id	sid	tid	token	lemma	pos
fcis-2970	1	1	frontiers	frontier	NOUN
fcis-2970	1	2	in	in	ADP
fcis-2970	1	3	computing	computing	NOUN
fcis-2970	1	4	and	and	CCONJ
fcis-2970	1	5	intelligent	intelligent	ADJ
fcis-2970	1	6	systems	system	NOUN
fcis-2970	1	7	issn	issn	VERB
fcis-2970	1	8	:	:	PUNCT
fcis-2970	1	9	2832	2832	NUM
fcis-2970	1	10	-	-	SYM
fcis-2970	1	11	6024	6024	NUM
fcis-2970	1	12	|	|	NOUN
fcis-2970	1	13	vol	vol	NOUN
fcis-2970	1	14	.	.	PROPN
fcis-2970	2	1	2	2	NUM
fcis-2970	2	2	,	,	PUNCT
fcis-2970	2	3	no	no	INTJ
fcis-2970	2	4	.	.	NOUN
fcis-2970	2	5	1	1	NUM
fcis-2970	2	6	,	,	PUNCT
fcis-2970	2	7	2022	2022	NUM
fcis-2970	2	8	70	70	NUM
fcis-2970	2	9	chinese	chinese	ADJ
fcis-2970	2	10	text	text	NOUN
fcis-2970	2	11	sentiment	sentiment	NOUN
fcis-2970	2	12	classification	classification	NOUN
fcis-2970	2	13	based	base	VERB
fcis-2970	2	14	on	on	ADP
fcis-2970	2	15	ernie	ernie	PROPN
fcis-2970	2	16	and	and	CCONJ
fcis-2970	2	17	bilstm	bilstm	NOUN
fcis-2970	2	18	-	-	PUNCT
fcis-2970	2	19	at	at	ADP
fcis-2970	2	20	jianrong	jianrong	PROPN
fcis-2970	2	21	wang	wang	PROPN
fcis-2970	2	22	,	,	PUNCT
fcis-2970	2	23	naiyi	naiyi	PROPN
fcis-2970	2	24	li	li	PROPN
fcis-2970	2	25	*	*	PROPN
fcis-2970	2	26	school	school	NOUN
fcis-2970	2	27	of	of	ADP
fcis-2970	2	28	mathematics	mathematic	NOUN
fcis-2970	2	29	and	and	CCONJ
fcis-2970	2	30	computer	computer	NOUN
fcis-2970	2	31	science	science	NOUN
fcis-2970	2	32	,	,	PUNCT
fcis-2970	2	33	guangdong	guangdong	PROPN
fcis-2970	2	34	ocean	ocean	PROPN
fcis-2970	2	35	university	university	PROPN
fcis-2970	2	36	,	,	PUNCT
fcis-2970	2	37	zhanjiang	zhanjiang	PROPN
fcis-2970	2	38	,	,	PUNCT
fcis-2970	2	39	guangdong	guangdong	PROPN
fcis-2970	2	40	,	,	PUNCT
fcis-2970	2	41	china	china	PROPN
fcis-2970	2	42	*	*	PUNCT
fcis-2970	2	43	corresponding	correspond	VERB
fcis-2970	2	44	author	author	NOUN
fcis-2970	2	45	:	:	PUNCT
fcis-2970	2	46	naiyi	naiyi	PROPN
fcis-2970	2	47	li	li	PROPN
fcis-2970	2	48	(	(	PUNCT
fcis-2970	2	49	email	email	NOUN
fcis-2970	2	50	:	:	PUNCT
fcis-2970	2	51	linaiyi1979@163.com	linaiyi1979@163.com	X
fcis-2970	2	52	)	)	PUNCT
fcis-2970	2	53	abstract	abstract	NOUN
fcis-2970	2	54	:	:	PUNCT
fcis-2970	2	55	chinese	chinese	ADJ
fcis-2970	2	56	text	text	NOUN
fcis-2970	2	57	sentiment	sentiment	NOUN
fcis-2970	2	58	classification	classification	NOUN
fcis-2970	2	59	is	be	AUX
fcis-2970	2	60	a	a	DET
fcis-2970	2	61	sub	sub	NOUN
fcis-2970	2	62	-	-	NOUN
fcis-2970	2	63	task	task	NOUN
fcis-2970	2	64	of	of	ADP
fcis-2970	2	65	natural	natural	ADJ
fcis-2970	2	66	language	language	NOUN
fcis-2970	2	67	processing	processing	NOUN
fcis-2970	2	68	.	.	PUNCT
fcis-2970	3	1	however	however	ADV
fcis-2970	3	2	,	,	PUNCT
fcis-2970	3	3	when	when	SCONJ
fcis-2970	3	4	text	text	NOUN
fcis-2970	3	5	representation	representation	NOUN
fcis-2970	3	6	is	be	AUX
fcis-2970	3	7	carried	carry	VERB
fcis-2970	3	8	out	out	ADP
fcis-2970	3	9	,	,	PUNCT
fcis-2970	3	10	the	the	DET
fcis-2970	3	11	polysemy	polysemy	NOUN
fcis-2970	3	12	of	of	ADP
fcis-2970	3	13	a	a	DET
fcis-2970	3	14	word	word	NOUN
fcis-2970	3	15	can	can	AUX
fcis-2970	3	16	not	not	PART
fcis-2970	3	17	be	be	AUX
fcis-2970	3	18	processed	process	VERB
fcis-2970	3	19	when	when	SCONJ
fcis-2970	3	20	using	use	VERB
fcis-2970	3	21	the	the	DET
fcis-2970	3	22	traditional	traditional	ADJ
fcis-2970	3	23	language	language	NOUN
fcis-2970	3	24	model	model	NOUN
fcis-2970	3	25	to	to	PART
fcis-2970	3	26	construct	construct	VERB
fcis-2970	3	27	the	the	DET
fcis-2970	3	28	word	word	NOUN
fcis-2970	3	29	vector	vector	NOUN
fcis-2970	3	30	,	,	PUNCT
fcis-2970	3	31	and	and	CCONJ
fcis-2970	3	32	the	the	DET
fcis-2970	3	33	long	long	ADJ
fcis-2970	3	34	-	-	PUNCT
fcis-2970	3	35	distance	distance	NOUN
fcis-2970	3	36	text	text	NOUN
fcis-2970	3	37	information	information	NOUN
fcis-2970	3	38	can	can	AUX
fcis-2970	3	39	not	not	PART
fcis-2970	3	40	be	be	AUX
fcis-2970	3	41	fully	fully	ADV
fcis-2970	3	42	extracted	extract	VERB
fcis-2970	3	43	when	when	SCONJ
fcis-2970	3	44	extracting	extract	VERB
fcis-2970	3	45	text	text	NOUN
fcis-2970	3	46	features	feature	NOUN
fcis-2970	3	47	.	.	PUNCT
fcis-2970	4	1	to	to	PART
fcis-2970	4	2	solve	solve	VERB
fcis-2970	4	3	this	this	DET
fcis-2970	4	4	problem	problem	NOUN
fcis-2970	4	5	,	,	PUNCT
fcis-2970	4	6	this	this	DET
fcis-2970	4	7	paper	paper	NOUN
fcis-2970	4	8	proposes	propose	VERB
fcis-2970	4	9	a	a	DET
fcis-2970	4	10	text	text	NOUN
fcis-2970	4	11	sentiment	sentiment	NOUN
fcis-2970	4	12	classification	classification	NOUN
fcis-2970	4	13	model	model	NOUN
fcis-2970	4	14	combining	combine	VERB
fcis-2970	4	15	ernie	ernie	PROPN
fcis-2970	4	16	and	and	CCONJ
fcis-2970	4	17	bilstm	bilstm	NOUN
fcis-2970	4	18	-	-	PUNCT
fcis-2970	4	19	at	at	NOUN
fcis-2970	4	20	.	.	PUNCT
fcis-2970	5	1	first	first	ADV
fcis-2970	5	2	,	,	PUNCT
fcis-2970	5	3	the	the	DET
fcis-2970	5	4	pre	pre	ADJ
fcis-2970	5	5	-	-	ADJ
fcis-2970	5	6	training	training	ADJ
fcis-2970	5	7	model	model	NOUN
fcis-2970	5	8	ernie	ernie	PROPN
fcis-2970	5	9	is	be	AUX
fcis-2970	5	10	used	use	VERB
fcis-2970	5	11	to	to	PART
fcis-2970	5	12	obtain	obtain	VERB
fcis-2970	5	13	the	the	DET
fcis-2970	5	14	word	word	NOUN
fcis-2970	5	15	vector	vector	NOUN
fcis-2970	5	16	representation	representation	NOUN
fcis-2970	5	17	of	of	ADP
fcis-2970	5	18	the	the	DET
fcis-2970	5	19	fused	fuse	VERB
fcis-2970	5	20	statement	statement	NOUN
fcis-2970	5	21	context	context	NOUN
fcis-2970	5	22	.	.	PUNCT
fcis-2970	6	1	then	then	ADV
fcis-2970	6	2	,	,	PUNCT
fcis-2970	6	3	the	the	DET
fcis-2970	6	4	bidirectional	bidirectional	ADJ
fcis-2970	6	5	long	long	ADJ
fcis-2970	6	6	-	-	PUNCT
fcis-2970	6	7	short	short	ADJ
fcis-2970	6	8	-	-	PUNCT
fcis-2970	6	9	term	term	NOUN
fcis-2970	6	10	memory	memory	NOUN
fcis-2970	6	11	neural	neural	ADJ
fcis-2970	6	12	network	network	NOUN
fcis-2970	6	13	is	be	AUX
fcis-2970	6	14	used	use	VERB
fcis-2970	6	15	to	to	PART
fcis-2970	6	16	extract	extract	VERB
fcis-2970	6	17	the	the	DET
fcis-2970	6	18	context	context	NOUN
fcis-2970	6	19	information	information	NOUN
fcis-2970	6	20	and	and	CCONJ
fcis-2970	6	21	depth	depth	NOUN
fcis-2970	6	22	semantic	semantic	ADJ
fcis-2970	6	23	information	information	NOUN
fcis-2970	6	24	of	of	ADP
fcis-2970	6	25	the	the	DET
fcis-2970	6	26	text	text	NOUN
fcis-2970	6	27	.	.	PUNCT
fcis-2970	7	1	then	then	ADV
fcis-2970	7	2	,	,	PUNCT
fcis-2970	7	3	the	the	DET
fcis-2970	7	4	attention	attention	NOUN
fcis-2970	7	5	mechanism	mechanism	NOUN
fcis-2970	7	6	is	be	AUX
fcis-2970	7	7	used	use	VERB
fcis-2970	7	8	to	to	PART
fcis-2970	7	9	assign	assign	VERB
fcis-2970	7	10	the	the	DET
fcis-2970	7	11	corresponding	correspond	VERB
fcis-2970	7	12	weights	weight	NOUN
fcis-2970	7	13	to	to	ADP
fcis-2970	7	14	the	the	DET
fcis-2970	7	15	hidden	hide	VERB
fcis-2970	7	16	layer	layer	NOUN
fcis-2970	7	17	vectors	vector	NOUN
fcis-2970	7	18	of	of	ADP
fcis-2970	7	19	each	each	DET
fcis-2970	7	20	time	time	NOUN
fcis-2970	7	21	step	step	NOUN
fcis-2970	7	22	output	output	NOUN
fcis-2970	7	23	by	by	ADP
fcis-2970	7	24	the	the	DET
fcis-2970	7	25	bilstm	bilstm	NOUN
fcis-2970	7	26	layer	layer	NOUN
fcis-2970	7	27	,	,	PUNCT
fcis-2970	7	28	and	and	CCONJ
fcis-2970	7	29	the	the	DET
fcis-2970	7	30	weighted	weight	VERB
fcis-2970	7	31	summation	summation	NOUN
fcis-2970	7	32	is	be	AUX
fcis-2970	7	33	integrated	integrate	VERB
fcis-2970	7	34	into	into	ADP
fcis-2970	7	35	the	the	DET
fcis-2970	7	36	sentence	sentence	NOUN
fcis-2970	7	37	features	feature	NOUN
fcis-2970	7	38	.	.	PUNCT
fcis-2970	8	1	finally	finally	ADV
fcis-2970	8	2	,	,	PUNCT
fcis-2970	8	3	the	the	DET
fcis-2970	8	4	softmax	softmax	NOUN
fcis-2970	8	5	function	function	NOUN
fcis-2970	8	6	is	be	AUX
fcis-2970	8	7	used	use	VERB
fcis-2970	8	8	to	to	PART
fcis-2970	8	9	calculate	calculate	VERB
fcis-2970	8	10	the	the	DET
fcis-2970	8	11	probability	probability	NOUN
fcis-2970	8	12	distribution	distribution	NOUN
fcis-2970	8	13	of	of	ADP
fcis-2970	8	14	the	the	DET
fcis-2970	8	15	emotional	emotional	ADJ
fcis-2970	8	16	category	category	NOUN
fcis-2970	8	17	of	of	ADP
fcis-2970	8	18	the	the	DET
fcis-2970	8	19	text	text	NOUN
fcis-2970	8	20	in	in	ADP
fcis-2970	8	21	the	the	DET
fcis-2970	8	22	output	output	NOUN
fcis-2970	8	23	layer	layer	NOUN
fcis-2970	8	24	.	.	PUNCT
fcis-2970	9	1	the	the	DET
fcis-2970	9	2	results	result	NOUN
fcis-2970	9	3	show	show	VERB
fcis-2970	9	4	that	that	SCONJ
fcis-2970	9	5	the	the	DET
fcis-2970	9	6	proposed	propose	VERB
fcis-2970	9	7	model	model	NOUN
fcis-2970	9	8	can	can	AUX
fcis-2970	9	9	achieve	achieve	VERB
fcis-2970	9	10	high	high	ADJ
fcis-2970	9	11	accuracy	accuracy	NOUN
fcis-2970	9	12	on	on	ADP
fcis-2970	9	13	both	both	DET
fcis-2970	9	14	hotel	hotel	NOUN
fcis-2970	9	15	reviews	review	NOUN
fcis-2970	9	16	and	and	CCONJ
fcis-2970	9	17	takeaway	takeaway	NOUN
fcis-2970	9	18	reviews	review	NOUN
fcis-2970	9	19	.	.	PUNCT
fcis-2970	10	1	based	base	VERB
fcis-2970	10	2	on	on	ADP
fcis-2970	10	3	the	the	DET
fcis-2970	10	4	pre	pre	ADJ
fcis-2970	10	5	-	-	ADJ
fcis-2970	10	6	training	training	ADJ
fcis-2970	10	7	model	model	NOUN
fcis-2970	10	8	,	,	PUNCT
fcis-2970	10	9	adding	add	VERB
fcis-2970	10	10	bidirectional	bidirectional	ADJ
fcis-2970	10	11	longterm	longterm	NOUN
fcis-2970	10	12	and	and	CCONJ
fcis-2970	10	13	short	short	ADJ
fcis-2970	10	14	-	-	PUNCT
fcis-2970	10	15	term	term	NOUN
fcis-2970	10	16	memory	memory	NOUN
fcis-2970	10	17	network	network	NOUN
fcis-2970	10	18	and	and	CCONJ
fcis-2970	10	19	attention	attention	NOUN
fcis-2970	10	20	mechanism	mechanism	NOUN
fcis-2970	10	21	is	be	AUX
fcis-2970	10	22	beneficial	beneficial	ADJ
fcis-2970	10	23	to	to	PART
fcis-2970	10	24	improve	improve	VERB
fcis-2970	10	25	the	the	DET
fcis-2970	10	26	classification	classification	NOUN
fcis-2970	10	27	effect	effect	NOUN
fcis-2970	10	28	of	of	ADP
fcis-2970	10	29	the	the	DET
fcis-2970	10	30	model	model	NOUN
fcis-2970	10	31	,	,	PUNCT
fcis-2970	10	32	and	and	CCONJ
fcis-2970	10	33	has	have	VERB
fcis-2970	10	34	certain	certain	ADJ
fcis-2970	10	35	practicability	practicability	NOUN
fcis-2970	10	36	in	in	ADP
fcis-2970	10	37	text	text	NOUN
fcis-2970	10	38	sentiment	sentiment	NOUN
fcis-2970	10	39	classification	classification	NOUN
fcis-2970	10	40	tasks	task	NOUN
fcis-2970	10	41	.	.	PUNCT
fcis-2970	11	1	keywords	keyword	NOUN
fcis-2970	11	2	:	:	PUNCT
fcis-2970	11	3	text	text	NOUN
fcis-2970	11	4	sentiment	sentiment	NOUN
fcis-2970	11	5	analysis	analysis	NOUN
fcis-2970	11	6	;	;	PUNCT
fcis-2970	11	7	pre	pre	ADJ
fcis-2970	11	8	-	-	ADJ
fcis-2970	11	9	training	training	ADJ
fcis-2970	11	10	model	model	NOUN
fcis-2970	11	11	;	;	PUNCT
fcis-2970	11	12	bidirectional	bidirectional	ADJ
fcis-2970	11	13	long	long	ADJ
fcis-2970	11	14	-	-	PUNCT
fcis-2970	11	15	term	term	NOUN
fcis-2970	11	16	and	and	CCONJ
fcis-2970	11	17	short	short	ADJ
fcis-2970	11	18	-	-	PUNCT
fcis-2970	11	19	term	term	NOUN
fcis-2970	11	20	memory	memory	NOUN
fcis-2970	11	21	network	network	NOUN
fcis-2970	11	22	;	;	PUNCT
fcis-2970	11	23	attention	attention	NOUN
fcis-2970	11	24	mechanism	mechanism	NOUN
fcis-2970	11	25	.	.	PUNCT
fcis-2970	12	1	1	1	X
fcis-2970	12	2	.	.	X
fcis-2970	12	3	introduction	introduction	NOUN
fcis-2970	12	4	with	with	ADP
fcis-2970	12	5	the	the	DET
fcis-2970	12	6	rapid	rapid	ADJ
fcis-2970	12	7	development	development	NOUN
fcis-2970	12	8	of	of	ADP
fcis-2970	12	9	information	information	NOUN
fcis-2970	12	10	technology	technology	NOUN
fcis-2970	12	11	and	and	CCONJ
fcis-2970	12	12	mobile	mobile	ADJ
fcis-2970	12	13	internet	internet	NOUN
fcis-2970	12	14	,	,	PUNCT
fcis-2970	12	15	china	china	PROPN
fcis-2970	12	16	has	have	AUX
fcis-2970	12	17	gradually	gradually	ADV
fcis-2970	12	18	stepped	step	VERB
fcis-2970	12	19	into	into	ADP
fcis-2970	12	20	the	the	DET
fcis-2970	12	21	comprehensive	comprehensive	ADJ
fcis-2970	12	22	internet	internet	NOUN
fcis-2970	12	23	era	era	NOUN
fcis-2970	12	24	.	.	PUNCT
fcis-2970	13	1	as	as	ADV
fcis-2970	13	2	early	early	ADV
fcis-2970	13	3	as	as	ADP
fcis-2970	13	4	the	the	DET
fcis-2970	13	5	49th	49th	ADJ
fcis-2970	13	6	statistical	statistical	ADJ
fcis-2970	13	7	report	report	NOUN
fcis-2970	13	8	on	on	ADP
fcis-2970	13	9	the	the	DET
fcis-2970	13	10	development	development	NOUN
fcis-2970	13	11	of	of	ADP
fcis-2970	13	12	internet	internet	NOUN
fcis-2970	13	13	in	in	ADP
fcis-2970	13	14	china	china	PROPN
fcis-2970	13	15	released	release	VERB
fcis-2970	13	16	by	by	ADP
fcis-2970	13	17	china	china	PROPN
fcis-2970	13	18	internet	internet	PROPN
fcis-2970	13	19	network	network	PROPN
fcis-2970	13	20	information	information	NOUN
fcis-2970	13	21	center	center	NOUN
fcis-2970	13	22	(	(	PUNCT
fcis-2970	13	23	cnnic	cnnic	ADJ
fcis-2970	13	24	)	)	PUNCT
fcis-2970	13	25	[	[	X
fcis-2970	13	26	1	1	NUM
fcis-2970	13	27	]	]	PUNCT
fcis-2970	13	28	,	,	PUNCT
fcis-2970	13	29	it	it	PRON
fcis-2970	13	30	was	be	AUX
fcis-2970	13	31	pointed	point	VERB
fcis-2970	13	32	out	out	ADP
fcis-2970	13	33	that	that	SCONJ
fcis-2970	13	34	by	by	ADP
fcis-2970	13	35	december	december	PROPN
fcis-2970	13	36	2021	2021	NUM
fcis-2970	13	37	,	,	PUNCT
fcis-2970	13	38	the	the	DET
fcis-2970	13	39	number	number	NOUN
fcis-2970	13	40	of	of	ADP
fcis-2970	13	41	chinese	chinese	ADJ
fcis-2970	13	42	netizens	netizen	NOUN
fcis-2970	13	43	had	have	AUX
fcis-2970	13	44	reached	reach	VERB
fcis-2970	13	45	1.032	1.032	NUM
fcis-2970	13	46	billion	billion	NUM
fcis-2970	13	47	.	.	PUNCT
fcis-2970	14	1	the	the	DET
fcis-2970	14	2	internet	internet	NOUN
fcis-2970	14	3	penetration	penetration	NOUN
fcis-2970	14	4	rate	rate	NOUN
fcis-2970	14	5	reached	reach	VERB
fcis-2970	14	6	73.0	73.0	NUM
fcis-2970	14	7	%	%	NOUN
fcis-2970	14	8	.	.	PUNCT
fcis-2970	15	1	users	user	NOUN
fcis-2970	15	2	are	be	AUX
fcis-2970	15	3	used	use	VERB
fcis-2970	15	4	to	to	ADP
fcis-2970	15	5	publishing	publish	VERB
fcis-2970	15	6	content	content	NOUN
fcis-2970	15	7	on	on	ADP
fcis-2970	15	8	the	the	DET
fcis-2970	15	9	internet	internet	NOUN
fcis-2970	15	10	,	,	PUNCT
fcis-2970	15	11	including	include	VERB
fcis-2970	15	12	opinions	opinion	NOUN
fcis-2970	15	13	on	on	ADP
fcis-2970	15	14	social	social	ADJ
fcis-2970	15	15	hot	hot	ADJ
fcis-2970	15	16	topics	topic	NOUN
fcis-2970	15	17	,	,	PUNCT
fcis-2970	15	18	national	national	ADJ
fcis-2970	15	19	policies	policy	NOUN
fcis-2970	15	20	,	,	PUNCT
fcis-2970	15	21	products	product	NOUN
fcis-2970	15	22	,	,	PUNCT
fcis-2970	15	23	services	service	NOUN
fcis-2970	15	24	,	,	PUNCT
fcis-2970	15	25	etc	etc	X
fcis-2970	15	26	.	.	X
fcis-2970	16	1	all	all	DET
fcis-2970	16	2	kinds	kind	NOUN
fcis-2970	16	3	of	of	ADP
fcis-2970	16	4	comments	comment	NOUN
fcis-2970	16	5	and	and	CCONJ
fcis-2970	16	6	opinions	opinion	NOUN
fcis-2970	16	7	are	be	AUX
fcis-2970	16	8	more	more	ADV
fcis-2970	16	9	or	or	CCONJ
fcis-2970	16	10	less	less	ADV
fcis-2970	16	11	attached	attach	VERB
fcis-2970	16	12	to	to	ADP
fcis-2970	16	13	users	user	NOUN
fcis-2970	16	14	'	'	PART
fcis-2970	16	15	emotional	emotional	ADJ
fcis-2970	16	16	attitudes	attitude	NOUN
fcis-2970	16	17	.	.	PUNCT
fcis-2970	17	1	quick	quick	ADJ
fcis-2970	17	2	and	and	CCONJ
fcis-2970	17	3	accurate	accurate	ADJ
fcis-2970	17	4	identification	identification	NOUN
fcis-2970	17	5	of	of	ADP
fcis-2970	17	6	emotion	emotion	NOUN
fcis-2970	17	7	types	type	NOUN
fcis-2970	17	8	plays	play	VERB
fcis-2970	17	9	an	an	DET
fcis-2970	17	10	important	important	ADJ
fcis-2970	17	11	role	role	NOUN
fcis-2970	17	12	in	in	ADP
fcis-2970	17	13	many	many	ADJ
fcis-2970	17	14	fields	field	NOUN
fcis-2970	17	15	.	.	PUNCT
fcis-2970	18	1	for	for	ADP
fcis-2970	18	2	example	example	NOUN
fcis-2970	18	3	,	,	PUNCT
fcis-2970	18	4	in	in	ADP
fcis-2970	18	5	terms	term	NOUN
fcis-2970	18	6	of	of	ADP
fcis-2970	18	7	public	public	ADJ
fcis-2970	18	8	opinion	opinion	NOUN
fcis-2970	18	9	,	,	PUNCT
fcis-2970	18	10	it	it	PRON
fcis-2970	18	11	is	be	AUX
fcis-2970	18	12	possible	possible	ADJ
fcis-2970	18	13	to	to	PART
fcis-2970	18	14	understand	understand	VERB
fcis-2970	18	15	public	public	ADJ
fcis-2970	18	16	opinions	opinion	NOUN
fcis-2970	18	17	by	by	ADP
fcis-2970	18	18	paying	pay	VERB
fcis-2970	18	19	attention	attention	NOUN
fcis-2970	18	20	to	to	ADP
fcis-2970	18	21	netizens	netizen	NOUN
fcis-2970	18	22	'	'	PART
fcis-2970	18	23	emotional	emotional	ADJ
fcis-2970	18	24	attitudes	attitude	NOUN
fcis-2970	18	25	toward	toward	ADP
fcis-2970	18	26	social	social	ADJ
fcis-2970	18	27	affairs	affair	NOUN
fcis-2970	18	28	,	,	PUNCT
fcis-2970	18	29	which	which	PRON
fcis-2970	18	30	can	can	AUX
fcis-2970	18	31	effectively	effectively	ADV
fcis-2970	18	32	prevent	prevent	VERB
fcis-2970	18	33	harmful	harmful	ADJ
fcis-2970	18	34	events	event	NOUN
fcis-2970	18	35	and	and	CCONJ
fcis-2970	18	36	facilitate	facilitate	VERB
fcis-2970	18	37	the	the	DET
fcis-2970	18	38	implementation	implementation	NOUN
fcis-2970	18	39	of	of	ADP
fcis-2970	18	40	policies	policy	NOUN
fcis-2970	18	41	.	.	PUNCT
fcis-2970	19	1	in	in	ADP
fcis-2970	19	2	the	the	DET
fcis-2970	19	3	business	business	NOUN
fcis-2970	19	4	service	service	NOUN
fcis-2970	19	5	,	,	PUNCT
fcis-2970	19	6	the	the	DET
fcis-2970	19	7	analysis	analysis	NOUN
fcis-2970	19	8	of	of	ADP
fcis-2970	19	9	user	user	NOUN
fcis-2970	19	10	evaluation	evaluation	NOUN
fcis-2970	19	11	can	can	AUX
fcis-2970	19	12	help	help	VERB
fcis-2970	19	13	improve	improve	VERB
fcis-2970	19	14	the	the	DET
fcis-2970	19	15	quality	quality	NOUN
fcis-2970	19	16	of	of	ADP
fcis-2970	19	17	goods	good	NOUN
fcis-2970	19	18	and	and	CCONJ
fcis-2970	19	19	services	service	NOUN
fcis-2970	19	20	,	,	PUNCT
fcis-2970	19	21	and	and	CCONJ
fcis-2970	19	22	do	do	VERB
fcis-2970	19	23	a	a	DET
fcis-2970	19	24	good	good	ADJ
fcis-2970	19	25	job	job	NOUN
fcis-2970	19	26	in	in	ADP
fcis-2970	19	27	customer	customer	NOUN
fcis-2970	19	28	relationship	relationship	NOUN
fcis-2970	19	29	management	management	NOUN
fcis-2970	19	30	;	;	PUNCT
fcis-2970	19	31	in	in	ADP
fcis-2970	19	32	the	the	DET
fcis-2970	19	33	aspect	aspect	NOUN
fcis-2970	19	34	of	of	ADP
fcis-2970	19	35	psychological	psychological	ADJ
fcis-2970	19	36	and	and	CCONJ
fcis-2970	19	37	emotional	emotional	ADJ
fcis-2970	19	38	counseling	counseling	NOUN
fcis-2970	19	39	,	,	PUNCT
fcis-2970	19	40	through	through	ADP
fcis-2970	19	41	the	the	DET
fcis-2970	19	42	analysis	analysis	NOUN
fcis-2970	19	43	of	of	ADP
fcis-2970	19	44	the	the	DET
fcis-2970	19	45	corpus	corpus	NOUN
fcis-2970	19	46	expression	expression	NOUN
fcis-2970	19	47	of	of	ADP
fcis-2970	19	48	the	the	DET
fcis-2970	19	49	object	object	NOUN
fcis-2970	19	50	,	,	PUNCT
fcis-2970	19	51	it	it	PRON
fcis-2970	19	52	can	can	AUX
fcis-2970	19	53	quickly	quickly	ADV
fcis-2970	19	54	identify	identify	VERB
fcis-2970	19	55	the	the	DET
fcis-2970	19	56	object	object	NOUN
fcis-2970	19	57	's	's	PART
fcis-2970	19	58	emotional	emotional	ADJ
fcis-2970	19	59	attitude	attitude	NOUN
fcis-2970	19	60	,	,	PUNCT
fcis-2970	19	61	and	and	CCONJ
fcis-2970	19	62	conduct	conduct	VERB
fcis-2970	19	63	targeted	target	VERB
fcis-2970	19	64	counseling	counseling	NOUN
fcis-2970	19	65	for	for	ADP
fcis-2970	19	66	users	user	NOUN
fcis-2970	19	67	with	with	ADP
fcis-2970	19	68	psychological	psychological	ADJ
fcis-2970	19	69	problems	problem	NOUN
fcis-2970	19	70	.	.	PUNCT
fcis-2970	20	1	nowadays	nowadays	ADV
fcis-2970	20	2	,	,	PUNCT
fcis-2970	20	3	according	accord	VERB
fcis-2970	20	4	to	to	ADP
fcis-2970	20	5	different	different	ADJ
fcis-2970	20	6	methods	method	NOUN
fcis-2970	20	7	used	use	VERB
fcis-2970	20	8	in	in	ADP
fcis-2970	20	9	classification	classification	NOUN
fcis-2970	20	10	,	,	PUNCT
fcis-2970	20	11	text	text	NOUN
fcis-2970	20	12	emotion	emotion	NOUN
fcis-2970	20	13	classification	classification	NOUN
fcis-2970	20	14	can	can	AUX
fcis-2970	20	15	be	be	AUX
fcis-2970	20	16	roughly	roughly	ADV
fcis-2970	20	17	divided	divide	VERB
fcis-2970	20	18	into	into	ADP
fcis-2970	20	19	three	three	NUM
fcis-2970	20	20	categories	category	NOUN
fcis-2970	20	21	,	,	PUNCT
fcis-2970	20	22	namely	namely	ADV
fcis-2970	20	23	emotion	emotion	VERB
fcis-2970	20	24	classification	classification	NOUN
fcis-2970	20	25	method	method	NOUN
fcis-2970	20	26	based	base	VERB
fcis-2970	20	27	on	on	ADP
fcis-2970	20	28	emotion	emotion	PROPN
fcis-2970	20	29	dictionary	dictionary	PROPN
fcis-2970	20	30	,	,	PUNCT
fcis-2970	20	31	emotion	emotion	NOUN
fcis-2970	20	32	classification	classification	NOUN
fcis-2970	20	33	method	method	NOUN
fcis-2970	20	34	based	base	VERB
fcis-2970	20	35	on	on	ADP
fcis-2970	20	36	traditional	traditional	ADJ
fcis-2970	20	37	machine	machine	NOUN
fcis-2970	20	38	learning	learning	NOUN
fcis-2970	20	39	,	,	PUNCT
fcis-2970	20	40	and	and	CCONJ
fcis-2970	20	41	emotion	emotion	NOUN
fcis-2970	20	42	classification	classification	NOUN
fcis-2970	20	43	method	method	NOUN
fcis-2970	20	44	based	base	VERB
fcis-2970	20	45	on	on	ADP
fcis-2970	20	46	deep	deep	ADJ
fcis-2970	20	47	learning	learning	NOUN
fcis-2970	21	1	[	[	X
fcis-2970	21	2	2	2	NUM
fcis-2970	21	3	]	]	PUNCT
fcis-2970	21	4	.	.	PUNCT
fcis-2970	22	1	the	the	DET
fcis-2970	22	2	text	text	NOUN
fcis-2970	22	3	emotion	emotion	NOUN
fcis-2970	22	4	classification	classification	NOUN
fcis-2970	22	5	method	method	NOUN
fcis-2970	22	6	based	base	VERB
fcis-2970	22	7	on	on	ADP
fcis-2970	22	8	the	the	DET
fcis-2970	22	9	emotion	emotion	NOUN
fcis-2970	22	10	dictionary	dictionary	NOUN
fcis-2970	22	11	is	be	AUX
fcis-2970	22	12	based	base	VERB
fcis-2970	22	13	on	on	ADP
fcis-2970	22	14	the	the	DET
fcis-2970	22	15	emotion	emotion	NOUN
fcis-2970	22	16	dictionary	dictionary	NOUN
fcis-2970	22	17	.	.	PUNCT
fcis-2970	23	1	the	the	DET
fcis-2970	23	2	pre	pre	ADJ
fcis-2970	23	3	-	-	ADJ
fcis-2970	23	4	processed	process	VERB
fcis-2970	23	5	text	text	NOUN
fcis-2970	23	6	data	datum	NOUN
fcis-2970	23	7	is	be	AUX
fcis-2970	23	8	used	use	VERB
fcis-2970	23	9	to	to	PART
