id	sid	tid	token	lemma	pos
fcis-7512	1	1	frontiers	frontier	NOUN
fcis-7512	1	2	in	in	ADP
fcis-7512	1	3	computing	computing	NOUN
fcis-7512	1	4	and	and	CCONJ
fcis-7512	1	5	intelligent	intelligent	ADJ
fcis-7512	1	6	systems	system	NOUN
fcis-7512	1	7	issn	issn	VERB
fcis-7512	1	8	:	:	PUNCT
fcis-7512	1	9	2832	2832	NUM
fcis-7512	1	10	-	-	SYM
fcis-7512	1	11	6024	6024	NUM
fcis-7512	1	12	|	|	NOUN
fcis-7512	1	13	vol	vol	NOUN
fcis-7512	1	14	.	.	PROPN
fcis-7512	2	1	3	3	NUM
fcis-7512	2	2	,	,	PUNCT
fcis-7512	2	3	no	no	INTJ
fcis-7512	2	4	.	.	NOUN
fcis-7512	2	5	2	2	NUM
fcis-7512	2	6	,	,	PUNCT
fcis-7512	2	7	2023	2023	NUM
fcis-7512	2	8	67	67	NUM
fcis-7512	2	9	multimodal	multimodal	NOUN
fcis-7512	2	10	emotion	emotion	NOUN
fcis-7512	2	11	analysis	analysis	NOUN
fcis-7512	2	12	model	model	NOUN
fcis-7512	2	13	based	base	VERB
fcis-7512	2	14	on	on	ADP
fcis-7512	2	15	interactive	interactive	ADJ
fcis-7512	2	16	attention	attention	NOUN
fcis-7512	2	17	mechanism	mechanism	NOUN
fcis-7512	2	18	bainan	bainan	NOUN
fcis-7512	2	19	zhou	zhou	PROPN
fcis-7512	2	20	1	1	NUM
fcis-7512	2	21	,	,	PUNCT
fcis-7512	2	22	xu	xu	PROPN
fcis-7512	2	23	li	li	PROPN
fcis-7512	2	24	2	2	NUM
fcis-7512	2	25	1	1	NUM
fcis-7512	2	26	school	school	NOUN
fcis-7512	2	27	of	of	ADP
fcis-7512	2	28	information	information	NOUN
fcis-7512	2	29	science	science	NOUN
fcis-7512	2	30	and	and	CCONJ
fcis-7512	2	31	engineering	engineering	NOUN
fcis-7512	2	32	,	,	PUNCT
fcis-7512	2	33	dalian	dalian	PROPN
fcis-7512	2	34	polytechnic	polytechnic	PROPN
fcis-7512	2	35	university	university	PROPN
fcis-7512	2	36	,	,	PUNCT
fcis-7512	2	37	dalian	dalian	PROPN
fcis-7512	2	38	,	,	PUNCT
fcis-7512	2	39	china	china	PROPN
fcis-7512	2	40	2	2	NUM
fcis-7512	2	41	engineering	engineering	NOUN
fcis-7512	2	42	training	training	NOUN
fcis-7512	2	43	center	center	NOUN
fcis-7512	2	44	,	,	PUNCT
fcis-7512	2	45	dalian	dalian	PROPN
fcis-7512	2	46	polytechnic	polytechnic	PROPN
fcis-7512	2	47	university	university	PROPN
fcis-7512	2	48	,	,	PUNCT
fcis-7512	2	49	dalian	dalian	PROPN
fcis-7512	2	50	,	,	PUNCT
fcis-7512	2	51	china	china	PROPN
fcis-7512	2	52	abstract	abstract	NOUN
fcis-7512	2	53	:	:	PUNCT
fcis-7512	2	54	in	in	ADP
fcis-7512	2	55	traditional	traditional	ADJ
fcis-7512	2	56	multi	multi	ADJ
fcis-7512	2	57	-	-	ADJ
fcis-7512	2	58	modal	modal	ADJ
fcis-7512	2	59	sentiment	sentiment	NOUN
fcis-7512	2	60	analysis	analysis	NOUN
fcis-7512	2	61	,	,	PUNCT
fcis-7512	2	62	feature	feature	NOUN
fcis-7512	2	63	fusion	fusion	NOUN
fcis-7512	2	64	is	be	AUX
fcis-7512	2	65	usually	usually	ADV
fcis-7512	2	66	achieved	achieve	VERB
fcis-7512	2	67	by	by	ADP
fcis-7512	2	68	simple	simple	ADJ
fcis-7512	2	69	splicing	splicing	NOUN
fcis-7512	2	70	,	,	PUNCT
fcis-7512	2	71	and	and	CCONJ
fcis-7512	2	72	multi	multi	ADJ
fcis-7512	2	73	-	-	ADJ
fcis-7512	2	74	modal	modal	ADJ
fcis-7512	2	75	sentiment	sentiment	NOUN
fcis-7512	2	76	analysis	analysis	NOUN
fcis-7512	2	77	is	be	AUX
fcis-7512	2	78	only	only	ADV
fcis-7512	2	79	trained	train	VERB
fcis-7512	2	80	as	as	ADP
fcis-7512	2	81	a	a	DET
fcis-7512	2	82	single	single	ADJ
fcis-7512	2	83	task	task	NOUN
fcis-7512	2	84	,	,	PUNCT
fcis-7512	2	85	without	without	ADP
fcis-7512	2	86	considering	consider	VERB
fcis-7512	2	87	the	the	DET
fcis-7512	2	88	contribution	contribution	NOUN
fcis-7512	2	89	of	of	ADP
fcis-7512	2	90	inter	inter	ADJ
fcis-7512	2	91	-	-	ADJ
fcis-7512	2	92	modal	modal	ADJ
fcis-7512	2	93	information	information	NOUN
fcis-7512	2	94	interaction	interaction	NOUN
fcis-7512	2	95	to	to	PART
fcis-7512	2	96	sentiment	sentiment	VERB
fcis-7512	2	97	analysis	analysis	NOUN
fcis-7512	2	98	and	and	CCONJ
fcis-7512	2	99	the	the	DET
fcis-7512	2	100	correlation	correlation	NOUN
fcis-7512	2	101	and	and	CCONJ
fcis-7512	2	102	constraint	constraint	NOUN
fcis-7512	2	103	relationship	relationship	NOUN
fcis-7512	2	104	between	between	ADP
fcis-7512	2	105	multi	multi	ADJ
fcis-7512	2	106	-	-	ADJ
fcis-7512	2	107	modal	modal	ADJ
fcis-7512	2	108	and	and	CCONJ
fcis-7512	2	109	single	single	ADJ
fcis-7512	2	110	-	-	PUNCT
fcis-7512	2	111	modal	modal	NOUN
fcis-7512	2	112	(	(	PUNCT
fcis-7512	2	113	text	text	NOUN
fcis-7512	2	114	,	,	PUNCT
fcis-7512	2	115	video	video	NOUN
fcis-7512	2	116	and	and	CCONJ
fcis-7512	2	117	audio	audio	ADJ
fcis-7512	2	118	)	)	PUNCT
fcis-7512	2	119	tasks	task	NOUN
fcis-7512	2	120	.	.	PUNCT
fcis-7512	3	1	therefore	therefore	ADV
fcis-7512	3	2	,	,	PUNCT
fcis-7512	3	3	a	a	DET
fcis-7512	3	4	multi	multi	ADJ
fcis-7512	3	5	-	-	ADJ
fcis-7512	3	6	task	task	ADJ
fcis-7512	3	7	model	model	NOUN
fcis-7512	3	8	based	base	VERB
fcis-7512	3	9	on	on	ADP
fcis-7512	3	10	interactive	interactive	ADJ
fcis-7512	3	11	attention	attention	NOUN
fcis-7512	3	12	mechanism	mechanism	NOUN
fcis-7512	3	13	is	be	AUX
fcis-7512	3	14	proposed	propose	VERB
fcis-7512	3	15	in	in	ADP
fcis-7512	3	16	this	this	DET
fcis-7512	3	17	paper	paper	NOUN
fcis-7512	3	18	,	,	PUNCT
fcis-7512	3	19	which	which	PRON
fcis-7512	3	20	uses	use	VERB
fcis-7512	3	21	inter	inter	ADJ
fcis-7512	3	22	-	-	ADJ
fcis-7512	3	23	modal	modal	ADJ
fcis-7512	3	24	attention	attention	NOUN
fcis-7512	3	25	mechanism	mechanism	NOUN
fcis-7512	3	26	and	and	CCONJ
fcis-7512	3	27	single	single	ADJ
fcis-7512	3	28	-	-	PUNCT
fcis-7512	3	29	modal	modal	NOUN
fcis-7512	3	30	self	self	NOUN
fcis-7512	3	31	-	-	PUNCT
fcis-7512	3	32	attention	attention	NOUN
fcis-7512	3	33	mechanism	mechanism	NOUN
fcis-7512	3	34	to	to	PART
fcis-7512	3	35	train	train	VERB
fcis-7512	3	36	multi	multi	ADJ
fcis-7512	3	37	-	-	ADJ
fcis-7512	3	38	modal	modal	ADJ
fcis-7512	3	39	sentiment	sentiment	NOUN
fcis-7512	3	40	analysis	analysis	NOUN
fcis-7512	3	41	and	and	CCONJ
fcis-7512	3	42	single	single	ADJ
fcis-7512	3	43	-	-	PUNCT
fcis-7512	3	44	modal	modal	ADJ
fcis-7512	3	45	sentiment	sentiment	NOUN
fcis-7512	3	46	analysis	analysis	NOUN
fcis-7512	3	47	together	together	ADV
fcis-7512	3	48	,	,	PUNCT
fcis-7512	3	49	so	so	SCONJ
fcis-7512	3	50	as	as	SCONJ
fcis-7512	3	51	to	to	PART
fcis-7512	3	52	make	make	VERB
fcis-7512	3	53	full	full	ADJ
fcis-7512	3	54	use	use	NOUN
fcis-7512	3	55	of	of	ADP
fcis-7512	3	56	inter	inter	ADJ
fcis-7512	3	57	-	-	ADJ
fcis-7512	3	58	modal	modal	ADJ
fcis-7512	3	59	and	and	CCONJ
fcis-7512	3	60	inter	inter	ADJ
fcis-7512	3	61	-	-	ADJ
fcis-7512	3	62	task	task	ADJ
fcis-7512	3	63	information	information	NOUN
fcis-7512	3	64	sharing	sharing	NOUN
fcis-7512	3	65	,	,	PUNCT
fcis-7512	3	66	mutual	mutual	ADJ
fcis-7512	3	67	complement	complement	NOUN
fcis-7512	3	68	,	,	PUNCT
fcis-7512	3	69	and	and	CCONJ
fcis-7512	3	70	reduce	reduce	VERB
fcis-7512	3	71	noise	noise	NOUN
fcis-7512	3	72	to	to	PART
fcis-7512	3	73	improve	improve	VERB
fcis-7512	3	74	the	the	DET
fcis-7512	3	75	overall	overall	ADJ
fcis-7512	3	76	recognition	recognition	NOUN
fcis-7512	3	77	performance	performance	NOUN
fcis-7512	3	78	.	.	PUNCT
fcis-7512	4	1	experiments	experiment	NOUN
fcis-7512	4	2	show	show	VERB
fcis-7512	4	3	that	that	SCONJ
fcis-7512	4	4	the	the	DET
fcis-7512	4	5	proposed	propose	VERB
fcis-7512	4	6	model	model	NOUN
fcis-7512	4	7	performs	perform	VERB
fcis-7512	4	8	well	well	ADV
fcis-7512	4	9	on	on	ADP
fcis-7512	4	10	mosi	mosi	NOUN
fcis-7512	4	11	and	and	CCONJ
fcis-7512	4	12	mosei	mosei	ADJ
fcis-7512	4	13	common	common	ADJ
fcis-7512	4	14	data	data	NOUN
fcis-7512	4	15	sets	set	NOUN
fcis-7512	4	16	for	for	ADP
fcis-7512	4	17	multimodal	multimodal	ADJ
fcis-7512	4	18	sentiment	sentiment	NOUN
fcis-7512	4	19	analysis	analysis	NOUN
fcis-7512	4	20	.	.	PUNCT
fcis-7512	5	1	keywords	keyword	NOUN
fcis-7512	5	2	:	:	PUNCT
fcis-7512	5	3	multimodal	multimodal	ADJ
fcis-7512	5	4	sentiment	sentiment	NOUN
fcis-7512	5	5	analysis	analysis	NOUN
fcis-7512	5	6	;	;	PUNCT
fcis-7512	5	7	attention	attention	NOUN
fcis-7512	5	8	mechanism	mechanism	NOUN
fcis-7512	5	9	;	;	PUNCT
fcis-7512	5	10	multi	multi	ADJ
fcis-7512	5	11	-	-	ADJ
fcis-7512	5	12	mission	mission	ADJ
fcis-7512	5	13	joint	joint	ADJ
fcis-7512	5	14	training	training	NOUN
fcis-7512	5	15	.	.	PUNCT
fcis-7512	6	1	1	1	X
fcis-7512	6	2	.	.	X
fcis-7512	6	3	introduction	introduction	NOUN
fcis-7512	6	4	traditional	traditional	ADJ
fcis-7512	6	5	sentiment	sentiment	NOUN
fcis-7512	6	6	analysis	analysis	NOUN
fcis-7512	6	7	usually	usually	ADV
fcis-7512	6	8	refers	refer	VERB
fcis-7512	6	9	to	to	ADP
fcis-7512	6	10	the	the	DET
fcis-7512	6	11	text	text	NOUN
fcis-7512	6	12	sentiment	sentiment	NOUN
fcis-7512	6	13	analysis	analysis	NOUN
fcis-7512	6	14	task	task	NOUN
fcis-7512	6	15	[	[	X
fcis-7512	6	16	1	1	NUM
fcis-7512	6	17	]	]	PUNCT
fcis-7512	6	18	,	,	PUNCT
fcis-7512	6	19	but	but	CCONJ
fcis-7512	6	20	with	with	ADP
fcis-7512	6	21	the	the	DET
fcis-7512	6	22	development	development	NOUN
fcis-7512	6	23	of	of	ADP
fcis-7512	6	24	social	social	ADJ
fcis-7512	6	25	media	medium	NOUN
fcis-7512	6	26	,	,	PUNCT
fcis-7512	6	27	the	the	DET
fcis-7512	6	28	types	type	NOUN
fcis-7512	6	29	and	and	CCONJ
fcis-7512	6	30	forms	form	NOUN
fcis-7512	6	31	of	of	ADP
fcis-7512	6	32	data	datum	NOUN
fcis-7512	6	33	are	be	AUX
fcis-7512	6	34	diversified	diversify	VERB
fcis-7512	6	35	.	.	PUNCT
fcis-7512	7	1	in	in	ADP
fcis-7512	7	2	recent	recent	ADJ
fcis-7512	7	3	years	year	NOUN
fcis-7512	7	4	,	,	PUNCT
fcis-7512	7	5	multimodal	multimodal	ADJ
fcis-7512	7	6	sentiment	sentiment	NOUN
fcis-7512	7	7	analysis	analysis	NOUN
fcis-7512	7	8	has	have	AUX
fcis-7512	7	9	become	become	VERB
fcis-7512	7	10	a	a	DET
fcis-7512	7	11	hot	hot	ADJ
fcis-7512	7	12	research	research	NOUN
fcis-7512	7	13	topic	topic	NOUN
fcis-7512	7	14	due	due	ADP
fcis-7512	7	15	to	to	ADP
fcis-7512	7	16	the	the	DET
fcis-7512	7	17	limitations	limitation	NOUN
fcis-7512	7	18	of	of	ADP
fcis-7512	7	19	traditional	traditional	ADJ
fcis-7512	7	20	text	text	NOUN
fcis-7512	7	21	sentiment	sentiment	NOUN
fcis-7512	7	22	analysis	analysis	NOUN
fcis-7512	7	23	only	only	ADV
fcis-7512	7	24	using	use	VERB
fcis-7512	7	25	text	text	NOUN
fcis-7512	7	26	data	datum	NOUN
fcis-7512	7	27	to	to	PART
fcis-7512	7	28	analyze	analyze	VERB
fcis-7512	7	29	emotional	emotional	ADJ
fcis-7512	7	30	information	information	NOUN
fcis-7512	7	31	.	.	PUNCT
fcis-7512	8	1	typically	typically	ADV
fcis-7512	8	2	,	,	PUNCT
fcis-7512	8	3	multimodal	multimodal	ADJ
fcis-7512	8	4	data	datum	NOUN
fcis-7512	8	5	consists	consist	VERB
fcis-7512	8	6	of	of	ADP
fcis-7512	8	7	three	three	NUM
fcis-7512	8	8	parts	part	NOUN
fcis-7512	8	9	:	:	PUNCT
fcis-7512	8	10	text	text	NOUN
fcis-7512	8	11	,	,	PUNCT
fcis-7512	8	12	video	video	NOUN
fcis-7512	8	13	,	,	PUNCT
fcis-7512	8	14	and	and	CCONJ
fcis-7512	8	15	audio	audio	NOUN
fcis-7512	8	16	.	.	PUNCT
fcis-7512	9	1	this	this	DET
fcis-7512	9	2	paper	paper	NOUN
fcis-7512	9	3	mainly	mainly	ADV
fcis-7512	9	4	studies	study	VERB
fcis-7512	9	5	the	the	DET
fcis-7512	9	6	multimodal	multimodal	ADJ
fcis-7512	9	7	sentiment	sentiment	NOUN
fcis-7512	9	8	analysis	analysis	NOUN
fcis-7512	9	9	task	task	NOUN
fcis-7512	9	10	of	of	ADP
fcis-7512	9	11	three	three	NUM
fcis-7512	9	12	modes	mode	NOUN
fcis-7512	9	13	.	.	PUNCT
fcis-7512	10	1	multimodal	multimodal	NOUN
fcis-7512	10	2	data	datum	NOUN
fcis-7512	10	3	representation	representation	NOUN
fcis-7512	10	4	and	and	CCONJ
fcis-7512	10	5	multimodal	multimodal	NOUN
fcis-7512	10	6	feature	feature	NOUN
fcis-7512	10	7	fusion	fusion	NOUN
fcis-7512	10	8	are	be	AUX
fcis-7512	10	9	two	two	NUM
fcis-7512	10	10	important	important	ADJ
fcis-7512	10	11	types	type	NOUN
fcis-7512	10	12	of	of	ADP
fcis-7512	10	13	work	work	NOUN
fcis-7512	10	14	in	in	ADP
fcis-7512	10	15	multimodal	multimodal	ADJ
fcis-7512	10	16	sentiment	sentiment	NOUN
fcis-7512	10	17	analysis	analysis	NOUN
fcis-7512	10	18	[	[	X
fcis-7512	10	19	2	2	NUM
fcis-7512	10	20	]	]	PUNCT
fcis-7512	10	21	.	.	PUNCT
fcis-7512	11	1	in	in	ADP
fcis-7512	11	2	recent	recent	ADJ
fcis-7512	11	3	years	year	NOUN
fcis-7512	11	4	,	,	PUNCT
fcis-7512	11	5	many	many	ADJ
fcis-7512	11	6	research	research	NOUN
fcis-7512	11	7	methods	method	NOUN
fcis-7512	11	8	have	have	AUX
fcis-7512	11	9	been	be	AUX
fcis-7512	11	10	proposed	propose	VERB
fcis-7512	11	11	for	for	ADP
fcis-7512	11	12	these	these	DET
fcis-7512	11	13	two	two	NUM
fcis-7512	11	14	types	type	NOUN
fcis-7512	11	15	of	of	ADP
fcis-7512	11	16	important	important	ADJ
fcis-7512	11	17	work	work	NOUN
fcis-7512	11	18	.	.	PUNCT
fcis-7512	12	1	for	for	ADP
fcis-7512	12	2	multi	multi	ADJ
fcis-7512	12	3	-	-	ADJ
fcis-7512	12	4	modal	modal	ADJ
fcis-7512	12	5	data	data	NOUN
fcis-7512	12	6	representation	representation	NOUN
fcis-7512	12	7	,	,	PUNCT
fcis-7512	12	8	neural	neural	ADJ
fcis-7512	12	9	networks	network	NOUN
fcis-7512	12	10	are	be	AUX
fcis-7512	12	11	usually	usually	ADV
fcis-7512	12	12	used	use	VERB
fcis-7512	12	13	to	to	PART
fcis-7512	12	14	represent	represent	VERB
fcis-7512	12	15	data	datum	NOUN
fcis-7512	12	16	of	of	ADP
fcis-7512	12	17	different	different	ADJ
fcis-7512	12	18	modes	mode	NOUN
fcis-7512	12	19	respectively	respectively	ADV
fcis-7512	12	20	.	.	PUNCT
fcis-7512	13	1	poria	poria	PROPN
fcis-7512	13	2	et	et	PROPN
fcis-7512	13	3	al	al	PROPN
fcis-7512	13	4	.	.	PUNCT
fcis-7512	14	1	[	[	X
fcis-7512	14	2	3	3	X
fcis-7512	14	3	]	]	PUNCT
fcis-7512	14	4	proposed	propose	VERB
fcis-7512	14	5	a	a	DET
fcis-7512	14	6	model	model	NOUN
fcis-7512	14	7	based	base	VERB
fcis-7512	14	8	on	on	ADP
fcis-7512	14	9	multicore	multicore	ADJ
fcis-7512	14	10	learning	learning	NOUN
fcis-7512	14	11	,	,	PUNCT
fcis-7512	14	12	which	which	PRON
fcis-7512	14	13	uses	use	VERB
fcis-7512	14	14	deep	deep	ADJ
fcis-7512	14	15	convolutional	convolutional	ADJ
fcis-7512	14	16	neural	neural	ADJ
fcis-7512	14	17	networks	network	NOUN
fcis-7512	14	18	to	to	PART
fcis-7512	14	19	extract	extract	VERB
fcis-7512	14	20	text	text	NOUN
fcis-7512	14	21	features	feature	NOUN
fcis-7512	14	22	and	and	CCONJ
fcis-7512	14	23	integrate	integrate	VERB
fcis-7512	14	24	them	they	PRON
fcis-7512	14	25	with	with	ADP
fcis-7512	14	26	other	other	ADJ
fcis-7512	14	27	(	(	PUNCT
fcis-7512	14	28	video	video	NOUN
fcis-7512	14	29	and	and	CCONJ
fcis-7512	14	30	audio	audio	NOUN
fcis-7512	14	31	)	)	PUNCT
fcis-7512	14	32	modes	mode	NOUN
fcis-7512	14	33	in	in	ADP
fcis-7512	14	34	the	the	DET
fcis-7512	14	35	form	form	NOUN
fcis-7512	14	36	of	of	ADP
fcis-7512	14	37	splicing	splicing	NOUN
fcis-7512	14	38	.	.	PUNCT
fcis-7512	15	1	cambria	cambria	PROPN
fcis-7512	15	2	et	et	PROPN
fcis-7512	15	3	al	al	PROPN
fcis-7512	15	4	.	.	PUNCT
fcis-7512	16	1	[	[	X
fcis-7512	16	2	4	4	X
fcis-7512	16	3	]	]	PUNCT
fcis-7512	16	4	proposed	propose	VERB
fcis-7512	16	5	a	a	DET
fcis-7512	16	6	general	general	ADJ
fcis-7512	16	7	multi	multi	ADJ
fcis-7512	16	8	-	-	ADJ
fcis-7512	16	9	modal	modal	ADJ
fcis-7512	16	10	sentiment	sentiment	NOUN
fcis-7512	16	11	analysis	analysis	NOUN
fcis-7512	16	12	framework	framework	NOUN
fcis-7512	16	13	,	,	PUNCT
fcis-7512	16	14	which	which	PRON
fcis-7512	16	15	consists	consist	VERB
fcis-7512	16	16	of	of	ADP
fcis-7512	16	17	feature	feature	NOUN
fcis-7512	16	18	learning	learn	VERB
fcis-7512	16	19	within	within	ADP
fcis-7512	16	20	modes	mode	NOUN
fcis-7512	16	21	and	and	CCONJ
fcis-7512	16	22	feature	feature	VERB
fcis-7512	16	23	splicing	splicing	NOUN
fcis-7512	16	24	of	of	ADP
fcis-7512	16	25	multi	multi	NOUN
fcis-7512	16	26	-	-	NOUN
fcis-7512	16	27	modes	mode	NOUN
fcis-7512	16	28	,	,	PUNCT
fcis-7512	16	29	and	and	CCONJ
fcis-7512	16	30	laid	lay	VERB
fcis-7512	16	31	a	a	DET
fcis-7512	16	32	foundation	foundation	NOUN
fcis-7512	16	33	for	for	ADP
fcis-7512	16	34	future	future	ADJ
fcis-7512	16	35	research	research	NOUN
fcis-7512	16	36	.	.	PUNCT
fcis-7512	17	1	zadeh	zadeh	PROPN
fcis-7512	17	2	et	et	PROPN
fcis-7512	17	3	al	al	PROPN
fcis-7512	17	4	.	.	PUNCT
fcis-7512	18	1	[	[	X
fcis-7512	18	2	5	5	NUM
fcis-7512	18	3	]	]	PUNCT
fcis-7512	18	4	introduced	introduce	VERB
fcis-7512	18	5	a	a	DET
fcis-7512	18	6	multimodal	multimodal	ADJ
fcis-7512	18	7	dictionary	dictionary	NOUN
fcis-7512	18	8	in	in	ADP
fcis-7512	18	9	order	order	NOUN
fcis-7512	18	10	to	to	PART
fcis-7512	18	11	better	well	ADV
fcis-7512	18	12	understand	understand	VERB
fcis-7512	18	13	the	the	DET
fcis-7512	18	14	interaction	interaction	NOUN
fcis-7512	18	15	between	between	ADP
fcis-7512	18	16	face	face	NOUN
fcis-7512	18	17	,	,	PUNCT
fcis-7512	18	18	gesture	gesture	NOUN
fcis-7512	18	19	and	and	CCONJ
fcis-7512	18	20	audio	audio	NOUN
fcis-7512	18	21	in	in	ADP
fcis-7512	18	22	expressing	express	VERB
fcis-7512	18	23	emotion	emotion	NOUN
fcis-7512	18	24	,	,	PUNCT
fcis-7512	18	25	and	and	CCONJ
fcis-7512	18	26	proposed	propose	VERB
fcis-7512	18	27	a	a	DET
fcis-7512	18	28	mosi	mosi	NOUN
fcis-7512	18	29	multimodal	multimodal	ADJ
fcis-7512	18	30	emotion	emotion	NOUN
fcis-7512	18	31	analysis	analysis	NOUN
fcis-7512	18	32	dataset	dataset	VERB
fcis-7512	18	33	.	.	PUNCT
fcis-7512	19	1	later	later	ADV
fcis-7512	19	2	,	,	PUNCT
fcis-7512	19	3	zadeh	zadeh	PROPN
fcis-7512	19	4	et	et	PROPN
fcis-7512	19	5	al	al	PROPN
fcis-7512	19	6	.	.	PUNCT
fcis-7512	20	1	[	[	X
fcis-7512	20	2	6	6	NUM
fcis-7512	20	3	]	]	PUNCT
fcis-7512	20	4	proposed	propose	VERB
fcis-7512	20	5	tensor	tensor	NOUN
fcis-7512	20	6	fusion	fusion	NOUN
fcis-7512	20	7	network	network	NOUN
fcis-7512	20	8	to	to	PART
fcis-7512	20	9	learn	learn	VERB
fcis-7512	20	10	features	feature	NOUN
fcis-7512	20	11	within	within	ADP
fcis-7512	20	12	and	and	CCONJ
fcis-7512	20	13	between	between	ADP
fcis-7512	20	14	modes	mode	NOUN
fcis-7512	20	15	.	.	PUNCT
fcis-7512	21	1	and	and	CCONJ
fcis-7512	21	2	the	the	DET
fcis-7512	21	3	precision	precision	NOUN
fcis-7512	21	4	of	of	ADP
fcis-7512	21	5	multimodal	multimodal	ADJ
fcis-7512	21	6	sentiment	sentiment	NOUN
fcis-7512	21	7	analysis	analysis	NOUN
fcis-7512	21	8	is	be	AUX
fcis-7512	21	9	improved	improve	VERB
fcis-7512	21	10	on	on	ADP
fcis-7512	21	11	mosi	mosi	PROPN
fcis-7512	21	12	data	datum	NOUN
fcis-7512	21	13	set	set	VERB
fcis-7512	21	14	.	.	PUNCT
fcis-7512	22	1	in	in	ADP
fcis-7512	22	2	order	order	NOUN
fcis-7512	22	3	to	to	PART
fcis-7512	22	4	consider	consider	VERB
fcis-7512	22	5	more	more	ADJ
fcis-7512	22	6	context	context	NOUN
fcis-7512	22	7	information	information	NOUN
fcis-7512	22	8	,	,	PUNCT
fcis-7512	22	9	poria	poria	PROPN
fcis-7512	22	10	et	et	PROPN
fcis-7512	22	11	al	al	PROPN
fcis-7512	22	12	.	.	PUNCT
fcis-7512	23	1	[	[	X
fcis-7512	23	2	7	7	NUM
fcis-7512	23	3	]	]	PUNCT
fcis-7512	23	4	developed	develop	VERB
fcis-7512	23	5	a	a	DET
fcis-7512	23	6	framework	framework	NOUN
fcis-7512	23	7	based	base	VERB
fcis-7512	23	8	on	on	ADP
fcis-7512	23	9	lstm	lstm	PROPN
fcis-7512	23	10	,	,	PUNCT
fcis-7512	23	11	which	which	PRON
fcis-7512	23	12	uses	use	VERB
fcis-7512	23	13	context	context	NOUN
fcis-7512	23	14	information	information	NOUN
fcis-7512	23	15	to	to	PART
fcis-7512	23	16	capture	capture	VERB
fcis-7512	23	17	the	the	DET
fcis-7512	23	18	dependency	dependency	NOUN
fcis-7512	23	19	between	between	ADP
fcis-7512	23	20	modes	mode	NOUN
fcis-7512	23	21	.	.	PUNCT
fcis-7512	24	1	in	in	ADP
fcis-7512	24	2	the	the	DET
fcis-7512	24	3	later	later	ADJ
fcis-7512	24	4	work	work	NOUN
fcis-7512	24	5	of	of	ADP
fcis-7512	24	6	zadeh	zadeh	PROPN
fcis-7512	24	7	et	et	PROPN
fcis-7512	24	8	al	al	PROPN
fcis-7512	24	9	.	.	PUNCT
fcis-7512	25	1	[	[	X
fcis-7512	25	2	8	8	NUM
fcis-7512	25	3	]	]	PUNCT
fcis-7512	25	4	,	,	PUNCT
fcis-7512	25	5	multiple	multiple	ADJ
fcis-7512	25	6	attention	attention	NOUN
fcis-7512	25	7	blocks	block	NOUN
fcis-7512	25	8	were	be	AUX
fcis-7512	25	9	used	use	VERB
fcis-7512	25	10	to	to	PART
fcis-7512	25	11	obtain	obtain	VERB
fcis-7512	25	12	the	the	DET
fcis-7512	25	13	information	information	NOUN
fcis-7512	25	14	between	between	ADP
fcis-7512	25	15	the	the	DET
fcis-7512	25	16	three	three	NUM
fcis-7512	25	17	modes	mode	NOUN
fcis-7512	25	18	and	and	CCONJ
fcis-7512	25	19	have	have	VERB
fcis-7512	25	20	good	good	ADJ
fcis-7512	25	21	performance	performance	NOUN
fcis-7512	25	22	on	on	ADP
fcis-7512	25	23	different	different	ADJ
fcis-7512	25	24	data	datum	NOUN
fcis-7512	25	25	sets	set	NOUN
fcis-7512	25	26	.	.	PUNCT
fcis-7512	26	1	after	after	ADP
fcis-7512	26	2	the	the	DET
fcis-7512	26	3	emergence	emergence	NOUN
fcis-7512	26	4	of	of	ADP
fcis-7512	26	5	transformer	transformer	NOUN
fcis-7512	26	6	[	[	X
fcis-7512	26	7	9	9	NUM
fcis-7512	26	8	]	]	PUNCT
fcis-7512	26	9	,	,	PUNCT
fcis-7512	26	10	tsai	tsai	PROPN
fcis-7512	26	11	et	et	PROPN
fcis-7512	26	12	al	al	PROPN
fcis-7512	26	13	.	.	PUNCT
fcis-7512	27	1	[	[	X
fcis-7512	27	2	10	10	NUM
fcis-7512	27	3	]	]	PUNCT
fcis-7512	27	4	proposed	propose	VERB
fcis-7512	27	5	a	a	DET
fcis-7512	27	6	cross	cross	ADJ
fcis-7512	27	7	-	-	ADJ
fcis-7512	27	8	modal	modal	ADJ
fcis-7512	27	9	transformer	transformer	NOUN
fcis-7512	27	10	model	model	NOUN
fcis-7512	27	11	to	to	PART
fcis-7512	27	12	strengthen	strengthen	VERB
fcis-7512	27	13	the	the	DET
fcis-7512	27	14	feature	feature	NOUN
fcis-7512	27	15	learning	learning	NOUN
fcis-7512	27	16	of	of	ADP
fcis-7512	27	17	target	target	NOUN
fcis-7512	27	18	modes	mode	NOUN
fcis-7512	27	19	through	through	ADP
fcis-7512	27	20	the	the	DET
fcis-7512	27	21	cross	cross	ADJ
fcis-7512	27	22	-	-	ADJ
fcis-7512	27	23	modal	modal	ADJ
fcis-7512	27	24	attention	attention	NOUN
fcis-7512	27	25	mechanism	mechanism	NOUN
fcis-7512	27	26	.	.	PUNCT
fcis-7512	28	1	however	however	ADV
fcis-7512	28	2	,	,	PUNCT
fcis-7512	28	3	the	the	DET
fcis-7512	28	4	above	above	ADJ
fcis-7512	28	5	research	research	NOUN
fcis-7512	28	6	on	on	ADP
fcis-7512	28	7	multimodal	multimodal	NOUN
fcis-7512	28	8	data	datum	NOUN
fcis-7512	28	9	representation	representation	NOUN
fcis-7512	28	10	does	do	AUX
fcis-7512	28	11	not	not	PART
fcis-7512	28	12	consider	consider	VERB
fcis-7512	28	13	the	the	DET
fcis-7512	28	14	correlation	correlation	NOUN
fcis-7512	28	15	between	between	ADP
fcis-7512	28	16	different	different	ADJ
fcis-7512	28	17	modal	modal	ADJ
fcis-7512	28	18	data	datum	NOUN
fcis-7512	28	19	in	in	ADP
fcis-7512	28	20	multimodal	multimodal	NOUN
fcis-7512	28	21	data	datum	NOUN
fcis-7512	28	22	representation	representation	NOUN
fcis-7512	28	23	,	,	PUNCT
fcis-7512	28	24	and	and	CCONJ
fcis-7512	28	25	only	only	ADV
fcis-7512	28	26	focuses	focus	VERB
fcis-7512	28	27	on	on	ADP
fcis-7512	28	28	the	the	DET
fcis-7512	28	29	representation	representation	NOUN
fcis-7512	28	30	of	of	ADP
fcis-7512	28	31	different	different	ADJ
fcis-7512	28	32	modal	modal	ADJ
fcis-7512	28	33	data	datum	NOUN
fcis-7512	28	34	respectively	respectively	ADV
fcis-7512	28	35	.	.	PUNCT
fcis-7512	29	1	in	in	ADP
fcis-7512	29	2	the	the	DET
fcis-7512	29	3	study	study	NOUN
fcis-7512	29	4	of	of	ADP
fcis-7512	29	5	multi	multi	ADJ
fcis-7512	29	6	-	-	ADJ
fcis-7512	29	7	modal	modal	ADJ
fcis-7512	29	8	sentiment	sentiment	NOUN
fcis-7512	29	9	analysis	analysis	NOUN
fcis-7512	29	10	,	,	PUNCT
fcis-7512	29	11	the	the	DET
fcis-7512	29	12	multimodal	multimodal	ADJ
fcis-7512	29	13	feature	feature	NOUN
fcis-7512	29	14	fusion	fusion	NOUN
fcis-7512	29	15	usually	usually	ADV
fcis-7512	29	16	adopts	adopt	VERB
fcis-7512	29	17	the	the	DET
fcis-7512	29	18	fusion	fusion	NOUN
fcis-7512	29	19	way	way	NOUN
fcis-7512	29	20	of	of	ADP
fcis-7512	29	21	early	early	ADJ
fcis-7512	29	22	fusion	fusion	NOUN
fcis-7512	29	23	(	(	PUNCT
fcis-7512	29	24	fusion	fusion	NOUN
fcis-7512	29	25	after	after	ADP
fcis-7512	29	26	feature	feature	NOUN
fcis-7512	29	27	learning	learning	NOUN
fcis-7512	29	28	)	)	PUNCT
fcis-7512	29	29	.	.	PUNCT
fcis-7512	30	1	poria	poria	PROPN
fcis-7512	30	2	et	et	PROPN
fcis-7512	30	3	al	al	PROPN
fcis-7512	30	4	.	.	PUNCT
fcis-7512	31	1	[	[	X
fcis-7512	31	2	3	3	X
fcis-7512	31	3	]	]	PUNCT
fcis-7512	31	4	and	and	CCONJ
fcis-7512	31	5	cambria	cambria	PROPN
fcis-7512	31	6	et	et	PROPN
fcis-7512	31	7	al	al	PROPN
fcis-7512	31	8	.	.	PUNCT
fcis-7512	32	1	[	[	X
fcis-7512	32	2	4	4	X
fcis-7512	32	3	]	]	PUNCT
fcis-7512	32	4	represent	represent	VERB
fcis-7512	32	5	the	the	DET
fcis-7512	32	6	multi	multi	ADJ
fcis-7512	32	7	-	-	ADJ
fcis-7512	32	8	modal	modal	ADJ
fcis-7512	32	9	data	datum	NOUN
fcis-7512	32	10	by	by	ADP
fcis-7512	32	11	neural	neural	ADJ
fcis-7512	32	12	network	network	NOUN
fcis-7512	32	13	,	,	PUNCT
fcis-7512	32	14	and	and	CCONJ
fcis-7512	32	15	then	then	ADV
fcis-7512	32	16	use	use	VERB
fcis-7512	32	17	a	a	DET
fcis-7512	32	18	simple	simple	ADJ
fcis-7512	32	19	splicing	splicing	NOUN
fcis-7512	32	20	way	way	NOUN
fcis-7512	32	21	to	to	PART
fcis-7512	32	22	carry	carry	VERB
fcis-7512	32	23	out	out	ADP
fcis-7512	32	24	feature	feature	NOUN
fcis-7512	32	25	fusion	fusion	NOUN
fcis-7512	32	26	.	.	PUNCT
fcis-7512	33	1	later	later	ADV
fcis-7512	33	2	,	,	PUNCT
fcis-7512	33	3	liu	liu	PROPN
fcis-7512	33	4	et	et	PROPN
fcis-7512	33	5	al	al	PROPN
fcis-7512	33	6	.	.	PUNCT
fcis-7512	34	1	[	[	X
fcis-7512	34	2	19	19	NUM
fcis-7512	34	3	]	]	PUNCT
fcis-7512	34	4	proposed	propose	VERB
fcis-7512	34	5	a	a	DET
fcis-7512	34	6	low	low	ADJ
fcis-7512	34	7	-	-	PUNCT
