id	sid	tid	token	lemma	pos
alkej-870	1	1	al	al	PROPN
alkej-870	1	2	-	-	PUNCT
alkej-870	1	3	khwarizmi	khwarizmi	PROPN
alkej-870	1	4	engineering	engineering	PROPN
alkej-870	1	5	journal	journal	PROPN
alkej-870	1	6	al	al	PROPN
alkej-870	1	7	-	-	PUNCT
alkej-870	1	8	khwarizmi	khwarizmi	PROPN
alkej-870	1	9	engineering	engineering	NOUN
alkej-870	1	10	journal	journal	PROPN
alkej-870	1	11	,	,	PUNCT
alkej-870	1	12	vol	vol	NOUN
alkej-870	1	13	.	.	PROPN
alkej-870	1	14	19	19	NUM
alkej-870	1	15	,	,	PUNCT
alkej-870	1	16	no	no	INTJ
alkej-870	1	17	.	.	NOUN
alkej-870	1	18	4	4	NUM
alkej-870	1	19	,	,	PUNCT
alkej-870	1	20	december	december	PROPN
alkej-870	1	21	,	,	PUNCT
alkej-870	1	22	(	(	PUNCT
alkej-870	1	23	2023	2023	NUM
alkej-870	1	24	)	)	PUNCT
alkej-870	2	1	p.	p.	NOUN
alkej-870	2	2	p.	p.	NOUN
alkej-870	2	3	4255	4255	NUM
alkej-870	3	1	an	an	DET
alkej-870	3	2	overview	overview	NOUN
alkej-870	3	3	of	of	ADP
alkej-870	3	4	audio	audio	ADJ
alkej-870	3	5	-	-	ADJ
alkej-870	3	6	visual	visual	ADJ
alkej-870	3	7	source	source	NOUN
alkej-870	3	8	separation	separation	NOUN
alkej-870	3	9	using	use	VERB
alkej-870	3	10	deep	deep	ADJ
alkej-870	3	11	learning	learning	NOUN
alkej-870	3	12	noorulhuda	noorulhuda	PROPN
alkej-870	3	13	mudhafar	mudhafar	PROPN
alkej-870	3	14	sulaiman	sulaiman	PROPN
alkej-870	3	15	*	*	PROPN
alkej-870	3	16	ahmed	ahmed	PROPN
alkej-870	3	17	al	al	PROPN
alkej-870	3	18	tmeme	tmeme	PROPN
alkej-870	3	19	*	*	PUNCT
alkej-870	3	20	*	*	PUNCT
alkej-870	3	21	mohammed	mohammed	PROPN
alkej-870	3	22	najah	najah	INTJ
alkej-870	3	23	mahdi	mahdi	PROPN
alkej-870	3	24	*	*	PUNCT
alkej-870	4	1	*	*	PUNCT
alkej-870	4	2	*	*	PUNCT
alkej-870	4	3	*	*	PUNCT
alkej-870	4	4	,	,	PUNCT
alkej-870	4	5	*	*	PUNCT
alkej-870	4	6	*	*	PUNCT
alkej-870	4	7	department	department	NOUN
alkej-870	4	8	of	of	ADP
alkej-870	4	9	information	information	NOUN
alkej-870	4	10	and	and	CCONJ
alkej-870	4	11	communications	communication	NOUN
alkej-870	4	12	engineering	engineering	NOUN
alkej-870	4	13	/	/	SYM
alkej-870	4	14	al	al	PROPN
alkej-870	4	15	-	-	PUNCT
alkej-870	4	16	khwarizmi	khwarizmi	PROPN
alkej-870	4	17	college	college	PROPN
alkej-870	4	18	of	of	ADP
alkej-870	4	19	engineering/	engineering/	NUM
alkej-870	4	20	university	university	NOUN
alkej-870	4	21	of	of	ADP
alkej-870	4	22	baghdad/	baghdad/	NUM
alkej-870	4	23	baghdad/	baghdad/	NUM
alkej-870	4	24	iraq	iraq	NOUN
alkej-870	4	25	*	*	PUNCT
alkej-870	5	1	*	*	PUNCT
alkej-870	5	2	*	*	PUNCT
alkej-870	5	3	adapt	adapt	VERB
alkej-870	5	4	centre/	centre/	NUM
alkej-870	5	5	school	school	NOUN
alkej-870	5	6	of	of	ADP
alkej-870	5	7	computing/	computing/	PROPN
alkej-870	5	8	dublin	dublin	PROPN
alkej-870	5	9	city	city	PROPN
alkej-870	5	10	university/	university/	PROPN
alkej-870	5	11	dublin	dublin	PROPN
alkej-870	5	12	d09	d09	PROPN
alkej-870	5	13	dxa0/	dxa0/	PROPN
alkej-870	5	14	ireland	ireland	PROPN
alkej-870	5	15	corresponding	correspond	VERB
alkej-870	5	16	author	author	NOUN
alkej-870	5	17	*	*	NOUN
alkej-870	5	18	email	email	NOUN
alkej-870	5	19	:	:	PUNCT
alkej-870	5	20	noor.abd2103m@kecbu.uobaghdad.edu.iq	noor.abd2103m@kecbu.uobaghdad.edu.iq	PROPN
alkej-870	5	21	*	*	PROPN
alkej-870	5	22	*	*	NOUN
alkej-870	5	23	email	email	NOUN
alkej-870	5	24	:	:	PUNCT
alkej-870	5	25	asattar@kecbu.uobaghdad.edu.iq	asattar@kecbu.uobaghdad.edu.iq	VERB
alkej-870	5	26	*	*	PUNCT
alkej-870	5	27	*	*	PUNCT
alkej-870	5	28	*	*	NOUN
alkej-870	5	29	email	email	NOUN
alkej-870	5	30	:	:	PUNCT
alkej-870	5	31	mohammed.mahdi@adaptcentre.ie	mohammed.mahdi@adaptcentre.ie	NOUN
alkej-870	5	32	(	(	PUNCT
alkej-870	5	33	received	receive	VERB
alkej-870	5	34	11	11	NUM
alkej-870	5	35	april	april	PROPN
alkej-870	5	36	2023	2023	NUM
alkej-870	5	37	;	;	PUNCT
alkej-870	5	38	accepted	accept	VERB
alkej-870	5	39	18	18	NUM
alkej-870	5	40	june	june	PROPN
alkej-870	5	41	2023	2023	NUM
alkej-870	5	42	)	)	PUNCT
alkej-870	5	43	3https://doi.org/10.22153/kej.2023.06.00	3https://doi.org/10.22153/kej.2023.06.00	PROPN
alkej-870	5	44	abstract	abstract	NOUN
alkej-870	5	45	in	in	ADP
alkej-870	5	46	this	this	DET
alkej-870	5	47	article	article	NOUN
alkej-870	5	48	,	,	PUNCT
alkej-870	5	49	the	the	DET
alkej-870	5	50	research	research	NOUN
alkej-870	5	51	presents	present	VERB
alkej-870	5	52	a	a	DET
alkej-870	5	53	general	general	ADJ
alkej-870	5	54	overview	overview	NOUN
alkej-870	5	55	of	of	ADP
alkej-870	5	56	deep	deep	ADJ
alkej-870	5	57	learning	learning	NOUN
alkej-870	5	58	-	-	PUNCT
alkej-870	5	59	based	base	VERB
alkej-870	5	60	avss	avss	NOUN
alkej-870	5	61	(	(	PUNCT
alkej-870	5	62	audio	audio	ADJ
alkej-870	5	63	-	-	ADJ
alkej-870	5	64	visual	visual	ADJ
alkej-870	5	65	source	source	NOUN
alkej-870	5	66	separation	separation	NOUN
alkej-870	5	67	)	)	PUNCT
alkej-870	5	68	systems	system	NOUN
alkej-870	5	69	.	.	PUNCT
alkej-870	6	1	avss	avss	NOUN
alkej-870	6	2	has	have	AUX
alkej-870	6	3	achieved	achieve	VERB
alkej-870	6	4	exceptional	exceptional	ADJ
alkej-870	6	5	results	result	NOUN
alkej-870	6	6	in	in	ADP
alkej-870	6	7	a	a	DET
alkej-870	6	8	number	number	NOUN
alkej-870	6	9	of	of	ADP
alkej-870	6	10	areas	area	NOUN
alkej-870	6	11	,	,	PUNCT
alkej-870	6	12	including	include	VERB
alkej-870	6	13	decreasing	decrease	VERB
alkej-870	6	14	noise	noise	NOUN
alkej-870	6	15	levels	level	NOUN
alkej-870	6	16	,	,	PUNCT
alkej-870	6	17	boosting	boost	VERB
alkej-870	6	18	speech	speech	NOUN
alkej-870	6	19	recognition	recognition	NOUN
alkej-870	6	20	,	,	PUNCT
alkej-870	6	21	and	and	CCONJ
alkej-870	6	22	improving	improve	VERB
alkej-870	6	23	audio	audio	ADJ
alkej-870	6	24	quality	quality	NOUN
alkej-870	6	25	.	.	PUNCT
alkej-870	7	1	the	the	DET
alkej-870	7	2	advantages	advantage	NOUN
alkej-870	7	3	and	and	CCONJ
alkej-870	7	4	disadvantages	disadvantage	NOUN
alkej-870	7	5	of	of	ADP
alkej-870	7	6	each	each	DET
alkej-870	7	7	deep	deep	ADJ
alkej-870	7	8	learning	learning	NOUN
alkej-870	7	9	model	model	NOUN
alkej-870	7	10	are	be	AUX
alkej-870	7	11	discussed	discuss	VERB
alkej-870	7	12	throughout	throughout	ADP
alkej-870	7	13	the	the	DET
alkej-870	7	14	research	research	NOUN
alkej-870	7	15	as	as	SCONJ
alkej-870	7	16	it	it	PRON
alkej-870	7	17	reviews	review	VERB
alkej-870	7	18	various	various	ADJ
alkej-870	7	19	current	current	ADJ
alkej-870	7	20	experiments	experiment	NOUN
alkej-870	7	21	on	on	ADP
alkej-870	7	22	avss	avss	NOUN
alkej-870	7	23	.	.	PUNCT
alkej-870	8	1	the	the	DET
alkej-870	8	2	tcd	tcd	PROPN
alkej-870	8	3	timit	timit	PROPN
alkej-870	8	4	dataset	dataset	NOUN
alkej-870	8	5	(	(	PUNCT
alkej-870	8	6	which	which	PRON
alkej-870	8	7	contains	contain	VERB
alkej-870	8	8	top	top	ADJ
alkej-870	8	9	-	-	PUNCT
alkej-870	8	10	notch	notch	NOUN
alkej-870	8	11	audio	audio	ADJ
alkej-870	8	12	and	and	CCONJ
alkej-870	8	13	video	video	NOUN
alkej-870	8	14	recordings	recording	NOUN
alkej-870	8	15	created	create	VERB
alkej-870	8	16	especially	especially	ADV
alkej-870	8	17	for	for	ADP
alkej-870	8	18	speech	speech	NOUN
alkej-870	8	19	recognition	recognition	NOUN
alkej-870	8	20	tasks	task	NOUN
alkej-870	8	21	)	)	PUNCT
alkej-870	8	22	and	and	CCONJ
alkej-870	8	23	the	the	DET
alkej-870	8	24	voxceleb	voxceleb	NOUN
alkej-870	8	25	dataset	dataset	NOUN
alkej-870	8	26	(	(	PUNCT
alkej-870	8	27	a	a	DET
alkej-870	8	28	sizable	sizable	ADJ
alkej-870	8	29	collection	collection	NOUN
alkej-870	8	30	of	of	ADP
alkej-870	8	31	brief	brief	ADJ
alkej-870	8	32	audio	audio	ADJ
alkej-870	8	33	-	-	ADJ
alkej-870	8	34	visual	visual	ADJ
alkej-870	8	35	clips	clip	NOUN
alkej-870	8	36	with	with	ADP
alkej-870	8	37	human	human	ADJ
alkej-870	8	38	speech	speech	NOUN
alkej-870	8	39	)	)	PUNCT
alkej-870	8	40	are	be	AUX
alkej-870	8	41	just	just	ADV
alkej-870	8	42	a	a	DET
alkej-870	8	43	couple	couple	NOUN
alkej-870	8	44	of	of	ADP
alkej-870	8	45	the	the	DET
alkej-870	8	46	useful	useful	ADJ
alkej-870	8	47	datasets	dataset	NOUN
alkej-870	8	48	summarized	summarize	VERB
alkej-870	8	49	in	in	ADP
alkej-870	8	50	the	the	DET
alkej-870	8	51	paper	paper	NOUN
alkej-870	8	52	that	that	PRON
alkej-870	8	53	can	can	AUX
alkej-870	8	54	be	be	AUX
alkej-870	8	55	used	use	VERB
alkej-870	8	56	to	to	PART
alkej-870	8	57	test	test	VERB
alkej-870	8	58	avss	avss	NOUN
alkej-870	8	59	systems	system	NOUN
alkej-870	8	60	.	.	PUNCT
alkej-870	9	1	in	in	ADP
alkej-870	9	2	its	its	PRON
alkej-870	9	3	basic	basic	ADJ
alkej-870	9	4	form	form	NOUN
alkej-870	9	5	,	,	PUNCT
alkej-870	9	6	this	this	DET
alkej-870	9	7	review	review	NOUN
alkej-870	9	8	aims	aim	VERB
alkej-870	9	9	to	to	PART
alkej-870	9	10	highlight	highlight	VERB
alkej-870	9	11	the	the	DET
alkej-870	9	12	growing	grow	VERB
alkej-870	9	13	importance	importance	NOUN
alkej-870	9	14	of	of	ADP
alkej-870	9	15	avss	avss	NOUN
alkej-870	9	16	in	in	ADP
alkej-870	9	17	improving	improve	VERB
alkej-870	9	18	the	the	DET
alkej-870	9	19	quality	quality	NOUN
alkej-870	9	20	of	of	ADP
alkej-870	9	21	audio	audio	ADJ
alkej-870	9	22	signals	signal	NOUN
alkej-870	9	23	.	.	PUNCT
alkej-870	10	1	keywords	keyword	NOUN
alkej-870	10	2	:	:	PUNCT
alkej-870	10	3	audio	audio	ADJ
alkej-870	10	4	-	-	ADJ
alkej-870	10	5	visual	visual	ADJ
alkej-870	10	6	,	,	PUNCT
alkej-870	10	7	source	source	NOUN
alkej-870	10	8	separation	separation	NOUN
alkej-870	10	9	,	,	PUNCT
alkej-870	10	10	deep	deep	ADJ
alkej-870	10	11	learning	learning	NOUN
alkej-870	10	12	,	,	PUNCT
alkej-870	10	13	cnn	cnn	PROPN
alkej-870	10	14	,	,	PUNCT
alkej-870	10	15	datasets	dataset	NOUN
alkej-870	10	16	.	.	PUNCT
alkej-870	11	1	1	1	X
alkej-870	11	2	.	.	X
alkej-870	11	3	introduction	introduction	NOUN
alkej-870	11	4	the	the	DET
alkej-870	11	5	process	process	NOUN
alkej-870	11	6	of	of	ADP
alkej-870	11	7	distinguishing	distinguish	VERB
alkej-870	11	8	particular	particular	ADJ
alkej-870	11	9	audio	audio	NOUN
alkej-870	11	10	sources	source	NOUN
alkej-870	11	11	from	from	ADP
alkej-870	11	12	a	a	DET
alkej-870	11	13	mixture	mixture	NOUN
alkej-870	11	14	of	of	ADP
alkej-870	11	15	audio	audio	ADJ
alkej-870	11	16	signals	signal	NOUN
alkej-870	11	17	using	use	VERB
alkej-870	11	18	visual	visual	ADJ
alkej-870	11	19	indicators	indicator	NOUN
alkej-870	11	20	as	as	SCONJ
alkej-870	11	21	further	further	ADJ
alkej-870	11	22	information	information	NOUN
alkej-870	11	23	is	be	AUX
alkej-870	11	24	known	know	VERB
alkej-870	11	25	as	as	ADP
alkej-870	11	26	audio	audio	ADJ
alkej-870	11	27	-	-	ADJ
alkej-870	11	28	visual	visual	ADJ
alkej-870	11	29	source	source	NOUN
alkej-870	11	30	separation	separation	NOUN
alkej-870	11	31	.	.	PUNCT
alkej-870	12	1	this	this	DET
alkej-870	12	2	method	method	NOUN
alkej-870	12	3	differs	differ	VERB
alkej-870	12	4	from	from	ADP
alkej-870	12	5	conventional	conventional	ADJ
alkej-870	12	6	ones	one	NOUN
alkej-870	12	7	that	that	PRON
alkej-870	12	8	exclusively	exclusively	ADV
alkej-870	12	9	rely	rely	VERB
alkej-870	12	10	on	on	ADP
alkej-870	12	11	the	the	DET
alkej-870	12	12	audio	audio	ADJ
alkej-870	12	13	stream	stream	NOUN
alkej-870	12	14	for	for	ADP
alkej-870	12	15	source	source	NOUN
alkej-870	12	16	separation	separation	NOUN
alkej-870	12	17	[	[	X
alkej-870	12	18	15	15	NUM
alkej-870	12	19	]	]	PUNCT
alkej-870	12	20	.	.	PUNCT
alkej-870	13	1	in	in	ADP
alkej-870	13	2	other	other	ADJ
alkej-870	13	3	words	word	NOUN
alkej-870	13	4	,	,	PUNCT
alkej-870	13	5	it	it	PRON
alkej-870	13	6	is	be	AUX
alkej-870	13	7	a	a	DET
