209	129	14	toutanova	toutanova	PROPN
fcis-5209	129	15	k.	k.	PROPN
fcis-5209	129	16	bert	bert	PROPN
fcis-5209	129	17	:	:	PUNCT
fcis-5209	129	18	pretraining	pretraine	VERB
fcis-5209	129	19	of	of	ADP
fcis-5209	129	20	deep	deep	ADJ
fcis-5209	129	21	bidirectional	bidirectional	ADJ
fcis-5209	129	22	transformers	transformer	NOUN
fcis-5209	129	23	for	for	ADP
fcis-5209	129	24	language	language	NOUN
fcis-5209	129	25	understanding[j	understanding[j	NOUN
fcis-5209	129	26	]	]	PUNCT
fcis-5209	129	27	.	.	PUNCT
fcis-5209	130	1	arxiv	arxiv	PROPN
fcis-5209	130	2	preprint	preprint	PROPN
fcis-5209	130	3	arxiv:1810.04805	arxiv:1810.04805	PROPN
fcis-5209	130	4	,	,	PUNCT
fcis-5209	130	5	2018	2018	NUM
fcis-5209	130	6	.	.	PUNCT
fcis-5209	131	1	[	[	X
fcis-5209	131	2	15	15	NUM
fcis-5209	131	3	]	]	X
fcis-5209	131	4	vaswani	vaswani	NOUN
fcis-5209	131	5	a	a	PRON
fcis-5209	131	6	,	,	PUNCT
fcis-5209	131	7	shazeer	shazeer	NOUN
fcis-5209	131	8	n	n	SYM
fcis-5209	131	9	,	,	PUNCT
fcis-5209	131	10	parmar	parmar	PROPN
fcis-5209	131	11	n	n	CCONJ
fcis-5209	131	12	,	,	PUNCT
fcis-5209	131	13	uszkoreit	uszkoreit	PROPN
fcis-5209	131	14	j	j	PROPN
fcis-5209	131	15	,	,	PUNCT
fcis-5209	131	16	jones	jones	PROPN
fcis-5209	131	17	l	l	PROPN
fcis-5209	131	18	,	,	PUNCT
fcis-5209	131	19	gomez	gomez	PROPN
fcis-5209	131	20	a	a	PROPN
fcis-5209	131	21	,	,	PUNCT
fcis-5209	131	22	et	et	PROPN
fcis-5209	131	23	al	al	PROPN
fcis-5209	131	24	.	.	PUNCT
fcis-5209	131	25	attention	attention	NOUN
fcis-5209	131	26	is	be	AUX
fcis-5209	131	27	all	all	PRON
fcis-5209	131	28	you	you	PRON
fcis-5209	131	29	need[j	need[j	VERB
fcis-5209	131	30	]	]	PUNCT
fcis-5209	131	31	.	.	PUNCT
fcis-5209	132	1	advances	advance	NOUN
fcis-5209	132	2	in	in	ADP
fcis-5209	132	3	neural	neural	ADJ
fcis-5209	132	4	information	information	NOUN
fcis-5209	132	5	processing	processing	NOUN
fcis-5209	132	6	systems	system	NOUN
fcis-5209	132	7	,	,	PUNCT
fcis-5209	132	8	2017	2017	NUM
fcis-5209	132	9	,	,	PUNCT
fcis-5209	132	10	30	30	NUM
fcis-5209	132	11	.	.	PUNCT
fcis-5209	133	1	[	[	X
fcis-5209	133	2	16	16	NUM
fcis-5209	133	3	]	]	X
fcis-5209	133	4	li	li	PROPN
fcis-5209	133	5	b	b	PROPN
fcis-5209	133	6	,	,	PUNCT
fcis-5209	133	7	zhou	zhou	PROPN
fcis-5209	133	8	h	h	PROPN
fcis-5209	133	9	,	,	PUNCT
fcis-5209	133	10	he	he	PRON
fcis-5209	133	11	j	j	PROPN
fcis-5209	133	12	,	,	PUNCT
fcis-5209	133	13	et	et	PROPN
fcis-5209	133	14	al	al	PROPN
fcis-5209	133	15	.	.	PROPN
fcis-5209	134	1	on	on	ADP
fcis-5209	134	2	the	the	DET
fcis-5209	134	3	sentence	sentence	NOUN
fcis-5209	134	4	embeddings	embedding	NOUN
fcis-5209	134	5	from	from	ADP
fcis-5209	134	6	pre	pre	ADJ
fcis-5209	134	7	-	-	ADJ
fcis-5209	134	8	trained	train	VERB
fcis-5209	134	9	language	language	NOUN
fcis-5209	134	10	models[c]//	models[c]//	PROPN