fcis-7512	34	8	rank	rank	NOUN
fcis-7512	34	9	multi	multi	ADJ
fcis-7512	34	10	-	-	ADJ
fcis-7512	34	11	mode	mode	ADJ
fcis-7512	34	12	fusion	fusion	NOUN
fcis-7512	34	13	method	method	NOUN
fcis-7512	34	14	,	,	PUNCT
fcis-7512	34	15	using	use	VERB
fcis-7512	34	16	low	low	ADJ
fcis-7512	34	17	-	-	PUNCT
fcis-7512	34	18	rank	rank	NOUN
fcis-7512	34	19	vector	vector	NOUN
fcis-7512	34	20	to	to	PART
fcis-7512	34	21	represent	represent	VERB
fcis-7512	34	22	multi	multi	ADJ
fcis-7512	34	23	-	-	ADJ
fcis-7512	34	24	mode	mode	ADJ
fcis-7512	34	25	data	datum	NOUN
fcis-7512	34	26	,	,	PUNCT
fcis-7512	34	27	and	and	CCONJ
fcis-7512	34	28	then	then	ADV
fcis-7512	34	29	using	use	VERB
fcis-7512	34	30	cartesian	cartesian	ADJ
fcis-7512	34	31	product	product	NOUN
fcis-7512	34	32	to	to	PART
fcis-7512	34	33	achieve	achieve	VERB
fcis-7512	34	34	feature	feature	NOUN
fcis-7512	34	35	fusion	fusion	NOUN
fcis-7512	34	36	.	.	PUNCT
fcis-7512	35	1	however	however	ADV
fcis-7512	35	2	,	,	PUNCT
fcis-7512	35	3	the	the	DET
fcis-7512	35	4	simple	simple	ADJ
fcis-7512	35	5	fusion	fusion	NOUN
fcis-7512	35	6	method	method	NOUN
fcis-7512	35	7	does	do	AUX
fcis-7512	35	8	not	not	PART
fcis-7512	35	9	consider	consider	VERB
fcis-7512	35	10	the	the	DET
fcis-7512	35	11	contribution	contribution	NOUN
fcis-7512	35	12	degree	degree	NOUN
fcis-7512	35	13	of	of	ADP
fcis-7512	35	14	different	different	ADJ
fcis-7512	35	15	modal	modal	ADJ
fcis-7512	35	16	data	datum	NOUN
fcis-7512	35	17	to	to	ADP
fcis-7512	35	18	the	the	DET
fcis-7512	35	19	results	result	NOUN
fcis-7512	35	20	of	of	ADP
fcis-7512	35	21	sentiment	sentiment	NOUN
fcis-7512	35	22	analysis	analysis	NOUN
fcis-7512	35	23	,	,	PUNCT
fcis-7512	35	24	and	and	CCONJ
fcis-7512	35	25	the	the	DET
fcis-7512	35	26	common	common	ADJ
fcis-7512	35	27	multi	multi	ADJ
fcis-7512	35	28	-	-	ADJ
fcis-7512	35	29	modal	modal	ADJ
fcis-7512	35	30	sentiment	sentiment	NOUN
fcis-7512	35	31	analysis	analysis	NOUN
fcis-7512	35	32	task	task	NOUN
fcis-7512	35	33	is	be	AUX
fcis-7512	35	34	usually	usually	ADV
fcis-7512	35	35	a	a	DET
fcis-7512	35	36	single	single	ADJ
fcis-7512	35	37	-	-	PUNCT
fcis-7512	35	38	task	task	NOUN
fcis-7512	35	39	training	training	NOUN
fcis-7512	35	40	method	method	NOUN
fcis-7512	35	41	,	,	PUNCT
fcis-7512	35	42	and	and	CCONJ
fcis-7512	35	43	does	do	AUX
fcis-7512	35	44	not	not	PART
fcis-7512	35	45	consider	consider	VERB
fcis-7512	35	46	the	the	DET
fcis-7512	35	47	correlation	correlation	NOUN
fcis-7512	35	48	and	and	CCONJ
fcis-7512	35	49	constraint	constraint	NOUN
fcis-7512	35	50	relationship	relationship	NOUN
fcis-7512	35	51	between	between	ADP
fcis-7512	35	52	multi	multi	ADJ
fcis-7512	35	53	-	-	ADJ
fcis-7512	35	54	modal	modal	ADJ
fcis-7512	35	55	sentiment	sentiment	NOUN
fcis-7512	35	56	analysis	analysis	NOUN
fcis-7512	35	57	and	and	CCONJ
fcis-7512	35	58	each	each	DET
fcis-7512	35	59	modal	modal	ADJ
fcis-7512	35	60	sentiment	sentiment	NOUN
fcis-7512	35	61	analysis	analysis	NOUN
fcis-7512	35	62	task	task	NOUN
fcis-7512	35	63	.	.	PUNCT
fcis-7512	36	1	to	to	PART
fcis-7512	36	2	solve	solve	VERB
fcis-7512	36	3	the	the	DET
fcis-7512	36	4	above	above	ADJ
fcis-7512	36	5	problems	problem	NOUN
fcis-7512	36	6	,	,	PUNCT
fcis-7512	36	7	this	this	DET
fcis-7512	36	8	paper	paper	NOUN
fcis-7512	36	9	proposes	propose	VERB
fcis-7512	36	10	a	a	DET
fcis-7512	36	11	multitask	multitask	ADJ
fcis-7512	36	12	emotion	emotion	NOUN
fcis-7512	36	13	analysis	analysis	NOUN
fcis-7512	36	14	model	model	NOUN
fcis-7512	36	15	based	base	VERB
fcis-7512	36	16	on	on	ADP
fcis-7512	36	17	attention	attention	NOUN
fcis-7512	36	18	mechanism	mechanism	NOUN
fcis-7512	36	19	[	[	X
fcis-7512	36	20	12	12	NUM
fcis-7512	36	21	]	]	PUNCT
fcis-7512	36	22	,	,	PUNCT
fcis-7512	36	23	which	which	PRON
fcis-7512	36	24	uses	use	VERB
fcis-7512	36	25	the	the	DET
fcis-7512	36	26	correlation	correlation	NOUN
fcis-7512	36	27	between	between	ADP
fcis-7512	36	28	single	single	ADJ
fcis-7512	36	29	-	-	PUNCT
fcis-7512	36	30	mode	mode	NOUN
fcis-7512	36	31	and	and	CCONJ
fcis-7512	36	32	multi	multi	ADJ
fcis-7512	36	33	-	-	ADJ
fcis-7512	36	34	mode	mode	ADJ
fcis-7512	36	35	emotion	emotion	NOUN
fcis-7512	36	36	analysis	analysis	NOUN
fcis-7512	36	37	tasks	task	NOUN
fcis-7512	36	38	to	to	PART
fcis-7512	36	39	achieve	achieve	VERB
fcis-7512	36	40	emotion	emotion	NOUN
fcis-7512	36	41	analysis	analysis	NOUN
fcis-7512	36	42	tasks	task	NOUN
fcis-7512	36	43	.	.	PUNCT
fcis-7512	37	1	in	in	ADP
fcis-7512	37	2	natural	natural	ADJ
fcis-7512	37	3	language	language	NOUN
fcis-7512	37	4	tasks	task	NOUN
fcis-7512	37	5	,	,	PUNCT
fcis-7512	37	6	context	context	NOUN
fcis-7512	37	7	information	information	NOUN
fcis-7512	37	8	has	have	VERB
fcis-7512	37	9	a	a	DET
fcis-7512	37	10	great	great	ADJ
fcis-7512	37	11	impact	impact	NOUN
fcis-7512	37	12	on	on	ADP
fcis-7512	37	13	the	the	DET
fcis-7512	37	14	overall	overall	ADJ
fcis-7512	37	15	recognition	recognition	NOUN
fcis-7512	37	16	performance	performance	NOUN
fcis-7512	37	17	[	[	X
fcis-7512	37	18	13	13	NUM
fcis-7512	37	19	]	]	PUNCT
fcis-7512	37	20	,	,	PUNCT
fcis-7512	37	21	so	so	ADV
fcis-7512	37	22	in	in	ADP
fcis-7512	37	23	single	single	ADJ
fcis-7512	37	24	-	-	PUNCT
fcis-7512	37	25	mode	mode	NOUN
fcis-7512	37	26	emotion	emotion	NOUN
fcis-7512	37	27	classification	classification	NOUN
fcis-7512	37	28	tasks	task	NOUN
fcis-7512	37	29	,	,	PUNCT
fcis-7512	37	30	in	in	ADP
fcis-7512	37	31	this	this	DET
fcis-7512	37	32	paper	paper	NOUN
fcis-7512	37	33	,	,	PUNCT
fcis-7512	37	34	the	the	DET
fcis-7512	37	35	self	self	NOUN
fcis-7512	37	36	-	-	PUNCT
fcis-7512	37	37	attention	attention	NOUN
fcis-7512	37	38	mechanism	mechanism	NOUN
fcis-7512	37	39	is	be	AUX
fcis-7512	37	40	used	use	VERB
fcis-7512	37	41	to	to	PART
fcis-7512	37	42	consider	consider	VERB
fcis-7512	37	43	the	the	DET
fcis-7512	37	44	semantic	semantic	ADJ
fcis-7512	37	45	information	information	NOUN
fcis-7512	37	46	of	of	ADP
fcis-7512	37	47	the	the	DET
fcis-7512	37	48	single	single	ADJ
fcis-7512	37	49	mode	mode	NOUN
fcis-7512	37	50	itself	itself	PRON
fcis-7512	37	51	,	,	PUNCT
fcis-7512	37	52	and	and	CCONJ
fcis-7512	37	53	the	the	DET
fcis-7512	37	54	interactive	interactive	ADJ
fcis-7512	37	55	attention	attention	NOUN
fcis-7512	37	56	mechanism	mechanism	NOUN
fcis-7512	37	57	is	be	AUX
fcis-7512	37	58	used	use	VERB
fcis-7512	37	59	to	to	PART
fcis-7512	37	60	achieve	achieve	VERB
fcis-7512	37	61	more	more	ADV
fcis-7512	37	62	fully	fully	ADV
fcis-7512	37	63	multi	multi	ADJ
fcis-7512	37	64	-	-	ADJ
fcis-7512	37	65	mode	mode	ADJ
fcis-7512	37	66	feature	feature	NOUN
fcis-7512	37	67	fusion	fusion	NOUN
fcis-7512	37	68	.	.	PUNCT
fcis-7512	38	1	the	the	DET
fcis-7512	38	2	innovation	innovation	NOUN
fcis-7512	38	3	of	of	ADP
fcis-7512	38	4	this	this	DET
fcis-7512	38	5	paper	paper	NOUN
fcis-7512	38	6	mainly	mainly	ADV
fcis-7512	38	7	includes	include	VERB
fcis-7512	38	8	two	two	NUM
fcis-7512	38	9	parts	part	NOUN
fcis-7512	38	10	:	:	PUNCT
fcis-7512	38	11	attention	attention	NOUN
fcis-7512	38	12	mechanism	mechanism	NOUN
fcis-7512	38	13	and	and	CCONJ
fcis-7512	38	14	multi	multi	ADJ
fcis-7512	38	15	-	-	ADJ
fcis-7512	38	16	task	task	ADJ
fcis-7512	38	17	joint	joint	ADJ
fcis-7512	38	18	training	training	NOUN
fcis-7512	38	19	.	.	PUNCT
fcis-7512	39	1	the	the	DET
fcis-7512	39	2	attention	attention	NOUN
fcis-7512	39	3	mechanism	mechanism	NOUN
fcis-7512	39	4	consists	consist	VERB
fcis-7512	39	5	of	of	ADP
fcis-7512	39	6	two	two	NUM
fcis-7512	39	7	parts	part	NOUN
fcis-7512	39	8	:	:	PUNCT
fcis-7512	39	9	intermodal	intermodal	ADJ
fcis-7512	39	10	attention	attention	NOUN
fcis-7512	39	11	mechanism	mechanism	NOUN
fcis-7512	39	12	and	and	CCONJ
fcis-7512	39	13	unimodal	unimodal	ADJ
fcis-7512	39	14	self	self	NOUN
fcis-7512	39	15	-	-	PUNCT
fcis-7512	39	16	attention	attention	NOUN
fcis-7512	39	17	mechanism	mechanism	NOUN
fcis-7512	39	18	.	.	PUNCT
fcis-7512	40	1	1	1	NUM
fcis-7512	40	2	)	)	PUNCT
fcis-7512	40	3	attention	attention	NOUN
fcis-7512	40	4	mechanism	mechanism	NOUN
fcis-7512	40	5	between	between	ADP
fcis-7512	40	6	modal	modal	NOUN
fcis-7512	40	7	:	:	PUNCT
fcis-7512	40	8	through	through	ADP
fcis-7512	40	9	two	two	NUM
fcis-7512	40	10	attention	attention	NOUN
fcis-7512	40	11	mechanism	mechanism	NOUN
fcis-7512	40	12	between	between	ADP
fcis-7512	40	13	modal	modal	NOUN
fcis-7512	40	14	corresponds	correspond	NOUN
fcis-7512	40	15	to	to	ADP
fcis-7512	40	16	a	a	DET
fcis-7512	40	17	weight	weight	NOUN
fcis-7512	40	18	of	of	ADP
fcis-7512	40	19	the	the	DET
fcis-7512	40	20	other	other	ADJ
fcis-7512	40	21	two	two	NUM
fcis-7512	40	22	kinds	kind	NOUN
fcis-7512	40	23	of	of	ADP
fcis-7512	40	24	mode	mode	NOUN
fcis-7512	40	25	,	,	PUNCT
fcis-7512	40	26	give	give	VERB
fcis-7512	40	27	full	full	ADJ
fcis-7512	40	28	consideration	consideration	NOUN
fcis-7512	40	29	to	to	ADP
fcis-7512	40	30	the	the	DET
fcis-7512	40	31	modal	modal	ADJ
fcis-7512	40	32	semantic	semantic	ADJ
fcis-7512	40	33	association	association	NOUN
fcis-7512	40	34	between	between	ADP
fcis-7512	40	35	the	the	DET
fcis-7512	40	36	information	information	NOUN
fcis-7512	40	37	interaction	interaction	NOUN
fcis-7512	40	38	,	,	PUNCT
fcis-7512	40	39	making	make	VERB
fcis-7512	40	40	models	model	NOUN
fcis-7512	40	41	focus	focus	VERB
fcis-7512	40	42	on	on	ADP
fcis-7512	40	43	the	the	DET
fcis-7512	40	44	more	more	ADV
fcis-7512	40	45	important	important	ADJ
fcis-7512	40	46	68	68	NUM
fcis-7512	40	47	information	information	NOUN
fcis-7512	40	48	,	,	PUNCT
fcis-7512	40	49	achieve	achieve	VERB
fcis-7512	40	50	the	the	DET
fcis-7512	40	51	purpose	purpose	NOUN
fcis-7512	40	52	of	of	ADP
fcis-7512	40	53	reducing	reduce	VERB
fcis-7512	40	54	noise	noise	NOUN
fcis-7512	40	55	.	.	PUNCT
fcis-7512	41	1	2	2	NUM
fcis-7512	41	2	)	)	PUNCT
fcis-7512	41	3	single	single	ADJ
fcis-7512	41	4	-	-	PUNCT
fcis-7512	41	5	mode	mode	NOUN
fcis-7512	41	6	state	state	NOUN
fcis-7512	41	7	since	since	SCONJ
fcis-7512	41	8	the	the	DET
fcis-7512	41	9	attention	attention	NOUN
fcis-7512	41	10	mechanism	mechanism	NOUN
fcis-7512	41	11	:	:	PUNCT
fcis-7512	41	12	by	by	ADP
fcis-7512	41	13	using	use	VERB
fcis-7512	41	14	single	single	ADJ
fcis-7512	41	15	mode	mode	NOUN
fcis-7512	41	16	state	state	NOUN
fcis-7512	41	17	context	context	NOUN
fcis-7512	41	18	information	information	NOUN
fcis-7512	41	19	,	,	PUNCT
fcis-7512	41	20	in	in	ADP
fcis-7512	41	21	order	order	NOUN
fcis-7512	41	22	to	to	PART
fcis-7512	41	23	realize	realize	VERB
fcis-7512	41	24	the	the	DET
fcis-7512	41	25	purpose	purpose	NOUN
fcis-7512	41	26	of	of	ADP
fcis-7512	41	27	information	information	NOUN
fcis-7512	41	28	gain	gain	NOUN
fcis-7512	41	29	.	.	PUNCT
fcis-7512	42	1	3	3	X
fcis-7512	42	2	)	)	PUNCT
fcis-7512	42	3	multitasking	multitaske	VERB
fcis-7512	42	4	joint	joint	ADJ
fcis-7512	42	5	training	training	NOUN
fcis-7512	42	6	:	:	PUNCT
fcis-7512	42	7	learn	learn	VERB
fcis-7512	42	8	by	by	ADP
fcis-7512	42	9	multitasking	multitaske	VERB
fcis-7512	42	10	to	to	AUX
fcis-7512	42	11	multimodal	multimodal	VERB
fcis-7512	42	12	emotional	emotional	ADJ
fcis-7512	42	13	analysis	analysis	NOUN
fcis-7512	42	14	tasks	task	NOUN
fcis-7512	42	15	and	and	CCONJ
fcis-7512	42	16	single	single	ADJ
fcis-7512	42	17	-	-	PUNCT
fcis-7512	42	18	mode	mode	NOUN
fcis-7512	42	19	state	state	NOUN
fcis-7512	42	20	analysis	analysis	NOUN
fcis-7512	42	21	task	task	NOUN
fcis-7512	42	22	joint	joint	ADJ
fcis-7512	42	23	training	training	NOUN
fcis-7512	42	24	in	in	ADP
fcis-7512	42	25	order	order	NOUN
fcis-7512	42	26	to	to	PART
fcis-7512	42	27	realize	realize	VERB
fcis-7512	42	28	the	the	DET
fcis-7512	42	29	related	related	ADJ
fcis-7512	42	30	tasks	task	NOUN
fcis-7512	42	31	associated	associate	VERB
fcis-7512	42	32	with	with	ADP
fcis-7512	42	33	constraints	constraint	NOUN
fcis-7512	42	34	,	,	PUNCT
fcis-7512	42	35	and	and	CCONJ
fcis-7512	42	36	the	the	DET
fcis-7512	42	37	weight	weight	NOUN
fcis-7512	42	38	given	give	VERB
fcis-7512	42	39	dynamic	dynamic	ADJ
fcis-7512	42	40	loss	loss	NOUN
fcis-7512	42	41	for	for	ADP
fcis-7512	42	42	different	different	ADJ
fcis-7512	42	43	mission	mission	NOUN
fcis-7512	42	44	,	,	PUNCT
fcis-7512	42	45	make	make	VERB
fcis-7512	42	46	model	model	NOUN
fcis-7512	42	47	can	can	AUX
fcis-7512	42	48	focus	focus	VERB
fcis-7512	42	49	on	on	ADP
fcis-7512	42	50	loss	loss	NOUN
fcis-7512	42	51	value	value	NOUN
fcis-7512	42	52	larger	large	ADJ
fcis-7512	42	53	task	task	NOUN
fcis-7512	42	54	,	,	PUNCT
fcis-7512	42	55	optimizing	optimize	VERB
fcis-7512	42	56	the	the	DET
fcis-7512	42	57	training	training	NOUN
fcis-7512	42	58	process	process	NOUN
fcis-7512	42	59	,	,	PUNCT
fcis-7512	42	60	speed	speed	VERB
fcis-7512	42	61	up	up	ADP
fcis-7512	42	62	the	the	DET
fcis-7512	42	63	model	model	NOUN
fcis-7512	42	64	convergence	convergence	NOUN
fcis-7512	42	65	.	.	PUNCT
fcis-7512	43	1	2	2	X
fcis-7512	43	2	.	.	X
fcis-7512	43	3	method	method	PROPN
fcis-7512	43	4	2.1	2.1	NUM
fcis-7512	43	5	.	.	PUNCT
fcis-7512	44	1	problem	problem	NOUN
fcis-7512	44	2	definition	definition	NOUN
fcis-7512	44	3	in	in	ADP
fcis-7512	44	4	the	the	DET
fcis-7512	44	5	multimodal	multimodal	ADJ
fcis-7512	44	6	sentiment	sentiment	NOUN
fcis-7512	44	7	analysis	analysis	NOUN
fcis-7512	44	8	task	task	NOUN
fcis-7512	44	9	,	,	PUNCT
fcis-7512	44	10	the	the	DET
fcis-7512	44	11	main	main	ADJ
fcis-7512	44	12	task	task	NOUN
fcis-7512	44	13	is	be	AUX
fcis-7512	44	14	to	to	PART
fcis-7512	44	15	determine	determine	VERB
fcis-7512	44	16	the	the	DET
fcis-7512	44	17	emotional	emotional	ADJ
fcis-7512	44	18	polarity	polarity	NOUN
fcis-7512	44	19	or	or	CCONJ
fcis-7512	44	20	intensity	intensity	NOUN
fcis-7512	44	21	of	of	ADP
fcis-7512	44	22	the	the	DET
fcis-7512	44	23	video	video	NOUN
fcis-7512	44	24	segment	segment	NOUN
fcis-7512	44	25	,	,	PUNCT
fcis-7512	44	26	so	so	ADV
fcis-7512	44	27	generally	generally	ADV
fcis-7512	44	28	,	,	PUNCT
fcis-7512	44	29	multimodal	multimodal	ADJ
fcis-7512	44	30	sentiment	sentiment	NOUN
fcis-7512	44	31	analysis	analysis	NOUN
fcis-7512	44	32	can	can	AUX
fcis-7512	44	33	be	be	AUX
fcis-7512	44	34	regarded	regard	VERB
fcis-7512	44	35	as	as	ADP
fcis-7512	44	36	either	either	CCONJ
fcis-7512	44	37	a	a	DET
fcis-7512	44	38	classification	classification	NOUN
fcis-7512	44	39	task	task	NOUN
fcis-7512	44	40	or	or	CCONJ
fcis-7512	44	41	a	a	DET
fcis-7512	44	42	regression	regression	NOUN
fcis-7512	44	43	task	task	NOUN
fcis-7512	44	44	.	.	PUNCT
fcis-7512	45	1	in	in	ADP
fcis-7512	45	2	this	this	DET
fcis-7512	45	3	paper	paper	NOUN
fcis-7512	45	4	,	,	PUNCT
fcis-7512	45	5	multimodal	multimodal	ADJ
fcis-7512	45	6	emotion	emotion	NOUN
fcis-7512	45	7	analysis	analysis	NOUN
fcis-7512	45	8	is	be	AUX
fcis-7512	45	9	regarded	regard	VERB
fcis-7512	45	10	as	as	ADP
fcis-7512	45	11	a	a	DET
fcis-7512	45	12	regression	regression	NOUN
fcis-7512	45	13	task	task	NOUN
fcis-7512	45	14	to	to	PART
fcis-7512	45	15	judge	judge	VERB
fcis-7512	45	16	the	the	DET
fcis-7512	45	17	emotional	emotional	ADJ
fcis-7512	45	18	intensity	intensity	NOUN
fcis-7512	45	19	of	of	ADP
fcis-7512	45	20	video	video	NOUN
fcis-7512	45	21	clips	clip	NOUN
fcis-7512	45	22	.	.	PUNCT
fcis-7512	46	1	the	the	DET
fcis-7512	46	2	model	model	NOUN
fcis-7512	46	3	input	input	NOUN
fcis-7512	46	4	in	in	ADP
fcis-7512	46	5	this	this	DET
fcis-7512	46	6	paper	paper	NOUN
fcis-7512	46	7	is	be	AUX
fcis-7512	46	8	unimodal	unimodal	ADJ
fcis-7512	46	9	original	original	ADJ
fcis-7512	46	10	sequence	sequence	NOUN
fcis-7512	46	11	𝑋𝑚	𝑋𝑚	PROPN
fcis-7512	46	12	∈	∈	PROPN
fcis-7512	46	13	ℝ𝑙𝑚×𝑑𝑚	ℝ𝑙𝑚×𝑑𝑚	NOUN
fcis-7512	46	14	,	,	PUNCT
fcis-7512	46	15	where	where	SCONJ
fcis-7512	46	16	𝑋𝑚	𝑋𝑚	PROPN
fcis-7512	46	17	represents	represent	VERB
fcis-7512	46	18	the	the	DET
fcis-7512	46	19	original	original	ADJ
fcis-7512	46	20	sequence	sequence	NOUN
fcis-7512	46	21	of	of	ADP
fcis-7512	46	22	input	input	NOUN
fcis-7512	46	23	,	,	PUNCT
fcis-7512	46	24	𝑙𝑚	𝑙𝑚	PROPN
fcis-7512	46	25	represents	represent	VERB
fcis-7512	46	26	the	the	DET
fcis-7512	46	27	length	length	NOUN
fcis-7512	46	28	of	of	ADP
fcis-7512	46	29	the	the	DET
fcis-7512	46	30	sequence	sequence	NOUN
fcis-7512	46	31	,	,	PUNCT
fcis-7512	46	32	and	and	CCONJ
fcis-7512	46	33	𝑑𝑚	𝑑𝑚	NOUN
fcis-7512	46	34	represents	represent	VERB
fcis-7512	46	35	the	the	DET
fcis-7512	46	36	dimension	dimension	NOUN
fcis-7512	46	37	represented	represent	VERB
fcis-7512	46	38	by	by	ADP
fcis-7512	46	39	the	the	DET
fcis-7512	46	40	vector	vector	NOUN
fcis-7512	46	41	of	of	ADP
fcis-7512	46	42	the	the	DET
fcis-7512	46	43	input	input	NOUN
fcis-7512	46	44	sequence	sequence	NOUN
fcis-7512	46	45	.	.	PUNCT
fcis-7512	47	1	in	in	ADP
fcis-7512	47	2	this	this	DET
fcis-7512	47	3	paper	paper	NOUN
fcis-7512	47	4	,	,	PUNCT
fcis-7512	47	5	𝑚	𝑚	PROPN
fcis-7512	47	6	∈	∈	PROPN
fcis-7512	47	7	{	{	PUNCT
fcis-7512	47	8	a	a	PRON
fcis-7512	47	9	,	,	PUNCT
fcis-7512	47	10	v	v	NOUN
fcis-7512	47	11	,	,	PUNCT
fcis-7512	47	12	t	t	PROPN
fcis-7512	47	13	}	}	PUNCT
fcis-7512	47	14	,	,	PUNCT
fcis-7512	47	15	where	where	SCONJ
fcis-7512	47	16	a	a	PRON
fcis-7512	47	17	is	be	AUX
fcis-7512	47	18	an	an	DET
fcis-7512	47	19	audio	audio	ADJ
fcis-7512	47	20	sequence	sequence	NOUN
fcis-7512	47	21	,	,	PUNCT
fcis-7512	47	22	v	v	NOUN
fcis-7512	47	23	is	be	AUX
fcis-7512	47	24	an	an	DET
fcis-7512	47	25	image	image	NOUN
fcis-7512	47	26	sequence	sequence	NOUN
fcis-7512	47	27	,	,	PUNCT
fcis-7512	47	28	and	and	CCONJ
fcis-7512	47	29	t	t	PROPN
fcis-7512	47	30	is	be	AUX
fcis-7512	47	31	a	a	DET
fcis-7512	47	32	text	text	NOUN
fcis-7512	47	33	sequence	sequence	NOUN
fcis-7512	47	34	.	.	PUNCT
fcis-7512	48	1	output	output	NOUN
fcis-7512	48	2	for	for	ADP
fcis-7512	48	3	multiple	multiple	ADJ
fcis-7512	48	4	modal	modal	ADJ
fcis-7512	48	5	emotional	emotional	ADJ
fcis-7512	48	6	intensity	intensity	NOUN
fcis-7512	48	7	�	�	PROPN
fcis-7512	48	8	̂	̂	NOUN
fcis-7512	48	9	�	�	NOUN
fcis-7512	48	10	m	m	NOUN
fcis-7512	48	11	∈	∈	PROPN
fcis-7512	48	12	𝑅	𝑅	PROPN
fcis-7512	48	13	,	,	PUNCT
fcis-7512	48	14	in	in	ADP
fcis-7512	48	15	the	the	DET
fcis-7512	48	16	stage	stage	NOUN
fcis-7512	48	17	of	of	ADP
fcis-7512	48	18	training	train	VERB
fcis-7512	48	19	this	this	DET
fcis-7512	48	20	model	model	NOUN
fcis-7512	48	21	with	with	ADP
fcis-7512	48	22	the	the	DET
fcis-7512	48	23	other	other	ADJ
fcis-7512	48	24	three	three	NUM
fcis-7512	48	25	output	output	NOUN
fcis-7512	48	26	�	�	NOUN
fcis-7512	48	27	̂	̂	NOUN
fcis-7512	48	28	�	�	NOUN
fcis-7512	48	29	𝑛	𝑛	PRON
fcis-7512	48	30	∈	∈	PROPN
fcis-7512	48	31	𝑅	𝑅	PROPN
fcis-7512	48	32	where	where	SCONJ
fcis-7512	48	33	𝑛	𝑛	DET
fcis-7512	48	34	∈	∈	PROPN
fcis-7512	48	35	{	{	PUNCT
fcis-7512	48	36	a	a	PRON
fcis-7512	48	37	,	,	PUNCT
fcis-7512	48	38	v	v	NOUN
fcis-7512	48	39	,	,	PUNCT
fcis-7512	48	40	t	t	PROPN
fcis-7512	48	41	}	}	PUNCT
fcis-7512	48	42	,	,	PUNCT
fcis-7512	48	43	this	this	DET
fcis-7512	48	44	paper	paper	NOUN
fcis-7512	48	45	uses	use	VERB
fcis-7512	48	46	only	only	ADV
fcis-7512	48	47	�	�	PROPN
fcis-7512	48	48	̂	̂	NOUN
fcis-7512	48	49	�	�	NOUN
fcis-7512	48	50	𝑚	𝑚	NOUN
fcis-7512	48	51	as	as	ADP
fcis-7512	48	52	the	the	DET
fcis-7512	48	53	final	final	ADJ
fcis-7512	48	54	prediction	prediction	NOUN
fcis-7512	48	55	results	result	NOUN
fcis-7512	48	56	.	.	PUNCT
fcis-7512	49	1	2.2	2.2	NUM
fcis-7512	49	2	.	.	PUNCT
fcis-7512	49	3	integral	integral	ADJ
fcis-7512	49	4	structure	structure	NOUN
fcis-7512	49	5	the	the	DET
fcis-7512	49	6	overall	overall	ADJ
fcis-7512	49	7	structure	structure	NOUN
fcis-7512	49	8	of	of	ADP
fcis-7512	49	9	the	the	DET
fcis-7512	49	10	model	model	NOUN
fcis-7512	49	11	in	in	ADP
fcis-7512	49	12	this	this	DET
fcis-7512	49	13	paper	paper	NOUN
fcis-7512	49	14	is	be	AUX
fcis-7512	49	15	shown	show	VERB
fcis-7512	49	16	in	in	ADP
fcis-7512	49	17	figure	figure	NOUN
fcis-7512	49	18	1	1	NUM
fcis-7512	49	19	.	.	PUNCT
fcis-7512	50	1	the	the	DET
fcis-7512	50	2	model	model	NOUN
fcis-7512	50	3	is	be	AUX
fcis-7512	50	4	mainly	mainly	ADV
fcis-7512	50	5	divided	divide	VERB
fcis-7512	50	6	into	into	ADP
fcis-7512	50	7	two	two	NUM
fcis-7512	50	8	parts	part	NOUN
fcis-7512	50	9	:	:	PUNCT
fcis-7512	50	10	multimodal	multimodal	ADJ
fcis-7512	50	11	sentiment	sentiment	NOUN
fcis-7512	50	12	analysis	analysis	NOUN
fcis-7512	50	13	and	and	CCONJ
fcis-7512	50	14	three	three	NUM
fcis-7512	50	15	independent	independent	ADJ
fcis-7512	50	16	single	single	ADJ
fcis-7512	50	17	-	-	PUNCT
fcis-7512	50	18	modal	modal	ADJ
fcis-7512	50	19	sentiment	sentiment	NOUN
fcis-7512	50	20	analysis	analysis	NOUN
fcis-7512	50	21	.	.	PUNCT
fcis-7512	51	1	multi	multi	ADJ
fcis-7512	51	2	-	-	NOUN
fcis-7512	51	3	task	task	NOUN
fcis-7512	51	4	[	[	X
fcis-7512	51	5	11	11	NUM
fcis-7512	51	6	]	]	X
fcis-7512	51	7	learning	learn	VERB
fcis-7512	51	8	mode	mode	NOUN
fcis-7512	51	9	is	be	AUX
fcis-7512	51	10	adopted	adopt	VERB
fcis-7512	51	11	.	.	PUNCT
fcis-7512	52	1	multi	multi	ADJ
fcis-7512	52	2	-	-	ADJ
fcis-7512	52	3	modal	modal	ADJ
fcis-7512	52	4	sentiment	sentiment	NOUN
fcis-7512	52	5	analysis	analysis	NOUN
fcis-7512	52	6	is	be	AUX
fcis-7512	52	7	used	use	VERB
fcis-7512	52	8	as	as	ADP
fcis-7512	52	9	the	the	DET
fcis-7512	52	10	main	main	ADJ
fcis-7512	52	11	task	task	NOUN
fcis-7512	52	12	in	in	ADP
fcis-7512	52	13	multi	multi	ADJ
fcis-7512	52	14	-	-	ADJ
fcis-7512	52	15	task	task	ADJ
fcis-7512	52	16	learning	learning	NOUN
fcis-7512	52	17	,	,	PUNCT
fcis-7512	52	18	and	and	CCONJ
fcis-7512	52	19	three	three	NUM
fcis-7512	52	20	independent	independent	ADJ
fcis-7512	52	21	single	single	ADJ
fcis-7512	52	22	-	-	PUNCT
fcis-7512	52	23	modal	modal	ADJ
fcis-7512	52	24	sentiment	sentiment	NOUN
fcis-7512	52	25	analysis	analysis	NOUN
fcis-7512	52	26	is	be	AUX
fcis-7512	52	27	used	use	VERB
fcis-7512	52	28	as	as	ADP
fcis-7512	52	29	the	the	DET
fcis-7512	52	30	auxiliary	auxiliary	ADJ
fcis-7512	52	31	task	task	NOUN
fcis-7512	52	32	in	in	ADP
fcis-7512	52	33	multi	multi	ADJ
fcis-7512	52	34	-	-	ADJ
fcis-7512	52	35	task	task	ADJ
fcis-7512	52	36	learning	learning	NOUN
fcis-7512	52	37	.	.	PUNCT
fcis-7512	53	1	the	the	DET
fcis-7512	53	2	learning	learn	VERB
fcis-7512	53	3	mode	mode	NOUN
fcis-7512	53	4	of	of	ADP
fcis-7512	53	5	sharing	share	VERB
fcis-7512	53	6	underlying	underlying	ADJ
fcis-7512	53	7	representation	representation	NOUN
fcis-7512	53	8	is	be	AUX
fcis-7512	53	9	adopted	adopt	VERB
fcis-7512	53	10	among	among	ADP
fcis-7512	53	11	different	different	ADJ
fcis-7512	53	12	tasks	task	NOUN
fcis-7512	53	13	.	.	PUNCT
fcis-7512	54	1	text	text	NOUN
fcis-7512	54	2	video	video	NOUN
fcis-7512	54	3	audio	audio	PROPN
fcis-7512	54	4	bert	bert	PROPN
fcis-7512	54	5	openface	openface	PROPN
fcis-7512	54	6	librosa	librosa	PROPN
fcis-7512	54	7	bi	bi	ADJ
fcis-7512	54	8	-	-	ADJ
fcis-7512	54	9	lstm	lstm	ADJ
fcis-7512	54	10	intermodal	intermodal	ADJ
fcis-7512	54	11	attention	attention	NOUN
fcis-7512	54	12	intermodal	intermodal	NOUN
fcis-7512	54	13	attention	attention	NOUN
fcis-7512	54	14	intermodal	intermodal	ADJ
fcis-7512	54	15	attention	attention	NOUN
fcis-7512	54	16	fully	fully	ADV
fcis-7512	54	17	connected	connect	VERB
fcis-7512	54	18	layer	layer	NOUN
fcis-7512	54	19	relu	relu	NOUN
fcis-7512	54	20	self	self	NOUN
fcis-7512	54	21	-	-	PUNCT
fcis-7512	54	22	attention	attention	NOUN
fcis-7512	54	23	mechanism	mechanism	NOUN
fcis-7512	54	24	relu	relu	NOUN
fcis-7512	54	25	ulgm	ulgm	ADJ
fcis-7512	54	26	self	self	NOUN
fcis-7512	54	27	-	-	PUNCT
fcis-7512	54	28	attention	attention	NOUN
fcis-7512	54	29	mechanism	mechanism	NOUN
fcis-7512	54	30	self	self	NOUN
fcis-7512	54	31	-	-	PUNCT
fcis-7512	54	32	attention	attention	NOUN
fcis-7512	54	33	mechanism	mechanism	NOUN
fcis-7512	54	34	ulgm	ulgm	ADJ
fcis-7512	54	35	relu	relu	NOUN
fcis-7512	54	36	relu	relu	NOUN
fcis-7512	54	37	ulgm	ulgm	ADJ
fcis-7512	54	38	weight	weight	NOUN
fcis-7512	54	39	loss	loss	NOUN
fcis-7512	54	40	bi	bi	ADJ
fcis-7512	54	41	-	-	ADJ
fcis-7512	54	42	lstm	lstm	ADJ
fcis-7512	54	43	multimodal	multimodal	NOUN
fcis-7512	54	44	affective	affective	NOUN
fcis-7512	54	45	intensity	intensity	NOUN
fcis-7512	54	46	f	f	NOUN
fcis-7512	54	47	f	f	X
fcis-7512	54	48	*	*	PUNCT
fcis-7512	54	49	figure	figure	NOUN
fcis-7512	54	50	1	1	NUM
fcis-7512	54	51	.	.	PUNCT
fcis-7512	55	1	overall	overall	ADJ
fcis-7512	55	2	model	model	NOUN
fcis-7512	55	3	structure	structure	NOUN
fcis-7512	55	4	2.3	2.3	NUM
fcis-7512	55	5	.	.	PUNCT
fcis-7512	56	1	multimodal	multimodal	NOUN
fcis-7512	56	2	data	datum	NOUN
fcis-7512	56	3	representation	representation	NOUN
fcis-7512	56	4	firstly	firstly	ADV
fcis-7512	56	5	,	,	PUNCT
fcis-7512	56	6	the	the	DET
fcis-7512	56	7	data	data	NOUN
fcis-7512	56	8	information	information	NOUN
fcis-7512	56	9	of	of	ADP
fcis-7512	56	10	three	three	NUM
fcis-7512	56	11	modes	mode	NOUN
fcis-7512	56	12	(	(	PUNCT
fcis-7512	56	13	text	text	NOUN
fcis-7512	56	14	,	,	PUNCT
fcis-7512	56	15	video	video	NOUN
fcis-7512	56	16	and	and	CCONJ