alkej-870	13	8	technique	technique	NOUN
alkej-870	13	9	that	that	PRON
alkej-870	13	10	utilizes	utilize	VERB
alkej-870	13	11	both	both	DET
alkej-870	13	12	auditory	auditory	ADJ
alkej-870	13	13	and	and	CCONJ
alkej-870	13	14	visual	visual	ADJ
alkej-870	13	15	information	information	NOUN
alkej-870	13	16	to	to	PART
alkej-870	13	17	separate	separate	VERB
alkej-870	13	18	individual	individual	ADJ
alkej-870	13	19	sound	sound	NOUN
alkej-870	13	20	sources	source	NOUN
alkej-870	13	21	from	from	ADP
alkej-870	13	22	a	a	DET
alkej-870	13	23	mixed	mixed	ADJ
alkej-870	13	24	audio	audio	ADJ
alkej-870	13	25	signal	signal	NOUN
alkej-870	13	26	.	.	PUNCT
alkej-870	14	1	avss	avss	NOUN
alkej-870	14	2	algorithms	algorithm	NOUN
alkej-870	14	3	generally	generally	ADV
alkej-870	14	4	take	take	VERB
alkej-870	14	5	an	an	DET
alkej-870	14	6	input	input	NOUN
alkej-870	14	7	signal	signal	NOUN
alkej-870	14	8	containing	contain	VERB
alkej-870	14	9	multiple	multiple	ADJ
alkej-870	14	10	sound	sound	NOUN
alkej-870	14	11	sources	source	NOUN
alkej-870	14	12	and	and	CCONJ
alkej-870	14	13	utilize	utilize	VERB
alkej-870	14	14	visual	visual	ADJ
alkej-870	14	15	information	information	NOUN
alkej-870	14	16	,	,	PUNCT
alkej-870	14	17	such	such	ADJ
alkej-870	14	18	as	as	ADP
alkej-870	14	19	lip	lip	NOUN
alkej-870	14	20	movement	movement	NOUN
alkej-870	14	21	or	or	CCONJ
alkej-870	14	22	facial	facial	ADJ
alkej-870	14	23	expressions	expression	NOUN
alkej-870	14	24	,	,	PUNCT
alkej-870	14	25	to	to	PART
alkej-870	14	26	separate	separate	VERB
alkej-870	14	27	the	the	DET
alkej-870	14	28	individual	individual	ADJ
alkej-870	14	29	sources	source	NOUN
alkej-870	14	30	.	.	PUNCT
alkej-870	15	1	the	the	DET
alkej-870	15	2	visual	visual	ADJ
alkej-870	15	3	information	information	NOUN
alkej-870	15	4	can	can	AUX
alkej-870	15	5	be	be	AUX
alkej-870	15	6	captured	capture	VERB
alkej-870	15	7	from	from	ADP
alkej-870	15	8	various	various	ADJ
alkej-870	15	9	sources	source	NOUN
alkej-870	15	10	,	,	PUNCT
alkej-870	15	11	such	such	ADJ
alkej-870	15	12	as	as	ADP
alkej-870	15	13	video	video	NOUN
alkej-870	15	14	recordings	recording	NOUN
alkej-870	15	15	or	or	CCONJ
alkej-870	15	16	real	real	ADJ
alkej-870	15	17	-	-	PUNCT
alkej-870	15	18	time	time	NOUN
alkej-870	15	19	cameras	camera	NOUN
alkej-870	15	20	.	.	PUNCT
alkej-870	16	1	the	the	DET
alkej-870	16	2	main	main	ADJ
alkej-870	16	3	advantage	advantage	NOUN
alkej-870	16	4	of	of	ADP
alkej-870	16	5	audiovisual	audiovisual	ADJ
alkej-870	16	6	source	source	NOUN
alkej-870	16	7	separation	separation	NOUN
alkej-870	16	8	over	over	ADP
alkej-870	16	9	traditional	traditional	ADJ
alkej-870	16	10	audioonly	audioonly	ADJ
alkej-870	16	11	approaches	approach	NOUN
alkej-870	16	12	is	be	AUX
alkej-870	16	13	using	use	VERB
alkej-870	16	14	visual	visual	ADJ
alkej-870	16	15	signals	signal	NOUN
alkej-870	16	16	to	to	PART
alkej-870	16	17	improve	improve	VERB
alkej-870	16	18	the	the	DET
alkej-870	16	19	separation	separation	NOUN
alkej-870	16	20	quality	quality	NOUN
alkej-870	16	21	.	.	PUNCT
alkej-870	17	1	for	for	ADP
alkej-870	17	2	example	example	NOUN
alkej-870	17	3	,	,	PUNCT
alkej-870	17	4	visual	visual	ADJ
alkej-870	17	5	cues	cue	NOUN
alkej-870	17	6	can	can	AUX
alkej-870	17	7	help	help	VERB
alkej-870	17	8	separate	separate	VERB
alkej-870	17	9	speech	speech	NOUN
alkej-870	17	10	from	from	ADP
alkej-870	17	11	background	background	NOUN
alkej-870	17	12	noise	noise	NOUN
alkej-870	17	13	[	[	X
alkej-870	17	14	46	46	NUM
alkej-870	17	15	]	]	PUNCT
alkej-870	17	16	or	or	CCONJ
alkej-870	17	17	separate	separate	ADJ
alkej-870	17	18	multiple	multiple	ADJ
alkej-870	17	19	speakers	speaker	NOUN
alkej-870	17	20	in	in	ADP
alkej-870	17	21	a	a	DET
alkej-870	17	22	video	video	NOUN
alkej-870	17	23	recording	recording	NOUN
alkej-870	17	24	[	[	X
alkej-870	17	25	16	16	NUM
alkej-870	17	26	-	-	SYM
alkej-870	17	27	17	17	NUM
alkej-870	17	28	]	]	PUNCT
alkej-870	17	29	.	.	PUNCT
alkej-870	18	1	one	one	NUM
alkej-870	18	2	advanced	advanced	ADJ
alkej-870	18	3	method	method	NOUN
alkej-870	18	4	for	for	ADP
alkej-870	18	5	audio	audio	ADJ
alkej-870	18	6	-	-	ADJ
alkej-870	18	7	visual	visual	ADJ
alkej-870	18	8	source	source	NOUN
alkej-870	18	9	separation	separation	NOUN
alkej-870	18	10	involves	involve	VERB
alkej-870	18	11	the	the	DET
alkej-870	18	12	use	use	NOUN
alkej-870	18	13	of	of	ADP
alkej-870	18	14	deep	deep	ADJ
alkej-870	18	15	learning	learning	NOUN
alkej-870	18	16	techniques	technique	NOUN
alkej-870	18	17	[	[	X
alkej-870	18	18	10	10	NUM
alkej-870	18	19	-	-	SYM
alkej-870	18	20	12,14,22,34,37	12,14,22,34,37	NUM
alkej-870	18	21	-	-	PUNCT
alkej-870	18	22	38	38	NUM
alkej-870	18	23	]	]	PUNCT
alkej-870	18	24	.	.	PUNCT
alkej-870	19	1	here	here	ADV
alkej-870	19	2	are	be	AUX
alkej-870	19	3	some	some	DET
alkej-870	19	4	specific	specific	ADJ
alkej-870	19	5	reasons	reason	NOUN
alkej-870	19	6	why	why	SCONJ
alkej-870	19	7	audio	audio	ADJ
alkej-870	19	8	-	-	ADJ
alkej-870	19	9	visual	visual	ADJ
alkej-870	19	10	source	source	NOUN
alkej-870	19	11	separation	separation	NOUN
alkej-870	19	12	is	be	AUX
alkej-870	19	13	important	important	ADJ
alkej-870	19	14	:	:	PUNCT
alkej-870	19	15	1	1	NUM
alkej-870	19	16	-	-	PUNCT
alkej-870	19	17	enhancing	enhance	VERB
alkej-870	19	18	speech	speech	NOUN
alkej-870	19	19	intelligibility	intelligibility	NOUN
alkej-870	19	20	:	:	PUNCT
alkej-870	19	21	in	in	ADP
alkej-870	19	22	scenarios	scenario	NOUN
alkej-870	19	23	where	where	SCONJ
alkej-870	19	24	speech	speech	NOUN
alkej-870	19	25	is	be	AUX
alkej-870	19	26	degraded	degrade	VERB
alkej-870	19	27	by	by	ADP
alkej-870	19	28	noise	noise	NOUN
alkej-870	19	29	or	or	CCONJ
alkej-870	19	30	overlapping	overlap	VERB
alkej-870	19	31	sources	source	NOUN
alkej-870	19	32	,	,	PUNCT
alkej-870	19	33	audio	audio	ADJ
alkej-870	19	34	-	-	ADJ
alkej-870	19	35	visual	visual	ADJ
alkej-870	19	36	source	source	NOUN
alkej-870	19	37	separation	separation	NOUN
alkej-870	19	38	can	can	AUX
alkej-870	19	39	be	be	AUX
alkej-870	19	40	used	use	VERB
alkej-870	19	41	to	to	PART
alkej-870	19	42	isolate	isolate	VERB
alkej-870	19	43	the	the	DET
alkej-870	19	44	target	target	NOUN
alkej-870	19	45	speaker	speaker	NOUN
alkej-870	19	46	's	's	PART
alkej-870	19	47	voice	voice	NOUN
alkej-870	19	48	and	and	CCONJ
alkej-870	19	49	improve	improve	VERB
alkej-870	19	50	speech	speech	NOUN
alkej-870	19	51	intelligibility	intelligibility	NOUN
alkej-870	19	52	[	[	X
alkej-870	19	53	10	10	NUM
alkej-870	19	54	-	-	SYM
alkej-870	19	55	12	12	NUM
alkej-870	19	56	,	,	PUNCT
alkej-870	19	57	17	17	NUM
alkej-870	19	58	,	,	PUNCT
alkej-870	19	59	20	20	NUM
alkej-870	19	60	]	]	PUNCT
alkej-870	19	61	.	.	PUNCT
alkej-870	20	1	2	2	NUM
alkej-870	20	2	-	-	PUNCT
alkej-870	20	3	improving	improve	VERB
alkej-870	20	4	speech	speech	NOUN
alkej-870	20	5	recognition	recognition	NOUN
alkej-870	20	6	:	:	PUNCT
alkej-870	20	7	audio	audio	ADJ
alkej-870	20	8	-	-	ADJ
alkej-870	20	9	visual	visual	ADJ
alkej-870	20	10	source	source	NOUN
alkej-870	20	11	separation	separation	NOUN
alkej-870	20	12	can	can	AUX
alkej-870	20	13	also	also	ADV
alkej-870	20	14	improve	improve	VERB
alkej-870	20	15	speech	speech	NOUN
alkej-870	20	16	recognition	recognition	NOUN
alkej-870	20	17	systems	system	NOUN
alkej-870	20	18	'	'	PART
alkej-870	20	19	accuracy	accuracy	NOUN
alkej-870	20	20	,	,	PUNCT
alkej-870	20	21	by	by	ADP
alkej-870	20	22	separating	separate	VERB
alkej-870	20	23	out	out	ADP
alkej-870	20	24	individual	individual	ADJ
alkej-870	20	25	speakers	speaker	NOUN
alkej-870	20	26	and	and	CCONJ
alkej-870	20	27	reducing	reduce	VERB
alkej-870	20	28	interference	interference	NOUN
alkej-870	20	29	mailto:noor.abd2103m@kecbu.uobaghdad.edu.iq	mailto:noor.abd2103m@kecbu.uobaghdad.edu.iq	NOUN
alkej-870	20	30	mailto:asattar@kecbu.uobaghdad.edu.iq	mailto:asattar@kecbu.uobaghdad.edu.iq	PROPN
alkej-870	20	31	mailto:mohammed.mahdi@adaptcentre.ie	mailto:mohammed.mahdi@adaptcentre.ie	PROPN
alkej-870	20	32	https://doi.org/10.22153/kej.2023.06.003	https://doi.org/10.22153/kej.2023.06.003	NOUN
alkej-870	20	33	nooralhuda	nooralhuda	NOUN
alkej-870	20	34	mudhafar	mudhafar	PROPN
alkej-870	20	35	sleman	sleman	PROPN
alkej-870	20	36	al	al	PROPN
alkej-870	20	37	-	-	PUNCT
alkej-870	20	38	khwarizmi	khwarizmi	PROPN
alkej-870	20	39	engineering	engineering	NOUN
alkej-870	20	40	journal	journal	PROPN
alkej-870	20	41	,	,	PUNCT
alkej-870	20	42	vol	vol	NOUN
alkej-870	20	43	.	.	PROPN
alkej-870	20	44	19	19	NUM
alkej-870	20	45	,	,	PUNCT
alkej-870	20	46	no	no	INTJ
alkej-870	20	47	.	.	NOUN
alkej-870	20	48	4	4	NUM
alkej-870	20	49	,	,	PUNCT
alkej-870	20	50	p.p	p.p	PROPN
alkej-870	20	51	.	.	NOUN
alkej-870	20	52	4255	4255	NUM
alkej-870	20	53	(	(	PUNCT
alkej-870	20	54	2023	2023	NUM
alkej-870	20	55	)	)	PUNCT
alkej-870	20	56	43	43	NUM
alkej-870	20	57	from	from	ADP
alkej-870	20	58	other	other	ADJ
alkej-870	20	59	sources	source	NOUN
alkej-870	20	60	.	.	PUNCT
alkej-870	21	1	in	in	ADP
alkej-870	21	2	[	[	X
alkej-870	21	3	25	25	NUM
alkej-870	21	4	]	]	PUNCT
alkej-870	21	5	the	the	DET
alkej-870	21	6	authors	author	NOUN
alkej-870	21	7	utilized	utilize	VERB
alkej-870	21	8	a	a	DET
alkej-870	21	9	separation	separation	NOUN
alkej-870	21	10	system	system	NOUN
alkej-870	21	11	to	to	PART
alkej-870	21	12	separate	separate	VERB
alkej-870	21	13	voices	voice	NOUN
alkej-870	21	14	from	from	ADP
alkej-870	21	15	a	a	DET
alkej-870	21	16	mixed	mixed	ADJ
alkej-870	21	17	audio	audio	NOUN
alkej-870	21	18	signal	signal	NOUN
alkej-870	21	19	,	,	PUNCT
alkej-870	21	20	then	then	ADV
alkej-870	21	21	fed	feed	VERB
alkej-870	21	22	the	the	DET
alkej-870	21	23	separated	separate	VERB
alkej-870	21	24	voices	voice	NOUN
alkej-870	21	25	into	into	ADP
alkej-870	21	26	an	an	DET
alkej-870	21	27	automatic	automatic	ADJ
alkej-870	21	28	speech	speech	NOUN
alkej-870	21	29	recognition	recognition	NOUN
alkej-870	21	30	(	(	PUNCT
alkej-870	21	31	asr	asr	NOUN
alkej-870	21	32	)	)	PUNCT
alkej-870	21	33	system	system	NOUN
alkej-870	21	34	for	for	ADP
alkej-870	21	35	further	further	ADJ
alkej-870	21	36	processing	processing	NOUN
alkej-870	21	37	.	.	PUNCT
alkej-870	22	1	in	in	ADP
alkej-870	22	2	[	[	X
alkej-870	22	3	40	40	NUM
alkej-870	22	4	]	]	PUNCT
alkej-870	22	5	the	the	DET
alkej-870	22	6	authors	author	NOUN
alkej-870	22	7	combined	combine	VERB
alkej-870	22	8	the	the	DET
alkej-870	22	9	spatial	spatial	ADJ
alkej-870	22	10	information	information	NOUN
alkej-870	22	11	from	from	ADP
alkej-870	22	12	a	a	DET
alkej-870	22	13	microphone	microphone	NOUN
alkej-870	22	14	array	array	NOUN
alkej-870	22	15	with	with	ADP
alkej-870	22	16	visual	visual	ADJ
alkej-870	22	17	information	information	NOUN
alkej-870	22	18	from	from	ADP
alkej-870	22	19	a	a	DET
alkej-870	22	20	camera	camera	NOUN
alkej-870	22	21	to	to	PART
alkej-870	22	22	improve	improve	VERB
alkej-870	22	23	the	the	DET
alkej-870	22	24	separation	separation	NOUN
alkej-870	22	25	and	and	CCONJ
alkej-870	22	26	recognition	recognition	NOUN
alkej-870	22	27	of	of	ADP
alkej-870	22	28	multiple	multiple	ADJ
alkej-870	22	29	speakers	speaker	NOUN
alkej-870	22	30	in	in	ADP
alkej-870	22	31	a	a	DET
alkej-870	22	32	noisy	noisy	ADJ
alkej-870	22	33	and	and	CCONJ
alkej-870	22	34	reverberant	reverberant	ADJ
alkej-870	22	35	environment	environment	NOUN
alkej-870	22	36	.	.	PUNCT
alkej-870	23	1	3	3	X