fcis-2970	23	10	match	match	VERB
fcis-2970	23	11	the	the	DET
fcis-2970	23	12	emotion	emotion	NOUN
fcis-2970	23	13	words	word	NOUN
fcis-2970	23	14	and	and	CCONJ
fcis-2970	23	15	emotion	emotion	NOUN
fcis-2970	23	16	polarity	polarity	NOUN
fcis-2970	23	17	contained	contain	VERB
fcis-2970	23	18	in	in	ADP
fcis-2970	23	19	the	the	DET
fcis-2970	23	20	emotion	emotion	NOUN
fcis-2970	23	21	dictionary	dictionary	NOUN
fcis-2970	23	22	,	,	PUNCT
fcis-2970	23	23	and	and	CCONJ
fcis-2970	23	24	the	the	DET
fcis-2970	23	25	emotion	emotion	NOUN
fcis-2970	23	26	polarity	polarity	NOUN
fcis-2970	23	27	is	be	AUX
fcis-2970	23	28	classified	classify	VERB
fcis-2970	23	29	according	accord	VERB
fcis-2970	23	30	to	to	ADP
fcis-2970	23	31	different	different	ADJ
fcis-2970	23	32	granularity	granularity	NOUN
fcis-2970	23	33	.	.	PUNCT
fcis-2970	24	1	li	li	PROPN
fcis-2970	24	2	yuqing	yuqe	VERB
fcis-2970	24	3	[	[	X
fcis-2970	24	4	3	3	NUM
fcis-2970	24	5	]	]	PUNCT
fcis-2970	24	6	et	et	PROPN
fcis-2970	24	7	al	al	PROPN
fcis-2970	24	8	.	.	PROPN
fcis-2970	24	9	built	build	VERB
fcis-2970	24	10	a	a	DET
fcis-2970	24	11	bilingual	bilingual	ADJ
fcis-2970	24	12	multi	multi	ADJ
fcis-2970	24	13	-	-	ADJ
fcis-2970	24	14	class	class	ADJ
fcis-2970	24	15	emotion	emotion	NOUN
fcis-2970	24	16	dictionary	dictionary	NOUN
fcis-2970	24	17	based	base	VERB
fcis-2970	24	18	on	on	ADP
fcis-2970	24	19	the	the	DET
fcis-2970	24	20	bilingual	bilingual	ADJ
fcis-2970	24	21	dictionary	dictionary	PROPN
fcis-2970	24	22	method	method	NOUN
fcis-2970	24	23	and	and	CCONJ
fcis-2970	24	24	conducted	conduct	VERB
fcis-2970	24	25	a	a	DET
fcis-2970	24	26	multi	multi	ADJ
fcis-2970	24	27	-	-	ADJ
fcis-2970	24	28	class	class	ADJ
fcis-2970	24	29	emotion	emotion	NOUN
fcis-2970	24	30	classification	classification	NOUN
fcis-2970	24	31	experiment	experiment	NOUN
fcis-2970	24	32	,	,	PUNCT
fcis-2970	24	33	and	and	CCONJ
fcis-2970	24	34	the	the	DET
fcis-2970	24	35	model	model	NOUN
fcis-2970	24	36	experiment	experiment	NOUN
fcis-2970	24	37	achieved	achieve	VERB
fcis-2970	24	38	good	good	ADJ
fcis-2970	24	39	results	result	NOUN
fcis-2970	24	40	.	.	PUNCT
fcis-2970	25	1	however	however	ADV
fcis-2970	25	2	,	,	PUNCT
fcis-2970	25	3	this	this	DET
fcis-2970	25	4	classification	classification	NOUN
fcis-2970	25	5	method	method	NOUN
fcis-2970	25	6	is	be	AUX
fcis-2970	25	7	limited	limit	VERB
fcis-2970	25	8	by	by	ADP
fcis-2970	25	9	the	the	DET
fcis-2970	25	10	scale	scale	NOUN
fcis-2970	25	11	of	of	ADP
fcis-2970	25	12	the	the	DET
fcis-2970	25	13	sentiment	sentiment	NOUN
fcis-2970	25	14	dictionary	dictionary	NOUN
fcis-2970	25	15	,	,	PUNCT
fcis-2970	25	16	so	so	CCONJ
fcis-2970	25	17	the	the	DET
fcis-2970	25	18	dictionary	dictionary	ADJ
fcis-2970	25	19	base	base	NOUN
fcis-2970	25	20	should	should	AUX
fcis-2970	25	21	be	be	AUX
fcis-2970	25	22	continuously	continuously	ADV
fcis-2970	25	23	expanded	expand	VERB
fcis-2970	25	24	.	.	PUNCT
fcis-2970	26	1	the	the	DET
fcis-2970	26	2	sentiment	sentiment	NOUN
fcis-2970	26	3	analysis	analysis	NOUN
fcis-2970	26	4	method	method	NOUN
fcis-2970	26	5	based	base	VERB
fcis-2970	26	6	on	on	ADP
fcis-2970	26	7	machine	machine	NOUN
fcis-2970	26	8	learning	learning	NOUN
fcis-2970	26	9	needs	need	VERB
fcis-2970	26	10	to	to	PART
fcis-2970	26	11	select	select	VERB
fcis-2970	26	12	the	the	DET
fcis-2970	26	13	classification	classification	NOUN
fcis-2970	26	14	algorithm	algorithm	NOUN
fcis-2970	26	15	,	,	PUNCT
fcis-2970	26	16	obtain	obtain	VERB
fcis-2970	26	17	the	the	DET
fcis-2970	26	18	model	model	NOUN
fcis-2970	26	19	parameters	parameter	NOUN
fcis-2970	26	20	through	through	ADP
fcis-2970	26	21	data	datum	NOUN
fcis-2970	26	22	training	training	NOUN
fcis-2970	26	23	,	,	PUNCT
fcis-2970	26	24	and	and	CCONJ
fcis-2970	26	25	then	then	ADV
fcis-2970	26	26	use	use	VERB
fcis-2970	26	27	the	the	DET
fcis-2970	26	28	trained	train	VERB
fcis-2970	26	29	model	model	NOUN
fcis-2970	26	30	to	to	PART
fcis-2970	26	31	predict	predict	VERB
fcis-2970	26	32	the	the	DET
fcis-2970	26	33	results	result	NOUN
fcis-2970	26	34	.	.	PUNCT
fcis-2970	27	1	tang	tang	PROPN
fcis-2970	27	2	huifeng	huifeng	NOUN
fcis-2970	28	1	[	[	X
fcis-2970	28	2	4	4	NUM
fcis-2970	28	3	]	]	PUNCT
fcis-2970	28	4	et	et	PROPN
fcis-2970	28	5	al	al	PROPN
fcis-2970	28	6	.	.	PROPN
fcis-2970	28	7	used	use	VERB
fcis-2970	28	8	common	common	ADJ
fcis-2970	28	9	machine	machine	NOUN
fcis-2970	28	10	learning	learning	NOUN
fcis-2970	28	11	methods	method	NOUN
fcis-2970	28	12	(	(	PUNCT
fcis-2970	28	13	svm	svm	PROPN
fcis-2970	28	14	,	,	PUNCT
fcis-2970	28	15	knn	knn	PROPN
fcis-2970	28	16	,	,	PUNCT
fcis-2970	28	17	etc	etc	X
fcis-2970	28	18	.	.	X
fcis-2970	28	19	)	)	PUNCT
fcis-2970	28	20	to	to	PART
fcis-2970	28	21	conduct	conduct	VERB
fcis-2970	28	22	the	the	DET
fcis-2970	28	23	experiment	experiment	NOUN
fcis-2970	28	24	of	of	ADP
fcis-2970	28	25	chinese	chinese	ADJ
fcis-2970	28	26	text	text	NOUN
fcis-2970	28	27	sentiment	sentiment	NOUN
fcis-2970	28	28	classification	classification	NOUN
fcis-2970	28	29	,	,	PUNCT
fcis-2970	28	30	and	and	CCONJ
fcis-2970	28	31	to	to	PART
fcis-2970	28	32	select	select	VERB
fcis-2970	28	33	appropriate	appropriate	ADJ
fcis-2970	28	34	feature	feature	NOUN
fcis-2970	28	35	representation	representation	NOUN
fcis-2970	28	36	and	and	CCONJ
fcis-2970	28	37	selection	selection	NOUN
fcis-2970	28	38	methods	method	NOUN
fcis-2970	28	39	,	,	PUNCT
fcis-2970	28	40	svm	svm	NOUN
fcis-2970	28	41	can	can	AUX
fcis-2970	28	42	achieve	achieve	VERB
fcis-2970	28	43	the	the	DET
fcis-2970	28	44	optimal	optimal	ADJ
fcis-2970	28	45	classification	classification	NOUN
fcis-2970	28	46	effect	effect	NOUN
fcis-2970	28	47	.	.	PUNCT
fcis-2970	29	1	however	however	ADV
fcis-2970	29	2	,	,	PUNCT
fcis-2970	29	3	this	this	DET
fcis-2970	29	4	classification	classification	NOUN
fcis-2970	29	5	method	method	NOUN
fcis-2970	29	6	fails	fail	VERB
fcis-2970	29	7	to	to	PART
fcis-2970	29	8	fully	fully	ADV
fcis-2970	29	9	consider	consider	VERB
fcis-2970	29	10	the	the	DET
fcis-2970	29	11	position	position	NOUN
fcis-2970	29	12	information	information	NOUN
fcis-2970	29	13	of	of	ADP
fcis-2970	29	14	the	the	DET
fcis-2970	29	15	words	word	NOUN
fcis-2970	29	16	in	in	ADP
fcis-2970	29	17	the	the	DET
fcis-2970	29	18	sentence	sentence	NOUN
fcis-2970	29	19	,	,	PUNCT
fcis-2970	29	20	which	which	PRON
fcis-2970	29	21	will	will	AUX
fcis-2970	29	22	lose	lose	VERB
fcis-2970	29	23	the	the	DET
fcis-2970	29	24	text	text	NOUN
fcis-2970	29	25	context	context	NOUN
fcis-2970	29	26	information	information	NOUN
fcis-2970	29	27	and	and	CCONJ
fcis-2970	29	28	affect	affect	VERB
fcis-2970	29	29	the	the	DET
fcis-2970	29	30	classification	classification	NOUN
fcis-2970	29	31	effect	effect	NOUN
fcis-2970	29	32	.	.	PUNCT
fcis-2970	30	1	since	since	SCONJ
fcis-2970	30	2	the	the	DET
fcis-2970	30	3	deep	deep	ADJ
fcis-2970	30	4	learning	learning	NOUN
fcis-2970	30	5	-	-	PUNCT
fcis-2970	30	6	based	base	VERB
fcis-2970	30	7	sentiment	sentiment	NOUN
fcis-2970	30	8	classification	classification	NOUN
fcis-2970	30	9	method	method	NOUN
fcis-2970	30	10	will	will	AUX
fcis-2970	30	11	make	make	VERB
fcis-2970	30	12	use	use	NOUN
fcis-2970	30	13	of	of	ADP
fcis-2970	30	14	the	the	DET
fcis-2970	30	15	word	word	NOUN
fcis-2970	30	16	order	order	NOUN
fcis-2970	30	17	information	information	NOUN
fcis-2970	30	18	of	of	ADP
fcis-2970	30	19	the	the	DET
fcis-2970	30	20	text	text	NOUN
fcis-2970	30	21	,	,	PUNCT
fcis-2970	30	22	extract	extract	VERB
fcis-2970	30	23	the	the	DET
fcis-2970	30	24	semantic	semantic	ADJ
fcis-2970	30	25	information	information	NOUN
fcis-2970	30	26	of	of	ADP
fcis-2970	30	27	the	the	DET
fcis-2970	30	28	words	word	NOUN
fcis-2970	30	29	,	,	PUNCT
fcis-2970	30	30	and	and	CCONJ
fcis-2970	30	31	take	take	VERB
fcis-2970	30	32	full	full	ADJ
fcis-2970	30	33	account	account	NOUN
fcis-2970	30	34	of	of	ADP
fcis-2970	30	35	the	the	DET
fcis-2970	30	36	advantages	advantage	NOUN
fcis-2970	30	37	of	of	ADP
fcis-2970	30	38	contextual	contextual	ADJ
fcis-2970	30	39	information	information	NOUN
fcis-2970	30	40	[	[	X
fcis-2970	30	41	2	2	NUM
fcis-2970	30	42	]	]	PUNCT
fcis-2970	30	43	,	,	PUNCT
fcis-2970	30	44	many	many	ADJ
fcis-2970	30	45	scholars	scholar	NOUN
fcis-2970	30	46	have	have	AUX
fcis-2970	30	47	carried	carry	VERB
fcis-2970	30	48	out	out	ADP
fcis-2970	30	49	relevant	relevant	ADJ
fcis-2970	30	50	studies	study	NOUN
fcis-2970	30	51	.	.	PUNCT
fcis-2970	31	1	among	among	ADP
fcis-2970	31	2	the	the	DET
fcis-2970	31	3	deep	deep	ADJ
fcis-2970	31	4	learning	learning	NOUN
fcis-2970	31	5	-	-	PUNCT
fcis-2970	31	6	based	base	VERB
fcis-2970	31	7	emotion	emotion	NOUN
fcis-2970	31	8	classification	classification	NOUN
fcis-2970	31	9	methods	method	NOUN
fcis-2970	31	10	,	,	PUNCT
fcis-2970	31	11	some	some	DET
fcis-2970	31	12	scholars	scholar	NOUN
fcis-2970	31	13	use	use	VERB
fcis-2970	31	14	a	a	DET
fcis-2970	31	15	single	single	ADJ
fcis-2970	31	16	neural	neural	ADJ
fcis-2970	31	17	network	network	NOUN
fcis-2970	31	18	for	for	ADP
fcis-2970	31	19	emotion	emotion	NOUN
fcis-2970	31	20	classification	classification	NOUN
fcis-2970	31	21	.	.	PUNCT
fcis-2970	32	1	for	for	ADP
fcis-2970	32	2	example	example	NOUN
fcis-2970	32	3	,	,	PUNCT
fcis-2970	32	4	jelodar	jelodar	NOUN
fcis-2970	33	1	[	[	X
fcis-2970	33	2	5	5	NUM
fcis-2970	33	3	]	]	PUNCT
fcis-2970	33	4	et	et	PROPN
fcis-2970	33	5	al	al	PROPN
fcis-2970	33	6	.	.	PROPN
fcis-2970	33	7	used	use	VERB
fcis-2970	33	8	the	the	DET
fcis-2970	33	9	lstm	lstm	PROPN
fcis-2970	33	10	model	model	NOUN
fcis-2970	33	11	when	when	SCONJ
fcis-2970	33	12	analyzing	analyze	VERB
fcis-2970	33	13	comments	comment	NOUN
fcis-2970	33	14	on	on	ADP
fcis-2970	33	15	covid-19	covid-19	PROPN
fcis-2970	33	16	,	,	PUNCT
fcis-2970	33	17	and	and	CCONJ
fcis-2970	33	18	the	the	DET
fcis-2970	33	19	experimental	experimental	ADJ
fcis-2970	33	20	results	result	NOUN
fcis-2970	33	21	can	can	AUX
fcis-2970	33	22	provide	provide	VERB
fcis-2970	33	23	data	data	NOUN
fcis-2970	33	24	support	support	NOUN
fcis-2970	33	25	for	for	ADP
fcis-2970	33	26	relevant	relevant	ADJ
fcis-2970	33	27	decisions	decision	NOUN
fcis-2970	33	28	.	.	PUNCT
fcis-2970	34	1	some	some	DET
fcis-2970	34	2	scholars	scholar	NOUN
fcis-2970	34	3	have	have	AUX
fcis-2970	34	4	considered	consider	VERB
fcis-2970	34	5	the	the	DET
fcis-2970	34	6	advantages	advantage	NOUN
fcis-2970	34	7	and	and	CCONJ
fcis-2970	34	8	disadvantages	disadvantage	NOUN
fcis-2970	34	9	of	of	ADP
fcis-2970	34	10	different	different	ADJ
fcis-2970	34	11	neural	neural	ADJ
fcis-2970	34	12	network	network	NOUN
fcis-2970	34	13	models	model	NOUN
fcis-2970	34	14	,	,	PUNCT
fcis-2970	34	15	and	and	CCONJ
fcis-2970	34	16	then	then	ADV
fcis-2970	34	17	improved	improve	VERB
fcis-2970	34	18	and	and	CCONJ
fcis-2970	34	19	mixed	mix	VERB
fcis-2970	34	20	the	the	DET
fcis-2970	34	21	models	model	NOUN
fcis-2970	34	22	,	,	PUNCT
fcis-2970	34	23	and	and	CCONJ
fcis-2970	34	24	achieved	achieve	VERB
fcis-2970	34	25	good	good	ADJ
fcis-2970	34	26	experimental	experimental	ADJ
fcis-2970	34	27	results	result	NOUN
fcis-2970	34	28	in	in	ADP
fcis-2970	34	29	the	the	DET
fcis-2970	34	30	task	task	NOUN
fcis-2970	34	31	of	of	ADP
fcis-2970	34	32	emotion	emotion	NOUN
fcis-2970	34	33	classification	classification	NOUN
fcis-2970	34	34	.	.	PUNCT
fcis-2970	35	1	for	for	ADP
fcis-2970	35	2	example	example	NOUN
fcis-2970	35	3	,	,	PUNCT
fcis-2970	35	4	luo	luo	PROPN
fcis-2970	35	5	fan	fan	PROPN
fcis-2970	35	6	et	et	PROPN
fcis-2970	35	7	al	al	PROPN
fcis-2970	35	8	.	.	PUNCT
fcis-2970	36	1	[	[	X
fcis-2970	36	2	6	6	NUM
fcis-2970	36	3	]	]	PUNCT
fcis-2970	36	4	proposed	propose	VERB
fcis-2970	36	5	a	a	DET
fcis-2970	36	6	multi	multi	ADJ
fcis-2970	36	7	-	-	ADJ
fcis-2970	36	8	layer	layer	ADJ
fcis-2970	36	9	network	network	NOUN
fcis-2970	36	10	h	h	NOUN
fcis-2970	36	11	-	-	PUNCT
fcis-2970	36	12	rnn	rnn	PROPN
fcis-2970	36	13	-	-	PUNCT
fcis-2970	36	14	cnn	cnn	PROPN
fcis-2970	36	15	,	,	PUNCT
fcis-2970	36	16	combined	combine	VERB
fcis-2970	36	17	71	71	NUM
fcis-2970	36	18	the	the	DET
fcis-2970	36	19	advantages	advantage	NOUN
fcis-2970	36	20	of	of	ADP
fcis-2970	36	21	the	the	DET
fcis-2970	36	22	two	two	NUM
fcis-2970	36	23	models	model	NOUN
fcis-2970	36	24	,	,	PUNCT
fcis-2970	36	25	used	use	VERB
fcis-2970	36	26	rnn	rnn	NOUN
fcis-2970	36	27	to	to	ADP
fcis-2970	36	28	model	model	NOUN
fcis-2970	36	29	text	text	NOUN
fcis-2970	36	30	sequences	sequence	NOUN
fcis-2970	36	31	,	,	PUNCT
fcis-2970	36	32	and	and	CCONJ
fcis-2970	36	33	used	use	VERB
fcis-2970	36	34	cnn	cnn	PROPN
fcis-2970	36	35	to	to	PART
fcis-2970	36	36	identify	identify	VERB
fcis-2970	36	37	information	information	NOUN
fcis-2970	36	38	across	across	ADP
fcis-2970	36	39	sentences	sentence	NOUN
fcis-2970	36	40	.	.	PUNCT
fcis-2970	37	1	the	the	DET
fcis-2970	37	2	model	model	NOUN
fcis-2970	37	3	has	have	AUX
fcis-2970	37	4	obtained	obtain	VERB
fcis-2970	37	5	good	good	ADJ
fcis-2970	37	6	experimental	experimental	ADJ
fcis-2970	37	7	results	result	NOUN
fcis-2970	37	8	.	.	PUNCT
fcis-2970	38	1	after	after	ADP
fcis-2970	38	2	seeing	see	VERB
fcis-2970	38	3	important	important	ADJ
fcis-2970	38	4	achievements	achievement	NOUN
fcis-2970	38	5	in	in	ADP
fcis-2970	38	6	the	the	DET
fcis-2970	38	7	application	application	NOUN
fcis-2970	38	8	of	of	ADP
fcis-2970	38	9	attention	attention	NOUN
fcis-2970	38	10	mechanisms	mechanism	NOUN
fcis-2970	38	11	in	in	ADP
fcis-2970	38	12	the	the	DET
fcis-2970	38	13	field	field	NOUN
fcis-2970	38	14	of	of	ADP
fcis-2970	38	15	visual	visual	ADJ
fcis-2970	38	16	images	image	NOUN
fcis-2970	38	17	,	,	PUNCT
fcis-2970	38	18	some	some	DET
fcis-2970	38	19	scholars	scholar	NOUN
fcis-2970	38	20	tried	try	VERB
fcis-2970	38	21	to	to	PART
fcis-2970	38	22	use	use	VERB
fcis-2970	38	23	attention	attention	NOUN
fcis-2970	38	24	mechanisms	mechanism	NOUN
fcis-2970	38	25	in	in	ADP
fcis-2970	38	26	the	the	DET
fcis-2970	38	27	field	field	NOUN
fcis-2970	38	28	of	of	ADP
fcis-2970	38	29	natural	natural	ADJ
fcis-2970	38	30	language	language	NOUN
fcis-2970	38	31	processing	processing	NOUN
fcis-2970	38	32	.	.	PUNCT
fcis-2970	39	1	bahdanau	bahdanau	PROPN
fcis-2970	39	2	et	et	PROPN
fcis-2970	39	3	al	al	PROPN
fcis-2970	39	4	.	.	PUNCT
fcis-2970	40	1	[	[	X
fcis-2970	40	2	7	7	X
fcis-2970	40	3	]	]	PUNCT
fcis-2970	40	4	added	add	VERB
fcis-2970	40	5	an	an	DET
fcis-2970	40	6	attention	attention	NOUN
fcis-2970	40	7	mechanism	mechanism	NOUN
fcis-2970	40	8	to	to	ADP
fcis-2970	40	9	the	the	DET
fcis-2970	40	10	machine	machine	NOUN
fcis-2970	40	11	translation	translation	NOUN
fcis-2970	40	12	task	task	NOUN
fcis-2970	40	13	,	,	PUNCT
fcis-2970	40	14	and	and	CCONJ
fcis-2970	40	15	the	the	DET
fcis-2970	40	16	success	success	NOUN
fcis-2970	40	17	of	of	ADP
fcis-2970	40	18	his	his	PRON
fcis-2970	40	19	experiment	experiment	NOUN
fcis-2970	40	20	meant	mean	VERB
fcis-2970	40	21	that	that	SCONJ
fcis-2970	40	22	the	the	DET
fcis-2970	40	23	attention	attention	NOUN
fcis-2970	40	24	mechanism	mechanism	NOUN
fcis-2970	40	25	began	begin	VERB
fcis-2970	40	26	to	to	PART
fcis-2970	40	27	be	be	AUX
fcis-2970	40	28	applied	apply	VERB
fcis-2970	40	29	in	in	ADP
fcis-2970	40	30	the	the	DET
fcis-2970	40	31	field	field	NOUN
fcis-2970	40	32	of	of	ADP
fcis-2970	40	33	natural	natural	ADJ
fcis-2970	40	34	language	language	NOUN
fcis-2970	40	35	.	.	PUNCT
fcis-2970	41	1	scholars	scholar	NOUN
fcis-2970	41	2	have	have	AUX
fcis-2970	41	3	successively	successively	ADV
fcis-2970	41	4	applied	apply	VERB
fcis-2970	41	5	attention	attention	NOUN
fcis-2970	41	6	mechanisms	mechanism	NOUN
fcis-2970	41	7	to	to	ADP
fcis-2970	41	8	subtasks	subtask	NOUN
fcis-2970	41	9	in	in	ADP
fcis-2970	41	10	the	the	DET
fcis-2970	41	11	field	field	NOUN
fcis-2970	41	12	,	,	PUNCT
fcis-2970	41	13	such	such	ADJ
fcis-2970	41	14	as	as	ADP
fcis-2970	41	15	the	the	DET
fcis-2970	41	16	two	two	NUM
fcis-2970	41	17	-	-	PUNCT
fcis-2970	41	18	layer	layer	NOUN
fcis-2970	41	19	cnn	cnn	PROPN
fcis-2970	41	20	-	-	PUNCT
fcis-2970	41	21	bilstm	bilstm	NOUN
fcis-2970	41	22	proposed	propose	VERB
fcis-2970	41	23	by	by	ADP
fcis-2970	41	24	liu	liu	PROPN
fcis-2970	41	25	fishing	fishing	PROPN
fcis-2970	41	26	et	et	PROPN
fcis-2970	41	27	al	al	PROPN
fcis-2970	41	28	.	.	PUNCT
fcis-2970	42	1	[	[	X
fcis-2970	42	2	8	8	NUM
fcis-2970	42	3	]	]	PUNCT
fcis-2970	42	4	,	,	PUNCT
fcis-2970	42	5	which	which	PRON
fcis-2970	42	6	added	add	VERB
fcis-2970	42	7	sentence	sentence	NOUN
fcis-2970	42	8	emotion	emotion	NOUN
fcis-2970	42	9	polarity	polarity	NOUN
fcis-2970	42	10	ordering	ordering	NOUN
fcis-2970	42	11	and	and	CCONJ
fcis-2970	42	12	attention	attention	NOUN
fcis-2970	42	13	mechanism	mechanism	NOUN
fcis-2970	42	14	.	.	PUNCT
fcis-2970	43	1	the	the	DET
fcis-2970	43	2	model	model	NOUN
fcis-2970	43	3	can	can	AUX
fcis-2970	43	4	fully	fully	ADV
fcis-2970	43	5	extract	extract	VERB
fcis-2970	43	6	text	text	NOUN
fcis-2970	43	7	features	feature	NOUN
fcis-2970	43	8	and	and	CCONJ
fcis-2970	43	9	optimize	optimize	VERB
fcis-2970	43	10	input	input	NOUN
fcis-2970	43	11	text	text	NOUN
fcis-2970	43	12	features	feature	NOUN
fcis-2970	43	13	,	,	PUNCT
fcis-2970	43	14	and	and	CCONJ
fcis-2970	43	15	experiments	experiment	NOUN
fcis-2970	43	16	show	show	VERB
fcis-2970	43	17	that	that	SCONJ
fcis-2970	43	18	the	the	DET
fcis-2970	43	19	model	model	NOUN
fcis-2970	43	20	has	have	VERB
fcis-2970	43	21	certain	certain	ADJ
fcis-2970	43	22	effectiveness	effectiveness	NOUN
fcis-2970	43	23	and	and	CCONJ
fcis-2970	43	24	feasibility	feasibility	NOUN
fcis-2970	43	25	.	.	PUNCT
fcis-2970	44	1	with	with	ADP
fcis-2970	44	2	the	the	DET
fcis-2970	44	3	emergence	emergence	NOUN
fcis-2970	44	4	of	of	ADP
fcis-2970	44	5	the	the	DET
fcis-2970	44	6	natural	natural	ADJ
fcis-2970	44	7	language	language	NOUN
fcis-2970	44	8	pretraining	pretraine	VERB
fcis-2970	44	9	model	model	NOUN
fcis-2970	44	10	,	,	PUNCT
fcis-2970	44	11	scholars	scholar	NOUN
fcis-2970	44	12	have	have	AUX
fcis-2970	44	13	been	be	AUX
fcis-2970	44	14	using	use	VERB
fcis-2970	44	15	the	the	DET
fcis-2970	44	16	pre	pre	ADJ
fcis-2970	44	17	-	-	ADJ
fcis-2970	44	18	training	training	ADJ
fcis-2970	44	19	model	model	NOUN
fcis-2970	44	20	in	in	ADP
fcis-2970	44	21	the	the	DET
fcis-2970	44	22	task	task	NOUN
fcis-2970	44	23	of	of	ADP
fcis-2970	44	24	emotion	emotion	NOUN
fcis-2970	44	25	classification	classification	NOUN
fcis-2970	44	26	,	,	PUNCT
fcis-2970	44	27	and	and	CCONJ
fcis-2970	44	28	have	have	AUX
fcis-2970	44	29	achieved	achieve	VERB
fcis-2970	44	30	better	well	ADJ
fcis-2970	44	31	results	result	NOUN
fcis-2970	44	32	in	in	ADP
fcis-2970	44	33	emotion	emotion	NOUN
fcis-2970	44	34	classification	classification	NOUN
fcis-2970	44	35	.	.	PUNCT
fcis-2970	45	1	pre	pre	ADJ
fcis-2970	45	2	-	-	ADJ
fcis-2970	45	3	training	training	ADJ
fcis-2970	45	4	models	model	NOUN
fcis-2970	45	5	[	[	X
fcis-2970	45	6	9	9	NUM
fcis-2970	45	7	]	]	PUNCT
fcis-2970	45	8	are	be	AUX
fcis-2970	45	9	divided	divide	VERB
fcis-2970	45	10	into	into	ADP
fcis-2970	45	11	static	static	ADJ
fcis-2970	45	12	models	model	NOUN
fcis-2970	45	13	and	and	CCONJ
fcis-2970	45	14	dynamic	dynamic	ADJ
fcis-2970	45	15	models	model	NOUN
fcis-2970	45	16	,	,	PUNCT
fcis-2970	45	17	among	among	ADP
fcis-2970	45	18	which	which	PRON
fcis-2970	45	19	the	the	DET
fcis-2970	45	20	word2vec	word2vec	PROPN
fcis-2970	45	21	model	model	NOUN
fcis-2970	45	22	[	[	X
fcis-2970	45	23	10	10	NUM
fcis-2970	45	24	]	]	PUNCT
fcis-2970	45	25	and	and	CCONJ
fcis-2970	45	26	glove	glove	NOUN
fcis-2970	45	27	model	model	NOUN
fcis-2970	45	28	are	be	AUX
fcis-2970	45	29	static	static	ADJ
fcis-2970	45	30	models	model	NOUN
fcis-2970	45	31	,	,	PUNCT
fcis-2970	45	32	and	and	CCONJ
fcis-2970	45	33	the	the	DET
fcis-2970	45	34	elmo	elmo	PROPN
fcis-2970	45	35	model	model	NOUN
fcis-2970	45	36	[	[	X
fcis-2970	45	37	11	11	NUM
fcis-2970	45	38	]	]	PUNCT
fcis-2970	45	39	,	,	PUNCT
fcis-2970	45	40	gpt	gpt	NOUN
fcis-2970	45	41	model	model	NOUN
fcis-2970	45	42	[	[	X
fcis-2970	45	43	12	12	NUM
fcis-2970	45	44	]	]	PUNCT
fcis-2970	45	45	,	,	PUNCT
fcis-2970	45	46	bert	bert	PROPN
fcis-2970	45	47	model	model	PROPN
fcis-2970	45	48	and	and	CCONJ
fcis-2970	45	49	ernie	ernie	PROPN
fcis-2970	45	50	model	model	NOUN
fcis-2970	45	51	[	[	X
fcis-2970	45	52	13	13	NUM
fcis-2970	45	53	]	]	PUNCT
fcis-2970	45	54	are	be	AUX
fcis-2970	45	55	dynamic	dynamic	ADJ
fcis-2970	45	56	models	model	NOUN
fcis-2970	45	57	.	.	PUNCT
fcis-2970	46	1	because	because	SCONJ
fcis-2970	46	2	static	static	ADJ
fcis-2970	46	3	word	word	NOUN
fcis-2970	46	4	vectors	vector	NOUN
fcis-2970	46	5	ca	can	AUX
fcis-2970	46	6	n't	not	PART
fcis-2970	46	7	represent	represent	VERB
fcis-2970	46	8	polysemous	polysemous	ADJ
fcis-2970	46	9	words	word	NOUN
fcis-2970	46	10	well	well	ADV
fcis-2970	46	11	,	,	PUNCT
fcis-2970	46	12	text	text	NOUN
fcis-2970	46	13	representation	representation	NOUN
fcis-2970	46	14	is	be	AUX
fcis-2970	46	15	limited	limited	ADJ
fcis-2970	46	16	.	.	PUNCT
fcis-2970	47	1	with	with	ADP
fcis-2970	47	2	the	the	DET
fcis-2970	47	3	introduction	introduction	NOUN
fcis-2970	47	4	of	of	ADP
fcis-2970	47	5	dynamic	dynamic	ADJ
fcis-2970	47	6	word	word	NOUN
fcis-2970	47	7	vectors	vector	NOUN
fcis-2970	47	8	,	,	PUNCT
fcis-2970	47	9	this	this	DET
fcis-2970	47	10	problem	problem	NOUN
fcis-2970	47	11	has	have	AUX
fcis-2970	47	12	been	be	AUX
fcis-2970	47	13	effectively	effectively	ADV
fcis-2970	47	14	solved	solve	VERB
fcis-2970	47	15	.	.	PUNCT
fcis-2970	48	1	the	the	DET
fcis-2970	48	2	emergence	emergence	NOUN
fcis-2970	48	3	of	of	ADP
fcis-2970	48	4	dynamic	dynamic	ADJ
fcis-2970	48	5	models	model	NOUN
fcis-2970	48	6	gpt	gpt	NOUN
fcis-2970	48	7	and	and	CCONJ
fcis-2970	48	8	bert	bert	PROPN
fcis-2970	48	9	,	,	PUNCT
fcis-2970	48	10	both	both	PRON
fcis-2970	48	11	based	base	VERB