fcis-7512	56	17	audio	audio	NOUN
fcis-7512	56	18	)	)	PUNCT
fcis-7512	56	19	is	be	AUX
fcis-7512	56	20	extracted	extract	VERB
fcis-7512	56	21	from	from	ADP
fcis-7512	56	22	the	the	DET
fcis-7512	56	23	video	video	NOUN
fcis-7512	56	24	segment	segment	NOUN
fcis-7512	56	25	,	,	PUNCT
fcis-7512	56	26	and	and	CCONJ
fcis-7512	56	27	the	the	DET
fcis-7512	56	28	input	input	NOUN
fcis-7512	56	29	sequence	sequence	NOUN
fcis-7512	56	30	𝑋𝑚	𝑋𝑚	PROPN
fcis-7512	56	31	of	of	ADP
fcis-7512	56	32	the	the	DET
fcis-7512	56	33	three	three	NUM
fcis-7512	56	34	modes	mode	NOUN
fcis-7512	56	35	is	be	AUX
fcis-7512	56	36	encoded	encode	VERB
fcis-7512	56	37	into	into	ADP
fcis-7512	56	38	a	a	DET
fcis-7512	56	39	vector	vector	NOUN
fcis-7512	56	40	representing	represent	VERB
fcis-7512	56	41	ℎ𝑚	ℎ𝑚	NOUN
fcis-7512	56	42	,	,	PUNCT
fcis-7512	56	43	where	where	SCONJ
fcis-7512	56	44	𝑚	𝑚	PROPN
fcis-7512	56	45	∈	∈	PROPN
fcis-7512	56	46	{	{	PUNCT
fcis-7512	56	47	a	a	PRON
fcis-7512	56	48	,	,	PUNCT
fcis-7512	56	49	v	v	NOUN
fcis-7512	56	50	,	,	PUNCT
fcis-7512	56	51	t	t	PROPN
fcis-7512	56	52	}	}	PUNCT
fcis-7512	56	53	.	.	PUNCT
fcis-7512	57	1	specific	specific	ADJ
fcis-7512	57	2	implementation	implementation	NOUN
fcis-7512	57	3	method	method	NOUN
fcis-7512	57	4	:	:	PUNCT
fcis-7512	57	5	for	for	ADP
fcis-7512	57	6	text	text	NOUN
fcis-7512	57	7	-	-	PUNCT
fcis-7512	57	8	like	like	ADJ
fcis-7512	57	9	data	datum	NOUN
fcis-7512	57	10	,	,	PUNCT
fcis-7512	57	11	bert	bert	PROPN
fcis-7512	58	1	[	[	X
fcis-7512	58	2	13	13	NUM
fcis-7512	58	3	]	]	PUNCT
fcis-7512	58	4	is	be	AUX
fcis-7512	58	5	used	use	VERB
fcis-7512	58	6	in	in	ADP
fcis-7512	58	7	this	this	DET
fcis-7512	58	8	paper	paper	NOUN
fcis-7512	58	9	to	to	PART
fcis-7512	58	10	encode	encode	VERB
fcis-7512	58	11	the	the	DET
fcis-7512	58	12	input	input	NOUN
fcis-7512	58	13	sequence	sequence	NOUN
fcis-7512	58	14	,	,	PUNCT
fcis-7512	58	15	and	and	CCONJ
fcis-7512	58	16	the	the	DET
fcis-7512	58	17	vector	vector	NOUN
fcis-7512	58	18	representation	representation	NOUN
fcis-7512	58	19	of	of	ADP
fcis-7512	58	20	the	the	PRON
fcis-7512	58	21	[	[	X
fcis-7512	58	22	cls	cls	NOUN
fcis-7512	58	23	]	]	X
fcis-7512	58	24	flag	flag	NOUN
fcis-7512	58	25	bit	bit	NOUN
fcis-7512	58	26	in	in	ADP
fcis-7512	58	27	the	the	DET
fcis-7512	58	28	last	last	ADJ
fcis-7512	58	29	layer	layer	NOUN
fcis-7512	58	30	is	be	AUX
fcis-7512	58	31	taken	take	VERB
fcis-7512	58	32	69	69	NUM
fcis-7512	58	33	as	as	ADP
fcis-7512	58	34	the	the	DET
fcis-7512	58	35	vector	vector	NOUN
fcis-7512	58	36	representation	representation	NOUN
fcis-7512	58	37	of	of	ADP
fcis-7512	58	38	the	the	DET
fcis-7512	58	39	whole	whole	ADJ
fcis-7512	58	40	sentence	sentence	NOUN
fcis-7512	58	41	.	.	PUNCT
fcis-7512	59	1	for	for	ADP
fcis-7512	59	2	audio	audio	ADJ
fcis-7512	59	3	data	datum	NOUN
fcis-7512	59	4	and	and	CCONJ
fcis-7512	59	5	image	image	NOUN
fcis-7512	59	6	data	datum	NOUN
fcis-7512	59	7	,	,	PUNCT
fcis-7512	59	8	bidirectional	bidirectional	ADJ
fcis-7512	59	9	lstm	lstm	NOUN
fcis-7512	59	10	[	[	X
fcis-7512	59	11	16	16	NUM
fcis-7512	59	12	]	]	PUNCT
fcis-7512	59	13	is	be	AUX
fcis-7512	59	14	used	use	VERB
fcis-7512	59	15	to	to	PART
fcis-7512	59	16	encode	encode	VERB
fcis-7512	59	17	non	non	ADJ
fcis-7512	59	18	-	-	ADJ
fcis-7512	59	19	text	text	ADJ
fcis-7512	59	20	data	datum	NOUN
fcis-7512	59	21	features	feature	NOUN
fcis-7512	59	22	according	accord	VERB
fcis-7512	59	23	to	to	ADP
fcis-7512	59	24	previous	previous	ADJ
fcis-7512	59	25	work	work	NOUN
fcis-7512	60	1	[	[	X
fcis-7512	60	2	14,15	14,15	NUM
fcis-7512	60	3	]	]	PUNCT
fcis-7512	60	4	.	.	PUNCT
fcis-7512	61	1	the	the	DET
fcis-7512	61	2	specific	specific	ADJ
fcis-7512	61	3	expression	expression	NOUN
fcis-7512	61	4	is	be	AUX
fcis-7512	61	5	as	as	ADP
fcis-7512	61	6	formula	formula	NOUN
fcis-7512	61	7	(	(	PUNCT
fcis-7512	61	8	1	1	NUM
fcis-7512	61	9	)	)	PUNCT
fcis-7512	61	10	,	,	PUNCT
fcis-7512	61	11	formula	formula	NOUN
fcis-7512	61	12	(	(	PUNCT
fcis-7512	61	13	2	2	NUM
fcis-7512	61	14	)	)	PUNCT
fcis-7512	61	15	.	.	PUNCT
fcis-7512	62	1	ht	ht	NOUN
fcis-7512	62	2	=	=	PUNCT
fcis-7512	62	3	bert(xt	bert(xt	NOUN
fcis-7512	62	4	;	;	PUNCT
fcis-7512	62	5	θt	θt	PROPN
fcis-7512	62	6	bert	bert	PROPN
fcis-7512	62	7	)	)	PUNCT
fcis-7512	62	8	.	.	PUNCT
fcis-7512	63	1	(	(	PUNCT
fcis-7512	63	2	1	1	X
fcis-7512	63	3	)	)	PUNCT
fcis-7512	63	4	hm	hm	NOUN
fcis-7512	63	5	=	=	SYM
fcis-7512	63	6	bilstm(xm	bilstm(xm	PROPN
fcis-7512	63	7	;	;	PUNCT
fcis-7512	63	8	θm	θm	PROPN
fcis-7512	63	9	bilstm	bilstm	NOUN
fcis-7512	63	10	)	)	PUNCT
fcis-7512	63	11	m	m	VERB
fcis-7512	63	12	∈	∈	NOUN
fcis-7512	63	13	{	{	PUNCT
fcis-7512	63	14	v	v	NOUN
fcis-7512	63	15	,	,	PUNCT
fcis-7512	63	16	a	a	PRON
fcis-7512	63	17	}	}	PUNCT
fcis-7512	63	18	.	.	PUNCT
fcis-7512	64	1	(	(	PUNCT
fcis-7512	64	2	2	2	X
fcis-7512	64	3	)	)	PUNCT
fcis-7512	64	4	ℎt	ℎt	PRON
fcis-7512	64	5	is	be	AUX
fcis-7512	64	6	the	the	DET
fcis-7512	64	7	text	text	NOUN
fcis-7512	64	8	data	datum	NOUN
fcis-7512	64	9	representation	representation	NOUN
fcis-7512	64	10	,	,	PUNCT
fcis-7512	64	11	and	and	CCONJ
fcis-7512	64	12	ℎ𝑚	ℎ𝑚	NOUN
fcis-7512	64	13	is	be	AUX
fcis-7512	64	14	the	the	DET
fcis-7512	64	15	non	non	ADJ
fcis-7512	64	16	-	-	ADJ
fcis-7512	64	17	text	text	ADJ
fcis-7512	64	18	data	datum	NOUN
fcis-7512	64	19	representation	representation	NOUN
fcis-7512	64	20	?	?	PUNCT
fcis-7512	65	1	three	three	NUM
fcis-7512	65	2	types	type	NOUN
fcis-7512	65	3	of	of	ADP
fcis-7512	65	4	data	datum	NOUN
fcis-7512	65	5	representation	representation	NOUN
fcis-7512	65	6	are	be	AUX
fcis-7512	65	7	the	the	DET
fcis-7512	65	8	underlying	underlying	ADJ
fcis-7512	65	9	representation	representation	NOUN
fcis-7512	65	10	shared	share	VERB
fcis-7512	65	11	by	by	ADP
fcis-7512	65	12	multitasking	multitaske	VERB
fcis-7512	65	13	learning	learning	NOUN
fcis-7512	65	14	.	.	PUNCT
fcis-7512	66	1	2.4	2.4	NUM
fcis-7512	66	2	.	.	PUNCT
fcis-7512	66	3	single	single	ADJ
fcis-7512	66	4	-	-	PUNCT
fcis-7512	66	5	mode	mode	NOUN
fcis-7512	66	6	self	self	NOUN
fcis-7512	66	7	-	-	PUNCT
fcis-7512	66	8	attention	attention	NOUN
fcis-7512	66	9	mechanism	mechanism	NOUN
fcis-7512	66	10	single	single	ADJ
fcis-7512	66	11	-	-	PUNCT
fcis-7512	66	12	mode	mode	NOUN
fcis-7512	66	13	attention	attention	NOUN
fcis-7512	66	14	mechanism	mechanism	NOUN
fcis-7512	66	15	is	be	AUX
fcis-7512	66	16	applied	apply	VERB
fcis-7512	66	17	to	to	ADP
fcis-7512	66	18	the	the	DET
fcis-7512	66	19	auxiliary	auxiliary	ADJ
fcis-7512	66	20	task	task	NOUN
fcis-7512	66	21	part	part	NOUN
fcis-7512	66	22	of	of	ADP
fcis-7512	66	23	multi	multi	ADJ
fcis-7512	66	24	-	-	NOUN
fcis-7512	66	25	task	task	ADJ
fcis-7512	66	26	learning	learning	NOUN
fcis-7512	66	27	.	.	PUNCT
fcis-7512	67	1	it	it	PRON
fcis-7512	67	2	mainly	mainly	ADV
fcis-7512	67	3	uses	use	VERB
fcis-7512	67	4	the	the	DET
fcis-7512	67	5	context	context	NOUN
fcis-7512	67	6	information	information	NOUN
fcis-7512	67	7	of	of	ADP
fcis-7512	67	8	single	single	ADJ
fcis-7512	67	9	-	-	PUNCT
fcis-7512	67	10	mode	mode	NOUN
fcis-7512	67	11	and	and	CCONJ
fcis-7512	67	12	only	only	ADV
fcis-7512	67	13	considers	consider	VERB
fcis-7512	67	14	the	the	DET
fcis-7512	67	15	importance	importance	NOUN
fcis-7512	67	16	of	of	ADP
fcis-7512	67	17	single	single	ADJ
fcis-7512	67	18	-	-	PUNCT
fcis-7512	67	19	mode	mode	NOUN
fcis-7512	67	20	to	to	PART
fcis-7512	67	21	emotion	emotion	VERB
fcis-7512	67	22	classification	classification	NOUN
fcis-7512	67	23	,	,	PUNCT
fcis-7512	67	24	without	without	ADP
fcis-7512	67	25	considering	consider	VERB
fcis-7512	67	26	the	the	DET
fcis-7512	67	27	relationship	relationship	NOUN
fcis-7512	67	28	between	between	ADP
fcis-7512	67	29	modes	mode	NOUN
fcis-7512	67	30	.	.	PUNCT
fcis-7512	68	1	concrete	concrete	ADJ
fcis-7512	68	2	implementation	implementation	NOUN
fcis-7512	68	3	way	way	NOUN
fcis-7512	68	4	:	:	PUNCT
fcis-7512	68	5	with	with	ADP
fcis-7512	68	6	specific	specific	ADJ
fcis-7512	68	7	data	datum	NOUN
fcis-7512	68	8	representation	representation	NOUN
fcis-7512	68	9	of	of	ADP
fcis-7512	68	10	single	single	ADJ
fcis-7512	68	11	-	-	PUNCT
fcis-7512	68	12	mode	mode	NOUN
fcis-7512	68	13	states	state	NOUN
fcis-7512	68	14	ℎ𝑛	ℎ𝑛	PROPN
fcis-7512	68	15	∈	∈	PROPN
fcis-7512	68	16	ℝ𝑙𝑛×ⅆ𝑛	ℝ𝑙𝑛×ⅆ𝑛	NOUN
fcis-7512	69	1	(	(	PUNCT
fcis-7512	69	2	ℎt	ℎt	ADP
fcis-7512	69	3	text	text	NOUN
fcis-7512	69	4	data	datum	NOUN
fcis-7512	69	5	representation	representation	NOUN
fcis-7512	69	6	and	and	CCONJ
fcis-7512	69	7	ℎ𝑚	ℎ𝑚	PRON
fcis-7512	69	8	non	non	ADJ
fcis-7512	69	9	-	-	ADJ
fcis-7512	69	10	text	text	ADJ
fcis-7512	69	11	data	datum	NOUN
fcis-7512	69	12	representation	representation	NOUN
fcis-7512	69	13	)	)	PUNCT
fcis-7512	69	14	which	which	PRON
fcis-7512	69	15	𝑙n	𝑙n	VERB
fcis-7512	69	16	sequence	sequence	NOUN
fcis-7512	69	17	length	length	NOUN
fcis-7512	69	18	,	,	PUNCT
fcis-7512	69	19	says	say	VERB
fcis-7512	69	20	𝑑𝑛	𝑑𝑛	PROPN
fcis-7512	69	21	said	say	VERB
fcis-7512	69	22	input	input	NOUN
fcis-7512	69	23	sequence	sequence	NOUN
fcis-7512	69	24	data	datum	NOUN
fcis-7512	69	25	representation	representation	NOUN
fcis-7512	69	26	vector	vector	NOUN
fcis-7512	69	27	dimensions	dimension	NOUN
fcis-7512	69	28	.	.	PUNCT
fcis-7512	70	1	the	the	DET
fcis-7512	70	2	original	original	ADJ
fcis-7512	70	3	input	input	NOUN
fcis-7512	70	4	unimodal	unimodal	ADJ
fcis-7512	70	5	data	datum	NOUN
fcis-7512	70	6	represents	represent	VERB
fcis-7512	70	7	ℎ𝑛.	ℎ𝑛.	NOUN
fcis-7512	70	8	by	by	ADP
fcis-7512	70	9	taking	take	VERB
fcis-7512	70	10	advantage	advantage	NOUN
fcis-7512	70	11	of	of	ADP
fcis-7512	70	12	its	its	PRON
fcis-7512	70	13	own	own	ADJ
fcis-7512	70	14	semantic	semantic	ADJ
fcis-7512	70	15	correlation	correlation	NOUN
fcis-7512	70	16	and	and	CCONJ
fcis-7512	70	17	transposing	transpose	VERB
fcis-7512	70	18	the	the	DET
fcis-7512	70	19	dot	dot	NOUN
fcis-7512	70	20	product	product	NOUN
fcis-7512	70	21	with	with	ADP
fcis-7512	70	22	its	its	PRON
fcis-7512	70	23	own	own	ADJ
fcis-7512	70	24	unimodal	unimodal	ADJ
fcis-7512	70	25	data	datum	NOUN
fcis-7512	70	26	representation	representation	NOUN
fcis-7512	70	27	,	,	PUNCT
fcis-7512	70	28	the	the	DET
fcis-7512	70	29	semantic	semantic	ADJ
fcis-7512	70	30	correlation	correlation	NOUN
fcis-7512	70	31	matrix	matrix	NOUN
fcis-7512	70	32	𝑀	𝑀	PROPN
fcis-7512	70	33	is	be	AUX
fcis-7512	70	34	calculated	calculate	VERB
fcis-7512	70	35	,	,	PUNCT
fcis-7512	70	36	and	and	CCONJ
fcis-7512	70	37	higher	high	ADJ
fcis-7512	70	38	weight	weight	NOUN
fcis-7512	70	39	is	be	AUX
fcis-7512	70	40	given	give	VERB
fcis-7512	70	41	to	to	ADP
fcis-7512	70	42	the	the	DET
fcis-7512	70	43	part	part	NOUN
fcis-7512	70	44	with	with	ADP
fcis-7512	70	45	greater	great	ADJ
fcis-7512	70	46	semantic	semantic	ADJ
fcis-7512	70	47	correlation	correlation	NOUN
fcis-7512	70	48	.	.	PUNCT
fcis-7512	71	1	after	after	ADP
fcis-7512	71	2	the	the	DET
fcis-7512	71	3	normalization	normalization	NOUN
fcis-7512	71	4	of	of	ADP
fcis-7512	71	5	the	the	DET
fcis-7512	71	6	semantic	semantic	ADJ
fcis-7512	71	7	correlation	correlation	NOUN
fcis-7512	71	8	matrix	matrix	NOUN
fcis-7512	71	9	,	,	PUNCT
fcis-7512	71	10	the	the	DET
fcis-7512	71	11	unimodal	unimodal	ADJ
fcis-7512	71	12	self	self	NOUN
fcis-7512	71	13	-	-	PUNCT
fcis-7512	71	14	attention	attention	NOUN
fcis-7512	71	15	weight	weight	NOUN
fcis-7512	71	16	matrix	matrix	NOUN
fcis-7512	71	17	𝐴	𝐴	PROPN
fcis-7512	71	18	is	be	AUX
fcis-7512	71	19	obtained	obtain	VERB
fcis-7512	71	20	.	.	PUNCT
fcis-7512	72	1	𝑀	𝑀	PROPN
fcis-7512	72	2	=	=	PUNCT
fcis-7512	72	3	ℎ𝑛	ℎ𝑛	PROPN
fcis-7512	72	4	⋅	⋅	PROPN
fcis-7512	72	5	ℎ𝑛	ℎ𝑛	PROPN
fcis-7512	72	6	𝑇.	𝑇.	PROPN
fcis-7512	72	7	(	(	PUNCT
fcis-7512	72	8	3	3	NUM
fcis-7512	72	9	)	)	PUNCT
fcis-7512	72	10	𝐴	𝐴	NOUN
fcis-7512	72	11	=	=	SYM
fcis-7512	72	12	𝑠𝑜𝑓𝑡𝑚𝑎𝑥(𝑀	𝑠𝑜𝑓𝑡𝑚𝑎𝑥(𝑀	PROPN
fcis-7512	72	13	)	)	PUNCT
fcis-7512	72	14	.	.	PUNCT
fcis-7512	73	1	(	(	PUNCT
fcis-7512	73	2	4	4	X
fcis-7512	73	3	)	)	PUNCT
fcis-7512	73	4	the	the	DET
fcis-7512	73	5	dot	dot	NOUN
fcis-7512	73	6	product	product	NOUN
fcis-7512	73	7	of	of	ADP
fcis-7512	73	8	the	the	DET
fcis-7512	73	9	attention	attention	NOUN
fcis-7512	73	10	weight	weight	NOUN
fcis-7512	73	11	matrix	matrix	NOUN
fcis-7512	73	12	a	a	DET
fcis-7512	73	13	containing	contain	VERB
fcis-7512	73	14	semantic	semantic	ADJ
fcis-7512	73	15	relevance	relevance	NOUN
fcis-7512	73	16	and	and	CCONJ
fcis-7512	73	17	the	the	DET
fcis-7512	73	18	original	original	ADJ
fcis-7512	73	19	unimodal	unimodal	ADJ
fcis-7512	73	20	data	datum	NOUN
fcis-7512	73	21	representation	representation	NOUN
fcis-7512	73	22	ℎ𝑛	ℎ𝑛	PROPN
fcis-7512	73	23	is	be	AUX
fcis-7512	73	24	obtained	obtain	VERB
fcis-7512	73	25	to	to	PART
fcis-7512	73	26	obtain	obtain	VERB
fcis-7512	73	27	a	a	DET
fcis-7512	73	28	new	new	ADJ
fcis-7512	73	29	unimodal	unimodal	ADJ
fcis-7512	73	30	data	datum	NOUN
fcis-7512	73	31	representation	representation	NOUN
fcis-7512	74	1	𝐻𝑛	𝐻𝑛	ADJ
fcis-7512	74	2	containing	contain	VERB
fcis-7512	74	3	contextual	contextual	ADJ
fcis-7512	74	4	information	information	NOUN
fcis-7512	74	5	.	.	PUNCT
fcis-7512	75	1	𝐻𝑛	𝐻𝑛	ADJ
fcis-7512	75	2	=	=	SYM
fcis-7512	75	3	𝐴	𝐴	PROPN
fcis-7512	75	4	⋅	⋅	PROPN
fcis-7512	75	5	ℎ𝑛.	ℎ𝑛.	X
fcis-7512	75	6	(	(	PUNCT
fcis-7512	75	7	5	5	NUM
fcis-7512	75	8	)	)	PUNCT
fcis-7512	75	9	finally	finally	ADV
fcis-7512	75	10	,	,	PUNCT
fcis-7512	75	11	the	the	DET
fcis-7512	75	12	resulting	result	VERB
fcis-7512	75	13	new	new	ADJ
fcis-7512	75	14	data	datum	NOUN
fcis-7512	75	15	representation	representation	NOUN
fcis-7512	75	16	is	be	AUX
fcis-7512	75	17	used	use	VERB
fcis-7512	75	18	in	in	ADP
fcis-7512	75	19	the	the	DET
fcis-7512	75	20	auxiliary	auxiliary	ADJ
fcis-7512	75	21	task	task	NOUN
fcis-7512	75	22	section	section	NOUN
fcis-7512	75	23	.	.	PUNCT
fcis-7512	76	1	2.5	2.5	NUM
fcis-7512	76	2	.	.	PUNCT
fcis-7512	77	1	intermodal	intermodal	ADJ
fcis-7512	77	2	attention	attention	NOUN
fcis-7512	77	3	mechanism	mechanism	NOUN
fcis-7512	77	4	intermodal	intermodal	ADJ
fcis-7512	77	5	attention	attention	NOUN
fcis-7512	77	6	mechanism	mechanism	NOUN
fcis-7512	77	7	is	be	AUX
fcis-7512	77	8	applied	apply	VERB
fcis-7512	77	9	to	to	ADP
fcis-7512	77	10	the	the	DET
fcis-7512	77	11	main	main	ADJ
fcis-7512	77	12	task	task	NOUN
fcis-7512	77	13	of	of	ADP
fcis-7512	77	14	multi	multi	ADJ
fcis-7512	77	15	-	-	NOUN
fcis-7512	77	16	task	task	ADJ
fcis-7512	77	17	learning	learning	NOUN
fcis-7512	77	18	to	to	PART
fcis-7512	77	19	better	well	ADV
fcis-7512	77	20	learn	learn	VERB
fcis-7512	77	21	the	the	DET
fcis-7512	77	22	hidden	hide	VERB
fcis-7512	77	23	information	information	NOUN
fcis-7512	77	24	between	between	ADP
fcis-7512	77	25	modes	mode	NOUN
fcis-7512	77	26	according	accord	VERB
fcis-7512	77	27	to	to	ADP
fcis-7512	77	28	the	the	DET
fcis-7512	77	29	information	information	NOUN
fcis-7512	77	30	interaction	interaction	NOUN
fcis-7512	77	31	between	between	ADP
fcis-7512	77	32	modes	mode	NOUN
fcis-7512	77	33	.	.	PUNCT
fcis-7512	78	1	taking	take	VERB
fcis-7512	78	2	text	text	NOUN
fcis-7512	78	3	data	datum	NOUN
fcis-7512	78	4	as	as	ADP
fcis-7512	78	5	an	an	DET
fcis-7512	78	6	example	example	NOUN
fcis-7512	78	7	,	,	PUNCT
fcis-7512	78	8	the	the	DET
fcis-7512	78	9	specific	specific	ADJ
fcis-7512	78	10	implementation	implementation	NOUN
fcis-7512	78	11	method	method	NOUN
fcis-7512	78	12	is	be	AUX
fcis-7512	78	13	as	as	SCONJ
fcis-7512	78	14	follows	follow	VERB
fcis-7512	78	15	:	:	PUNCT
fcis-7512	78	16	after	after	SCONJ
fcis-7512	78	17	the	the	DET
fcis-7512	78	18	text	text	NOUN
fcis-7512	78	19	data	datum	NOUN
fcis-7512	78	20	representation	representation	NOUN
fcis-7512	78	21	ℎt	ℎt	PRON
fcis-7512	78	22	is	be	AUX
fcis-7512	78	23	obtained	obtain	VERB
fcis-7512	78	24	,	,	PUNCT
fcis-7512	78	25	the	the	DET
fcis-7512	78	26	semantic	semantic	ADJ
fcis-7512	78	27	correlation	correlation	NOUN
fcis-7512	78	28	between	between	ADP
fcis-7512	78	29	modes	mode	NOUN
fcis-7512	78	30	is	be	AUX
fcis-7512	78	31	used	use	VERB
fcis-7512	78	32	to	to	PART
fcis-7512	78	33	calculate	calculate	VERB
fcis-7512	78	34	the	the	DET
fcis-7512	78	35	correlation	correlation	NOUN
fcis-7512	78	36	degree	degree	NOUN
fcis-7512	78	37	of	of	ADP
fcis-7512	78	38	text	text	NOUN
fcis-7512	78	39	corresponding	correspond	VERB
fcis-7512	78	40	to	to	ADP
fcis-7512	78	41	audio	audio	NOUN
fcis-7512	78	42	and	and	CCONJ
fcis-7512	78	43	video	video	NOUN
fcis-7512	78	44	respectively	respectively	ADV
fcis-7512	78	45	(	(	PUNCT
fcis-7512	78	46	denoted	denote	VERB
fcis-7512	78	47	by	by	ADP
fcis-7512	78	48	semantic	semantic	ADJ
fcis-7512	78	49	correlation	correlation	NOUN
fcis-7512	78	50	matrix	matrix	NOUN
fcis-7512	78	51	𝑀1	𝑀1	NOUN
fcis-7512	78	52	and	and	CCONJ
fcis-7512	78	53	𝑀2	𝑀2	PROPN
fcis-7512	78	54	,	,	PUNCT
fcis-7512	78	55	and	and	CCONJ
fcis-7512	78	56	the	the	DET
fcis-7512	78	57	part	part	NOUN
fcis-7512	78	58	corresponding	correspond	VERB
fcis-7512	78	59	to	to	ADP
fcis-7512	78	60	audio	audio	ADJ
fcis-7512	78	61	or	or	CCONJ
fcis-7512	78	62	video	video	NOUN
fcis-7512	78	63	correlation	correlation	NOUN
fcis-7512	78	64	degree	degree	NOUN
fcis-7512	78	65	is	be	AUX
fcis-7512	78	66	given	give	VERB
fcis-7512	78	67	high	high	ADJ
fcis-7512	78	68	weight	weight	NOUN
fcis-7512	78	69	)	)	PUNCT
fcis-7512	78	70	,	,	PUNCT
fcis-7512	78	71	and	and	CCONJ
fcis-7512	78	72	the	the	DET
fcis-7512	78	73	two	two	NUM
fcis-7512	78	74	semantic	semantic	ADJ
fcis-7512	78	75	correlation	correlation	NOUN
fcis-7512	78	76	matrices	matrix	NOUN
fcis-7512	78	77	are	be	AUX
fcis-7512	78	78	normalized	normalize	VERB
fcis-7512	78	79	.	.	PUNCT
fcis-7512	79	1	the	the	DET
fcis-7512	79	2	attention	attention	NOUN
fcis-7512	79	3	weight	weight	NOUN
fcis-7512	79	4	matrices	matrix	NOUN
fcis-7512	79	5	𝐴1	𝐴1	PROPN
fcis-7512	79	6	and	and	CCONJ
fcis-7512	79	7	𝐴2	𝐴2	PROPN
fcis-7512	79	8	corresponding	correspond	VERB
fcis-7512	79	9	to	to	ADP
fcis-7512	79	10	video	video	NOUN
fcis-7512	79	11	and	and	CCONJ
fcis-7512	79	12	audio	audio	NOUN
fcis-7512	79	13	are	be	AUX
fcis-7512	79	14	obtained	obtain	VERB
fcis-7512	79	15	.	.	PUNCT
fcis-7512	80	1	𝑀1	𝑀1	PROPN
fcis-7512	80	2	=	=	SYM
fcis-7512	80	3	ℎt	ℎt	ADP
fcis-7512	80	4	⋅	⋅	PROPN
fcis-7512	80	5	(	(	PUNCT
fcis-7512	80	6	ℎa)t	ℎa)t	ADJ
fcis-7512	80	7	.	.	PUNCT
fcis-7512	81	1	(	(	PUNCT
fcis-7512	81	2	6	6	NUM
fcis-7512	81	3	)	)	PUNCT
fcis-7512	81	4	𝑀2	𝑀2	NOUN
fcis-7512	81	5	=	=	SYM
fcis-7512	81	6	ℎt	ℎt	ADP
fcis-7512	81	7	⋅	⋅	PROPN
fcis-7512	81	8	(	(	PUNCT
fcis-7512	81	9	ℎv)t	ℎv)t	PROPN
fcis-7512	81	10	.	.	PUNCT
fcis-7512	82	1	(	(	PUNCT
fcis-7512	82	2	7	7	X
fcis-7512	82	3	)	)	PUNCT
fcis-7512	82	4	𝐻𝑛	𝐻𝑛	NOUN
fcis-7512	82	5	=	=	NOUN
fcis-7512	82	6	𝑠𝑜𝑓𝑡𝑚𝑎𝑥(𝑀1	𝑠𝑜𝑓𝑡𝑚𝑎𝑥(𝑀1	X
fcis-7512	82	7	)	)	PUNCT
fcis-7512	82	8	.	.	PUNCT
fcis-7512	83	1	(	(	PUNCT
fcis-7512	83	2	8)	8)	NUM
fcis-7512	83	3	𝐴2	𝐴2	NOUN
fcis-7512	83	4	=	=	PUNCT
fcis-7512	83	5	𝑠𝑜𝑓𝑡𝑚𝑎𝑥(𝑀2	𝑠𝑜𝑓𝑡𝑚𝑎𝑥(𝑀2	NUM
fcis-7512	83	6	)	)	PUNCT
fcis-7512	83	7	.	.	PUNCT
fcis-7512	84	1	(	(	PUNCT
fcis-7512	84	2	9	9	X
fcis-7512	84	3	)	)	PUNCT
fcis-7512	84	4	the	the	DET
fcis-7512	84	5	attention	attention	NOUN
fcis-7512	84	6	weight	weight	NOUN
fcis-7512	84	7	matrix	matrix	NOUN
fcis-7512	84	8	𝐴1	𝐴1	PROPN
fcis-7512	84	9	and	and	CCONJ
fcis-7512	84	10	𝐴2	𝐴2	PROPN
fcis-7512	84	11	are	be	AUX
fcis-7512	84	12	dot	dot	NOUN
fcis-7512	84	13	product	product	NOUN
fcis-7512	84	14	with	with	ADP
fcis-7512	84	15	the	the	DET
fcis-7512	84	16	original	original	ADJ
fcis-7512	84	17	single	single	ADJ
fcis-7512	84	18	-	-	PUNCT
fcis-7512	84	19	mode	mode	NOUN
fcis-7512	84	20	text	text	NOUN
fcis-7512	84	21	data	datum	NOUN
fcis-7512	84	22	representation	representation	NOUN
fcis-7512	84	23	,	,	PUNCT
fcis-7512	84	24	so	so	SCONJ
fcis-7512	84	25	that	that	SCONJ
fcis-7512	84	26	the	the	DET
fcis-7512	84	27	new	new	ADJ
fcis-7512	84	28	data	data	NOUN
fcis-7512	84	29	representation	representation	NOUN
fcis-7512	84	30	contains	contain	VERB
fcis-7512	84	31	the	the	DET
fcis-7512	84	32	semantic	semantic	ADJ
fcis-7512	84	33	relevance	relevance	NOUN
fcis-7512	84	34	weight	weight	NOUN
fcis-7512	84	35	information	information	NOUN
fcis-7512	84	36	of	of	ADP
fcis-7512	84	37	text	text	NOUN
fcis-7512	84	38	data	datum	NOUN
fcis-7512	84	39	,	,	PUNCT
fcis-7512	84	40	audio	audio	ADJ
fcis-7512	84	41	data	datum	NOUN
fcis-7512	84	42	and	and	CCONJ
fcis-7512	84	43	video	video	NOUN
fcis-7512	84	44	data	datum	NOUN
fcis-7512	84	45	,	,	PUNCT
fcis-7512	84	46	and	and	CCONJ
fcis-7512	84	47	two	two	NUM
fcis-7512	84	48	data	datum	NOUN
fcis-7512	84	49	representations	representation	NOUN
fcis-7512	84	50	corresponding	correspond	VERB
fcis-7512	84	51	to	to	ADP
fcis-7512	84	52	video	video	NOUN
fcis-7512	84	53	and	and	CCONJ
fcis-7512	84	54	audio	audio	NOUN
fcis-7512	84	55	text	text	NOUN
fcis-7512	84	56	are	be	AUX
fcis-7512	84	57	obtained	obtain	VERB
fcis-7512	84	58	.	.	PUNCT
fcis-7512	85	1	𝐻tv	𝐻tv	PROPN
fcis-7512	85	2	=	=	SYM
fcis-7512	85	3	𝐴1	𝐴1	PROPN
fcis-7512	85	4	⋅	⋅	PROPN
fcis-7512	85	5	ℎt	ℎt	X
fcis-7512	85	6	.	.	PUNCT
fcis-7512	86	1	(	(	PUNCT
fcis-7512	86	2	10	10	NUM
fcis-7512	86	3	)	)	PUNCT
fcis-7512	86	4	𝐻ta	𝐻ta	PROPN
fcis-7512	86	5	=	=	PROPN
fcis-7512	86	6	𝐴2	𝐴2	PROPN
fcis-7512	86	7	⋅	⋅	PROPN
fcis-7512	86	8	ℎt	ℎt	X
fcis-7512	86	9	.	.	PUNCT
fcis-7512	87	1	(	(	PUNCT
fcis-7512	87	2	11	11	NUM
fcis-7512	87	3	)	)	PUNCT
fcis-7512	87	4	in	in	ADP
fcis-7512	87	5	order	order	NOUN
fcis-7512	87	6	to	to	PART
fcis-7512	87	7	make	make	VERB
fcis-7512	87	8	the	the	DET
fcis-7512	87	9	text	text	NOUN
fcis-7512	87	10	data	datum	NOUN
fcis-7512	87	11	representation	representation	NOUN
fcis-7512	87	12	contain	contain	VERB
fcis-7512	87	13	more	more	ADJ
fcis-7512	87	14	intermodal	intermodal	ADJ
fcis-7512	87	15	information	information	NOUN
fcis-7512	87	16	,	,	PUNCT
fcis-7512	87	17	the	the	DET
fcis-7512	87	18	two	two	NUM
fcis-7512	87	19	data	datum	NOUN
fcis-7512	87	20	representations	representation	NOUN
fcis-7512	87	21	are	be	AUX
fcis-7512	87	22	fused	fuse	VERB
fcis-7512	87	23	in	in	ADP
fcis-7512	87	24	the	the	DET
fcis-7512	87	25	way	way	NOUN
fcis-7512	87	26	of	of	ADP
fcis-7512	87	27	concatenation	concatenation	NOUN
fcis-7512	87	28	,	,	PUNCT
fcis-7512	87	29	and	and	CCONJ
fcis-7512	87	30	the	the	DET
fcis-7512	87	31	new	new	ADJ
fcis-7512	87	32	text	text	NOUN
fcis-7512	87	33	data	datum	NOUN
fcis-7512	87	34	representation	representation	NOUN
fcis-7512	87	35	𝐻t	𝐻t	ADP
fcis-7512	87	36	is	be	AUX
fcis-7512	87	37	obtained	obtain	VERB
fcis-7512	87	38	.	.	PUNCT
fcis-7512	88	1	𝐻t	𝐻t	ADP
fcis-7512	88	2	=	=	SYM
fcis-7512	88	3	𝐻tv	𝐻tv	PROPN
fcis-7512	88	4	+	+	CCONJ
fcis-7512	88	5	𝐻ta	𝐻ta	PROPN
fcis-7512	88	6	.	.	PUNCT
fcis-7512	89	1	(	(	PUNCT
fcis-7512	89	2	12	12	NUM
fcis-7512	89	3	)	)	PUNCT
fcis-7512	89	4	similarly	similarly	ADV
fcis-7512	89	5	,	,	PUNCT
fcis-7512	89	6	for	for	SCONJ
fcis-7512	89	7	audio	audio	NOUN
fcis-7512	89	8	and	and	CCONJ
fcis-7512	89	9	video	video	NOUN
fcis-7512	89	10	data	datum	NOUN
fcis-7512	89	11	and	and	CCONJ
fcis-7512	89	12	text	text	NOUN
fcis-7512	89	13	data	datum	NOUN
fcis-7512	89	14	processing	process	VERB
fcis-7512	89	15	the	the	DET
fcis-7512	89	16	same	same	ADJ
fcis-7512	89	17	way	way	NOUN
fcis-7512	89	18	,	,	PUNCT
fcis-7512	89	19	respectively	respectively	ADV
fcis-7512	89	20	by	by	ADP
fcis-7512	89	21	computing	compute	VERB
fcis-7512	89	22	the	the	DET
fcis-7512	89	23	semantic	semantic	ADJ
fcis-7512	89	24	correlation	correlation	NOUN
fcis-7512	89	25	matrix	matrix	NOUN
fcis-7512	89	26	and	and	CCONJ
fcis-7512	89	27	weight	weight	NOUN
fcis-7512	89	28	matrix	matrix	NOUN
fcis-7512	89	29	,	,	PUNCT
fcis-7512	89	30	attention	attention	NOUN
fcis-7512	89	31	to	to	PART
fcis-7512	89	32	get	get	VERB
fcis-7512	89	33	new	new	ADJ
fcis-7512	89	34	data	datum	NOUN
fcis-7512	89	35	respectively	respectively	ADV
fcis-7512	89	36	said	say	VERB
fcis-7512	89	37	𝐻a	𝐻a	NOUN
fcis-7512	89	38	and	and	CCONJ
fcis-7512	89	39	𝐻a	𝐻a	PROPN
fcis-7512	89	40	,	,	PUNCT
fcis-7512	89	41	finally	finally	ADV
fcis-7512	89	42	said	say	VERB
fcis-7512	89	43	the	the	DET
fcis-7512	89	44	new	new	ADJ
fcis-7512	89	45	data	datum	NOUN
fcis-7512	89	46	for	for	ADP
fcis-7512	89	47	main	main	ADJ
fcis-7512	89	48	task	task	NOUN
fcis-7512	89	49	.	.	PUNCT
fcis-7512	90	1	2.6	2.6	NUM
fcis-7512	90	2	.	.	PUNCT
fcis-7512	91	1	multi	multi	ADJ
fcis-7512	91	2	-	-	ADJ
fcis-7512	91	3	mission	mission	ADJ
fcis-7512	91	4	joint	joint	ADJ
fcis-7512	91	5	training	training	NOUN
fcis-7512	91	6	for	for	ADP
fcis-7512	91	7	multi	multi	ADJ
fcis-7512	91	8	-	-	ADJ
fcis-7512	91	9	task	task	ADJ
fcis-7512	91	10	joint	joint	ADJ
fcis-7512	91	11	training	training	NOUN
fcis-7512	91	12	,	,	PUNCT
fcis-7512	91	13	multi	multi	ADJ
fcis-7512	91	14	-	-	ADJ
fcis-7512	91	15	modal	modal	ADJ
fcis-7512	91	16	sentiment	sentiment	NOUN
fcis-7512	91	17	analysis	analysis	NOUN
fcis-7512	91	18	is	be	AUX
fcis-7512	91	19	taken	take	VERB
fcis-7512	91	20	as	as	ADP