alkej-870	23	2	-	-	PUNCT
alkej-870	23	3	enabling	enable	VERB
alkej-870	23	4	better	well	ADJ
alkej-870	23	5	audio	audio	ADJ
alkej-870	23	6	and	and	CCONJ
alkej-870	23	7	video	video	NOUN
alkej-870	23	8	processing	processing	NOUN
alkej-870	23	9	:	:	PUNCT
alkej-870	23	10	audio	audio	ADJ
alkej-870	23	11	-	-	ADJ
alkej-870	23	12	visual	visual	ADJ
alkej-870	23	13	source	source	NOUN
alkej-870	23	14	separation	separation	NOUN
alkej-870	23	15	can	can	AUX
alkej-870	23	16	be	be	AUX
alkej-870	23	17	used	use	VERB
alkej-870	23	18	to	to	PART
alkej-870	23	19	separate	separate	VERB
alkej-870	23	20	out	out	ADP
alkej-870	23	21	different	different	ADJ
alkej-870	23	22	sources	source	NOUN
alkej-870	23	23	of	of	ADP
alkej-870	23	24	sound	sound	NOUN
alkej-870	23	25	in	in	ADP
alkej-870	23	26	a	a	DET
alkej-870	23	27	video	video	NOUN
alkej-870	23	28	recording	recording	NOUN
alkej-870	23	29	,	,	PUNCT
alkej-870	23	30	which	which	PRON
alkej-870	23	31	can	can	AUX
alkej-870	23	32	enable	enable	VERB
alkej-870	23	33	better	well	ADJ
alkej-870	23	34	postprocessing	postprocessing	NOUN
alkej-870	23	35	of	of	ADP
alkej-870	23	36	both	both	DET
alkej-870	23	37	the	the	DET
alkej-870	23	38	audio	audio	NOUN
alkej-870	23	39	and	and	CCONJ
alkej-870	23	40	video	video	NOUN
alkej-870	23	41	[	[	X
alkej-870	23	42	23	23	NUM
alkej-870	23	43	,	,	PUNCT
alkej-870	23	44	28	28	NUM
alkej-870	23	45	,	,	PUNCT
alkej-870	23	46	47	47	NUM
alkej-870	23	47	]	]	PUNCT
alkej-870	23	48	.	.	PUNCT
alkej-870	24	1	overall	overall	ADJ
alkej-870	24	2	,	,	PUNCT
alkej-870	24	3	audio	audio	ADJ
alkej-870	24	4	-	-	ADJ
alkej-870	24	5	visual	visual	ADJ
alkej-870	24	6	source	source	NOUN
alkej-870	24	7	separation	separation	NOUN
alkej-870	24	8	can	can	AUX
alkej-870	24	9	help	help	VERB
alkej-870	24	10	address	address	VERB
alkej-870	24	11	some	some	PRON
alkej-870	24	12	of	of	ADP
alkej-870	24	13	the	the	DET
alkej-870	24	14	challenges	challenge	NOUN
alkej-870	24	15	in	in	ADP
alkej-870	24	16	speech	speech	NOUN
alkej-870	24	17	recognition	recognition	NOUN
alkej-870	24	18	by	by	ADP
alkej-870	24	19	improving	improve	VERB
alkej-870	24	20	the	the	DET
alkej-870	24	21	quality	quality	NOUN
alkej-870	24	22	and	and	CCONJ
alkej-870	24	23	intelligibility	intelligibility	NOUN
alkej-870	24	24	of	of	ADP
alkej-870	24	25	speech	speech	NOUN
alkej-870	24	26	signals	signal	NOUN
alkej-870	24	27	.	.	PUNCT
alkej-870	25	1	additionally	additionally	ADV
alkej-870	25	2	,	,	PUNCT
alkej-870	25	3	the	the	DET
alkej-870	25	4	development	development	NOUN
alkej-870	25	5	of	of	ADP
alkej-870	25	6	audio	audio	ADJ
alkej-870	25	7	-	-	ADJ
alkej-870	25	8	visual	visual	ADJ
alkej-870	25	9	source	source	NOUN
alkej-870	25	10	separation	separation	NOUN
alkej-870	25	11	methods	method	NOUN
alkej-870	25	12	has	have	AUX
alkej-870	25	13	led	lead	VERB
alkej-870	25	14	to	to	ADP
alkej-870	25	15	the	the	DET
alkej-870	25	16	creation	creation	NOUN
alkej-870	25	17	of	of	ADP
alkej-870	25	18	large	large	ADJ
alkej-870	25	19	-	-	PUNCT
alkej-870	25	20	scale	scale	NOUN
alkej-870	25	21	datasets	dataset	NOUN
alkej-870	25	22	that	that	PRON
alkej-870	25	23	can	can	AUX
alkej-870	25	24	be	be	AUX
alkej-870	25	25	used	use	VERB
alkej-870	25	26	for	for	ADP
alkej-870	25	27	further	further	ADJ
alkej-870	25	28	research	research	NOUN
alkej-870	25	29	in	in	ADP
alkej-870	25	30	related	related	ADJ
alkej-870	25	31	fields	field	NOUN
alkej-870	25	32	[	[	X
alkej-870	25	33	17	17	NUM
alkej-870	25	34	,	,	PUNCT
alkej-870	25	35	50	50	NUM
alkej-870	25	36	-	-	SYM
alkej-870	25	37	60	60	NUM
alkej-870	25	38	]	]	PUNCT
alkej-870	25	39	.	.	PUNCT
alkej-870	26	1	2	2	X
alkej-870	26	2	.	.	X
alkej-870	26	3	deep	deep	ADJ
alkej-870	26	4	-	-	PUNCT
alkej-870	26	5	learning	learning	NOUN
alkej-870	26	6	for	for	ADP
alkej-870	26	7	audio	audio	ADJ
alkej-870	26	8	-	-	ADJ
alkej-870	26	9	visual	visual	ADJ
alkej-870	26	10	source	source	NOUN
alkej-870	26	11	separation	separation	NOUN
alkej-870	26	12	deep	deep	ADJ
alkej-870	26	13	learning	learning	NOUN
alkej-870	26	14	techniques	technique	NOUN
alkej-870	26	15	have	have	AUX
alkej-870	26	16	emerged	emerge	VERB
alkej-870	26	17	as	as	ADP
alkej-870	26	18	powerful	powerful	ADJ
alkej-870	26	19	tools	tool	NOUN
alkej-870	26	20	in	in	ADP
alkej-870	26	21	a	a	DET
alkej-870	26	22	wide	wide	ADJ
alkej-870	26	23	range	range	NOUN
alkej-870	26	24	of	of	ADP
alkej-870	26	25	applications	application	NOUN
alkej-870	26	26	[	[	X
alkej-870	26	27	69	69	NUM
alkej-870	26	28	]	]	PUNCT
alkej-870	26	29	.	.	PUNCT
alkej-870	27	1	deep	deep	ADJ
alkej-870	27	2	learning	learning	NOUN
alkej-870	27	3	models	model	NOUN
alkej-870	27	4	consist	consist	VERB
alkej-870	27	5	of	of	ADP
alkej-870	27	6	multiple	multiple	ADJ
alkej-870	27	7	processing	processing	NOUN
alkej-870	27	8	layers	layer	NOUN
alkej-870	27	9	that	that	PRON
alkej-870	27	10	learn	learn	VERB
alkej-870	27	11	different	different	ADJ
alkej-870	27	12	representations	representation	NOUN
alkej-870	27	13	of	of	ADP
alkej-870	27	14	data	datum	NOUN
alkej-870	27	15	to	to	PART
alkej-870	27	16	solve	solve	VERB
alkej-870	27	17	various	various	ADJ
alkej-870	27	18	problems	problem	NOUN
alkej-870	27	19	.	.	PUNCT
alkej-870	28	1	the	the	DET
alkej-870	28	2	field	field	NOUN
alkej-870	28	3	of	of	ADP
alkej-870	28	4	audio	audio	ADJ
alkej-870	28	5	-	-	ADJ
alkej-870	28	6	visual	visual	ADJ
alkej-870	28	7	source	source	NOUN
alkej-870	28	8	separation	separation	NOUN
alkej-870	28	9	witnessed	witness	VERB
alkej-870	28	10	the	the	DET
alkej-870	28	11	development	development	NOUN
alkej-870	28	12	of	of	ADP
alkej-870	28	13	various	various	ADJ
alkej-870	28	14	deep	deep	ADJ
alkej-870	28	15	learning	learning	NOUN
alkej-870	28	16	approaches	approach	NOUN
alkej-870	28	17	to	to	PART
alkej-870	28	18	help	help	VERB
alkej-870	28	19	solve	solve	VERB
alkej-870	28	20	the	the	DET
alkej-870	28	21	audio	audio	ADJ
alkej-870	28	22	separation	separation	NOUN
alkej-870	28	23	problem	problem	NOUN
alkej-870	28	24	in	in	ADP
alkej-870	28	25	the	the	DET
alkej-870	28	26	presence	presence	NOUN
alkej-870	28	27	of	of	ADP
alkej-870	28	28	visual	visual	ADJ
alkej-870	28	29	cues	cue	NOUN
alkej-870	28	30	.	.	PUNCT
alkej-870	29	1	each	each	DET
alkej-870	29	2	approach	approach	NOUN
alkej-870	29	3	has	have	VERB
alkej-870	29	4	its	its	PRON
alkej-870	29	5	advantages	advantage	NOUN
alkej-870	29	6	and	and	CCONJ
alkej-870	29	7	limitations	limitation	NOUN
alkej-870	29	8	,	,	PUNCT
alkej-870	29	9	depending	depend	VERB
alkej-870	29	10	on	on	ADP
alkej-870	29	11	the	the	DET
alkej-870	29	12	application	application	NOUN
alkej-870	29	13	and	and	CCONJ
alkej-870	29	14	data	datum	NOUN
alkej-870	29	15	availability	availability	NOUN
alkej-870	29	16	.	.	PUNCT
alkej-870	30	1	in	in	ADP
alkej-870	30	2	figure	figure	NOUN
alkej-870	30	3	1	1	NUM
alkej-870	30	4	,	,	PUNCT
alkej-870	30	5	deep	deep	ADJ
alkej-870	30	6	learning	learning	NOUN
alkej-870	30	7	methods	method	NOUN
alkej-870	30	8	that	that	PRON
alkej-870	30	9	are	be	AUX
alkej-870	30	10	commonly	commonly	ADV
alkej-870	30	11	used	use	VERB
alkej-870	30	12	for	for	ADP
alkej-870	30	13	avss	avss	NOUN
alkej-870	30	14	are	be	AUX
alkej-870	30	15	presented	present	VERB
alkej-870	30	16	and	and	CCONJ
alkej-870	30	17	explained	explain	VERB
alkej-870	30	18	as	as	SCONJ
alkej-870	30	19	follows	follow	VERB
alkej-870	30	20	:	:	PUNCT
alkej-870	30	21			NOUN
alkej-870	30	22	deep	deep	ADJ
alkej-870	30	23	neural	neural	ADJ
alkej-870	30	24	network	network	NOUN
alkej-870	30	25	(	(	PUNCT
alkej-870	30	26	dnn	dnn	PROPN
alkej-870	30	27	):	):	PUNCT
alkej-870	30	28	this	this	DET
alkej-870	30	29	deep	deep	ADJ
alkej-870	30	30	learning	learning	NOUN
alkej-870	30	31	architecture	architecture	NOUN
alkej-870	30	32	consists	consist	VERB
alkej-870	30	33	of	of	ADP
alkej-870	30	34	multiple	multiple	ADJ
alkej-870	30	35	hidden	hide	VERB
alkej-870	30	36	layers	layer	NOUN
alkej-870	30	37	for	for	ADP
alkej-870	30	38	data	data	NOUN
alkej-870	30	39	processing	processing	NOUN
alkej-870	30	40	.	.	PUNCT
alkej-870	31	1	it	it	PRON
alkej-870	31	2	is	be	AUX
alkej-870	31	3	widely	widely	ADV
alkej-870	31	4	used	use	VERB
alkej-870	31	5	for	for	ADP
alkej-870	31	6	image	image	NOUN
alkej-870	31	7	classification	classification	NOUN
alkej-870	31	8	,	,	PUNCT
alkej-870	31	9	speech	speech	NOUN
alkej-870	31	10	recognition	recognition	NOUN
alkej-870	31	11	,	,	PUNCT
alkej-870	31	12	or	or	CCONJ
alkej-870	31	13	natural	natural	ADJ
alkej-870	31	14	language	language	NOUN
alkej-870	31	15	processing	processing	NOUN
alkej-870	31	16	applications	application	NOUN
alkej-870	31	17	.	.	PUNCT
alkej-870	32	1			X
alkej-870	32	2	convolutional	convolutional	ADJ
alkej-870	32	3	neural	neural	ADJ
alkej-870	32	4	network	network	NOUN
alkej-870	32	5	(	(	PUNCT
alkej-870	32	6	cnn	cnn	PROPN
alkej-870	32	7	):	):	PUNCT
alkej-870	32	8	this	this	DET
alkej-870	32	9	deep	deep	ADJ
alkej-870	32	10	learning	learning	NOUN
alkej-870	32	11	architecture	architecture	NOUN
alkej-870	32	12	has	have	VERB
alkej-870	32	13	the	the	DET
alkej-870	32	14	ability	ability	NOUN
alkej-870	32	15	to	to	PART
alkej-870	32	16	detect	detect	VERB
alkej-870	32	17	patterns	pattern	NOUN
alkej-870	32	18	in	in	ADP
alkej-870	32	19	images	image	NOUN
alkej-870	32	20	.	.	PUNCT
alkej-870	33	1	it	it	PRON
alkej-870	33	2	is	be	AUX
alkej-870	33	3	widely	widely	ADV
alkej-870	33	4	used	use	VERB
alkej-870	33	5	for	for	ADP
alkej-870	33	6	image	image	NOUN
alkej-870	33	7	processing	processing	NOUN
alkej-870	33	8	and	and	CCONJ
alkej-870	33	9	object	object	NOUN
alkej-870	33	10	detection	detection	NOUN
alkej-870	33	11	.	.	PUNCT
alkej-870	34	1			X
alkej-870	35	1	long	long	ADJ
alkej-870	35	2	short	short	ADJ
alkej-870	35	3	term	term	NOUN
alkej-870	35	4	memory	memory	NOUN
alkej-870	35	5	(	(	PUNCT
alkej-870	35	6	lstm	lstm	NOUN
alkej-870	35	7	):	):	PUNCT
alkej-870	35	8	this	this	DET
alkej-870	35	9	deep	deep	ADJ
alkej-870	35	10	learning	learning	NOUN
alkej-870	35	11	architecture	architecture	NOUN
alkej-870	35	12	can	can	AUX
alkej-870	35	13	learn	learn	VERB
alkej-870	35	14	long	long	ADJ
alkej-870	35	15	-	-	PUNCT
alkej-870	35	16	term	term	NOUN
alkej-870	35	17	dependencies	dependency	NOUN
alkej-870	35	18	between	between	ADP
alkej-870	35	19	time	time	NOUN
alkej-870	35	20	-	-	PUNCT
alkej-870	35	21	sequenced	sequence	VERB
alkej-870	35	22	data	datum	NOUN
alkej-870	35	23	.	.	PUNCT
alkej-870	36	1	it	it	PRON
alkej-870	36	2	is	be	AUX
alkej-870	36	3	widely	widely	ADV
alkej-870	36	4	used	use	VERB
alkej-870	36	5	for	for	ADP
alkej-870	36	6	speech	speech	NOUN
alkej-870	36	7	recognition	recognition	NOUN
alkej-870	36	8	and	and	CCONJ
alkej-870	36	9	data	datum	NOUN
alkej-870	36	10	analysis	analysis	NOUN
alkej-870	36	11	.	.	PUNCT
alkej-870	37	1			X
alkej-870	37	2	bidirectional	bidirectional	ADJ
alkej-870	37	3	long	long	ADJ
alkej-870	37	4	short	short	ADJ
alkej-870	37	5	term	term	NOUN
alkej-870	37	6	memory	memory	NOUN
alkej-870	37	7	(	(	PUNCT
alkej-870	37	8	blstm	blstm	PROPN
alkej-870	37	9	):	):	PUNCT
alkej-870	37	10	this	this	DET
alkej-870	37	11	deep	deep	ADJ
alkej-870	37	12	learning	learning	NOUN
alkej-870	37	13	architecture	architecture	NOUN
alkej-870	37	14	can	can	AUX
alkej-870	37	15	learn	learn	VERB
alkej-870	37	16	long	long	ADJ
alkej-870	37	17	-	-	PUNCT
alkej-870	37	18	term	term	NOUN
alkej-870	37	19	dependencies	dependency	NOUN