fcis-2970	48	12	on	on	ADP
fcis-2970	48	13	the	the	DET
fcis-2970	48	14	transformer	transformer	NOUN
fcis-2970	48	15	model	model	NOUN
fcis-2970	48	16	,	,	PUNCT
fcis-2970	48	17	provides	provide	VERB
fcis-2970	48	18	a	a	DET
fcis-2970	48	19	new	new	ADJ
fcis-2970	48	20	way	way	NOUN
fcis-2970	48	21	of	of	ADP
fcis-2970	48	22	thinking	thinking	NOUN
fcis-2970	48	23	when	when	SCONJ
fcis-2970	48	24	dealing	deal	VERB
fcis-2970	48	25	with	with	ADP
fcis-2970	48	26	natural	natural	ADJ
fcis-2970	48	27	language	language	NOUN
fcis-2970	48	28	processing	processing	NOUN
fcis-2970	48	29	tasks	task	NOUN
fcis-2970	48	30	in	in	ADP
fcis-2970	48	31	the	the	DET
fcis-2970	48	32	future	future	NOUN
fcis-2970	48	33	.	.	PUNCT
fcis-2970	49	1	the	the	DET
fcis-2970	49	2	bert	bert	PROPN
fcis-2970	49	3	model	model	NOUN
fcis-2970	49	4	has	have	VERB
fcis-2970	49	5	a	a	DET
fcis-2970	49	6	good	good	ADJ
fcis-2970	49	7	performance	performance	NOUN
fcis-2970	49	8	in	in	ADP
fcis-2970	49	9	many	many	ADJ
fcis-2970	49	10	tasks	task	NOUN
fcis-2970	49	11	.	.	PUNCT
fcis-2970	50	1	it	it	PRON
fcis-2970	50	2	performs	perform	VERB
fcis-2970	50	3	the	the	DET
fcis-2970	50	4	vector	vector	NOUN
fcis-2970	50	5	representation	representation	NOUN
fcis-2970	50	6	of	of	ADP
fcis-2970	50	7	text	text	NOUN
fcis-2970	50	8	according	accord	VERB
fcis-2970	50	9	to	to	ADP
fcis-2970	50	10	the	the	DET
fcis-2970	50	11	word	word	NOUN
fcis-2970	50	12	level	level	NOUN
fcis-2970	50	13	,	,	PUNCT
fcis-2970	50	14	extracts	extract	VERB
fcis-2970	50	15	semantic	semantic	ADJ
fcis-2970	50	16	information	information	NOUN
fcis-2970	50	17	combined	combine	VERB
fcis-2970	50	18	with	with	ADP
fcis-2970	50	19	context	context	NOUN
fcis-2970	50	20	,	,	PUNCT
fcis-2970	50	21	and	and	CCONJ
fcis-2970	50	22	deals	deal	NOUN
fcis-2970	50	23	with	with	ADP
fcis-2970	50	24	polysemous	polysemous	ADJ
fcis-2970	50	25	words	word	NOUN
fcis-2970	50	26	well	well	ADV
fcis-2970	50	27	.	.	PUNCT
fcis-2970	51	1	however	however	ADV
fcis-2970	51	2	,	,	PUNCT
fcis-2970	51	3	the	the	DET
fcis-2970	51	4	bert	bert	PROPN
fcis-2970	51	5	model	model	NOUN
fcis-2970	51	6	does	do	AUX
fcis-2970	51	7	not	not	PART
fcis-2970	51	8	make	make	VERB
fcis-2970	51	9	use	use	NOUN
fcis-2970	51	10	of	of	ADP
fcis-2970	51	11	lexical	lexical	ADJ
fcis-2970	51	12	,	,	PUNCT
fcis-2970	51	13	grammatical	grammatical	ADJ
fcis-2970	51	14	structure	structure	NOUN
fcis-2970	51	15	,	,	PUNCT
fcis-2970	51	16	and	and	CCONJ
fcis-2970	51	17	semantic	semantic	ADJ
fcis-2970	51	18	information	information	NOUN
fcis-2970	51	19	in	in	ADP
fcis-2970	51	20	sentences	sentence	NOUN
fcis-2970	51	21	to	to	PART
fcis-2970	51	22	learn	learn	VERB
fcis-2970	51	23	modeling	modeling	NOUN
fcis-2970	51	24	,	,	PUNCT
fcis-2970	51	25	which	which	PRON
fcis-2970	51	26	is	be	AUX
fcis-2970	51	27	difficult	difficult	ADJ
fcis-2970	51	28	to	to	PART
fcis-2970	51	29	provide	provide	VERB
fcis-2970	51	30	a	a	DET
fcis-2970	51	31	good	good	ADJ
fcis-2970	51	32	vector	vector	NOUN
fcis-2970	51	33	representation	representation	NOUN
fcis-2970	51	34	of	of	ADP
fcis-2970	51	35	newly	newly	ADV
fcis-2970	51	36	emerged	emerge	VERB
fcis-2970	51	37	words	word	NOUN
fcis-2970	51	38	,	,	PUNCT
fcis-2970	51	39	while	while	SCONJ
fcis-2970	51	40	the	the	DET
fcis-2970	51	41	ernie	ernie	PROPN
fcis-2970	51	42	model	model	NOUN
fcis-2970	51	43	[	[	X
fcis-2970	51	44	14	14	NUM
fcis-2970	51	45	]	]	PUNCT
fcis-2970	51	46	fully	fully	ADV
fcis-2970	51	47	considers	consider	VERB
fcis-2970	51	48	the	the	DET
fcis-2970	51	49	lexical	lexical	ADJ
fcis-2970	51	50	,	,	PUNCT
fcis-2970	51	51	grammatical	grammatical	ADJ
fcis-2970	51	52	structure	structure	NOUN
fcis-2970	51	53	and	and	CCONJ
fcis-2970	51	54	semantic	semantic	ADJ
fcis-2970	51	55	information	information	NOUN
fcis-2970	51	56	of	of	ADP
fcis-2970	51	57	text	text	NOUN
fcis-2970	51	58	for	for	ADP
fcis-2970	51	59	modeling	modeling	NOUN
fcis-2970	51	60	,	,	PUNCT
fcis-2970	51	61	improving	improve	VERB
fcis-2970	51	62	the	the	DET
fcis-2970	51	63	universality	universality	NOUN
fcis-2970	51	64	of	of	ADP
fcis-2970	51	65	semantic	semantic	ADJ
fcis-2970	51	66	expression	expression	NOUN
fcis-2970	51	67	.	.	PUNCT
fcis-2970	52	1	in	in	ADP
fcis-2970	52	2	natural	natural	ADJ
fcis-2970	52	3	language	language	NOUN
fcis-2970	52	4	processing	processing	NOUN
fcis-2970	52	5	tasks	task	NOUN
fcis-2970	52	6	using	use	VERB
fcis-2970	52	7	the	the	DET
fcis-2970	52	8	neural	neural	ADJ
fcis-2970	52	9	network	network	NOUN
fcis-2970	52	10	method	method	NOUN
fcis-2970	52	11	,	,	PUNCT
fcis-2970	52	12	how	how	SCONJ
fcis-2970	52	13	to	to	PART
fcis-2970	52	14	convert	convert	VERB
fcis-2970	52	15	text	text	NOUN
fcis-2970	52	16	characters	character	NOUN
fcis-2970	52	17	into	into	ADP
fcis-2970	52	18	digital	digital	ADJ
fcis-2970	52	19	features	feature	NOUN
fcis-2970	52	20	combined	combine	VERB
fcis-2970	52	21	with	with	ADP
fcis-2970	52	22	text	text	NOUN
fcis-2970	52	23	information	information	NOUN
fcis-2970	52	24	will	will	AUX
fcis-2970	52	25	determine	determine	VERB
fcis-2970	52	26	the	the	DET
fcis-2970	52	27	upper	upper	ADJ
fcis-2970	52	28	limit	limit	NOUN
fcis-2970	52	29	of	of	ADP
fcis-2970	52	30	model	model	NOUN
fcis-2970	52	31	performance	performance	NOUN
fcis-2970	52	32	.	.	PUNCT
fcis-2970	53	1	if	if	SCONJ
fcis-2970	53	2	a	a	DET
fcis-2970	53	3	static	static	ADJ
fcis-2970	53	4	pre	pre	ADJ
fcis-2970	53	5	-	-	ADJ
fcis-2970	53	6	training	training	ADJ
fcis-2970	53	7	model	model	NOUN
fcis-2970	53	8	is	be	AUX
fcis-2970	53	9	used	use	VERB
fcis-2970	53	10	,	,	PUNCT
fcis-2970	53	11	it	it	PRON
fcis-2970	53	12	will	will	AUX
fcis-2970	53	13	only	only	ADV
fcis-2970	53	14	learn	learn	VERB
fcis-2970	53	15	the	the	DET
fcis-2970	53	16	static	static	ADJ
fcis-2970	53	17	vector	vector	NOUN
fcis-2970	53	18	of	of	ADP
fcis-2970	53	19	words	word	NOUN
fcis-2970	53	20	and	and	CCONJ
fcis-2970	53	21	ignore	ignore	VERB
fcis-2970	53	22	the	the	DET
fcis-2970	53	23	multi	multi	NOUN
fcis-2970	53	24	-	-	NOUN
fcis-2970	53	25	semantics	semantic	NOUN
fcis-2970	53	26	of	of	ADP
fcis-2970	53	27	polysemy	polysemy	NOUN
fcis-2970	53	28	in	in	ADP
fcis-2970	53	29	the	the	DET
fcis-2970	53	30	text	text	NOUN
fcis-2970	53	31	.	.	PUNCT
fcis-2970	54	1	therefore	therefore	ADV
fcis-2970	54	2	,	,	PUNCT
fcis-2970	54	3	the	the	DET
fcis-2970	54	4	dynamic	dynamic	ADJ
fcis-2970	54	5	pre	pre	ADJ
fcis-2970	54	6	-	-	ADJ
fcis-2970	54	7	training	training	ADJ
fcis-2970	54	8	model	model	NOUN
fcis-2970	54	9	ernie	ernie	PROPN
fcis-2970	54	10	will	will	AUX
fcis-2970	54	11	be	be	AUX
fcis-2970	54	12	used	use	VERB
fcis-2970	54	13	in	in	ADP
fcis-2970	54	14	this	this	DET
fcis-2970	54	15	paper	paper	NOUN
fcis-2970	54	16	.	.	PUNCT
fcis-2970	55	1	this	this	DET
fcis-2970	55	2	model	model	NOUN
fcis-2970	55	3	can	can	AUX
fcis-2970	55	4	effectively	effectively	ADV
fcis-2970	55	5	solve	solve	VERB
fcis-2970	55	6	the	the	DET
fcis-2970	55	7	polysemy	polysemy	NOUN
fcis-2970	55	8	representation	representation	NOUN
fcis-2970	55	9	problem	problem	NOUN
fcis-2970	55	10	through	through	ADP
fcis-2970	55	11	the	the	DET
fcis-2970	55	12	sentence	sentence	NOUN
fcis-2970	55	13	word	word	NOUN
fcis-2970	55	14	vector	vector	NOUN
fcis-2970	55	15	generated	generate	VERB
fcis-2970	55	16	after	after	ADP
fcis-2970	55	17	training	training	NOUN
fcis-2970	55	18	.	.	PUNCT
fcis-2970	56	1	the	the	DET
fcis-2970	56	2	overall	overall	ADJ
fcis-2970	56	3	process	process	NOUN
fcis-2970	56	4	is	be	AUX
fcis-2970	56	5	to	to	PART
fcis-2970	56	6	use	use	VERB
fcis-2970	56	7	each	each	DET
fcis-2970	56	8	word	word	NOUN
fcis-2970	56	9	segmentation	segmentation	NOUN
fcis-2970	56	10	vector	vector	NOUN
fcis-2970	56	11	of	of	ADP
fcis-2970	56	12	the	the	DET
fcis-2970	56	13	output	output	NOUN
fcis-2970	56	14	matrix	matrix	NOUN
fcis-2970	56	15	of	of	ADP
fcis-2970	56	16	the	the	DET
fcis-2970	56	17	ernie	ernie	NOUN
fcis-2970	56	18	pre	pre	ADJ
fcis-2970	56	19	-	-	ADJ
fcis-2970	56	20	training	training	ADJ
fcis-2970	56	21	model	model	NOUN
fcis-2970	56	22	as	as	ADP
fcis-2970	56	23	the	the	DET
fcis-2970	56	24	input	input	NOUN
fcis-2970	56	25	of	of	ADP
fcis-2970	56	26	the	the	DET
fcis-2970	56	27	bidirectional	bidirectional	ADJ
fcis-2970	56	28	long	long	ADJ
fcis-2970	56	29	and	and	CCONJ
fcis-2970	56	30	short	short	ADJ
fcis-2970	56	31	-	-	PUNCT
fcis-2970	56	32	term	term	NOUN
fcis-2970	56	33	memory	memory	NOUN
fcis-2970	56	34	neural	neural	ADJ
fcis-2970	56	35	network	network	NOUN
fcis-2970	56	36	,	,	PUNCT
fcis-2970	56	37	and	and	CCONJ
fcis-2970	56	38	then	then	ADV
fcis-2970	56	39	use	use	VERB
fcis-2970	56	40	the	the	DET
fcis-2970	56	41	attention	attention	NOUN
fcis-2970	56	42	mechanism	mechanism	NOUN
fcis-2970	56	43	to	to	PART
fcis-2970	56	44	weigh	weigh	VERB
fcis-2970	56	45	the	the	DET
fcis-2970	56	46	output	output	NOUN
fcis-2970	56	47	through	through	ADP
fcis-2970	56	48	the	the	DET
fcis-2970	56	49	neural	neural	ADJ
fcis-2970	56	50	network	network	NOUN
fcis-2970	56	51	.	.	PUNCT
fcis-2970	57	1	more	more	ADJ
fcis-2970	57	2	weight	weight	NOUN
fcis-2970	57	3	is	be	AUX
fcis-2970	57	4	given	give	VERB
fcis-2970	57	5	to	to	ADP
fcis-2970	57	6	the	the	DET
fcis-2970	57	7	time	time	NOUN
fcis-2970	57	8	step	step	NOUN
fcis-2970	57	9	output	output	NOUN
fcis-2970	57	10	that	that	PRON
fcis-2970	57	11	has	have	VERB
fcis-2970	57	12	a	a	DET
fcis-2970	57	13	greater	great	ADJ
fcis-2970	57	14	impact	impact	NOUN
fcis-2970	57	15	on	on	ADP
fcis-2970	57	16	the	the	DET
fcis-2970	57	17	emotional	emotional	ADJ
fcis-2970	57	18	label	label	NOUN
fcis-2970	57	19	,	,	PUNCT
fcis-2970	57	20	and	and	CCONJ
fcis-2970	57	21	more	more	ADJ
fcis-2970	57	22	emphasis	emphasis	NOUN
fcis-2970	57	23	is	be	AUX
fcis-2970	57	24	placed	place	VERB
fcis-2970	57	25	on	on	ADP
fcis-2970	57	26	the	the	DET
fcis-2970	57	27	words	word	NOUN
fcis-2970	57	28	in	in	ADP
fcis-2970	57	29	the	the	DET
fcis-2970	57	30	sentence	sentence	NOUN
fcis-2970	57	31	to	to	PART
fcis-2970	57	32	improve	improve	VERB
fcis-2970	57	33	the	the	DET
fcis-2970	57	34	overall	overall	ADJ
fcis-2970	57	35	performance	performance	NOUN
fcis-2970	57	36	of	of	ADP
fcis-2970	57	37	the	the	DET
fcis-2970	57	38	model	model	NOUN
fcis-2970	57	39	.	.	PUNCT
fcis-2970	58	1	2	2	X
fcis-2970	58	2	.	.	X
fcis-2970	58	3	text	text	NOUN
fcis-2970	58	4	emotion	emotion	NOUN
fcis-2970	58	5	classification	classification	NOUN
fcis-2970	58	6	model	model	NOUN
fcis-2970	58	7	based	base	VERB
fcis-2970	58	8	on	on	ADP
fcis-2970	58	9	ernie	ernie	PROPN
fcis-2970	58	10	and	and	CCONJ
fcis-2970	58	11	bilstm	bilstm	NOUN
fcis-2970	58	12	-	-	PUNCT
fcis-2970	58	13	at	at	ADP
fcis-2970	58	14	the	the	DET
fcis-2970	58	15	text	text	NOUN
fcis-2970	58	16	emotion	emotion	NOUN
fcis-2970	58	17	classification	classification	NOUN
fcis-2970	58	18	model	model	NOUN
fcis-2970	58	19	combined	combine	VERB
fcis-2970	58	20	with	with	ADP
fcis-2970	58	21	ernie	ernie	NOUN
fcis-2970	58	22	and	and	CCONJ
fcis-2970	58	23	bilstm	bilstm	NOUN
fcis-2970	58	24	-	-	PUNCT
fcis-2970	58	25	at	at	NOUN
fcis-2970	58	26	is	be	AUX
fcis-2970	58	27	shown	show	VERB
fcis-2970	58	28	in	in	ADP
fcis-2970	58	29	figure	figure	NOUN
fcis-2970	58	30	1	1	NUM
fcis-2970	58	31	below	below	ADV
fcis-2970	58	32	.	.	PUNCT
fcis-2970	59	1	the	the	DET
fcis-2970	59	2	model	model	NOUN
fcis-2970	59	3	has	have	VERB
fcis-2970	59	4	six	six	NUM
fcis-2970	59	5	layers	layer	NOUN
fcis-2970	59	6	,	,	PUNCT
fcis-2970	59	7	which	which	PRON
fcis-2970	59	8	are	be	AUX
fcis-2970	59	9	data	datum	NOUN
fcis-2970	59	10	preprocessing	preprocesse	VERB
fcis-2970	59	11	layer	layer	NOUN
fcis-2970	59	12	,	,	PUNCT
fcis-2970	59	13	word	word	NOUN
fcis-2970	59	14	embedding	embed	VERB
fcis-2970	59	15	layer	layer	NOUN
fcis-2970	59	16	,	,	PUNCT
fcis-2970	59	17	ernie	ernie	NOUN
fcis-2970	59	18	layer	layer	NOUN
fcis-2970	59	19	,	,	PUNCT
fcis-2970	59	20	bilstm	bilstm	NOUN
fcis-2970	59	21	semantic	semantic	ADJ
fcis-2970	59	22	extraction	extraction	NOUN
fcis-2970	59	23	layer	layer	NOUN
fcis-2970	59	24	,	,	PUNCT
fcis-2970	59	25	attention	attention	NOUN
fcis-2970	59	26	mechanism	mechanism	NOUN
fcis-2970	59	27	layer	layer	NOUN
fcis-2970	59	28	and	and	CCONJ
fcis-2970	59	29	the	the	DET
fcis-2970	59	30	final	final	ADJ
fcis-2970	59	31	output	output	NOUN
fcis-2970	59	32	layer	layer	NOUN
fcis-2970	59	33	.	.	PUNCT
fcis-2970	60	1	among	among	ADP
fcis-2970	60	2	them	they	PRON
fcis-2970	60	3	,	,	PUNCT
fcis-2970	60	4	the	the	DET
fcis-2970	60	5	data	data	NOUN
fcis-2970	60	6	preprocessing	preprocesse	VERB
fcis-2970	60	7	layer	layer	NOUN
fcis-2970	60	8	will	will	AUX
fcis-2970	60	9	remove	remove	VERB
fcis-2970	60	10	the	the	DET
fcis-2970	60	11	useless	useless	ADJ
fcis-2970	60	12	data	datum	NOUN
fcis-2970	60	13	in	in	ADP
fcis-2970	60	14	the	the	DET
fcis-2970	60	15	text	text	NOUN
fcis-2970	60	16	,	,	PUNCT
fcis-2970	60	17	leaving	leave	VERB
fcis-2970	60	18	only	only	ADV
fcis-2970	60	19	the	the	DET
fcis-2970	60	20	text	text	NOUN
fcis-2970	60	21	that	that	PRON
fcis-2970	60	22	can	can	AUX
fcis-2970	60	23	express	express	VERB
fcis-2970	60	24	the	the	DET
fcis-2970	60	25	semantic	semantic	ADJ
fcis-2970	60	26	information	information	NOUN
fcis-2970	60	27	;	;	PUNCT
fcis-2970	60	28	in	in	ADP
fcis-2970	60	29	the	the	DET
fcis-2970	60	30	word	word	NOUN
fcis-2970	60	31	embedding	embed	VERB
fcis-2970	60	32	layer	layer	NOUN
fcis-2970	60	33	,	,	PUNCT
fcis-2970	60	34	the	the	DET
fcis-2970	60	35	word	word	NOUN
fcis-2970	60	36	segmentation	segmentation	NOUN
fcis-2970	60	37	is	be	AUX
fcis-2970	60	38	mapped	map	VERB
fcis-2970	60	39	to	to	ADP
fcis-2970	60	40	the	the	DET
fcis-2970	60	41	corresponding	corresponding	ADJ
fcis-2970	60	42	word	word	NOUN
fcis-2970	60	43	vector	vector	NOUN
fcis-2970	60	44	in	in	ADP
fcis-2970	60	45	the	the	DET
fcis-2970	60	46	word	word	NOUN
fcis-2970	60	47	direction	direction	NOUN
fcis-2970	60	48	scale	scale	NOUN
fcis-2970	60	49	;	;	PUNCT
fcis-2970	60	50	in	in	ADP
fcis-2970	60	51	ernie	ernie	NOUN
fcis-2970	60	52	layer	layer	NOUN
fcis-2970	60	53	,	,	PUNCT
fcis-2970	60	54	the	the	DET
fcis-2970	60	55	text	text	NOUN
fcis-2970	60	56	word	word	NOUN
fcis-2970	60	57	vector	vector	NOUN
fcis-2970	60	58	is	be	AUX
fcis-2970	60	59	transformed	transform	VERB
fcis-2970	60	60	into	into	ADP
fcis-2970	60	61	the	the	DET
fcis-2970	60	62	dynamic	dynamic	ADJ
fcis-2970	60	63	vector	vector	NOUN
fcis-2970	60	64	representation	representation	NOUN
fcis-2970	60	65	of	of	ADP
fcis-2970	60	66	the	the	DET
fcis-2970	60	67	sentence	sentence	NOUN
fcis-2970	60	68	by	by	ADP
fcis-2970	60	69	combining	combine	VERB
fcis-2970	60	70	the	the	DET
fcis-2970	60	71	semantic	semantic	ADJ
fcis-2970	60	72	information	information	NOUN
fcis-2970	60	73	of	of	ADP
fcis-2970	60	74	the	the	DET
fcis-2970	60	75	text	text	NOUN
fcis-2970	60	76	.	.	PUNCT
fcis-2970	61	1	in	in	ADP
fcis-2970	61	2	the	the	DET
fcis-2970	61	3	bilstm	bilstm	NOUN
fcis-2970	61	4	layer	layer	NOUN
fcis-2970	61	5	,	,	PUNCT
fcis-2970	61	6	the	the	DET
fcis-2970	61	7	text	text	NOUN
fcis-2970	61	8	context	context	NOUN
fcis-2970	61	9	semantic	semantic	ADJ
fcis-2970	61	10	information	information	NOUN
fcis-2970	61	11	will	will	AUX
fcis-2970	61	12	be	be	AUX
fcis-2970	61	13	further	far	ADV
fcis-2970	61	14	extracted	extract	VERB
fcis-2970	61	15	.	.	PUNCT
fcis-2970	62	1	at	at	ADP
fcis-2970	62	2	the	the	DET
fcis-2970	62	3	level	level	NOUN
fcis-2970	62	4	of	of	ADP
fcis-2970	62	5	attention	attention	NOUN
fcis-2970	62	6	mechanism	mechanism	NOUN
fcis-2970	62	7	,	,	PUNCT
fcis-2970	62	8	different	different	ADJ
fcis-2970	62	9	weights	weight	NOUN
fcis-2970	62	10	are	be	AUX
fcis-2970	62	11	given	give	VERB
fcis-2970	62	12	according	accord	VERB
fcis-2970	62	13	to	to	ADP
fcis-2970	62	14	the	the	DET
fcis-2970	62	15	importance	importance	NOUN
fcis-2970	62	16	of	of	ADP
fcis-2970	62	17	each	each	DET
fcis-2970	62	18	participle	participle	NOUN
fcis-2970	62	19	to	to	ADP
fcis-2970	62	20	the	the	DET
fcis-2970	62	21	sentence	sentence	NOUN
fcis-2970	62	22	emotion	emotion	NOUN
fcis-2970	62	23	classification	classification	NOUN
fcis-2970	62	24	,	,	PUNCT
fcis-2970	62	25	and	and	CCONJ
fcis-2970	62	26	the	the	DET
fcis-2970	62	27	weighted	weight	VERB
fcis-2970	62	28	summation	summation	NOUN
fcis-2970	62	29	can	can	AUX
fcis-2970	62	30	get	get	VERB
fcis-2970	62	31	the	the	DET
fcis-2970	62	32	semantic	semantic	ADJ
fcis-2970	62	33	information	information	NOUN
fcis-2970	62	34	at	at	ADP
fcis-2970	62	35	the	the	DET
fcis-2970	62	36	sentence	sentence	NOUN
fcis-2970	62	37	level	level	NOUN
fcis-2970	62	38	.	.	PUNCT
fcis-2970	63	1	in	in	ADP
fcis-2970	63	2	the	the	DET
fcis-2970	63	3	output	output	NOUN
fcis-2970	63	4	layer	layer	NOUN
fcis-2970	63	5	,	,	PUNCT
fcis-2970	63	6	the	the	DET
fcis-2970	63	7	fully	fully	ADV
fcis-2970	63	8	connected	connected	ADJ
fcis-2970	63	9	network	network	NOUN
fcis-2970	63	10	will	will	AUX
fcis-2970	63	11	be	be	AUX
fcis-2970	63	12	used	use	VERB
fcis-2970	63	13	to	to	PART
fcis-2970	63	14	obtain	obtain	VERB
fcis-2970	63	15	the	the	DET
fcis-2970	63	16	probability	probability	NOUN
fcis-2970	63	17	of	of	ADP
fcis-2970	63	18	each	each	DET
fcis-2970	63	19	category	category	NOUN
fcis-2970	63	20	,	,	PUNCT
fcis-2970	63	21	and	and	CCONJ
fcis-2970	63	22	then	then	ADV
fcis-2970	63	23	the	the	DET
fcis-2970	63	24	classification	classification	NOUN
fcis-2970	63	25	results	result	NOUN
fcis-2970	63	26	will	will	AUX
fcis-2970	63	27	be	be	AUX
fcis-2970	63	28	obtained	obtain	VERB
fcis-2970	63	29	.	.	PUNCT
fcis-2970	64	1	figure	figure	NOUN
fcis-2970	64	2	1	1	NUM
fcis-2970	64	3	.	.	PUNCT
fcis-2970	65	1	text	text	NOUN
fcis-2970	65	2	emotion	emotion	NOUN
fcis-2970	65	3	classification	classification	NOUN
fcis-2970	65	4	model	model	NOUN
fcis-2970	65	5	based	base	VERB
fcis-2970	65	6	on	on	ADP
fcis-2970	65	7	ernie	ernie	PROPN
fcis-2970	65	8	and	and	CCONJ
fcis-2970	65	9	bilstm	bilstm	NOUN
fcis-2970	65	10	-	-	PUNCT
fcis-2970	65	11	at	at	ADP
fcis-2970	65	12	2.1	2.1	NUM
fcis-2970	65	13	.	.	PUNCT
fcis-2970	66	1	data	datum	NOUN
fcis-2970	66	2	preprocessing	preprocessing	NOUN
fcis-2970	66	3	and	and	CCONJ
fcis-2970	66	4	word	word	NOUN
fcis-2970	66	5	static	static	ADJ
fcis-2970	66	6	vector	vector	NOUN
fcis-2970	66	7	representation	representation	NOUN
fcis-2970	66	8	due	due	ADP
fcis-2970	66	9	to	to	ADP
fcis-2970	66	10	the	the	DET
fcis-2970	66	11	large	large	ADJ
fcis-2970	66	12	amount	amount	NOUN
fcis-2970	66	13	of	of	ADP
fcis-2970	66	14	noise	noise	NOUN
fcis-2970	66	15	in	in	ADP
fcis-2970	66	16	the	the	DET
fcis-2970	66	17	comment	comment	NOUN
fcis-2970	66	18	text	text	NOUN
fcis-2970	66	19	data	datum	NOUN
fcis-2970	66	20	,	,	PUNCT
fcis-2970	66	21	it	it	PRON
fcis-2970	66	22	will	will	AUX
fcis-2970	66	23	affect	affect	VERB
fcis-2970	66	24	the	the	DET
fcis-2970	66	25	extraction	extraction	NOUN
fcis-2970	66	26	of	of	ADP
fcis-2970	66	27	subsequent	subsequent	ADJ
fcis-2970	66	28	sentence	sentence	NOUN
fcis-2970	66	29	features	feature	NOUN
fcis-2970	66	30	,	,	PUNCT
fcis-2970	66	31	so	so	SCONJ
fcis-2970	66	32	it	it	PRON
fcis-2970	66	33	is	be	AUX
fcis-2970	66	34	necessary	necessary	ADJ
fcis-2970	66	35	to	to	AUX
fcis-2970	66	36	de	de	NOUN
fcis-2970	66	37	-	-	VERB
fcis-2970	66	38	noise	noise	VERB
fcis-2970	66	39	the	the	DET
fcis-2970	66	40	text	text	NOUN
fcis-2970	66	41	before	before	ADP
fcis-2970	66	42	the	the	DET
fcis-2970	66	43	extraction	extraction	NOUN
fcis-2970	66	44	of	of	ADP
fcis-2970	66	45	sentence	sentence	NOUN
fcis-2970	66	46	features	feature	NOUN
fcis-2970	66	47	.	.	PUNCT
fcis-2970	67	1	regular	regular	ADJ
fcis-2970	67	2	expressions	expression	NOUN
fcis-2970	67	3	can	can	AUX
fcis-2970	67	4	be	be	AUX
fcis-2970	67	5	used	use	VERB
fcis-2970	67	6	to	to	PART
fcis-2970	67	7	remove	remove	VERB
fcis-2970	67	8	punctuation	punctuation	NOUN
fcis-2970	67	9	marks	mark	NOUN
fcis-2970	67	10	and	and	CCONJ
fcis-2970	67	11	expressions	expression	NOUN
fcis-2970	67	12	,	,	PUNCT
fcis-2970	67	13	etc	etc	X
fcis-2970	67	14	.	.	X
fcis-2970	67	15	,	,	PUNCT
fcis-2970	67	16	and	and	CCONJ
fcis-2970	67	17	then	then	ADV
fcis-2970	67	18	word	word	NOUN
fcis-2970	67	19	segmentation	segmentation	NOUN
fcis-2970	67	20	is	be	AUX
fcis-2970	67	21	carried	carry	VERB
fcis-2970	67	22	out	out	ADP
fcis-2970	67	23	to	to	PART
fcis-2970	67	24	limit	limit	VERB
fcis-2970	67	25	the	the	DET
fcis-2970	67	26	sentence	sentence	NOUN
fcis-2970	67	27	length	length	NOUN
fcis-2970	67	28	to	to	ADP
fcis-2970	67	29	no	no	DET
fcis-2970	67	30	more	more	ADJ
fcis-2970	67	31	than	than	ADP
fcis-2970	67	32	the	the	DET
fcis-2970	67	33	maximum	maximum	ADJ
fcis-2970	67	34	length	length	NOUN
fcis-2970	67	35	minus	minus	CCONJ
fcis-2970	67	36	2	2	NUM
fcis-2970	67	37	,	,	PUNCT
fcis-2970	67	38	and	and	CCONJ
fcis-2970	67	39	then	then	ADV
fcis-2970	67	40	[	[	X
fcis-2970	67	41	cls	cls	X
fcis-2970	67	42	]	]	PUNCT
fcis-2970	67	43	and	and	CCONJ
fcis-2970	67	44	[	[	X
fcis-2970	67	45	sep	sep	X
fcis-2970	67	46	]	]	X
fcis-2970	67	47	are	be	AUX
fcis-2970	67	48	added	add	VERB
fcis-2970	67	49	at	at	ADP
fcis-2970	67	50	the	the	DET
fcis-2970	67	51	beginning	beginning	NOUN
fcis-2970	67	52	and	and	CCONJ
fcis-2970	67	53	end	end	NOUN
fcis-2970	67	54	of	of	ADP
fcis-2970	67	55	the	the	DET
fcis-2970	67	56	sentence	sentence	NOUN
fcis-2970	67	57	respectively	respectively	ADV
fcis-2970	67	58	.	.	PUNCT
fcis-2970	68	1	then	then	ADV
fcis-2970	68	2	,	,	PUNCT
fcis-2970	68	3	each	each	DET
fcis-2970	68	4	character	character	NOUN
fcis-2970	68	5	in	in	ADP
fcis-2970	68	6	the	the	DET
fcis-2970	68	7	sentence	sentence	NOUN
fcis-2970	68	8	is	be	AUX
fcis-2970	68	9	converted	convert	VERB
fcis-2970	68	10	into	into	ADP
fcis-2970	68	11	a	a	DET
fcis-2970	68	12	corresponding	correspond	VERB
fcis-2970	68	13	vector	vector	NOUN
fcis-2970	68	14	.	.	PUNCT
fcis-2970	69	1	without	without	ADP