fcis-7512	91	21	the	the	DET
fcis-7512	91	22	main	main	ADJ
fcis-7512	91	23	task	task	NOUN
fcis-7512	91	24	,	,	PUNCT
fcis-7512	91	25	single	single	ADJ
fcis-7512	91	26	-	-	PUNCT
fcis-7512	91	27	modal	modal	ADJ
fcis-7512	91	28	sentiment	sentiment	NOUN
fcis-7512	91	29	analysis	analysis	NOUN
fcis-7512	91	30	as	as	ADP
fcis-7512	91	31	the	the	DET
fcis-7512	91	32	auxiliary	auxiliary	ADJ
fcis-7512	91	33	task	task	NOUN
fcis-7512	91	34	,	,	PUNCT
fcis-7512	91	35	and	and	CCONJ
fcis-7512	91	36	the	the	DET
fcis-7512	91	37	auxiliary	auxiliary	ADJ
fcis-7512	91	38	task	task	NOUN
fcis-7512	91	39	only	only	ADV
fcis-7512	91	40	exists	exist	VERB
fcis-7512	91	41	in	in	ADP
fcis-7512	91	42	the	the	DET
fcis-7512	91	43	training	training	NOUN
fcis-7512	91	44	stage	stage	NOUN
fcis-7512	91	45	.	.	PUNCT
fcis-7512	92	1	the	the	DET
fcis-7512	92	2	main	main	ADJ
fcis-7512	92	3	task	task	NOUN
fcis-7512	92	4	and	and	CCONJ
fcis-7512	92	5	the	the	DET
fcis-7512	92	6	auxiliary	auxiliary	ADJ
fcis-7512	92	7	task	task	NOUN
fcis-7512	92	8	share	share	VERB
fcis-7512	92	9	the	the	DET
fcis-7512	92	10	underlying	underlie	VERB
fcis-7512	92	11	data	datum	NOUN
fcis-7512	92	12	representation	representation	NOUN
fcis-7512	92	13	part	part	NOUN
fcis-7512	92	14	,	,	PUNCT
fcis-7512	92	15	and	and	CCONJ
fcis-7512	92	16	the	the	DET
fcis-7512	92	17	loss	loss	NOUN
fcis-7512	92	18	of	of	ADP
fcis-7512	92	19	the	the	DET
fcis-7512	92	20	main	main	ADJ
fcis-7512	92	21	task	task	NOUN
fcis-7512	92	22	and	and	CCONJ
fcis-7512	92	23	the	the	DET
fcis-7512	92	24	loss	loss	NOUN
fcis-7512	92	25	of	of	ADP
fcis-7512	92	26	the	the	DET
fcis-7512	92	27	auxiliary	auxiliary	ADJ
fcis-7512	92	28	task	task	NOUN
fcis-7512	92	29	are	be	AUX
fcis-7512	92	30	taken	take	VERB
fcis-7512	92	31	as	as	ADP
fcis-7512	92	32	the	the	DET
fcis-7512	92	33	total	total	ADJ
fcis-7512	92	34	loss	loss	NOUN
fcis-7512	92	35	of	of	ADP
fcis-7512	92	36	the	the	DET
fcis-7512	92	37	model	model	NOUN
fcis-7512	92	38	to	to	PART
fcis-7512	92	39	optimize	optimize	VERB
fcis-7512	92	40	together	together	ADV
fcis-7512	92	41	.	.	PUNCT
fcis-7512	93	1	as	as	ADP
fcis-7512	93	2	for	for	ADP
fcis-7512	93	3	the	the	DET
fcis-7512	93	4	chief	chief	ADJ
fcis-7512	93	5	officer	officer	NOUN
fcis-7512	93	6	,	,	PUNCT
fcis-7512	93	7	after	after	ADP
fcis-7512	93	8	obtaining	obtain	VERB
fcis-7512	93	9	the	the	DET
fcis-7512	93	10	single	single	ADJ
fcis-7512	93	11	mode	mode	NOUN
fcis-7512	93	12	data	datum	NOUN
fcis-7512	93	13	representation	representation	PROPN
fcis-7512	93	14	ℎ𝑛	ℎ𝑛	PROPN
fcis-7512	93	15	,	,	PUNCT
fcis-7512	93	16	due	due	ADP
fcis-7512	93	17	to	to	ADP
fcis-7512	93	18	the	the	DET
fcis-7512	93	19	different	different	ADJ
fcis-7512	93	20	dimension	dimension	NOUN
fcis-7512	93	21	of	of	ADP
fcis-7512	93	22	feature	feature	NOUN
fcis-7512	93	23	vectors	vector	NOUN
fcis-7512	93	24	extracted	extract	VERB
fcis-7512	93	25	from	from	ADP
fcis-7512	93	26	the	the	DET
fcis-7512	93	27	multi	multi	ADJ
fcis-7512	93	28	-	-	ADJ
fcis-7512	93	29	mode	mode	ADJ
fcis-7512	93	30	data	datum	NOUN
fcis-7512	93	31	,	,	PUNCT
fcis-7512	93	32	this	this	DET
fcis-7512	93	33	paper	paper	NOUN
fcis-7512	93	34	puts	put	VERB
fcis-7512	93	35	three	three	NUM
fcis-7512	93	36	multi	multi	ADJ
fcis-7512	93	37	-	-	ADJ
fcis-7512	93	38	mode	mode	ADJ
fcis-7512	93	39	feature	feature	NOUN
fcis-7512	93	40	vectors	vector	NOUN
fcis-7512	93	41	through	through	ADP
fcis-7512	93	42	the	the	DET
fcis-7512	93	43	full	full	ADJ
fcis-7512	93	44	connection	connection	NOUN
fcis-7512	93	45	layer	layer	NOUN
fcis-7512	93	46	to	to	PART
fcis-7512	93	47	avoid	avoid	VERB
fcis-7512	93	48	the	the	DET
fcis-7512	93	49	influence	influence	NOUN
fcis-7512	93	50	of	of	ADP
fcis-7512	93	51	spatial	spatial	ADJ
fcis-7512	93	52	features	feature	NOUN
fcis-7512	93	53	of	of	ADP
fcis-7512	93	54	vectors	vector	NOUN
fcis-7512	93	55	on	on	ADP
fcis-7512	93	56	the	the	DET
fcis-7512	93	57	calculation	calculation	NOUN
fcis-7512	93	58	,	,	PUNCT
fcis-7512	93	59	and	and	CCONJ
fcis-7512	93	60	then	then	ADV
fcis-7512	93	61	makes	make	VERB
fcis-7512	93	62	use	use	NOUN
fcis-7512	93	63	of	of	ADP
fcis-7512	93	64	semantic	semantic	ADJ
fcis-7512	93	65	correlation	correlation	NOUN
fcis-7512	93	66	between	between	ADP
fcis-7512	93	67	modes	mode	NOUN
fcis-7512	93	68	to	to	PART
fcis-7512	93	69	make	make	VERB
fcis-7512	93	70	inter	inter	ADJ
fcis-7512	93	71	-	-	ADJ
fcis-7512	93	72	modal	modal	ADJ
fcis-7512	93	73	attention	attention	NOUN
fcis-7512	93	74	mechanism	mechanism	NOUN
fcis-7512	93	75	for	for	ADP
fcis-7512	93	76	the	the	DET
fcis-7512	93	77	three	three	NUM
fcis-7512	93	78	modal	modal	ADJ
fcis-7512	93	79	data	datum	NOUN
fcis-7512	93	80	respectively	respectively	ADV
fcis-7512	93	81	.	.	PUNCT
fcis-7512	94	1	the	the	DET
fcis-7512	94	2	new	new	ADJ
fcis-7512	94	3	data	datum	NOUN
fcis-7512	94	4	representation	representation	NOUN
fcis-7512	94	5	𝐻𝑛	𝐻𝑛	PROPN
fcis-7512	94	6	of	of	ADP
fcis-7512	94	7	three	three	NUM
fcis-7512	94	8	modes	mode	NOUN
fcis-7512	94	9	is	be	AUX
fcis-7512	94	10	obtained	obtain	VERB
fcis-7512	94	11	,	,	PUNCT
fcis-7512	94	12	where	where	SCONJ
fcis-7512	94	13	𝑛	𝑛	DET
fcis-7512	94	14	∈	∈	PROPN
fcis-7512	94	15	{	{	PUNCT
fcis-7512	94	16	a	a	PRON
fcis-7512	94	17	,	,	PUNCT
fcis-7512	94	18	v	v	NOUN
fcis-7512	94	19	,	,	PUNCT
fcis-7512	94	20	t	t	PROPN
fcis-7512	94	21	}	}	PUNCT
fcis-7512	94	22	,	,	PUNCT
fcis-7512	94	23	and	and	CCONJ
fcis-7512	94	24	the	the	DET
fcis-7512	94	25	new	new	ADJ
fcis-7512	94	26	single	single	ADJ
fcis-7512	94	27	-	-	PUNCT
fcis-7512	94	28	mode	mode	NOUN
fcis-7512	94	29	data	datum	NOUN
fcis-7512	94	30	representation	representation	NOUN
fcis-7512	94	31	is	be	AUX
fcis-7512	94	32	connected	connect	VERB
fcis-7512	94	33	with	with	ADP
fcis-7512	94	34	the	the	DET
fcis-7512	94	35	original	original	ADJ
fcis-7512	94	36	data	datum	NOUN
fcis-7512	94	37	representation	representation	NOUN
fcis-7512	94	38	to	to	PART
fcis-7512	94	39	achieve	achieve	VERB
fcis-7512	94	40	multi	multi	ADJ
fcis-7512	94	41	-	-	ADJ
fcis-7512	94	42	mode	mode	ADJ
fcis-7512	94	43	feature	feature	NOUN
fcis-7512	94	44	fusion	fusion	NOUN
fcis-7512	94	45	to	to	PART
fcis-7512	94	46	get𝐹.	get𝐹.	VERB
fcis-7512	94	47	𝐹	𝐹	PROPN
fcis-7512	94	48	=	=	PUNCT
fcis-7512	95	1	𝐻a	𝐻a	PROPN
fcis-7512	96	1	+	+	CCONJ
fcis-7512	97	1	𝐻t	𝐻t	ADP
fcis-7512	97	2	+	+	CCONJ
fcis-7512	97	3	𝐻v	𝐻v	PROPN
fcis-7512	97	4	+	+	CCONJ
fcis-7512	97	5	ℎt	ℎt	VERB
fcis-7512	98	1	+	+	CCONJ
fcis-7512	98	2	ℎa	ℎa	NOUN
fcis-7512	98	3	+	+	CCONJ
fcis-7512	98	4	ℎv	ℎv	PROPN
fcis-7512	98	5	(	(	PUNCT
fcis-7512	98	6	13	13	NUM
fcis-7512	98	7	)	)	PUNCT
fcis-7512	98	8	the	the	DET
fcis-7512	98	9	fused	fuse	VERB
fcis-7512	98	10	feature	feature	NOUN
fcis-7512	98	11	vector	vector	NOUN
fcis-7512	98	12	f	f	PROPN
fcis-7512	98	13	contains	contain	VERB
fcis-7512	98	14	two	two	NUM
fcis-7512	98	15	parts	part	NOUN
fcis-7512	98	16	of	of	ADP
fcis-7512	98	17	information	information	NOUN
fcis-7512	98	18	:	:	PUNCT
fcis-7512	98	19	the	the	DET
fcis-7512	98	20	first	first	ADJ
fcis-7512	98	21	is	be	AUX
fcis-7512	98	22	the	the	DET
fcis-7512	98	23	original	original	ADJ
fcis-7512	98	24	data	datum	NOUN
fcis-7512	98	25	representation	representation	NOUN
fcis-7512	98	26	,	,	PUNCT
fcis-7512	98	27	which	which	PRON
fcis-7512	98	28	contains	contain	VERB
fcis-7512	98	29	the	the	DET
fcis-7512	98	30	context	context	NOUN
fcis-7512	98	31	information	information	NOUN
fcis-7512	98	32	of	of	ADP
fcis-7512	98	33	the	the	DET
fcis-7512	98	34	three	three	NUM
fcis-7512	98	35	modes	mode	NOUN
fcis-7512	98	36	after	after	ADP
fcis-7512	98	37	the	the	DET
fcis-7512	98	38	time	time	NOUN
fcis-7512	98	39	series	series	PROPN
fcis-7512	98	40	modeling	modeling	NOUN
fcis-7512	98	41	of	of	ADP
fcis-7512	98	42	the	the	DET
fcis-7512	98	43	three	three	NUM
fcis-7512	98	44	-	-	PUNCT
fcis-7512	98	45	modal	modal	NOUN
fcis-7512	98	46	data	datum	NOUN
fcis-7512	98	47	by	by	ADP
fcis-7512	98	48	using	use	VERB
fcis-7512	98	49	the	the	DET
fcis-7512	98	50	bidirectional	bidirectional	ADJ
fcis-7512	98	51	lstm	lstm	ADJ
fcis-7512	98	52	neural	neural	ADJ
fcis-7512	98	53	network	network	NOUN
fcis-7512	98	54	.	.	PUNCT
fcis-7512	99	1	on	on	ADP
fcis-7512	99	2	the	the	DET
fcis-7512	99	3	other	other	ADJ
fcis-7512	99	4	hand	hand	NOUN
fcis-7512	99	5	,	,	PUNCT
fcis-7512	99	6	using	use	VERB
fcis-7512	99	7	the	the	DET
fcis-7512	99	8	intermodal	intermodal	ADJ
fcis-7512	99	9	attention	attention	NOUN
fcis-7512	99	10	mechanism	mechanism	NOUN
fcis-7512	99	11	,	,	PUNCT
fcis-7512	99	12	the	the	DET
fcis-7512	99	13	new	new	ADJ
fcis-7512	99	14	data	data	NOUN
fcis-7512	99	15	representation	representation	NOUN
fcis-7512	99	16	includes	include	VERB
fcis-7512	99	17	semantic	semantic	ADJ
fcis-7512	99	18	correlations	correlation	NOUN
fcis-7512	99	19	between	between	ADP
fcis-7512	99	20	the	the	DET
fcis-7512	99	21	three	three	NUM
fcis-7512	99	22	modes	mode	NOUN
fcis-7512	99	23	.	.	PUNCT
fcis-7512	100	1	therefore	therefore	ADV
fcis-7512	100	2	,	,	PUNCT
fcis-7512	100	3	the	the	DET
fcis-7512	100	4	feature	feature	NOUN
fcis-7512	100	5	vector	vector	NOUN
fcis-7512	100	6	f	f	PROPN
fcis-7512	100	7	can	can	AUX
fcis-7512	100	8	contain	contain	VERB
fcis-7512	100	9	more	more	ADJ
fcis-7512	100	10	information	information	NOUN
fcis-7512	100	11	and	and	CCONJ
fcis-7512	100	12	reduce	reduce	VERB
fcis-7512	100	13	noise	noise	NOUN
fcis-7512	100	14	compared	compare	VERB
fcis-7512	100	15	with	with	ADP
fcis-7512	100	16	the	the	DET
fcis-7512	100	17	previous	previous	ADJ
fcis-7512	100	18	fusion	fusion	NOUN
fcis-7512	100	19	method	method	NOUN
fcis-7512	100	20	.	.	PUNCT
fcis-7512	101	1	relu	relu	NOUN
fcis-7512	101	2	activation	activation	NOUN
fcis-7512	101	3	function	function	NOUN
fcis-7512	101	4	is	be	AUX
fcis-7512	101	5	sparse	sparse	ADJ
fcis-7512	101	6	,	,	PUNCT
fcis-7512	101	7	which	which	PRON
fcis-7512	101	8	enables	enable	VERB
fcis-7512	101	9	the	the	DET
fcis-7512	101	10	sparse	sparse	ADJ
fcis-7512	101	11	model	model	NOUN
fcis-7512	101	12	to	to	AUX
fcis-7512	101	13	better	well	ADV
fcis-7512	101	14	mine	mine	VERB
fcis-7512	101	15	relevant	relevant	ADJ
fcis-7512	101	16	features	feature	NOUN
fcis-7512	101	17	and	and	CCONJ
fcis-7512	101	18	fit	fit	ADJ
fcis-7512	101	19	training	training	NOUN
fcis-7512	101	20	data	datum	NOUN
fcis-7512	101	21	.	.	PUNCT
fcis-7512	102	1	therefore	therefore	ADV
fcis-7512	102	2	,	,	PUNCT
fcis-7512	102	3	relu	relu	NOUN
fcis-7512	102	4	activation	activation	NOUN
fcis-7512	102	5	function	function	NOUN
fcis-7512	102	6	is	be	AUX
fcis-7512	102	7	selected	select	VERB
fcis-7512	102	8	as	as	ADP
fcis-7512	102	9	the	the	DET
fcis-7512	102	10	activation	activation	NOUN
fcis-7512	102	11	function	function	NOUN
fcis-7512	102	12	in	in	ADP
fcis-7512	102	13	this	this	DET
fcis-7512	102	14	paper	paper	NOUN
fcis-7512	102	15	.	.	PUNCT
fcis-7512	103	1	the	the	DET
fcis-7512	103	2	fused	fuse	VERB
fcis-7512	103	3	data	data	NOUN
fcis-7512	103	4	represents	represent	VERB
fcis-7512	103	5	f	f	PROPN
fcis-7512	103	6	through	through	ADP
fcis-7512	103	7	the	the	DET
fcis-7512	103	8	relu	relu	NOUN
fcis-7512	103	9	activation	activation	NOUN
fcis-7512	103	10	function	function	VERB
fcis-7512	103	11	to	to	PART
fcis-7512	103	12	get	get	VERB
fcis-7512	103	13	𝐹∗.	𝐹∗.	NOUN
fcis-7512	103	14	𝐹∗	𝐹∗	PUNCT
fcis-7512	103	15	=	=	SYM
fcis-7512	103	16	𝑅𝑒𝑙𝑢(𝑊1	𝑅𝑒𝑙𝑢(𝑊1	PUNCT
fcis-7512	103	17	𝑇𝐹	𝑇𝐹	PROPN
fcis-7512	103	18	+	+	CCONJ
fcis-7512	103	19	𝑏1	𝑏1	ADJ
fcis-7512	103	20	)	)	PUNCT
fcis-7512	103	21	.	.	PUNCT
fcis-7512	104	1	(	(	PUNCT
fcis-7512	104	2	14	14	NUM
fcis-7512	104	3	)	)	PUNCT
fcis-7512	104	4	where	where	SCONJ
fcis-7512	104	5	𝑊1	𝑊1	PROPN
fcis-7512	104	6	∈	∈	PROPN
fcis-7512	104	7	𝑅𝑙n×2(𝑑a+𝑑v+𝑑t	𝑅𝑙n×2(𝑑a+𝑑v+𝑑t	PROPN
fcis-7512	104	8	)	)	PUNCT
fcis-7512	104	9	,	,	PUNCT
fcis-7512	104	10	𝐹∗	𝐹∗	X
fcis-7512	104	11	is	be	AUX
fcis-7512	104	12	used	use	VERB
fcis-7512	104	13	to	to	PART
fcis-7512	104	14	predict	predict	VERB
fcis-7512	104	15	the	the	DET
fcis-7512	104	16	emotional	emotional	ADJ
fcis-7512	104	17	intensity	intensity	NOUN
fcis-7512	104	18	.	.	PUNCT
fcis-7512	105	1	where	where	SCONJ
fcis-7512	105	2	𝑊2	𝑊2	PROPN
fcis-7512	105	3	∈	∈	PROPN
fcis-7512	105	4	𝑅2(𝑑a+𝑑v+𝑑t)×1	𝑅2(𝑑a+𝑑v+𝑑t)×1	NOUN
fcis-7512	105	5	.	.	PUNCT
fcis-7512	105	6	ŷm	ŷm	NOUN
fcis-7512	105	7	=	=	SYM
fcis-7512	105	8	w2	w2	PROPN
fcis-7512	105	9	tf∗	tf∗	NOUN
fcis-7512	105	10	+	+	CCONJ
fcis-7512	105	11	b2	b2	NOUN
fcis-7512	105	12	.	.	PUNCT
fcis-7512	106	1	(	(	PUNCT
fcis-7512	106	2	15	15	NUM
fcis-7512	106	3	)	)	PUNCT
fcis-7512	106	4	𝑦𝑛	𝑦𝑛	NOUN
fcis-7512	106	5	=	=	NOUN
fcis-7512	106	6	ulgm(𝑦m	ulgm(𝑦m	NOUN
fcis-7512	106	7	,	,	PUNCT
fcis-7512	106	8	𝐹∗	𝐹∗	NUM
fcis-7512	106	9	,	,	PUNCT
fcis-7512	106	10	𝐻𝑛	𝐻𝑛	NOUN
fcis-7512	106	11	)	)	PUNCT
fcis-7512	106	12	.	.	PUNCT
fcis-7512	107	1	(	(	PUNCT
fcis-7512	107	2	16	16	NUM
fcis-7512	107	3	)	)	PUNCT
fcis-7512	107	4	for	for	ADP
fcis-7512	107	5	auxiliary	auxiliary	ADJ
fcis-7512	107	6	tasks	task	NOUN
fcis-7512	107	7	,	,	PUNCT
fcis-7512	107	8	in	in	ADP
fcis-7512	107	9	order	order	NOUN
fcis-7512	107	10	to	to	PART
fcis-7512	107	11	reduce	reduce	VERB
fcis-7512	107	12	the	the	DET
fcis-7512	107	13	internal	internal	ADJ
fcis-7512	107	14	noise	noise	NOUN
fcis-7512	107	15	of	of	ADP
fcis-7512	107	16	single	single	ADJ
fcis-7512	107	17	-	-	PUNCT
fcis-7512	107	18	mode	mode	NOUN
fcis-7512	107	19	data	datum	NOUN
fcis-7512	107	20	,	,	PUNCT
fcis-7512	107	21	a	a	DET
fcis-7512	107	22	new	new	ADJ
fcis-7512	107	23	single	single	ADJ
fcis-7512	107	24	-	-	PUNCT
fcis-7512	107	25	mode	mode	NOUN
fcis-7512	107	26	data	datum	NOUN
fcis-7512	107	27	representation	representation	NOUN
fcis-7512	107	28	𝐻𝑛	𝐻𝑛	PROPN
fcis-7512	107	29	is	be	AUX
fcis-7512	107	30	obtained	obtain	VERB
fcis-7512	107	31	by	by	ADP
fcis-7512	107	32	using	use	VERB
fcis-7512	107	33	the	the	DET
fcis-7512	107	34	single	single	ADJ
fcis-7512	107	35	-	-	PUNCT
fcis-7512	107	36	mode	mode	NOUN
fcis-7512	107	37	self	self	NOUN
fcis-7512	107	38	-	-	PUNCT
fcis-7512	107	39	attention	attention	NOUN
fcis-7512	107	40	mechanism	mechanism	NOUN
fcis-7512	107	41	.	.	PUNCT
fcis-7512	108	1	similarly	similarly	ADV
fcis-7512	108	2	,	,	PUNCT
fcis-7512	108	3	relu	relu	NOUN
fcis-7512	108	4	activation	activation	NOUN
fcis-7512	108	5	function	function	NOUN
fcis-7512	108	6	is	be	AUX
fcis-7512	108	7	used	use	VERB
fcis-7512	108	8	in	in	ADP
fcis-7512	108	9	this	this	DET
fcis-7512	108	10	paper	paper	NOUN
fcis-7512	108	11	to	to	PART
fcis-7512	108	12	make	make	VERB
fcis-7512	108	13	the	the	DET
fcis-7512	108	14	single	single	ADJ
fcis-7512	108	15	-	-	PUNCT
fcis-7512	108	16	mode	mode	NOUN
fcis-7512	108	17	data	datum	NOUN
fcis-7512	108	18	representation	representation	NOUN
fcis-7512	108	19	non	non	ADJ
fcis-7512	108	20	-	-	ADJ
fcis-7512	108	21	linear	linear	ADJ
fcis-7512	108	22	.	.	PUNCT
fcis-7512	109	1	the	the	DET
fcis-7512	109	2	results	result	NOUN
fcis-7512	109	3	of	of	ADP
fcis-7512	109	4	activation	activation	NOUN
fcis-7512	109	5	function	function	NOUN
fcis-7512	109	6	are	be	AUX
fcis-7512	109	7	used	use	VERB
fcis-7512	109	8	to	to	PART
fcis-7512	109	9	predict	predict	VERB
fcis-7512	109	10	emotion	emotion	NOUN
fcis-7512	109	11	.	.	PUNCT
fcis-7512	110	1	�	�	PROPN
fcis-7512	110	2	̂	̂	X
fcis-7512	110	3	�	�	NOUN
fcis-7512	110	4	𝑛	𝑛	PRON
fcis-7512	110	5	=	=	PUNCT
fcis-7512	110	6	𝑊2	𝑊2	PROPN
fcis-7512	110	7	𝑇	𝑇	PROPN
fcis-7512	110	8	·	·	PUNCT
fcis-7512	111	1	[	[	X
fcis-7512	111	2	relu(𝑊1	relu(𝑊1	X
fcis-7512	111	3	𝑇ℎ𝑛	𝑇ℎ𝑛	PROPN
fcis-7512	111	4	+	+	CCONJ
fcis-7512	111	5	𝑏1)]+𝑏2	𝑏1)]+𝑏2	PROPN
fcis-7512	111	6	)	)	PUNCT
fcis-7512	111	7	,	,	PUNCT
fcis-7512	111	8	𝑊1	𝑊1	PROPN
fcis-7512	111	9	∈	∈	PROPN
fcis-7512	111	10	𝑅𝑙𝑛×ⅆ𝑛	𝑅𝑙𝑛×ⅆ𝑛	PROPN
fcis-7512	111	11	、	、	PROPN
fcis-7512	111	12	𝑊2	𝑊2	PROPN
fcis-7512	111	13	∈	∈	PROPN
fcis-7512	111	14	𝑅ⅆ𝑛×1	𝑅ⅆ𝑛×1	NOUN
fcis-7512	111	15	where	where	SCONJ
fcis-7512	111	16	𝑛	𝑛	DET
fcis-7512	111	17	∈	∈	PROPN
fcis-7512	111	18	{	{	PUNCT
fcis-7512	111	19	a	a	PRON
fcis-7512	111	20	,	,	PUNCT
fcis-7512	111	21	v	v	NOUN
fcis-7512	111	22	,	,	PUNCT
fcis-7512	111	23	t	t	PROPN
fcis-7512	111	24	}	}	PUNCT
fcis-7512	111	25	,	,	PUNCT
fcis-7512	111	26	according	accord	VERB
fcis-7512	111	27	to	to	ADP
fcis-7512	111	28	the	the	DET
fcis-7512	111	29	yu	yu	PROPN
fcis-7512	111	30	's	's	PART
fcis-7512	111	31	work	work	NOUN
fcis-7512	111	32	before	before	ADP
fcis-7512	111	33	[	[	X
fcis-7512	111	34	13	13	NUM
fcis-7512	111	35	]	]	PUNCT
fcis-7512	111	36	,	,	PUNCT
fcis-7512	111	37	the	the	DET
fcis-7512	111	38	self	self	NOUN
fcis-7512	111	39	generated	generate	VERB
fcis-7512	111	40	label	label	NOUN
fcis-7512	111	41	method	method	NOUN
fcis-7512	111	42	is	be	AUX
fcis-7512	111	43	used	use	VERB
fcis-7512	111	44	to	to	PART
fcis-7512	111	45	generate	generate	VERB
fcis-7512	111	46	single	single	ADJ
fcis-7512	111	47	mode	mode	NOUN
fcis-7512	111	48	label	label	NOUN
fcis-7512	111	49	.	.	PUNCT
fcis-7512	112	1	70	70	NUM
fcis-7512	112	2	2.7	2.7	NUM
fcis-7512	112	3	.	.	PUNCT
fcis-7512	113	1	weight	weight	NOUN
fcis-7512	113	2	loss	loss	NOUN
fcis-7512	113	3	allocation	allocation	NOUN
fcis-7512	113	4	finally	finally	ADV
fcis-7512	113	5	,	,	PUNCT
fcis-7512	113	6	multi	multi	ADJ
fcis-7512	113	7	-	-	ADJ
fcis-7512	113	8	modal	modal	ADJ
fcis-7512	113	9	sentiment	sentiment	NOUN
fcis-7512	113	10	analysis	analysis	NOUN
fcis-7512	113	11	is	be	AUX
fcis-7512	113	12	trained	train	VERB
fcis-7512	113	13	jointly	jointly	ADV
fcis-7512	113	14	with	with	ADP
fcis-7512	113	15	three	three	NUM
fcis-7512	113	16	single	single	ADJ
fcis-7512	113	17	-	-	PUNCT
fcis-7512	113	18	modal	modal	ADJ
fcis-7512	113	19	sentiment	sentiment	NOUN
fcis-7512	113	20	analysis	analysis	NOUN
fcis-7512	113	21	.	.	PUNCT
fcis-7512	114	1	the	the	DET
fcis-7512	114	2	absolute	absolute	ADJ
fcis-7512	114	3	value	value	NOUN
fcis-7512	114	4	of	of	ADP
fcis-7512	114	5	the	the	DET
fcis-7512	114	6	difference	difference	NOUN
fcis-7512	114	7	between	between	ADP
fcis-7512	114	8	the	the	DET
fcis-7512	114	9	label	label	NOUN
fcis-7512	114	10	generated	generate	VERB
fcis-7512	114	11	by	by	ADP
fcis-7512	114	12	single	single	ADJ
fcis-7512	114	13	mode	mode	NOUN
fcis-7512	114	14	and	and	CCONJ
fcis-7512	114	15	the	the	DET
fcis-7512	114	16	predicted	predict	VERB
fcis-7512	114	17	label	label	NOUN
fcis-7512	114	18	is	be	AUX
fcis-7512	114	19	taken	take	VERB
fcis-7512	114	20	as	as	ADP
fcis-7512	114	21	the	the	DET
fcis-7512	114	22	loss	loss	NOUN
fcis-7512	114	23	value	value	NOUN
fcis-7512	114	24	of	of	ADP
fcis-7512	114	25	single	single	ADJ
fcis-7512	114	26	mode	mode	NOUN
fcis-7512	114	27	,	,	PUNCT
fcis-7512	114	28	and	and	CCONJ
fcis-7512	114	29	the	the	DET
fcis-7512	114	30	loss	loss	NOUN
fcis-7512	114	31	of	of	ADP
fcis-7512	114	32	single	single	ADJ
fcis-7512	114	33	-	-	PUNCT
fcis-7512	114	34	modal	modal	ADJ
fcis-7512	114	35	sentiment	sentiment	NOUN
fcis-7512	114	36	analysis	analysis	NOUN
fcis-7512	114	37	subtask	subtask	NOUN
fcis-7512	114	38	is	be	AUX
fcis-7512	114	39	assigned	assign	VERB
fcis-7512	114	40	a	a	DET
fcis-7512	114	41	weight	weight	NOUN
fcis-7512	114	42	𝑤𝑆	𝑤𝑆	PROPN
fcis-7512	114	43	𝑖	𝑖	SYM
fcis-7512	114	44	,	,	PUNCT
fcis-7512	114	45	where	where	SCONJ
fcis-7512	114	46	𝑠	𝑠	PROPN
fcis-7512	114	47	∈	∈	PROPN
fcis-7512	114	48	{	{	PUNCT
fcis-7512	114	49	a	a	PRON
fcis-7512	114	50	,	,	PUNCT
fcis-7512	114	51	v	v	NOUN
fcis-7512	114	52	,	,	PUNCT
fcis-7512	114	53	t	t	PROPN
fcis-7512	114	54	}	}	PUNCT
fcis-7512	114	55	,	,	PUNCT
fcis-7512	114	56	𝑤𝑆	𝑤𝑆	PROPN
fcis-7512	114	57	𝑖	𝑖	SYM
fcis-7512	114	58	=	=	SYM
fcis-7512	114	59	𝑡𝑎𝑛ℎ(|𝑦𝑠	𝑡𝑎𝑛ℎ(|𝑦𝑠	ADJ
fcis-7512	114	60	(	(	PUNCT
fcis-7512	114	61	𝑖	𝑖	NOUN
fcis-7512	114	62	)	)	PUNCT
fcis-7512	114	63	−	−	PROPN
fcis-7512	114	64	𝑦m|	𝑦m|	NOUN
fcis-7512	114	65	)	)	PUNCT
fcis-7512	114	66	said	say	VERB
fcis-7512	114	67	the	the	DET
fcis-7512	114	68	subtasks	subtask	NOUN
fcis-7512	114	69	s	s	PART
fcis-7512	114	70	first	first	ADJ
fcis-7512	114	71	i	i	PRON
fcis-7512	114	72	samples	sample	VERB
fcis-7512	114	73	,	,	PUNCT
fcis-7512	114	74	when	when	SCONJ
fcis-7512	114	75	the	the	DET
fcis-7512	114	76	tag	tag	NOUN
fcis-7512	114	77	model	model	NOUN
fcis-7512	114	78	focuses	focus	VERB
fcis-7512	114	79	more	more	ADJ
fcis-7512	114	80	on	on	ADP
fcis-7512	114	81	multimodal	multimodal	ADJ
fcis-7512	114	82	emotional	emotional	ADJ
fcis-7512	114	83	labels	label	NOUN
fcis-7512	114	84	are	be	AUX
fcis-7512	114	85	different	different	ADJ
fcis-7512	114	86	parts	part	NOUN
fcis-7512	114	87	.	.	PUNCT
fcis-7512	115	1	the	the	DET
fcis-7512	115	2	loss	loss	NOUN
fcis-7512	115	3	of	of	ADP
fcis-7512	115	4	the	the	DET
fcis-7512	115	5	single	single	ADJ
fcis-7512	115	6	-	-	PUNCT
fcis-7512	115	7	mode	mode	NOUN
fcis-7512	115	8	sentiment	sentiment	NOUN
fcis-7512	115	9	analysis	analysis	NOUN
fcis-7512	115	10	subtask	subtask	NOUN
fcis-7512	115	11	is	be	AUX
fcis-7512	115	12	:	:	PUNCT
fcis-7512	115	13	𝐿𝑠	𝐿𝑠	PROPN
fcis-7512	115	14	=	=	SYM
fcis-7512	115	15	𝑤𝑆	𝑤𝑆	PROPN
fcis-7512	115	16	𝑖	𝑖	SYM
fcis-7512	115	17	×	×	PROPN
fcis-7512	115	18	|	|	ADV
fcis-7512	115	19	�	�	NOUN
fcis-7512	115	20	̂	̂	NOUN
fcis-7512	115	21	�	�	PROPN
fcis-7512	115	22	𝑠	𝑠	PROPN
fcis-7512	115	23	(	(	PUNCT
fcis-7512	115	24	𝑖	𝑖	NOUN
fcis-7512	115	25	)	)	PUNCT
fcis-7512	115	26	−	−	PROPN
fcis-7512	115	27	𝑦𝑠	𝑦𝑠	INTJ
fcis-7512	115	28	(	(	PUNCT
fcis-7512	115	29	𝑖	𝑖	NOUN
fcis-7512	115	30	)	)	PUNCT
fcis-7512	115	31	|	|	ADV
fcis-7512	115	32	.	.	PUNCT
fcis-7512	116	1	(	(	PUNCT
fcis-7512	116	2	17	17	NUM
fcis-7512	116	3	)	)	PUNCT
fcis-7512	116	4	for	for	ADP
fcis-7512	116	5	the	the	DET
fcis-7512	116	6	loss	loss	NOUN
fcis-7512	116	7	of	of	ADP
fcis-7512	116	8	multimodal	multimodal	ADJ
fcis-7512	116	9	sentiment	sentiment	NOUN
fcis-7512	116	10	analysis	analysis	NOUN
fcis-7512	116	11	of	of	ADP
fcis-7512	116	12	the	the	DET
fcis-7512	116	13	main	main	ADJ
fcis-7512	116	14	task	task	NOUN
fcis-7512	116	15	,	,	PUNCT
fcis-7512	116	16	this	this	DET
fcis-7512	116	17	paper	paper	NOUN
fcis-7512	116	18	adopts	adopt	VERB
fcis-7512	116	19	the	the	DET
fcis-7512	116	20	mean	mean	ADJ
fcis-7512	116	21	value	value	NOUN
fcis-7512	116	22	of	of	ADP
fcis-7512	116	23	the	the	DET
fcis-7512	116	24	total	total	ADJ
fcis-7512	116	25	loss	loss	NOUN
fcis-7512	116	26	,	,	PUNCT
fcis-7512	116	27	where	where	SCONJ
fcis-7512	116	28	n	n	PRON
fcis-7512	116	29	represents	represent	VERB
fcis-7512	116	30	the	the	DET
fcis-7512	116	31	total	total	ADJ
fcis-7512	116	32	number	number	NOUN
fcis-7512	116	33	of	of	ADP
fcis-7512	116	34	training	training	NOUN
fcis-7512	116	35	samples	sample	NOUN
fcis-7512	116	36	.	.	PUNCT
fcis-7512	117	1	𝐿m	𝐿m	PROPN
fcis-7512	117	2	=	=	NOUN
fcis-7512	117	3	1	1	NUM
fcis-7512	117	4	𝑁	𝑁	PROPN
fcis-7512	117	5	×	×	NOUN
fcis-7512	117	6	∑	∑	PUNCT
fcis-7512	117	7	|	|	ADV
fcis-7512	117	8	�	�	PROPN
fcis-7512	117	9	̂	̂	NOUN
fcis-7512	117	10	�	�	NOUN
fcis-7512	117	11	m	m	NOUN
fcis-7512	117	12	𝑖	𝑖	NOUN
fcis-7512	117	13	−	−	PROPN
fcis-7512	117	14	𝑦m	𝑦m	NOUN
fcis-7512	117	15	𝑖	𝑖	PRON
fcis-7512	117	16	|	|	NOUN
fcis-7512	117	17	𝑛	𝑛	PRON
fcis-7512	117	18	𝑖	𝑖	X
fcis-7512	117	19	.	.	PUNCT
fcis-7512	118	1	(	(	PUNCT
fcis-7512	118	2	18	18	NUM
fcis-7512	118	3	)	)	PUNCT
fcis-7512	118	4	for	for	ADP
fcis-7512	118	5	the	the	DET
fcis-7512	118	6	loss	loss	NOUN
fcis-7512	118	7	of	of	ADP
fcis-7512	118	8	each	each	DET
fcis-7512	118	9	task	task	NOUN
fcis-7512	118	10	,	,	PUNCT
fcis-7512	118	11	this	this	DET
fcis-7512	118	12	paper	paper	NOUN
fcis-7512	118	13	calculates	calculate	VERB
fcis-7512	118	14	a	a	DET
fcis-7512	118	15	loss	loss	NOUN
fcis-7512	118	16	weight	weight	NOUN
fcis-7512	118	17	𝑤_𝐿𝑠	𝑤_𝐿𝑠	NOUN
fcis-7512	118	18	where	where	SCONJ
fcis-7512	118	19	𝑠	𝑠	PROPN
fcis-7512	118	20	∈	∈	PROPN
fcis-7512	118	21	{	{	PUNCT
fcis-7512	118	22	a	a	PRON
fcis-7512	118	23	,	,	PUNCT
fcis-7512	118	24	v	v	NOUN
fcis-7512	118	25	,	,	PUNCT
fcis-7512	118	26	t	t	PROPN
fcis-7512	118	27	,	,	PUNCT
fcis-7512	118	28	m	m	NOUN
fcis-7512	118	29	}	}	PUNCT
fcis-7512	118	30	and	and	CCONJ
fcis-7512	118	31	assigns	assign	VERB
fcis-7512	118	32	each	each	DET
fcis-7512	118	33	task	task	NOUN
fcis-7512	118	34	separately	separately	ADV
fcis-7512	118	35	.	.	PUNCT
fcis-7512	119	1	𝑤_𝐿𝑠	𝑤_𝐿𝑠	NOUN
fcis-7512	119	2	=	=	PUNCT
fcis-7512	119	3	𝐿𝑠	𝐿𝑠	PROPN
fcis-7512	119	4	∑	∑	PUNCT
fcis-7512	119	5	𝐿s	𝐿s	PROPN
fcis-7512	119	6	{	{	PUNCT
fcis-7512	119	7	a	a	PRON
fcis-7512	119	8	,	,	PUNCT
fcis-7512	119	9	v	v	NOUN
fcis-7512	119	10	,	,	PUNCT
fcis-7512	119	11	t	t	PROPN
fcis-7512	119	12	,	,	PUNCT
fcis-7512	119	13	m	m	NOUN
fcis-7512	119	14	}	}	PUNCT
fcis-7512	119	15	𝑠	𝑠	PROPN
fcis-7512	119	16	𝐿s	𝐿s	PROPN
fcis-7512	119	17	.	.	PUNCT
fcis-7512	120	1	(	(	PUNCT
fcis-7512	120	2	19	19	NUM
fcis-7512	120	3	)	)	PUNCT
fcis-7512	120	4	it	it	PRON
fcis-7512	120	5	indicates	indicate	VERB
fcis-7512	120	6	that	that	SCONJ
fcis-7512	120	7	the	the	DET
fcis-7512	120	8	larger	large	ADJ