alkej-870	37	20	between	between	ADP
alkej-870	37	21	time	time	NOUN
alkej-870	37	22	sequential	sequential	ADJ
alkej-870	37	23	data	datum	NOUN
alkej-870	37	24	in	in	ADP
alkej-870	37	25	both	both	CCONJ
alkej-870	37	26	backward	backward	ADJ
alkej-870	37	27	and	and	CCONJ
alkej-870	37	28	forward	forward	ADJ
alkej-870	37	29	directions	direction	NOUN
alkej-870	37	30	,	,	PUNCT
alkej-870	37	31	making	make	VERB
alkej-870	37	32	it	it	PRON
alkej-870	37	33	more	more	ADV
alkej-870	37	34	useful	useful	ADJ
alkej-870	37	35	in	in	ADP
alkej-870	37	36	sequential	sequential	ADJ
alkej-870	37	37	pattern	pattern	NOUN
alkej-870	37	38	analysis	analysis	NOUN
alkej-870	37	39	.	.	PUNCT
alkej-870	38	1	deep	deep	ADJ
alkej-870	38	2	learning	learning	NOUN
alkej-870	38	3	approaches	approach	NOUN
alkej-870	38	4	are	be	AUX
alkej-870	38	5	used	use	VERB
alkej-870	38	6	to	to	PART
alkej-870	38	7	extract	extract	VERB
alkej-870	38	8	audio	audio	ADJ
alkej-870	38	9	and	and	CCONJ
alkej-870	38	10	visual	visual	ADJ
alkej-870	38	11	features	feature	NOUN
alkej-870	38	12	and	and	CCONJ
alkej-870	38	13	map	map	VERB
alkej-870	38	14	them	they	PRON
alkej-870	38	15	to	to	ADP
alkej-870	38	16	different	different	ADJ
alkej-870	38	17	output	output	NOUN
alkej-870	38	18	signals	signal	NOUN
alkej-870	38	19	.	.	PUNCT
alkej-870	39	1	one	one	NUM
alkej-870	39	2	approach	approach	NOUN
alkej-870	39	3	is	be	AUX
alkej-870	39	4	joint	joint	ADJ
alkej-870	39	5	audio	audio	ADJ
alkej-870	39	6	-	-	ADJ
alkej-870	39	7	visual	visual	ADJ
alkej-870	39	8	processing	processing	NOUN
alkej-870	39	9	,	,	PUNCT
alkej-870	39	10	which	which	PRON
alkej-870	39	11	processes	process	VERB
alkej-870	39	12	audio	audio	ADJ
alkej-870	39	13	and	and	CCONJ
alkej-870	39	14	visual	visual	ADJ
alkej-870	39	15	models	model	NOUN
alkej-870	39	16	jointly	jointly	ADV
alkej-870	39	17	and	and	CCONJ
alkej-870	39	18	uses	use	VERB
alkej-870	39	19	the	the	DET
alkej-870	39	20	output	output	NOUN
alkej-870	39	21	to	to	PART
alkej-870	39	22	separate	separate	VERB
alkej-870	39	23	the	the	DET
alkej-870	39	24	target	target	NOUN
alkej-870	39	25	audio	audio	NOUN
alkej-870	39	26	source	source	NOUN
alkej-870	39	27	,	,	PUNCT
alkej-870	39	28	such	such	ADJ
alkej-870	39	29	as	as	ADP
alkej-870	39	30	in	in	ADP
alkej-870	39	31	[	[	X
alkej-870	39	32	16	16	NUM
alkej-870	39	33	,	,	PUNCT
alkej-870	39	34	19	19	NUM
alkej-870	39	35	,	,	PUNCT
alkej-870	39	36	40	40	NUM
alkej-870	39	37	]	]	PUNCT
alkej-870	39	38	.	.	PUNCT
alkej-870	40	1	examples	example	NOUN
alkej-870	40	2	of	of	ADP
alkej-870	40	3	joint	joint	ADJ
alkej-870	40	4	audio	audio	ADJ
alkej-870	40	5	-	-	ADJ
alkej-870	40	6	visual	visual	ADJ
alkej-870	40	7	processing	processing	NOUN
alkej-870	40	8	approaches	approach	NOUN
alkej-870	40	9	include	include	VERB
alkej-870	40	10	cnns	cnn	NOUN
alkej-870	40	11	and	and	CCONJ
alkej-870	40	12	recurrent	recurrent	ADJ
alkej-870	40	13	neural	neural	ADJ
alkej-870	40	14	networks	network	NOUN
alkej-870	40	15	(	(	PUNCT
alkej-870	40	16	rnns	rnns	PROPN
alkej-870	40	17	)	)	PUNCT
alkej-870	40	18	that	that	PRON
alkej-870	40	19	input	input	VERB
alkej-870	40	20	audio	audio	ADJ
alkej-870	40	21	and	and	CCONJ
alkej-870	40	22	visual	visual	ADJ
alkej-870	40	23	data	datum	NOUN
alkej-870	41	1	[	[	X
alkej-870	41	2	16	16	NUM
alkej-870	41	3	]	]	PUNCT
alkej-870	41	4	.	.	PUNCT
alkej-870	42	1	another	another	DET
alkej-870	42	2	approach	approach	NOUN
alkej-870	42	3	is	be	AUX
alkej-870	42	4	the	the	DET
alkej-870	42	5	multi	multi	ADJ
alkej-870	42	6	-	-	ADJ
alkej-870	42	7	task	task	ADJ
alkej-870	42	8	learning	learning	NOUN
alkej-870	42	9	approach	approach	NOUN
alkej-870	42	10	,	,	PUNCT
alkej-870	42	11	where	where	SCONJ
alkej-870	42	12	audio	audio	ADJ
alkej-870	42	13	and	and	CCONJ
alkej-870	42	14	visual	visual	ADJ
alkej-870	42	15	modalities	modality	NOUN
alkej-870	42	16	are	be	AUX
alkej-870	42	17	treated	treat	VERB
alkej-870	42	18	as	as	ADP
alkej-870	42	19	separate	separate	ADJ
alkej-870	42	20	tasks	task	NOUN
alkej-870	42	21	,	,	PUNCT
alkej-870	42	22	and	and	CCONJ
alkej-870	42	23	a	a	DET
alkej-870	42	24	deep	deep	ADJ
alkej-870	42	25	neural	neural	ADJ
alkej-870	42	26	network	network	NOUN
alkej-870	42	27	is	be	AUX
alkej-870	42	28	trained	train	VERB
alkej-870	42	29	to	to	PART
alkej-870	42	30	perform	perform	VERB
alkej-870	42	31	both	both	DET
alkej-870	42	32	tasks	task	NOUN
alkej-870	42	33	simultaneously	simultaneously	ADV
alkej-870	42	34	.	.	PUNCT
alkej-870	43	1	fig	fig	NOUN
alkej-870	43	2	.	.	PUNCT
alkej-870	44	1	1	1	X
alkej-870	44	2	.	.	X
alkej-870	44	3	deep	deep	ADJ
alkej-870	44	4	-	-	PUNCT
alkej-870	44	5	learning	learn	VERB
alkej-870	44	6	methods	method	NOUN
alkej-870	44	7	for	for	ADP
alkej-870	44	8	audio	audio	ADJ
alkej-870	44	9	visual	visual	ADJ
alkej-870	44	10	source	source	NOUN
alkej-870	44	11	separation	separation	NOUN
alkej-870	44	12	.	.	PUNCT
alkej-870	45	1	nooralhuda	nooralhuda	PROPN
alkej-870	45	2	mudhafar	mudhafar	PROPN
alkej-870	45	3	sleman	sleman	PROPN
alkej-870	45	4	al	al	PROPN
alkej-870	45	5	-	-	PUNCT
alkej-870	45	6	khwarizmi	khwarizmi	PROPN
alkej-870	45	7	engineering	engineering	NOUN
alkej-870	45	8	journal	journal	PROPN
alkej-870	45	9	,	,	PUNCT
alkej-870	45	10	vol	vol	NOUN
alkej-870	45	11	.	.	PROPN
alkej-870	45	12	19	19	NUM
alkej-870	45	13	,	,	PUNCT
alkej-870	45	14	no	no	INTJ
alkej-870	45	15	.	.	NOUN
alkej-870	45	16	4	4	NUM
alkej-870	45	17	,	,	PUNCT
alkej-870	45	18	p.p	p.p	PROPN
alkej-870	45	19	.	.	NOUN
alkej-870	45	20	4255	4255	NUM
alkej-870	45	21	(	(	PUNCT
alkej-870	45	22	2023	2023	NUM
alkej-870	45	23	)	)	PUNCT
alkej-870	45	24	44	44	NUM
alkej-870	45	25	the	the	DET
alkej-870	45	26	pros	pro	NOUN
alkej-870	45	27	and	and	CCONJ
alkej-870	45	28	cons	con	NOUN
alkej-870	45	29	of	of	ADP
alkej-870	45	30	the	the	DET
alkej-870	45	31	deep	deep	ADJ
alkej-870	45	32	learning	learning	NOUN
alkej-870	45	33	approach	approach	NOUN
alkej-870	45	34	are	be	AUX
alkej-870	45	35	listed	list	VERB
alkej-870	45	36	in	in	ADP
alkej-870	45	37	table	table	NOUN
alkej-870	45	38	1	1	NUM
alkej-870	45	39	.	.	PUNCT
alkej-870	46	1	it	it	PRON
alkej-870	46	2	can	can	AUX
alkej-870	46	3	be	be	AUX
alkej-870	46	4	observed	observe	VERB
alkej-870	46	5	from	from	ADP
alkej-870	46	6	table	table	NOUN
alkej-870	46	7	1	1	NUM
alkej-870	46	8	that	that	SCONJ
alkej-870	46	9	the	the	DET
alkej-870	46	10	pros	pro	NOUN
alkej-870	46	11	and	and	CCONJ
alkej-870	46	12	cons	con	NOUN
alkej-870	46	13	reflect	reflect	VERB
alkej-870	46	14	common	common	ADJ
alkej-870	46	15	observations	observation	NOUN
alkej-870	46	16	and	and	CCONJ
alkej-870	46	17	considerations	consideration	NOUN
alkej-870	46	18	in	in	ADP
alkej-870	46	19	the	the	DET
alkej-870	46	20	field	field	NOUN
alkej-870	46	21	.	.	PUNCT
alkej-870	47	1	however	however	ADV
alkej-870	47	2	,	,	PUNCT
alkej-870	47	3	it	it	PRON
alkej-870	47	4	's	be	AUX
alkej-870	47	5	important	important	ADJ
alkej-870	47	6	to	to	PART
alkej-870	47	7	note	note	VERB
alkej-870	47	8	that	that	SCONJ
alkej-870	47	9	variations	variation	NOUN
alkej-870	47	10	and	and	CCONJ
alkej-870	47	11	specific	specific	ADJ
alkej-870	47	12	trade	trade	NOUN
alkej-870	47	13	-	-	PUNCT
alkej-870	47	14	offs	off	NOUN
alkej-870	47	15	can	can	AUX
alkej-870	47	16	vary	vary	VERB
alkej-870	47	17	depending	depend	VERB
alkej-870	47	18	on	on	ADP
alkej-870	47	19	the	the	DET
alkej-870	47	20	deep	deep	ADJ
alkej-870	47	21	learning	learning	NOUN
alkej-870	47	22	architectures	architecture	NOUN
alkej-870	47	23	,	,	PUNCT
alkej-870	47	24	datasets	dataset	NOUN
alkej-870	47	25	,	,	PUNCT
alkej-870	47	26	and	and	CCONJ
alkej-870	47	27	problem	problem	NOUN
alkej-870	47	28	settings	setting	NOUN
alkej-870	47	29	used	use	VERB
alkej-870	47	30	in	in	ADP
alkej-870	47	31	different	different	ADJ
alkej-870	47	32	studies	study	NOUN
alkej-870	47	33	.	.	PUNCT
alkej-870	48	1	table	table	NOUN
alkej-870	48	2	1	1	NUM
alkej-870	48	3	.	.	PUNCT
alkej-870	49	1	pros	pro	NOUN
alkej-870	49	2	and	and	CCONJ
alkej-870	49	3	cons	con	NOUN
alkej-870	49	4	of	of	ADP
alkej-870	49	5	deep	deep	ADJ
alkej-870	49	6	learning	learning	NOUN
alkej-870	49	7	approaches	approach	NOUN
alkej-870	49	8	pros	pro	NOUN
alkej-870	49	9	cons	con	NOUN
alkej-870	49	10	can	can	AUX
alkej-870	49	11	handle	handle	VERB
alkej-870	49	12	complex	complex	ADJ
alkej-870	49	13	and	and	CCONJ
alkej-870	49	14	high	high	ADV
alkej-870	49	15	-	-	PUNCT
alkej-870	49	16	dimensional	dimensional	ADJ
alkej-870	49	17	data	datum	NOUN
alkej-870	49	18	requires	require	VERB
alkej-870	49	19	large	large	ADJ
alkej-870	49	20	amounts	amount	NOUN
alkej-870	49	21	of	of	ADP
alkej-870	49	22	data	datum	NOUN
alkej-870	49	23	for	for	ADP
alkej-870	49	24	training	training	NOUN
alkej-870	49	25	can	can	AUX
alkej-870	49	26	learn	learn	VERB
alkej-870	49	27	complex	complex	ADJ
alkej-870	49	28	relationships	relationship	NOUN
alkej-870	49	29	between	between	ADP
alkej-870	49	30	audio	audio	ADJ
alkej-870	49	31	and	and	CCONJ
alkej-870	49	32	visual	visual	ADJ
alkej-870	49	33	modalities	modality	NOUN
alkej-870	49	34	it	it	PRON
alkej-870	49	35	can	can	AUX
alkej-870	49	36	be	be	AUX
alkej-870	49	37	computationally	computationally	ADV
alkej-870	49	38	expensive	expensive	ADJ
alkej-870	49	39	and	and	CCONJ
alkej-870	49	40	require	require	VERB
alkej-870	49	41	powerful	powerful	ADJ
alkej-870	49	42	hardware	hardware	NOUN
alkej-870	49	43	can	can	AUX
alkej-870	49	44	generalize	generalize	VERB
alkej-870	49	45	well	well	ADV
alkej-870	49	46	to	to	ADP
alkej-870	49	47	new	new	ADJ
alkej-870	49	48	data	datum	NOUN
alkej-870	49	49	can	can	AUX
alkej-870	49	50	suffer	suffer	VERB
alkej-870	49	51	from	from	ADP
alkej-870	49	52	overfitting	overfitte	VERB
alkej-870	49	53	to	to	ADP
alkej-870	49	54	training	training	NOUN
alkej-870	49	55	data	datum	NOUN
alkej-870	49	56	can	can	AUX
alkej-870	49	57	capture	capture	VERB
alkej-870	49	58	both	both	PRON
alkej-870	49	59	lowlevel	lowlevel	VERB
alkej-870	49	60	and	and	CCONJ
alkej-870	49	61	high	high	ADJ
alkej-870	49	62	-	-	PUNCT
alkej-870	49	63	level	level	NOUN
alkej-870	49	64	features	feature	NOUN
alkej-870	49	65	it	it	PRON
alkej-870	49	66	may	may	AUX
alkej-870	49	67	require	require	VERB
alkej-870	49	68	significant	significant	ADJ
alkej-870	49	69	effort	effort	NOUN
alkej-870	49	70	for	for	ADP
alkej-870	49	71	hyperparameter	hyperparameter	NOUN
alkej-870	49	72	tuning	tuning	NOUN
alkej-870	49	73	and	and	CCONJ
alkej-870	49	74	model	model	NOUN
alkej-870	49	75	selection	selection	NOUN
alkej-870	49	76	3	3	NUM
alkej-870	49	77	.	.	PUNCT
alkej-870	50	1	recent	recent	ADJ
alkej-870	50	2	avss	avss	NOUN
alkej-870	50	3	works	work	VERB
alkej-870	50	4	different	different	ADJ
alkej-870	50	5	systems	system	NOUN
alkej-870	50	6	for	for	ADP
alkej-870	50	7	distinguishing	distinguish	VERB
alkej-870	50	8	audio	audio	ADJ
alkej-870	50	9	-	-	ADJ
alkej-870	50	10	visual	visual	ADJ
alkej-870	50	11	inputs	input	NOUN
alkej-870	50	12	have	have	AUX
alkej-870	50	13	been	be	AUX
alkej-870	50	14	developed	develop	VERB
alkej-870	50	15	as	as	ADP
alkej-870	50	16	a	a	DET
alkej-870	50	17	result	result	NOUN
alkej-870	50	18	of	of	ADP
alkej-870	50	19	advancements	advancement	NOUN
alkej-870	50	20	in	in	ADP
alkej-870	50	21	deep	deep	ADJ
alkej-870	50	22	learning	learning	NOUN
alkej-870	50	23	.	.	PUNCT
alkej-870	51	1	these	these	DET
alkej-870	51	2	systems	system	NOUN