fcis-2970	69	2	input	input	NOUN
fcis-2970	69	3	into	into	ADP
fcis-2970	69	4	the	the	DET
fcis-2970	69	5	model	model	NOUN
fcis-2970	69	6	for	for	ADP
fcis-2970	69	7	training	training	NOUN
fcis-2970	69	8	,	,	PUNCT
fcis-2970	69	9	these	these	DET
fcis-2970	69	10	vectors	vector	NOUN
fcis-2970	69	11	are	be	AUX
fcis-2970	69	12	just	just	ADV
fcis-2970	69	13	static	static	ADJ
fcis-2970	69	14	vectors	vector	NOUN
fcis-2970	69	15	that	that	PRON
fcis-2970	69	16	can	can	AUX
fcis-2970	69	17	not	not	PART
fcis-2970	69	18	solve	solve	VERB
fcis-2970	69	19	the	the	DET
fcis-2970	69	20	polysemous	polysemous	ADJ
fcis-2970	69	21	problem	problem	NOUN
fcis-2970	69	22	.	.	PUNCT
fcis-2970	70	1	after	after	ADP
fcis-2970	70	2	the	the	DET
fcis-2970	70	3	training	training	NOUN
fcis-2970	70	4	,	,	PUNCT
fcis-2970	70	5	the	the	DET
fcis-2970	70	6	dynamic	dynamic	ADJ
fcis-2970	70	7	vector	vector	NOUN
fcis-2970	70	8	representation	representation	NOUN
fcis-2970	70	9	of	of	ADP
fcis-2970	70	10	the	the	DET
fcis-2970	70	11	text	text	NOUN
fcis-2970	70	12	that	that	PRON
fcis-2970	70	13	integrates	integrate	VERB
fcis-2970	70	14	the	the	DET
fcis-2970	70	15	sentence	sentence	NOUN
fcis-2970	70	16	information	information	NOUN
fcis-2970	70	17	will	will	AUX
fcis-2970	70	18	be	be	AUX
fcis-2970	70	19	obtained	obtain	VERB
fcis-2970	70	20	,	,	PUNCT
fcis-2970	70	21	which	which	PRON
fcis-2970	70	22	can	can	AUX
fcis-2970	70	23	better	well	ADV
fcis-2970	70	24	solve	solve	VERB
fcis-2970	70	25	the	the	DET
fcis-2970	70	26	polysemous	polysemous	ADJ
fcis-2970	70	27	problem	problem	NOUN
fcis-2970	70	28	.	.	PUNCT
fcis-2970	71	1	72	72	NUM
fcis-2970	71	2	2.2	2.2	NUM
fcis-2970	71	3	.	.	PUNCT
fcis-2970	72	1	ernie	ernie	PROPN
fcis-2970	72	2	pre	pre	ADJ
fcis-2970	72	3	-	-	ADJ
fcis-2970	72	4	training	training	ADJ
fcis-2970	72	5	model	model	NOUN
fcis-2970	72	6	ernie	ernie	PROPN
fcis-2970	72	7	model	model	PROPN
fcis-2970	72	8	is	be	AUX
fcis-2970	72	9	a	a	DET
fcis-2970	72	10	kind	kind	NOUN
fcis-2970	72	11	of	of	ADP
fcis-2970	72	12	enhanced	enhanced	ADJ
fcis-2970	72	13	presentation	presentation	NOUN
fcis-2970	72	14	pretraining	pretraine	VERB
fcis-2970	72	15	model	model	NOUN
fcis-2970	72	16	which	which	PRON
fcis-2970	72	17	realizes	realize	VERB
fcis-2970	72	18	knowledge	knowledge	NOUN
fcis-2970	72	19	integration	integration	NOUN
fcis-2970	72	20	through	through	ADP
fcis-2970	72	21	mask	mask	NOUN
fcis-2970	72	22	mechanism	mechanism	NOUN
fcis-2970	72	23	.	.	PUNCT
fcis-2970	73	1	it	it	PRON
fcis-2970	73	2	includes	include	VERB
fcis-2970	73	3	two	two	NUM
fcis-2970	73	4	parts	part	NOUN
fcis-2970	73	5	:	:	PUNCT
fcis-2970	73	6	text	text	NOUN
fcis-2970	73	7	encoder	encoder	NOUN
fcis-2970	73	8	and	and	CCONJ
fcis-2970	73	9	knowledge	knowledge	NOUN
fcis-2970	73	10	integration	integration	NOUN
fcis-2970	73	11	.	.	PUNCT
fcis-2970	74	1	the	the	DET
fcis-2970	74	2	former	former	ADJ
fcis-2970	74	3	has	have	VERB
fcis-2970	74	4	the	the	DET
fcis-2970	74	5	same	same	ADJ
fcis-2970	74	6	structure	structure	NOUN
fcis-2970	74	7	as	as	ADP
fcis-2970	74	8	the	the	DET
fcis-2970	74	9	encoder	encoder	NOUN
fcis-2970	74	10	in	in	ADP
fcis-2970	74	11	transformer	transformer	NOUN
fcis-2970	74	12	.	.	PUNCT
fcis-2970	75	1	it	it	PRON
fcis-2970	75	2	captures	capture	VERB
fcis-2970	75	3	the	the	DET
fcis-2970	75	4	context	context	NOUN
fcis-2970	75	5	information	information	NOUN
fcis-2970	75	6	of	of	ADP
fcis-2970	75	7	each	each	DET
fcis-2970	75	8	mark	mark	NOUN
fcis-2970	75	9	in	in	ADP
fcis-2970	75	10	the	the	DET
fcis-2970	75	11	sentence	sentence	NOUN
fcis-2970	75	12	through	through	ADP
fcis-2970	75	13	the	the	DET
fcis-2970	75	14	self	self	NOUN
fcis-2970	75	15	-	-	PUNCT
fcis-2970	75	16	attention	attention	NOUN
fcis-2970	75	17	mechanism	mechanism	NOUN
fcis-2970	75	18	,	,	PUNCT
fcis-2970	75	19	inputs	input	VERB
fcis-2970	75	20	the	the	DET
fcis-2970	75	21	result	result	NOUN
fcis-2970	75	22	into	into	ADP
fcis-2970	75	23	the	the	DET
fcis-2970	75	24	feed	feed	NOUN
fcis-2970	75	25	-	-	PUNCT
fcis-2970	75	26	forward	forward	ADV
fcis-2970	75	27	neural	neural	ADJ
fcis-2970	75	28	network	network	NOUN
fcis-2970	75	29	,	,	PUNCT
fcis-2970	75	30	and	and	CCONJ
fcis-2970	75	31	generates	generate	VERB
fcis-2970	75	32	the	the	DET
fcis-2970	75	33	corresponding	corresponding	ADJ
fcis-2970	75	34	word	word	NOUN
fcis-2970	75	35	vector	vector	NOUN
fcis-2970	75	36	representation	representation	NOUN
fcis-2970	75	37	at	at	ADP
fcis-2970	75	38	last	last	ADJ
fcis-2970	75	39	.	.	PUNCT
fcis-2970	76	1	the	the	DET
fcis-2970	76	2	encoder	encoder	NOUN
fcis-2970	76	3	structure	structure	NOUN
fcis-2970	76	4	is	be	AUX
fcis-2970	76	5	shown	show	VERB
fcis-2970	76	6	in	in	ADP
fcis-2970	76	7	figure	figure	NOUN
fcis-2970	76	8	2	2	NUM
fcis-2970	76	9	below	below	ADV
fcis-2970	76	10	.	.	PUNCT
fcis-2970	77	1	figure	figure	NOUN
fcis-2970	77	2	2	2	NUM
fcis-2970	77	3	.	.	PUNCT
fcis-2970	77	4	transformer	transformer	ADJ
fcis-2970	77	5	encoder	encoder	NOUN
fcis-2970	77	6	structure	structure	NOUN
fcis-2970	77	7	diagram	diagram	PROPN
fcis-2970	77	8	knowledge	knowledge	NOUN
fcis-2970	77	9	integration	integration	NOUN
fcis-2970	77	10	is	be	AUX
fcis-2970	77	11	accomplished	accomplish	VERB
fcis-2970	77	12	through	through	ADP
fcis-2970	77	13	the	the	DET
fcis-2970	77	14	multi	multi	ADJ
fcis-2970	77	15	-	-	ADJ
fcis-2970	77	16	stage	stage	ADJ
fcis-2970	77	17	knowledge	knowledge	NOUN
fcis-2970	77	18	mask	mask	NOUN
fcis-2970	77	19	strategy	strategy	NOUN
fcis-2970	77	20	to	to	PART
fcis-2970	77	21	obtain	obtain	VERB
fcis-2970	77	22	the	the	DET
fcis-2970	77	23	language	language	NOUN
fcis-2970	77	24	representation	representation	NOUN
fcis-2970	77	25	of	of	ADP
fcis-2970	77	26	the	the	DET
fcis-2970	77	27	fusion	fusion	NOUN
fcis-2970	77	28	phrase	phrase	NOUN
fcis-2970	77	29	level	level	NOUN
fcis-2970	77	30	and	and	CCONJ
fcis-2970	77	31	entity	entity	NOUN
fcis-2970	77	32	level	level	NOUN
fcis-2970	77	33	[	[	X
fcis-2970	77	34	15	15	NUM
fcis-2970	77	35	]	]	PUNCT
fcis-2970	77	36	.	.	PUNCT
fcis-2970	78	1	the	the	DET
fcis-2970	78	2	ernie	ernie	PROPN
fcis-2970	78	3	model	model	PROPN
fcis-2970	78	4	structure	structure	NOUN
fcis-2970	78	5	is	be	AUX
fcis-2970	78	6	shown	show	VERB
fcis-2970	78	7	in	in	ADP
fcis-2970	78	8	figure	figure	NOUN
fcis-2970	78	9	3	3	NUM
fcis-2970	78	10	below	below	ADV
fcis-2970	78	11	.	.	PUNCT
fcis-2970	79	1	figure	figure	VERB
fcis-2970	79	2	3	3	NUM
fcis-2970	79	3	.	.	PUNCT
fcis-2970	80	1	network	network	NOUN
fcis-2970	80	2	structure	structure	NOUN
fcis-2970	80	3	of	of	ADP
fcis-2970	80	4	ernie	ernie	PROPN
fcis-2970	80	5	model	model	PROPN
fcis-2970	80	6	the	the	DET
fcis-2970	80	7	input	input	NOUN
fcis-2970	80	8	vector	vector	NOUN
fcis-2970	80	9	is	be	AUX
fcis-2970	80	10	formed	form	VERB
fcis-2970	80	11	by	by	ADP
fcis-2970	80	12	the	the	DET
fcis-2970	80	13	corresponding	corresponding	ADJ
fcis-2970	80	14	addition	addition	NOUN
fcis-2970	80	15	of	of	ADP
fcis-2970	80	16	the	the	DET
fcis-2970	80	17	position	position	NOUN
fcis-2970	80	18	vector	vector	NOUN
fcis-2970	80	19	,	,	PUNCT
fcis-2970	80	20	sentence	sentence	NOUN
fcis-2970	80	21	vector	vector	NOUN
fcis-2970	80	22	,	,	PUNCT
fcis-2970	80	23	and	and	CCONJ
fcis-2970	80	24	word	word	NOUN
fcis-2970	80	25	embedding	embed	VERB
fcis-2970	80	26	vector	vector	NOUN
fcis-2970	80	27	,	,	PUNCT
fcis-2970	80	28	as	as	SCONJ
fcis-2970	80	29	shown	show	VERB
fcis-2970	80	30	in	in	ADP
fcis-2970	80	31	figure	figure	NOUN
fcis-2970	80	32	4	4	NUM
fcis-2970	80	33	below	below	ADV
fcis-2970	80	34	.	.	PUNCT
fcis-2970	81	1	figure	figure	VERB
fcis-2970	81	2	4	4	NUM
fcis-2970	81	3	.	.	PUNCT
fcis-2970	81	4	input	input	NOUN
fcis-2970	81	5	structure	structure	NOUN
fcis-2970	81	6	of	of	ADP
fcis-2970	81	7	the	the	DET
fcis-2970	81	8	ernie	ernie	PROPN
fcis-2970	81	9	model	model	NOUN
fcis-2970	81	10	as	as	SCONJ
fcis-2970	81	11	can	can	AUX
fcis-2970	81	12	be	be	AUX
fcis-2970	81	13	seen	see	VERB
fcis-2970	81	14	from	from	ADP
fcis-2970	81	15	the	the	DET
fcis-2970	81	16	ernie	ernie	PROPN
fcis-2970	81	17	model	model	PROPN
fcis-2970	81	18	structure	structure	PROPN
fcis-2970	81	19	diagram	diagram	PROPN
fcis-2970	81	20	,	,	PUNCT
fcis-2970	81	21	the	the	DET
fcis-2970	81	22	output	output	NOUN
fcis-2970	81	23	part	part	NOUN
fcis-2970	81	24	is	be	AUX
fcis-2970	81	25	the	the	DET
fcis-2970	81	26	word	word	NOUN
fcis-2970	81	27	vector	vector	NOUN
fcis-2970	81	28	combined	combine	VERB
fcis-2970	81	29	with	with	ADP
fcis-2970	81	30	the	the	DET
fcis-2970	81	31	sentence	sentence	NOUN
fcis-2970	81	32	context	context	NOUN
fcis-2970	81	33	.	.	PUNCT
fcis-2970	82	1	because	because	SCONJ
fcis-2970	82	2	the	the	DET
fcis-2970	82	3	bidirectional	bidirectional	ADJ
fcis-2970	82	4	language	language	NOUN
fcis-2970	82	5	model	model	NOUN
fcis-2970	82	6	will	will	AUX
fcis-2970	82	7	use	use	VERB
fcis-2970	82	8	contextual	contextual	ADJ
fcis-2970	82	9	information	information	NOUN
fcis-2970	82	10	when	when	SCONJ
fcis-2970	82	11	predicting	predict	VERB
fcis-2970	82	12	words	word	NOUN
fcis-2970	82	13	,	,	PUNCT
fcis-2970	82	14	and	and	CCONJ
fcis-2970	82	15	the	the	DET
fcis-2970	82	16	transformer	transformer	NOUN
fcis-2970	82	17	has	have	VERB
fcis-2970	82	18	its	its	PRON
fcis-2970	82	19	selfattention	selfattention	NOUN
fcis-2970	82	20	,	,	PUNCT
fcis-2970	82	21	when	when	SCONJ
fcis-2970	82	22	predicting	predict	VERB
fcis-2970	82	23	other	other	ADJ
fcis-2970	82	24	words	word	NOUN
fcis-2970	82	25	,	,	PUNCT
fcis-2970	82	26	the	the	DET
fcis-2970	82	27	information	information	NOUN
fcis-2970	82	28	of	of	ADP
fcis-2970	82	29	the	the	DET
fcis-2970	82	30	word	word	NOUN
fcis-2970	82	31	will	will	AUX
fcis-2970	82	32	also	also	ADV
fcis-2970	82	33	be	be	AUX
fcis-2970	82	34	included	include	VERB
fcis-2970	82	35	in	in	ADP
fcis-2970	82	36	the	the	DET
fcis-2970	82	37	network	network	NOUN
fcis-2970	82	38	parameters	parameter	NOUN
fcis-2970	82	39	of	of	ADP
fcis-2970	82	40	the	the	DET
fcis-2970	82	41	previous	previous	ADJ
fcis-2970	82	42	layer	layer	NOUN
fcis-2970	82	43	,	,	PUNCT
fcis-2970	82	44	resulting	result	VERB
fcis-2970	82	45	in	in	ADP
fcis-2970	82	46	leakage	leakage	NOUN
fcis-2970	82	47	of	of	ADP
fcis-2970	82	48	information	information	NOUN
fcis-2970	82	49	.	.	PUNCT
fcis-2970	83	1	to	to	PART
fcis-2970	83	2	solve	solve	VERB
fcis-2970	83	3	this	this	DET
fcis-2970	83	4	problem	problem	NOUN
fcis-2970	83	5	,	,	PUNCT
fcis-2970	83	6	the	the	DET
fcis-2970	83	7	bert	bert	PROPN
fcis-2970	83	8	model	model	PROPN
fcis-2970	83	9	randomly	randomly	ADV
fcis-2970	83	10	masks	mask	NOUN
fcis-2970	83	11	out	out	ADV
fcis-2970	83	12	about	about	ADP
fcis-2970	83	13	15	15	NUM
fcis-2970	83	14	%	%	NOUN
fcis-2970	83	15	of	of	ADP
fcis-2970	83	16	the	the	DET
fcis-2970	83	17	words	word	NOUN
fcis-2970	83	18	and	and	CCONJ
fcis-2970	83	19	hides	hide	VERB
fcis-2970	83	20	part	part	NOUN
fcis-2970	83	21	of	of	ADP
fcis-2970	83	22	the	the	DET
fcis-2970	83	23	input	input	NOUN
fcis-2970	83	24	sequence	sequence	NOUN
fcis-2970	83	25	.	.	PUNCT
fcis-2970	84	1	ernie	ernie	PROPN
fcis-2970	84	2	model	model	PROPN
fcis-2970	84	3	improves	improve	VERB
fcis-2970	84	4	the	the	DET
fcis-2970	84	5	bert	bert	PROPN
fcis-2970	84	6	model	model	NOUN
fcis-2970	84	7	,	,	PUNCT
fcis-2970	84	8	which	which	PRON
fcis-2970	84	9	can	can	AUX
fcis-2970	84	10	only	only	ADV
fcis-2970	84	11	cover	cover	VERB
fcis-2970	84	12	words	word	NOUN
fcis-2970	84	13	.	.	PUNCT
fcis-2970	85	1	in	in	ADP
fcis-2970	85	2	addition	addition	NOUN
fcis-2970	85	3	to	to	ADP
fcis-2970	85	4	covering	cover	VERB
fcis-2970	85	5	sequence	sequence	NOUN
fcis-2970	85	6	words	word	NOUN
fcis-2970	85	7	,	,	PUNCT
fcis-2970	85	8	the	the	DET
fcis-2970	85	9	ernie	ernie	PROPN
fcis-2970	85	10	model	model	PROPN
fcis-2970	85	11	also	also	ADV
fcis-2970	85	12	uses	use	VERB
fcis-2970	85	13	a	a	DET
fcis-2970	85	14	multi	multi	ADJ
fcis-2970	85	15	-	-	ADJ
fcis-2970	85	16	stage	stage	ADJ
fcis-2970	85	17	knowledge	knowledge	NOUN
fcis-2970	85	18	mask	mask	NOUN
fcis-2970	85	19	strategy	strategy	NOUN
fcis-2970	85	20	to	to	PART
fcis-2970	85	21	cover	cover	VERB
fcis-2970	85	22	phrases	phrase	NOUN
fcis-2970	85	23	and	and	CCONJ
fcis-2970	85	24	entities	entity	NOUN
fcis-2970	85	25	of	of	ADP
fcis-2970	85	26	sentences	sentence	NOUN
fcis-2970	85	27	.	.	PUNCT
fcis-2970	86	1	2.2.1	2.2.1	NUM
fcis-2970	86	2	.	.	PUNCT
fcis-2970	86	3	transformer	transformer	NOUN
fcis-2970	86	4	encoder	encoder	NOUN
fcis-2970	86	5	the	the	DET
fcis-2970	86	6	basic	basic	ADJ
fcis-2970	86	7	structure	structure	NOUN
fcis-2970	86	8	of	of	ADP
fcis-2970	86	9	the	the	DET
fcis-2970	86	10	ernie	ernie	PROPN
fcis-2970	86	11	model	model	NOUN
fcis-2970	86	12	is	be	AUX
fcis-2970	86	13	a	a	DET
fcis-2970	86	14	multi	multi	ADJ
fcis-2970	86	15	-	-	ADJ
fcis-2970	86	16	layer	layer	ADJ
fcis-2970	86	17	bidirectional	bidirectional	ADJ
fcis-2970	86	18	transformer	transformer	NOUN
fcis-2970	86	19	encoder	encoder	NOUN
fcis-2970	86	20	stack	stack	NOUN
fcis-2970	86	21	,	,	PUNCT
fcis-2970	86	22	where	where	SCONJ
fcis-2970	86	23	the	the	DET
fcis-2970	86	24	encoder	encoder	NOUN
fcis-2970	86	25	structure	structure	NOUN
fcis-2970	86	26	is	be	AUX
fcis-2970	86	27	the	the	DET
fcis-2970	86	28	same	same	ADJ
fcis-2970	86	29	,	,	PUNCT
fcis-2970	86	30	and	and	CCONJ
fcis-2970	86	31	the	the	DET
fcis-2970	86	32	weight	weight	NOUN
fcis-2970	86	33	is	be	AUX
fcis-2970	86	34	not	not	PART
fcis-2970	86	35	shared	share	VERB
fcis-2970	86	36	.	.	PUNCT
fcis-2970	87	1	the	the	DET
fcis-2970	87	2	selfattention	selfattention	NOUN
fcis-2970	87	3	mechanism	mechanism	NOUN
fcis-2970	87	4	will	will	AUX
fcis-2970	87	5	be	be	AUX
fcis-2970	87	6	used	use	VERB
fcis-2970	87	7	in	in	ADP
fcis-2970	87	8	the	the	DET
fcis-2970	87	9	encoder	encoder	NOUN
fcis-2970	87	10	to	to	PART
fcis-2970	87	11	integrate	integrate	VERB
fcis-2970	87	12	the	the	DET
fcis-2970	87	13	sentence	sentence	NOUN
fcis-2970	87	14	information	information	NOUN
fcis-2970	87	15	according	accord	VERB
fcis-2970	87	16	to	to	ADP
fcis-2970	87	17	the	the	DET
fcis-2970	87	18	importance	importance	NOUN
fcis-2970	87	19	of	of	ADP
fcis-2970	87	20	the	the	DET
fcis-2970	87	21	words	word	NOUN
fcis-2970	87	22	in	in	ADP
fcis-2970	87	23	the	the	DET
fcis-2970	87	24	sentence	sentence	NOUN
fcis-2970	87	25	,	,	PUNCT
fcis-2970	87	26	to	to	PART
fcis-2970	87	27	improve	improve	VERB
fcis-2970	87	28	the	the	DET
fcis-2970	87	29	utilization	utilization	NOUN
fcis-2970	87	30	of	of	ADP
fcis-2970	87	31	the	the	DET
fcis-2970	87	32	features	feature	NOUN
fcis-2970	87	33	.	.	PUNCT
fcis-2970	88	1	the	the	DET
fcis-2970	88	2	specific	specific	ADJ
fcis-2970	88	3	calculation	calculation	NOUN
fcis-2970	88	4	formula	formula	NOUN
fcis-2970	88	5	is	be	AUX
fcis-2970	88	6	as	as	SCONJ
fcis-2970	88	7	follows	follow	VERB
fcis-2970	88	8	:	:	PUNCT
fcis-2970	88	9	(	(	PUNCT
fcis-2970	88	10	,	,	PUNCT
fcis-2970	88	11	,	,	PUNCT
fcis-2970	88	12	)	)	PUNCT
fcis-2970	89	1	(	(	PUNCT
fcis-2970	89	2	)	)	PUNCT
fcis-2970	89	3	t	t	PROPN
fcis-2970	89	4	k	k	PROPN
fcis-2970	89	5	qk	qk	ADP
fcis-2970	89	6	attention	attention	NOUN
fcis-2970	89	7	q	q	PROPN
fcis-2970	90	1	k	k	PROPN
fcis-2970	90	2	v	v	X
fcis-2970	90	3	softmax	softmax	NOUN
fcis-2970	90	4	v	v	NOUN
fcis-2970	90	5	d	d	X
fcis-2970	90	6	=	=	PUNCT
fcis-2970	90	7	(	(	PUNCT
fcis-2970	90	8	1	1	NUM
fcis-2970	90	9	)	)	PUNCT
fcis-2970	90	10	where	where	SCONJ
fcis-2970	90	11	:	:	PUNCT
fcis-2970	90	12	,	,	PUNCT
fcis-2970	90	13	,	,	PUNCT
fcis-2970	90	14	q	q	PROPN
fcis-2970	90	15	k	k	NOUN
fcis-2970	90	16	v	v	ADJ
fcis-2970	90	17	input	input	NOUN
fcis-2970	90	18	word	word	NOUN
fcis-2970	90	19	vector	vector	NOUN
fcis-2970	90	20	matrix	matrix	NOUN
fcis-2970	90	21	;	;	PUNCT
fcis-2970	90	22	kd	kd	PROPN
fcis-2970	90	23	input	input	VERB
fcis-2970	90	24	vector	vector	NOUN
fcis-2970	90	25	dimensions	dimension	NOUN
fcis-2970	90	26	.	.	PUNCT
fcis-2970	91	1	2.2.2	2.2.2	X
fcis-2970	91	2	.	.	PUNCT
fcis-2970	91	3	knowledge	knowledge	NOUN
fcis-2970	91	4	integration	integration	NOUN
fcis-2970	91	5	different	different	ADJ
fcis-2970	91	6	from	from	ADP
fcis-2970	91	7	bert	bert	PROPN
fcis-2970	91	8	's	's	PART
fcis-2970	91	9	pre	pre	ADJ
fcis-2970	91	10	-	-	ADJ
fcis-2970	91	11	training	training	ADJ
fcis-2970	91	12	model	model	NOUN
fcis-2970	91	13	,	,	PUNCT
fcis-2970	91	14	ernie	ernie	PROPN
fcis-2970	92	1	[	[	X
fcis-2970	92	2	14	14	NUM
fcis-2970	92	3	]	]	PUNCT
fcis-2970	92	4	can	can	AUX
fcis-2970	92	5	make	make	VERB
fcis-2970	92	6	good	good	ADJ
fcis-2970	92	7	use	use	NOUN
fcis-2970	92	8	of	of	ADP
fcis-2970	92	9	the	the	DET
fcis-2970	92	10	lexical	lexical	ADJ
fcis-2970	92	11	,	,	PUNCT
fcis-2970	92	12	grammatical	grammatical	ADJ
fcis-2970	92	13	structure	structure	NOUN
fcis-2970	92	14	,	,	PUNCT
fcis-2970	92	15	and	and	CCONJ
fcis-2970	92	16	semantic	semantic	ADJ
fcis-2970	92	17	information	information	NOUN
fcis-2970	92	18	in	in	ADP
fcis-2970	92	19	the	the	DET
fcis-2970	92	20	training	training	NOUN
fcis-2970	92	21	data	datum	NOUN
fcis-2970	92	22	and	and	CCONJ
fcis-2970	92	23	propose	propose	VERB
fcis-2970	92	24	multi	multi	ADJ
fcis-2970	92	25	-	-	ADJ
fcis-2970	92	26	stage	stage	ADJ
fcis-2970	92	27	knowledge	knowledge	NOUN
fcis-2970	92	28	masking	masking	NOUN
fcis-2970	92	29	strategies	strategy	NOUN
fcis-2970	92	30	to	to	PART
fcis-2970	92	31	integrate	integrate	VERB
fcis-2970	92	32	the	the	DET
fcis-2970	92	33	phrase	phrase	NOUN
fcis-2970	92	34	and	and	CCONJ
fcis-2970	92	35	entity	entity	NOUN
fcis-2970	92	36	level	level	NOUN
fcis-2970	92	37	knowledge	knowledge	NOUN
fcis-2970	92	38	into	into	ADP
fcis-2970	92	39	the	the	DET
fcis-2970	92	40	language	language	NOUN
fcis-2970	92	41	representation	representation	NOUN
fcis-2970	92	42	.	.	PUNCT
fcis-2970	93	1	instead	instead	ADV
fcis-2970	93	2	of	of	ADP
fcis-2970	93	3	directly	directly	ADV
fcis-2970	93	4	adding	add	VERB
fcis-2970	93	5	knowledge	knowledge	NOUN
fcis-2970	93	6	embedding	embed	VERB
fcis-2970	93	7	,	,	PUNCT
fcis-2970	93	8	this	this	PRON
fcis-2970	93	9	greatly	greatly	ADV
fcis-2970	93	10	enhances	enhance	VERB
fcis-2970	93	11	the	the	DET
fcis-2970	93	12	syntactic	syntactic	ADJ
fcis-2970	93	13	and	and	CCONJ
fcis-2970	93	14	grammatical	grammatical	ADJ
fcis-2970	93	15	representation	representation	NOUN
fcis-2970	93	16	ability	ability	NOUN
fcis-2970	93	17	of	of	ADP
fcis-2970	93	18	word	word	NOUN
fcis-2970	93	19	vectors	vector	NOUN
fcis-2970	93	20	.	.	PUNCT
fcis-2970	94	1	in	in	ADP
fcis-2970	94	2	addition	addition	NOUN
fcis-2970	94	3	to	to	ADP
fcis-2970	94	4	the	the	DET
fcis-2970	94	5	bert	bert	PROPN
fcis-2970	94	6	model	model	NOUN
fcis-2970	94	7	which	which	PRON
fcis-2970	94	8	only	only	ADV
fcis-2970	94	9	provides	provide	VERB
fcis-2970	94	10	basic	basic	ADJ
fcis-2970	94	11	-	-	PUNCT
fcis-2970	94	12	level	level	NOUN
fcis-2970	94	13	masking	masking	NOUN
fcis-2970	94	14	,	,	PUNCT
fcis-2970	94	15	it	it	PRON
fcis-2970	94	16	adds	add	VERB
fcis-2970	94	17	phrase	phrase	NOUN
fcis-2970	94	18	-	-	PUNCT
fcis-2970	94	19	based	base	VERB
fcis-2970	94	20	masking	masking	NOUN
fcis-2970	94	21	and	and	CCONJ
fcis-2970	94	22	entity	entity	NOUN
fcis-2970	94	23	-	-	PUNCT
fcis-2970	94	24	level	level	NOUN
fcis-2970	94	25	masking	masking	NOUN
fcis-2970	94	26	.	.	PUNCT
fcis-2970	95	1	the	the	DET
fcis-2970	95	2	example	example	NOUN
fcis-2970	95	3	of	of	ADP
fcis-2970	95	4	the	the	DET
fcis-2970	95	5	knowledge	knowledge	NOUN
fcis-2970	95	6	mask	mask	NOUN
fcis-2970	95	7	policy	policy	NOUN
fcis-2970	95	8	used	use	VERB
fcis-2970	95	9	in	in	ADP
fcis-2970	95	10	ernie	ernie	PROPN
fcis-2970	95	11	is	be	AUX
fcis-2970	95	12	shown	show	VERB
fcis-2970	95	13	in	in	ADP
fcis-2970	95	14	figure	figure	NOUN
fcis-2970	95	15	5[17	5[17	NUM
fcis-2970	95	16	]	]	PUNCT
fcis-2970	95	17	.	.	PUNCT
fcis-2970	96	1	figure	figure	NOUN
fcis-2970	96	2	5	5	NUM
fcis-2970	96	3	.	.	NOUN
fcis-2970	96	4	example	example	NOUN
fcis-2970	96	5	of	of	ADP
fcis-2970	96	6	ernie	ernie	PROPN
fcis-2970	96	7	knowledge	knowledge	PROPN
fcis-2970	96	8	mask	mask	PROPN
fcis-2970	96	9	strategy	strategy	NOUN
fcis-2970	96	10	2.3	2.3	NUM
fcis-2970	96	11	.	.	PUNCT
fcis-2970	97	1	text	text	NOUN
fcis-2970	97	2	semantic	semantic	ADJ
fcis-2970	97	3	extraction	extraction	NOUN
fcis-2970	97	4	bilstm	bilstm	NOUN
fcis-2970	97	5	layer	layer	NOUN
fcis-2970	97	6	and	and	CCONJ
fcis-2970	97	7	attention	attention	NOUN
fcis-2970	97	8	mechanism	mechanism	NOUN
fcis-2970	97	9	the	the	DET
fcis-2970	97	10	main	main	ADJ
fcis-2970	97	11	function	function	NOUN
fcis-2970	97	12	of	of	ADP
fcis-2970	97	13	the	the	DET
fcis-2970	97	14	bilstm	bilstm	NOUN
fcis-2970	97	15	layer	layer	NOUN
fcis-2970	97	16	is	be	AUX
fcis-2970	97	17	to	to	PART
fcis-2970	97	18	extract	extract	VERB
fcis-2970	97	19	text	text	NOUN
fcis-2970	97	20	features	feature	NOUN
fcis-2970	97	21	and	and	CCONJ
fcis-2970	97	22	construct	construct	VERB
fcis-2970	97	23	two	two	NUM
fcis-2970	97	24	lstm	lstm	NOUN
fcis-2970	97	25	networks	network	NOUN
fcis-2970	97	26	that	that	PRON
fcis-2970	97	27	are	be	AUX
fcis-2970	97	28	good	good	ADJ
fcis-2970	97	29	at	at	ADP
fcis-2970	97	30	dealing	deal	VERB
fcis-2970	97	31	with	with	ADP
fcis-2970	97	32	long	long	ADJ
fcis-2970	97	33	dependencies	dependency	NOUN
fcis-2970	97	34	to	to	PART
fcis-2970	97	35	obtain	obtain	VERB
fcis-2970	97	36	forward	forward	ADV
fcis-2970	97	37	and	and	CCONJ
fcis-2970	97	38	reverse	reverse	VERB
fcis-2970	97	39	text	text	NOUN
fcis-2970	97	40	information	information	NOUN
fcis-2970	97	41	,	,	PUNCT
fcis-2970	97	42	which	which	PRON
fcis-2970	97	43	is	be	AUX
fcis-2970	97	44	easier	easy	ADJ
fcis-2970	97	45	to	to	PART
fcis-2970	97	46	extract	extract	VERB
fcis-2970	97	47	the	the	DET
fcis-2970	97	48	deep	deep	ADJ
fcis-2970	97	49	semantic	semantic	ADJ
fcis-2970	97	50	expression	expression	NOUN
fcis-2970	97	51	hidden	hide	VERB