fcis-7512	120	9	weight	weight	NOUN
fcis-7512	120	10	loss	loss	NOUN
fcis-7512	120	11	is	be	AUX
fcis-7512	120	12	given	give	VERB
fcis-7512	120	13	to	to	ADP
fcis-7512	120	14	the	the	DET
fcis-7512	120	15	part	part	NOUN
fcis-7512	120	16	with	with	ADP
fcis-7512	120	17	larger	large	ADJ
fcis-7512	120	18	loss	loss	NOUN
fcis-7512	120	19	to	to	PART
fcis-7512	120	20	accelerate	accelerate	VERB
fcis-7512	120	21	the	the	DET
fcis-7512	120	22	convergence	convergence	NOUN
fcis-7512	120	23	of	of	ADP
fcis-7512	120	24	the	the	DET
fcis-7512	120	25	model	model	NOUN
fcis-7512	120	26	and	and	CCONJ
fcis-7512	120	27	accelerate	accelerate	VERB
fcis-7512	120	28	the	the	DET
fcis-7512	120	29	training	training	NOUN
fcis-7512	120	30	process	process	NOUN
fcis-7512	120	31	.	.	PUNCT
fcis-7512	121	1	the	the	DET
fcis-7512	121	2	total	total	ADJ
fcis-7512	121	3	loss	loss	NOUN
fcis-7512	121	4	of	of	ADP
fcis-7512	121	5	the	the	DET
fcis-7512	121	6	model	model	NOUN
fcis-7512	121	7	is	be	AUX
fcis-7512	121	8	:	:	PUNCT
fcis-7512	121	9	𝐿	𝐿	PROPN
fcis-7512	121	10	=	=	SYM
fcis-7512	121	11	∑	∑	PUNCT
fcis-7512	121	12	𝑤_𝐿𝑠	𝑤_𝐿𝑠	ADJ
fcis-7512	121	13	{	{	PUNCT
fcis-7512	121	14	a	a	DET
fcis-7512	121	15	,	,	PUNCT
fcis-7512	121	16	v	v	NOUN
fcis-7512	121	17	,	,	PUNCT
fcis-7512	121	18	t	t	PROPN
fcis-7512	121	19	,	,	PUNCT
fcis-7512	121	20	m	m	NOUN
fcis-7512	121	21	}	}	PUNCT
fcis-7512	121	22	𝑠	𝑠	PROPN
fcis-7512	121	23	.	.	PUNCT
fcis-7512	122	1	(	(	PUNCT
fcis-7512	122	2	19	19	NUM
fcis-7512	122	3	)	)	PUNCT
fcis-7512	122	4	3	3	NUM
fcis-7512	122	5	.	.	X
fcis-7512	123	1	experimental	experimental	ADJ
fcis-7512	123	2	setting	set	VERB
fcis-7512	123	3	3.1	3.1	NUM
fcis-7512	123	4	.	.	PUNCT
fcis-7512	124	1	dataset	dataset	VERB
fcis-7512	124	2	the	the	DET
fcis-7512	124	3	datasets	dataset	NOUN
fcis-7512	124	4	used	use	VERB
fcis-7512	124	5	in	in	ADP
fcis-7512	124	6	this	this	DET
fcis-7512	124	7	paper	paper	NOUN
fcis-7512	124	8	are	be	AUX
fcis-7512	124	9	mosi	mosi	NOUN
fcis-7512	124	10	and	and	CCONJ
fcis-7512	124	11	mosei	mosei	ADJ
fcis-7512	124	12	common	common	ADJ
fcis-7512	124	13	data	data	NOUN
fcis-7512	124	14	sets	set	NOUN
fcis-7512	124	15	for	for	ADP
fcis-7512	124	16	multimodal	multimodal	ADJ
fcis-7512	124	17	sentiment	sentiment	NOUN
fcis-7512	124	18	analysis	analysis	NOUN
fcis-7512	124	19	.	.	PUNCT
fcis-7512	125	1	the	the	DET
fcis-7512	125	2	basic	basic	ADJ
fcis-7512	125	3	information	information	NOUN
fcis-7512	125	4	of	of	ADP
fcis-7512	125	5	the	the	DET
fcis-7512	125	6	two	two	NUM
fcis-7512	125	7	data	data	NOUN
fcis-7512	125	8	sets	set	NOUN
fcis-7512	125	9	is	be	AUX
fcis-7512	125	10	shown	show	VERB
fcis-7512	125	11	in	in	ADP
fcis-7512	125	12	table	table	NOUN
fcis-7512	125	13	1	1	NUM
fcis-7512	125	14	.	.	PUNCT
fcis-7512	125	15	table	table	NOUN
fcis-7512	125	16	1	1	NUM
fcis-7512	125	17	.	.	PUNCT
fcis-7512	125	18	basic	basic	ADJ
fcis-7512	125	19	dataset	dataset	NOUN
fcis-7512	125	20	information	information	NOUN
fcis-7512	125	21	dataset	dataset	VERB
fcis-7512	125	22	train	train	NOUN
fcis-7512	125	23	valid	valid	ADJ
fcis-7512	125	24	test	test	NOUN
fcis-7512	125	25	all	all	DET
fcis-7512	125	26	mosi	mosi	PROPN
fcis-7512	125	27	1284	1284	NUM
fcis-7512	125	28	299	299	NUM
fcis-7512	125	29	686	686	NUM
fcis-7512	125	30	2199	2199	NUM
fcis-7512	126	1	mosei	mosei	ADJ
fcis-7512	126	2	16323	16323	NUM
fcis-7512	126	3	1871	1871	NUM
fcis-7512	126	4	4659	4659	NUM
fcis-7512	126	5	22856	22856	NUM
fcis-7512	126	6	mosi	mosi	PROPN
fcis-7512	126	7	data	datum	NOUN
fcis-7512	126	8	set	set	NOUN
fcis-7512	126	9	:	:	PUNCT
fcis-7512	126	10	mosi	mosi	NUM
fcis-7512	126	11	data	datum	NOUN
fcis-7512	126	12	set	set	VERB
fcis-7512	126	13	is	be	AUX
fcis-7512	126	14	a	a	DET
fcis-7512	126	15	commonly	commonly	ADV
fcis-7512	126	16	used	use	VERB
fcis-7512	126	17	benchmark	benchmark	NOUN
fcis-7512	126	18	data	datum	NOUN
fcis-7512	126	19	set	set	VERB
fcis-7512	126	20	for	for	ADP
fcis-7512	126	21	multimodal	multimodal	ADJ
fcis-7512	126	22	sentiment	sentiment	NOUN
fcis-7512	126	23	analysis	analysis	NOUN
fcis-7512	126	24	.	.	PUNCT
fcis-7512	127	1	it	it	PRON
fcis-7512	127	2	is	be	AUX
fcis-7512	127	3	a	a	DET
fcis-7512	127	4	video	video	NOUN
fcis-7512	127	5	blog	blog	NOUN
fcis-7512	127	6	mainly	mainly	ADV
fcis-7512	127	7	about	about	ADP
fcis-7512	127	8	film	film	NOUN
fcis-7512	127	9	review	review	NOUN
fcis-7512	127	10	videos	video	NOUN
fcis-7512	127	11	on	on	ADP
fcis-7512	127	12	youtube	youtube	NOUN
fcis-7512	127	13	collected	collect	VERB
fcis-7512	127	14	by	by	ADP
fcis-7512	127	15	zadeh	zadeh	PROPN
fcis-7512	127	16	et	et	PROPN
fcis-7512	127	17	al	al	PROPN
fcis-7512	127	18	.	.	PUNCT
fcis-7512	128	1	[	[	X
fcis-7512	128	2	35	35	NUM
fcis-7512	128	3	]	]	PUNCT
fcis-7512	128	4	.	.	PUNCT
fcis-7512	129	1	a	a	DET
fcis-7512	129	2	total	total	NOUN
fcis-7512	129	3	of	of	ADP
fcis-7512	129	4	93	93	NUM
fcis-7512	129	5	videos	video	NOUN
fcis-7512	129	6	including	include	VERB
fcis-7512	129	7	2199	2199	NUM
fcis-7512	129	8	video	video	NOUN
fcis-7512	129	9	clips	clip	NOUN
fcis-7512	129	10	were	be	AUX
fcis-7512	129	11	randomly	randomly	ADV
fcis-7512	129	12	collected	collect	VERB
fcis-7512	129	13	.	.	PUNCT
fcis-7512	130	1	the	the	DET
fcis-7512	130	2	labels	label	NOUN
fcis-7512	130	3	of	of	ADP
fcis-7512	130	4	these	these	DET
fcis-7512	130	5	videos	video	NOUN
fcis-7512	130	6	were	be	AUX
fcis-7512	130	7	marked	mark	VERB
fcis-7512	130	8	and	and	CCONJ
fcis-7512	130	9	averaged	average	VERB
fcis-7512	130	10	by	by	ADP
fcis-7512	130	11	five	five	NUM
fcis-7512	130	12	markers	marker	NOUN
fcis-7512	130	13	from	from	ADP
fcis-7512	130	14	amazon	amazon	NOUN
fcis-7512	130	15	's	's	PART
fcis-7512	130	16	crowdsourcing	crowdsourcing	NOUN
fcis-7512	130	17	platform	platform	NOUN
fcis-7512	130	18	,	,	PUNCT
fcis-7512	130	19	and	and	CCONJ
fcis-7512	130	20	marked	mark	VERB
fcis-7512	130	21	as	as	ADP
fcis-7512	130	22	seven	seven	NUM
fcis-7512	130	23	emotional	emotional	ADJ
fcis-7512	130	24	tendencies	tendency	NOUN
fcis-7512	130	25	ranging	range	VERB
fcis-7512	130	26	from	from	ADP
fcis-7512	130	27	-3	-3	PUNCT
fcis-7512	130	28	(	(	PUNCT
fcis-7512	130	29	negative	negative	ADJ
fcis-7512	130	30	)	)	PUNCT
fcis-7512	130	31	to	to	ADP
fcis-7512	130	32	+3	+3	PROPN
fcis-7512	130	33	(	(	PUNCT
fcis-7512	130	34	positive	positive	ADJ
fcis-7512	130	35	)	)	PUNCT
fcis-7512	130	36	.	.	PUNCT
fcis-7512	131	1	mosei	mosei	ADJ
fcis-7512	131	2	dataset	dataset	NOUN
fcis-7512	131	3	:	:	PUNCT
fcis-7512	131	4	the	the	DET
fcis-7512	131	5	mosei	mosei	ADJ
fcis-7512	131	6	[	[	X
fcis-7512	131	7	36	36	NUM
fcis-7512	131	8	]	]	X
fcis-7512	131	9	dataset	dataset	NOUN
fcis-7512	131	10	expands	expand	VERB
fcis-7512	131	11	the	the	DET
fcis-7512	131	12	content	content	NOUN
fcis-7512	131	13	of	of	ADP
fcis-7512	131	14	the	the	DET
fcis-7512	131	15	data	datum	NOUN
fcis-7512	131	16	,	,	PUNCT
fcis-7512	131	17	including	include	VERB
fcis-7512	131	18	3228	3228	NUM
fcis-7512	131	19	videos	video	NOUN
fcis-7512	131	20	,	,	PUNCT
fcis-7512	131	21	23,453	23,453	NUM
fcis-7512	131	22	sentences	sentence	NOUN
fcis-7512	131	23	,	,	PUNCT
fcis-7512	131	24	1000	1000	NUM
fcis-7512	131	25	narrators	narrator	NOUN
fcis-7512	131	26	,	,	PUNCT
fcis-7512	131	27	250	250	NUM
fcis-7512	131	28	topics	topic	NOUN
fcis-7512	131	29	,	,	PUNCT
fcis-7512	131	30	with	with	ADP
fcis-7512	131	31	a	a	DET
fcis-7512	131	32	total	total	ADJ
fcis-7512	131	33	duration	duration	NOUN
fcis-7512	131	34	of	of	ADP
fcis-7512	131	35	65	65	NUM
fcis-7512	131	36	hours	hour	NOUN
fcis-7512	131	37	and	and	CCONJ
fcis-7512	131	38	a	a	DET
fcis-7512	131	39	total	total	NOUN
fcis-7512	131	40	of	of	ADP
fcis-7512	131	41	22,856	22,856	NUM
fcis-7512	131	42	video	video	NOUN
fcis-7512	131	43	segments	segment	NOUN
fcis-7512	131	44	.	.	PUNCT
fcis-7512	132	1	this	this	DET
fcis-7512	132	2	data	datum	NOUN
fcis-7512	132	3	set	set	VERB
fcis-7512	132	4	provides	provide	VERB
fcis-7512	132	5	data	datum	NOUN
fcis-7512	132	6	annotation	annotation	NOUN
fcis-7512	132	7	for	for	ADP
fcis-7512	132	8	emotion	emotion	NOUN
fcis-7512	132	9	classification	classification	NOUN
fcis-7512	132	10	of	of	ADP
fcis-7512	132	11	2	2	NUM
fcis-7512	132	12	,	,	PUNCT
fcis-7512	132	13	5	5	NUM
fcis-7512	132	14	and	and	CCONJ
fcis-7512	132	15	7	7	NUM
fcis-7512	132	16	categories	category	NOUN
fcis-7512	132	17	and	and	CCONJ
fcis-7512	132	18	also	also	ADV
fcis-7512	132	19	marks	mark	VERB
fcis-7512	132	20	emotion	emotion	NOUN
fcis-7512	132	21	intensity	intensity	NOUN
fcis-7512	132	22	,	,	PUNCT
fcis-7512	132	23	which	which	PRON
fcis-7512	132	24	is	be	AUX
fcis-7512	132	25	marked	mark	VERB
fcis-7512	132	26	as	as	ADP
fcis-7512	132	27	seven	seven	NUM
fcis-7512	132	28	emotional	emotional	ADJ
fcis-7512	132	29	tendencies	tendency	NOUN
fcis-7512	132	30	from	from	ADP
fcis-7512	132	31	-3	-3	PUNCT
fcis-7512	132	32	(	(	PUNCT
fcis-7512	132	33	negative	negative	ADJ
fcis-7512	132	34	)	)	PUNCT
fcis-7512	132	35	to	to	ADP
fcis-7512	132	36	+3	+3	PROPN
fcis-7512	132	37	(	(	PUNCT
fcis-7512	132	38	positive	positive	ADJ
fcis-7512	132	39	)	)	PUNCT
fcis-7512	132	40	.	.	PUNCT
fcis-7512	133	1	3.2	3.2	NUM
fcis-7512	133	2	.	.	PUNCT
fcis-7512	134	1	contrast	contrast	NOUN
fcis-7512	134	2	model	model	NOUN
fcis-7512	134	3	in	in	ADP
fcis-7512	134	4	order	order	NOUN
fcis-7512	134	5	to	to	PART
fcis-7512	134	6	fully	fully	ADV
fcis-7512	134	7	verify	verify	VERB
fcis-7512	134	8	the	the	DET
fcis-7512	134	9	performance	performance	NOUN
fcis-7512	134	10	of	of	ADP
fcis-7512	134	11	the	the	DET
fcis-7512	134	12	model	model	NOUN
fcis-7512	135	1	,	,	PUNCT
fcis-7512	135	2	the	the	DET
fcis-7512	135	3	recent	recent	ADJ
fcis-7512	135	4	multimodal	multimodal	ADJ
fcis-7512	135	5	sentiment	sentiment	NOUN
fcis-7512	135	6	analysis	analysis	NOUN
fcis-7512	135	7	benchmark	benchmark	NOUN
fcis-7512	135	8	model	model	NOUN
fcis-7512	135	9	is	be	AUX
fcis-7512	135	10	compared	compare	VERB
fcis-7512	135	11	with	with	ADP
fcis-7512	135	12	the	the	DET
fcis-7512	135	13	proposed	propose	VERB
fcis-7512	135	14	model	model	NOUN
fcis-7512	135	15	under	under	ADP
fcis-7512	135	16	the	the	DET
fcis-7512	135	17	same	same	ADJ
fcis-7512	135	18	data	datum	NOUN
fcis-7512	135	19	set	set	VERB
fcis-7512	135	20	.	.	PUNCT
fcis-7512	136	1	the	the	DET
fcis-7512	136	2	relevant	relevant	ADJ
fcis-7512	136	3	benchmark	benchmark	NOUN
fcis-7512	136	4	model	model	NOUN
fcis-7512	136	5	is	be	AUX
fcis-7512	136	6	as	as	SCONJ
fcis-7512	136	7	follows	follow	VERB
fcis-7512	136	8	:	:	PUNCT
fcis-7512	136	9	tfn	tfn	NOUN
fcis-7512	136	10	[	[	X
fcis-7512	136	11	6	6	NUM
fcis-7512	136	12	]	]	NOUN
fcis-7512	136	13	:	:	PUNCT
fcis-7512	136	14	tensor	tensor	NOUN
fcis-7512	136	15	fusion	fusion	NOUN
fcis-7512	136	16	network	network	NOUN
fcis-7512	136	17	,	,	PUNCT
fcis-7512	136	18	by	by	ADP
fcis-7512	136	19	learning	learn	VERB
fcis-7512	136	20	the	the	DET
fcis-7512	136	21	data	data	NOUN
fcis-7512	136	22	representation	representation	NOUN
fcis-7512	136	23	of	of	ADP
fcis-7512	136	24	three	three	NUM
fcis-7512	136	25	modes	mode	NOUN
fcis-7512	136	26	respectively	respectively	ADV
fcis-7512	136	27	,	,	PUNCT
fcis-7512	136	28	carries	carry	VERB
fcis-7512	136	29	out	out	ADP
fcis-7512	136	30	feature	feature	NOUN
fcis-7512	136	31	fusion	fusion	NOUN
fcis-7512	136	32	through	through	ADP
fcis-7512	136	33	the	the	DET
fcis-7512	136	34	form	form	NOUN
fcis-7512	136	35	of	of	ADP
fcis-7512	136	36	cartesian	cartesian	ADJ
fcis-7512	136	37	product	product	NOUN
fcis-7512	136	38	,	,	PUNCT
fcis-7512	136	39	and	and	CCONJ
fcis-7512	136	40	finally	finally	ADV
fcis-7512	136	41	realizes	realize	VERB
fcis-7512	136	42	the	the	DET
fcis-7512	136	43	task	task	NOUN
fcis-7512	136	44	of	of	ADP
fcis-7512	136	45	multi	multi	ADJ
fcis-7512	136	46	-	-	ADJ
fcis-7512	136	47	mode	mode	ADJ
fcis-7512	136	48	sentiment	sentiment	NOUN
fcis-7512	136	49	analysis	analysis	NOUN
fcis-7512	136	50	through	through	ADP
fcis-7512	136	51	classifier	classifier	NOUN
fcis-7512	136	52	.	.	PUNCT
fcis-7512	137	1	lmf	lmf	PROPN
fcis-7512	138	1	[	[	X
fcis-7512	138	2	19	19	NUM
fcis-7512	138	3	]	]	NOUN
fcis-7512	138	4	:	:	PUNCT
fcis-7512	138	5	low	low	ADJ
fcis-7512	138	6	-	-	PUNCT
fcis-7512	138	7	rank	rank	NOUN
fcis-7512	138	8	fusion	fusion	NOUN
fcis-7512	138	9	network	network	NOUN
fcis-7512	138	10	,	,	PUNCT
fcis-7512	138	11	which	which	PRON
fcis-7512	138	12	is	be	AUX
fcis-7512	138	13	an	an	DET
fcis-7512	138	14	improvement	improvement	NOUN
fcis-7512	138	15	of	of	ADP
fcis-7512	138	16	tensor	tensor	NOUN
fcis-7512	138	17	fusion	fusion	NOUN
fcis-7512	138	18	network	network	NOUN
fcis-7512	138	19	,	,	PUNCT
fcis-7512	138	20	reduces	reduce	VERB
fcis-7512	138	21	the	the	DET
fcis-7512	138	22	computation	computation	NOUN
fcis-7512	138	23	amount	amount	NOUN
fcis-7512	138	24	of	of	ADP
fcis-7512	138	25	cartesian	cartesian	ADJ
fcis-7512	138	26	product	product	NOUN
fcis-7512	138	27	of	of	ADP
fcis-7512	138	28	tensor	tensor	NOUN
fcis-7512	138	29	fusion	fusion	NOUN
fcis-7512	138	30	network	network	NOUN
fcis-7512	138	31	through	through	ADP
fcis-7512	138	32	low	low	ADJ
fcis-7512	138	33	-	-	PUNCT
fcis-7512	138	34	dimensional	dimensional	ADJ
fcis-7512	138	35	vector	vector	NOUN
fcis-7512	138	36	representation	representation	NOUN
fcis-7512	138	37	.	.	PUNCT
fcis-7512	139	1	mfn	mfn	PROPN
fcis-7512	140	1	[	[	X
fcis-7512	140	2	20	20	NUM
fcis-7512	140	3	]	]	NOUN
fcis-7512	140	4	:	:	PUNCT
fcis-7512	140	5	memory	memory	NOUN
fcis-7512	140	6	fusion	fusion	NOUN
fcis-7512	140	7	network	network	NOUN
fcis-7512	140	8	,	,	PUNCT
fcis-7512	140	9	which	which	PRON
fcis-7512	140	10	uses	use	VERB
fcis-7512	140	11	lstm	lstm	NOUN
fcis-7512	140	12	to	to	PART
fcis-7512	140	13	learn	learn	VERB
fcis-7512	140	14	single	single	ADJ
fcis-7512	140	15	-	-	PUNCT
fcis-7512	140	16	mode	mode	NOUN
fcis-7512	140	17	characteristic	characteristic	ADJ
fcis-7512	140	18	information	information	NOUN
fcis-7512	140	19	and	and	CCONJ
fcis-7512	140	20	then	then	ADV
fcis-7512	140	21	uses	use	VERB
fcis-7512	140	22	the	the	DET
fcis-7512	140	23	memory	memory	NOUN
fcis-7512	140	24	attention	attention	NOUN
fcis-7512	140	25	mechanism	mechanism	NOUN
fcis-7512	140	26	to	to	PART
fcis-7512	140	27	realize	realize	VERB
fcis-7512	140	28	the	the	DET
fcis-7512	140	29	interaction	interaction	NOUN
fcis-7512	140	30	between	between	ADP
fcis-7512	140	31	modes	mode	NOUN
fcis-7512	140	32	to	to	PART
fcis-7512	140	33	achieve	achieve	VERB
fcis-7512	140	34	multi	multi	ADJ
fcis-7512	140	35	-	-	ADJ
fcis-7512	140	36	mode	mode	ADJ
fcis-7512	140	37	emotion	emotion	NOUN
fcis-7512	140	38	analysis	analysis	NOUN
fcis-7512	140	39	.	.	PUNCT
fcis-7512	141	1	misa	misa	NOUN
fcis-7512	141	2	[	[	X
fcis-7512	141	3	14	14	NUM
fcis-7512	141	4	]	]	X
fcis-7512	141	5	:	:	PUNCT
fcis-7512	141	6	multi	multi	ADJ
fcis-7512	141	7	-	-	ADJ
fcis-7512	141	8	modal	modal	ADJ
fcis-7512	141	9	data	datum	NOUN
fcis-7512	141	10	is	be	AUX
fcis-7512	141	11	mapped	map	VERB
fcis-7512	141	12	to	to	ADP
fcis-7512	141	13	two	two	NUM
fcis-7512	141	14	subspaces	subspace	NOUN
fcis-7512	141	15	,	,	PUNCT
fcis-7512	141	16	and	and	CCONJ
fcis-7512	141	17	then	then	ADV
fcis-7512	141	18	feature	feature	VERB
fcis-7512	141	19	vectors	vector	NOUN
fcis-7512	141	20	of	of	ADP
fcis-7512	141	21	different	different	ADJ
fcis-7512	141	22	subspaces	subspace	NOUN
fcis-7512	141	23	are	be	AUX
fcis-7512	141	24	fused	fuse	VERB
fcis-7512	141	25	.	.	PUNCT
fcis-7512	142	1	reven	reven	NOUN
fcis-7512	143	1	[	[	X
fcis-7512	143	2	21	21	NUM
fcis-7512	143	3	]	]	X
fcis-7512	143	4	:	:	PUNCT
fcis-7512	143	5	an	an	DET
fcis-7512	143	6	attention	attention	NOUN
fcis-7512	143	7	-	-	PUNCT
fcis-7512	143	8	based	base	VERB
fcis-7512	143	9	model	model	NOUN
fcis-7512	143	10	adjusts	adjust	VERB
fcis-7512	143	11	word	word	NOUN
fcis-7512	143	12	embedding	embed	VERB
fcis-7512	143	13	in	in	ADP
fcis-7512	143	14	textual	textual	ADJ
fcis-7512	143	15	data	datum	NOUN
fcis-7512	143	16	through	through	ADP
fcis-7512	143	17	non	non	ADJ
fcis-7512	143	18	-	-	ADJ
fcis-7512	143	19	textual	textual	ADJ
fcis-7512	143	20	data	datum	NOUN
fcis-7512	143	21	.	.	PUNCT
fcis-7512	144	1	mult	mult	PROPN
fcis-7512	145	1	[	[	X
fcis-7512	145	2	8	8	NUM
fcis-7512	145	3	]	]	SYM
fcis-7512	145	4	:	:	PUNCT
fcis-7512	145	5	two	two	NUM
fcis-7512	145	6	-	-	PUNCT
fcis-7512	145	7	way	way	NOUN
fcis-7512	145	8	transformer	transformer	NOUN
fcis-7512	145	9	is	be	AUX
fcis-7512	145	10	used	use	VERB
fcis-7512	145	11	to	to	PART
fcis-7512	145	12	realize	realize	VERB
fcis-7512	145	13	the	the	DET
fcis-7512	145	14	representation	representation	NOUN
fcis-7512	145	15	of	of	ADP
fcis-7512	145	16	one	one	NUM
fcis-7512	145	17	mode	mode	NOUN
fcis-7512	145	18	corresponding	correspond	VERB
fcis-7512	145	19	to	to	ADP
fcis-7512	145	20	another	another	DET
fcis-7512	145	21	mode	mode	NOUN
fcis-7512	145	22	,	,	PUNCT
fcis-7512	145	23	and	and	CCONJ
fcis-7512	145	24	then	then	ADV
fcis-7512	145	25	the	the	DET
fcis-7512	145	26	feature	feature	NOUN
fcis-7512	145	27	fusion	fusion	NOUN
fcis-7512	145	28	process	process	NOUN
fcis-7512	145	29	is	be	AUX
fcis-7512	145	30	carried	carry	VERB
fcis-7512	145	31	out	out	ADP
fcis-7512	145	32	.	.	PUNCT
fcis-7512	146	1	mag	mag	ADJ
fcis-7512	146	2	-	-	PUNCT
fcis-7512	146	3	bert	bert	NOUN
fcis-7512	147	1	[	[	X
fcis-7512	147	2	22	22	NUM
fcis-7512	147	3	]	]	X
fcis-7512	147	4	:	:	PUNCT
fcis-7512	147	5	a	a	DET
fcis-7512	147	6	multi	multi	ADJ
fcis-7512	147	7	-	-	ADJ
fcis-7512	147	8	modal	modal	ADJ
fcis-7512	147	9	adaptation	adaptation	NOUN
fcis-7512	147	10	gate	gate	NOUN
fcis-7512	147	11	(	(	PUNCT
fcis-7512	147	12	mag	mag	PROPN
fcis-7512	147	13	)	)	PUNCT
fcis-7512	147	14	is	be	AUX
fcis-7512	147	15	designed	design	VERB
fcis-7512	147	16	to	to	PART
fcis-7512	147	17	optimize	optimize	VERB
fcis-7512	147	18	the	the	DET
fcis-7512	147	19	fusion	fusion	NOUN
fcis-7512	147	20	process	process	NOUN
fcis-7512	147	21	of	of	ADP
fcis-7512	147	22	textual	textual	ADJ
fcis-7512	147	23	and	and	CCONJ
fcis-7512	147	24	nontextual	nontextual	ADJ
fcis-7512	147	25	data	datum	NOUN
fcis-7512	147	26	.	.	PUNCT
fcis-7512	148	1	self	self	NOUN
fcis-7512	148	2	-	-	PUNCT
fcis-7512	148	3	mm	mm	NOUN
fcis-7512	149	1	[	[	X
fcis-7512	149	2	15	15	NUM
fcis-7512	149	3	]	]	X
fcis-7512	149	4	:	:	PUNCT
fcis-7512	149	5	multi	multi	ADJ
fcis-7512	149	6	-	-	ADJ
fcis-7512	149	7	modal	modal	ADJ
fcis-7512	149	8	sentiment	sentiment	NOUN
fcis-7512	149	9	analysis	analysis	NOUN
fcis-7512	149	10	task	task	NOUN
fcis-7512	149	11	is	be	AUX
fcis-7512	149	12	realized	realize	VERB
fcis-7512	149	13	by	by	ADP
fcis-7512	149	14	using	use	VERB
fcis-7512	149	15	multi	multi	ADJ
fcis-7512	149	16	-	-	ADJ
fcis-7512	149	17	task	task	ADJ
fcis-7512	149	18	learning	learning	NOUN
fcis-7512	149	19	and	and	CCONJ
fcis-7512	149	20	single	single	ADJ
fcis-7512	149	21	-	-	PUNCT
fcis-7512	149	22	modal	modal	NOUN
fcis-7512	149	23	label	label	NOUN
fcis-7512	149	24	automatic	automatic	ADJ
fcis-7512	149	25	generation	generation	NOUN
fcis-7512	149	26	.	.	PUNCT
fcis-7512	150	1	3.3	3.3	NUM
fcis-7512	150	2	.	.	PUNCT
fcis-7512	151	1	hyperparameter	hyperparameter	NOUN
fcis-7512	151	2	and	and	CCONJ
fcis-7512	151	3	evaluation	evaluation	NOUN
fcis-7512	151	4	index	index	NOUN
fcis-7512	151	5	this	this	DET
fcis-7512	151	6	article	article	NOUN
fcis-7512	151	7	uses	use	VERB
fcis-7512	151	8	the	the	DET
fcis-7512	151	9	adam	adam	PROPN
fcis-7512	151	10	as	as	ADP
fcis-7512	151	11	optimizer	optimizer	NOUN
fcis-7512	151	12	,	,	PUNCT
fcis-7512	151	13	bert	bert	PROPN
fcis-7512	151	14	vector	vector	NOUN
fcis-7512	151	15	of	of	ADP
fcis-7512	151	16	the	the	DET
fcis-7512	151	17	initial	initial	ADJ
fcis-7512	151	18	5e−5	5e−5	NUM
fcis-7512	151	19	,	,	PUNCT
fcis-7512	151	20	dropout	dropout	NOUN
fcis-7512	151	21	value	value	NOUN
fcis-7512	151	22	of	of	ADP
fcis-7512	151	23	0.1	0.1	NUM
fcis-7512	151	24	,	,	PUNCT
fcis-7512	151	25	two	two	NUM
fcis-7512	151	26	-	-	PUNCT
fcis-7512	151	27	way	way	NOUN
fcis-7512	151	28	lstm	lstm	ADJ
fcis-7512	151	29	dropout	dropout	NOUN
fcis-7512	151	30	value	value	NOUN
fcis-7512	151	31	is	be	AUX
fcis-7512	151	32	0	0	NUM
fcis-7512	151	33	,	,	PUNCT
fcis-7512	151	34	vector	vector	NOUN
fcis-7512	151	35	other	other	ADJ
fcis-7512	151	36	parameters	parameter	NOUN
fcis-7512	151	37	as	as	SCONJ
fcis-7512	151	38	shown	show	VERB
fcis-7512	151	39	in	in	ADP
fcis-7512	151	40	table	table	NOUN
fcis-7512	151	41	2	2	NUM
fcis-7512	151	42	,	,	PUNCT
fcis-7512	151	43	batch_size	batch_size	NOUN
fcis-7512	151	44	(	(	PUNCT
fcis-7512	151	45	batch_size=32	batch_size=32	NOUN
fcis-7512	151	46	)	)	PUNCT
fcis-7512	151	47	was	be	AUX
fcis-7512	151	48	used	use	VERB
fcis-7512	151	49	for	for	ADP
fcis-7512	151	50	both	both	DET
fcis-7512	151	51	mosi	mosi	NOUN
fcis-7512	151	52	and	and	CCONJ
fcis-7512	151	53	mosei	mosei	ADJ
fcis-7512	151	54	data	datum	NOUN
fcis-7512	151	55	sets	set	NOUN
fcis-7512	151	56	.	.	PUNCT
fcis-7512	152	1	table	table	NOUN
fcis-7512	152	2	2	2	NUM
fcis-7512	152	3	.	.	PUNCT
fcis-7512	152	4	relevant	relevant	ADJ
fcis-7512	152	5	learning	learning	NOUN
fcis-7512	152	6	rates	rate	NOUN
fcis-7512	152	7	of	of	ADP
fcis-7512	152	8	mosi	mosi	NOUN
fcis-7512	152	9	and	and	CCONJ
fcis-7512	152	10	mosei	mosei	ADJ
fcis-7512	152	11	datasets	dataset	NOUN
fcis-7512	152	12	video	video	NOUN
fcis-7512	152	13	data	datum	NOUN
fcis-7512	152	14	learning	learn	VERB
fcis-7512	152	15	rate	rate	NOUN
fcis-7512	152	16	audio	audio	ADJ
fcis-7512	152	17	data	datum	NOUN
fcis-7512	152	18	learning	learn	VERB
fcis-7512	152	19	rate	rate	NOUN
fcis-7512	152	20	learning	learning	NOUN
fcis-7512	152	21	rate	rate	NOUN
fcis-7512	152	22	of	of	ADP
fcis-7512	152	23	other	other	ADJ
fcis-7512	152	24	parameters	parameter	NOUN
fcis-7512	152	25	video	video	NOUN
fcis-7512	152	26	data	datum	NOUN
fcis-7512	152	27	learning	learn	VERB
fcis-7512	152	28	rate	rate	NOUN
fcis-7512	152	29	mosi	mosi	NOUN
fcis-7512	152	30	0.6e−4	0.6e−4	NOUN
fcis-7512	153	1	0.6e−3	0.6e−3	NUM
fcis-7512	153	2	0.6e−3	0.6e−3	NUM
fcis-7512	153	3	mosei	mosei	ADJ
fcis-7512	154	1	0.6e−4	0.6e−4	ADV
fcis-7512	154	2	0.005	0.005	NUM
fcis-7512	154	3	0.6e−3	0.6e−3	NUM
fcis-7512	155	1	according	accord	VERB
fcis-7512	155	2	to	to	ADP
fcis-7512	155	3	previous	previous	ADJ
fcis-7512	155	4	work	work	NOUN
fcis-7512	155	5	[	[	X
fcis-7512	155	6	14,22	14,22	ADV
fcis-7512	155	7	]	]	X
fcis-7512	155	8	,	,	PUNCT
fcis-7512	155	9	this	this	DET
fcis-7512	155	10	paper	paper	NOUN
fcis-7512	155	11	evaluates	evaluate	VERB
fcis-7512	155	12	the	the	DET
fcis-7512	155	13	performance	performance	NOUN
fcis-7512	155	14	of	of	ADP
fcis-7512	155	15	the	the	DET
fcis-7512	155	16	proposed	propose	VERB
fcis-7512	155	17	model	model	NOUN
fcis-7512	155	18	in	in	ADP
fcis-7512	155	19	two	two	NUM
fcis-7512	155	20	ways	way	NOUN
fcis-7512	155	21	:	:	PUNCT
fcis-7512	155	22	classification	classification	NOUN
fcis-7512	155	23	and	and	CCONJ
fcis-7512	155	24	regression	regression	NOUN
fcis-7512	155	25	.	.	PUNCT
fcis-7512	156	1	for	for	ADP
fcis-7512	156	2	classification	classification	NOUN
fcis-7512	156	3	problems	problem	NOUN
fcis-7512	156	4	,	,	PUNCT
fcis-7512	156	5	accuracy	accuracy	NOUN
fcis-7512	156	6	(	(	PUNCT
fcis-7512	156	7	seven	seven	NUM
fcis-7512	156	8	classification	classification	NOUN
fcis-7512	156	9	tasks	task	NOUN
fcis-7512	156	10	and	and	CCONJ
fcis-7512	156	11	two	two	NUM
fcis-7512	156	12	classification	classification	NOUN
fcis-7512	156	13	tasks	task	NOUN
fcis-7512	156	14	)	)	PUNCT
fcis-7512	156	15	and	and	CCONJ
fcis-7512	156	16	f1	f1	NOUN
fcis-7512	156	17	-	-	PUNCT
fcis-7512	156	18	score	score	NOUN
fcis-7512	156	19	were	be	AUX
fcis-7512	156	20	used	use	VERB
fcis-7512	156	21	to	to	PART
fcis-7512	156	22	evaluate	evaluate	VERB
fcis-7512	156	23	the	the	DET
fcis-7512	156	24	model	model	NOUN
fcis-7512	156	25	performance	performance	NOUN
fcis-7512	156	26	.	.	PUNCT
fcis-7512	157	1	specifically	specifically	ADV
fcis-7512	157	2	,	,	PUNCT
fcis-7512	157	3	for	for	ADP
fcis-7512	157	4	mosi	mosi	NOUN
fcis-7512	157	5	and	and	CCONJ
fcis-7512	157	6	mosei	mosei	ADJ
fcis-7512	157	7	datasets	dataset	NOUN
fcis-7512	157	8	,	,	PUNCT
fcis-7512	157	9	binary	binary	ADJ
fcis-7512	157	10	accuracy	accuracy	NOUN
fcis-7512	157	11	and	and	CCONJ
fcis-7512	157	12	f1	f1	NOUN
fcis-7512	157	13	scores	score	NOUN
fcis-7512	157	14	are	be	AUX
fcis-7512	157	15	calculated	calculate	VERB
fcis-7512	157	16	in	in	ADP
fcis-7512	157	17	two	two	NUM
fcis-7512	157	18	ways	way	NOUN
fcis-7512	157	19	:	:	PUNCT
fcis-7512	157	20	negative	negative	ADJ
fcis-7512	157	21	/	/	SYM
fcis-7512	157	22	non	non	ADJ
fcis-7512	157	23	-	-	ADJ
fcis-7512	157	24	negative	negative	ADJ
fcis-7512	157	25	(	(	PUNCT
fcis-7512	157	26	without	without	ADP
fcis-7512	157	27	excluding	exclude	VERB
fcis-7512	157	28	zero	zero	NUM
fcis-7512	157	29	)	)	PUNCT
fcis-7512	157	30	(	(	PUNCT
fcis-7512	157	31	zadeh	zadeh	X