alkej-870	51	3	employ	employ	VERB
alkej-870	51	4	a	a	DET
alkej-870	51	5	blend	blend	NOUN
alkej-870	51	6	of	of	ADP
alkej-870	51	7	deep	deep	ADJ
alkej-870	51	8	learning	learning	NOUN
alkej-870	51	9	techniques	technique	NOUN
alkej-870	51	10	,	,	PUNCT
alkej-870	51	11	including	include	VERB
alkej-870	51	12	dnns	dnn	NOUN
alkej-870	51	13	,	,	PUNCT
alkej-870	51	14	cnns	cnn	NOUN
alkej-870	51	15	,	,	PUNCT
alkej-870	51	16	lstms	lstms	ADJ
alkej-870	51	17	,	,	PUNCT
alkej-870	51	18	and	and	CCONJ
alkej-870	51	19	blstms	blstms	NOUN
alkej-870	51	20	.	.	PUNCT
alkej-870	52	1	the	the	DET
alkej-870	52	2	next	next	ADJ
alkej-870	52	3	part	part	NOUN
alkej-870	52	4	of	of	ADP
alkej-870	52	5	this	this	DET
alkej-870	52	6	section	section	NOUN
alkej-870	52	7	is	be	AUX
alkej-870	52	8	devoted	devote	VERB
alkej-870	52	9	to	to	ADP
alkej-870	52	10	categorizing	categorize	VERB
alkej-870	52	11	each	each	PRON
alkej-870	52	12	of	of	ADP
alkej-870	52	13	these	these	DET
alkej-870	52	14	works	work	NOUN
alkej-870	52	15	based	base	VERB
alkej-870	52	16	on	on	ADP
alkej-870	52	17	the	the	DET
alkej-870	52	18	specific	specific	ADJ
alkej-870	52	19	deep	deep	ADJ
alkej-870	52	20	learning	learning	NOUN
alkej-870	52	21	technique	technique	NOUN
alkej-870	52	22	that	that	PRON
alkej-870	52	23	they	they	PRON
alkej-870	52	24	use	use	VERB
alkej-870	52	25	.	.	PUNCT
alkej-870	53	1	an	an	DET
alkej-870	53	2	organized	organize	VERB
alkej-870	53	3	method	method	NOUN
alkej-870	53	4	for	for	ADP
alkej-870	53	5	understanding	understanding	NOUN
alkej-870	53	6	and	and	CCONJ
alkej-870	53	7	comparing	compare	VERB
alkej-870	53	8	different	different	ADJ
alkej-870	53	9	approaches	approach	NOUN
alkej-870	53	10	to	to	ADP
alkej-870	53	11	audio	audio	ADJ
alkej-870	53	12	-	-	ADJ
alkej-870	53	13	visual	visual	ADJ
alkej-870	53	14	source	source	NOUN
alkej-870	53	15	separation	separation	NOUN
alkej-870	53	16	is	be	AUX
alkej-870	53	17	provided	provide	VERB
alkej-870	53	18	by	by	ADP
alkej-870	53	19	this	this	DET
alkej-870	53	20	classification	classification	NOUN
alkej-870	53	21	.	.	PUNCT
alkej-870	54	1	it	it	PRON
alkej-870	54	2	also	also	ADV
alkej-870	54	3	helps	help	VERB
alkej-870	54	4	evaluate	evaluate	VERB
alkej-870	54	5	how	how	SCONJ
alkej-870	54	6	certain	certain	ADJ
alkej-870	54	7	topologies	topology	NOUN
alkej-870	54	8	affect	affect	VERB
alkej-870	54	9	separation	separation	NOUN
alkej-870	54	10	approaches	approach	NOUN
alkej-870	54	11	,	,	PUNCT
alkej-870	54	12	which	which	PRON
alkej-870	54	13	helps	help	VERB
alkej-870	54	14	to	to	PART
alkej-870	54	15	clarify	clarify	VERB
alkej-870	54	16	the	the	DET
alkej-870	54	17	advantages	advantage	NOUN
alkej-870	54	18	and	and	CCONJ
alkej-870	54	19	disadvantages	disadvantage	NOUN
alkej-870	54	20	of	of	ADP
alkej-870	54	21	each	each	PRON
alkej-870	54	22	.	.	PUNCT
alkej-870	55	1	the	the	DET
alkej-870	55	2	dnn	dnn	PROPN
alkej-870	55	3	-	-	PUNCT
alkej-870	55	4	based	base	VERB
alkej-870	55	5	works	work	NOUN
alkej-870	55	6	will	will	AUX
alkej-870	55	7	be	be	AUX
alkej-870	55	8	covered	cover	VERB
alkej-870	55	9	in	in	ADP
alkej-870	55	10	the	the	DET
alkej-870	55	11	subsection	subsection	NOUN
alkej-870	55	12	that	that	PRON
alkej-870	55	13	follows	follow	VERB
alkej-870	55	14	.	.	PUNCT
alkej-870	56	1	takahashi	takahashi	PROPN
alkej-870	56	2	et	et	PROPN
alkej-870	56	3	al	al	PROPN
alkej-870	57	1	[	[	X
alkej-870	57	2	10	10	NUM
alkej-870	57	3	]	]	PUNCT
alkej-870	57	4	suggested	suggest	VERB
alkej-870	57	5	a	a	DET
alkej-870	57	6	unique	unique	ADJ
alkej-870	57	7	method	method	NOUN
alkej-870	57	8	that	that	PRON
alkej-870	57	9	included	include	VERB
alkej-870	57	10	end	end	NOUN
alkej-870	57	11	-	-	PUNCT
alkej-870	57	12	to	to	ADP
alkej-870	57	13	-	-	PUNCT
alkej-870	57	14	end	end	NOUN
alkej-870	57	15	speech	speech	NOUN
alkej-870	57	16	recognition	recognition	NOUN
alkej-870	57	17	in	in	ADP
alkej-870	57	18	order	order	NOUN
alkej-870	57	19	to	to	PART
alkej-870	57	20	enhance	enhance	VERB
alkej-870	57	21	voice	voice	NOUN
alkej-870	57	22	separation	separation	NOUN
alkej-870	57	23	performance	performance	NOUN
alkej-870	57	24	.	.	PUNCT
alkej-870	58	1	a	a	DET
alkej-870	58	2	dnn	dnn	PROPN
alkej-870	58	3	trained	train	VERB
alkej-870	58	4	on	on	ADP
alkej-870	58	5	mixed	mixed	ADJ
alkej-870	58	6	audio	audio	ADJ
alkej-870	58	7	signals	signal	NOUN
alkej-870	58	8	and	and	CCONJ
alkej-870	58	9	their	their	PRON
alkej-870	58	10	transcriptions	transcription	NOUN
alkej-870	58	11	is	be	AUX
alkej-870	58	12	used	use	VERB
alkej-870	58	13	in	in	ADP
alkej-870	58	14	the	the	DET
alkej-870	58	15	suggested	suggest	VERB
alkej-870	58	16	voice	voice	NOUN
alkej-870	58	17	separation	separation	NOUN
alkej-870	58	18	method	method	NOUN
alkej-870	58	19	,	,	PUNCT
alkej-870	58	20	which	which	PRON
alkej-870	58	21	incorporates	incorporate	VERB
alkej-870	58	22	end	end	NOUN
alkej-870	58	23	-	-	PUNCT
alkej-870	58	24	to	to	ADP
alkej-870	58	25	-	-	PUNCT
alkej-870	58	26	end	end	NOUN
alkej-870	58	27	speech	speech	NOUN
alkej-870	58	28	recognition	recognition	NOUN
alkej-870	58	29	to	to	PART
alkej-870	58	30	direct	direct	VERB
alkej-870	58	31	the	the	DET
alkej-870	58	32	separation	separation	NOUN
alkej-870	58	33	process	process	NOUN
alkej-870	58	34	.	.	PUNCT
alkej-870	59	1	the	the	DET
alkej-870	59	2	temporal	temporal	ADJ
alkej-870	59	3	dependencies	dependency	NOUN
alkej-870	59	4	are	be	AUX
alkej-870	59	5	modeled	model	VERB
alkej-870	59	6	and	and	CCONJ
alkej-870	59	7	the	the	DET
alkej-870	59	8	input	input	NOUN
alkej-870	59	9	signal	signal	NOUN
alkej-870	59	10	is	be	AUX
alkej-870	59	11	converted	convert	VERB
alkej-870	59	12	into	into	ADP
alkej-870	59	13	a	a	DET
alkej-870	59	14	feature	feature	NOUN
alkej-870	59	15	representation	representation	NOUN
alkej-870	59	16	using	use	VERB
alkej-870	59	17	deep	deep	ADJ
alkej-870	59	18	learning	learning	NOUN
alkej-870	59	19	layers	layer	NOUN
alkej-870	59	20	like	like	ADP
alkej-870	59	21	fully	fully	ADV
alkej-870	59	22	connected	connected	ADJ
alkej-870	59	23	and	and	CCONJ
alkej-870	59	24	lstm	lstm	ADJ
alkej-870	59	25	.	.	PUNCT
alkej-870	60	1	although	although	SCONJ
alkej-870	60	2	the	the	DET
alkej-870	60	3	technique	technique	NOUN
alkej-870	60	4	does	do	AUX
alkej-870	60	5	not	not	PART
alkej-870	60	6	necessitate	necessitate	VERB
alkej-870	60	7	the	the	DET
alkej-870	60	8	pre	pre	NOUN
alkej-870	60	9	-	-	NOUN
alkej-870	60	10	separation	separation	NOUN
alkej-870	60	11	of	of	ADP
alkej-870	60	12	audio	audio	ADJ
alkej-870	60	13	and	and	CCONJ
alkej-870	60	14	visual	visual	ADJ
alkej-870	60	15	elements	element	NOUN
alkej-870	60	16	,	,	PUNCT
alkej-870	60	17	its	its	PRON
alkej-870	60	18	effectiveness	effectiveness	NOUN
alkej-870	60	19	may	may	AUX
alkej-870	60	20	be	be	AUX
alkej-870	60	21	impacted	impact	VERB
alkej-870	60	22	by	by	ADP
alkej-870	60	23	failures	failure	NOUN
alkej-870	60	24	in	in	ADP
alkej-870	60	25	voice	voice	NOUN
alkej-870	60	26	recognition	recognition	NOUN
alkej-870	60	27	.	.	PUNCT
alkej-870	61	1	chung	chung	PROPN
alkej-870	61	2	et	et	PROPN
alkej-870	61	3	al	al	PROPN
alkej-870	61	4	.	.	PUNCT
alkej-870	62	1	[	[	X
alkej-870	62	2	11	11	NUM
alkej-870	62	3	]	]	PUNCT
alkej-870	62	4	proposed	propose	VERB
alkej-870	62	5	a	a	DET
alkej-870	62	6	facefilter	facefilter	NOUN
alkej-870	62	7	approach	approach	NOUN
alkej-870	62	8	that	that	PRON
alkej-870	62	9	enhances	enhance	VERB
alkej-870	62	10	speech	speech	NOUN
alkej-870	62	11	in	in	ADP
alkej-870	62	12	a	a	DET
alkej-870	62	13	noisy	noisy	ADJ
alkej-870	62	14	audio	audio	NOUN
alkej-870	62	15	recording	recording	NOUN
alkej-870	62	16	using	use	VERB
alkej-870	62	17	a	a	DET
alkej-870	62	18	single	single	ADJ
alkej-870	62	19	facial	facial	ADJ
alkej-870	62	20	image	image	NOUN
alkej-870	62	21	of	of	ADP
alkej-870	62	22	the	the	DET
alkej-870	62	23	speaker	speaker	NOUN
alkej-870	62	24	,	,	PUNCT
alkej-870	62	25	employing	employ	VERB
alkej-870	62	26	a	a	DET
alkej-870	62	27	dnn	dnn	NOUN
alkej-870	62	28	with	with	ADP
alkej-870	62	29	three	three	NUM
alkej-870	62	30	subnetworks	subnetwork	NOUN
alkej-870	62	31	that	that	PRON
alkej-870	62	32	extract	extract	VERB
alkej-870	62	33	and	and	CCONJ
alkej-870	62	34	combine	combine	VERB
alkej-870	62	35	visual	visual	ADJ
alkej-870	62	36	and	and	CCONJ
alkej-870	62	37	audio	audio	NOUN
alkej-870	62	38	features	feature	NOUN
alkej-870	62	39	to	to	PART
alkej-870	62	40	produce	produce	VERB
alkej-870	62	41	a	a	DET
alkej-870	62	42	mask	mask	NOUN
alkej-870	62	43	for	for	ADP
alkej-870	62	44	the	the	DET
alkej-870	62	45	target	target	NOUN
alkej-870	62	46	speaker	speaker	NOUN
alkej-870	62	47	's	's	PART
alkej-870	62	48	voice	voice	NOUN
alkej-870	62	49	.	.	PUNCT
alkej-870	63	1	the	the	DET
alkej-870	63	2	method	method	NOUN
alkej-870	63	3	is	be	AUX
alkej-870	63	4	advantageous	advantageous	ADJ
alkej-870	63	5	in	in	ADP
alkej-870	63	6	scenarios	scenario	NOUN
alkej-870	63	7	with	with	ADP
alkej-870	63	8	limited	limited	ADJ
alkej-870	63	9	visual	visual	ADJ
alkej-870	63	10	data	datum	NOUN
alkej-870	63	11	and	and	CCONJ
alkej-870	63	12	faster	fast	ADJ
alkej-870	63	13	processing	processing	NOUN
alkej-870	63	14	times	time	NOUN
alkej-870	63	15	,	,	PUNCT
alkej-870	63	16	but	but	CCONJ
alkej-870	63	17	may	may	AUX
alkej-870	63	18	be	be	AUX
alkej-870	63	19	limited	limit	VERB
alkej-870	63	20	by	by	ADP
alkej-870	63	21	assumptions	assumption	NOUN
alkej-870	63	22	of	of	ADP
alkej-870	63	23	visible	visible	ADJ
alkej-870	63	24	and	and	CCONJ
alkej-870	63	25	stationary	stationary	ADJ
alkej-870	63	26	faces	face	NOUN
alkej-870	63	27	,	,	PUNCT
alkej-870	63	28	leading	lead	VERB
alkej-870	63	29	to	to	ADP
alkej-870	63	30	negative	negative	ADJ
alkej-870	63	31	impacts	impact	NOUN
alkej-870	63	32	on	on	ADP
alkej-870	63	33	model	model	NOUN
alkej-870	63	34	performance	performance	NOUN
alkej-870	63	35	in	in	ADP
alkej-870	63	36	real	real	ADJ
alkej-870	63	37	-	-	PUNCT
alkej-870	63	38	world	world	NOUN
alkej-870	63	39	scenarios	scenario	NOUN
alkej-870	63	40	,	,	PUNCT
alkej-870	63	41	and	and	CCONJ
alkej-870	63	42	may	may	AUX
alkej-870	63	43	not	not	PART
alkej-870	63	44	be	be	AUX
alkej-870	63	45	effective	effective	ADJ
alkej-870	63	46	in	in	ADP
alkej-870	63	47	partially	partially	ADV
alkej-870	63	48	obscured	obscure	VERB
alkej-870	63	49	or	or	CCONJ
alkej-870	63	50	poor	poor	ADJ
alkej-870	63	51	lighting	lighting	NOUN
alkej-870	63	52	conditions	condition	NOUN
alkej-870	63	53	.	.	PUNCT
alkej-870	64	1	gu	gu	NOUN
alkej-870	64	2	et	et	PROPN
alkej-870	64	3	al	al	PROPN
alkej-870	64	4	.	.	PUNCT
alkej-870	65	1	[	[	X
alkej-870	65	2	12	12	NUM
alkej-870	65	3	]	]	PUNCT
alkej-870	65	4	proposed	propose	VERB
alkej-870	65	5	a	a	DET
alkej-870	65	6	deep	deep	ADJ
alkej-870	65	7	learning	learning	NOUN
alkej-870	65	8	-	-	PUNCT
alkej-870	65	9	based	base	VERB
alkej-870	65	10	approach	approach	NOUN
alkej-870	65	11	for	for	ADP
alkej-870	65	12	target	target	NOUN
alkej-870	65	13	speech	speech	NOUN
alkej-870	65	14	separation	separation	NOUN
alkej-870	65	15	from	from	ADP
alkej-870	65	16	multimodal	multimodal	NOUN
alkej-870	65	17	and	and	CCONJ
alkej-870	65	18	multi	multi	ADJ
alkej-870	65	19	-	-	ADJ
alkej-870	65	20	channel	channel	ADJ
alkej-870	65	21	mixtures	mixture	NOUN
alkej-870	65	22	of	of	ADP
alkej-870	65	23	speech	speech	NOUN
alkej-870	65	24	and	and	CCONJ