fcis-2970	97	52	in	in	ADP
fcis-2970	97	53	the	the	DET
fcis-2970	97	54	text	text	NOUN
fcis-2970	97	55	.	.	PUNCT
fcis-2970	98	1	compared	compare	VERB
fcis-2970	98	2	with	with	ADP
fcis-2970	98	3	traditional	traditional	ADJ
fcis-2970	98	4	rnn	rnn	NOUN
fcis-2970	98	5	,	,	PUNCT
fcis-2970	98	6	lstm	lstm	NOUN
fcis-2970	98	7	can	can	AUX
fcis-2970	98	8	deal	deal	VERB
fcis-2970	98	9	with	with	ADP
fcis-2970	98	10	long	long	ADJ
fcis-2970	98	11	-	-	PUNCT
fcis-2970	98	12	term	term	NOUN
fcis-2970	98	13	dependence	dependence	NOUN
fcis-2970	98	14	and	and	CCONJ
fcis-2970	98	15	gradient	gradient	ADJ
fcis-2970	98	16	disappearance	disappearance	NOUN
fcis-2970	98	17	mainly	mainly	ADV
fcis-2970	98	18	due	due	ADP
fcis-2970	98	19	to	to	ADP
fcis-2970	98	20	its	its	PRON
fcis-2970	98	21	internal	internal	ADJ
fcis-2970	98	22	three	three	NUM
fcis-2970	98	23	gate	gate	NOUN
fcis-2970	98	24	functions	function	NOUN
fcis-2970	98	25	:	:	PUNCT
fcis-2970	98	26	input	input	NOUN
fcis-2970	98	27	gate	gate	NOUN
fcis-2970	98	28	,	,	PUNCT
fcis-2970	98	29	forget	forget	VERB
fcis-2970	98	30	gate	gate	NOUN
fcis-2970	98	31	,	,	PUNCT
fcis-2970	98	32	and	and	CCONJ
fcis-2970	98	33	output	output	NOUN
fcis-2970	98	34	gate	gate	NOUN
fcis-2970	98	35	.	.	PUNCT
fcis-2970	99	1	these	these	DET
fcis-2970	99	2	three	three	NUM
fcis-2970	99	3	functions	function	NOUN
fcis-2970	99	4	are	be	AUX
fcis-2970	99	5	used	use	VERB
fcis-2970	99	6	to	to	PART
fcis-2970	99	7	control	control	VERB
fcis-2970	99	8	information	information	NOUN
fcis-2970	99	9	retention	retention	NOUN
fcis-2970	99	10	.	.	PUNCT
fcis-2970	100	1	the	the	DET
fcis-2970	100	2	model	model	NOUN
fcis-2970	100	3	structure	structure	NOUN
fcis-2970	100	4	is	be	AUX
fcis-2970	100	5	shown	show	VERB
fcis-2970	100	6	in	in	ADP
fcis-2970	100	7	figure	figure	NOUN
fcis-2970	100	8	6	6	NUM
fcis-2970	100	9	below	below	ADV
fcis-2970	100	10	.	.	PUNCT
fcis-2970	101	1	figure	figure	VERB
fcis-2970	101	2	6	6	NUM
fcis-2970	101	3	.	.	PUNCT
fcis-2970	102	1	lstm	lstm	NOUN
fcis-2970	102	2	model	model	NOUN
fcis-2970	102	3	structure	structure	NOUN
fcis-2970	102	4	the	the	DET
fcis-2970	102	5	forward	forward	ADJ
fcis-2970	102	6	calculation	calculation	NOUN
fcis-2970	102	7	process	process	NOUN
fcis-2970	102	8	of	of	ADP
fcis-2970	102	9	a	a	DET
fcis-2970	102	10	single	single	ADJ
fcis-2970	102	11	lstm	lstm	NOUN
fcis-2970	102	12	memory	memory	NOUN
fcis-2970	102	13	unit	unit	NOUN
fcis-2970	102	14	at	at	ADP
fcis-2970	102	15	the	the	DET
fcis-2970	102	16	moment	moment	NOUN
fcis-2970	102	17	is	be	AUX
fcis-2970	102	18	as	as	SCONJ
fcis-2970	102	19	follows	follow	VERB
fcis-2970	102	20	:	:	PUNCT
fcis-2970	102	21	(	(	PUNCT
fcis-2970	102	22	1	1	X
fcis-2970	102	23	)	)	PUNCT
fcis-2970	102	24	input	input	NOUN
fcis-2970	102	25	unit	unit	NOUN
fcis-2970	102	26	,	,	PUNCT
fcis-2970	102	27	processing	process	VERB
fcis-2970	102	28	the	the	DET
fcis-2970	102	29	input	input	NOUN
fcis-2970	102	30	of	of	ADP
fcis-2970	102	31	the	the	DET
fcis-2970	102	32	current	current	ADJ
fcis-2970	102	33	sequence	sequence	NOUN
fcis-2970	102	34	position	position	NOUN
fcis-2970	102	35	:	:	PUNCT
fcis-2970	102	36	1	1	NUM
fcis-2970	102	37	(	(	PUNCT
fcis-2970	102	38	[	[	PUNCT
fcis-2970	102	39	,	,	PUNCT
fcis-2970	102	40	]	]	PUNCT
fcis-2970	102	41	)	)	PUNCT
fcis-2970	103	1	t	t	NOUN
fcis-2970	104	1	i	i	PRON
fcis-2970	104	2	t	t	PROPN
fcis-2970	104	3	t	t	PROPN
fcis-2970	104	4	ii	ii	PROPN
fcis-2970	104	5	w	w	PROPN
fcis-2970	104	6	h	h	PROPN
fcis-2970	104	7	x	x	PROPN
fcis-2970	104	8	b	b	PROPN
fcis-2970	104	9	−=	−=	NOUN
fcis-2970	104	10			NOUN
fcis-2970	104	11	+	+	CCONJ
fcis-2970	104	12	(	(	PUNCT
fcis-2970	104	13	2	2	NUM
fcis-2970	104	14	)	)	PUNCT
fcis-2970	104	15	1	1	NUM
fcis-2970	104	16	1tanh	1tanh	NUM
fcis-2970	104	17	(	(	PUNCT
fcis-2970	104	18	[	[	PUNCT
fcis-2970	104	19	,	,	PUNCT
fcis-2970	104	20	]	]	PUNCT
fcis-2970	104	21	)	)	PUNCT
fcis-2970	104	22	,	,	PUNCT
fcis-2970	104	23	t	t	PROPN
fcis-2970	104	24	c	c	PROPN
fcis-2970	104	25	t	t	PROPN
fcis-2970	104	26	t	t	PROPN
fcis-2970	104	27	c	c	PROPN
fcis-2970	104	28	t	t	PROPN
fcis-2970	104	29	t	t	PROPN
fcis-2970	104	30	t	t	PROPN
fcis-2970	104	31	t	t	PROPN
fcis-2970	104	32	tc	tc	NUM
fcis-2970	104	33	w	w	PROPN
fcis-2970	104	34	h	h	NOUN
fcis-2970	104	35	x	x	X
fcis-2970	105	1	b	b	NOUN
fcis-2970	105	2	c	c	NOUN
fcis-2970	106	1	f	f	NOUN
fcis-2970	106	2	c	c	NOUN
fcis-2970	106	3	i	i	PRON
fcis-2970	106	4	c−	c−	VERB
fcis-2970	106	5	−=	−=	PRON
fcis-2970	106	6			NOUN
fcis-2970	107	1	+	+	CCONJ
fcis-2970	107	2	=	=	SYM
fcis-2970	107	3			PROPN
fcis-2970	107	4	+	+	CCONJ
fcis-2970	107	5			PROPN
fcis-2970	107	6	(	(	PUNCT
fcis-2970	107	7	3	3	NUM
fcis-2970	107	8	)	)	PUNCT
fcis-2970	107	9	(	(	PUNCT
fcis-2970	107	10	2	2	X
fcis-2970	107	11	)	)	PUNCT
fcis-2970	107	12	forgetting	forget	VERB
fcis-2970	107	13	unit	unit	NOUN
fcis-2970	107	14	,	,	PUNCT
fcis-2970	107	15	forgetting	forget	VERB
fcis-2970	107	16	information	information	NOUN
fcis-2970	107	17	:	:	PUNCT
fcis-2970	107	18	73	73	NUM
fcis-2970	107	19	1	1	NUM
fcis-2970	107	20	(	(	PUNCT
fcis-2970	107	21	[	[	PUNCT
fcis-2970	107	22	,	,	PUNCT
fcis-2970	107	23	]	]	PUNCT
fcis-2970	107	24	)	)	PUNCT
fcis-2970	108	1	t	t	PROPN
fcis-2970	108	2	f	f	PROPN
fcis-2970	108	3	t	t	PROPN
fcis-2970	108	4	t	t	PROPN
fcis-2970	108	5	ff	ff	NOUN
fcis-2970	108	6	w	w	PROPN
fcis-2970	108	7	h	h	PROPN
fcis-2970	108	8	x	x	PROPN
fcis-2970	108	9	b	b	PROPN
fcis-2970	108	10	−=	−=	NOUN
fcis-2970	108	11			NOUN
fcis-2970	108	12	+	+	CCONJ
fcis-2970	108	13	(	(	PUNCT
fcis-2970	108	14	4	4	NUM
fcis-2970	108	15	)	)	PUNCT
fcis-2970	108	16	(	(	PUNCT
fcis-2970	108	17	3	3	X
fcis-2970	108	18	)	)	PUNCT
fcis-2970	108	19	update	update	NOUN
fcis-2970	108	20	the	the	DET
fcis-2970	108	21	unit	unit	NOUN
fcis-2970	108	22	,	,	PUNCT
fcis-2970	108	23	update	update	VERB
fcis-2970	108	24	the	the	DET
fcis-2970	108	25	status	status	NOUN
fcis-2970	108	26	of	of	ADP
fcis-2970	108	27	the	the	DET
fcis-2970	108	28	unit	unit	NOUN
fcis-2970	108	29	after	after	ADP
fcis-2970	108	30	the	the	DET
fcis-2970	108	31	abandoned	abandon	VERB
fcis-2970	108	32	information	information	NOUN
fcis-2970	108	33	:	:	PUNCT
fcis-2970	108	34	1	1	NUM
fcis-2970	108	35	t	t	NOUN
fcis-2970	108	36	t	t	PROPN
fcis-2970	108	37	t	t	PROPN
fcis-2970	108	38	t	t	PROPN
fcis-2970	109	1	tc	tc	NOUN
fcis-2970	109	2	f	f	PROPN
fcis-2970	110	1	c	c	VERB
fcis-2970	110	2	i	i	PRON
fcis-2970	110	3	c−=	c−=	NOUN
fcis-2970	110	4			VERB
fcis-2970	110	5	+	+	CCONJ
fcis-2970	110	6			PROPN
fcis-2970	110	7	(	(	PUNCT
fcis-2970	110	8	5	5	NUM
fcis-2970	110	9	)	)	PUNCT
fcis-2970	110	10	(	(	PUNCT
fcis-2970	110	11	4	4	X
fcis-2970	110	12	)	)	PUNCT
fcis-2970	110	13	output	output	NOUN
fcis-2970	110	14	unit	unit	NOUN
fcis-2970	110	15	,	,	PUNCT
fcis-2970	110	16	determine	determine	VERB
fcis-2970	110	17	the	the	DET
fcis-2970	110	18	output	output	NOUN
fcis-2970	110	19	value	value	NOUN
fcis-2970	110	20	:	:	PUNCT
fcis-2970	110	21	1	1	NUM
fcis-2970	110	22	(	(	PUNCT
fcis-2970	110	23	[	[	PUNCT
fcis-2970	110	24	,	,	PUNCT
fcis-2970	110	25	]	]	PUNCT
fcis-2970	110	26	)	)	PUNCT
fcis-2970	111	1	t	t	NOUN
fcis-2970	111	2	o	o	X
fcis-2970	111	3	t	t	NOUN
fcis-2970	111	4	t	t	X
fcis-2970	111	5	oo	oo	INTJ
fcis-2970	111	6	w	w	PROPN
fcis-2970	111	7	h	h	NOUN
fcis-2970	111	8	x	x	PROPN
fcis-2970	111	9	b	b	PROPN
fcis-2970	111	10	−=	−=	NOUN
fcis-2970	111	11			NOUN
fcis-2970	111	12	+	+	CCONJ
fcis-2970	111	13	(	(	PUNCT
fcis-2970	111	14	6	6	NUM
fcis-2970	111	15	)	)	PUNCT
fcis-2970	111	16	tanh	tanh	NOUN
fcis-2970	111	17	(	(	PUNCT
fcis-2970	111	18	)	)	PUNCT
fcis-2970	111	19	t	t	NOUN
fcis-2970	111	20	t	t	NOUN
fcis-2970	111	21	tz	tz	NOUN
fcis-2970	111	22	o	o	NOUN
fcis-2970	111	23	c=	c=	NOUN
fcis-2970	111	24			NOUN
fcis-2970	111	25	(	(	PUNCT
fcis-2970	111	26	7	7	NUM
fcis-2970	111	27	)	)	PUNCT
fcis-2970	111	28	where	where	SCONJ
fcis-2970	111	29	:	:	PUNCT
fcis-2970	111	30			PROPN
fcis-2970	111	31	,w	,w	PROPN
fcis-2970	111	32	b	b	PROPN
fcis-2970	112	1			AUX
fcis-2970	112	2	the	the	DET
fcis-2970	112	3	parameters	parameter	NOUN
fcis-2970	112	4	to	to	PART
fcis-2970	112	5	be	be	AUX
fcis-2970	112	6	trained	train	VERB
fcis-2970	112	7	;	;	PUNCT
fcis-2970	112	8	1th	1th	ADJ
fcis-2970	112	9	−	−	NOUN
fcis-2970	112	10	:	:	PUNCT
fcis-2970	112	11	the	the	DET
fcis-2970	112	12	previous	previous	ADJ
fcis-2970	112	13	time	time	NOUN
fcis-2970	112	14	of	of	ADP
fcis-2970	112	15	t	t	PROPN
fcis-2970	112	16	hides	hide	VERB
fcis-2970	112	17	the	the	DET
fcis-2970	112	18	state	state	NOUN
fcis-2970	112	19	of	of	ADP
fcis-2970	112	20	the	the	DET
fcis-2970	112	21	layer	layer	NOUN
fcis-2970	112	22	;	;	PUNCT
fcis-2970	112	23	tx	tx	X
fcis-2970	112	24	:	:	PUNCT
fcis-2970	112	25	the	the	DET
fcis-2970	112	26	input	input	NOUN
fcis-2970	112	27	at	at	ADP
fcis-2970	112	28	time	time	NOUN
fcis-2970	112	29	t.	t.	PROPN
fcis-2970	112	30	in	in	ADP
fcis-2970	112	31	language	language	NOUN
fcis-2970	112	32	expression	expression	NOUN
fcis-2970	112	33	,	,	PUNCT
fcis-2970	112	34	the	the	DET
fcis-2970	112	35	semantic	semantic	ADJ
fcis-2970	112	36	meaning	meaning	NOUN
fcis-2970	112	37	of	of	ADP
fcis-2970	112	38	words	word	NOUN
fcis-2970	112	39	in	in	ADP
fcis-2970	112	40	sentences	sentence	NOUN
fcis-2970	112	41	is	be	AUX
fcis-2970	112	42	affected	affect	VERB
fcis-2970	112	43	by	by	ADP
fcis-2970	112	44	the	the	DET
fcis-2970	112	45	context	context	NOUN
fcis-2970	112	46	,	,	PUNCT
fcis-2970	112	47	so	so	ADV
fcis-2970	112	48	it	it	PRON
fcis-2970	112	49	is	be	AUX
fcis-2970	112	50	necessary	necessary	ADJ
fcis-2970	112	51	to	to	PART
fcis-2970	112	52	extract	extract	VERB
fcis-2970	112	53	sentence	sentence	NOUN
fcis-2970	112	54	features	feature	NOUN
fcis-2970	112	55	in	in	ADP
fcis-2970	112	56	both	both	CCONJ
fcis-2970	112	57	positive	positive	ADJ
fcis-2970	112	58	and	and	CCONJ
fcis-2970	112	59	negative	negative	ADJ
fcis-2970	112	60	directions	direction	NOUN
fcis-2970	112	61	.	.	PUNCT
fcis-2970	113	1	bilstm	bilstm	NOUN
fcis-2970	113	2	uses	use	VERB
fcis-2970	113	3	lstm	lstm	PROPN
fcis-2970	113	4	to	to	PART
fcis-2970	113	5	extract	extract	VERB
fcis-2970	113	6	features	feature	NOUN
fcis-2970	113	7	in	in	ADP
fcis-2970	113	8	both	both	CCONJ
fcis-2970	113	9	positive	positive	ADJ
fcis-2970	113	10	and	and	CCONJ
fcis-2970	113	11	negative	negative	ADJ
fcis-2970	113	12	directions	direction	NOUN
fcis-2970	113	13	of	of	ADP
fcis-2970	113	14	text	text	NOUN
fcis-2970	113	15	,	,	PUNCT
fcis-2970	113	16	and	and	CCONJ
fcis-2970	113	17	finally	finally	ADV
fcis-2970	113	18	integrates	integrate	NOUN
fcis-2970	113	19	and	and	CCONJ
fcis-2970	113	20	outputs	output	NOUN
fcis-2970	113	21	the	the	DET
fcis-2970	113	22	hidden	hide	VERB
fcis-2970	113	23	layer	layer	NOUN
fcis-2970	113	24	results	result	NOUN
fcis-2970	113	25	of	of	ADP
fcis-2970	113	26	both	both	DET
fcis-2970	113	27	directions	direction	NOUN
fcis-2970	113	28	.	.	PUNCT
fcis-2970	114	1	the	the	DET
fcis-2970	114	2	calculation	calculation	NOUN
fcis-2970	114	3	process	process	NOUN
fcis-2970	114	4	is	be	AUX
fcis-2970	114	5	as	as	SCONJ
fcis-2970	114	6	follows	follow	VERB
fcis-2970	114	7	:	:	PUNCT
fcis-2970	114	8	[	[	PUNCT
fcis-2970	114	9	,	,	PUNCT
fcis-2970	114	10	]	]	X
fcis-2970	114	11	t	t	X
fcis-2970	114	12	t	t	PROPN
fcis-2970	114	13	th	th	X
fcis-2970	114	14	h	h	PROPN
fcis-2970	114	15	h=	h=	X
fcis-2970	114	16	(	(	PUNCT
fcis-2970	114	17	8)	8)	NUM
fcis-2970	114	18			PROPN
fcis-2970	114	19	1	1	PROPN
fcis-2970	114	20	,	,	PUNCT
fcis-2970	114	21	....	....	PUNCT
fcis-2970	114	22	,	,	PUNCT
fcis-2970	114	23	,	,	PUNCT
fcis-2970	114	24	....	....	PUNCT
fcis-2970	114	25	,	,	PUNCT
fcis-2970	115	1	t	t	PROPN
fcis-2970	115	2	th	th	X
fcis-2970	115	3	h	h	PROPN
fcis-2970	115	4	h	h	PROPN
fcis-2970	115	5	h=	h=	NOUN
fcis-2970	115	6	(	(	PUNCT
fcis-2970	115	7	9	9	NUM
fcis-2970	115	8	)	)	PUNCT
fcis-2970	115	9	where	where	SCONJ
fcis-2970	115	10	:	:	PUNCT
fcis-2970	115	11	h	h	NOUN
fcis-2970	115	12	:	:	PUNCT
fcis-2970	115	13	forward	forward	ADV
fcis-2970	115	14	hidden	hide	VERB
fcis-2970	115	15	layer	layer	NOUN
fcis-2970	115	16	vector	vector	NOUN
fcis-2970	115	17	;	;	PUNCT
fcis-2970	115	18	h	h	NOUN
fcis-2970	115	19	:	:	PUNCT
fcis-2970	115	20	backward	backward	ADJ
fcis-2970	115	21	hidden	hide	VERB
fcis-2970	115	22	layer	layer	NOUN
fcis-2970	115	23	vector	vector	NOUN
fcis-2970	115	24	;	;	PUNCT
fcis-2970	115	25	th	th	X
fcis-2970	115	26	:	:	PUNCT
fcis-2970	115	27	hidden	hide	VERB
fcis-2970	115	28	layer	layer	NOUN
fcis-2970	115	29	vector	vector	NOUN
fcis-2970	115	30	at	at	ADP
fcis-2970	115	31	time	time	NOUN
fcis-2970	115	32	t	t	PROPN
fcis-2970	115	33	;	;	PUNCT
fcis-2970	115	34	h	h	NOUN
fcis-2970	115	35	:	:	PUNCT
fcis-2970	115	36	hidden	hide	VERB
fcis-2970	115	37	layer	layer	NOUN
fcis-2970	115	38	vector	vector	NOUN
fcis-2970	115	39	at	at	ADP
fcis-2970	115	40	all	all	DET
fcis-2970	115	41	times	time	NOUN
fcis-2970	115	42	.	.	PUNCT
fcis-2970	116	1	passing	pass	VERB
fcis-2970	116	2	through	through	ADP
fcis-2970	116	3	the	the	DET
fcis-2970	116	4	bilstm	bilstm	NOUN
fcis-2970	116	5	layer	layer	NOUN
fcis-2970	116	6	will	will	AUX
fcis-2970	116	7	obtain	obtain	VERB
fcis-2970	116	8	the	the	DET
fcis-2970	116	9	hidden	hide	VERB
fcis-2970	116	10	layer	layer	NOUN
fcis-2970	116	11	vector	vector	NOUN
fcis-2970	116	12	at	at	ADP
fcis-2970	116	13	all	all	DET
fcis-2970	116	14	times	time	NOUN
fcis-2970	116	15	as	as	ADP
fcis-2970	116	16	input	input	NOUN
fcis-2970	116	17	to	to	ADP
fcis-2970	116	18	the	the	DET
fcis-2970	116	19	attention	attention	NOUN
fcis-2970	116	20	mechanism	mechanism	NOUN
fcis-2970	116	21	layer	layer	NOUN
fcis-2970	116	22	.	.	PUNCT
fcis-2970	117	1	initialize	initialize	VERB
fcis-2970	117	2	gw	gw	PROPN
fcis-2970	117	3	,	,	PUNCT
fcis-2970	117	4	multiply	multiply	VERB
fcis-2970	117	5	gw	gw	PROPN
fcis-2970	117	6	by	by	ADP
fcis-2970	117	7	each	each	DET
fcis-2970	117	8	th	th	PRON
fcis-2970	117	9	respectively	respectively	ADV
fcis-2970	117	10	,	,	PUNCT
fcis-2970	117	11	and	and	CCONJ
fcis-2970	117	12	then	then	ADV
fcis-2970	117	13	obtain	obtain	VERB
fcis-2970	117	14	tu	tu	PROPN
fcis-2970	117	15	through	through	ADP
fcis-2970	117	16	the	the	DET
fcis-2970	117	17	tanh	tanh	PROPN
fcis-2970	117	18	function	function	NOUN
fcis-2970	117	19	.	.	PUNCT
fcis-2970	118	1	then	then	ADV
fcis-2970	118	2	obtain	obtain	VERB
fcis-2970	118	3	the	the	DET
fcis-2970	118	4	corresponding	corresponding	ADJ
fcis-2970	118	5	weight	weight	NOUN
fcis-2970	118	6	value	value	NOUN
fcis-2970	118	7	ta	ta	ADP
fcis-2970	118	8	of	of	ADP
fcis-2970	118	9	each	each	DET
fcis-2970	118	10	hidden	hide	VERB
fcis-2970	118	11	layer	layer	NOUN
fcis-2970	118	12	through	through	ADP
fcis-2970	118	13	weight	weight	NOUN
fcis-2970	118	14	calculation	calculation	NOUN
fcis-2970	118	15	.	.	PUNCT
fcis-2970	119	1	finally	finally	ADV
fcis-2970	119	2	,	,	PUNCT
fcis-2970	119	3	carry	carry	VERB
fcis-2970	119	4	out	out	ADP
fcis-2970	119	5	weighted	weight	VERB
fcis-2970	119	6	summation	summation	NOUN
fcis-2970	119	7	of	of	ADP
fcis-2970	119	8	the	the	DET
fcis-2970	119	9	hidden	hide	VERB
fcis-2970	119	10	layer	layer	NOUN
fcis-2970	119	11	vectors	vector	NOUN
fcis-2970	119	12	at	at	ADP
fcis-2970	119	13	all	all	DET
fcis-2970	119	14	times	time	NOUN
fcis-2970	119	15	to	to	PART
fcis-2970	119	16	obtain	obtain	VERB
fcis-2970	119	17	the	the	DET
fcis-2970	119	18	sentence	sentence	NOUN
fcis-2970	119	19	feature	feature	NOUN
fcis-2970	119	20	representing	represent	VERB
fcis-2970	119	21	r	r	NOUN
fcis-2970	119	22	.	.	PUNCT
fcis-2970	120	1	the	the	DET
fcis-2970	120	2	calculation	calculation	NOUN
fcis-2970	120	3	process	process	NOUN
fcis-2970	120	4	is	be	AUX
fcis-2970	120	5	as	as	SCONJ
fcis-2970	120	6	follows	follow	VERB
fcis-2970	120	7	:	:	PUNCT
fcis-2970	120	8	tanh	tanh	NOUN
fcis-2970	120	9	(	(	PUNCT
fcis-2970	120	10	)	)	PUNCT
fcis-2970	120	11	t	t	PROPN
fcis-2970	120	12	g	g	PROPN
fcis-2970	120	13	t	t	PROPN
fcis-2970	120	14	gu	gu	PROPN
fcis-2970	120	15	w	w	PROPN
fcis-2970	120	16	h	h	PROPN
fcis-2970	120	17	b=	b=	NOUN
fcis-2970	121	1			PROPN
fcis-2970	121	2	+	+	CCONJ
fcis-2970	121	3	(	(	PUNCT
fcis-2970	121	4	10	10	NUM
fcis-2970	121	5	)	)	PUNCT
fcis-2970	121	6	t	t	NOUN
fcis-2970	121	7	(	(	PUNCT
fcis-2970	121	8	)	)	PUNCT
fcis-2970	121	9	(	(	PUNCT
fcis-2970	121	10	u	u	NOUN
fcis-2970	121	11	)	)	PUNCT
fcis-2970	121	12	t	t	PROPN
fcis-2970	121	13	t	t	PROPN
fcis-2970	121	14	t	t	PROPN
fcis-2970	121	15	exp	exp	NOUN
fcis-2970	121	16	u	u	NOUN
fcis-2970	121	17	exp	exp	NOUN
fcis-2970	121	18			NOUN
fcis-2970	121	19	=	=	SYM
fcis-2970	121	20			X
fcis-2970	121	21	(	(	PUNCT
fcis-2970	121	22	11	11	NUM
fcis-2970	121	23	)	)	PUNCT
fcis-2970	121	24	t	t	NOUN
fcis-2970	121	25	t	t	PROPN
fcis-2970	121	26	t	t	PROPN
fcis-2970	121	27	r	r	PROPN
fcis-2970	121	28	h=	h=	PROPN
fcis-2970	121	29	(	(	PUNCT
fcis-2970	121	30	12	12	NUM
fcis-2970	121	31	)	)	PUNCT
fcis-2970	121	32	where	where	SCONJ
fcis-2970	121	33	,	,	PUNCT
fcis-2970	121	34	gw	gw	PROPN
fcis-2970	121	35	is	be	AUX
fcis-2970	121	36	the	the	DET
fcis-2970	121	37	parameter	parameter	NOUN
fcis-2970	121	38	matrix	matrix	NOUN
fcis-2970	121	39	of	of	ADP
fcis-2970	121	40	the	the	DET
fcis-2970	121	41	attention	attention	NOUN
fcis-2970	121	42	mechanism	mechanism	NOUN
fcis-2970	121	43	layer	layer	NOUN
fcis-2970	121	44	,	,	PUNCT
fcis-2970	121	45	and	and	CCONJ
fcis-2970	121	46	gb	gb	PRON
fcis-2970	121	47	is	be	AUX
fcis-2970	121	48	the	the	DET
fcis-2970	121	49	bias	bias	NOUN
fcis-2970	121	50	vector	vector	NOUN
fcis-2970	121	51	.	.	PUNCT
fcis-2970	122	1	both	both	PRON
fcis-2970	122	2	of	of	ADP
fcis-2970	122	3	them	they	PRON
fcis-2970	122	4	need	need	VERB
fcis-2970	122	5	to	to	PART
fcis-2970	122	6	be	be	AUX
fcis-2970	122	7	learned	learn	VERB
fcis-2970	122	8	during	during	ADP
fcis-2970	122	9	training	training	NOUN
fcis-2970	122	10	and	and	CCONJ
fcis-2970	122	11	(	(	PUNCT
fcis-2970	122	12	)	)	PUNCT
fcis-2970	122	13	exp	exp	NOUN
fcis-2970	122	14	is	be	AUX
fcis-2970	122	15	an	an	DET
fcis-2970	122	16	exponential	exponential	ADJ
fcis-2970	122	17	function	function	NOUN
fcis-2970	122	18	.	.	PUNCT
fcis-2970	123	1	the	the	DET
fcis-2970	123	2	weighted	weight	VERB
fcis-2970	123	3	sentence	sentence	NOUN
fcis-2970	123	4	representation	representation	NOUN
fcis-2970	123	5	is	be	AUX
fcis-2970	123	6	obtained	obtain	VERB
fcis-2970	123	7	through	through	ADP
fcis-2970	123	8	the	the	DET
fcis-2970	123	9	attention	attention	NOUN
fcis-2970	123	10	mechanism	mechanism	NOUN
fcis-2970	123	11	layer	layer	NOUN
fcis-2970	123	12	,	,	PUNCT
fcis-2970	123	13	and	and	CCONJ
fcis-2970	123	14	the	the	DET
fcis-2970	123	15	sentence	sentence	NOUN
fcis-2970	123	16	vector	vector	NOUN
fcis-2970	123	17	is	be	AUX
fcis-2970	123	18	input	input	VERB
fcis-2970	123	19	into	into	ADP
fcis-2970	123	20	the	the	DET
fcis-2970	123	21	fully	fully	ADV
fcis-2970	123	22	connected	connected	ADJ
fcis-2970	123	23	network	network	NOUN
fcis-2970	123	24	for	for	ADP
fcis-2970	123	25	spatial	spatial	ADJ
fcis-2970	123	26	mapping	mapping	NOUN
fcis-2970	123	27	.	.	PUNCT
fcis-2970	124	1	then	then	ADV
fcis-2970	124	2	,	,	PUNCT
fcis-2970	124	3	the	the	DET
fcis-2970	124	4	sentence	sentence	NOUN
fcis-2970	124	5	emotion	emotion	NOUN
fcis-2970	124	6	classification	classification	NOUN
fcis-2970	124	7	probability	probability	NOUN
fcis-2970	124	8	distribution	distribution	NOUN
fcis-2970	124	9	rp	rp	NOUN
fcis-2970	124	10	is	be	AUX
fcis-2970	124	11	obtained	obtain	VERB
fcis-2970	124	12	through	through	ADP
fcis-2970	124	13	the	the	DET
fcis-2970	124	14	softmax	softmax	NOUN
fcis-2970	124	15	function	function	NOUN
fcis-2970	124	16	,	,	PUNCT
fcis-2970	124	17	and	and	CCONJ
fcis-2970	124	18	the	the	DET
fcis-2970	124	19	formula	formula	NOUN
fcis-2970	124	20	is	be	AUX
fcis-2970	124	21	as	as	SCONJ
fcis-2970	124	22	follows	follow	VERB
fcis-2970	124	23	:	:	PUNCT
fcis-2970	124	24	(	(	PUNCT
fcis-2970	124	25	)	)	PUNCT
fcis-2970	124	26	r	r	NOUN
fcis-2970	124	27	s	s	PROPN
fcis-2970	124	28	sp	sp	ADP
fcis-2970	124	29	softmax	softmax	NOUN
fcis-2970	124	30	w	w	PROPN
fcis-2970	124	31	r	r	NOUN
fcis-2970	124	32	b=	b=	NOUN
fcis-2970	124	33	+	+	CCONJ
fcis-2970	124	34	(	(	PUNCT
fcis-2970	124	35	13	13	NUM
fcis-2970	124	36	)	)	PUNCT
fcis-2970	124	37	among	among	ADP
fcis-2970	124	38	them	they	PRON
fcis-2970	124	39	,	,	PUNCT
fcis-2970	124	40	sw	sw	PROPN
fcis-2970	124	41	and	and	CCONJ
fcis-2970	124	42	sb	sb	PROPN
fcis-2970	124	43	are	be	AUX
fcis-2970	124	44	the	the	DET
fcis-2970	124	45	parameter	parameter	NOUN
fcis-2970	124	46	matrix	matrix	NOUN
fcis-2970	124	47	and	and	CCONJ
fcis-2970	124	48	bias	bias	NOUN
fcis-2970	124	49	in	in	ADP
fcis-2970	124	50	the	the	DET
fcis-2970	124	51	fully	fully	ADV
fcis-2970	124	52	connected	connected	ADJ
fcis-2970	124	53	network	network	NOUN
fcis-2970	124	54	respectively	respectively	ADV
fcis-2970	124	55	,	,	PUNCT
fcis-2970	124	56	which	which	PRON
fcis-2970	124	57	need	need	VERB
fcis-2970	124	58	to	to	PART
fcis-2970	124	59	be	be	AUX
fcis-2970	124	60	learned	learn	VERB
fcis-2970	124	61	in	in	ADP
fcis-2970	124	62	the	the	DET
fcis-2970	124	63	training	training	NOUN
fcis-2970	124	64	process	process	NOUN
fcis-2970	124	65	.	.	PUNCT
fcis-2970	125	1	finally	finally	ADV
fcis-2970	125	2	,	,	PUNCT
fcis-2970	125	3	the	the	DET
fcis-2970	125	4	category	category	NOUN
fcis-2970	125	5	corresponding	correspond	VERB
fcis-2970	125	6	to	to	ADP
fcis-2970	125	7	the	the	DET
fcis-2970	125	8	maximum	maximum	ADJ
fcis-2970	125	9	probability	probability	NOUN
fcis-2970	125	10	is	be	AUX
fcis-2970	125	11	selected	select	VERB
fcis-2970	125	12	from	from	ADP
fcis-2970	125	13	the	the	DET
fcis-2970	125	14	output	output	NOUN
fcis-2970	125	15	as	as	SCONJ
fcis-2970	125	16	the	the	DET
fcis-2970	125	17	prediction	prediction	NOUN
fcis-2970	125	18	result	result	NOUN
fcis-2970	126	1	[	[	X