fcis-5209	134	11	proceedings	proceeding	NOUN
fcis-5209	134	12	of	of	ADP
fcis-5209	134	13	the	the	DET
fcis-5209	134	14	2020	2020	NUM
fcis-5209	134	15	conference	conference	NOUN
fcis-5209	134	16	on	on	ADP
fcis-5209	134	17	empirical	empirical	ADJ
fcis-5209	134	18	methods	method	NOUN
fcis-5209	134	19	in	in	ADP
fcis-5209	134	20	natural	natural	ADJ
fcis-5209	134	21	language	language	NOUN
fcis-5209	134	22	processing	processing	NOUN
fcis-5209	134	23	(	(	PUNCT
fcis-5209	134	24	emnlp	emnlp	ADJ
fcis-5209	134	25	)	)	PUNCT
fcis-5209	134	26	.	.	PUNCT
fcis-5209	135	1	2020	2020	NUM
fcis-5209	135	2	.	.	PUNCT
fcis-5209	136	1	[	[	X
fcis-5209	136	2	17	17	NUM
fcis-5209	136	3	]	]	PUNCT
fcis-5209	136	4	xia	xia	PROPN
fcis-5209	136	5	t	t	PROPN
fcis-5209	136	6	,	,	PUNCT
fcis-5209	136	7	wang	wang	PROPN
fcis-5209	136	8	y	y	PROPN
fcis-5209	136	9	,	,	PUNCT
fcis-5209	136	10	tian	tian	PROPN
fcis-5209	136	11	y	y	PROPN
fcis-5209	136	12	,	,	PUNCT
fcis-5209	136	13	chang	chang	PROPN
fcis-5209	136	14	y.	y.	PROPN
fcis-5209	136	15	using	use	VERB
fcis-5209	136	16	prior	prior	ADJ
fcis-5209	136	17	knowledge	knowledge	NOUN
fcis-5209	136	18	to	to	PART
fcis-5209	136	19	guide	guide	VERB
fcis-5209	136	20	bert	bert	PROPN
fcis-5209	136	21	's	's	PART
fcis-5209	136	22	attention	attention	NOUN
fcis-5209	136	23	in	in	ADP
fcis-5209	136	24	semantic	semantic	ADJ
fcis-5209	136	25	textual	textual	ADJ
fcis-5209	136	26	matching	matching	NOUN
fcis-5209	136	27	tasks[j	tasks[j	NOUN
fcis-5209	136	28	]	]	PUNCT
fcis-5209	136	29	.	.	PUNCT
fcis-5209	137	1	2021	2021	NUM
fcis-5209	137	2	.	.	PUNCT
fcis-5209	138	1	[	[	X
fcis-5209	138	2	18	18	NUM
fcis-5209	138	3	]	]	X
fcis-5209	138	4	meng	meng	PROPN
fcis-5209	138	5	jinxu	jinxu	PROPN
fcis-5209	138	6	,	,	PUNCT
fcis-5209	138	7	shan	shan	PROPN
fcis-5209	138	8	hongtao	hongtao	PROPN
fcis-5209	138	9	,	,	PUNCT
fcis-5209	138	10	wan	wan	PROPN
fcis-5209	138	11	junjie	junjie	PROPN
fcis-5209	138	12	,	,	PUNCT
fcis-5209	138	13	jia	jia	PROPN
fcis-5209	138	14	renxiang	renxiang	PROPN
fcis-5209	138	15	.	.	PUNCT
fcis-5209	138	16	bsla	bsla	PROPN
fcis-5209	138	17	:	:	PUNCT
fcis-5209	138	18	improved	improved	ADJ
fcis-5209	138	19	text	text	NOUN
fcis-5209	138	20	similarity	similarity	NOUN
fcis-5209	138	21	model	model	NOUN
fcis-5209	138	22	of	of	ADP
fcis-5209	138	23	siamese	siamese	ADJ
fcis-5209	138	24	lstm	lstm	PROPN
fcis-5209	138	25	[	[	X
fcis-5209	138	26	j	j	X
fcis-5209	138	27	]	]	X
fcis-5209	138	28	.	.	PUNCT
fcis-5209	139	1	computer	computer	NOUN
fcis-5209	139	2	engineering	engineering	NOUN
fcis-5209	139	3	and	and	CCONJ
fcis-5209	139	4	application	application	NOUN
fcis-5209	139	5	.	.	PUNCT
fcis-5209	140	1	2021	2021	NUM
fcis-5209	140	2	.	.	PUNCT
fcis-5209	141	1	[	[	X
fcis-5209	141	2	19	19	NUM
fcis-5209	141	3	]	]	PUNCT
fcis-5209	141	4	cui	cui	NOUN