alkej-870	65	25	noise	noise	NOUN
alkej-870	65	26	signals	signal	NOUN
alkej-870	65	27	using	use	VERB
alkej-870	65	28	audio	audio	NOUN
alkej-870	65	29	,	,	PUNCT
alkej-870	65	30	visual	visual	ADJ
alkej-870	65	31	,	,	PUNCT
alkej-870	65	32	and	and	CCONJ
alkej-870	65	33	spatial	spatial	ADJ
alkej-870	65	34	information	information	NOUN
alkej-870	65	35	to	to	PART
alkej-870	65	36	improve	improve	VERB
alkej-870	65	37	accuracy	accuracy	NOUN
alkej-870	65	38	.	.	PUNCT
alkej-870	66	1	dnn	dnn	PROPN
alkej-870	66	2	combines	combine	VERB
alkej-870	66	3	audio	audio	ADJ
alkej-870	66	4	spectrogram	spectrogram	NOUN
alkej-870	66	5	,	,	PUNCT
alkej-870	66	6	visual	visual	ADJ
alkej-870	66	7	feature	feature	NOUN
alkej-870	66	8	map	map	NOUN
alkej-870	66	9	,	,	PUNCT
alkej-870	66	10	and	and	CCONJ
alkej-870	66	11	spatial	spatial	ADJ
alkej-870	66	12	inter	inter	ADJ
alkej-870	66	13	-	-	ADJ
alkej-870	66	14	channel	channel	ADJ
alkej-870	66	15	phase	phase	NOUN
alkej-870	66	16	differences	difference	NOUN
alkej-870	66	17	to	to	PART
alkej-870	66	18	predict	predict	VERB
alkej-870	66	19	the	the	DET
alkej-870	66	20	ideal	ideal	ADJ
alkej-870	66	21	ratio	ratio	NOUN
alkej-870	66	22	mask	mask	NOUN
alkej-870	66	23	for	for	ADP
alkej-870	66	24	speech	speech	NOUN
alkej-870	66	25	separation	separation	NOUN
alkej-870	66	26	,	,	PUNCT
alkej-870	66	27	trained	train	VERB
alkej-870	66	28	end	end	NOUN
alkej-870	66	29	-	-	PUNCT
alkej-870	66	30	toend	toend	NOUN
alkej-870	66	31	for	for	ADP
alkej-870	66	32	efficient	efficient	ADJ
alkej-870	66	33	separation	separation	NOUN
alkej-870	66	34	without	without	ADP
alkej-870	66	35	complex	complex	ADJ
alkej-870	66	36	preprocessing	preprocessing	NOUN
alkej-870	66	37	.	.	PUNCT
alkej-870	67	1	zhang	zhang	PROPN
alkej-870	67	2	et	et	PROPN
alkej-870	67	3	al	al	PROPN
alkej-870	67	4	.	.	PUNCT
alkej-870	68	1	[	[	X
alkej-870	68	2	13	13	NUM
alkej-870	68	3	]	]	PUNCT
alkej-870	68	4	proposed	propose	VERB
alkej-870	68	5	an	an	DET
alkej-870	68	6	approach	approach	NOUN
alkej-870	68	7	to	to	PART
alkej-870	68	8	improve	improve	VERB
alkej-870	68	9	the	the	DET
alkej-870	68	10	mvdr	mvdr	NOUN
alkej-870	68	11	beamformer	beamformer	NOUN
alkej-870	68	12	for	for	ADP
alkej-870	68	13	multichannel	multichannel	ADJ
alkej-870	68	14	target	target	NOUN
alkej-870	68	15	speech	speech	NOUN
alkej-870	68	16	separation	separation	NOUN
alkej-870	68	17	by	by	ADP
alkej-870	68	18	replacing	replace	VERB
alkej-870	68	19	the	the	DET
alkej-870	68	20	conventional	conventional	ADJ
alkej-870	68	21	covariance	covariance	NOUN
alkej-870	68	22	matrix	matrix	NOUN
alkej-870	68	23	estimation	estimation	NOUN
alkej-870	68	24	with	with	ADP
alkej-870	68	25	a	a	DET
alkej-870	68	26	dnn	dnn	PROPN
alkej-870	68	27	trained	train	VERB
alkej-870	68	28	on	on	ADP
alkej-870	68	29	a	a	DET
alkej-870	68	30	large	large	ADJ
alkej-870	68	31	-	-	PUNCT
alkej-870	68	32	scale	scale	NOUN
alkej-870	68	33	speech	speech	NOUN
alkej-870	68	34	corpus	corpus	NOUN
alkej-870	68	35	.	.	PUNCT
alkej-870	69	1	the	the	DET
alkej-870	69	2	approach	approach	NOUN
alkej-870	69	3	is	be	AUX
alkej-870	69	4	more	more	ADV
alkej-870	69	5	robust	robust	ADJ
alkej-870	69	6	for	for	ADP
alkej-870	69	7	non	non	ADJ
alkej-870	69	8	-	-	ADJ
alkej-870	69	9	stationary	stationary	ADJ
alkej-870	69	10	noise	noise	NOUN
alkej-870	69	11	and	and	CCONJ
alkej-870	69	12	acoustic	acoustic	ADJ
alkej-870	69	13	environments	environment	NOUN
alkej-870	69	14	.	.	PUNCT
alkej-870	70	1	however	however	ADV
alkej-870	70	2	,	,	PUNCT
alkej-870	70	3	the	the	DET
alkej-870	70	4	computational	computational	ADJ
alkej-870	70	5	cost	cost	NOUN
alkej-870	70	6	of	of	ADP
alkej-870	70	7	training	train	VERB
alkej-870	70	8	the	the	DET
alkej-870	70	9	dnn	dnn	PROPN
alkej-870	70	10	and	and	CCONJ
alkej-870	70	11	computing	compute	VERB
alkej-870	70	12	the	the	DET
alkej-870	70	13	mvdr	mvdr	NOUN
alkej-870	70	14	beamformer	beamformer	NOUN
alkej-870	70	15	weights	weight	NOUN
alkej-870	70	16	can	can	AUX
alkej-870	70	17	be	be	AUX
alkej-870	70	18	high	high	ADJ
alkej-870	70	19	.	.	PUNCT
alkej-870	71	1	li	li	PROPN
alkej-870	71	2	et	et	PROPN
alkej-870	71	3	al	al	PROPN
alkej-870	71	4	.	.	PUNCT
alkej-870	72	1	[	[	X
alkej-870	72	2	14	14	NUM
alkej-870	72	3	]	]	PUNCT
alkej-870	72	4	proposed	propose	VERB
alkej-870	72	5	an	an	DET
alkej-870	72	6	av	av	PROPN
alkej-870	72	7	deep	deep	ADJ
alkej-870	72	8	learning	learning	NOUN
alkej-870	72	9	approach	approach	NOUN
alkej-870	72	10	for	for	ADP
alkej-870	72	11	multi	multi	ADJ
alkej-870	72	12	-	-	ADJ
alkej-870	72	13	channel	channel	NOUN
alkej-870	72	14	speech	speech	NOUN
alkej-870	72	15	separation	separation	NOUN
alkej-870	72	16	by	by	ADP
alkej-870	72	17	jointly	jointly	ADV
alkej-870	72	18	modeling	model	VERB
alkej-870	72	19	audio	audio	ADJ
alkej-870	72	20	-	-	ADJ
alkej-870	72	21	visual	visual	ADJ
alkej-870	72	22	cues	cue	NOUN
alkej-870	72	23	.	.	PUNCT
alkej-870	73	1	it	it	PRON
alkej-870	73	2	includes	include	VERB
alkej-870	73	3	a	a	DET
alkej-870	73	4	neural	neural	ADJ
alkej-870	73	5	network	network	NOUN
alkej-870	73	6	that	that	PRON
alkej-870	73	7	estimates	estimate	VERB
alkej-870	73	8	separation	separation	NOUN
alkej-870	73	9	filters	filter	VERB
alkej-870	73	10	for	for	ADP
alkej-870	73	11	target	target	NOUN
alkej-870	73	12	speech	speech	NOUN
alkej-870	73	13	from	from	ADP
alkej-870	73	14	multiple	multiple	ADJ
alkej-870	73	15	microphones	microphone	NOUN
alkej-870	73	16	and	and	CCONJ
alkej-870	73	17	video	video	NOUN
alkej-870	73	18	frames	frame	NOUN
alkej-870	73	19	,	,	PUNCT
alkej-870	73	20	and	and	CCONJ
alkej-870	73	21	a	a	DET
alkej-870	73	22	multi	multi	ADJ
alkej-870	73	23	-	-	ADJ
alkej-870	73	24	task	task	ADJ
alkej-870	73	25	framework	framework	NOUN
alkej-870	73	26	for	for	ADP
alkej-870	73	27	dereverberation	dereverberation	NOUN
alkej-870	73	28	and	and	CCONJ
alkej-870	73	29	speech	speech	NOUN
alkej-870	73	30	recognition	recognition	NOUN
alkej-870	73	31	.	.	PUNCT
alkej-870	74	1	advantage	advantage	NOUN
alkej-870	74	2	:	:	PUNCT
alkej-870	74	3	effective	effective	ADJ
alkej-870	74	4	in	in	ADP
alkej-870	74	5	removing	remove	VERB
alkej-870	74	6	noise	noise	NOUN
alkej-870	74	7	and	and	CCONJ
alkej-870	74	8	reverberation	reverberation	NOUN
alkej-870	74	9	.	.	PUNCT
alkej-870	75	1	disadvantages	disadvantage	NOUN
alkej-870	75	2	:	:	PUNCT
alkej-870	75	3	requires	require	VERB
alkej-870	75	4	multichannel	multichannel	ADJ
alkej-870	75	5	audio	audio	ADJ
alkej-870	75	6	signals	signal	NOUN
alkej-870	75	7	,	,	PUNCT
alkej-870	75	8	large	large	ADJ
alkej-870	75	9	amounts	amount	NOUN
alkej-870	75	10	of	of	ADP
alkej-870	75	11	training	training	NOUN
alkej-870	75	12	nooralhuda	nooralhuda	NOUN
alkej-870	75	13	mudhafar	mudhafar	PROPN
alkej-870	75	14	sleman	sleman	PROPN
alkej-870	75	15	al	al	PROPN
alkej-870	75	16	-	-	PUNCT
alkej-870	75	17	khwarizmi	khwarizmi	PROPN
alkej-870	75	18	engineering	engineering	NOUN
alkej-870	75	19	journal	journal	PROPN
alkej-870	75	20	,	,	PUNCT
alkej-870	75	21	vol	vol	NOUN
alkej-870	75	22	.	.	PROPN
alkej-870	76	1	19	19	NUM
alkej-870	76	2	,	,	PUNCT
alkej-870	76	3	no	no	INTJ
alkej-870	76	4	.	.	NOUN
alkej-870	76	5	4	4	NUM
alkej-870	76	6	,	,	PUNCT
alkej-870	76	7	p.p	p.p	PROPN
alkej-870	76	8	.	.	NOUN
alkej-870	76	9	4255	4255	NUM
alkej-870	76	10	(	(	PUNCT
alkej-870	76	11	2023	2023	NUM
alkej-870	76	12	)	)	PUNCT
alkej-870	76	13	45	45	NUM
alkej-870	76	14	data	datum	NOUN
alkej-870	76	15	,	,	PUNCT
alkej-870	76	16	and	and	CCONJ
alkej-870	76	17	computational	computational	ADJ
alkej-870	76	18	resources	resource	NOUN
alkej-870	76	19	for	for	ADP
alkej-870	76	20	deep	deep	ADJ
alkej-870	76	21	neural	neural	ADJ
alkej-870	76	22	networks	network	NOUN
alkej-870	76	23	.	.	PUNCT
alkej-870	77	1	ong	ong	INTJ
alkej-870	77	2	et	et	PROPN
alkej-870	77	3	al	al	PROPN
alkej-870	77	4	.	.	PUNCT
alkej-870	78	1	[	[	X
alkej-870	78	2	15	15	NUM
alkej-870	78	3	]	]	PUNCT
alkej-870	78	4	proposed	propose	VERB
alkej-870	78	5	a	a	DET
alkej-870	78	6	real	real	ADJ
alkej-870	78	7	-	-	PUNCT
alkej-870	78	8	time	time	NOUN
alkej-870	78	9	online	online	ADJ
alkej-870	78	10	multi	multi	ADJ
alkej-870	78	11	-	-	ADJ
alkej-870	78	12	source	source	NOUN
alkej-870	78	13	separation	separation	NOUN
alkej-870	78	14	method	method	NOUN
alkej-870	78	15	that	that	PRON
alkej-870	78	16	uses	use	VERB
alkej-870	78	17	both	both	DET
alkej-870	78	18	audio	audio	ADJ
alkej-870	78	19	and	and	CCONJ
alkej-870	78	20	visual	visual	ADJ
alkej-870	78	21	information	information	NOUN
alkej-870	78	22	to	to	PART
alkej-870	78	23	estimate	estimate	VERB
alkej-870	78	24	the	the	DET
alkej-870	78	25	position	position	NOUN
alkej-870	78	26	of	of	ADP
alkej-870	78	27	speakers	speaker	NOUN
alkej-870	78	28	and	and	CCONJ
alkej-870	78	29	separate	separate	ADJ
alkej-870	78	30	target	target	NOUN
alkej-870	78	31	speech	speech	NOUN
alkej-870	78	32	from	from	ADP
alkej-870	78	33	background	background	NOUN
alkej-870	78	34	noise	noise	NOUN
alkej-870	78	35	.	.	PUNCT
alkej-870	79	1	they	they	PRON
alkej-870	79	2	developed	develop	VERB
alkej-870	79	3	a	a	DET
alkej-870	79	4	deep	deep	ADJ
alkej-870	79	5	neural	neural	ADJ
alkej-870	79	6	network	network	NOUN
alkej-870	79	7	combining	combine	VERB
alkej-870	79	8	audio	audio	ADJ
alkej-870	79	9	and	and	CCONJ
alkej-870	79	10	visual	visual	ADJ
alkej-870	79	11	processing	processing	NOUN
alkej-870	79	12	features	feature	NOUN
alkej-870	79	13	.	.	PUNCT
alkej-870	80	1	this	this	DET
alkej-870	80	2	approach	approach	NOUN
alkej-870	80	3	shows	show	VERB
alkej-870	80	4	improved	improved	ADJ
alkej-870	80	5	performance	performance	NOUN
alkej-870	80	6	,	,	PUNCT
alkej-870	80	7	especially	especially	ADV
alkej-870	80	8	in	in	ADP
alkej-870	80	9	noisy	noisy	ADJ
alkej-870	80	10	and	and	CCONJ
alkej-870	80	11	reverberant	reverberant	ADJ
alkej-870	80	12	environments	environment	NOUN
alkej-870	80	13	.	.	PUNCT
alkej-870	81	1	however	however	ADV
alkej-870	81	2	,	,	PUNCT
alkej-870	81	3	it	it	PRON
alkej-870	81	4	requires	require	VERB
alkej-870	81	5	multiple	multiple	ADJ
alkej-870	81	6	microphones	microphone	NOUN
alkej-870	81	7	and	and	CCONJ
alkej-870	81	8	cameras	camera	NOUN
alkej-870	81	9	to	to	PART
alkej-870	81	10	capture	capture	VERB
alkej-870	81	11	the	the	DET
alkej-870	81	12	audio	audio	ADJ
alkej-870	81	13	and	and	CCONJ
alkej-870	81	14	visual	visual	ADJ
alkej-870	81	15	data	datum	NOUN
alkej-870	81	16	,	,	PUNCT
alkej-870	81	17	and	and	CCONJ
alkej-870	81	18	the	the	DET
alkej-870	81	19	quality	quality	NOUN
alkej-870	81	20	of	of	ADP
alkej-870	81	21	the	the	DET
alkej-870	81	22	separated	separate	VERB
alkej-870	81	23	speech	speech	NOUN
alkej-870	81	24	may	may	AUX
alkej-870	81	25	vary	vary	VERB
alkej-870	81	26	depending	depend	VERB
alkej-870	81	27	on	on	ADP
alkej-870	81	28	the	the	DET
alkej-870	81	29	quality	quality	NOUN
alkej-870	81	30	of	of	ADP
alkej-870	81	31	the	the	DET
alkej-870	81	32	captured	capture	VERB
alkej-870	81	33	audio	audio	NOUN
alkej-870	81	34	and	and	CCONJ
alkej-870	81	35	video	video	NOUN
alkej-870	81	36	.	.	PUNCT
alkej-870	82	1	furthermore	furthermore	ADV
alkej-870	82	2	,	,	PUNCT
alkej-870	82	3	in	in	ADP
alkej-870	82	4	this	this	DET
alkej-870	82	5	subsection	subsection	NOUN
alkej-870	82	6	,	,	PUNCT
alkej-870	82	7	the	the	DET
alkej-870	82	8	cnn	cnn	PROPN
alkej-870	82	9	-	-	PUNCT
alkej-870	82	10	based	base	VERB