fcis-2970	126	2	16	16	NUM
fcis-2970	126	3	]	]	PUNCT
fcis-2970	126	4	.	.	PUNCT
fcis-2970	127	1	3	3	X
fcis-2970	127	2	.	.	X
fcis-2970	127	3	experiment	experiment	NOUN
fcis-2970	127	4	and	and	CCONJ
fcis-2970	127	5	result	result	VERB
fcis-2970	127	6	analysis	analysis	NOUN
fcis-2970	127	7	3.1	3.1	NUM
fcis-2970	127	8	.	.	PUNCT
fcis-2970	127	9	experimental	experimental	ADJ
fcis-2970	127	10	environment	environment	NOUN
fcis-2970	127	11	and	and	CCONJ
fcis-2970	127	12	data	datum	NOUN
fcis-2970	127	13	set	set	VERB
fcis-2970	127	14	the	the	DET
fcis-2970	127	15	development	development	NOUN
fcis-2970	127	16	environment	environment	NOUN
fcis-2970	127	17	for	for	ADP
fcis-2970	127	18	the	the	DET
fcis-2970	127	19	experiment	experiment	NOUN
fcis-2970	127	20	in	in	ADP
fcis-2970	127	21	this	this	DET
fcis-2970	127	22	paper	paper	NOUN
fcis-2970	127	23	is	be	AUX
fcis-2970	127	24	the	the	DET
fcis-2970	127	25	paddle	paddle	NOUN
fcis-2970	127	26	,	,	PUNCT
fcis-2970	127	27	and	and	CCONJ
fcis-2970	127	28	the	the	DET
fcis-2970	127	29	development	development	NOUN
fcis-2970	127	30	tool	tool	NOUN
fcis-2970	127	31	is	be	AUX
fcis-2970	127	32	ai	ai	VERB
fcis-2970	127	33	studio	studio	NOUN
fcis-2970	127	34	based	base	VERB
fcis-2970	127	35	on	on	ADP
fcis-2970	127	36	baidu	baidu	PROPN
fcis-2970	127	37	's	's	PART
fcis-2970	127	38	deep	deep	ADJ
fcis-2970	127	39	learning	learn	VERB
fcis-2970	127	40	open	open	ADJ
fcis-2970	127	41	-	-	PUNCT
fcis-2970	127	42	source	source	NOUN
fcis-2970	127	43	platform	platform	NOUN
fcis-2970	127	44	fei	fei	PROPN
fcis-2970	127	45	paddle	paddle	PROPN
fcis-2970	127	46	.	.	PUNCT
fcis-2970	128	1	the	the	DET
fcis-2970	128	2	development	development	NOUN
fcis-2970	128	3	language	language	NOUN
fcis-2970	128	4	is	be	AUX
fcis-2970	128	5	python	python	NOUN
fcis-2970	128	6	,	,	PUNCT
fcis-2970	128	7	and	and	CCONJ
fcis-2970	128	8	the	the	DET
fcis-2970	128	9	free	free	ADJ
fcis-2970	128	10	gpu	gpu	PROPN
fcis-2970	128	11	-	-	PUNCT
fcis-2970	128	12	accelerated	accelerate	VERB
fcis-2970	128	13	program	program	NOUN
fcis-2970	128	14	provided	provide	VERB
fcis-2970	128	15	by	by	ADP
fcis-2970	128	16	the	the	DET
fcis-2970	128	17	online	online	ADJ
fcis-2970	128	18	platform	platform	NOUN
fcis-2970	128	19	is	be	AUX
fcis-2970	128	20	used	use	VERB
fcis-2970	128	21	.	.	PUNCT
fcis-2970	129	1	the	the	DET
fcis-2970	129	2	data	datum	NOUN
fcis-2970	129	3	used	use	VERB
fcis-2970	129	4	in	in	ADP
fcis-2970	129	5	the	the	DET
fcis-2970	129	6	experiment	experiment	NOUN
fcis-2970	129	7	are	be	AUX
fcis-2970	129	8	user	user	NOUN
fcis-2970	129	9	review	review	NOUN
fcis-2970	129	10	data	datum	NOUN
fcis-2970	129	11	set	set	VERB
fcis-2970	129	12	and	and	CCONJ
fcis-2970	129	13	hotel	hotel	NOUN
fcis-2970	129	14	review	review	NOUN
fcis-2970	129	15	data	datum	NOUN
fcis-2970	129	16	set	set	VERB
fcis-2970	129	17	of	of	ADP
fcis-2970	129	18	a	a	DET
fcis-2970	129	19	food	food	NOUN
fcis-2970	129	20	delivery	delivery	NOUN
fcis-2970	129	21	platform	platform	NOUN
fcis-2970	129	22	,	,	PUNCT
fcis-2970	129	23	among	among	ADP
fcis-2970	129	24	which	which	PRON
fcis-2970	129	25	there	there	PRON
fcis-2970	129	26	are	be	VERB
fcis-2970	129	27	11,987	11,987	NUM
fcis-2970	129	28	food	food	NOUN
fcis-2970	129	29	review	review	NOUN
fcis-2970	129	30	data	data	NOUN
fcis-2970	129	31	sets	set	NOUN
fcis-2970	129	32	,	,	PUNCT
fcis-2970	129	33	including	include	VERB
fcis-2970	129	34	4,000	4,000	NUM
fcis-2970	129	35	positive	positive	ADJ
fcis-2970	129	36	and	and	CCONJ
fcis-2970	129	37	7,987	7,987	NUM
fcis-2970	129	38	negative	negative	ADJ
fcis-2970	129	39	evaluation	evaluation	NOUN
fcis-2970	129	40	data	datum	NOUN
fcis-2970	129	41	.	.	PUNCT
fcis-2970	130	1	there	there	PRON
fcis-2970	130	2	are	be	VERB
fcis-2970	130	3	7766	7766	NUM
fcis-2970	130	4	pieces	piece	NOUN
fcis-2970	130	5	of	of	ADP
fcis-2970	130	6	hotel	hotel	NOUN
fcis-2970	130	7	evaluation	evaluation	NOUN
fcis-2970	130	8	data	datum	NOUN
fcis-2970	130	9	set	set	VERB
fcis-2970	130	10	,	,	PUNCT
fcis-2970	130	11	including	include	VERB
fcis-2970	130	12	5322	5322	NUM
fcis-2970	130	13	pieces	piece	NOUN
fcis-2970	130	14	of	of	ADP
fcis-2970	130	15	positive	positive	ADJ
fcis-2970	130	16	evaluation	evaluation	NOUN
fcis-2970	130	17	data	datum	NOUN
fcis-2970	130	18	and	and	CCONJ
fcis-2970	130	19	2444	2444	NUM
fcis-2970	130	20	pieces	piece	NOUN
fcis-2970	130	21	of	of	ADP
fcis-2970	130	22	negative	negative	ADJ
fcis-2970	130	23	and	and	CCONJ
fcis-2970	130	24	negative	negative	ADJ
fcis-2970	130	25	evaluation	evaluation	NOUN
fcis-2970	130	26	data	datum	NOUN
fcis-2970	130	27	.	.	PUNCT
fcis-2970	131	1	in	in	ADP
fcis-2970	131	2	the	the	DET
fcis-2970	131	3	experiment	experiment	NOUN
fcis-2970	131	4	,	,	PUNCT
fcis-2970	131	5	the	the	DET
fcis-2970	131	6	data	datum	NOUN
fcis-2970	131	7	set	set	VERB
fcis-2970	131	8	will	will	AUX
fcis-2970	131	9	be	be	AUX
fcis-2970	131	10	divided	divide	VERB
fcis-2970	131	11	into	into	ADP
fcis-2970	131	12	the	the	DET
fcis-2970	131	13	training	training	NOUN
fcis-2970	131	14	set	set	NOUN
fcis-2970	131	15	and	and	CCONJ
fcis-2970	131	16	the	the	DET
fcis-2970	131	17	test	test	NOUN
fcis-2970	131	18	set	set	VERB
fcis-2970	131	19	according	accord	VERB
fcis-2970	131	20	to	to	ADP
fcis-2970	131	21	9:1	9:1	NUM
fcis-2970	131	22	.	.	PUNCT
fcis-2970	132	1	table	table	NOUN
fcis-2970	132	2	1	1	NUM
fcis-2970	132	3	.	.	PUNCT
fcis-2970	133	1	partition	partition	NOUN
fcis-2970	133	2	of	of	ADP
fcis-2970	133	3	data	datum	NOUN
fcis-2970	133	4	set	set	VERB
fcis-2970	133	5	dataset	dataset	NOUN
fcis-2970	133	6	training	training	NOUN
fcis-2970	133	7	set	set	NOUN
fcis-2970	133	8	test	test	NOUN
fcis-2970	133	9	set	set	VERB
fcis-2970	133	10	positive	positive	ADJ
fcis-2970	133	11	negative	negative	ADJ
fcis-2970	133	12	positive	positive	ADJ
fcis-2970	133	13	negative	negative	ADJ
fcis-2970	133	14	takeaway	takeaway	NOUN
fcis-2970	133	15	review	review	NOUN
fcis-2970	133	16	date	date	NOUN
fcis-2970	133	17	set	set	VERB
fcis-2970	133	18	3600	3600	NUM
fcis-2970	133	19	7189	7189	NUM
fcis-2970	133	20	400	400	NUM
fcis-2970	133	21	798	798	NUM
fcis-2970	133	22	hotel	hotel	NOUN
fcis-2970	133	23	review	review	NOUN
fcis-2970	133	24	data	datum	NOUN
fcis-2970	133	25	set	set	VERB
fcis-2970	133	26	4730	4730	NUM
fcis-2970	133	27	2172	2172	NUM
fcis-2970	133	28	592	592	NUM
fcis-2970	133	29	272	272	NUM
fcis-2970	133	30	3.2	3.2	NUM
fcis-2970	133	31	.	.	PUNCT
fcis-2970	134	1	experimental	experimental	ADJ
fcis-2970	134	2	evaluation	evaluation	NOUN
fcis-2970	134	3	criteria	criterion	NOUN
fcis-2970	134	4	to	to	PART
fcis-2970	134	5	verify	verify	VERB
fcis-2970	134	6	the	the	DET
fcis-2970	134	7	effectiveness	effectiveness	NOUN
fcis-2970	134	8	of	of	ADP
fcis-2970	134	9	the	the	DET
fcis-2970	134	10	model	model	NOUN
fcis-2970	134	11	in	in	ADP
fcis-2970	134	12	text	text	NOUN
fcis-2970	134	13	classification	classification	NOUN
fcis-2970	134	14	experiments	experiment	NOUN
fcis-2970	134	15	,	,	PUNCT
fcis-2970	134	16	some	some	DET
fcis-2970	134	17	evaluation	evaluation	NOUN
fcis-2970	134	18	indexes	index	NOUN
fcis-2970	134	19	commonly	commonly	ADV
fcis-2970	134	20	used	use	VERB
fcis-2970	134	21	in	in	ADP
fcis-2970	134	22	classification	classification	NOUN
fcis-2970	134	23	models	model	NOUN
fcis-2970	134	24	are	be	AUX
fcis-2970	134	25	introduced	introduce	VERB
fcis-2970	134	26	to	to	PART
fcis-2970	134	27	measure	measure	VERB
fcis-2970	134	28	the	the	DET
fcis-2970	134	29	model	model	NOUN
fcis-2970	134	30	.	.	PUNCT
fcis-2970	135	1	the	the	DET
fcis-2970	135	2	evaluation	evaluation	NOUN
fcis-2970	135	3	indexes	index	NOUN
fcis-2970	135	4	used	use	VERB
fcis-2970	135	5	in	in	ADP
fcis-2970	135	6	the	the	DET
fcis-2970	135	7	binary	binary	ADJ
fcis-2970	135	8	experiment	experiment	NOUN
fcis-2970	135	9	include	include	VERB
fcis-2970	135	10	precision	precision	NOUN
fcis-2970	135	11	(	(	PUNCT
fcis-2970	135	12	p	p	NOUN
fcis-2970	135	13	-	-	PUNCT
fcis-2970	135	14	value	value	NOUN
fcis-2970	135	15	)	)	PUNCT
fcis-2970	135	16	,	,	PUNCT
fcis-2970	135	17	recall	recall	INTJ
fcis-2970	135	18	(	(	PUNCT
fcis-2970	135	19	r	r	NOUN
fcis-2970	135	20	-	-	PUNCT
fcis-2970	135	21	value	value	NOUN
fcis-2970	135	22	)	)	PUNCT
fcis-2970	135	23	,	,	PUNCT
fcis-2970	135	24	and	and	CCONJ
fcis-2970	135	25	f1	f1	NOUN
fcis-2970	135	26	value	value	NOUN
fcis-2970	135	27	.	.	PUNCT
fcis-2970	136	1	the	the	DET
fcis-2970	136	2	equivalent	equivalent	ADJ
fcis-2970	136	3	values	value	NOUN
fcis-2970	136	4	of	of	ADP
fcis-2970	136	5	tp	tp	PROPN
fcis-2970	136	6	,	,	PUNCT
fcis-2970	136	7	fp	fp	INTJ
fcis-2970	136	8	,	,	PUNCT
fcis-2970	136	9	fn	fn	NOUN
fcis-2970	136	10	,	,	PUNCT
fcis-2970	136	11	and	and	CCONJ
fcis-2970	136	12	tn	tn	NOUN
fcis-2970	136	13	used	use	VERB
fcis-2970	136	14	in	in	ADP
fcis-2970	136	15	the	the	DET
fcis-2970	136	16	calculation	calculation	NOUN
fcis-2970	136	17	are	be	AUX
fcis-2970	136	18	as	as	SCONJ
fcis-2970	136	19	follows	follow	VERB
fcis-2970	136	20	:	:	PUNCT
fcis-2970	136	21	table	table	NOUN
fcis-2970	136	22	2	2	NUM
fcis-2970	136	23	.	.	PUNCT
fcis-2970	137	1	classification	classification	NOUN
fcis-2970	137	2	discriminant	discriminant	ADJ
fcis-2970	137	3	confusion	confusion	NOUN
fcis-2970	137	4	matrix	matrix	NOUN
fcis-2970	137	5	confusion	confusion	NOUN
fcis-2970	137	6	matrix	matrix	NOUN
fcis-2970	137	7	real	real	ADJ
fcis-2970	137	8	label	label	NOUN
fcis-2970	137	9	positive	positive	ADJ
fcis-2970	137	10	negative	negative	ADJ
fcis-2970	137	11	predictive	predictive	ADJ
fcis-2970	137	12	label	label	NOUN
fcis-2970	137	13	positive	positive	ADJ
fcis-2970	137	14	tp	tp	ADP
fcis-2970	137	15	fp	fp	PROPN
fcis-2970	137	16	negative	negative	PROPN
fcis-2970	137	17	fn	fn	PROPN
fcis-2970	137	18	tn	tn	PROPN
fcis-2970	137	19	accuracy	accuracy	NOUN
fcis-2970	137	20	,	,	PUNCT
fcis-2970	137	21	recall	recall	NOUN
fcis-2970	137	22	rate	rate	NOUN
fcis-2970	137	23	and	and	CCONJ
fcis-2970	137	24	value	value	NOUN
fcis-2970	137	25	are	be	AUX
fcis-2970	137	26	calculated	calculate	VERB
fcis-2970	137	27	as	as	SCONJ
fcis-2970	137	28	follows	follow	VERB
fcis-2970	137	29	:	:	PUNCT
fcis-2970	137	30	tp	tp	ADP
fcis-2970	137	31	p	p	NOUN
fcis-2970	137	32	tp	tp	ADP
fcis-2970	137	33	fp	fp	PROPN
fcis-2970	137	34	=	=	PUNCT
fcis-2970	138	1	+	+	CCONJ
fcis-2970	138	2	(	(	PUNCT
fcis-2970	138	3	14	14	NUM
fcis-2970	138	4	)	)	PUNCT
fcis-2970	138	5	tp	tp	ADP
fcis-2970	138	6	r	r	NOUN
fcis-2970	138	7	tp	tp	NOUN
fcis-2970	138	8	fn	fn	NOUN
fcis-2970	139	1	=	=	PUNCT
fcis-2970	140	1	+	+	CCONJ
fcis-2970	140	2	(	(	PUNCT
fcis-2970	140	3	15	15	NUM
fcis-2970	140	4	)	)	SYM
fcis-2970	140	5	1	1	NUM
fcis-2970	140	6	2	2	NUM
fcis-2970	140	7	p	p	NOUN
fcis-2970	140	8	r	r	NOUN
fcis-2970	140	9	f	f	NOUN
fcis-2970	140	10	p	p	NOUN
fcis-2970	140	11	r	r	NOUN
fcis-2970	140	12			PROPN
fcis-2970	140	13			PROPN
fcis-2970	140	14	=	=	PUNCT
fcis-2970	141	1	+	+	PUNCT
fcis-2970	141	2	(	(	PUNCT
fcis-2970	141	3	16	16	NUM
fcis-2970	141	4	)	)	PUNCT
fcis-2970	141	5	3.3	3.3	NUM
fcis-2970	141	6	.	.	PUNCT
fcis-2970	142	1	selection	selection	NOUN
fcis-2970	142	2	of	of	ADP
fcis-2970	142	3	experimental	experimental	ADJ
fcis-2970	142	4	parameters	parameter	NOUN
fcis-2970	142	5	the	the	DET
fcis-2970	142	6	model	model	NOUN
fcis-2970	142	7	experiment	experiment	NOUN
fcis-2970	142	8	results	result	NOUN
fcis-2970	142	9	will	will	AUX
fcis-2970	142	10	be	be	AUX
fcis-2970	142	11	related	relate	VERB
fcis-2970	142	12	to	to	ADP
fcis-2970	142	13	the	the	DET
fcis-2970	142	14	parameter	parameter	NOUN
fcis-2970	142	15	settings	setting	NOUN
fcis-2970	142	16	.	.	PUNCT
fcis-2970	143	1	after	after	ADP
fcis-2970	143	2	several	several	ADJ
fcis-2970	143	3	comparative	comparative	ADJ
fcis-2970	143	4	experiments	experiment	NOUN
fcis-2970	143	5	,	,	PUNCT
fcis-2970	143	6	the	the	DET
fcis-2970	143	7	model	model	NOUN
fcis-2970	143	8	parameter	parameter	NOUN
fcis-2970	143	9	values	value	NOUN
fcis-2970	143	10	are	be	AUX
fcis-2970	143	11	set	set	VERB
fcis-2970	143	12	as	as	SCONJ
fcis-2970	143	13	follows	follow	VERB
fcis-2970	143	14	:	:	PUNCT
fcis-2970	143	15	the	the	DET
fcis-2970	143	16	maximum	maximum	ADJ
fcis-2970	143	17	length	length	NOUN
fcis-2970	143	18	of	of	ADP
fcis-2970	143	19	the	the	DET
fcis-2970	143	20	input	input	NOUN
fcis-2970	143	21	text	text	NOUN
fcis-2970	143	22	is	be	AUX
fcis-2970	143	23	128	128	NUM
fcis-2970	143	24	;	;	PUNCT
fcis-2970	143	25	the	the	DET
fcis-2970	143	26	pre	pre	ADJ
fcis-2970	143	27	-	-	ADJ
fcis-2970	143	28	training	training	ADJ
fcis-2970	143	29	model	model	NOUN
fcis-2970	143	30	output	output	NOUN
fcis-2970	143	31	word	word	NOUN
fcis-2970	143	32	vector	vector	NOUN
fcis-2970	143	33	dimension	dimension	NOUN
fcis-2970	143	34	is	be	AUX
fcis-2970	143	35	768	768	NUM
fcis-2970	143	36	dimensions	dimension	NOUN
fcis-2970	143	37	;	;	PUNCT
fcis-2970	143	38	the	the	DET
fcis-2970	143	39	output	output	NOUN
fcis-2970	143	40	dimension	dimension	NOUN
fcis-2970	143	41	of	of	ADP
fcis-2970	143	42	bilstm	bilstm	NOUN
fcis-2970	143	43	is	be	AUX
fcis-2970	143	44	256	256	NUM
fcis-2970	143	45	dimensions	dimension	NOUN
fcis-2970	143	46	.	.	PUNCT
fcis-2970	144	1	dropout	dropout	NOUN
fcis-2970	144	2	set	set	VERB
fcis-2970	144	3	to	to	ADP
fcis-2970	144	4	0.5	0.5	NUM
fcis-2970	144	5	;	;	PUNCT
fcis-2970	144	6	choose	choose	VERB
fcis-2970	144	7	the	the	DET
fcis-2970	144	8	cross	cross	NOUN
fcis-2970	144	9	entropy	entropy	NOUN
fcis-2970	144	10	as	as	ADP
fcis-2970	144	11	the	the	DET
fcis-2970	144	12	loss	loss	NOUN
fcis-2970	144	13	function	function	NOUN
fcis-2970	144	14	;	;	PUNCT
fcis-2970	144	15	adamw	adamw	PROPN
fcis-2970	144	16	was	be	AUX
fcis-2970	144	17	selected	select	VERB
fcis-2970	144	18	as	as	ADP
fcis-2970	144	19	the	the	DET
fcis-2970	144	20	optimizer	optimizer	NOUN
fcis-2970	144	21	,	,	PUNCT
fcis-2970	144	22	and	and	CCONJ
fcis-2970	144	23	the	the	DET
fcis-2970	144	24	learning	learning	NOUN
fcis-2970	144	25	rate	rate	NOUN
fcis-2970	144	26	was	be	AUX
fcis-2970	144	27	set	set	VERB
fcis-2970	144	28	to	to	ADP
fcis-2970	144	29	5e5	5e5	NUM
fcis-2970	144	30	;	;	PUNCT
fcis-2970	144	31	set	set	VERB
fcis-2970	144	32	the	the	DET
fcis-2970	144	33	epoch	epoch	NOUN
fcis-2970	144	34	of	of	ADP
fcis-2970	144	35	training	training	NOUN
fcis-2970	144	36	set	set	VERB
fcis-2970	144	37	to	to	ADP
fcis-2970	144	38	5	5	NUM
fcis-2970	144	39	and	and	CCONJ
fcis-2970	144	40	batch_size	batch_size	VERB
fcis-2970	144	41	to	to	ADP
fcis-2970	144	42	32	32	NUM
fcis-2970	144	43	.	.	PUNCT
fcis-2970	145	1	3.4	3.4	NUM
fcis-2970	145	2	.	.	PUNCT
fcis-2970	146	1	experimental	experimental	ADJ
fcis-2970	146	2	comparison	comparison	NOUN
fcis-2970	146	3	and	and	CCONJ
fcis-2970	146	4	analysis	analysis	NOUN
fcis-2970	146	5	to	to	PART
fcis-2970	146	6	verify	verify	VERB
fcis-2970	146	7	the	the	DET
fcis-2970	146	8	effectiveness	effectiveness	NOUN
fcis-2970	146	9	of	of	ADP
fcis-2970	146	10	the	the	DET
fcis-2970	146	11	model	model	NOUN
fcis-2970	146	12	in	in	ADP
fcis-2970	146	13	this	this	DET
fcis-2970	146	14	paper	paper	NOUN
fcis-2970	146	15	,	,	PUNCT
fcis-2970	146	16	the	the	DET
fcis-2970	146	17	same	same	ADJ
fcis-2970	146	18	data	datum	NOUN
fcis-2970	146	19	set	set	VERB
fcis-2970	146	20	and	and	CCONJ
fcis-2970	146	21	the	the	DET
fcis-2970	146	22	same	same	ADJ
fcis-2970	146	23	experimental	experimental	ADJ
fcis-2970	146	24	environment	environment	NOUN
fcis-2970	146	25	are	be	AUX
fcis-2970	146	26	used	use	VERB
fcis-2970	146	27	to	to	PART
fcis-2970	146	28	compare	compare	VERB
fcis-2970	146	29	the	the	DET
fcis-2970	146	30	emotion	emotion	NOUN
fcis-2970	146	31	classification	classification	NOUN
fcis-2970	146	32	model	model	NOUN
fcis-2970	146	33	in	in	ADP
fcis-2970	146	34	this	this	DET
fcis-2970	146	35	paper	paper	NOUN
fcis-2970	146	36	with	with	ADP
fcis-2970	146	37	other	other	ADJ
fcis-2970	146	38	emotion	emotion	NOUN
fcis-2970	146	39	classification	classification	NOUN
fcis-2970	146	40	models	model	NOUN
fcis-2970	146	41	,	,	PUNCT
fcis-2970	146	42	train	train	NOUN
fcis-2970	146	43	on	on	ADP
fcis-2970	146	44	the	the	DET
fcis-2970	146	45	hotel	hotel	NOUN
fcis-2970	146	46	evaluation	evaluation	NOUN
fcis-2970	146	47	training	training	NOUN
fcis-2970	146	48	set	set	NOUN
fcis-2970	146	49	,	,	PUNCT
fcis-2970	146	50	and	and	CCONJ
fcis-2970	146	51	finally	finally	ADV
fcis-2970	146	52	test	test	VERB
fcis-2970	146	53	on	on	ADP
fcis-2970	146	54	the	the	DET
fcis-2970	146	55	corresponding	corresponding	ADJ
fcis-2970	146	56	test	test	NOUN
fcis-2970	146	57	set	set	NOUN
fcis-2970	146	58	.	.	PUNCT
fcis-2970	147	1	the	the	DET
fcis-2970	147	2	test	test	NOUN
fcis-2970	147	3	results	result	NOUN
fcis-2970	147	4	are	be	AUX
fcis-2970	147	5	taken	take	VERB
fcis-2970	147	6	as	as	ADP
fcis-2970	147	7	the	the	DET
fcis-2970	147	8	macro	macro	ADJ
fcis-2970	147	9	average	average	ADJ
fcis-2970	147	10	value	value	NOUN
fcis-2970	147	11	of	of	ADP
fcis-2970	147	12	the	the	DET
fcis-2970	147	13	evaluation	evaluation	NOUN
fcis-2970	147	14	index	index	NOUN
fcis-2970	147	15	,	,	PUNCT
fcis-2970	147	16	as	as	SCONJ
fcis-2970	147	17	shown	show	VERB
fcis-2970	147	18	in	in	ADP
fcis-2970	147	19	table	table	NOUN
fcis-2970	147	20	3	3	NUM
fcis-2970	147	21	:	:	SYM
fcis-2970	147	22	74	74	NUM
fcis-2970	147	23	table	table	NOUN
fcis-2970	147	24	3	3	NUM
fcis-2970	147	25	.	.	PUNCT
fcis-2970	147	26	comparison	comparison	NOUN
fcis-2970	147	27	of	of	ADP
fcis-2970	147	28	experimental	experimental	ADJ
fcis-2970	147	29	results	result	NOUN
fcis-2970	147	30	of	of	ADP
fcis-2970	147	31	hotel	hotel	NOUN
fcis-2970	147	32	review	review	NOUN
fcis-2970	147	33	test	test	NOUN
fcis-2970	147	34	set	set	VERB
fcis-2970	147	35	number	number	NOUN
fcis-2970	147	36	of	of	ADP
fcis-2970	147	37	the	the	DET
fcis-2970	147	38	model	model	NOUN
fcis-2970	147	39	model	model	PROPN
fcis-2970	147	40	precision	precision	PROPN
fcis-2970	147	41	recall	recall	PROPN
fcis-2970	147	42	f1	f1	PROPN
fcis-2970	147	43	1	1	NUM
fcis-2970	147	44	word2vec	word2vec	ADV
fcis-2970	147	45	-	-	PUNCT
fcis-2970	147	46	lstm	lstm	ADJ
fcis-2970	147	47	0.7930	0.7930	NUM
fcis-2970	147	48	0.7224	0.7224	NUM
fcis-2970	147	49	0.7382	0.7382	NUM
fcis-2970	147	50	2	2	NUM
fcis-2970	147	51	word2vec	word2vec	NUM
fcis-2970	147	52	-	-	PUNCT
fcis-2970	147	53	lstm	lstm	NOUN
fcis-2970	147	54	-	-	PUNCT
fcis-2970	147	55	at	at	ADP
fcis-2970	147	56	0.8549	0.8549	NUM
fcis-2970	147	57	0.8288	0.8288	NUM
fcis-2970	147	58	0.8399	0.8399	NUM
fcis-2970	147	59	3	3	NUM
fcis-2970	147	60	word2vec	word2vec	ADV
fcis-2970	147	61	-	-	PUNCT
fcis-2970	147	62	bilstm	bilstm	NOUN
fcis-2970	147	63	0.8521	0.8521	NUM
fcis-2970	147	64	0.8398	0.8398	NUM
fcis-2970	147	65	0.8455	0.8455	NUM
fcis-2970	147	66	4	4	NUM
fcis-2970	147	67	word2vec	word2vec	NOUN
fcis-2970	147	68	-	-	PUNCT
fcis-2970	147	69	birnn	birnn	NOUN
fcis-2970	147	70	-	-	PUNCT
fcis-2970	147	71	at	at	ADP
fcis-2970	147	72	0.8590	0.8590	NUM
fcis-2970	147	73	0.8360	0.8360	NUM
fcis-2970	147	74	0.8462	0.8462	NUM
fcis-2970	147	75	5	5	NUM
fcis-2970	147	76	word2vec	word2vec	ADV
fcis-2970	147	77	-	-	PUNCT
fcis-2970	147	78	bilstm	bilstm	NOUN
fcis-2970	147	79	-	-	PUNCT
fcis-2970	147	80	at	at	ADP
fcis-2970	147	81	0.8818	0.8818	NUM
fcis-2970	147	82	0.8527	0.8527	NUM
fcis-2970	147	83	0.8647	0.8647	NUM
fcis-2970	147	84	6	6	NUM
fcis-2970	147	85	ernie	ernie	NOUN
fcis-2970	147	86	0.8952	0.8952	NUM
fcis-2970	147	87	0.8995	0.8995	NUM
fcis-2970	147	88	0.8973	0.8973	NUM
fcis-2970	147	89	7	7	NUM
fcis-2970	147	90	ernie	ernie	NOUN
fcis-2970	147	91	-	-	PUNCT
fcis-2970	147	92	bilstm	bilstm	NOUN
fcis-2970	147	93	0.8979	0.8979	NUM
fcis-2970	147	94	0.8972	0.8972	NUM
fcis-2970	147	95	0.8976	0.8976	NUM
fcis-2970	147	96	8	8	NUM
fcis-2970	147	97	ernie	ernie	NOUN
fcis-2970	147	98	-	-	PUNCT
fcis-2970	147	99	bilstm	bilstm	NOUN
fcis-2970	147	100	-	-	PUNCT
fcis-2970	147	101	at	at	ADP
fcis-2970	147	102	0.9058	0.9058	NUM
fcis-2970	147	103	0.8869	0.8869	NUM
fcis-2970	147	104	0.8953	0.8953	NUM
fcis-2970	147	105	in	in	ADP
fcis-2970	147	106	the	the	DET
fcis-2970	147	107	table	table	NOUN
fcis-2970	147	108	,	,	PUNCT
fcis-2970	147	109	model	model	NOUN
fcis-2970	147	110	1	1	NUM
fcis-2970	147	111	introduces	introduce	NOUN
fcis-2970	147	112	the	the	DET
fcis-2970	147	113	word	word	NOUN
fcis-2970	147	114	vector	vector	NOUN
fcis-2970	147	115	obtained	obtain	VERB
fcis-2970	147	116	by	by	ADP
fcis-2970	147	117	using	use	VERB
fcis-2970	147	118	word2vec	word2vec	PRON
fcis-2970	147	119	in	in	ADP
fcis-2970	147	120	the	the	DET
fcis-2970	147	121	external	external	ADJ
fcis-2970	147	122	corpus	corpus	NOUN
fcis-2970	147	123	training	training	NOUN
fcis-2970	147	124	,	,	PUNCT
fcis-2970	147	125	then	then	ADV
fcis-2970	147	126	inputs	input	VERB
fcis-2970	147	127	the	the	DET
fcis-2970	147	128	text	text	NOUN
fcis-2970	147	129	sentence	sentence	NOUN
fcis-2970	147	130	represented	represent	VERB
fcis-2970	147	131	by	by	ADP
fcis-2970	147	132	the	the	DET
fcis-2970	147	133	word	word	NOUN
fcis-2970	147	134	vector	vector	NOUN
fcis-2970	147	135	in	in	ADP
fcis-2970	147	136	the	the	DET
fcis-2970	147	137	word	word	NOUN
fcis-2970	147	138	embedding	embed	VERB
fcis-2970	147	139	layer	layer	NOUN
fcis-2970	147	140	into	into	ADP
fcis-2970	147	141	the	the	DET
fcis-2970	147	142	lstm	lstm	PROPN
fcis-2970	147	143	model	model	NOUN
fcis-2970	147	144	to	to	PART
fcis-2970	147	145	extract	extract	VERB
fcis-2970	147	146	the	the	DET
fcis-2970	147	147	text	text	NOUN
fcis-2970	147	148	depth	depth	NOUN
fcis-2970	147	149	feature	feature	NOUN
fcis-2970	147	150	,	,	PUNCT
fcis-2970	147	151	and	and	CCONJ
fcis-2970	147	152	finally	finally	ADV
fcis-2970	147	153	gets	get	VERB
fcis-2970	147	154	the	the	DET
fcis-2970	147	155	prediction	prediction	NOUN
fcis-2970	147	156	result	result	NOUN
fcis-2970	147	157	through	through	ADP