fcis-7512	157	32	et	et	PROPN
fcis-7512	157	33	al	al	PROPN
fcis-7512	157	34	.	.	PUNCT
fcis-7512	158	1	[	[	X
fcis-7512	158	2	6	6	NUM
fcis-7512	158	3	]	]	PUNCT
fcis-7512	158	4	)	)	PUNCT
fcis-7512	158	5	and	and	CCONJ
fcis-7512	158	6	negative	negative	ADJ
fcis-7512	158	7	/	/	SYM
fcis-7512	158	8	positive	positive	ADJ
fcis-7512	158	9	(	(	PUNCT
fcis-7512	158	10	excluding	exclude	VERB
fcis-7512	158	11	zero	zero	NUM
fcis-7512	158	12	)	)	PUNCT
fcis-7512	158	13	(	(	PUNCT
fcis-7512	158	14	tsai	tsai	PROPN
fcis-7512	158	15	et	et	PROPN
fcis-7512	158	16	al	al	PROPN
fcis-7512	158	17	.	.	PUNCT
fcis-7512	159	1	[	[	X
fcis-7512	159	2	14	14	NUM
fcis-7512	159	3	]	]	SYM
fcis-7512	159	4	)	)	PUNCT
fcis-7512	159	5	.	.	PUNCT
fcis-7512	160	1	the	the	DET
fcis-7512	160	2	average	average	ADJ
fcis-7512	160	3	absolute	absolute	ADJ
fcis-7512	160	4	error	error	NOUN
fcis-7512	160	5	and	and	CCONJ
fcis-7512	160	6	correlation	correlation	NOUN
fcis-7512	160	7	coefficient	coefficient	NOUN
fcis-7512	160	8	are	be	AUX
fcis-7512	160	9	used	use	VERB
fcis-7512	160	10	as	as	ADP
fcis-7512	160	11	evaluation	evaluation	NOUN
fcis-7512	160	12	indexes	index	NOUN
fcis-7512	160	13	for	for	ADP
fcis-7512	160	14	regression	regression	NOUN
fcis-7512	160	15	tasks	task	NOUN
fcis-7512	160	16	.	.	PUNCT
fcis-7512	161	1	3.4	3.4	NUM
fcis-7512	161	2	.	.	PUNCT
fcis-7512	162	1	experimental	experimental	ADJ
fcis-7512	162	2	result	result	VERB
fcis-7512	162	3	the	the	DET
fcis-7512	162	4	specific	specific	ADJ
fcis-7512	162	5	experimental	experimental	ADJ
fcis-7512	162	6	results	result	NOUN
fcis-7512	162	7	are	be	AUX
fcis-7512	162	8	shown	show	VERB
fcis-7512	162	9	in	in	ADP
fcis-7512	162	10	table	table	NOUN
fcis-7512	162	11	3	3	NUM
fcis-7512	162	12	and	and	CCONJ
fcis-7512	162	13	table	table	NOUN
fcis-7512	162	14	4	4	NUM
fcis-7512	162	15	,	,	PUNCT
fcis-7512	162	16	where	where	SCONJ
fcis-7512	162	17	the	the	DET
fcis-7512	162	18	data	datum	NOUN
fcis-7512	162	19	in	in	ADP
fcis-7512	162	20	the	the	DET
fcis-7512	162	21	table	table	NOUN
fcis-7512	162	22	comes	come	VERB
fcis-7512	162	23	from	from	ADP
fcis-7512	162	24	the	the	DET
fcis-7512	162	25	experimental	experimental	ADJ
fcis-7512	162	26	data	datum	NOUN
fcis-7512	162	27	of	of	ADP
fcis-7512	162	28	wei	wei	PROPN
fcis-7512	162	29	et	et	PROPN
fcis-7512	162	30	al	al	PROPN
fcis-7512	162	31	.	.	PUNCT
fcis-7512	163	1	[	[	X
fcis-7512	163	2	23	23	NUM
fcis-7512	163	3	]	]	PUNCT
fcis-7512	163	4	,	,	PUNCT
fcis-7512	163	5	and	and	CCONJ
fcis-7512	163	6	the	the	DET
fcis-7512	163	7	model	model	NOUN
fcis-7512	163	8	is	be	AUX
fcis-7512	163	9	the	the	DET
fcis-7512	163	10	experimental	experimental	ADJ
fcis-7512	163	11	results	result	NOUN
fcis-7512	163	12	of	of	ADP
fcis-7512	163	13	the	the	DET
fcis-7512	163	14	model	model	NOUN
fcis-7512	163	15	in	in	ADP
fcis-7512	163	16	this	this	DET
fcis-7512	163	17	paper	paper	NOUN
fcis-7512	163	18	.	.	PUNCT
fcis-7512	164	1	for	for	ADP
fcis-7512	164	2	binary	binary	ADJ
fcis-7512	164	3	classification	classification	NOUN
fcis-7512	164	4	accuracy	accuracy	NOUN
fcis-7512	164	5	and	and	CCONJ
fcis-7512	164	6	f1	f1	NOUN
fcis-7512	164	7	score	score	NOUN
fcis-7512	164	8	,	,	PUNCT
fcis-7512	164	9	the	the	DET
fcis-7512	164	10	left	left	ADJ
fcis-7512	164	11	side	side	NOUN
fcis-7512	164	12	is	be	AUX
fcis-7512	164	13	the	the	DET
fcis-7512	164	14	negative	negative	ADJ
fcis-7512	164	15	/	/	SYM
fcis-7512	164	16	non	non	ADJ
fcis-7512	164	17	-	-	ADJ
fcis-7512	164	18	negative	negative	ADJ
fcis-7512	164	19	experimental	experimental	ADJ
fcis-7512	164	20	results	result	NOUN
fcis-7512	164	21	,	,	PUNCT
fcis-7512	164	22	and	and	CCONJ
fcis-7512	164	23	the	the	DET
fcis-7512	164	24	right	right	ADJ
fcis-7512	164	25	side	side	NOUN
fcis-7512	164	26	is	be	AUX
fcis-7512	164	27	the	the	DET
fcis-7512	164	28	negative	negative	ADJ
fcis-7512	164	29	/	/	SYM
fcis-7512	164	30	positive	positive	ADJ
fcis-7512	164	31	experimental	experimental	ADJ
fcis-7512	164	32	results	result	NOUN
fcis-7512	164	33	(	(	PUNCT
fcis-7512	164	34	excluding	exclude	VERB
fcis-7512	164	35	zero	zero	NUM
fcis-7512	164	36	)	)	PUNCT
fcis-7512	164	37	.	.	PUNCT
fcis-7512	165	1	the	the	DET
fcis-7512	165	2	overall	overall	ADJ
fcis-7512	165	3	experimental	experimental	ADJ
fcis-7512	165	4	results	result	NOUN
fcis-7512	165	5	are	be	AUX
fcis-7512	165	6	based	base	VERB
fcis-7512	165	7	on	on	ADP
fcis-7512	165	8	the	the	DET
fcis-7512	165	9	percentage	percentage	NOUN
fcis-7512	165	10	system	system	NOUN
fcis-7512	165	11	.	.	PUNCT
fcis-7512	166	1	the	the	DET
fcis-7512	166	2	training	training	NOUN
fcis-7512	166	3	loss	loss	NOUN
fcis-7512	166	4	curves	curve	NOUN
fcis-7512	166	5	of	of	ADP
fcis-7512	166	6	mosi	mosi	PROPN
fcis-7512	166	7	data	datum	NOUN
fcis-7512	166	8	set	set	VERB
fcis-7512	166	9	and	and	CCONJ
fcis-7512	166	10	mosei	mosei	ADJ
fcis-7512	166	11	data	datum	NOUN
fcis-7512	166	12	set	set	VERB
fcis-7512	166	13	are	be	AUX
fcis-7512	166	14	shown	show	VERB
fcis-7512	166	15	in	in	ADP
fcis-7512	166	16	figure	figure	NOUN
fcis-7512	166	17	.	.	PUNCT
fcis-7512	167	1	2	2	NUM
fcis-7512	167	2	and	and	CCONJ
fcis-7512	167	3	figure	figure	NOUN
fcis-7512	167	4	.	.	PUNCT
fcis-7512	168	1	3	3	X
fcis-7512	168	2	.	.	X
fcis-7512	168	3	71	71	NUM
fcis-7512	168	4	figure	figure	NOUN
fcis-7512	168	5	2	2	NUM
fcis-7512	168	6	.	.	PUNCT
fcis-7512	168	7	loss	loss	NOUN
fcis-7512	168	8	of	of	ADP
fcis-7512	168	9	mosi	mosi	NOUN
fcis-7512	168	10	training	training	NOUN
fcis-7512	168	11	figure	figure	NOUN
fcis-7512	168	12	3	3	NUM
fcis-7512	168	13	.	.	PUNCT
fcis-7512	169	1	loss	loss	NOUN
fcis-7512	169	2	of	of	ADP
fcis-7512	169	3	mosei	mosei	ADJ
fcis-7512	169	4	training	training	NOUN
fcis-7512	169	5	table	table	NOUN
fcis-7512	169	6	3	3	NUM
fcis-7512	169	7	.	.	PUNCT
fcis-7512	170	1	experimental	experimental	ADJ
fcis-7512	170	2	results	result	NOUN
fcis-7512	170	3	of	of	ADP
fcis-7512	170	4	mosi	mosi	PROPN
fcis-7512	170	5	dataset	dataset	ADJ
fcis-7512	170	6	model	model	NOUN
fcis-7512	170	7	mosi	mosi	PROPN
fcis-7512	170	8	mae	mae	PROPN
fcis-7512	170	9	corr	corr	PROPN
fcis-7512	170	10	acc7	acc7	PROPN
fcis-7512	170	11	acc2	acc2	PROPN
fcis-7512	170	12	f1	f1	PROPN
fcis-7512	170	13	tfn	tfn	NOUN
fcis-7512	170	14	90.10	90.10	NUM
fcis-7512	170	15	96.80	96.80	NUM
fcis-7512	170	16	34.90	34.90	NUM
fcis-7512	170	17	-/80.80	-/80.80	PUNCT
fcis-7512	170	18	-/80.70	-/80.70	PROPN
fcis-7512	170	19	lmf	lmf	PROPN
fcis-7512	170	20	91.70	91.70	NUM
fcis-7512	170	21	69.50	69.50	NUM
fcis-7512	170	22	33.20	33.20	NUM
fcis-7512	170	23	-/82.50	-/82.50	PROPN
fcis-7512	170	24	-/82.40	-/82.40	PROPN
fcis-7512	170	25	mfm	mfm	NOUN
fcis-7512	170	26	87.70	87.70	NUM
fcis-7512	170	27	70.60	70.60	NUM
fcis-7512	170	28	35.40	35.40	NUM
fcis-7512	170	29	-/81.70	-/81.70	PROPN
fcis-7512	170	30	-/81.60	-/81.60	PROPN
fcis-7512	170	31	raven	raven	VERB
fcis-7512	170	32	86.20	86.20	NUM
fcis-7512	170	33	71.40	71.40	NUM
fcis-7512	170	34	39.00	39.00	NUM
fcis-7512	170	35	-/83.00	-/83.00	PROPN
fcis-7512	170	36	-/83.00	-/83.00	PROPN
fcis-7512	170	37	mult	mult	PROPN
fcis-7512	170	38	86.10	86.10	NUM
fcis-7512	170	39	71.10	71.10	NUM
fcis-7512	170	40	81.50/84.10	81.50/84.10	NUM
fcis-7512	170	41	80.60/83.90	80.60/83.90	NUM
fcis-7512	170	42	misa	misa	NOUN
fcis-7512	170	43	80.40	80.40	NUM
fcis-7512	170	44	76.40	76.40	NUM
fcis-7512	170	45	80.79/82.10	80.79/82.10	NUM
fcis-7512	170	46	80.77/82.03	80.77/82.03	NUM
fcis-7512	170	47	magbert	magbert	NOUN
fcis-7512	170	48	72.70	72.70	NUM
fcis-7512	170	49	78.10	78.10	NUM
fcis-7512	170	50	43.62	43.62	NUM
fcis-7512	170	51	82.37/84.43	82.37/84.43	NUM
fcis-7512	170	52	82.50/84.61	82.50/84.61	NUM
fcis-7512	170	53	self	self	NOUN
fcis-7512	170	54	-	-	PUNCT
fcis-7512	170	55	mm	mm	PROPN
fcis-7512	170	56	71.20	71.20	NUM
fcis-7512	170	57	79.50	79.50	NUM
fcis-7512	170	58	45.79	45.79	NUM
fcis-7512	170	59	84.14/84.77	84.14/84.77	NUM
fcis-7512	170	60	82.68/84.91	82.68/84.91	NUM
fcis-7512	170	61	model	model	NOUN
fcis-7512	170	62	72.10	72.10	NUM
fcis-7512	170	63	79.60	79.60	NUM
fcis-7512	170	64	45.33	45.33	NUM
fcis-7512	170	65	83.21/85.18	83.21/85.18	NUM
fcis-7512	170	66	83.11/85.15	83.11/85.15	NUM
fcis-7512	170	67	table	table	NOUN
fcis-7512	170	68	4	4	NUM
fcis-7512	170	69	.	.	PUNCT
fcis-7512	171	1	experimental	experimental	ADJ
fcis-7512	171	2	results	result	NOUN
fcis-7512	171	3	of	of	ADP
fcis-7512	171	4	moesi	moesi	PROPN
fcis-7512	171	5	dataset	dataset	PROPN
fcis-7512	171	6	model	model	PROPN
fcis-7512	171	7	mosei	mosei	PROPN
fcis-7512	171	8	mae	mae	PROPN
fcis-7512	171	9	corr	corr	PROPN
fcis-7512	171	10	acc7	acc7	PROPN
fcis-7512	171	11	acc2	acc2	PROPN
fcis-7512	171	12	f1	f1	PROPN
fcis-7512	171	13	tfn	tfn	NOUN
fcis-7512	171	14	59.30	59.30	NUM
fcis-7512	171	15	70.00	70.00	NUM
fcis-7512	171	16	50.20	50.20	NUM
fcis-7512	171	17	-/82.50	-/82.50	PROPN
fcis-7512	171	18	-/82.10	-/82.10	PROPN
fcis-7512	171	19	lmf	lmf	PROPN
fcis-7512	171	20	62.30	62.30	NUM
fcis-7512	171	21	67.70	67.70	NUM
fcis-7512	171	22	48.00	48.00	NUM
fcis-7512	171	23	-/82.00	-/82.00	PROPN
fcis-7512	171	24	-/82.10	-/82.10	PROPN
fcis-7512	171	25	mfm	mfm	NOUN
fcis-7512	171	26	56.80	56.80	NUM
fcis-7512	171	27	71.70	71.70	NUM
fcis-7512	171	28	51.30	51.30	NUM
fcis-7512	171	29	-/84.40	-/84.40	PROPN
fcis-7512	171	30	-/84.30	-/84.30	PROPN
fcis-7512	171	31	raven	raven	VERB
fcis-7512	171	32	56.50	56.50	NUM
fcis-7512	171	33	71.30	71.30	NUM
fcis-7512	171	34	-/81.70	-/81.70	PROPN
fcis-7512	171	35	-/84.20	-/84.20	PROPN
fcis-7512	171	36	-/84.20	-/84.20	PROPN
fcis-7512	171	37	mult	mult	PROPN
fcis-7512	171	38	58.00	58.00	NUM
fcis-7512	171	39	70.30	70.30	NUM
fcis-7512	171	40	-/82.50	-/82.50	PROPN
fcis-7512	171	41	-/82.30	-/82.30	PROPN
fcis-7512	171	42	misa	misa	PROPN
fcis-7512	172	1	56.80	56.80	NUM
fcis-7512	172	2	72.40	72.40	NUM
fcis-7512	172	3	82.59/84.23	82.59/84.23	NUM
fcis-7512	172	4	82.53/85.30	82.53/85.30	NUM
fcis-7512	172	5	magbert	magbert	NOUN
fcis-7512	172	6	54.30	54.30	NUM
fcis-7512	172	7	75.50	75.50	NUM
fcis-7512	172	8	52.67	52.67	NUM
fcis-7512	172	9	82.51/84.82	82.51/84.82	NUM
fcis-7512	172	10	82.77/84.71	82.77/84.71	NUM
fcis-7512	172	11	self	self	NOUN
fcis-7512	172	12	-	-	PUNCT
fcis-7512	172	13	mm	mm	PROPN
fcis-7512	172	14	52.90	52.90	NUM
fcis-7512	172	15	76.70	76.70	NUM
fcis-7512	172	16	53.46	53.46	NUM
fcis-7512	172	17	82.68/84.96	82.68/84.96	NUM
fcis-7512	172	18	82.95/84.93	82.95/84.93	NUM
fcis-7512	172	19	model	model	NOUN
fcis-7512	172	20	52.40	52.40	NUM
fcis-7512	172	21	77.20	77.20	NUM
fcis-7512	172	22	54.31	54.31	NUM
fcis-7512	172	23	82.18/85.30	82.18/85.30	NUM
fcis-7512	172	24	83.01/85.24	83.01/85.24	NUM
fcis-7512	172	25	3.5	3.5	NUM
fcis-7512	172	26	.	.	PUNCT
fcis-7512	173	1	analysis	analysis	NOUN
fcis-7512	173	2	of	of	ADP
fcis-7512	173	3	experimental	experimental	ADJ
fcis-7512	173	4	results	result	NOUN
fcis-7512	173	5	according	accord	VERB
fcis-7512	173	6	to	to	ADP
fcis-7512	173	7	the	the	DET
fcis-7512	173	8	data	datum	NOUN
fcis-7512	173	9	results	result	NOUN
fcis-7512	173	10	in	in	ADP
fcis-7512	173	11	table	table	NOUN
fcis-7512	173	12	3	3	NUM
fcis-7512	173	13	and	and	CCONJ
fcis-7512	173	14	table	table	NOUN
fcis-7512	173	15	4	4	NUM
fcis-7512	173	16	,	,	PUNCT
fcis-7512	173	17	for	for	ADP
fcis-7512	173	18	classification	classification	NOUN
fcis-7512	173	19	tasks	task	NOUN
fcis-7512	173	20	,	,	PUNCT
fcis-7512	173	21	compared	compare	VERB
fcis-7512	173	22	with	with	ADP
fcis-7512	173	23	the	the	DET
fcis-7512	173	24	earlier	early	ADJ
fcis-7512	173	25	tfn	tfn	NOUN
fcis-7512	173	26	,	,	PUNCT
fcis-7512	173	27	lfn	lfn	PROPN
fcis-7512	173	28	and	and	CCONJ
fcis-7512	173	29	other	other	ADJ
fcis-7512	173	30	reference	reference	NOUN
fcis-7512	173	31	models	model	NOUN
fcis-7512	173	32	,	,	PUNCT
fcis-7512	173	33	the	the	DET
fcis-7512	173	34	model	model	NOUN
fcis-7512	173	35	proposed	propose	VERB
fcis-7512	173	36	in	in	ADP
fcis-7512	173	37	this	this	DET
fcis-7512	173	38	paper	paper	NOUN
fcis-7512	173	39	has	have	VERB
fcis-7512	173	40	a	a	DET
fcis-7512	173	41	3%-5	3%-5	NUM
fcis-7512	173	42	%	%	NOUN
fcis-7512	173	43	improvement	improvement	NOUN
fcis-7512	173	44	in	in	ADP
fcis-7512	173	45	accuracy	accuracy	NOUN
fcis-7512	173	46	and	and	CCONJ
fcis-7512	173	47	f1	f1	NOUN
fcis-7512	173	48	score	score	NOUN
fcis-7512	173	49	,	,	PUNCT
fcis-7512	173	50	which	which	PRON
fcis-7512	173	51	is	be	AUX
fcis-7512	173	52	still	still	ADV
fcis-7512	173	53	better	well	ADJ
fcis-7512	173	54	than	than	ADP
fcis-7512	173	55	the	the	DET
fcis-7512	173	56	reference	reference	NOUN
fcis-7512	173	57	model	model	NOUN
fcis-7512	173	58	that	that	PRON
fcis-7512	173	59	also	also	ADV
fcis-7512	173	60	uses	use	VERB
fcis-7512	173	61	bert	bert	PROPN
fcis-7512	173	62	as	as	ADP
fcis-7512	173	63	the	the	DET
fcis-7512	173	64	feature	feature	NOUN
fcis-7512	173	65	extraction	extraction	NOUN
fcis-7512	173	66	of	of	ADP
fcis-7512	173	67	text	text	NOUN
fcis-7512	173	68	data	datum	NOUN
fcis-7512	173	69	.	.	PUNCT
fcis-7512	174	1	for	for	ADP
fcis-7512	174	2	regression	regression	NOUN
fcis-7512	174	3	tasks	task	NOUN
fcis-7512	174	4	,	,	PUNCT
fcis-7512	174	5	the	the	DET
fcis-7512	174	6	model	model	NOUN
fcis-7512	174	7	proposed	propose	VERB
fcis-7512	174	8	in	in	ADP
fcis-7512	174	9	this	this	DET
fcis-7512	174	10	paper	paper	NOUN
fcis-7512	174	11	is	be	AUX
fcis-7512	174	12	better	well	ADJ
fcis-7512	174	13	than	than	ADP
fcis-7512	174	14	the	the	DET
fcis-7512	174	15	benchmark	benchmark	NOUN
fcis-7512	174	16	model	model	NOUN
fcis-7512	174	17	in	in	ADP
fcis-7512	174	18	terms	term	NOUN
fcis-7512	174	19	of	of	ADP
fcis-7512	174	20	mean	mean	ADJ
fcis-7512	174	21	absolute	absolute	ADJ
fcis-7512	174	22	error	error	NOUN
fcis-7512	174	23	mae	mae	PROPN
fcis-7512	174	24	and	and	CCONJ
fcis-7512	174	25	correlation	correlation	NOUN
fcis-7512	174	26	corr	corr	PROPN
fcis-7512	174	27	.	.	PUNCT
fcis-7512	175	1	in	in	ADP
fcis-7512	175	2	terms	term	NOUN
fcis-7512	175	3	of	of	ADP
fcis-7512	175	4	the	the	DET
fcis-7512	175	5	two	two	NUM
fcis-7512	175	6	common	common	ADJ
fcis-7512	175	7	multimodal	multimodal	NOUN
fcis-7512	175	8	sentiment	sentiment	NOUN
fcis-7512	175	9	analysis	analysis	NOUN
fcis-7512	175	10	data	data	NOUN
fcis-7512	175	11	sets	set	NOUN
fcis-7512	175	12	,	,	PUNCT
fcis-7512	175	13	as	as	SCONJ
fcis-7512	175	14	the	the	DET
fcis-7512	175	15	amount	amount	NOUN
fcis-7512	175	16	of	of	ADP
fcis-7512	175	17	data	datum	NOUN
fcis-7512	175	18	in	in	ADP
fcis-7512	175	19	mosei	mosei	ADJ
fcis-7512	175	20	data	datum	NOUN
fcis-7512	175	21	set	set	VERB
fcis-7512	175	22	is	be	AUX
fcis-7512	175	23	much	much	ADV
fcis-7512	175	24	larger	large	ADJ
fcis-7512	175	25	than	than	ADP
fcis-7512	175	26	that	that	PRON
fcis-7512	175	27	in	in	ADP
fcis-7512	175	28	mosi	mosi	PROPN
fcis-7512	175	29	data	datum	NOUN
fcis-7512	175	30	set	set	VERB
fcis-7512	175	31	,	,	PUNCT
fcis-7512	175	32	experimental	experimental	ADJ
fcis-7512	175	33	results	result	NOUN
fcis-7512	175	34	show	show	VERB
fcis-7512	175	35	that	that	SCONJ
fcis-7512	175	36	the	the	DET
fcis-7512	175	37	model	model	NOUN
fcis-7512	175	38	proposed	propose	VERB
fcis-7512	175	39	in	in	ADP
fcis-7512	175	40	this	this	DET
fcis-7512	175	41	paper	paper	NOUN
fcis-7512	175	42	has	have	AUX
fcis-7512	175	43	improved	improve	VERB
fcis-7512	175	44	in	in	ADP
fcis-7512	175	45	correlation	correlation	NOUN
fcis-7512	175	46	,	,	PUNCT
fcis-7512	175	47	accuracy	accuracy	NOUN
fcis-7512	175	48	and	and	CCONJ
fcis-7512	175	49	f1	f1	NOUN
fcis-7512	175	50	score	score	NOUN
fcis-7512	175	51	compared	compare	VERB
fcis-7512	175	52	with	with	ADP
fcis-7512	175	53	before	before	ADV
fcis-7512	175	54	in	in	ADP
fcis-7512	175	55	mosi	mosi	PROPN
fcis-7512	175	56	data	datum	NOUN
fcis-7512	175	57	set	set	VERB
fcis-7512	175	58	,	,	PUNCT
fcis-7512	175	59	while	while	SCONJ
fcis-7512	175	60	its	its	PRON
fcis-7512	175	61	effect	effect	NOUN
fcis-7512	175	62	in	in	ADP
fcis-7512	175	63	average	average	ADJ
fcis-7512	175	64	absolute	absolute	ADJ
fcis-7512	175	65	error	error	NOUN
fcis-7512	175	66	is	be	AUX
fcis-7512	175	67	not	not	PART
fcis-7512	175	68	so	so	ADV
fcis-7512	175	69	good	good	ADJ
fcis-7512	175	70	.	.	PUNCT
fcis-7512	176	1	it	it	PRON
fcis-7512	176	2	indicates	indicate	VERB
fcis-7512	176	3	that	that	SCONJ
fcis-7512	176	4	a	a	DET
fcis-7512	176	5	small	small	ADJ
fcis-7512	176	6	amount	amount	NOUN
fcis-7512	176	7	of	of	ADP
fcis-7512	176	8	data	datum	NOUN
fcis-7512	176	9	will	will	AUX
fcis-7512	176	10	affect	affect	VERB
fcis-7512	176	11	the	the	DET
fcis-7512	176	12	overall	overall	ADJ
fcis-7512	176	13	fitting	fitting	ADJ
fcis-7512	176	14	effect	effect	NOUN
fcis-7512	176	15	of	of	ADP
fcis-7512	176	16	the	the	DET
fcis-7512	176	17	data	datum	NOUN
fcis-7512	176	18	and	and	CCONJ
fcis-7512	176	19	lead	lead	VERB
fcis-7512	176	20	to	to	ADP
fcis-7512	176	21	larger	large	ADJ
fcis-7512	176	22	overall	overall	ADJ
fcis-7512	176	23	error	error	NOUN
fcis-7512	176	24	.	.	PUNCT
fcis-7512	177	1	however	however	ADV
fcis-7512	177	2	,	,	PUNCT
fcis-7512	177	3	for	for	ADP
fcis-7512	177	4	the	the	DET
fcis-7512	177	5	mosei	mosei	ADJ
fcis-7512	177	6	data	datum	NOUN
fcis-7512	177	7	set	set	VERB
fcis-7512	177	8	,	,	PUNCT
fcis-7512	177	9	the	the	DET
fcis-7512	177	10	data	datum	NOUN
fcis-7512	177	11	set	set	NOUN
fcis-7512	177	12	has	have	VERB
fcis-7512	177	13	a	a	DET
fcis-7512	177	14	relatively	relatively	ADV
fcis-7512	177	15	large	large	ADJ
fcis-7512	177	16	amount	amount	NOUN
fcis-7512	177	17	of	of	ADP
fcis-7512	177	18	data	datum	NOUN
fcis-7512	177	19	,	,	PUNCT
fcis-7512	177	20	so	so	CCONJ
fcis-7512	177	21	the	the	DET
fcis-7512	177	22	improvement	improvement	NOUN
fcis-7512	177	23	effect	effect	NOUN
fcis-7512	177	24	is	be	AUX
fcis-7512	177	25	obvious	obvious	ADJ
fcis-7512	177	26	in	in	ADP
fcis-7512	177	27	all	all	DET
fcis-7512	177	28	aspects	aspect	NOUN
fcis-7512	177	29	.	.	PUNCT
fcis-7512	178	1	to	to	PART
fcis-7512	178	2	sum	sum	VERB
fcis-7512	178	3	up	up	ADP
fcis-7512	178	4	,	,	PUNCT
fcis-7512	178	5	the	the	DET
fcis-7512	178	6	experimental	experimental	ADJ
fcis-7512	178	7	improvement	improvement	NOUN
fcis-7512	178	8	effect	effect	NOUN
fcis-7512	178	9	of	of	ADP
fcis-7512	178	10	the	the	DET
fcis-7512	178	11	model	model	NOUN
fcis-7512	178	12	proposed	propose	VERB
fcis-7512	178	13	in	in	ADP
fcis-7512	178	14	this	this	DET
fcis-7512	178	15	paper	paper	NOUN
fcis-7512	178	16	on	on	ADP
fcis-7512	178	17	the	the	DET
fcis-7512	178	18	mosei	mosei	ADJ
fcis-7512	178	19	data	datum	NOUN
fcis-7512	178	20	set	set	VERB
fcis-7512	178	21	is	be	AUX
fcis-7512	178	22	better	well	ADJ
fcis-7512	178	23	than	than	ADP
fcis-7512	178	24	that	that	PRON
fcis-7512	178	25	on	on	ADP
fcis-7512	178	26	the	the	DET
fcis-7512	178	27	mosi	mosi	PROPN
fcis-7512	178	28	data	datum	NOUN
fcis-7512	178	29	.	.	PUNCT
fcis-7512	179	1	it	it	PRON
fcis-7512	179	2	can	can	AUX
fcis-7512	179	3	be	be	AUX
fcis-7512	179	4	shown	show	VERB
fcis-7512	179	5	that	that	SCONJ
fcis-7512	179	6	data	datum	NOUN
fcis-7512	179	7	is	be	AUX
fcis-7512	179	8	one	one	NUM
fcis-7512	179	9	of	of	ADP
fcis-7512	179	10	the	the	DET
fcis-7512	179	11	important	important	ADJ
fcis-7512	179	12	factors	factor	NOUN
fcis-7512	179	13	affecting	affect	VERB
fcis-7512	179	14	the	the	DET
fcis-7512	179	15	model	model	NOUN
fcis-7512	179	16	effect	effect	NOUN
fcis-7512	179	17	.	.	PUNCT
fcis-7512	180	1	3.6	3.6	NUM
fcis-7512	180	2	.	.	PUNCT
fcis-7512	180	3	ablation	ablation	NOUN
fcis-7512	180	4	experiment	experiment	NOUN
fcis-7512	180	5	according	accord	VERB
fcis-7512	180	6	to	to	ADP
fcis-7512	180	7	the	the	DET
fcis-7512	180	8	experimental	experimental	ADJ
fcis-7512	180	9	results	result	NOUN
fcis-7512	180	10	,	,	PUNCT
fcis-7512	180	11	it	it	PRON
fcis-7512	180	12	can	can	AUX
fcis-7512	180	13	be	be	AUX
fcis-7512	180	14	found	find	VERB
fcis-7512	180	15	that	that	SCONJ
fcis-7512	180	16	the	the	DET
fcis-7512	180	17	performance	performance	NOUN
fcis-7512	180	18	of	of	ADP
fcis-7512	180	19	the	the	DET
fcis-7512	180	20	model	model	NOUN
fcis-7512	180	21	is	be	AUX
fcis-7512	180	22	improved	improve	VERB
fcis-7512	180	23	by	by	ADP
fcis-7512	180	24	introducing	introduce	VERB
fcis-7512	180	25	intermodal	intermodal	ADJ
fcis-7512	180	26	self	self	NOUN
fcis-7512	180	27	-	-	PUNCT
fcis-7512	180	28	attention	attention	NOUN
fcis-7512	180	29	mechanism	mechanism	NOUN
fcis-7512	180	30	,	,	PUNCT
fcis-7512	180	31	single	single	ADJ
fcis-7512	180	32	-	-	PUNCT
fcis-7512	180	33	mode	mode	NOUN
fcis-7512	180	34	selfattention	selfattention	NOUN
fcis-7512	180	35	mechanism	mechanism	NOUN
fcis-7512	180	36	and	and	CCONJ
fcis-7512	180	37	loss	loss	NOUN
fcis-7512	180	38	weight	weight	NOUN
fcis-7512	180	39	allocation	allocation	NOUN
fcis-7512	180	40	.	.	PUNCT
fcis-7512	181	1	the	the	DET
fcis-7512	181	2	following	follow	VERB
fcis-7512	181	3	ablation	ablation	NOUN
fcis-7512	181	4	experiments	experiment	NOUN
fcis-7512	181	5	were	be	AUX
fcis-7512	181	6	conducted	conduct	VERB
fcis-7512	181	7	for	for	ADP
fcis-7512	181	8	these	these	DET
fcis-7512	181	9	three	three	NUM
fcis-7512	181	10	aspects	aspect	NOUN
fcis-7512	181	11	,	,	PUNCT
fcis-7512	181	12	and	and	CCONJ
fcis-7512	181	13	the	the	DET
fcis-7512	181	14	experimental	experimental	ADJ
fcis-7512	181	15	results	result	NOUN
fcis-7512	181	16	are	be	AUX
fcis-7512	181	17	shown	show	VERB
fcis-7512	181	18	in	in	ADP
fcis-7512	181	19	table	table	NOUN
fcis-7512	181	20	5	5	NUM
fcis-7512	181	21	and	and	CCONJ
fcis-7512	181	22	table	table	NOUN
fcis-7512	181	23	6	6	NUM
fcis-7512	181	24	.	.	PUNCT
fcis-7512	182	1	note	note	NOUN
fcis-7512	182	2	:	:	PUNCT
fcis-7512	182	3	bi	bi	NOUN
fcis-7512	182	4	represents	represent	VERB
fcis-7512	182	5	the	the	DET
fcis-7512	182	6	intermodal	intermodal	ADJ
fcis-7512	182	7	attention	attention	NOUN
fcis-7512	182	8	72	72	NUM
fcis-7512	182	9	mechanism	mechanism	NOUN
fcis-7512	182	10	,	,	PUNCT
fcis-7512	182	11	uni	uni	PROPN
fcis-7512	182	12	represents	represent	VERB
fcis-7512	182	13	the	the	DET
fcis-7512	182	14	unimodal	unimodal	ADJ
fcis-7512	182	15	attention	attention	NOUN
fcis-7512	182	16	mechanism	mechanism	NOUN
fcis-7512	182	17	,	,	PUNCT
fcis-7512	182	18	and	and	CCONJ
fcis-7512	182	19	weight	weight	NOUN
fcis-7512	182	20	represents	represent	VERB
fcis-7512	182	21	the	the	DET
fcis-7512	182	22	weight	weight	NOUN
fcis-7512	182	23	loss	loss	NOUN
fcis-7512	182	24	mechanism	mechanism	NOUN
fcis-7512	182	25	.	.	PUNCT
fcis-7512	183	1	other	other	ADJ
fcis-7512	183	2	data	datum	NOUN
fcis-7512	183	3	descriptions	description	NOUN
fcis-7512	183	4	are	be	AUX
fcis-7512	183	5	the	the	DET
fcis-7512	183	6	same	same	ADJ
fcis-7512	183	7	as	as	ADP
fcis-7512	183	8	in	in	ADP
fcis-7512	183	9	table	table	NOUN
fcis-7512	183	10	3	3	NUM
fcis-7512	183	11	.	.	PUNCT
fcis-7512	184	1	the	the	DET
fcis-7512	184	2	whole	whole	ADJ
fcis-7512	184	3	experimental	experimental	ADJ
fcis-7512	184	4	results	result	NOUN
fcis-7512	184	5	are	be	AUX
fcis-7512	184	6	based	base	VERB
fcis-7512	184	7	on	on	ADP
fcis-7512	184	8	the	the	DET
fcis-7512	184	9	percentage	percentage	NOUN
fcis-7512	184	10	system	system	NOUN
fcis-7512	184	11	.	.	PUNCT
fcis-7512	185	1	table	table	NOUN
fcis-7512	185	2	5	5	NUM
fcis-7512	185	3	.	.	PUNCT
fcis-7512	185	4	ablation	ablation	NOUN
fcis-7512	185	5	study	study	NOUN
fcis-7512	185	6	results	result	NOUN
fcis-7512	185	7	of	of	ADP
fcis-7512	185	8	mosi	mosi	NOUN
fcis-7512	185	9	dataset	dataset	ADJ
fcis-7512	185	10	method	method	NOUN
fcis-7512	185	11	mosi	mosi	PROPN
fcis-7512	185	12	mae	mae	PROPN
fcis-7512	185	13	corr	corr	PROPN
fcis-7512	185	14	acc7	acc7	PROPN
fcis-7512	185	15	acc2	acc2	PROPN
fcis-7512	185	16	f1	f1	NOUN
fcis-7512	185	17	bi+uni	bi+uni	VERB
fcis-7512	185	18	72.70	72.70	NUM
fcis-7512	185	19	76.40	76.40	NUM
fcis-7512	185	20	45.54	45.54	NUM
fcis-7512	185	21	82.54/84.36	82.54/84.36	NUM
fcis-7512	185	22	82.54/84.39	82.54/84.39	NUM
fcis-7512	185	23	bi+weight	bi+weight	NOUN
fcis-7512	185	24	72.40	72.40	NUM
fcis-7512	185	25	79.40	79.40	NUM
fcis-7512	185	26	44.92	44.92	NUM
fcis-7512	185	27	82.32/84.42	82.32/84.42	NUM
fcis-7512	185	28	82.59/84.44	82.59/84.44	NUM
fcis-7512	185	29	uni+weight	uni+weight	NOUN
fcis-7512	185	30	73.60	73.60	NUM
fcis-7512	185	31	78.80	78.80	NUM
fcis-7512	185	32	44.75	44.75	NUM
fcis-7512	185	33	82.68/84.39	82.68/84.39	NUM
fcis-7512	185	34	82.65/84.41	82.65/84.41	NUM
fcis-7512	185	35	bi	bi	NOUN
fcis-7512	185	36	72.70	72.70	NUM
fcis-7512	185	37	76.40	76.40	NUM
fcis-7512	185	38	45.28	45.28	NUM
fcis-7512	185	39	82.62/84.27	82.62/84.27	NUM
fcis-7512	185	40	82.60/84.30	82.60/84.30	NUM
fcis-7512	185	41	uni	uni	PROPN
fcis-7512	185	42	72.30	72.30	NUM
fcis-7512	185	43	79.10	79.10	NUM
fcis-7512	185	44	46.30	46.30	NUM
fcis-7512	185	45	82.98/84.85	82.98/84.85	NUM
fcis-7512	185	46	82.9/84.84	82.9/84.84	NUM
fcis-7512	185	47	weight	weight	NOUN
fcis-7512	185	48	73.60	73.60	NUM