fcis-5209	141	5	y	y	PROPN
fcis-5209	141	6	,	,	PUNCT
fcis-5209	141	7	w	w	PROPN
fcis-5209	141	8	che	che	PROPN
fcis-5209	141	9	,	,	PUNCT
fcis-5209	141	10	t	t	PROPN
fcis-5209	141	11	liu	liu	PROPN
fcis-5209	141	12	,	,	PUNCT
fcis-5209	141	13	b	b	PROPN
fcis-5209	141	14	qin	qin	X
fcis-5209	141	15	,	,	PUNCT
fcis-5209	141	16	z	z	PROPN
fcis-5209	141	17	yang	yang	PROPN
fcis-5209	141	18	,	,	PUNCT
fcis-5209	141	19	s	s	PROPN
fcis-5209	141	20	wang	wang	PROPN
fcis-5209	141	21	and	and	CCONJ
fcis-5209	141	22	g	g	PROPN
fcis-5209	141	23	hu	hu	PROPN
fcis-5209	141	24	,	,	PUNCT
fcis-5209	141	25	“	"	PUNCT
fcis-5209	141	26	pretraining	pretraine	VERB
fcis-5209	141	27	with	with	ADP
fcis-5209	141	28	whole	whole	ADJ
fcis-5209	141	29	word	word	NOUN
fcis-5209	141	30	masking	mask	VERB
fcis-5209	141	31	for	for	ADP
fcis-5209	141	32	chinese	chinese	ADJ
fcis-5209	141	33	bert	bert	PROPN
fcis-5209	141	34	,	,	PUNCT
fcis-5209	141	35	”	"	PUNCT
fcis-5209	141	36	arxiv	arxiv	PROPN
fcis-5209	141	37	preprint	preprint	VERB
fcis-5209	141	38	arxiv:1906.08101	arxiv:1906.08101	PROPN
fcis-5209	141	39	,	,	PUNCT
fcis-5209	141	40	2019	2019	NUM
fcis-5209	141	41	.	.	PUNCT
fcis-5209	142	1	[	[	X
fcis-5209	142	2	20	20	NUM
fcis-5209	142	3	]	]	PUNCT
fcis-5209	142	4	cui	cui	PROPN
fcis-5209	142	5	y	y	PROPN
fcis-5209	142	6	,	,	PUNCT
fcis-5209	142	7	che	che	PROPN
fcis-5209	142	8	w	w	PROPN
fcis-5209	142	9	,	,	PUNCT
fcis-5209	142	10	liu	liu	PROPN
fcis-5209	142	11	t	t	PROPN
fcis-5209	142	12	,	,	PUNCT
fcis-5209	142	13	qin	qin	PROPN
fcis-5209	142	14	b	b	PROPN
fcis-5209	142	15	,	,	PUNCT
fcis-5209	142	16	yang	yang	PROPN
fcis-5209	142	17	z.	z.	PROPN
fcis-5209	142	18	pre	pre	VERB
fcis-5209	142	19	-	-	VERB
fcis-5209	142	20	training	training	NOUN
fcis-5209	142	21	with	with	ADP
fcis-5209	142	22	whole	whole	ADJ
fcis-5209	142	23	word	word	NOUN
fcis-5209	142	24	masking	mask	VERB
fcis-5209	142	25	for	for	ADP
fcis-5209	142	26	chinese	chinese	ADJ
fcis-5209	142	27	bert[j	bert[j	NOUN
fcis-5209	142	28	]	]	PUNCT
fcis-5209	142	29	.	.	PUNCT
fcis-5209	142	30	ieee	ieee	PROPN
fcis-5209	142	31	/	/	SYM
fcis-5209	142	32	acm	acm	PROPN
fcis-5209	142	33	transactions	transaction	NOUN
fcis-5209	142	34	on	on	ADP
fcis-5209	142	35	audio	audio	NOUN
fcis-5209	142	36	,	,	PUNCT
fcis-5209	142	37	speech	speech	NOUN
fcis-5209	142	38	,	,	PUNCT
fcis-5209	142	39	and	and	CCONJ
fcis-5209	142	40	language	language	NOUN
fcis-5209	142	41	processing	processing	NOUN
fcis-5209	142	42	,	,	PUNCT
fcis-5209	142	43	2021	2021	NUM
fcis-5209	142	44	,	,	PUNCT
fcis-5209	142	45	29	29	NUM
fcis-5209	142	46	:	:	SYM
fcis-5209	142	47	3504	3504	NUM
fcis-5209	142	48	-	-	SYM
fcis-5209	142	49	3514	3514	NUM
fcis-5209	142	50	.	.	PUNCT
fcis-5209	143	1	[	[	X
fcis-5209	143	2	21	21	NUM
fcis-5209	143	3	]	]	X