alkej-870	82	11	works	work	NOUN
alkej-870	82	12	will	will	AUX
alkej-870	82	13	be	be	AUX
alkej-870	82	14	discussed	discuss	VERB
alkej-870	82	15	.	.	PUNCT
alkej-870	83	1	afouras	afouras	PROPN
alkej-870	83	2	et	et	PROPN
alkej-870	83	3	al	al	PROPN
alkej-870	83	4	.	.	PUNCT
alkej-870	84	1	[	[	X
alkej-870	84	2	16	16	NUM
alkej-870	84	3	]	]	PUNCT
alkej-870	84	4	proposed	propose	VERB
alkej-870	84	5	a	a	DET
alkej-870	84	6	deep	deep	ADJ
alkej-870	84	7	learning	learning	NOUN
alkej-870	84	8	model	model	NOUN
alkej-870	84	9	for	for	ADP
alkej-870	84	10	audio	audio	ADJ
alkej-870	84	11	-	-	ADJ
alkej-870	84	12	visual	visual	ADJ
alkej-870	84	13	speech	speech	NOUN
alkej-870	84	14	enhancement	enhancement	NOUN
alkej-870	84	15	using	use	VERB
alkej-870	84	16	a	a	DET
alkej-870	84	17	two	two	NUM
alkej-870	84	18	-	-	PUNCT
alkej-870	84	19	stream	stream	NOUN
alkej-870	84	20	cnn	cnn	NOUN
alkej-870	84	21	architecture	architecture	NOUN
alkej-870	84	22	to	to	PART
alkej-870	84	23	predict	predict	VERB
alkej-870	84	24	a	a	DET
alkej-870	84	25	cleaned	clean	VERB
alkej-870	84	26	audio	audio	NOUN
alkej-870	84	27	signal	signal	NOUN
alkej-870	84	28	by	by	ADP
alkej-870	84	29	taking	take	VERB
alkej-870	84	30	audio	audio	ADJ
alkej-870	84	31	and	and	CCONJ
alkej-870	84	32	visual	visual	ADJ
alkej-870	84	33	inputs	input	NOUN
alkej-870	84	34	.	.	PUNCT
alkej-870	85	1	advantages	advantage	NOUN
alkej-870	85	2	include	include	VERB
alkej-870	85	3	using	use	VERB
alkej-870	85	4	both	both	DET
alkej-870	85	5	modalities	modality	NOUN
alkej-870	85	6	,	,	PUNCT
alkej-870	85	7	improving	improve	VERB
alkej-870	85	8	accuracy	accuracy	NOUN
alkej-870	85	9	and	and	CCONJ
alkej-870	85	10	outperforming	outperform	VERB
alkej-870	85	11	existing	exist	VERB
alkej-870	85	12	models	model	NOUN
alkej-870	85	13	.	.	PUNCT
alkej-870	86	1	disadvantages	disadvantage	NOUN
alkej-870	86	2	include	include	VERB
alkej-870	86	3	the	the	DET
alkej-870	86	4	requirement	requirement	NOUN
alkej-870	86	5	for	for	ADP
alkej-870	86	6	both	both	DET
alkej-870	86	7	inputs	input	NOUN
alkej-870	86	8	and	and	CCONJ
alkej-870	86	9	the	the	DET
alkej-870	86	10	need	need	NOUN
alkej-870	86	11	for	for	ADP
alkej-870	86	12	a	a	DET
alkej-870	86	13	large	large	ADJ
alkej-870	86	14	amount	amount	NOUN
alkej-870	86	15	of	of	ADP
alkej-870	86	16	training	training	NOUN
alkej-870	86	17	data	datum	NOUN
alkej-870	86	18	and	and	CCONJ
alkej-870	86	19	computational	computational	ADJ
alkej-870	86	20	resources	resource	NOUN
alkej-870	86	21	.	.	PUNCT
alkej-870	87	1	ephrat	ephrat	PROPN
alkej-870	87	2	et	et	PROPN
alkej-870	87	3	al	al	PROPN
alkej-870	87	4	.	.	PUNCT
alkej-870	88	1	[	[	X
alkej-870	88	2	17	17	NUM
alkej-870	88	3	]	]	PUNCT
alkej-870	88	4	proposed	propose	VERB
alkej-870	88	5	an	an	DET
alkej-870	88	6	audio	audio	ADJ
alkej-870	88	7	-	-	ADJ
alkej-870	88	8	visual	visual	ADJ
alkej-870	88	9	deep	deep	ADJ
alkej-870	88	10	neural	neural	ADJ
alkej-870	88	11	network	network	NOUN
alkej-870	88	12	approach	approach	NOUN
alkej-870	88	13	for	for	ADP
alkej-870	88	14	separating	separate	VERB
alkej-870	88	15	multiple	multiple	ADJ
alkej-870	88	16	speakers	speaker	NOUN
alkej-870	88	17	in	in	ADP
alkej-870	88	18	noisy	noisy	ADJ
alkej-870	88	19	environments	environment	NOUN
alkej-870	88	20	.	.	PUNCT
alkej-870	89	1	the	the	DET
alkej-870	89	2	approach	approach	NOUN
alkej-870	89	3	consists	consist	VERB
alkej-870	89	4	of	of	ADP
alkej-870	89	5	two	two	NUM
alkej-870	89	6	streams	stream	NOUN
alkej-870	89	7	:	:	PUNCT
alkej-870	89	8	an	an	DET
alkej-870	89	9	audio	audio	ADJ
alkej-870	89	10	stream	stream	NOUN
alkej-870	89	11	that	that	PRON
alkej-870	89	12	predicts	predict	VERB
alkej-870	89	13	ideal	ideal	ADJ
alkej-870	89	14	ratio	ratio	NOUN
alkej-870	89	15	masks	mask	NOUN
alkej-870	89	16	(	(	PUNCT
alkej-870	89	17	irm	irm	PROPN
alkej-870	89	18	)	)	PUNCT
alkej-870	89	19	and	and	CCONJ
alkej-870	89	20	a	a	DET
alkej-870	89	21	video	video	NOUN
alkej-870	89	22	stream	stream	NOUN
alkej-870	89	23	that	that	PRON
alkej-870	89	24	predicts	predict	VERB
alkej-870	89	25	attention	attention	NOUN
alkej-870	89	26	masks	mask	NOUN
alkej-870	89	27	to	to	PART
alkej-870	89	28	capture	capture	VERB
alkej-870	89	29	the	the	DET
alkej-870	89	30	spatial	spatial	ADJ
alkej-870	89	31	distribution	distribution	NOUN
alkej-870	89	32	of	of	ADP
alkej-870	89	33	speakers	speaker	NOUN
alkej-870	89	34	.	.	PUNCT
alkej-870	90	1	these	these	DET
alkej-870	90	2	streams	stream	NOUN
alkej-870	90	3	are	be	AUX
alkej-870	90	4	then	then	ADV
alkej-870	90	5	combined	combine	VERB
alkej-870	90	6	using	use	VERB
alkej-870	90	7	a	a	DET
alkej-870	90	8	multi	multi	ADJ
alkej-870	90	9	-	-	ADJ
alkej-870	90	10	modal	modal	ADJ
alkej-870	90	11	fusion	fusion	NOUN
alkej-870	90	12	network	network	NOUN
alkej-870	90	13	to	to	PART
alkej-870	90	14	generate	generate	VERB
alkej-870	90	15	the	the	DET
alkej-870	90	16	final	final	ADJ
alkej-870	90	17	separated	separate	VERB
alkej-870	90	18	speech	speech	NOUN
alkej-870	90	19	signals	signal	NOUN
alkej-870	90	20	.	.	PUNCT
alkej-870	91	1	the	the	DET
alkej-870	91	2	network	network	NOUN
alkej-870	91	3	utilizes	utilize	VERB
alkej-870	91	4	cnn	cnn	PROPN
alkej-870	91	5	and	and	CCONJ
alkej-870	91	6	blstm	blstm	PROPN
alkej-870	91	7	layers	layer	NOUN
alkej-870	91	8	to	to	PART
alkej-870	91	9	extract	extract	VERB
alkej-870	91	10	meaningful	meaningful	ADJ
alkej-870	91	11	features	feature	NOUN
alkej-870	91	12	and	and	CCONJ
alkej-870	91	13	model	model	NOUN
alkej-870	91	14	temporal	temporal	ADJ
alkej-870	91	15	dynamics	dynamic	NOUN
alkej-870	91	16	.	.	PUNCT
alkej-870	92	1	an	an	DET
alkej-870	92	2	advantage	advantage	NOUN
alkej-870	92	3	of	of	ADP
alkej-870	92	4	this	this	DET
alkej-870	92	5	approach	approach	NOUN
alkej-870	92	6	is	be	AUX
alkej-870	92	7	its	its	PRON
alkej-870	92	8	ability	ability	NOUN
alkej-870	92	9	to	to	PART
alkej-870	92	10	separate	separate	VERB
alkej-870	92	11	speech	speech	NOUN
alkej-870	92	12	even	even	ADV
alkej-870	92	13	when	when	SCONJ
alkej-870	92	14	the	the	DET
alkej-870	92	15	target	target	NOUN
alkej-870	92	16	speaker	speaker	NOUN
alkej-870	92	17	is	be	AUX
alkej-870	92	18	not	not	PART
alkej-870	92	19	visible	visible	ADJ
alkej-870	92	20	in	in	ADP
alkej-870	92	21	the	the	DET
alkej-870	92	22	camera	camera	NOUN
alkej-870	92	23	's	's	PART
alkej-870	92	24	field	field	NOUN
alkej-870	92	25	of	of	ADP
alkej-870	92	26	view	view	NOUN
alkej-870	92	27	.	.	PUNCT
alkej-870	93	1	however	however	ADV
alkej-870	93	2	,	,	PUNCT
alkej-870	93	3	it	it	PRON
alkej-870	93	4	relies	rely	VERB
alkej-870	93	5	on	on	ADP
alkej-870	93	6	synchronized	synchronize	VERB
alkej-870	93	7	audio	audio	ADJ
alkej-870	93	8	and	and	CCONJ
alkej-870	93	9	video	video	NOUN
alkej-870	93	10	data	datum	NOUN
alkej-870	93	11	,	,	PUNCT
alkej-870	93	12	and	and	CCONJ
alkej-870	93	13	its	its	PRON
alkej-870	93	14	performance	performance	NOUN
alkej-870	93	15	may	may	AUX
alkej-870	93	16	be	be	AUX
alkej-870	93	17	limited	limit	VERB
alkej-870	93	18	in	in	ADP
alkej-870	93	19	extremely	extremely	ADV
alkej-870	93	20	noisy	noisy	ADJ
alkej-870	93	21	environments	environment	NOUN
alkej-870	93	22	or	or	CCONJ
alkej-870	93	23	when	when	SCONJ
alkej-870	93	24	multiple	multiple	ADJ
alkej-870	93	25	speakers	speaker	NOUN
alkej-870	93	26	are	be	AUX
alkej-870	93	27	speaking	speak	VERB
alkej-870	93	28	simultaneously	simultaneously	ADV
alkej-870	93	29	.	.	PUNCT
alkej-870	94	1	hou	hou	NOUN
alkej-870	94	2	et	et	PROPN
alkej-870	94	3	al	al	PROPN
alkej-870	94	4	.	.	PUNCT
alkej-870	95	1	[	[	X
alkej-870	95	2	18	18	NUM
alkej-870	95	3	]	]	PUNCT
alkej-870	95	4	proposed	propose	VERB
alkej-870	95	5	deep	deep	ADJ
alkej-870	95	6	learning	learning	NOUN
alkej-870	95	7	-	-	PUNCT
alkej-870	95	8	based	base	VERB
alkej-870	95	9	audio	audio	ADJ
alkej-870	95	10	-	-	ADJ
alkej-870	95	11	visual	visual	ADJ
alkej-870	95	12	speech	speech	NOUN
alkej-870	95	13	enhancement	enhancement	NOUN
alkej-870	95	14	using	use	VERB
alkej-870	95	15	cnns	cnn	NOUN
alkej-870	95	16	for	for	ADP
alkej-870	95	17	audio	audio	ADJ
alkej-870	95	18	and	and	CCONJ
alkej-870	95	19	visual	visual	ADJ
alkej-870	95	20	processing	processing	NOUN
alkej-870	95	21	.	.	PUNCT
alkej-870	96	1	the	the	DET
alkej-870	96	2	multimodal	multimodal	ADJ
alkej-870	96	3	fusion	fusion	NOUN
alkej-870	96	4	network	network	NOUN
alkej-870	96	5	combines	combine	VERB
alkej-870	96	6	the	the	DET
alkej-870	96	7	two	two	NUM
alkej-870	96	8	branches	branch	NOUN
alkej-870	96	9	to	to	PART
alkej-870	96	10	enhance	enhance	VERB
alkej-870	96	11	speech	speech	NOUN
alkej-870	96	12	intelligibility	intelligibility	NOUN
alkej-870	96	13	with	with	ADP
alkej-870	96	14	lip	lip	NOUN
alkej-870	96	15	movements	movement	NOUN
alkej-870	96	16	.	.	PUNCT
alkej-870	97	1	advantages	advantage	NOUN
alkej-870	97	2	include	include	VERB
alkej-870	97	3	handling	handle	VERB
alkej-870	97	4	noisy	noisy	ADJ
alkej-870	97	5	speech	speech	NOUN
alkej-870	97	6	signals	signal	NOUN
alkej-870	97	7	and	and	CCONJ
alkej-870	97	8	avoiding	avoid	VERB
alkej-870	97	9	handcrafted	handcrafted	ADJ
alkej-870	97	10	features	feature	NOUN
alkej-870	97	11	.	.	PUNCT
alkej-870	98	1	however	however	ADV
alkej-870	98	2	,	,	PUNCT
alkej-870	98	3	using	use	VERB
alkej-870	98	4	lip	lip	NOUN
alkej-870	98	5	movements	movement	NOUN
alkej-870	98	6	as	as	ADP
alkej-870	98	7	a	a	DET
alkej-870	98	8	visual	visual	ADJ
alkej-870	98	9	modality	modality	NOUN
alkej-870	98	10	may	may	AUX
alkej-870	98	11	not	not	PART
alkej-870	98	12	be	be	AUX
alkej-870	98	13	suitable	suitable	ADJ
alkej-870	98	14	for	for	ADP
alkej-870	98	15	all	all	DET
alkej-870	98	16	scenarios	scenario	NOUN
alkej-870	98	17	.	.	PUNCT
alkej-870	99	1	lu	lu	VERB
alkej-870	99	2	et	et	PROPN
alkej-870	99	3	al	al	PROPN
alkej-870	99	4	.	.	PUNCT
alkej-870	100	1	[	[	X
alkej-870	100	2	19	19	NUM
alkej-870	100	3	]	]	PUNCT
alkej-870	100	4	proposed	propose	VERB
alkej-870	100	5	an	an	DET
alkej-870	100	6	approach	approach	NOUN
alkej-870	100	7	that	that	PRON
alkej-870	100	8	uses	use	VERB
alkej-870	100	9	deep	deep	ADJ
alkej-870	100	10	neural	neural	ADJ
alkej-870	100	11	networks	network	NOUN
alkej-870	100	12	for	for	ADP
alkej-870	100	13	audio	audio	ADJ
alkej-870	100	14	-	-	ADJ
alkej-870	100	15	visual	visual	ADJ
alkej-870	100	16	speech	speech	NOUN
alkej-870	100	17	separation	separation	NOUN
alkej-870	100	18	.	.	PUNCT
alkej-870	101	1	the	the	DET
alkej-870	101	2	network	network	NOUN
alkej-870	101	3	includes	include	VERB
alkej-870	101	4	cnn	cnn	PROPN
alkej-870	101	5	and	and	CCONJ
alkej-870	101	6	rnn	rnn	VERB
alkej-870	101	7	layers	layer	NOUN
alkej-870	101	8	for	for	ADP
alkej-870	101	9	audio	audio	NOUN
alkej-870	101	10	and	and	CCONJ
alkej-870	101	11	cnn	cnn	NOUN
alkej-870	101	12	layers	layer	NOUN
alkej-870	101	13	for	for	ADP
alkej-870	101	14	visual	visual	ADJ
alkej-870	101	15	processing	processing	NOUN
alkej-870	101	16	.	.	PUNCT
alkej-870	102	1	an	an	DET
alkej-870	102	2	attention	attention	NOUN
alkej-870	102	3	-	-	PUNCT
alkej-870	102	4	based	base	VERB
alkej-870	102	5	multimodal	multimodal	NOUN
alkej-870	102	6	fusion	fusion	NOUN
alkej-870	102	7	network	network	NOUN
alkej-870	102	8	is	be	AUX
alkej-870	102	9	estimated	estimate	VERB
alkej-870	102	10	by	by	ADP
alkej-870	102	11	combining	combine	VERB
alkej-870	102	12	the	the	DET
alkej-870	102	13	two	two	NUM