fcis-2970	147	158	the	the	DET
fcis-2970	147	159	output	output	NOUN
fcis-2970	147	160	of	of	ADP
fcis-2970	147	161	the	the	DET
fcis-2970	147	162	full	full	ADJ
fcis-2970	147	163	connection	connection	NOUN
fcis-2970	147	164	layer	layer	NOUN
fcis-2970	147	165	.	.	PUNCT
fcis-2970	148	1	in	in	ADP
fcis-2970	148	2	model	model	NOUN
fcis-2970	148	3	2	2	NUM
fcis-2970	148	4	,	,	PUNCT
fcis-2970	148	5	the	the	DET
fcis-2970	148	6	same	same	ADJ
fcis-2970	148	7	external	external	ADJ
fcis-2970	148	8	word	word	NOUN
fcis-2970	148	9	vector	vector	NOUN
fcis-2970	148	10	model	model	NOUN
fcis-2970	148	11	is	be	AUX
fcis-2970	148	12	introduced	introduce	VERB
fcis-2970	148	13	,	,	PUNCT
fcis-2970	148	14	and	and	CCONJ
fcis-2970	148	15	then	then	ADV
fcis-2970	148	16	the	the	DET
fcis-2970	148	17	sentence	sentence	NOUN
fcis-2970	148	18	vector	vector	NOUN
fcis-2970	148	19	is	be	AUX
fcis-2970	148	20	input	input	VERB
fcis-2970	148	21	into	into	ADP
fcis-2970	148	22	lstm	lstm	NOUN
fcis-2970	148	23	-	-	PUNCT
fcis-2970	148	24	at	at	ADP
fcis-2970	148	25	,	,	PUNCT
fcis-2970	148	26	and	and	CCONJ
fcis-2970	148	27	the	the	DET
fcis-2970	148	28	prediction	prediction	NOUN
fcis-2970	148	29	result	result	NOUN
fcis-2970	148	30	is	be	AUX
fcis-2970	148	31	obtained	obtain	VERB
fcis-2970	148	32	through	through	ADP
fcis-2970	148	33	the	the	DET
fcis-2970	148	34	full	full	ADJ
fcis-2970	148	35	-	-	PUNCT
fcis-2970	148	36	connection	connection	NOUN
fcis-2970	148	37	layer	layer	NOUN
fcis-2970	148	38	output	output	NOUN
fcis-2970	148	39	.	.	PUNCT
fcis-2970	149	1	in	in	ADP
fcis-2970	149	2	model	model	NOUN
fcis-2970	149	3	3	3	NUM
fcis-2970	149	4	,	,	PUNCT
fcis-2970	149	5	the	the	DET
fcis-2970	149	6	pre	pre	ADJ
fcis-2970	149	7	-	-	ADJ
fcis-2970	149	8	trained	train	VERB
fcis-2970	149	9	word	word	NOUN
fcis-2970	149	10	vector	vector	NOUN
fcis-2970	149	11	model	model	NOUN
fcis-2970	149	12	was	be	AUX
fcis-2970	149	13	introduced	introduce	VERB
fcis-2970	149	14	,	,	PUNCT
fcis-2970	149	15	and	and	CCONJ
fcis-2970	149	16	then	then	ADV
fcis-2970	149	17	the	the	DET
fcis-2970	149	18	text	text	NOUN
fcis-2970	149	19	sentence	sentence	NOUN
fcis-2970	149	20	vector	vector	NOUN
fcis-2970	149	21	was	be	AUX
fcis-2970	149	22	input	input	VERB
fcis-2970	149	23	into	into	ADP
fcis-2970	149	24	the	the	DET
fcis-2970	149	25	model	model	NOUN
fcis-2970	149	26	bilstm	bilstm	NOUN
fcis-2970	149	27	to	to	PART
fcis-2970	149	28	extract	extract	VERB
fcis-2970	149	29	the	the	DET
fcis-2970	149	30	depth	depth	NOUN
fcis-2970	149	31	features	feature	NOUN
fcis-2970	149	32	,	,	PUNCT
fcis-2970	149	33	and	and	CCONJ
fcis-2970	149	34	the	the	DET
fcis-2970	149	35	prediction	prediction	NOUN
fcis-2970	149	36	results	result	NOUN
fcis-2970	149	37	were	be	AUX
fcis-2970	149	38	output	output	NOUN
fcis-2970	149	39	in	in	ADP
fcis-2970	149	40	the	the	DET
fcis-2970	149	41	full	full	ADJ
fcis-2970	149	42	-	-	PUNCT
fcis-2970	149	43	connection	connection	NOUN
fcis-2970	149	44	layer	layer	NOUN
fcis-2970	149	45	.	.	PUNCT
fcis-2970	150	1	in	in	ADP
fcis-2970	150	2	model	model	NOUN
fcis-2970	150	3	4	4	NUM
fcis-2970	150	4	,	,	PUNCT
fcis-2970	150	5	the	the	DET
fcis-2970	150	6	pre	pre	ADJ
fcis-2970	150	7	-	-	ADJ
fcis-2970	150	8	trained	train	VERB
fcis-2970	150	9	word	word	NOUN
fcis-2970	150	10	vector	vector	NOUN
fcis-2970	150	11	is	be	AUX
fcis-2970	150	12	used	use	VERB
fcis-2970	150	13	to	to	PART
fcis-2970	150	14	input	input	VERB
fcis-2970	150	15	the	the	DET
fcis-2970	150	16	obtained	obtain	VERB
fcis-2970	150	17	text	text	NOUN
fcis-2970	150	18	segmentation	segmentation	NOUN
fcis-2970	150	19	vector	vector	NOUN
fcis-2970	150	20	into	into	ADP
fcis-2970	150	21	birnn	birnn	NOUN
fcis-2970	150	22	-	-	PUNCT
fcis-2970	150	23	at	at	ADP
fcis-2970	150	24	,	,	PUNCT
fcis-2970	150	25	and	and	CCONJ
fcis-2970	150	26	then	then	ADV
fcis-2970	150	27	the	the	DET
fcis-2970	150	28	prediction	prediction	NOUN
fcis-2970	150	29	result	result	NOUN
fcis-2970	150	30	is	be	AUX
fcis-2970	150	31	obtained	obtain	VERB
fcis-2970	150	32	through	through	ADP
fcis-2970	150	33	the	the	DET
fcis-2970	150	34	full	full	ADJ
fcis-2970	150	35	-	-	PUNCT
fcis-2970	150	36	connection	connection	NOUN
fcis-2970	150	37	layer	layer	NOUN
fcis-2970	150	38	output	output	NOUN
fcis-2970	150	39	.	.	PUNCT
fcis-2970	151	1	in	in	ADP
fcis-2970	151	2	model	model	NOUN
fcis-2970	151	3	5	5	NUM
fcis-2970	151	4	,	,	PUNCT
fcis-2970	151	5	the	the	DET
fcis-2970	151	6	same	same	ADJ
fcis-2970	151	7	pre	pre	ADJ
fcis-2970	151	8	-	-	ADJ
fcis-2970	151	9	trained	train	VERB
fcis-2970	151	10	word	word	NOUN
fcis-2970	151	11	vector	vector	NOUN
fcis-2970	151	12	was	be	AUX
fcis-2970	151	13	used	use	VERB
fcis-2970	151	14	to	to	PART
fcis-2970	151	15	input	input	VERB
fcis-2970	151	16	the	the	DET
fcis-2970	151	17	embedded	embed	VERB
fcis-2970	151	18	word	word	NOUN
fcis-2970	151	19	vector	vector	NOUN
fcis-2970	151	20	into	into	ADP
fcis-2970	151	21	the	the	DET
fcis-2970	151	22	downstream	downstream	ADJ
fcis-2970	151	23	bilstm	bilstm	NOUN
fcis-2970	151	24	-	-	PUNCT
fcis-2970	151	25	at	at	NOUN
fcis-2970	151	26	,	,	PUNCT
fcis-2970	151	27	and	and	CCONJ
fcis-2970	151	28	the	the	DET
fcis-2970	151	29	prediction	prediction	NOUN
fcis-2970	151	30	result	result	NOUN
fcis-2970	151	31	was	be	AUX
fcis-2970	151	32	obtained	obtain	VERB
fcis-2970	151	33	through	through	ADP
fcis-2970	151	34	the	the	DET
fcis-2970	151	35	fullconnection	fullconnection	NOUN
fcis-2970	151	36	layer	layer	NOUN
fcis-2970	151	37	output	output	NOUN
fcis-2970	151	38	.	.	PUNCT
fcis-2970	152	1	model	model	NOUN
fcis-2970	152	2	6	6	NUM
fcis-2970	152	3	inputs	input	NOUN
fcis-2970	152	4	the	the	DET
fcis-2970	152	5	first	first	ADJ
fcis-2970	152	6	label	label	NOUN
fcis-2970	152	7	[	[	X
fcis-2970	152	8	cls	cls	X
fcis-2970	152	9	]	]	X
fcis-2970	152	10	vector	vector	NOUN
fcis-2970	152	11	of	of	ADP
fcis-2970	152	12	the	the	DET
fcis-2970	152	13	ernie	ernie	PROPN
fcis-2970	152	14	model	model	PROPN
fcis-2970	152	15	output	output	NOUN
fcis-2970	152	16	sentence	sentence	NOUN
fcis-2970	152	17	into	into	ADP
fcis-2970	152	18	the	the	DET
fcis-2970	152	19	full	full	ADJ
fcis-2970	152	20	connection	connection	NOUN
fcis-2970	152	21	layer	layer	NOUN
fcis-2970	152	22	and	and	CCONJ
fcis-2970	152	23	softmax	softmax	NOUN
fcis-2970	152	24	to	to	PART
fcis-2970	152	25	get	get	VERB
fcis-2970	152	26	the	the	DET
fcis-2970	152	27	prediction	prediction	NOUN
fcis-2970	152	28	result	result	NOUN
fcis-2970	152	29	;	;	PUNCT
fcis-2970	152	30	in	in	ADP
fcis-2970	152	31	model	model	NOUN
fcis-2970	152	32	7	7	NUM
fcis-2970	152	33	,	,	PUNCT
fcis-2970	152	34	the	the	DET
fcis-2970	152	35	hidden	hidden	ADJ
fcis-2970	152	36	state	state	NOUN
fcis-2970	152	37	sequence	sequence	NOUN
fcis-2970	152	38	at	at	ADP
fcis-2970	152	39	the	the	DET
fcis-2970	152	40	last	last	ADJ
fcis-2970	152	41	layer	layer	NOUN
fcis-2970	152	42	of	of	ADP
fcis-2970	152	43	the	the	DET
fcis-2970	152	44	ernie	ernie	PROPN
fcis-2970	152	45	model	model	NOUN
fcis-2970	152	46	is	be	AUX
fcis-2970	152	47	used	use	VERB
fcis-2970	152	48	as	as	ADP
fcis-2970	152	49	the	the	DET
fcis-2970	152	50	word	word	NOUN
fcis-2970	152	51	vector	vector	NOUN
fcis-2970	152	52	of	of	ADP
fcis-2970	152	53	the	the	DET
fcis-2970	152	54	sentence	sentence	NOUN
fcis-2970	152	55	,	,	PUNCT
fcis-2970	152	56	and	and	CCONJ
fcis-2970	152	57	the	the	DET
fcis-2970	152	58	word	word	NOUN
fcis-2970	152	59	vector	vector	NOUN
fcis-2970	152	60	is	be	AUX
fcis-2970	152	61	input	input	VERB
fcis-2970	152	62	into	into	ADP
fcis-2970	152	63	the	the	DET
fcis-2970	152	64	bilstm	bilstm	NOUN
fcis-2970	152	65	model	model	NOUN
fcis-2970	152	66	to	to	PART
fcis-2970	152	67	get	get	VERB
fcis-2970	152	68	the	the	DET
fcis-2970	152	69	prediction	prediction	NOUN
fcis-2970	152	70	result	result	NOUN
fcis-2970	152	71	.	.	PUNCT
fcis-2970	153	1	model	model	NOUN
fcis-2970	153	2	8	8	NUM
fcis-2970	153	3	used	use	VERB
fcis-2970	153	4	ernie	ernie	PROPN
fcis-2970	153	5	pre	pre	ADJ
fcis-2970	153	6	-	-	ADJ
fcis-2970	153	7	training	training	ADJ
fcis-2970	153	8	technology	technology	NOUN
fcis-2970	153	9	to	to	PART
fcis-2970	153	10	get	get	VERB
fcis-2970	153	11	the	the	DET
fcis-2970	153	12	word	word	NOUN
fcis-2970	153	13	vector	vector	NOUN
fcis-2970	153	14	of	of	ADP
fcis-2970	153	15	the	the	DET
fcis-2970	153	16	text	text	NOUN
fcis-2970	153	17	,	,	PUNCT
fcis-2970	153	18	and	and	CCONJ
fcis-2970	153	19	then	then	ADV
fcis-2970	153	20	input	input	VERB
fcis-2970	153	21	the	the	DET
fcis-2970	153	22	word	word	NOUN
fcis-2970	153	23	vector	vector	NOUN
fcis-2970	153	24	into	into	ADP
fcis-2970	153	25	the	the	DET
fcis-2970	153	26	downstream	downstream	ADJ
fcis-2970	153	27	model	model	NOUN
fcis-2970	153	28	bilstm	bilstm	NOUN
fcis-2970	153	29	-	-	PUNCT
fcis-2970	153	30	at	at	NOUN
fcis-2970	153	31	to	to	PART
fcis-2970	153	32	get	get	VERB
fcis-2970	153	33	the	the	DET
fcis-2970	153	34	prediction	prediction	NOUN
fcis-2970	153	35	result	result	NOUN
fcis-2970	153	36	.	.	PUNCT
fcis-2970	154	1	as	as	SCONJ
fcis-2970	154	2	can	can	AUX
fcis-2970	154	3	be	be	AUX
fcis-2970	154	4	seen	see	VERB
fcis-2970	154	5	from	from	ADP
fcis-2970	154	6	the	the	DET
fcis-2970	154	7	table	table	NOUN
fcis-2970	154	8	,	,	PUNCT
fcis-2970	154	9	comparing	compare	VERB
fcis-2970	154	10	experiments	experiment	NOUN
fcis-2970	154	11	1	1	NUM
fcis-2970	154	12	,	,	PUNCT
fcis-2970	154	13	3	3	NUM
fcis-2970	154	14	,	,	PUNCT
fcis-2970	154	15	2	2	NUM
fcis-2970	154	16	,	,	PUNCT
fcis-2970	154	17	and	and	CCONJ
fcis-2970	154	18	4	4	NUM
fcis-2970	154	19	,	,	PUNCT
fcis-2970	154	20	it	it	PRON
fcis-2970	154	21	can	can	AUX
fcis-2970	154	22	be	be	AUX
fcis-2970	154	23	found	find	VERB
fcis-2970	154	24	that	that	SCONJ
fcis-2970	154	25	bidirectional	bidirectional	ADJ
fcis-2970	154	26	lstm	lstm	NOUN
fcis-2970	154	27	can	can	AUX
fcis-2970	154	28	achieve	achieve	VERB
fcis-2970	154	29	a	a	DET
fcis-2970	154	30	better	well	ADJ
fcis-2970	154	31	classification	classification	NOUN
fcis-2970	154	32	effect	effect	NOUN
fcis-2970	154	33	than	than	ADP
fcis-2970	154	34	one	one	NUM
fcis-2970	154	35	-	-	PUNCT
fcis-2970	154	36	way	way	NOUN
fcis-2970	154	37	lstm	lstm	NOUN
fcis-2970	154	38	,	,	PUNCT
fcis-2970	154	39	mainly	mainly	ADV
fcis-2970	154	40	because	because	SCONJ
fcis-2970	154	41	the	the	DET
fcis-2970	154	42	bidirectional	bidirectional	ADJ
fcis-2970	154	43	structure	structure	NOUN
fcis-2970	154	44	can	can	AUX
fcis-2970	154	45	extract	extract	VERB
fcis-2970	154	46	text	text	NOUN
fcis-2970	154	47	context	context	NOUN
fcis-2970	154	48	information	information	NOUN
fcis-2970	154	49	better	well	ADV
fcis-2970	154	50	.	.	PUNCT
fcis-2970	155	1	by	by	ADP
fcis-2970	155	2	comparing	compare	VERB
fcis-2970	155	3	experiments	experiment	NOUN
fcis-2970	155	4	1	1	NUM
fcis-2970	155	5	,	,	PUNCT
fcis-2970	155	6	2,3	2,3	NUM
fcis-2970	155	7	,	,	PUNCT
fcis-2970	155	8	5	5	NUM
fcis-2970	155	9	and	and	CCONJ
fcis-2970	155	10	7	7	NUM
fcis-2970	155	11	,	,	PUNCT
fcis-2970	155	12	and	and	CCONJ
fcis-2970	155	13	8	8	NUM
fcis-2970	155	14	,	,	PUNCT
fcis-2970	155	15	it	it	PRON
fcis-2970	155	16	can	can	AUX
fcis-2970	155	17	be	be	AUX
fcis-2970	155	18	found	find	VERB
fcis-2970	155	19	that	that	SCONJ
fcis-2970	155	20	by	by	ADP
fcis-2970	155	21	adding	add	VERB
fcis-2970	155	22	the	the	DET
fcis-2970	155	23	attention	attention	NOUN
fcis-2970	155	24	mechanism	mechanism	NOUN
fcis-2970	155	25	to	to	ADP
fcis-2970	155	26	the	the	DET
fcis-2970	155	27	original	original	ADJ
fcis-2970	155	28	model	model	NOUN
fcis-2970	155	29	,	,	PUNCT
fcis-2970	155	30	the	the	DET
fcis-2970	155	31	evaluation	evaluation	NOUN
fcis-2970	155	32	index	index	NOUN
fcis-2970	155	33	of	of	ADP
fcis-2970	155	34	classification	classification	NOUN
fcis-2970	155	35	results	result	NOUN
fcis-2970	155	36	is	be	AUX
fcis-2970	155	37	improved	improve	VERB
fcis-2970	155	38	,	,	PUNCT
fcis-2970	155	39	mainly	mainly	ADV
fcis-2970	155	40	because	because	SCONJ
fcis-2970	155	41	the	the	DET
fcis-2970	155	42	introduction	introduction	NOUN
fcis-2970	155	43	of	of	ADP
fcis-2970	155	44	the	the	DET
fcis-2970	155	45	attention	attention	NOUN
fcis-2970	155	46	mechanism	mechanism	NOUN
fcis-2970	155	47	can	can	AUX
fcis-2970	155	48	make	make	VERB
fcis-2970	155	49	the	the	DET
fcis-2970	155	50	model	model	NOUN
fcis-2970	155	51	focus	focus	VERB
fcis-2970	155	52	on	on	ADP
fcis-2970	155	53	words	word	NOUN
fcis-2970	155	54	that	that	PRON
fcis-2970	155	55	are	be	AUX
fcis-2970	155	56	more	more	ADV
fcis-2970	155	57	important	important	ADJ
fcis-2970	155	58	for	for	ADP
fcis-2970	155	59	emotion	emotion	NOUN
fcis-2970	155	60	analysis	analysis	NOUN
fcis-2970	155	61	.	.	PUNCT
fcis-2970	156	1	in	in	ADP
fcis-2970	156	2	contrast	contrast	NOUN
fcis-2970	156	3	,	,	PUNCT
fcis-2970	156	4	in	in	ADP
fcis-2970	156	5	experiment	experiment	NOUN
fcis-2970	156	6	6	6	NUM
fcis-2970	156	7	,	,	PUNCT
fcis-2970	156	8	the	the	DET
fcis-2970	156	9	first	first	ADJ
fcis-2970	156	10	sentence	sentence	NOUN
fcis-2970	156	11	flag	flag	NOUN
fcis-2970	156	12	[	[	X
fcis-2970	156	13	cls	cls	NOUN
fcis-2970	156	14	]	]	PUNCT
fcis-2970	156	15	was	be	AUX
fcis-2970	156	16	directly	directly	ADV
fcis-2970	156	17	used	use	VERB
fcis-2970	156	18	,	,	PUNCT
fcis-2970	156	19	and	and	CCONJ
fcis-2970	156	20	in	in	ADP
fcis-2970	156	21	experiment	experiment	NOUN
fcis-2970	156	22	7	7	NUM
fcis-2970	156	23	,	,	PUNCT
fcis-2970	156	24	the	the	DET
fcis-2970	156	25	text	text	NOUN
fcis-2970	156	26	word	word	NOUN
fcis-2970	156	27	vector	vector	NOUN
fcis-2970	156	28	obtained	obtain	VERB
fcis-2970	156	29	by	by	ADP
fcis-2970	156	30	ernie	ernie	PROPN
fcis-2970	156	31	pre	pre	NOUN
fcis-2970	156	32	-	-	NOUN
fcis-2970	156	33	training	training	NOUN
fcis-2970	156	34	was	be	AUX
fcis-2970	156	35	used	use	VERB
fcis-2970	156	36	as	as	ADP
fcis-2970	156	37	text	text	NOUN
fcis-2970	156	38	representation	representation	NOUN
fcis-2970	156	39	and	and	CCONJ
fcis-2970	156	40	input	input	NOUN
fcis-2970	156	41	into	into	ADP
fcis-2970	156	42	bilstm	bilstm	NOUN
fcis-2970	156	43	for	for	ADP
fcis-2970	156	44	further	further	ADJ
fcis-2970	156	45	feature	feature	NOUN
fcis-2970	156	46	extraction	extraction	NOUN
fcis-2970	156	47	.	.	PUNCT
fcis-2970	157	1	then	then	ADV
fcis-2970	157	2	,	,	PUNCT
fcis-2970	157	3	classification	classification	NOUN
fcis-2970	157	4	results	result	NOUN
fcis-2970	157	5	were	be	AUX
fcis-2970	157	6	output	output	VERB
fcis-2970	157	7	through	through	ADP
fcis-2970	157	8	the	the	DET
fcis-2970	157	9	full	full	ADJ
fcis-2970	157	10	connection	connection	NOUN
fcis-2970	157	11	layer	layer	NOUN
fcis-2970	157	12	.	.	PUNCT
fcis-2970	158	1	the	the	DET
fcis-2970	158	2	accuracy	accuracy	NOUN
fcis-2970	158	3	rate	rate	NOUN
fcis-2970	158	4	and	and	CCONJ
fcis-2970	158	5	f1	f1	NOUN
fcis-2970	158	6	value	value	NOUN
fcis-2970	158	7	of	of	ADP
fcis-2970	158	8	the	the	DET
fcis-2970	158	9	classification	classification	NOUN
fcis-2970	158	10	index	index	NOUN
fcis-2970	158	11	were	be	AUX
fcis-2970	158	12	slightly	slightly	ADV
fcis-2970	158	13	improved	improve	VERB
fcis-2970	158	14	.	.	PUNCT
fcis-2970	159	1	in	in	ADP
fcis-2970	159	2	experiment	experiment	NOUN
fcis-2970	159	3	8	8	NUM
fcis-2970	159	4	,	,	PUNCT
fcis-2970	159	5	an	an	DET
fcis-2970	159	6	attention	attention	NOUN
fcis-2970	159	7	mechanism	mechanism	NOUN
fcis-2970	159	8	was	be	AUX
fcis-2970	159	9	added	add	VERB
fcis-2970	159	10	to	to	PART
fcis-2970	159	11	increase	increase	VERB
fcis-2970	159	12	the	the	DET
fcis-2970	159	13	weight	weight	NOUN
fcis-2970	159	14	of	of	ADP
fcis-2970	159	15	more	more	ADV
fcis-2970	159	16	important	important	ADJ
fcis-2970	159	17	words	word	NOUN
fcis-2970	159	18	for	for	ADP
fcis-2970	159	19	emotional	emotional	ADJ
fcis-2970	159	20	labels	label	NOUN
fcis-2970	159	21	in	in	ADP
fcis-2970	159	22	text	text	NOUN
fcis-2970	159	23	statements	statement	NOUN
fcis-2970	159	24	,	,	PUNCT
fcis-2970	159	25	and	and	CCONJ
fcis-2970	159	26	the	the	DET
fcis-2970	159	27	accuracy	accuracy	NOUN
fcis-2970	159	28	rate	rate	NOUN
fcis-2970	159	29	of	of	ADP
fcis-2970	159	30	the	the	DET
fcis-2970	159	31	evaluation	evaluation	NOUN
fcis-2970	159	32	index	index	NOUN
fcis-2970	159	33	was	be	AUX
fcis-2970	159	34	improved	improve	VERB
fcis-2970	159	35	to	to	ADP
fcis-2970	159	36	a	a	DET
fcis-2970	159	37	certain	certain	ADJ
fcis-2970	159	38	extent	extent	NOUN
fcis-2970	159	39	.	.	PUNCT
fcis-2970	160	1	to	to	PART
fcis-2970	160	2	test	test	VERB
fcis-2970	160	3	the	the	DET
fcis-2970	160	4	effectiveness	effectiveness	NOUN
fcis-2970	160	5	of	of	ADP
fcis-2970	160	6	the	the	DET
fcis-2970	160	7	model	model	NOUN
fcis-2970	160	8	in	in	ADP
fcis-2970	160	9	this	this	DET
fcis-2970	160	10	paper	paper	NOUN
fcis-2970	160	11	,	,	PUNCT
fcis-2970	160	12	the	the	DET
fcis-2970	160	13	takeout	takeout	NOUN
fcis-2970	160	14	review	review	NOUN
fcis-2970	160	15	data	datum	NOUN
fcis-2970	160	16	set	set	VERB
fcis-2970	160	17	was	be	AUX
fcis-2970	160	18	used	use	VERB
fcis-2970	160	19	for	for	ADP
fcis-2970	160	20	training	training	NOUN
fcis-2970	160	21	and	and	CCONJ
fcis-2970	160	22	testing	testing	NOUN
fcis-2970	160	23	under	under	ADP
fcis-2970	160	24	the	the	DET
fcis-2970	160	25	same	same	ADJ
fcis-2970	160	26	environment	environment	NOUN
fcis-2970	160	27	.	.	PUNCT
fcis-2970	161	1	compared	compare	VERB
fcis-2970	161	2	with	with	ADP
fcis-2970	161	3	model	model	NOUN
fcis-2970	161	4	1	1	NUM
fcis-2970	161	5	and	and	CCONJ
fcis-2970	161	6	model	model	NOUN
fcis-2970	161	7	2	2	NUM
fcis-2970	161	8	,	,	PUNCT
fcis-2970	161	9	various	various	ADJ
fcis-2970	161	10	classification	classification	NOUN
fcis-2970	161	11	evaluation	evaluation	NOUN
fcis-2970	161	12	indexes	index	NOUN
fcis-2970	161	13	were	be	AUX
fcis-2970	161	14	improved	improve	VERB
fcis-2970	161	15	by	by	ADP
fcis-2970	161	16	adding	add	VERB
fcis-2970	161	17	bilstm	bilstm	NOUN
fcis-2970	161	18	.	.	PUNCT
fcis-2970	162	1	model	model	NOUN
fcis-2970	162	2	3	3	NUM
fcis-2970	162	3	incorporated	incorporate	VERB
fcis-2970	162	4	an	an	DET
fcis-2970	162	5	attention	attention	NOUN
fcis-2970	162	6	mechanism	mechanism	NOUN
fcis-2970	162	7	on	on	ADP
fcis-2970	162	8	the	the	DET
fcis-2970	162	9	basis	basis	NOUN
fcis-2970	162	10	of	of	ADP
fcis-2970	162	11	model	model	NOUN
fcis-2970	162	12	2	2	NUM
fcis-2970	162	13	,	,	PUNCT
fcis-2970	162	14	and	and	CCONJ
fcis-2970	162	15	the	the	DET
fcis-2970	162	16	indexes	index	NOUN
fcis-2970	162	17	were	be	AUX
fcis-2970	162	18	further	far	ADV
fcis-2970	162	19	improved	improve	VERB
fcis-2970	162	20	.	.	PUNCT
fcis-2970	163	1	specific	specific	ADJ
fcis-2970	163	2	experimental	experimental	ADJ
fcis-2970	163	3	results	result	NOUN
fcis-2970	163	4	are	be	AUX
fcis-2970	163	5	shown	show	VERB
fcis-2970	163	6	in	in	ADP
fcis-2970	163	7	table	table	NOUN
fcis-2970	163	8	4	4	NUM
fcis-2970	163	9	:	:	PUNCT
fcis-2970	163	10	table	table	NOUN
fcis-2970	163	11	4	4	NUM
fcis-2970	163	12	.	.	PUNCT
fcis-2970	163	13	three	three	NUM
fcis-2970	163	14	scheme	scheme	NOUN
fcis-2970	163	15	comparing	compare	VERB
fcis-2970	163	16	number	number	NOUN
fcis-2970	163	17	of	of	ADP
fcis-2970	163	18	the	the	DET
fcis-2970	163	19	model	model	NOUN
fcis-2970	163	20	model	model	PROPN
fcis-2970	163	21	precision	precision	PROPN
fcis-2970	163	22	recall	recall	PROPN
fcis-2970	163	23	f1	f1	PROPN
fcis-2970	163	24	1	1	NUM
fcis-2970	163	25	ernie	ernie	PROPN
fcis-2970	163	26	0.8929	0.8929	NUM
fcis-2970	163	27	0.8841	0.8841	NUM
fcis-2970	163	28	0.8883	0.8883	NUM
fcis-2970	163	29	2	2	NUM
fcis-2970	163	30	ernie	ernie	NOUN
fcis-2970	163	31	-	-	PUNCT
fcis-2970	163	32	bilstm	bilstm	NOUN
fcis-2970	163	33	0.8993	0.8993	NUM
fcis-2970	163	34	0.8890	0.8890	NUM
fcis-2970	163	35	0.8937	0.8937	NUM
fcis-2970	163	36	3	3	NUM
fcis-2970	163	37	ernie	ernie	NOUN
fcis-2970	163	38	-	-	PUNCT
fcis-2970	163	39	bilstm	bilstm	NOUN
fcis-2970	163	40	-	-	PUNCT
fcis-2970	163	41	at	at	ADP
fcis-2970	163	42	0.9061	0.9061	NUM
fcis-2970	163	43	0.9020	0.9020	NUM
fcis-2970	163	44	0.9040	0.9040	NUM
fcis-2970	163	45	4	4	NUM
fcis-2970	163	46	.	.	PUNCT
fcis-2970	163	47	conclusion	conclusion	NOUN
fcis-2970	163	48	in	in	ADP
fcis-2970	163	49	this	this	DET
fcis-2970	163	50	paper	paper	NOUN
fcis-2970	163	51	,	,	PUNCT
fcis-2970	163	52	we	we	PRON
fcis-2970	163	53	propose	propose	VERB
fcis-2970	163	54	a	a	DET
fcis-2970	163	55	text	text	NOUN
fcis-2970	163	56	emotion	emotion	NOUN
fcis-2970	163	57	classification	classification	NOUN
fcis-2970	163	58	model	model	NOUN
fcis-2970	163	59	combining	combine	VERB
fcis-2970	163	60	ernie	ernie	PROPN
fcis-2970	163	61	and	and	CCONJ
fcis-2970	163	62	bilstm	bilstm	NOUN
fcis-2970	163	63	-	-	PUNCT
fcis-2970	163	64	at	at	NOUN
fcis-2970	163	65	to	to	PART
fcis-2970	163	66	achieve	achieve	VERB
fcis-2970	163	67	sentence	sentence	NOUN
fcis-2970	163	68	-	-	PUNCT
fcis-2970	163	69	level	level	NOUN
fcis-2970	163	70	emotion	emotion	NOUN
fcis-2970	163	71	classification	classification	NOUN
fcis-2970	163	72	.	.	PUNCT
fcis-2970	164	1	we	we	PRON
fcis-2970	164	2	use	use	VERB
fcis-2970	164	3	the	the	DET
fcis-2970	164	4	ernie	ernie	NOUN
fcis-2970	164	5	pretraining	pretraine	VERB
fcis-2970	164	6	model	model	NOUN
fcis-2970	164	7	to	to	PART
fcis-2970	164	8	get	get	VERB
fcis-2970	164	9	word	word	NOUN
fcis-2970	164	10	vector	vector	NOUN
fcis-2970	164	11	representation	representation	NOUN
fcis-2970	164	12	integrating	integrate	VERB
fcis-2970	164	13	text	text	NOUN
fcis-2970	164	14	context	context	NOUN
fcis-2970	164	15	.	.	PUNCT
fcis-2970	165	1	we	we	PRON
fcis-2970	165	2	use	use	VERB
fcis-2970	165	3	neural	neural	ADJ
fcis-2970	165	4	networks	network	NOUN
fcis-2970	165	5	and	and	CCONJ
fcis-2970	165	6	attention	attention	NOUN
fcis-2970	165	7	mechanisms	mechanism	NOUN
fcis-2970	165	8	to	to	PART
fcis-2970	165	9	extract	extract	VERB