fcis-7512	185	49	79.00	79.00	NUM
fcis-7512	185	50	44.05	44.05	NUM
fcis-7512	185	51	82.80/84.61	82.80/84.61	NUM
fcis-7512	185	52	82.75/84.61	82.75/84.61	NUM
fcis-7512	185	53	model	model	NOUN
fcis-7512	185	54	72.10	72.10	NUM
fcis-7512	185	55	79.60	79.60	NUM
fcis-7512	185	56	45.33	45.33	NUM
fcis-7512	185	57	83.21/85.18	83.21/85.18	NUM
fcis-7512	185	58	83.11/85.15	83.11/85.15	NUM
fcis-7512	185	59	table	table	NOUN
fcis-7512	185	60	6	6	NUM
fcis-7512	185	61	.	.	PUNCT
fcis-7512	185	62	ablation	ablation	NOUN
fcis-7512	185	63	study	study	NOUN
fcis-7512	185	64	results	result	NOUN
fcis-7512	185	65	of	of	ADP
fcis-7512	185	66	mosi	mosi	NOUN
fcis-7512	185	67	dataset	dataset	ADJ
fcis-7512	185	68	method	method	NOUN
fcis-7512	185	69	mosi	mosi	PROPN
fcis-7512	185	70	mae	mae	PROPN
fcis-7512	185	71	corr	corr	PROPN
fcis-7512	185	72	acc7	acc7	PROPN
fcis-7512	185	73	acc2	acc2	PROPN
fcis-7512	185	74	f1	f1	NOUN
fcis-7512	185	75	bi+uni	bi+uni	VERB
fcis-7512	185	76	53.50	53.50	NUM
fcis-7512	185	77	76.00	76.00	NUM
fcis-7512	185	78	53.35	53.35	NUM
fcis-7512	185	79	81.27/84.52	81.27/84.52	NUM
fcis-7512	185	80	81.67/84.44	81.67/84.44	NUM
fcis-7512	185	81	bi+weight	bi+weight	NOUN
fcis-7512	185	82	52.60	52.60	NUM
fcis-7512	185	83	77.10	77.10	NUM
fcis-7512	185	84	53.84	53.84	NUM
fcis-7512	185	85	77.92/84.16	77.92/84.16	NUM
fcis-7512	185	86	78.81/84.29	78.81/84.29	NUM
fcis-7512	185	87	uni+weight	uni+weight	NOUN
fcis-7512	185	88	53.10	53.10	NUM
fcis-7512	185	89	76.40	76.40	NUM
fcis-7512	185	90	53.91	53.91	NUM
fcis-7512	185	91	81.95/85.00	81.95/85.00	NUM
fcis-7512	185	92	82.34/84.92	82.34/84.92	NUM
fcis-7512	185	93	bi	bi	NOUN
fcis-7512	185	94	53.20	53.20	NUM
fcis-7512	185	95	76.40	76.40	NUM
fcis-7512	185	96	53.18	53.18	NUM
fcis-7512	185	97	81.76/84.65	81.76/84.65	NUM
fcis-7512	185	98	82.08/84.51	82.08/84.51	NUM
fcis-7512	185	99	uni	uni	PROPN
fcis-7512	185	100	53.60	53.60	NUM
fcis-7512	185	101	76.00	76.00	NUM
fcis-7512	185	102	53.41	53.41	NUM
fcis-7512	185	103	81.95/84.63	81.95/84.63	NUM
fcis-7512	185	104	82.22/84.45	82.22/84.45	NUM
fcis-7512	185	105	weight	weight	NOUN
fcis-7512	185	106	53.40	53.40	NUM
fcis-7512	185	107	76.20	76.20	NUM
fcis-7512	185	108	53.40	53.40	NUM
fcis-7512	185	109	77.72/83.14	77.72/83.14	NUM
fcis-7512	185	110	78.52/83.24	78.52/83.24	NUM
fcis-7512	185	111	model	model	NOUN
fcis-7512	185	112	52.40	52.40	NUM
fcis-7512	185	113	77.20	77.20	NUM
fcis-7512	185	114	54.31	54.31	NUM
fcis-7512	185	115	82.18/85.30	82.18/85.30	NUM
fcis-7512	185	116	83.01/85.24	83.01/85.24	NUM
fcis-7512	185	117	3.7	3.7	NUM
fcis-7512	185	118	.	.	PUNCT
fcis-7512	186	1	analysis	analysis	NOUN
fcis-7512	186	2	of	of	ADP
fcis-7512	186	3	ablation	ablation	NOUN
fcis-7512	186	4	results	result	NOUN
fcis-7512	186	5	according	accord	VERB
fcis-7512	186	6	to	to	ADP
fcis-7512	186	7	the	the	DET
fcis-7512	186	8	experimental	experimental	ADJ
fcis-7512	186	9	results	result	NOUN
fcis-7512	186	10	of	of	ADP
fcis-7512	186	11	ablation	ablation	NOUN
fcis-7512	186	12	implementation	implementation	NOUN
fcis-7512	186	13	in	in	ADP
fcis-7512	186	14	table	table	NOUN
fcis-7512	186	15	5	5	NUM
fcis-7512	186	16	and	and	CCONJ
fcis-7512	186	17	table	table	NOUN
fcis-7512	186	18	6	6	NUM
fcis-7512	186	19	,	,	PUNCT
fcis-7512	186	20	for	for	ADP
fcis-7512	186	21	classification	classification	NOUN
fcis-7512	186	22	tasks	task	NOUN
fcis-7512	186	23	,	,	PUNCT
fcis-7512	186	24	using	use	VERB
fcis-7512	186	25	only	only	ADV
fcis-7512	186	26	the	the	DET
fcis-7512	186	27	attention	attention	NOUN
fcis-7512	186	28	mechanism	mechanism	NOUN
fcis-7512	186	29	on	on	ADP
fcis-7512	186	30	a	a	DET
fcis-7512	186	31	relatively	relatively	ADV
fcis-7512	186	32	small	small	ADJ
fcis-7512	186	33	mosi	mosi	NOUN
fcis-7512	186	34	dataset	dataset	NOUN
fcis-7512	186	35	can	can	AUX
fcis-7512	186	36	better	well	ADV
fcis-7512	186	37	improve	improve	VERB
fcis-7512	186	38	the	the	DET
fcis-7512	186	39	accuracy	accuracy	NOUN
fcis-7512	186	40	of	of	ADP
fcis-7512	186	41	classification	classification	NOUN
fcis-7512	186	42	.	.	PUNCT
fcis-7512	187	1	in	in	ADP
fcis-7512	187	2	addition	addition	NOUN
fcis-7512	187	3	,	,	PUNCT
fcis-7512	187	4	the	the	DET
fcis-7512	187	5	mean	mean	ADJ
fcis-7512	187	6	absolute	absolute	ADJ
fcis-7512	187	7	error	error	NOUN
fcis-7512	187	8	mae	mae	PROPN
fcis-7512	187	9	and	and	CCONJ
fcis-7512	187	10	correlation	correlation	NOUN
fcis-7512	187	11	values	value	NOUN
fcis-7512	187	12	in	in	ADP
fcis-7512	187	13	the	the	DET
fcis-7512	187	14	regression	regression	NOUN
fcis-7512	187	15	task	task	NOUN
fcis-7512	187	16	are	be	AUX
fcis-7512	187	17	only	only	ADV
fcis-7512	187	18	0.2%~0.7	0.2%~0.7	NUM
fcis-7512	187	19	%	%	NOUN
fcis-7512	187	20	difference	difference	NOUN
fcis-7512	187	21	.	.	PUNCT
fcis-7512	188	1	for	for	ADP
fcis-7512	188	2	mosei	mosei	ADJ
fcis-7512	188	3	dataset	dataset	NOUN
fcis-7512	188	4	with	with	ADP
fcis-7512	188	5	large	large	ADJ
fcis-7512	188	6	data	datum	NOUN
fcis-7512	188	7	set	set	VERB
fcis-7512	188	8	size	size	NOUN
fcis-7512	188	9	,	,	PUNCT
fcis-7512	188	10	the	the	DET
fcis-7512	188	11	effect	effect	NOUN
fcis-7512	188	12	of	of	ADP
fcis-7512	188	13	the	the	DET
fcis-7512	188	14	combined	combine	VERB
fcis-7512	188	15	use	use	NOUN
fcis-7512	188	16	of	of	ADP
fcis-7512	188	17	attention	attention	NOUN
fcis-7512	188	18	mechanism	mechanism	NOUN
fcis-7512	188	19	and	and	CCONJ
fcis-7512	188	20	weight	weight	NOUN
fcis-7512	188	21	loss	loss	NOUN
fcis-7512	188	22	mechanism	mechanism	NOUN
fcis-7512	188	23	was	be	AUX
fcis-7512	188	24	significantly	significantly	ADV
fcis-7512	188	25	improved	improve	VERB
fcis-7512	188	26	.	.	PUNCT
fcis-7512	189	1	for	for	ADP
fcis-7512	189	2	classification	classification	NOUN
fcis-7512	189	3	tasks	task	NOUN
fcis-7512	189	4	,	,	PUNCT
fcis-7512	189	5	the	the	DET
fcis-7512	189	6	accuracy	accuracy	NOUN
fcis-7512	189	7	of	of	ADP
fcis-7512	189	8	seven	seven	NUM
fcis-7512	189	9	classification	classification	NOUN
fcis-7512	189	10	tasks	task	NOUN
fcis-7512	189	11	,	,	PUNCT
fcis-7512	189	12	the	the	DET
fcis-7512	189	13	accuracy	accuracy	NOUN
fcis-7512	189	14	of	of	ADP
fcis-7512	189	15	two	two	NUM
fcis-7512	189	16	classification	classification	NOUN
fcis-7512	189	17	tasks	task	NOUN
fcis-7512	189	18	and	and	CCONJ
fcis-7512	189	19	f1	f1	NOUN
fcis-7512	189	20	score	score	NOUN
fcis-7512	189	21	were	be	AUX
fcis-7512	189	22	increased	increase	VERB
fcis-7512	189	23	by	by	ADP
fcis-7512	189	24	1%-4	1%-4	NUM
fcis-7512	189	25	%	%	NOUN
fcis-7512	189	26	.	.	PUNCT
fcis-7512	190	1	in	in	ADP
fcis-7512	190	2	regression	regression	NOUN
fcis-7512	190	3	tasks	task	NOUN
fcis-7512	190	4	,	,	PUNCT
fcis-7512	190	5	attention	attention	NOUN
fcis-7512	190	6	mechanism	mechanism	NOUN
fcis-7512	190	7	and	and	CCONJ
fcis-7512	190	8	weight	weight	NOUN
fcis-7512	190	9	loss	loss	NOUN
fcis-7512	190	10	mechanism	mechanism	NOUN
fcis-7512	190	11	were	be	AUX
fcis-7512	190	12	used	use	VERB
fcis-7512	190	13	together	together	ADV
fcis-7512	190	14	.	.	PUNCT
fcis-7512	191	1	mean	mean	VERB
fcis-7512	191	2	absolute	absolute	ADJ
fcis-7512	191	3	error	error	NOUN
fcis-7512	191	4	mae	mae	PROPN
fcis-7512	191	5	and	and	CCONJ
fcis-7512	191	6	correlation	correlation	NOUN
fcis-7512	191	7	are	be	AUX
fcis-7512	191	8	also	also	ADV
fcis-7512	191	9	better	well	ADJ
fcis-7512	191	10	than	than	ADP
fcis-7512	191	11	the	the	DET
fcis-7512	191	12	two	two	NUM
fcis-7512	191	13	alone	alone	ADV
fcis-7512	191	14	.	.	PUNCT
fcis-7512	192	1	as	as	ADP
fcis-7512	192	2	a	a	DET
fcis-7512	192	3	whole	whole	NOUN
fcis-7512	192	4	,	,	PUNCT
fcis-7512	192	5	the	the	DET
fcis-7512	192	6	attention	attention	NOUN
fcis-7512	192	7	mechanism	mechanism	NOUN
fcis-7512	192	8	and	and	CCONJ
fcis-7512	192	9	weight	weight	NOUN
fcis-7512	192	10	loss	loss	NOUN
fcis-7512	192	11	mechanism	mechanism	NOUN
fcis-7512	192	12	proposed	propose	VERB
fcis-7512	192	13	in	in	ADP
fcis-7512	192	14	this	this	DET
fcis-7512	192	15	paper	paper	NOUN
fcis-7512	192	16	show	show	VERB
fcis-7512	192	17	obvious	obvious	ADJ
fcis-7512	192	18	performance	performance	NOUN
fcis-7512	192	19	on	on	ADP
fcis-7512	192	20	mosei	mosei	ADJ
fcis-7512	192	21	with	with	ADP
fcis-7512	192	22	large	large	ADJ
fcis-7512	192	23	data	datum	NOUN
fcis-7512	192	24	scale	scale	NOUN
fcis-7512	192	25	.	.	PUNCT
fcis-7512	193	1	it	it	PRON
fcis-7512	193	2	can	can	AUX
fcis-7512	193	3	be	be	AUX
fcis-7512	193	4	seen	see	VERB
fcis-7512	193	5	that	that	SCONJ
fcis-7512	193	6	when	when	SCONJ
fcis-7512	193	7	the	the	DET
fcis-7512	193	8	attention	attention	NOUN
fcis-7512	193	9	mechanism	mechanism	NOUN
fcis-7512	193	10	is	be	AUX
fcis-7512	193	11	removed	remove	VERB
fcis-7512	193	12	,	,	PUNCT
fcis-7512	193	13	the	the	DET
fcis-7512	193	14	overall	overall	ADJ
fcis-7512	193	15	classification	classification	NOUN
fcis-7512	193	16	effect	effect	NOUN
fcis-7512	193	17	decreases	decrease	VERB
fcis-7512	193	18	significantly	significantly	ADV
fcis-7512	193	19	in	in	ADP
fcis-7512	193	20	the	the	DET
fcis-7512	193	21	accuracy	accuracy	NOUN
fcis-7512	193	22	of	of	ADP
fcis-7512	193	23	binary	binary	ADJ
fcis-7512	193	24	classification	classification	NOUN
fcis-7512	193	25	,	,	PUNCT
fcis-7512	193	26	reaching	reach	VERB
fcis-7512	193	27	1.2%~4.4	1.2%~4.4	NUM
fcis-7512	193	28	%	%	NOUN
fcis-7512	193	29	.	.	PUNCT
fcis-7512	194	1	it	it	PRON
fcis-7512	194	2	fully	fully	ADV
fcis-7512	194	3	demonstrates	demonstrate	VERB
fcis-7512	194	4	the	the	DET
fcis-7512	194	5	importance	importance	NOUN
fcis-7512	194	6	of	of	ADP
fcis-7512	194	7	attention	attention	NOUN
fcis-7512	194	8	mechanism	mechanism	NOUN
fcis-7512	194	9	in	in	ADP
fcis-7512	194	10	natural	natural	ADJ
fcis-7512	194	11	language	language	NOUN
fcis-7512	194	12	tasks	task	NOUN
fcis-7512	194	13	.	.	PUNCT
fcis-7512	195	1	it	it	PRON
fcis-7512	195	2	can	can	AUX
fcis-7512	195	3	effectively	effectively	ADV
fcis-7512	195	4	help	help	VERB
fcis-7512	195	5	the	the	DET
fcis-7512	195	6	model	model	NOUN
fcis-7512	195	7	understand	understand	VERB
fcis-7512	195	8	semantic	semantic	ADJ
fcis-7512	195	9	and	and	CCONJ
fcis-7512	195	10	contextual	contextual	ADJ
fcis-7512	195	11	information	information	NOUN
fcis-7512	195	12	.	.	PUNCT
fcis-7512	196	1	4	4	X
fcis-7512	196	2	.	.	X
fcis-7512	196	3	conclusion	conclusion	NOUN
fcis-7512	196	4	this	this	DET
fcis-7512	196	5	paper	paper	NOUN
fcis-7512	196	6	introduces	introduce	VERB
fcis-7512	196	7	a	a	DET
fcis-7512	196	8	multi	multi	NOUN
fcis-7512	196	9	-	-	NOUN
fcis-7512	196	10	task	task	ADJ
fcis-7512	196	11	and	and	CCONJ
fcis-7512	196	12	multi	multi	ADJ
fcis-7512	196	13	-	-	ADJ
fcis-7512	196	14	mode	mode	ADJ
fcis-7512	196	15	emotion	emotion	NOUN
fcis-7512	196	16	analysis	analysis	NOUN
fcis-7512	196	17	model	model	NOUN
fcis-7512	196	18	based	base	VERB
fcis-7512	196	19	on	on	ADP
fcis-7512	196	20	attention	attention	NOUN
fcis-7512	196	21	mechanism	mechanism	NOUN
fcis-7512	196	22	,	,	PUNCT
fcis-7512	196	23	which	which	PRON
fcis-7512	196	24	can	can	AUX
fcis-7512	196	25	be	be	AUX
fcis-7512	196	26	divided	divide	VERB
fcis-7512	196	27	into	into	ADP
fcis-7512	196	28	two	two	NUM
fcis-7512	196	29	types	type	NOUN
fcis-7512	196	30	:	:	PUNCT
fcis-7512	196	31	intermodal	intermodal	ADJ
fcis-7512	196	32	attention	attention	NOUN
fcis-7512	196	33	mechanism	mechanism	NOUN
fcis-7512	196	34	and	and	CCONJ
fcis-7512	196	35	single	single	ADJ
fcis-7512	196	36	-	-	PUNCT
fcis-7512	196	37	mode	mode	NOUN
fcis-7512	196	38	self	self	NOUN
fcis-7512	196	39	-	-	PUNCT
fcis-7512	196	40	attention	attention	NOUN
fcis-7512	196	41	mechanism	mechanism	NOUN
fcis-7512	196	42	.	.	PUNCT
fcis-7512	197	1	the	the	DET
fcis-7512	197	2	intermodal	intermodal	ADJ
fcis-7512	197	3	attention	attention	NOUN
fcis-7512	197	4	mechanism	mechanism	NOUN
fcis-7512	197	5	can	can	AUX
fcis-7512	197	6	fully	fully	ADV
fcis-7512	197	7	consider	consider	VERB
fcis-7512	197	8	the	the	DET
fcis-7512	197	9	intermodal	intermodal	ADJ
fcis-7512	197	10	relationship	relationship	NOUN
fcis-7512	197	11	and	and	CCONJ
fcis-7512	197	12	complement	complement	VERB
fcis-7512	197	13	each	each	DET
fcis-7512	197	14	other	other	ADJ
fcis-7512	197	15	.	.	PUNCT
fcis-7512	198	1	compared	compare	VERB
fcis-7512	198	2	with	with	ADP
fcis-7512	198	3	the	the	DET
fcis-7512	198	4	previous	previous	ADJ
fcis-7512	198	5	work	work	NOUN
fcis-7512	198	6	,	,	PUNCT
fcis-7512	198	7	more	more	ADJ
fcis-7512	198	8	feature	feature	NOUN
fcis-7512	198	9	information	information	NOUN
fcis-7512	198	10	can	can	AUX
fcis-7512	198	11	be	be	AUX
fcis-7512	198	12	learned	learn	VERB
fcis-7512	198	13	.	.	PUNCT
fcis-7512	199	1	for	for	ADP
fcis-7512	199	2	the	the	DET
fcis-7512	199	3	single	single	ADJ
fcis-7512	199	4	-	-	PUNCT
fcis-7512	199	5	mode	mode	NOUN
fcis-7512	199	6	self	self	NOUN
fcis-7512	199	7	-	-	PUNCT
fcis-7512	199	8	attention	attention	NOUN
fcis-7512	199	9	mechanism	mechanism	NOUN
fcis-7512	199	10	,	,	PUNCT
fcis-7512	199	11	the	the	DET
fcis-7512	199	12	context	context	NOUN
fcis-7512	199	13	information	information	NOUN
fcis-7512	199	14	of	of	ADP
fcis-7512	199	15	single	single	ADJ
fcis-7512	199	16	-	-	PUNCT
fcis-7512	199	17	mode	mode	NOUN
fcis-7512	199	18	can	can	AUX
fcis-7512	199	19	be	be	AUX
fcis-7512	199	20	fully	fully	ADV
fcis-7512	199	21	learned	learn	VERB
fcis-7512	199	22	,	,	PUNCT
fcis-7512	199	23	and	and	CCONJ
fcis-7512	199	24	the	the	DET
fcis-7512	199	25	information	information	NOUN
fcis-7512	199	26	gain	gain	NOUN
fcis-7512	199	27	between	between	ADP
fcis-7512	199	28	modes	mode	NOUN
fcis-7512	199	29	can	can	AUX
fcis-7512	199	30	be	be	AUX
fcis-7512	199	31	fully	fully	ADV
fcis-7512	199	32	utilized	utilize	VERB
fcis-7512	199	33	to	to	PART
fcis-7512	199	34	reduce	reduce	VERB
fcis-7512	199	35	noise	noise	NOUN
fcis-7512	199	36	through	through	ADP
fcis-7512	199	37	the	the	DET
fcis-7512	199	38	combination	combination	NOUN
fcis-7512	199	39	of	of	ADP
fcis-7512	199	40	multi	multi	ADJ
fcis-7512	199	41	-	-	NOUN
fcis-7512	199	42	task	task	ADJ
fcis-7512	199	43	learning	learning	NOUN
fcis-7512	199	44	.	.	PUNCT
fcis-7512	200	1	however	however	ADV
fcis-7512	200	2	,	,	PUNCT
fcis-7512	200	3	the	the	DET
fcis-7512	200	4	model	model	NOUN
fcis-7512	200	5	in	in	ADP
fcis-7512	200	6	this	this	DET
fcis-7512	200	7	paper	paper	NOUN
fcis-7512	200	8	still	still	ADV
fcis-7512	200	9	needs	need	VERB
fcis-7512	200	10	to	to	PART
fcis-7512	200	11	be	be	AUX
fcis-7512	200	12	improved	improve	VERB
fcis-7512	200	13	.	.	PUNCT
fcis-7512	201	1	the	the	DET
fcis-7512	201	2	model	model	NOUN
fcis-7512	201	3	proposed	propose	VERB
fcis-7512	201	4	in	in	ADP
fcis-7512	201	5	this	this	DET
fcis-7512	201	6	paper	paper	NOUN
fcis-7512	201	7	can	can	AUX
fcis-7512	201	8	only	only	ADV
fcis-7512	201	9	deal	deal	VERB
fcis-7512	201	10	with	with	ADP
fcis-7512	201	11	multimodal	multimodal	ADJ
fcis-7512	201	12	emotion	emotion	NOUN
fcis-7512	201	13	task	task	NOUN
fcis-7512	201	14	segmentation	segmentation	NOUN
fcis-7512	201	15	,	,	PUNCT
fcis-7512	201	16	and	and	CCONJ
fcis-7512	201	17	has	have	VERB
fcis-7512	201	18	no	no	DET
fcis-7512	201	19	adaptive	adaptive	ADJ
fcis-7512	201	20	ability	ability	NOUN
fcis-7512	201	21	to	to	PART
fcis-7512	201	22	single	single	ADJ
fcis-7512	201	23	mode	mode	NOUN
fcis-7512	201	24	and	and	CCONJ
fcis-7512	201	25	double	double	ADJ
fcis-7512	201	26	mode	mode	NOUN
fcis-7512	201	27	emotion	emotion	NOUN
fcis-7512	201	28	analysis	analysis	NOUN
fcis-7512	201	29	task	task	NOUN
fcis-7512	201	30	,	,	PUNCT
fcis-7512	201	31	and	and	CCONJ
fcis-7512	201	32	the	the	DET
fcis-7512	201	33	multi	multi	ADJ
fcis-7512	201	34	-	-	ADJ
fcis-7512	201	35	task	task	ADJ
fcis-7512	201	36	weight	weight	NOUN
fcis-7512	201	37	loss	loss	NOUN
fcis-7512	201	38	is	be	AUX
fcis-7512	201	39	a	a	DET
fcis-7512	201	40	part	part	NOUN
fcis-7512	201	41	without	without	ADP
fcis-7512	201	42	parameter	parameter	NOUN
fcis-7512	201	43	learning	learning	NOUN
fcis-7512	201	44	.	.	PUNCT
fcis-7512	202	1	it	it	PRON
fcis-7512	202	2	needs	need	VERB
fcis-7512	202	3	to	to	PART
fcis-7512	202	4	be	be	AUX
fcis-7512	202	5	improved	improve	VERB
fcis-7512	202	6	in	in	ADP
fcis-7512	202	7	the	the	DET
fcis-7512	202	8	future	future	NOUN
fcis-7512	202	9	.	.	PUNCT
fcis-7512	203	1	acknowledgment	acknowledgment	NOUN
fcis-7512	203	2	this	this	DET
fcis-7512	203	3	paper	paper	NOUN
fcis-7512	203	4	was	be	AUX
fcis-7512	203	5	supported	support	VERB
fcis-7512	203	6	by	by	ADP
fcis-7512	203	7	"	"	PUNCT
fcis-7512	203	8	seedling	seedle	VERB
fcis-7512	203	9	cultivation	cultivation	NOUN
fcis-7512	203	10	"	"	PUNCT
fcis-7512	203	11	project	project	NOUN
fcis-7512	203	12	for	for	ADP
fcis-7512	203	13	young	young	ADJ
fcis-7512	203	14	scientific	scientific	ADJ
fcis-7512	203	15	and	and	CCONJ
fcis-7512	203	16	technological	technological	ADJ
fcis-7512	203	17	talents	talent	NOUN
fcis-7512	203	18	of	of	ADP
fcis-7512	203	19	education	education	PROPN
fcis-7512	203	20	department	department	PROPN
fcis-7512	203	21	of	of	ADP
fcis-7512	203	22	liaoning	liaoning	PROPN
fcis-7512	203	23	province	province	PROPN
fcis-7512	203	24	,	,	PUNCT
fcis-7512	203	25	no	no	INTJ
fcis-7512	203	26	.	.	PUNCT
fcis-7512	204	1	j2020113	j2020113	PROPN
fcis-7512	204	2	.	.	PUNCT
fcis-7512	205	1	references	reference	NOUN
fcis-7512	205	2	[	[	X
fcis-7512	205	3	1	1	X
fcis-7512	205	4	]	]	PUNCT
fcis-7512	205	5	yang	yang	PROPN
fcis-7512	205	6	li	li	PROPN
fcis-7512	205	7	-	-	PROPN
fcis-7512	205	8	gong	gong	PROPN
fcis-7512	205	9	,	,	PUNCT
fcis-7512	205	10	zhu	zhu	PROPN
fcis-7512	205	11	jian	jian	PROPN
fcis-7512	205	12	,	,	PUNCT
fcis-7512	205	13	tang	tang	PROPN
fcis-7512	205	14	shi	shi	PROPN
fcis-7512	205	15	-	-	PUNCT
fcis-7512	205	16	ping	ping	PROPN
fcis-7512	205	17	.	.	PUNCT
fcis-7512	206	1	a	a	DET
fcis-7512	206	2	review	review	NOUN
fcis-7512	206	3	of	of	ADP
fcis-7512	206	4	text	text	NOUN
fcis-7512	206	5	sentiment	sentiment	NOUN
fcis-7512	206	6	analysis	analysis	NOUN
fcis-7512	206	7	[	[	X
fcis-7512	206	8	j	j	X
fcis-7512	206	9	]	]	X
fcis-7512	206	10	.	.	PUNCT
fcis-7512	207	1	journal	journal	PROPN
fcis-7512	207	2	of	of	ADP
fcis-7512	207	3	computer	computer	NOUN
fcis-7512	207	4	applications	application	NOUN
fcis-7512	207	5	,	,	PUNCT
fcis-7512	207	6	2013	2013	NUM
fcis-7512	207	7	,	,	PUNCT
fcis-7512	207	8	33(6):1574	33(6):1574	NUM
fcis-7512	207	9	-	-	SYM
fcis-7512	207	10	1607	1607	NUM
fcis-7512	207	11	.	.	PUNCT
fcis-7512	208	1	[	[	X
fcis-7512	208	2	2	2	NUM
fcis-7512	208	3	]	]	X
fcis-7512	208	4	liu	liu	PROPN
fcis-7512	208	5	jiming	jiming	PROPN
fcis-7512	208	6	,	,	PUNCT
fcis-7512	208	7	zhang	zhang	PROPN
fcis-7512	208	8	peixiang	peixiang	PROPN
fcis-7512	208	9	,	,	PUNCT
fcis-7512	208	10	liu	liu	PROPN
fcis-7512	208	11	ying	ying	PROPN
fcis-7512	208	12	,	,	PUNCT
fcis-7512	208	13	et	et	PROPN
fcis-7512	208	14	al	al	PROPN
fcis-7512	208	15	.	.	PUNCT
fcis-7512	208	16	multimodal	multimodal	ADJ
fcis-7512	208	17	sentiment	sentiment	NOUN
fcis-7512	208	18	analysis	analysis	NOUN
fcis-7512	208	19	[	[	X
fcis-7512	208	20	j	j	X
fcis-7512	208	21	]	]	X
fcis-7512	208	22	.	.	PUNCT
fcis-7512	209	1	journal	journal	PROPN
fcis-7512	209	2	of	of	ADP
fcis-7512	209	3	computer	computer	NOUN
fcis-7512	209	4	science	science	NOUN
fcis-7512	209	5	and	and	CCONJ
fcis-7512	209	6	exploration	exploration	NOUN
fcis-7512	209	7	,	,	PUNCT
fcis-7512	209	8	2021	2021	NUM
fcis-7512	209	9	,	,	PUNCT
fcis-7512	209	10	15(7	15(7	NUM
fcis-7512	209	11	):	):	PUNCT
fcis-7512	209	12	1165	1165	NUM
fcis-7512	209	13	.	.	PUNCT
fcis-7512	210	1	[	[	X
fcis-7512	210	2	3	3	X
fcis-7512	210	3	]	]	X
fcis-7512	210	4	poria	poria	PROPN
fcis-7512	210	5	s	s	PROPN
fcis-7512	210	6	,	,	PUNCT
fcis-7512	210	7	chaturvedi	chaturvedi	PROPN
fcis-7512	210	8	i	i	PROPN
fcis-7512	210	9	,	,	PUNCT
fcis-7512	210	10	cambria	cambria	PROPN
fcis-7512	210	11	e	e	PROPN
fcis-7512	210	12	,	,	PUNCT
fcis-7512	210	13	et	et	PROPN
fcis-7512	210	14	al	al	PROPN
fcis-7512	210	15	.	.	PUNCT
fcis-7512	210	16	convolutional	convolutional	ADJ
fcis-7512	210	17	mkl	mkl	PROPN
fcis-7512	210	18	based	base	VERB
fcis-7512	210	19	multimodal	multimodal	NOUN
fcis-7512	210	20	emotion	emotion	NOUN
fcis-7512	210	21	recognition	recognition	NOUN
fcis-7512	210	22	and	and	CCONJ
fcis-7512	210	23	sentiment	sentiment	NOUN
fcis-7512	210	24	analysis[c]//2016	analysis[c]//2016	DET
fcis-7512	210	25	ieee	ieee	NOUN
fcis-7512	210	26	16th	16th	ADJ
fcis-7512	210	27	international	international	ADJ
fcis-7512	210	28	conference	conference	NOUN
fcis-7512	210	29	on	on	ADP
fcis-7512	210	30	data	datum	NOUN
fcis-7512	210	31	mining	mining	NOUN
fcis-7512	210	32	(	(	PUNCT
fcis-7512	210	33	icdm	icdm	NOUN
fcis-7512	210	34	)	)	PUNCT
fcis-7512	210	35	.	.	PUNCT
fcis-7512	211	1	ieee	ieee	PROPN
fcis-7512	211	2	,	,	PUNCT
fcis-7512	211	3	2016	2016	NUM
fcis-7512	211	4	:	:	PUNCT
fcis-7512	211	5	439	439	NUM
fcis-7512	211	6	-	-	SYM
fcis-7512	211	7	448	448	NUM
fcis-7512	211	8	.	.	PUNCT
fcis-7512	212	1	[	[	X
fcis-7512	212	2	4	4	NUM
fcis-7512	212	3	]	]	X
fcis-7512	212	4	erik	erik	PROPN
fcis-7512	212	5	cambria	cambria	PROPN
fcis-7512	212	6	,	,	PUNCT
fcis-7512	212	7	devamanyu	devamanyu	VERB
fcis-7512	212	8	hazarika	hazarika	NOUN
fcis-7512	212	9	,	,	PUNCT
fcis-7512	212	10	soujanya	soujanya	NOUN
fcis-7512	212	11	poria	poria	NOUN
fcis-7512	212	12	,	,	PUNCT
fcis-7512	212	13	amir	amir	PROPN
fcis-7512	212	14	hussain	hussain	PROPN
fcis-7512	212	15	,	,	PUNCT
fcis-7512	212	16	and	and	CCONJ
fcis-7512	212	17	rbv	rbv	VERB
fcis-7512	212	18	subramanyam	subramanyam	NOUN
fcis-7512	212	19	.	.	PUNCT
fcis-7512	213	1	2017.benchmarking	2017.benchmarke	VERB
fcis-7512	213	2	multimodal	multimodal	ADJ
fcis-7512	213	3	sentiment	sentiment	NOUN
fcis-7512	213	4	analysis	analysis	NOUN
fcis-7512	213	5	.	.	PUNCT
fcis-7512	214	1	in	in	ADP
fcis-7512	214	2	international	international	ADJ
fcis-7512	214	3	conference	conference	NOUN
fcis-7512	214	4	on	on	ADP
fcis-7512	214	5	computational	computational	ADJ
fcis-7512	214	6	linguistics	linguistic	NOUN
fcis-7512	214	7	and	and	CCONJ
fcis-7512	214	8	intelligent	intelligent	ADJ
fcis-7512	214	9	text	text	NOUN
fcis-7512	214	10	processing	processing	NOUN
fcis-7512	214	11	,	,	PUNCT
fcis-7512	214	12	pages	page	NOUN
fcis-7512	214	13	166–179.springer	166–179.springer	NUM
fcis-7512	214	14	.	.	PUNCT
fcis-7512	215	1	[	[	X
fcis-7512	215	2	5	5	NUM
fcis-7512	215	3	]	]	PUNCT
fcis-7512	215	4	zadeh	zadeh	PROPN
fcis-7512	215	5	a	a	X
fcis-7512	215	6	,	,	PUNCT
fcis-7512	215	7	zellers	zeller	NOUN
fcis-7512	215	8	r	r	NOUN
fcis-7512	215	9	,	,	PUNCT
fcis-7512	215	10	pincus	pincus	PROPN
fcis-7512	215	11	e	e	PROPN
fcis-7512	215	12	,	,	PUNCT
fcis-7512	215	13	et	et	PROPN
fcis-7512	215	14	al	al	PROPN
fcis-7512	215	15	.	.	PUNCT
fcis-7512	215	16	multimodal	multimodal	ADJ
fcis-7512	215	17	sentiment	sentiment	NOUN
fcis-7512	215	18	intensity	intensity	NOUN
fcis-7512	215	19	analysis	analysis	NOUN
fcis-7512	215	20	in	in	ADP
fcis-7512	215	21	videos	video	NOUN
fcis-7512	215	22	:	:	PUNCT
fcis-7512	215	23	facial	facial	ADJ
fcis-7512	215	24	gestures	gesture	NOUN
fcis-7512	215	25	and	and	CCONJ
fcis-7512	215	26	verbal	verbal	ADJ
fcis-7512	215	27	messages[j	messages[j	PROPN
fcis-7512	215	28	]	]	PUNCT
fcis-7512	215	29	.	.	PUNCT
fcis-7512	216	1	ieee	ieee	PROPN
fcis-7512	216	2	intelligent	intelligent	ADJ
fcis-7512	216	3	systems	system	NOUN
fcis-7512	216	4	,	,	PUNCT
fcis-7512	216	5	2016	2016	NUM
fcis-7512	216	6	,	,	PUNCT
fcis-7512	216	7	31(6):82	31(6):82	NUM
fcis-7512	216	8	-	-	SYM
fcis-7512	216	9	88	88	NUM
fcis-7512	216	10	.	.	PUNCT
fcis-7512	217	1	[	[	X
fcis-7512	217	2	6	6	NUM
fcis-7512	217	3	]	]	PUNCT
fcis-7512	217	4	amir	amir	PROPN
fcis-7512	217	5	zadeh	zadeh	PROPN
fcis-7512	217	6	,	,	PUNCT
fcis-7512	217	7	minghai	minghai	PROPN
fcis-7512	217	8	chen	chen	PROPN
fcis-7512	217	9	,	,	PUNCT
fcis-7512	217	10	soujanya	soujanya	PROPN
fcis-7512	217	11	poria	poria	PROPN
fcis-7512	217	12	,	,	PUNCT
fcis-7512	217	13	erik	erik	PROPN
fcis-7512	217	14	cambria	cambria	PROPN
fcis-7512	217	15	,	,	PUNCT
fcis-7512	217	16	and	and	CCONJ
fcis-7512	217	17	louis	louis	NOUN
fcis-7512	217	18	-	-	PUNCT
fcis-7512	217	19	philippe	philippe	PROPN
fcis-7512	217	20	morency	morency	NOUN
fcis-7512	217	21	.	.	PUNCT
fcis-7512	218	1	2017	2017	NUM
fcis-7512	218	2	.	.	PUNCT
fcis-7512	218	3	tensor	tensor	NOUN
fcis-7512	218	4	fusion	fusion	NOUN
fcis-7512	218	5	network	network	NOUN
fcis-7512	218	6	for	for	ADP
fcis-7512	218	7	multimodal	multimodal	ADJ
fcis-7512	218	8	sentiment	sentiment	NOUN
fcis-7512	218	9	analysis.arxiv	analysis.arxiv	PROPN
fcis-7512	218	10	preprint	preprint	NOUN
fcis-7512	218	11	arxiv	arxiv	PROPN
fcis-7512	218	12	:	:	PUNCT
fcis-7512	218	13	1707	1707	NUM
fcis-7512	218	14	.	.	PUNCT
fcis-7512	218	15	07250	07250	NUM
fcis-7512	218	16	.	.	PUNCT
fcis-7512	219	1	73	73	NUM
fcis-7512	220	1	[	[	X
fcis-7512	220	2	7	7	NUM
fcis-7512	220	3	]	]	X
fcis-7512	220	4	poria	poria	PROPN