fcis-5209	143	4	liu	liu	PROPN
fcis-5209	143	5	x	x	PROPN
fcis-5209	143	6	,	,	PUNCT
fcis-5209	143	7	chen	chen	PROPN
fcis-5209	143	8	q	q	PROPN
fcis-5209	143	9	,	,	PUNCT
fcis-5209	143	10	deng	deng	PROPN
fcis-5209	143	11	c	c	PROPN
fcis-5209	143	12	,	,	PUNCT
fcis-5209	143	13	zeng	zeng	PROPN
fcis-5209	143	14	h	h	PROPN
fcis-5209	143	15	,	,	PUNCT
fcis-5209	143	16	chen	chen	PROPN
fcis-5209	143	17	j	j	PROPN
fcis-5209	143	18	,	,	PUNCT
fcis-5209	143	19	li	li	PROPN
fcis-5209	143	20	d	d	PROPN
fcis-5209	143	21	,	,	PUNCT
fcis-5209	143	22	et	et	PROPN
fcis-5209	143	23	al	al	PROPN
fcis-5209	143	24	.	.	PROPN
fcis-5209	143	25	lcqmc	lcqmc	PROPN
fcis-5209	143	26	:	:	PUNCT
fcis-5209	143	27	a	a	DET
fcis-5209	143	28	large	large	ADJ
fcis-5209	143	29	-	-	PUNCT
fcis-5209	143	30	scale	scale	NOUN
fcis-5209	143	31	chinese	chinese	ADJ
fcis-5209	143	32	question	question	NOUN
fcis-5209	143	33	matching	match	VERB
fcis-5209	143	34	corpus[c]//proceedings	corpus[c]//proceeding	NOUN
fcis-5209	143	35	of	of	ADP
fcis-5209	143	36	the	the	DET
fcis-5209	143	37	27th	27th	ADJ
fcis-5209	143	38	international	international	ADJ
fcis-5209	143	39	conference	conference	NOUN
fcis-5209	143	40	on	on	ADP
fcis-5209	143	41	computational	computational	ADJ
fcis-5209	143	42	linguistics	linguistic	NOUN
fcis-5209	143	43	.	.	PUNCT
fcis-5209	143	44	2018	2018	NUM
fcis-5209	143	45	:	:	PUNCT
fcis-5209	143	46	1952	1952	NUM
fcis-5209	143	47	-	-	SYM
fcis-5209	143	48	1962	1962	NUM
fcis-5209	143	49	.	.	PUNCT
fcis-5209	144	1	[	[	X
fcis-5209	144	2	22	22	NUM
fcis-5209	144	3	]	]	X
fcis-5209	144	4	zhang	zhang	PROPN
fcis-5209	144	5	wenhui	wenhui	PROPN
fcis-5209	144	6	,	,	PUNCT
fcis-5209	144	7	wang	wang	PROPN
fcis-5209	144	8	meiling	meiling	PROPN
fcis-5209	144	9	,	,	PUNCT
fcis-5209	144	10	hou	hou	PROPN
fcis-5209	144	11	zhirong	zhirong	PROPN
fcis-5209	144	12	.	.	PUNCT
fcis-5209	145	1	a	a	DET
fcis-5209	145	2	short	short	ADJ
fcis-5209	145	3	text	text	NOUN
fcis-5209	145	4	matching	matching	NOUN
fcis-5209	145	5	model	model	NOUN
fcis-5209	145	6	incorporating	incorporate	VERB
fcis-5209	145	7	contextual	contextual	ADJ
fcis-5209	145	8	semantic	semantic	ADJ
fcis-5209	145	9	differences	difference	NOUN
fcis-5209	145	10	[	[	X
fcis-5209	145	11	j	j	X
fcis-5209	145	12	/	/	SYM
fcis-5209	145	13	ol	ol	PROPN
fcis-5209	145	14	]	]	PUNCT
fcis-5209	145	15	.	.	PUNCT
fcis-5209	146	1	journal	journal	PROPN
fcis-5209	146	2	of	of	ADP
fcis-5209	146	3	peking	peking	PROPN
fcis-5209	146	4	university	university	PROPN
fcis-5209	146	5	(	(	PUNCT
fcis-5209	146	6	natural	natural	ADJ
fcis-5209	146	7	science	science	NOUN
fcis-5209	146	8	edition	edition	NOUN
fcis-5209	146	9	)	)	PUNCT
fcis-5209	147	1	https://doi.org/10.13209/j.0479-8023.2022.071	https://doi.org/10.13209/j.0479-8023.2022.071	PROPN
fcis-5209	147	2	.	.	PUNCT