fcis-2970	165	10	text	text	NOUN
fcis-2970	165	11	word	word	NOUN
fcis-2970	165	12	vector	vector	NOUN
fcis-2970	165	13	features	feature	VERB
fcis-2970	165	14	to	to	PART
fcis-2970	165	15	better	well	ADV
fcis-2970	165	16	understand	understand	VERB
fcis-2970	165	17	text	text	NOUN
fcis-2970	165	18	semantic	semantic	ADJ
fcis-2970	165	19	information	information	NOUN
fcis-2970	165	20	.	.	PUNCT
fcis-2970	166	1	the	the	DET
fcis-2970	166	2	model	model	NOUN
fcis-2970	166	3	in	in	ADP
fcis-2970	166	4	this	this	DET
fcis-2970	166	5	paper	paper	NOUN
fcis-2970	166	6	is	be	AUX
fcis-2970	166	7	tested	test	VERB
fcis-2970	166	8	on	on	ADP
fcis-2970	166	9	the	the	DET
fcis-2970	166	10	test	test	NOUN
fcis-2970	166	11	set	set	NOUN
fcis-2970	166	12	,	,	PUNCT
fcis-2970	166	13	and	and	CCONJ
fcis-2970	166	14	some	some	DET
fcis-2970	166	15	evaluation	evaluation	NOUN
fcis-2970	166	16	indexes	index	NOUN
fcis-2970	166	17	are	be	AUX
fcis-2970	166	18	improved	improve	VERB
fcis-2970	166	19	to	to	ADP
fcis-2970	166	20	a	a	DET
fcis-2970	166	21	certain	certain	ADJ
fcis-2970	166	22	extent	extent	NOUN
fcis-2970	166	23	,	,	PUNCT
fcis-2970	166	24	indicating	indicate	VERB
fcis-2970	166	25	that	that	SCONJ
fcis-2970	166	26	the	the	DET
fcis-2970	166	27	model	model	NOUN
fcis-2970	166	28	has	have	VERB
fcis-2970	166	29	a	a	DET
fcis-2970	166	30	certain	certain	ADJ
fcis-2970	166	31	effect	effect	NOUN
fcis-2970	166	32	.	.	PUNCT
fcis-2970	167	1	only	only	ADV
fcis-2970	167	2	binary	binary	ADJ
fcis-2970	167	3	task	task	NOUN
fcis-2970	167	4	is	be	AUX
fcis-2970	167	5	considered	consider	VERB
fcis-2970	167	6	in	in	ADP
fcis-2970	167	7	this	this	DET
fcis-2970	167	8	model	model	NOUN
fcis-2970	167	9	.	.	PUNCT
fcis-2970	168	1	in	in	ADP
fcis-2970	168	2	future	future	ADJ
fcis-2970	168	3	work	work	NOUN
fcis-2970	168	4	,	,	PUNCT
fcis-2970	168	5	multi	multi	ADJ
fcis-2970	168	6	-	-	ADJ
fcis-2970	168	7	emotion	emotion	ADJ
fcis-2970	168	8	text	text	NOUN
fcis-2970	168	9	classification	classification	NOUN
fcis-2970	168	10	experiments	experiment	NOUN
fcis-2970	168	11	can	can	AUX
fcis-2970	168	12	be	be	AUX
fcis-2970	168	13	considered	consider	VERB
fcis-2970	168	14	,	,	PUNCT
fcis-2970	168	15	and	and	CCONJ
fcis-2970	168	16	the	the	DET
fcis-2970	168	17	application	application	NOUN
fcis-2970	168	18	of	of	ADP
fcis-2970	168	19	the	the	DET
fcis-2970	168	20	latest	late	ADJ
fcis-2970	168	21	ernie3.0	ernie3.0	PUNCT
fcis-2970	168	22	model	model	NOUN
fcis-2970	168	23	in	in	ADP
fcis-2970	168	24	emotion	emotion	NOUN
fcis-2970	168	25	classification	classification	NOUN
fcis-2970	168	26	tasks	task	NOUN
fcis-2970	168	27	can	can	AUX
fcis-2970	168	28	also	also	ADV
fcis-2970	168	29	be	be	AUX
fcis-2970	168	30	tried	try	VERB
fcis-2970	168	31	.	.	PUNCT
fcis-2970	169	1	acknowledgements	acknowledgement	NOUN
fcis-2970	169	2	this	this	DET
fcis-2970	169	3	work	work	NOUN
fcis-2970	169	4	was	be	AUX
fcis-2970	169	5	partially	partially	ADV
fcis-2970	169	6	supported	support	VERB
fcis-2970	169	7	by	by	ADP
fcis-2970	169	8	natural	natural	ADJ
fcis-2970	169	9	science	science	NOUN
fcis-2970	169	10	foundation	foundation	PROPN
fcis-2970	169	11	of	of	ADP
fcis-2970	169	12	china	china	PROPN
fcis-2970	169	13	(	(	PUNCT
fcis-2970	169	14	12161074	12161074	NUM
fcis-2970	169	15	)	)	PUNCT
fcis-2970	169	16	,	,	PUNCT
fcis-2970	169	17	natural	natural	ADJ
fcis-2970	169	18	science	science	NOUN
fcis-2970	169	19	foundation	foundation	PROPN
fcis-2970	169	20	of	of	ADP
fcis-2970	169	21	guangdong	guangdong	PROPN
fcis-2970	169	22	province	province	PROPN
fcis-2970	169	23	(	(	PUNCT
fcis-2970	169	24	2022a1515010978	2022a1515010978	NOUN
fcis-2970	169	25	)	)	PUNCT
fcis-2970	169	26	and	and	CCONJ
fcis-2970	169	27	natural	natural	ADJ
fcis-2970	169	28	science	science	NOUN
fcis-2970	169	29	foundation	foundation	NOUN
fcis-2970	169	30	of	of	ADP
fcis-2970	169	31	guangdong	guangdong	PROPN
fcis-2970	169	32	ocean	ocean	PROPN
fcis-2970	169	33	university	university	PROPN
fcis-2970	169	34	(	(	PUNCT
fcis-2970	169	35	r17083	r17083	PROPN
fcis-2970	169	36	,	,	PUNCT
fcis-2970	169	37	c17201	c17201	PROPN
fcis-2970	169	38	,	,	PUNCT
fcis-2970	169	39	p16091	p16091	PROPN
fcis-2970	169	40	)	)	PUNCT
fcis-2970	169	41	.	.	PUNCT
fcis-2970	170	1	references	reference	NOUN
fcis-2970	170	2	[	[	X
fcis-2970	170	3	1	1	NUM
fcis-2970	170	4	]	]	PUNCT
fcis-2970	170	5	cnnic	cnnic	ADJ
fcis-2970	170	6	internet	internet	NOUN
fcis-2970	170	7	research	research	NOUN
fcis-2970	170	8	.	.	PUNCT
fcis-2970	171	1	the	the	DET
fcis-2970	171	2	49th	49th	ADJ
fcis-2970	171	3	statistical	statistical	ADJ
fcis-2970	171	4	report	report	NOUN
fcis-2970	171	5	on	on	ADP
fcis-2970	171	6	the	the	DET
fcis-2970	171	7	development	development	NOUN
fcis-2970	171	8	of	of	ADP
fcis-2970	171	9	internet	internet	NOUN
fcis-2970	171	10	in	in	ADP
fcis-2970	171	11	china	china	PROPN
fcis-2970	172	1	[	[	X
fcis-2970	172	2	r	r	X
fcis-2970	172	3	]	]	PUNCT
fcis-2970	172	4	.	.	PUNCT
fcis-2970	173	1	beijing	beijing	PROPN
fcis-2970	173	2	:	:	PUNCT
fcis-2970	174	1	china	china	PROPN
fcis-2970	174	2	internet	internet	PROPN
fcis-2970	174	3	network	network	PROPN
fcis-2970	174	4	information	information	NOUN
fcis-2970	174	5	center	center	NOUN
fcis-2970	174	6	,	,	PUNCT
fcis-2970	174	7	2022	2022	NUM
fcis-2970	174	8	.	.	PUNCT
fcis-2970	175	1	[	[	X
fcis-2970	175	2	2	2	NUM
fcis-2970	175	3	]	]	X
fcis-2970	175	4	wang	wang	PROPN
fcis-2970	175	5	ting	ting	PROPN
fcis-2970	175	6	,	,	PUNCT
fcis-2970	175	7	yang	yang	PROPN
fcis-2970	175	8	wenzhong	wenzhong	PROPN
fcis-2970	175	9	.	.	PUNCT
fcis-2970	176	1	review	review	NOUN
fcis-2970	176	2	on	on	ADP
fcis-2970	176	3	text	text	NOUN
fcis-2970	176	4	sentiment	sentiment	NOUN
fcis-2970	176	5	analysis	analysis	NOUN
fcis-2970	176	6	methods	method	NOUN
fcis-2970	176	7	[	[	X
fcis-2970	176	8	j	j	X
fcis-2970	176	9	]	]	X
fcis-2970	176	10	.	.	PUNCT
fcis-2970	177	1	computer	computer	NOUN
fcis-2970	177	2	engineering	engineering	NOUN
fcis-2970	177	3	and	and	CCONJ
fcis-2970	177	4	applications,2021,57(12):11	applications,2021,57(12):11	PROPN
fcis-2970	177	5	-	-	PUNCT
fcis-2970	177	6	24	24	NUM
fcis-2970	177	7	.	.	PUNCT
fcis-2970	178	1	[	[	X
fcis-2970	178	2	3	3	NUM
fcis-2970	178	3	]	]	X
fcis-2970	178	4	li	li	PROPN
fcis-2970	178	5	yuqing	yuqing	PROPN
fcis-2970	178	6	,	,	PUNCT
fcis-2970	178	7	li	li	PROPN
fcis-2970	178	8	xin	xin	PROPN
fcis-2970	178	9	,	,	PUNCT
fcis-2970	178	10	han	han	PROPN
fcis-2970	178	11	xu	xu	PROPN
fcis-2970	178	12	,	,	PUNCT
fcis-2970	178	13	song	song	PROPN
fcis-2970	178	14	dandan	dandan	PROPN
fcis-2970	178	15	,	,	PUNCT
fcis-2970	178	16	liao	liao	PROPN
fcis-2970	178	17	lejian	lejian	PROPN
fcis-2970	178	18	.	.	PUNCT
fcis-2970	179	1	a	a	DET
fcis-2970	179	2	bilingual	bilingual	ADJ
fcis-2970	179	3	lexicon	lexicon	NOUN
fcis-2970	179	4	-	-	PUNCT
fcis-2970	179	5	based	base	VERB
fcis-2970	179	6	multi	multi	ADJ
fcis-2970	179	7	-	-	ADJ
fcis-2970	179	8	class	class	ADJ
fcis-2970	179	9	semantic	semantic	ADJ
fcis-2970	179	10	orientation	orientation	NOUN
fcis-2970	179	11	analysis	analysis	NOUN
fcis-2970	179	12	for	for	ADP
fcis-2970	179	13	microblogs[j	microblogs[j	NOUN
fcis-2970	179	14	]	]	PUNCT
fcis-2970	179	15	.	.	PUNCT
fcis-2970	180	1	acta	acta	PROPN
fcis-2970	180	2	electronica	electronica	PROPN
fcis-2970	180	3	sinica,2016,44(09):2068	sinica,2016,44(09):2068	PROPN
fcis-2970	180	4	-	-	PUNCT
fcis-2970	180	5	2073	2073	NUM
fcis-2970	180	6	.	.	PUNCT
fcis-2970	181	1	[	[	X
fcis-2970	181	2	4	4	NUM
fcis-2970	181	3	]	]	X
fcis-2970	181	4	tang	tang	X
fcis-2970	181	5	huifeng	huifeng	PROPN
fcis-2970	181	6	,	,	PUNCT
fcis-2970	181	7	tan	tan	PROPN
fcis-2970	181	8	songbo	songbo	PROPN
fcis-2970	181	9	,	,	PUNCT
fcis-2970	181	10	cheng	cheng	PROPN
fcis-2970	181	11	xueqi	xueqi	PROPN
fcis-2970	181	12	.	.	PUNCT
fcis-2970	182	1	research	research	NOUN
fcis-2970	182	2	on	on	ADP
fcis-2970	182	3	sentiment	sentiment	NOUN
fcis-2970	182	4	classification	classification	NOUN
fcis-2970	182	5	of	of	ADP
fcis-2970	182	6	chinese	chinese	ADJ
fcis-2970	182	7	reviews	review	NOUN
fcis-2970	182	8	based	base	VERB
fcis-2970	182	9	on	on	ADP
fcis-2970	182	10	supervised	supervised	ADJ
fcis-2970	182	11	machine	machine	NOUN
fcis-2970	182	12	learning	learn	VERB
fcis-2970	182	13	techniques	technique	NOUN
fcis-2970	182	14	.	.	PUNCT
fcis-2970	182	15	,	,	PUNCT
fcis-2970	182	16	2007	2007	NUM
fcis-2970	182	17	,	,	PUNCT
fcis-2970	182	18	21(6):88	21(6):88	NUM
fcis-2970	182	19	-	-	SYM
fcis-2970	182	20	94	94	NUM
fcis-2970	182	21	.	.	PUNCT
fcis-2970	183	1	[	[	X
fcis-2970	183	2	5	5	NUM
fcis-2970	183	3	]	]	PUNCT
fcis-2970	183	4	h	h	NOUN
fcis-2970	183	5	jelodar	jelodar	NOUN
fcis-2970	183	6	,	,	PUNCT
fcis-2970	183	7	wang	wang	PROPN
fcis-2970	183	8	y	y	PROPN
fcis-2970	183	9	,	,	PUNCT
fcis-2970	183	10	orji	orji	PROPN
fcis-2970	183	11	r	r	PROPN
fcis-2970	183	12	,	,	PUNCT
fcis-2970	183	13	et	et	PROPN
fcis-2970	183	14	al	al	PROPN
fcis-2970	183	15	.	.	PUNCT
fcis-2970	184	1	deep	deep	ADJ
fcis-2970	184	2	sentiment	sentiment	NOUN
fcis-2970	184	3	classification	classification	NOUN
fcis-2970	184	4	and	and	CCONJ
fcis-2970	184	5	topic	topic	NOUN
fcis-2970	184	6	discovery	discovery	NOUN
fcis-2970	184	7	on	on	ADP
fcis-2970	184	8	novel	novel	ADJ
fcis-2970	184	9	coronavirusor	coronavirusor	NOUN
fcis-2970	184	10	covid-19	covid-19	PROPN
fcis-2970	184	11	online	online	ADJ
fcis-2970	184	12	discussions	discussion	NOUN
fcis-2970	184	13	:	:	PUNCT
fcis-2970	184	14	nlp	nlp	NOUN
fcis-2970	184	15	using	use	VERB
fcis-2970	184	16	lstm	lstm	PROPN
fcis-2970	184	17	recurrent	recurrent	ADJ
fcis-2970	184	18	neural	neural	ADJ
fcis-2970	184	19	network	network	NOUN
fcis-2970	184	20	approach[j	approach[j	PROPN
fcis-2970	184	21	]	]	PUNCT
fcis-2970	184	22	.	.	PUNCT
fcis-2970	185	1	arxiv,2020	arxiv,2020	ADP
fcis-2970	185	2	.	.	PUNCT
fcis-2970	186	1	[	[	X
fcis-2970	186	2	6	6	NUM
fcis-2970	186	3	]	]	X
fcis-2970	186	4	luo	luo	PROPN
fcis-2970	186	5	fan	fan	PROPN
fcis-2970	186	6	,	,	PUNCT
fcis-2970	186	7	wang	wang	PROPN
fcis-2970	186	8	houfeng	houfeng	PROPN
fcis-2970	186	9	.	.	PUNCT
fcis-2970	187	1	chinese	chinese	ADJ
fcis-2970	187	2	text	text	NOUN
fcis-2970	187	3	sentiment	sentiment	NOUN
fcis-2970	187	4	classification	classification	NOUN
fcis-2970	187	5	based	base	VERB
fcis-2970	187	6	on	on	ADP
fcis-2970	187	7	hierarchical	hierarchical	ADJ
fcis-2970	187	8	network	network	NOUN
fcis-2970	187	9	of	of	ADP
fcis-2970	187	10	rnn	rnn	PROPN
fcis-2970	187	11	and	and	CCONJ
fcis-2970	187	12	cnn	cnn	PROPN
fcis-2970	187	13	.	.	PUNCT
fcis-2970	188	1	journal	journal	PROPN
fcis-2970	188	2	of	of	ADP
fcis-2970	188	3	beijing	beijing	PROPN
fcis-2970	188	4	university	university	PROPN
fcis-2970	188	5	(	(	PUNCT
fcis-2970	188	6	natural	natural	ADJ
fcis-2970	188	7	science	science	NOUN
fcis-2970	188	8	)	)	PUNCT
fcis-2970	188	9	,	,	PUNCT
fcis-2970	188	10	2018	2018	NUM
fcis-2970	188	11	,	,	PUNCT
fcis-2970	188	12	v.54	v.54	ADP
fcis-2970	188	13	;	;	PUNCT
fcis-2970	188	14	no.287(03):4	no.287(03):4	NOUN
fcis-2970	188	15	-	-	PUNCT
fcis-2970	188	16	10	10	NUM
fcis-2970	188	17	.	.	PUNCT
fcis-2970	189	1	[	[	X
fcis-2970	189	2	7	7	NUM
fcis-2970	189	3	]	]	SYM
fcis-2970	189	4	bahdanau	bahdanau	NOUN
fcis-2970	189	5	,	,	PUNCT
fcis-2970	189	6	dzmitry	dzmitry	NOUN
fcis-2970	189	7	et	et	PROPN
fcis-2970	189	8	al	al	PROPN
fcis-2970	189	9	.	.	PUNCT
fcis-2970	190	1	“	"	PUNCT
fcis-2970	190	2	neural	neural	ADJ
fcis-2970	190	3	machine	machine	NOUN
fcis-2970	190	4	translation	translation	NOUN
fcis-2970	190	5	by	by	ADP
fcis-2970	190	6	jointly	jointly	ADV
fcis-2970	190	7	learning	learn	VERB
fcis-2970	190	8	to	to	PART
fcis-2970	190	9	align	align	VERB
fcis-2970	190	10	and	and	CCONJ
fcis-2970	190	11	translate[j	translate[j	NOUN
fcis-2970	190	12	]	]	PUNCT
fcis-2970	190	13	.	.	PUNCT
fcis-2970	190	14	”	"	PUNCT
fcis-2970	191	1	corrabs/1409.0473	corrabs/1409.0473	VERB
fcis-2970	191	2	(	(	PUNCT
fcis-2970	191	3	2015	2015	NUM
fcis-2970	191	4	):	):	PUNCT
fcis-2970	191	5	n.	n.	PROPN
fcis-2970	191	6	pag	pag	PROPN
fcis-2970	191	7	.	.	PUNCT
fcis-2970	192	1	[	[	X
fcis-2970	192	2	8	8	NUM
fcis-2970	192	3	]	]	X
fcis-2970	192	4	liu	liu	PROPN
fcis-2970	192	5	fasheng	fasheng	PROPN
fcis-2970	192	6	,	,	PUNCT
fcis-2970	192	7	xu	xu	PROPN
fcis-2970	192	8	minlin	minlin	PROPN
fcis-2970	192	9	,	,	PUNCT
fcis-2970	192	10	deng	deng	PROPN
fcis-2970	192	11	xiaohong	xiaohong	PROPN
fcis-2970	192	12	.	.	PUNCT
fcis-2970	193	1	research	research	NOUN
fcis-2970	193	2	on	on	ADP
fcis-2970	193	3	emotion	emotion	NOUN
fcis-2970	193	4	analysis	analysis	NOUN
fcis-2970	193	5	combining	combine	VERB
fcis-2970	193	6	attention	attention	NOUN
fcis-2970	193	7	mechanism	mechanism	NOUN
fcis-2970	193	8	and	and	CCONJ
fcis-2970	193	9	sentence	sentence	NOUN
fcis-2970	193	10	ordering	order	VERB
fcis-2970	193	11	[	[	X
fcis-2970	193	12	j	j	X
fcis-2970	193	13	]	]	X
fcis-2970	193	14	.	.	PUNCT
fcis-2970	194	1	computer	computer	NOUN
fcis-2970	194	2	engineering	engineering	NOUN
fcis-2970	194	3	and	and	CCONJ
fcis-2970	194	4	applications,20,56(13):12	applications,20,56(13):12	PROPN
fcis-2970	194	5	-	-	PROPN
fcis-2970	194	6	19	19	NUM
fcis-2970	194	7	.	.	PUNCT
fcis-2970	194	8	75	75	NUM
fcis-2970	195	1	[	[	SYM
fcis-2970	195	2	9	9	NUM
fcis-2970	195	3	]	]	SYM
fcis-2970	195	4	li	li	PROPN
fcis-2970	195	5	zhou	zhou	PROPN
fcis-2970	195	6	-	-	PUNCT
fcis-2970	195	7	jun	jun	PROPN
fcis-2970	195	8	,	,	PUNCT
fcis-2970	195	9	fan	fan	PROPN
fcis-2970	195	10	yu	yu	PROPN
fcis-2970	195	11	,	,	PUNCT
fcis-2970	195	12	wu	wu	PROPN
fcis-2970	195	13	xian	xian	PROPN
fcis-2970	195	14	-	-	PUNCT
fcis-2970	195	15	jie	jie	PROPN
fcis-2970	195	16	.	.	PUNCT
fcis-2970	196	1	a	a	DET
fcis-2970	196	2	review	review	NOUN
fcis-2970	196	3	of	of	ADP
fcis-2970	196	4	pre	pre	ADJ
fcis-2970	196	5	-	-	ADJ
fcis-2970	196	6	training	training	ADJ
fcis-2970	196	7	techniques	technique	NOUN
fcis-2970	196	8	for	for	ADP
fcis-2970	196	9	natural	natural	ADJ
fcis-2970	196	10	language	language	NOUN
fcis-2970	196	11	processing	processing	NOUN
fcis-2970	196	12	[	[	X
fcis-2970	196	13	j	j	X
fcis-2970	196	14	]	]	X
fcis-2970	196	15	.	.	PUNCT
fcis-2970	197	1	computer	computer	NOUN
fcis-2970	197	2	science,20,47(03):162	science,20,47(03):162	NOUN
fcis-2970	197	3	-	-	PUNCT
fcis-2970	197	4	173	173	NUM
fcis-2970	197	5	.	.	PUNCT
fcis-2970	198	1	[	[	X
fcis-2970	198	2	10	10	NUM
fcis-2970	198	3	]	]	X
fcis-2970	198	4	tomas	tomas	PROPN
fcis-2970	198	5	mikolov	mikolov	PROPN
fcis-2970	198	6	et	et	PROPN
fcis-2970	198	7	al	al	PROPN
fcis-2970	198	8	.	.	PROPN
fcis-2970	198	9	efficient	efficient	ADJ
fcis-2970	198	10	estimation	estimation	NOUN
fcis-2970	198	11	of	of	ADP
fcis-2970	198	12	word	word	NOUN
fcis-2970	198	13	representations	representation	NOUN
fcis-2970	198	14	in	in	ADP
fcis-2970	198	15	vector	vector	NOUN
fcis-2970	198	16	space[j	space[j	NOUN
fcis-2970	198	17	]	]	PUNCT
fcis-2970	198	18	.	.	PUNCT
fcis-2970	199	1	corr	corr	PROPN
fcis-2970	199	2	,	,	PUNCT
fcis-2970	199	3	2013	2013	NUM
fcis-2970	199	4	,	,	PUNCT
fcis-2970	199	5	abs/1301.3781	abs/1301.3781	PROPN
fcis-2970	200	1	[	[	X
fcis-2970	200	2	11	11	NUM
fcis-2970	200	3	]	]	X
fcis-2970	200	4	peter	peter	PROPN
fcis-2970	200	5	m	m	PROPN
fcis-2970	200	6	e	e	PROPN
fcis-2970	200	7	,	,	PUNCT
fcis-2970	200	8	neumann	neumann	PROPN
fcis-2970	200	9	m	m	PROPN
fcis-2970	200	10	,	,	PUNCT
fcis-2970	200	11	iyyer	iyyer	PROPN
fcis-2970	200	12	m	m	PROPN
fcis-2970	200	13	,	,	PUNCT
fcis-2970	200	14	et	et	PROPN
fcis-2970	200	15	al	al	PROPN
fcis-2970	200	16	.	.	PUNCT
fcis-2970	201	1	deep	deep	ADJ
fcis-2970	201	2	contextualized	contextualized	ADJ
fcis-2970	201	3	word	word	NOUN
fcis-2970	201	4	representations[j	representations[j	NOUN
fcis-2970	201	5	]	]	PUNCT
fcis-2970	201	6	.	.	PUNCT
fcis-2970	202	1	arxiv：1802	arxiv：1802	PROPN
fcis-2970	202	2	.	.	PUNCT
fcis-2970	202	3	05365，2018	05365，2018	PROPN
fcis-2970	202	4	.	.	PUNCT
fcis-2970	203	1	[	[	X
fcis-2970	203	2	12	12	NUM
fcis-2970	203	3	]	]	X
fcis-2970	203	4	radford	radford	PROPN
fcis-2970	203	5	a	a	PROPN
fcis-2970	203	6	,	,	PUNCT
fcis-2970	203	7	narasimhan	narasimhan	PROPN
fcis-2970	203	8	k	k	PROPN
fcis-2970	203	9	,	,	PUNCT
fcis-2970	203	10	salimans	salimans	PROPN
fcis-2970	203	11	t	t	PROPN
fcis-2970	203	12	,	,	PUNCT
fcis-2970	203	13	et	et	PROPN
fcis-2970	203	14	al	al	PROPN
fcis-2970	203	15	.	.	PUNCT
fcis-2970	204	1	improving	improve	VERB
fcis-2970	204	2	language	language	NOUN
fcis-2970	204	3	understanding	understanding	NOUN
fcis-2970	204	4	by	by	ADP
fcis-2970	204	5	generative	generative	ADJ
fcis-2970	204	6	pretraining[j	pretraining[j	NOUN
fcis-2970	204	7	/	/	SYM
fcis-2970	204	8	ol].[2020	ol].[2020	PROPN
fcis-2970	204	9	-	-	PUNCT
fcis-2970	204	10	0701].https://www.cs.ubc.ca/~amuham01	0701].https://www.cs.ubc.ca/~amuham01	NOUN
fcis-2970	204	11	/	/	SYM
fcis-2970	204	12	ling530	ling530	PROPN
fcis-2970	204	13	/	/	SYM
fcis-2970	204	14	papers	paper	NOUN
fcis-2970	204	15	/	/	SYM
fcis-2970	204	16	radf	radf	NOUN
fcis-2970	204	17	ord2018improving.pdf	ord2018improving.pdf	NOUN
fcis-2970	204	18	.	.	PUNCT
fcis-2970	205	1	[	[	X
fcis-2970	205	2	13	13	NUM
fcis-2970	205	3	]	]	X
fcis-2970	205	4	gao	gao	PROPN
fcis-2970	205	5	z	z	PROPN
fcis-2970	205	6	j	j	PROPN
fcis-2970	205	7	,	,	PUNCT
fcis-2970	205	8	feng	feng	PROPN
fcis-2970	205	9	a	a	X
fcis-2970	205	10	,	,	PUNCT
fcis-2970	205	11	song	song	NOUN
fcis-2970	205	12	x	x	SYM
fcis-2970	205	13	y	y	PROPN
fcis-2970	205	14	,	,	PUNCT
fcis-2970	205	15	et	et	PROPN
fcis-2970	205	16	al	al	PROPN
fcis-2970	205	17	.	.	PROPN
fcis-2970	205	18	target	target	NOUN
fcis-2970	205	19	-	-	PUNCT
fcis-2970	205	20	dependent	dependent	ADJ
fcis-2970	205	21	sentiment	sentiment	NOUN
fcis-2970	205	22	classification	classification	NOUN
fcis-2970	205	23	with	with	ADP
fcis-2970	205	24	bert[j	bert[j	NOUN
fcis-2970	205	25	]	]	PUNCT
fcis-2970	205	26	.	.	PUNCT
fcis-2970	206	1	ieee	ieee	NOUN
fcis-2970	206	2	access	access	NOUN
fcis-2970	206	3	,	,	PUNCT
fcis-2970	206	4	2019,7	2019,7	NUM
fcis-2970	206	5	:	:	PUNCT
fcis-2970	206	6	154290154299	154290154299	NUM
fcis-2970	206	7	.	.	PUNCT
fcis-2970	207	1	[	[	X
fcis-2970	207	2	14	14	NUM
fcis-2970	207	3	]	]	X
fcis-2970	207	4	yu	yu	PROPN
fcis-2970	207	5	sun	sun	PROPN
fcis-2970	207	6	et	et	PROPN
fcis-2970	207	7	al	al	PROPN
fcis-2970	207	8	.	.	PROPN
fcis-2970	207	9	ernie	ernie	PROPN
fcis-2970	207	10	:	:	PUNCT
fcis-2970	207	11	enhanced	enhance	VERB
fcis-2970	207	12	representation	representation	NOUN
fcis-2970	207	13	through	through	ADP
fcis-2970	207	14	knowledge	knowledge	NOUN
fcis-2970	207	15	integration.[j	integration.[j	PROPN
fcis-2970	207	16	]	]	PUNCT
fcis-2970	207	17	.	.	PUNCT
fcis-2970	208	1	corr	corr	PROPN
fcis-2970	208	2	,	,	PUNCT
fcis-2970	208	3	2019	2019	NUM
fcis-2970	208	4	,	,	PUNCT
fcis-2970	208	5	abs/1904.09223	abs/1904.09223	ADP
fcis-2970	209	1	[	[	X
fcis-2970	209	2	15	15	NUM
fcis-2970	209	3	]	]	X
fcis-2970	209	4	jingsheng	jingsheng	PROPN
fcis-2970	209	5	lei	lei	PROPN
fcis-2970	209	6	,	,	PUNCT
fcis-2970	209	7	ye	ye	PROPN
fcis-2970	209	8	qian	qian	PROPN
fcis-2970	209	9	.	.	PUNCT
fcis-2970	209	10	"	"	PUNCT
fcis-2970	210	1	chinese	chinese	ADJ
fcis-2970	210	2	-	-	PUNCT
fcis-2970	210	3	text	text	NOUN
fcis-2970	210	4	classification	classification	NOUN
fcis-2970	210	5	method	method	NOUN
fcis-2970	210	6	based	base	VERB
fcis-2970	210	7	on	on	ADP
fcis-2970	210	8	ernie	ernie	NOUN
fcis-2970	210	9	-	-	PUNCT
fcis-2970	210	10	bigru	bigru	NOUN
fcis-2970	210	11	.	.	PUNCT
fcis-2970	210	12	"	"	PUNCT
fcis-2970	211	1	journal	journal	PROPN
fcis-2970	211	2	of	of	ADP
fcis-2970	211	3	shanghai	shanghai	PROPN
fcis-2970	211	4	university	university	PROPN
fcis-2970	211	5	of	of	ADP
fcis-2970	211	6	electric	electric	ADJ
fcis-2970	211	7	power	power	NOUN
fcis-2970	211	8	36.04(2020):329	36.04(2020):329	PROPN
fcis-2970	211	9	-	-	PUNCT
fcis-2970	211	10	335	335	NUM
fcis-2970	211	11	+	+	NOUN
fcis-2970	211	12	350	350	NUM
fcis-2970	211	13	.	.	PUNCT
fcis-2970	212	1	[	[	X
fcis-2970	212	2	16	16	NUM
fcis-2970	212	3	]	]	X
fcis-2970	212	4	zemin	zemin	PROPN
fcis-2970	212	5	huang	huang	PROPN
fcis-2970	212	6	,	,	PUNCT
fcis-2970	212	7	xiaoling	xiaoling	PROPN
fcis-2970	212	8	wu	wu	PROPN
fcis-2970	212	9	,	,	PUNCT
fcis-2970	212	10	yinggang	yinggang	PROPN
fcis-2970	212	11	wu	wu	PROPN
fcis-2970	212	12	,	,	PUNCT
fcis-2970	212	13	jie	jie	PROPN
fcis-2970	212	14	ling	ling	PROPN
fcis-2970	212	15	.	.	PUNCT
fcis-2970	213	1	analysis	analysis	NOUN
fcis-2970	213	2	of	of	ADP
fcis-2970	213	3	chinese	chinese	ADJ
fcis-2970	213	4	text	text	NOUN
fcis-2970	213	5	emotions	emotion	NOUN
fcis-2970	213	6	combining	combine	VERB
fcis-2970	213	7	bert	bert	PROPN
fcis-2970	213	8	and	and	CCONJ
fcis-2970	213	9	bisru	bisru	NOUN
fcis-2970	213	10	-	-	PUNCT
fcis-2970	213	11	at[j	at[j	PROPN
fcis-2970	213	12	]	]	PUNCT
fcis-2970	213	13	.	.	PUNCT
fcis-2970	214	1	computer	computer	NOUN
fcis-2970	214	2	engineering	engineering	NOUN
fcis-2970	214	3	and	and	CCONJ
fcis-2970	214	4	science	science	NOUN
fcis-2970	214	5	,	,	PUNCT
fcis-2970	214	6	201,43(09):1668	201,43(09):1668	NUM
fcis-2970	214	7	-	-	SYM
fcis-2970	214	8	1675	1675	NUM
fcis-2970	214	9	.	.	PUNCT
fcis-2970	215	1	[	[	X
fcis-2970	215	2	17	17	NUM
fcis-2970	215	3	]	]	X
fcis-2970	215	4	chen	chen	PROPN
fcis-2970	215	5	jie	jie	PROPN
fcis-2970	215	6	,	,	PUNCT
fcis-2970	215	7	ma	ma	PROPN
fcis-2970	215	8	jing	jing	PROPN
fcis-2970	215	9	,	,	PUNCT
fcis-2970	215	10	li	li	PROPN
fcis-2970	215	11	xiaofeng	xiaofeng	PROPN
fcis-2970	215	12	.	.	PUNCT
fcis-2970	215	13	data	datum	NOUN
fcis-2970	215	14	analysis	analysis	NOUN
fcis-2970	215	15	and	and	CCONJ
fcis-2970	215	16	knowledge	knowledge	NOUN
fcis-2970	215	17	discovery	discovery	NOUN
fcis-2970	215	18	,	,	PUNCT
fcis-2970	215	19	201,5(09):21	201,5(09):21	NOUN
fcis-2970	215	20	-	-	SYM
fcis-2970	215	21	30	30	NUM
fcis-2970	215	22	.	.	PUNCT