fcis-7512	220	5	s	s	PROPN
fcis-7512	220	6	,	,	PUNCT
fcis-7512	220	7	cambria	cambria	PROPN
fcis-7512	220	8	e	e	PROPN
fcis-7512	220	9	,	,	PUNCT
fcis-7512	220	10	hazarika	hazarika	NOUN
fcis-7512	220	11	d	d	NOUN
fcis-7512	220	12	,	,	PUNCT
fcis-7512	220	13	et	et	PROPN
fcis-7512	220	14	al	al	PROPN
fcis-7512	220	15	.	.	PUNCT
fcis-7512	221	1	context	context	NOUN
fcis-7512	221	2	-	-	PUNCT
fcis-7512	221	3	dependent	dependent	ADJ
fcis-7512	221	4	sentiment	sentiment	NOUN
fcis-7512	221	5	analysis	analysis	NOUN
fcis-7512	221	6	in	in	ADP
fcis-7512	221	7	user	user	NOUN
fcis-7512	221	8	-	-	PUNCT
fcis-7512	221	9	generated	generate	VERB
fcis-7512	221	10	videos[c]//proceedings	videos[c]//proceeding	NOUN
fcis-7512	221	11	of	of	ADP
fcis-7512	221	12	the	the	DET
fcis-7512	221	13	55th	55th	ADJ
fcis-7512	221	14	annual	annual	ADJ
fcis-7512	221	15	meeting	meeting	NOUN
fcis-7512	221	16	of	of	ADP
fcis-7512	221	17	the	the	DET
fcis-7512	221	18	association	association	NOUN
fcis-7512	221	19	for	for	ADP
fcis-7512	221	20	computational	computational	ADJ
fcis-7512	221	21	linguistics	linguistic	NOUN
fcis-7512	221	22	(	(	PUNCT
fcis-7512	221	23	volume	volume	NOUN
fcis-7512	221	24	1	1	NUM
fcis-7512	221	25	:	:	PUNCT
fcis-7512	221	26	long	long	ADJ
fcis-7512	221	27	papers	paper	NOUN
fcis-7512	221	28	)	)	PUNCT
fcis-7512	221	29	.	.	PUNCT
fcis-7512	222	1	2017	2017	NUM
fcis-7512	222	2	:	:	PUNCT
fcis-7512	222	3	873	873	NUM
fcis-7512	222	4	-	-	NUM
fcis-7512	222	5	883	883	NUM
fcis-7512	222	6	.	.	PUNCT
fcis-7512	223	1	[	[	X
fcis-7512	223	2	8	8	NUM
fcis-7512	223	3	]	]	X
fcis-7512	223	4	zadeh	zadeh	PROPN
fcis-7512	223	5	a	a	PROPN
fcis-7512	223	6	,	,	PUNCT
fcis-7512	223	7	liang	liang	PROPN
fcis-7512	223	8	p	p	PROPN
fcis-7512	223	9	p	p	PROPN
fcis-7512	223	10	,	,	PUNCT
fcis-7512	223	11	poria	poria	PROPN
fcis-7512	223	12	s	s	PROPN
fcis-7512	223	13	,	,	PUNCT
fcis-7512	223	14	et	et	PROPN
fcis-7512	223	15	al	al	PROPN
fcis-7512	223	16	.	.	PUNCT
fcis-7512	224	1	multi	multi	ADJ
fcis-7512	224	2	-	-	ADJ
fcis-7512	224	3	attention	attention	ADJ
fcis-7512	224	4	recurrent	recurrent	ADJ
fcis-7512	224	5	network	network	NOUN
fcis-7512	224	6	for	for	ADP
fcis-7512	224	7	human	human	ADJ
fcis-7512	224	8	communication	communication	NOUN
fcis-7512	224	9	comprehension[c	comprehension[c	PROPN
fcis-7512	224	10	]	]	PUNCT
fcis-7512	225	1	//	//	NUM
fcis-7512	225	2	proceedings	proceeding	NOUN
fcis-7512	225	3	of	of	ADP
fcis-7512	225	4	the	the	DET
fcis-7512	225	5	aaai	aaai	PROPN
fcis-7512	225	6	conference	conference	NOUN
fcis-7512	225	7	on	on	ADP
fcis-7512	225	8	artificial	artificial	ADJ
fcis-7512	225	9	intelligence	intelligence	NOUN
fcis-7512	225	10	.	.	PUNCT
fcis-7512	226	1	2018	2018	NUM
fcis-7512	226	2	,	,	PUNCT
fcis-7512	226	3	32(1	32(1	NUM
fcis-7512	226	4	)	)	PUNCT
fcis-7512	226	5	.	.	PUNCT
fcis-7512	227	1	[	[	X
fcis-7512	227	2	9	9	NUM
fcis-7512	227	3	]	]	SYM
fcis-7512	227	4	vaswani	vaswani	NOUN
fcis-7512	227	5	a	a	DET
fcis-7512	227	6	,	,	PUNCT
fcis-7512	227	7	shazeer	shazeer	NOUN
fcis-7512	227	8	n	n	NOUN
fcis-7512	227	9	,	,	PUNCT
fcis-7512	227	10	parmar	parmar	PROPN
fcis-7512	227	11	n	n	PROPN
fcis-7512	227	12	,	,	PUNCT
fcis-7512	227	13	et	et	PROPN
fcis-7512	227	14	al	al	PROPN
fcis-7512	227	15	.	.	PUNCT
fcis-7512	227	16	attention	attention	NOUN
fcis-7512	227	17	is	be	AUX
fcis-7512	227	18	all	all	PRON
fcis-7512	227	19	you	you	PRON
fcis-7512	227	20	need[j	need[j	VERB
fcis-7512	227	21	]	]	PUNCT
fcis-7512	227	22	.	.	PUNCT
fcis-7512	228	1	arxiv	arxiv	PROPN
fcis-7512	228	2	,	,	PUNCT
fcis-7512	228	3	2017	2017	NUM
fcis-7512	228	4	.	.	PUNCT
fcis-7512	229	1	[	[	X
fcis-7512	229	2	10	10	NUM
fcis-7512	229	3	]	]	X
fcis-7512	229	4	tsai	tsai	PROPN
fcis-7512	229	5	,	,	PUNCT
fcis-7512	229	6	y	y	PROPN
fcis-7512	229	7	.-h	.-h	PROPN
fcis-7512	229	8	.	.	PUNCT
fcis-7512	230	1	h.	h.	PROPN
fcis-7512	230	2	;	;	PUNCT
fcis-7512	230	3	bai	bai	PROPN
fcis-7512	230	4	,	,	PUNCT
fcis-7512	230	5	s.	s.	PROPN
fcis-7512	230	6	;	;	PUNCT
fcis-7512	230	7	liang	liang	PROPN
fcis-7512	230	8	,	,	PUNCT
fcis-7512	230	9	p	p	X
fcis-7512	230	10	.	.	PUNCT
fcis-7512	231	1	p	p	X
fcis-7512	231	2	.	.	PUNCT
fcis-7512	232	1	;	;	PUNCT
fcis-7512	232	2	kolter	kolter	PROPN
fcis-7512	232	3	,	,	PUNCT
fcis-7512	232	4	j.	j.	PROPN
fcis-7512	232	5	z.	z.	PROPN
fcis-7512	232	6	;	;	PUNCT
fcis-7512	232	7	morency	morency	NOUN
fcis-7512	232	8	,	,	PUNCT
fcis-7512	232	9	l.p	l.p	PROPN
fcis-7512	232	10	.	.	PUNCT
fcis-7512	232	11	;	;	PUNCT
fcis-7512	232	12	and	and	CCONJ
fcis-7512	232	13	salakhutdinov	salakhutdinov	PROPN
fcis-7512	232	14	,	,	PUNCT
fcis-7512	232	15	r.	r.	PROPN
fcis-7512	232	16	2019	2019	NUM
fcis-7512	232	17	.	.	PUNCT
fcis-7512	233	1	multimodal	multimodal	NOUN
fcis-7512	233	2	transformer	transformer	NOUN
fcis-7512	233	3	for	for	ADP
fcis-7512	233	4	unaligned	unaligned	ADJ
fcis-7512	233	5	multimodal	multimodal	NOUN
fcis-7512	233	6	language	language	NOUN
fcis-7512	233	7	sequences	sequence	NOUN
fcis-7512	233	8	.	.	PUNCT
fcis-7512	234	1	in	in	ADP
fcis-7512	234	2	proceedings	proceeding	NOUN
fcis-7512	234	3	of	of	ADP
fcis-7512	234	4	the	the	DET
fcis-7512	234	5	conference	conference	NOUN
fcis-7512	234	6	.	.	PUNCT
fcis-7512	235	1	association	association	NOUN
fcis-7512	235	2	for	for	ADP
fcis-7512	235	3	computational	computational	ADJ
fcis-7512	235	4	linguistics	linguistic	NOUN
fcis-7512	235	5	.	.	PUNCT
fcis-7512	236	1	meeting	meeting	NOUN
fcis-7512	236	2	,	,	PUNCT
fcis-7512	236	3	volume	volume	NOUN
fcis-7512	236	4	2019	2019	NUM
fcis-7512	236	5	,	,	PUNCT
fcis-7512	236	6	6558	6558	NUM
fcis-7512	236	7	.	.	PUNCT
fcis-7512	237	1	nih	nih	PROPN
fcis-7512	237	2	public	public	ADJ
fcis-7512	237	3	access	access	NOUN
fcis-7512	237	4	.	.	PUNCT
fcis-7512	238	1	[	[	X
fcis-7512	238	2	11	11	NUM
fcis-7512	238	3	]	]	X
fcis-7512	238	4	yu	yu	PROPN
fcis-7512	238	5	z	z	PROPN
fcis-7512	238	6	,	,	PUNCT
fcis-7512	238	7	qiang	qiang	PROPN
fcis-7512	238	8	y	y	PROPN
fcis-7512	238	9	.	.	PUNCT
fcis-7512	239	1	a	a	DET
fcis-7512	239	2	survey	survey	NOUN
fcis-7512	239	3	on	on	ADP
fcis-7512	239	4	multi	multi	ADJ
fcis-7512	239	5	-	-	ADJ
fcis-7512	239	6	task	task	ADJ
fcis-7512	239	7	learning[j	learning[j	NOUN
fcis-7512	239	8	]	]	PUNCT
fcis-7512	239	9	.	.	PUNCT
fcis-7512	240	1	2017	2017	NUM
fcis-7512	240	2	.	.	PUNCT
fcis-7512	241	1	[	[	X
fcis-7512	241	2	12	12	NUM
fcis-7512	241	3	]	]	X
fcis-7512	241	4	davoodi	davoodi	ADJ
fcis-7512	241	5	e	e	PROPN
fcis-7512	241	6	,	,	PUNCT
fcis-7512	241	7	kosseim	kosseim	PROPN
fcis-7512	241	8	l	l	PROPN
fcis-7512	241	9	,	,	PUNCT
fcis-7512	241	10	mongrain	mongrain	PROPN
fcis-7512	241	11	m	m	INTJ
fcis-7512	241	12	.	.	PUNCT
fcis-7512	242	1	on	on	ADP
fcis-7512	242	2	the	the	DET
fcis-7512	242	3	influence	influence	NOUN
fcis-7512	242	4	of	of	ADP
fcis-7512	242	5	contextual	contextual	ADJ
fcis-7512	242	6	features	feature	NOUN
fcis-7512	242	7	for	for	ADP
fcis-7512	242	8	the	the	DET
fcis-7512	242	9	identification	identification	NOUN
fcis-7512	242	10	of	of	ADP
fcis-7512	242	11	complex	complex	ADJ
fcis-7512	242	12	words[j	words[j	PROPN
fcis-7512	242	13	]	]	PUNCT
fcis-7512	242	14	.	.	PUNCT
fcis-7512	243	1	international	international	ADJ
fcis-7512	243	2	journal	journal	PROPN
fcis-7512	243	3	of	of	ADP
fcis-7512	243	4	semantic	semantic	ADJ
fcis-7512	243	5	computing	computing	NOUN
fcis-7512	243	6	,	,	PUNCT
fcis-7512	243	7	2017	2017	NUM
fcis-7512	243	8	,	,	PUNCT
fcis-7512	243	9	11(04):497	11(04):497	NUM
fcis-7512	243	10	-	-	SYM
fcis-7512	243	11	511	511	NUM
fcis-7512	243	12	.	.	PUNCT
fcis-7512	244	1	v	v	X
fcis-7512	244	2	,	,	PUNCT
fcis-7512	244	3	et	et	PROPN
fcis-7512	244	4	al	al	PROPN
fcis-7512	244	5	.	.	PROPN
fcis-7512	244	6	channel	channel	NOUN
fcis-7512	244	7	models	model	NOUN
fcis-7512	244	8	for	for	ADP
fcis-7512	244	9	fixed	fix	VERB
fcis-7512	244	10	wireless	wireless	ADJ
fcis-7512	244	11	applications	application	NOUN
fcis-7512	244	12	.	.	PUNCT
fcis-7512	245	1	ieee	ieee	NOUN
fcis-7512	245	2	802.16a	802.16a	PROPN
fcis-7512	245	3	cont	cont	PROPN
fcis-7512	245	4	.	.	PUNCT
fcis-7512	246	1	ieee	ieee	PROPN
fcis-7512	246	2	802.16.3c-01/29r4	802.16.3c-01/29r4	NUM
fcis-7512	246	3	,	,	PUNCT
fcis-7512	246	4	2003	2003	NUM
fcis-7512	247	1	[	[	X
fcis-7512	247	2	13	13	NUM
fcis-7512	247	3	]	]	X
fcis-7512	247	4	devlin	devlin	PROPN
fcis-7512	247	5	j	j	PROPN
fcis-7512	247	6	,	,	PUNCT
fcis-7512	247	7	chang	chang	PROPN
fcis-7512	247	8	m	m	PROPN
fcis-7512	247	9	w	w	PROPN
fcis-7512	247	10	,	,	PUNCT
fcis-7512	247	11	lee	lee	PROPN
fcis-7512	247	12	k	k	PROPN
fcis-7512	247	13	,	,	PUNCT
fcis-7512	247	14	et	et	PROPN
fcis-7512	247	15	al	al	PROPN
fcis-7512	247	16	.	.	PUNCT
fcis-7512	247	17	bert	bert	PROPN
fcis-7512	247	18	:	:	PUNCT
fcis-7512	247	19	pre	pre	ADJ
fcis-7512	247	20	-	-	NOUN
fcis-7512	247	21	training	training	NOUN
fcis-7512	247	22	of	of	ADP
fcis-7512	247	23	deep	deep	ADJ
fcis-7512	247	24	bidirectional	bidirectional	ADJ
fcis-7512	247	25	transformers	transformer	NOUN
fcis-7512	247	26	for	for	ADP
fcis-7512	247	27	language	language	NOUN
fcis-7512	247	28	understanding	understanding	NOUN
fcis-7512	247	29	[	[	X
fcis-7512	247	30	j	j	X
fcis-7512	247	31	]	]	X
fcis-7512	247	32	.	.	PUNCT
fcis-7512	247	33	2018	2018	NUM
fcis-7512	247	34	.	.	PUNCT
fcis-7512	248	1	[	[	X
fcis-7512	248	2	14	14	NUM
fcis-7512	248	3	]	]	PUNCT
fcis-7512	248	4	devamanyu	devamanyu	NOUN
fcis-7512	248	5	hazarika	hazarika	NOUN
fcis-7512	248	6	,	,	PUNCT
fcis-7512	248	7	roger	roger	PROPN
fcis-7512	248	8	zimmermann	zimmermann	PROPN
fcis-7512	248	9	,	,	PUNCT
fcis-7512	248	10	and	and	CCONJ
fcis-7512	248	11	soujanya	soujanya	PROPN
fcis-7512	248	12	poria	poria	NOUN
fcis-7512	248	13	.	.	PUNCT
fcis-7512	249	1	2020	2020	NUM
fcis-7512	249	2	.	.	PUNCT
fcis-7512	250	1	misa	misa	NOUN
fcis-7512	250	2	:	:	PUNCT
fcis-7512	250	3	modality	modality	NOUN
fcis-7512	250	4	-	-	PUNCT
fcis-7512	250	5	invariant	invariant	ADJ
fcis-7512	250	6	and	and	CCONJ
fcis-7512	250	7	-	-	PUNCT
fcis-7512	250	8	specific	specific	ADJ
fcis-7512	250	9	representations	representation	NOUN
fcis-7512	250	10	for	for	ADP
fcis-7512	250	11	multimodal	multimodal	ADJ
fcis-7512	250	12	sentiment	sentiment	NOUN
fcis-7512	250	13	analysis	analysis	NOUN
fcis-7512	250	14	.	.	PUNCT
fcis-7512	251	1	in	in	ADP
fcis-7512	251	2	proceedings	proceeding	NOUN
fcis-7512	251	3	of	of	ADP
fcis-7512	251	4	the	the	DET
fcis-7512	251	5	28th	28th	ADJ
fcis-7512	251	6	acm	acm	PROPN
fcis-7512	251	7	international	international	ADJ
fcis-7512	251	8	conference	conference	NOUN
fcis-7512	251	9	on	on	ADP
fcis-7512	251	10	multimedia	multimedia	NOUN
fcis-7512	251	11	,	,	PUNCT
fcis-7512	251	12	pages	page	NOUN
fcis-7512	251	13	1122–1131	1122–1131	NUM
fcis-7512	251	14	.	.	PUNCT
fcis-7512	252	1	[	[	X
fcis-7512	252	2	15	15	NUM
fcis-7512	252	3	]	]	X
fcis-7512	252	4	wenmeng	wenmeng	PROPN
fcis-7512	252	5	yu	yu	PROPN
fcis-7512	252	6	,	,	PUNCT
fcis-7512	252	7	hua	hua	PROPN
fcis-7512	252	8	xu	xu	PROPN
fcis-7512	252	9	,	,	PUNCT
fcis-7512	252	10	ziqi	ziqi	NOUN
fcis-7512	252	11	yuan	yuan	PROPN
fcis-7512	252	12	,	,	PUNCT
fcis-7512	252	13	and	and	CCONJ
fcis-7512	252	14	jiele	jiele	PROPN
fcis-7512	252	15	wu.2021	wu.2021	PROPN
fcis-7512	252	16	.	.	PUNCT
fcis-7512	253	1	learning	learn	VERB
fcis-7512	253	2	modality	modality	NOUN
fcis-7512	253	3	-	-	PUNCT
fcis-7512	253	4	specific	specific	ADJ
fcis-7512	253	5	representa	representa	NOUN
fcis-7512	253	6	-	-	PUNCT
fcis-7512	253	7	tions	tion	NOUN
fcis-7512	253	8	with	with	ADP
fcis-7512	253	9	selfsupervised	selfsupervise	VERB
fcis-7512	253	10	multi	multi	ADJ
fcis-7512	253	11	-	-	NOUN
fcis-7512	253	12	task	task	ADJ
fcis-7512	253	13	learning	learning	NOUN
fcis-7512	253	14	for	for	ADP
fcis-7512	253	15	multimodal	multimodal	ADJ
fcis-7512	253	16	sentiment	sentiment	NOUN
fcis-7512	253	17	analysis	analysis	NOUN
fcis-7512	253	18	.	.	PUNCT
fcis-7512	254	1	arxiv	arxiv	PROPN
fcis-7512	254	2	preprintarxiv	preprintarxiv	PROPN
fcis-7512	254	3	:	:	PUNCT
fcis-7512	255	1	2102.04830	2102.04830	NUM
fcis-7512	255	2	.	.	PUNCT
fcis-7512	256	1	[	[	X
fcis-7512	256	2	16	16	NUM
fcis-7512	256	3	]	]	X
fcis-7512	256	4	hochreiter	hochreiter	PROPN
fcis-7512	256	5	s	s	PART
fcis-7512	256	6	,	,	PUNCT
fcis-7512	256	7	schmidhuber	schmidhuber	PROPN
fcis-7512	256	8	j	j	PROPN
fcis-7512	256	9	.	.	PUNCT
fcis-7512	257	1	long	long	ADJ
fcis-7512	257	2	short	short	ADJ
fcis-7512	257	3	-	-	PUNCT
fcis-7512	257	4	term	term	NOUN
fcis-7512	257	5	memory[j	memory[j	NOUN
fcis-7512	257	6	]	]	PUNCT
fcis-7512	257	7	.	.	PUNCT
fcis-7512	258	1	neural	neural	ADJ
fcis-7512	258	2	computation	computation	NOUN
fcis-7512	258	3	,	,	PUNCT
fcis-7512	258	4	1997	1997	NUM
fcis-7512	258	5	,	,	PUNCT
fcis-7512	258	6	9(8):1735	9(8):1735	NUM
fcis-7512	258	7	-	-	PUNCT
fcis-7512	258	8	1780	1780	NUM
fcis-7512	258	9	.	.	PUNCT
fcis-7512	259	1	[	[	X
fcis-7512	259	2	17	17	NUM
fcis-7512	259	3	]	]	X
fcis-7512	259	4	amir	amir	PROPN
fcis-7512	259	5	zadeh	zadeh	PROPN
fcis-7512	259	6	,	,	PUNCT
fcis-7512	259	7	rowan	rowan	PROPN
fcis-7512	259	8	zellers	zeller	NOUN
fcis-7512	259	9	,	,	PUNCT
fcis-7512	259	10	eli	eli	PROPN
fcis-7512	259	11	pincus	pincus	PROPN
fcis-7512	259	12	,	,	PUNCT
fcis-7512	259	13	and	and	CCONJ
fcis-7512	259	14	louis	louis	NOUN
fcis-7512	259	15	-	-	PUNCT
fcis-7512	259	16	philippe	philippe	PROPN
fcis-7512	259	17	morency	morency	NOUN
fcis-7512	259	18	.	.	PUNCT
fcis-7512	260	1	2016	2016	NUM
fcis-7512	260	2	.	.	PUNCT
fcis-7512	261	1	multimodal	multimodal	ADJ
fcis-7512	261	2	sentiment	sentiment	NOUN
fcis-7512	261	3	intensity	intensity	NOUN
fcis-7512	261	4	analysis	analysis	NOUN
fcis-7512	261	5	in	in	ADP
fcis-7512	261	6	videos	video	NOUN
fcis-7512	261	7	:	:	PUNCT
fcis-7512	261	8	facial	facial	ADJ
fcis-7512	261	9	gestures	gesture	NOUN
fcis-7512	261	10	and	and	CCONJ
fcis-7512	261	11	verbal	verbal	ADJ
fcis-7512	261	12	messages	message	NOUN
fcis-7512	261	13	.	.	PUNCT
fcis-7512	262	1	ieee	ieee	NOUN
fcis-7512	262	2	intelligent	intelligent	ADJ
fcis-7512	262	3	systems	system	NOUN
fcis-7512	262	4	,	,	PUNCT
fcis-7512	262	5	31(6):82–88	31(6):82–88	NUM
fcis-7512	262	6	.	.	PUNCT
fcis-7512	263	1	[	[	X
fcis-7512	263	2	18	18	NUM
fcis-7512	263	3	]	]	X
fcis-7512	263	4	amirali	amirali	PROPN
fcis-7512	263	5	bagher	bagher	PROPN
fcis-7512	263	6	zadeh	zadeh	PROPN
fcis-7512	263	7	,	,	PUNCT
fcis-7512	263	8	paul	paul	PROPN
fcis-7512	263	9	pu	pu	PROPN
fcis-7512	263	10	liang	liang	PROPN
fcis-7512	263	11	,	,	PUNCT
fcis-7512	263	12	soujanya	soujanya	PROPN
fcis-7512	263	13	poria	poria	PROPN
fcis-7512	263	14	,	,	PUNCT
fcis-7512	263	15	erik	erik	PROPN
fcis-7512	263	16	cambria	cambria	PROPN
fcis-7512	263	17	,	,	PUNCT
fcis-7512	263	18	and	and	CCONJ
fcis-7512	263	19	louis	louis	NOUN
fcis-7512	263	20	-	-	PUNCT
fcis-7512	263	21	philippe	philippe	NOUN
fcis-7512	263	22	morency	morency	NOUN
fcis-7512	263	23	.	.	PUNCT
fcis-7512	264	1	2018.multimodal	2018.multimodal	ADJ
fcis-7512	264	2	language	language	NOUN
fcis-7512	264	3	analysis	analysis	NOUN
fcis-7512	264	4	in	in	ADP
fcis-7512	264	5	the	the	DET
fcis-7512	264	6	wild	wild	NOUN
fcis-7512	264	7	:	:	PUNCT
fcis-7512	264	8	cmu	cmu	ADJ
fcis-7512	264	9	-	-	PUNCT
fcis-7512	264	10	mosei	mosei	ADJ
fcis-7512	264	11	dataset	dataset	NOUN
fcis-7512	264	12	and	and	CCONJ
fcis-7512	264	13	interpretable	interpretable	ADJ
fcis-7512	264	14	dynamic	dynamic	ADJ
fcis-7512	264	15	fusion	fusion	NOUN
fcis-7512	264	16	graph	graph	NOUN
fcis-7512	264	17	.	.	PUNCT
fcis-7512	265	1	in	in	ADP
fcis-7512	265	2	proceedings	proceeding	NOUN
fcis-7512	265	3	of	of	ADP
fcis-7512	265	4	the	the	DET
fcis-7512	265	5	56th	56th	ADJ
fcis-7512	265	6	annual	annual	ADJ
fcis-7512	265	7	meeting	meeting	NOUN
fcis-7512	265	8	of	of	ADP
fcis-7512	265	9	the	the	DET
fcis-7512	265	10	association	association	NOUN
fcis-7512	265	11	for	for	ADP
fcis-7512	265	12	computational	computational	ADJ
fcis-7512	265	13	linguistics	linguistic	NOUN
fcis-7512	265	14	(	(	PUNCT
fcis-7512	265	15	volume	volume	NOUN
fcis-7512	265	16	1	1	NUM
fcis-7512	265	17	:	:	PUNCT
fcis-7512	265	18	long	long	ADJ
fcis-7512	265	19	papers	paper	NOUN
fcis-7512	265	20	)	)	PUNCT
fcis-7512	265	21	,	,	PUNCT
fcis-7512	265	22	pages	page	NOUN
fcis-7512	265	23	2236–2246	2236–2246	NUM
fcis-7512	265	24	.	.	PUNCT
fcis-7512	266	1	[	[	X
fcis-7512	266	2	19	19	NUM
fcis-7512	266	3	]	]	X
fcis-7512	266	4	liu	liu	PROPN
fcis-7512	266	5	,	,	PUNCT
fcis-7512	266	6	z.	z.	PROPN
fcis-7512	266	7	;	;	PUNCT
fcis-7512	266	8	shen	shen	PROPN
fcis-7512	266	9	,	,	PUNCT
fcis-7512	266	10	y	y	PROPN
fcis-7512	266	11	.	.	PUNCT
fcis-7512	266	12	;	;	PUNCT
fcis-7512	266	13	lakshminarasimhan	lakshminarasimhan	NOUN
fcis-7512	266	14	,	,	PUNCT
fcis-7512	266	15	v	v	NOUN
fcis-7512	266	16	.	.	PUNCT
fcis-7512	266	17	b.	b.	PROPN
fcis-7512	266	18	;	;	PUNCT
fcis-7512	266	19	liang	liang	PROPN
fcis-7512	266	20	,	,	PUNCT
fcis-7512	266	21	p	p	X
fcis-7512	266	22	.	.	PUNCT
fcis-7512	267	1	p	p	X
fcis-7512	267	2	.;zadeh	.;zadeh	PROPN
fcis-7512	267	3	,	,	PUNCT
fcis-7512	267	4	a.	a.	PROPN
fcis-7512	267	5	b.	b.	PROPN
fcis-7512	267	6	;	;	PUNCT
fcis-7512	267	7	and	and	CCONJ
fcis-7512	267	8	morency	morency	NOUN
fcis-7512	267	9	,	,	PUNCT
fcis-7512	267	10	l.-p	l.-p	PROPN
fcis-7512	267	11	.	.	PUNCT
fcis-7512	268	1	2018	2018	NUM
fcis-7512	268	2	.	.	PUNCT
fcis-7512	269	1	efficient	efficient	ADJ
fcis-7512	269	2	low	low	ADJ
fcis-7512	269	3	-	-	PUNCT
fcis-7512	269	4	rank	rank	NOUN
fcis-7512	269	5	multimodal	multimodal	NOUN
fcis-7512	269	6	fusion	fusion	NOUN
fcis-7512	269	7	with	with	ADP
fcis-7512	269	8	modality	modality	NOUN
fcis-7512	269	9	-	-	PUNCT
fcis-7512	269	10	specific	specific	ADJ
fcis-7512	269	11	factors	factor	NOUN
fcis-7512	269	12	.	.	PUNCT
fcis-7512	270	1	in	in	ADP
fcis-7512	270	2	proceedings	proceeding	NOUN
fcis-7512	270	3	of	of	ADP
fcis-7512	270	4	the	the	DET
fcis-7512	270	5	56th	56th	ADJ
fcis-7512	270	6	annual	annual	ADJ
fcis-7512	270	7	meeting	meeting	NOUN
fcis-7512	270	8	of	of	ADP
fcis-7512	270	9	the	the	DET
fcis-7512	270	10	association	association	NOUN
fcis-7512	270	11	for	for	ADP
fcis-7512	270	12	computational	computational	ADJ
fcis-7512	270	13	linguistics	linguistic	NOUN
fcis-7512	270	14	(	(	PUNCT
fcis-7512	270	15	v	v	NOUN
fcis-7512	270	16	olume	olume	NOUN
fcis-7512	270	17	1	1	NUM
fcis-7512	270	18	:	:	PUNCT
fcis-7512	270	19	long	long	ADJ
fcis-7512	270	20	papers	paper	NOUN
fcis-7512	270	21	)	)	PUNCT
fcis-7512	270	22	,	,	PUNCT
fcis-7512	270	23	2247	2247	NUM
fcis-7512	270	24	–	–	PUNCT
fcis-7512	270	25	2256	2256	NUM
fcis-7512	270	26	.	.	PUNCT
fcis-7512	271	1	[	[	X
fcis-7512	271	2	20	20	NUM
fcis-7512	271	3	]	]	SYM
fcis-7512	271	4	zadeh	zadeh	PROPN
fcis-7512	271	5	,	,	PUNCT
fcis-7512	271	6	a.	a.	PROPN
fcis-7512	271	7	;	;	PUNCT
fcis-7512	271	8	liang	liang	PROPN
fcis-7512	271	9	,	,	PUNCT
fcis-7512	271	10	p	p	X
fcis-7512	271	11	.	.	PUNCT
fcis-7512	272	1	p	p	X
fcis-7512	272	2	.	.	PUNCT
fcis-7512	273	1	;	;	PUNCT
fcis-7512	273	2	mazumder	mazumder	NOUN
fcis-7512	273	3	,	,	PUNCT
fcis-7512	273	4	n.	n.	NOUN
fcis-7512	273	5	;	;	PUNCT
fcis-7512	273	6	poria	poria	PROPN
fcis-7512	273	7	,	,	PUNCT
fcis-7512	273	8	s.	s.	PROPN
fcis-7512	273	9	;	;	PUNCT
fcis-7512	273	10	cambria	cambria	PROPN
fcis-7512	273	11	,	,	PUNCT
fcis-7512	273	12	e.	e.	PROPN
fcis-7512	273	13	;	;	PUNCT
fcis-7512	273	14	and	and	CCONJ
fcis-7512	273	15	morency	morency	NOUN
fcis-7512	273	16	,	,	PUNCT
fcis-7512	273	17	l.-p	l.-p	PROPN
fcis-7512	273	18	.	.	PUNCT
fcis-7512	274	1	2018a	2018a	NUM
fcis-7512	274	2	.	.	PUNCT
fcis-7512	274	3	memory	memory	NOUN
fcis-7512	274	4	fusion	fusion	NOUN
fcis-7512	274	5	network	network	NOUN
fcis-7512	274	6	for	for	ADP
fcis-7512	274	7	multiview	multiview	PROPN
fcis-7512	274	8	sequential	sequential	ADJ
fcis-7512	274	9	learning	learning	NOUN
fcis-7512	274	10	.	.	PUNCT
fcis-7512	275	1	arxiv	arxiv	PROPN
fcis-7512	275	2	preprintarxiv:1802.00927	preprintarxiv:1802.00927	NUM
fcis-7512	275	3	.	.	PUNCT
fcis-7512	276	1	[	[	X
fcis-7512	276	2	21	21	NUM
fcis-7512	276	3	]	]	X
fcis-7512	276	4	wang	wang	PROPN
fcis-7512	276	5	,	,	PUNCT
fcis-7512	276	6	y	y	PROPN
fcis-7512	276	7	.	.	PUNCT
fcis-7512	276	8	;	;	PUNCT
fcis-7512	276	9	shen	shen	PROPN
fcis-7512	276	10	,	,	PUNCT
fcis-7512	276	11	y	y	PROPN
fcis-7512	276	12	.	.	PUNCT
fcis-7512	276	13	;	;	PUNCT
fcis-7512	276	14	liu	liu	PROPN
fcis-7512	276	15	,	,	PUNCT
fcis-7512	276	16	z.	z.	PROPN
fcis-7512	276	17	;	;	PUNCT
fcis-7512	276	18	liang	liang	PROPN
fcis-7512	276	19	,	,	PUNCT
fcis-7512	276	20	p	p	X
fcis-7512	276	21	.	.	PUNCT
fcis-7512	277	1	p	p	X
fcis-7512	277	2	.	.	PUNCT
fcis-7512	277	3	;	;	PUNCT
fcis-7512	277	4	zadeh	zadeh	PROPN
fcis-7512	277	5	,	,	PUNCT
fcis-7512	277	6	a.	a.	NOUN
fcis-7512	277	7	;	;	PUNCT
fcis-7512	277	8	and	and	CCONJ
fcis-7512	277	9	morency	morency	NOUN
fcis-7512	277	10	,	,	PUNCT
fcis-7512	277	11	l.-p	l.-p	PROPN
fcis-7512	277	12	.	.	PUNCT
fcis-7512	278	1	2019	2019	NUM
fcis-7512	278	2	.	.	PUNCT
fcis-7512	278	3	words	word	NOUN
fcis-7512	278	4	can	can	AUX
fcis-7512	278	5	shift	shift	VERB
fcis-7512	278	6	:	:	PUNCT
fcis-7512	278	7	dynamically	dynamically	ADV
fcis-7512	278	8	adjusting	adjust	VERB
fcis-7512	278	9	word	word	NOUN
fcis-7512	278	10	representations	representation	NOUN
fcis-7512	278	11	using	use	VERB
fcis-7512	278	12	nonverbal	nonverbal	ADJ
fcis-7512	278	13	behaviors	behavior	NOUN
fcis-7512	278	14	.	.	PUNCT
fcis-7512	279	1	in	in	ADP
fcis-7512	279	2	proceedings	proceeding	NOUN
fcis-7512	279	3	of	of	ADP
fcis-7512	279	4	the	the	DET
fcis-7512	279	5	aaai	aaai	PROPN
fcis-7512	279	6	conference	conference	NOUN
fcis-7512	279	7	on	on	ADP
fcis-7512	279	8	artificial	artificial	ADJ
fcis-7512	279	9	intelligence	intelligence	NOUN
fcis-7512	279	10	,	,	PUNCT
fcis-7512	279	11	volume	volume	NOUN
fcis-7512	279	12	33	33	NUM
fcis-7512	279	13	,	,	PUNCT
fcis-7512	279	14	7216–7223	7216–7223	NUM
fcis-7512	279	15	.	.	PUNCT
fcis-7512	280	1	[	[	X
fcis-7512	280	2	22	22	NUM
fcis-7512	280	3	]	]	X
fcis-7512	280	4	rahman	rahman	PROPN
fcis-7512	280	5	,	,	PUNCT
fcis-7512	280	6	w.	w.	PROPN
fcis-7512	280	7	;	;	PUNCT
fcis-7512	280	8	hasan	hasan	PROPN
fcis-7512	280	9	,	,	PUNCT
fcis-7512	280	10	m.	m.	PROPN
fcis-7512	280	11	k.	k.	PROPN
fcis-7512	280	12	;	;	PUNCT
fcis-7512	280	13	lee	lee	PROPN
fcis-7512	280	14	,	,	PUNCT
fcis-7512	280	15	s.	s.	PROPN
fcis-7512	280	16	;	;	PUNCT
fcis-7512	280	17	zadeh	zadeh	PROPN
fcis-7512	280	18	,	,	PUNCT
fcis-7512	280	19	a.	a.	PROPN
fcis-7512	280	20	b.	b.	PROPN
fcis-7512	280	21	;	;	PUNCT
fcis-7512	280	22	mao	mao	PROPN
fcis-7512	280	23	,	,	PUNCT
fcis-7512	280	24	c.;morency	c.;morency	NOUN
fcis-7512	280	25	,	,	PUNCT
fcis-7512	280	26	l.-p	l.-p	PROPN
fcis-7512	280	27	.	.	PUNCT
fcis-7512	280	28	;	;	PUNCT
fcis-7512	280	29	and	and	CCONJ
fcis-7512	280	30	hoque	hoque	NOUN
fcis-7512	280	31	,	,	PUNCT
fcis-7512	280	32	e.	e.	PROPN
fcis-7512	280	33	2020	2020	NUM
fcis-7512	280	34	.	.	PUNCT
fcis-7512	281	1	integrating	integrate	VERB
fcis-7512	281	2	multimodal	multimodal	NOUN
fcis-7512	281	3	information	information	NOUN
fcis-7512	281	4	in	in	ADP
fcis-7512	281	5	large	large	ADJ
fcis-7512	281	6	pretrained	pretraine	VERB
fcis-7512	281	7	transformers	transformer	NOUN
fcis-7512	281	8	.	.	PUNCT
fcis-7512	282	1	in	in	ADP
fcis-7512	282	2	proceedings	proceeding	NOUN
fcis-7512	282	3	of	of	ADP
fcis-7512	282	4	the	the	DET
fcis-7512	282	5	58th	58th	ADJ
fcis-7512	282	6	annual	annual	ADJ
fcis-7512	282	7	meeting	meeting	NOUN
fcis-7512	282	8	of	of	ADP
fcis-7512	282	9	the	the	DET
fcis-7512	282	10	association	association	NOUN
fcis-7512	282	11	for	for	ADP
fcis-7512	282	12	computational	computational	ADJ
fcis-7512	282	13	linguistics	linguistic	NOUN
fcis-7512	282	14	,	,	PUNCT
fcis-7512	282	15	2359–2369	2359–2369	NUM
fcis-7512	282	16	.	.	PUNCT
fcis-7512	283	1	[	[	X
fcis-7512	283	2	23	23	NUM
fcis-7512	283	3	]	]	X
fcis-7512	283	4	han	han	PROPN
fcis-7512	283	5	w	w	PROPN
fcis-7512	283	6	,	,	PUNCT
fcis-7512	283	7	chen	chen	PROPN
fcis-7512	283	8	h	h	PROPN
fcis-7512	283	9	,	,	PUNCT
fcis-7512	283	10	poria	poria	PROPN
fcis-7512	283	11	s	s	PART
fcis-7512	283	12	.	.	PUNCT
fcis-7512	283	13	improving	improve	VERB
fcis-7512	283	14	multimodal	multimodal	ADJ
fcis-7512	283	15	fusion	fusion	NOUN
fcis-7512	283	16	with	with	ADP
fcis-7512	283	17	hierarchical	hierarchical	ADJ
fcis-7512	283	18	mutual	mutual	ADJ
fcis-7512	283	19	information	information	NOUN
fcis-7512	283	20	maximization	maximization	NOUN
fcis-7512	283	21	for	for	ADP
fcis-7512	283	22	multimodal	multimodal	ADJ
fcis-7512	283	23	sentiment	sentiment	NOUN
fcis-7512	283	24	analysis[j	analysis[j	PROPN
fcis-7512	283	25	]	]	PUNCT
fcis-7512	283	26	.	.	PUNCT
fcis-7512	283	27	2021	2021	NUM
fcis-7512	283	28	.	.	PUNCT
fcis-7512	284	1	arxiv	arxiv	PROPN
fcis-7512	284	2	preprintarxiv	preprintarxiv	PROPN
fcis-7512	284	3	:	:	PUNCT
fcis-7512	284	4	2102	2102	NUM
fcis-7512	284	5	.	.	PUNCT
fcis-7512	285	1	04830.2109.0041	04830.2109.0041	NUM
fcis-7512	285	2	.	.	PUNCT