alkej-870	102	14	branches	branch	NOUN
alkej-870	102	15	,	,	PUNCT
alkej-870	102	16	and	and	CCONJ
alkej-870	102	17	estimate	estimate	VERB
alkej-870	102	18	the	the	DET
alkej-870	102	19	ideal	ideal	ADJ
alkej-870	102	20	ratio	ratio	NOUN
alkej-870	102	21	mask	mask	NOUN
alkej-870	102	22	for	for	ADP
alkej-870	102	23	speech	speech	NOUN
alkej-870	102	24	separation	separation	NOUN
alkej-870	102	25	.	.	PUNCT
alkej-870	103	1	the	the	DET
alkej-870	103	2	main	main	ADJ
alkej-870	103	3	disadvantage	disadvantage	NOUN
alkej-870	103	4	is	be	AUX
alkej-870	103	5	the	the	DET
alkej-870	103	6	need	need	NOUN
alkej-870	103	7	for	for	ADP
alkej-870	103	8	aligned	aligned	ADJ
alkej-870	103	9	audio	audio	ADJ
alkej-870	103	10	-	-	ADJ
alkej-870	103	11	visual	visual	ADJ
alkej-870	103	12	data	datum	NOUN
alkej-870	103	13	.	.	PUNCT
alkej-870	104	1	zhao	zhao	PROPN
alkej-870	104	2	et	et	PROPN
alkej-870	104	3	al	al	PROPN
alkej-870	104	4	.	.	PUNCT
alkej-870	105	1	[	[	X
alkej-870	105	2	20	20	NUM
alkej-870	105	3	]	]	PUNCT
alkej-870	105	4	proposed	propose	VERB
alkej-870	105	5	an	an	DET
alkej-870	105	6	audio	audio	ADJ
alkej-870	105	7	-	-	ADJ
alkej-870	105	8	visual	visual	ADJ
alkej-870	105	9	source	source	NOUN
alkej-870	105	10	separation	separation	NOUN
alkej-870	105	11	method	method	NOUN
alkej-870	105	12	using	use	VERB
alkej-870	105	13	a	a	DET
alkej-870	105	14	cnn	cnn	NOUN
alkej-870	105	15	.	.	PUNCT
alkej-870	106	1	it	it	PRON
alkej-870	106	2	uses	use	VERB
alkej-870	106	3	a	a	DET
alkej-870	106	4	two	two	NUM
alkej-870	106	5	-	-	PUNCT
alkej-870	106	6	stage	stage	NOUN
alkej-870	106	7	approach	approach	NOUN
alkej-870	106	8	:	:	PUNCT
alkej-870	106	9	first	first	ADV
alkej-870	106	10	,	,	PUNCT
alkej-870	106	11	the	the	DET
alkej-870	106	12	cnn	cnn	PROPN
alkej-870	106	13	predicts	predict	VERB
alkej-870	106	14	the	the	DET
alkej-870	106	15	magnitude	magnitude	NOUN
alkej-870	106	16	spectrogram	spectrogram	NOUN
alkej-870	106	17	of	of	ADP
alkej-870	106	18	the	the	DET
alkej-870	106	19	target	target	NOUN
alkej-870	106	20	speech	speech	NOUN
alkej-870	106	21	signal	signal	NOUN
alkej-870	106	22	from	from	ADP
alkej-870	106	23	visual	visual	ADJ
alkej-870	106	24	features	feature	NOUN
alkej-870	106	25	;	;	PUNCT
alkej-870	106	26	second	second	X
alkej-870	106	27	,	,	PUNCT
alkej-870	106	28	the	the	DET
alkej-870	106	29	predicted	predict	VERB
alkej-870	106	30	spectrogram	spectrogram	NOUN
alkej-870	106	31	is	be	AUX
alkej-870	106	32	used	use	VERB
alkej-870	106	33	to	to	PART
alkej-870	106	34	estimate	estimate	VERB
alkej-870	106	35	the	the	DET
alkej-870	106	36	speech	speech	NOUN
alkej-870	106	37	signal	signal	NOUN
alkej-870	106	38	.	.	PUNCT
alkej-870	107	1	visual	visual	ADJ
alkej-870	107	2	features	feature	NOUN
alkej-870	107	3	are	be	AUX
alkej-870	107	4	extracted	extract	VERB
alkej-870	107	5	from	from	ADP
alkej-870	107	6	a	a	DET
alkej-870	107	7	pre	pre	ADJ
alkej-870	107	8	-	-	ADJ
alkej-870	107	9	trained	train	VERB
alkej-870	107	10	network	network	NOUN
alkej-870	107	11	and	and	CCONJ
alkej-870	107	12	fine	fine	ADV
alkej-870	107	13	-	-	PUNCT
alkej-870	107	14	tuned	tune	VERB
alkej-870	107	15	on	on	ADP
alkej-870	107	16	mixture	mixture	NOUN
alkej-870	107	17	frames	frame	NOUN
alkej-870	107	18	.	.	PUNCT
alkej-870	108	1	it	it	PRON
alkej-870	108	2	can	can	AUX
alkej-870	108	3	separate	separate	VERB
alkej-870	108	4	speech	speech	NOUN
alkej-870	108	5	even	even	ADV
alkej-870	108	6	in	in	ADP
alkej-870	108	7	noisy	noisy	ADJ
alkej-870	108	8	or	or	CCONJ
alkej-870	108	9	reverberant	reverberant	ADJ
alkej-870	108	10	environments	environment	NOUN
alkej-870	108	11	.	.	PUNCT
alkej-870	109	1	however	however	ADV
alkej-870	109	2	,	,	PUNCT
alkej-870	109	3	it	it	PRON
alkej-870	109	4	requires	require	VERB
alkej-870	109	5	a	a	DET
alkej-870	109	6	large	large	ADJ
alkej-870	109	7	amount	amount	NOUN
alkej-870	109	8	of	of	ADP
alkej-870	109	9	training	training	NOUN
alkej-870	109	10	data	datum	NOUN
alkej-870	109	11	.	.	PUNCT
alkej-870	110	1	gabbay	gabbay	NOUN
alkej-870	110	2	et	et	PROPN
alkej-870	110	3	al	al	PROPN
alkej-870	110	4	.	.	PUNCT
alkej-870	111	1	[	[	X
alkej-870	111	2	21	21	NUM
alkej-870	111	3	]	]	PUNCT
alkej-870	111	4	proposed	propose	VERB
alkej-870	111	5	a	a	DET
alkej-870	111	6	method	method	NOUN
alkej-870	111	7	for	for	ADP
alkej-870	111	8	enhancing	enhance	VERB
alkej-870	111	9	speech	speech	NOUN
alkej-870	111	10	signals	signal	NOUN
alkej-870	111	11	in	in	ADP
alkej-870	111	12	noisy	noisy	ADJ
alkej-870	111	13	environments	environment	NOUN
alkej-870	111	14	using	use	VERB
alkej-870	111	15	visual	visual	ADJ
alkej-870	111	16	information	information	NOUN
alkej-870	111	17	from	from	ADP
alkej-870	111	18	a	a	DET
alkej-870	111	19	video	video	NOUN
alkej-870	111	20	recording	recording	NOUN
alkej-870	111	21	of	of	ADP
alkej-870	111	22	speakers	speaker	NOUN
alkej-870	111	23	.	.	PUNCT
alkej-870	112	1	they	they	PRON
alkej-870	112	2	used	use	VERB
alkej-870	112	3	deep	deep	ADJ
alkej-870	112	4	learning	learning	NOUN
alkej-870	112	5	to	to	PART
alkej-870	112	6	extract	extract	VERB
alkej-870	112	7	facial	facial	ADJ
alkej-870	112	8	features	feature	NOUN
alkej-870	112	9	from	from	ADP
alkej-870	112	10	a	a	DET
alkej-870	112	11	cnn	cnn	NOUN
alkej-870	112	12	and	and	CCONJ
alkej-870	112	13	integrated	integrate	VERB
alkej-870	112	14	them	they	PRON
alkej-870	112	15	into	into	ADP
alkej-870	112	16	a	a	DET
alkej-870	112	17	deep	deep	ADJ
alkej-870	112	18	rnn	rnn	NOUN
alkej-870	112	19	for	for	ADP
alkej-870	112	20	speech	speech	NOUN
alkej-870	112	21	separation	separation	NOUN
alkej-870	112	22	.	.	PUNCT
alkej-870	113	1	achieved	achieve	VERB
alkej-870	113	2	state	state	NOUN
alkej-870	113	3	-	-	PUNCT
alkej-870	113	4	of	of	ADP
alkej-870	113	5	-	-	PUNCT
alkej-870	113	6	the	the	DET
alkej-870	113	7	-	-	PUNCT
alkej-870	113	8	art	art	NOUN
alkej-870	113	9	performance	performance	NOUN
alkej-870	113	10	,	,	PUNCT
alkej-870	113	11	but	but	CCONJ
alkej-870	113	12	limitations	limitation	NOUN
alkej-870	113	13	include	include	VERB
alkej-870	113	14	quality	quality	NOUN
alkej-870	113	15	of	of	ADP
alkej-870	113	16	video	video	NOUN
alkej-870	113	17	,	,	PUNCT
alkej-870	113	18	accuracy	accuracy	NOUN
alkej-870	113	19	of	of	ADP
alkej-870	113	20	facial	facial	ADJ
alkej-870	113	21	feature	feature	NOUN
alkej-870	113	22	extraction	extraction	NOUN
alkej-870	113	23	,	,	PUNCT
alkej-870	113	24	and	and	CCONJ
alkej-870	113	25	computational	computational	ADJ
alkej-870	113	26	resources	resource	NOUN
alkej-870	113	27	required	require	VERB
alkej-870	113	28	.	.	PUNCT
alkej-870	114	1	gogate	gogate	PROPN
alkej-870	114	2	et	et	PROPN
alkej-870	114	3	al	al	PROPN
alkej-870	114	4	.	.	PUNCT
alkej-870	115	1	[	[	X
alkej-870	115	2	22	22	NUM
alkej-870	115	3	]	]	PUNCT
alkej-870	115	4	proposed	propose	VERB
alkej-870	115	5	a	a	DET
alkej-870	115	6	method	method	NOUN
alkej-870	115	7	that	that	PRON
alkej-870	115	8	uses	use	VERB
alkej-870	115	9	a	a	DET
alkej-870	115	10	cnn	cnn	NOUN
alkej-870	115	11	to	to	PART
alkej-870	115	12	extract	extract	VERB
alkej-870	115	13	visual	visual	ADJ
alkej-870	115	14	features	feature	NOUN
alkej-870	115	15	and	and	CCONJ
alkej-870	115	16	an	an	DET
alkej-870	115	17	rnn	rnn	NOUN
alkej-870	115	18	to	to	PART
alkej-870	115	19	process	process	VERB
alkej-870	115	20	audio	audio	ADJ
alkej-870	115	21	signals	signal	NOUN
alkej-870	115	22	to	to	PART
alkej-870	115	23	estimate	estimate	VERB
alkej-870	115	24	a	a	DET
alkej-870	115	25	mask	mask	NOUN
alkej-870	115	26	separating	separate	VERB
alkej-870	115	27	the	the	DET
alkej-870	115	28	speaker	speaker	NOUN
alkej-870	115	29	's	's	PART
alkej-870	115	30	voice	voice	NOUN
alkej-870	115	31	from	from	ADP
alkej-870	115	32	background	background	NOUN
alkej-870	115	33	noise	noise	NOUN
alkej-870	115	34	.	.	PUNCT
alkej-870	116	1	deep	deep	ADJ
alkej-870	116	2	learning	learning	NOUN
alkej-870	116	3	improves	improve	VERB
alkej-870	116	4	mask	mask	NOUN
alkej-870	116	5	estimation	estimation	NOUN
alkej-870	116	6	and	and	CCONJ
alkej-870	116	7	speech	speech	NOUN
alkej-870	116	8	separation	separation	NOUN
alkej-870	116	9	.	.	PUNCT
alkej-870	117	1	the	the	DET
alkej-870	117	2	method	method	NOUN
alkej-870	117	3	is	be	AUX
alkej-870	117	4	speakerindependent	speakerindependent	NOUN
alkej-870	117	5	but	but	CCONJ
alkej-870	117	6	computationally	computationally	ADV
alkej-870	117	7	expensive	expensive	ADJ
alkej-870	117	8	for	for	ADP
alkej-870	117	9	real	real	ADJ
alkej-870	117	10	-	-	PUNCT
alkej-870	117	11	time	time	NOUN
alkej-870	117	12	use	use	NOUN
alkej-870	117	13	.	.	PUNCT
alkej-870	118	1	morrone	morrone	NOUN
alkej-870	118	2	et	et	PROPN
alkej-870	118	3	al	al	PROPN
alkej-870	118	4	.	.	PUNCT
alkej-870	119	1	[	[	X
alkej-870	119	2	23	23	NUM
alkej-870	119	3	]	]	PUNCT
alkej-870	119	4	proposed	propose	VERB
alkej-870	119	5	a	a	DET
alkej-870	119	6	method	method	NOUN
alkej-870	119	7	for	for	ADP
alkej-870	119	8	enhancing	enhance	VERB
alkej-870	119	9	speech	speech	NOUN
alkej-870	119	10	using	use	VERB
alkej-870	119	11	audio	audio	ADJ
alkej-870	119	12	and	and	CCONJ
alkej-870	119	13	visual	visual	ADJ
alkej-870	119	14	information	information	NOUN
alkej-870	119	15	in	in	ADP
alkej-870	119	16	multi	multi	ADJ
alkej-870	119	17	-	-	ADJ
alkej-870	119	18	talker	talker	ADJ
alkej-870	119	19	environments	environment	NOUN
alkej-870	119	20	.	.	PUNCT
alkej-870	120	1	the	the	DET
alkej-870	120	2	deep	deep	ADJ
alkej-870	120	3	neural	neural	ADJ
alkej-870	120	4	network	network	NOUN
alkej-870	120	5	architecture	architecture	NOUN
alkej-870	120	6	includes	include	VERB
alkej-870	120	7	audio	audio	ADJ
alkej-870	120	8	and	and	CCONJ
alkej-870	120	9	visual	visual	ADJ
alkej-870	120	10	branches	branch	NOUN
alkej-870	120	11	,	,	PUNCT
alkej-870	120	12	with	with	ADP
alkej-870	120	13	cnn	cnn	PROPN
alkej-870	120	14	and	and	CCONJ
alkej-870	120	15	rnn	rnn	VERB
alkej-870	120	16	in	in	ADP
alkej-870	120	17	the	the	DET
alkej-870	120	18	audio	audio	ADJ
alkej-870	120	19	branch	branch	NOUN
alkej-870	120	20	and	and	CCONJ
alkej-870	120	21	face	face	NOUN
alkej-870	120	22	detection	detection	NOUN
alkej-870	120	23	and	and	CCONJ
alkej-870	120	24	landmark	landmark	NOUN
alkej-870	120	25	localization	localization	NOUN
alkej-870	120	26	algorithms	algorithm	NOUN
alkej-870	120	27	in	in	ADP
alkej-870	120	28	the	the	DET
alkej-870	120	29	visual	visual	ADJ
alkej-870	120	30	branch	branch	NOUN
alkej-870	120	31	.	.	PUNCT
alkej-870	121	1	the	the	DET
alkej-870	121	2	attention	attention	NOUN
alkej-870	121	3	mechanism	mechanism	NOUN
alkej-870	121	4	combines	combine	VERB
alkej-870	121	5	the	the	DET
alkej-870	121	6	two	two	NUM
alkej-870	121	7	branches	branch	NOUN
alkej-870	121	8	for	for	ADP
alkej-870	121	9	final	final	ADJ
alkej-870	121	10	results	result	NOUN
alkej-870	121	11	.	.	PUNCT
alkej-870	122	1	the	the	DET
alkej-870	122	2	disadvantage	disadvantage	NOUN
alkej-870	122	3	is	be	AUX
alkej-870	122	4	that	that	SCONJ
alkej-870	122	5	the	the	DET
alkej-870	122	6	accuracy	accuracy	NOUN
alkej-870	122	7	of	of	ADP
alkej-870	122	8	facial	facial	ADJ
alkej-870	122	9	landmark	landmark	NOUN
alkej-870	122	10	detection	detection	NOUN
alkej-870	122	11	can	can	AUX
alkej-870	122	12	affect	affect	VERB
alkej-870	122	13	performance	performance	NOUN
alkej-870	122	14	.	.	PUNCT
alkej-870	123	1	wu	wu	PROPN
alkej-870	123	2	et	et	PROPN
alkej-870	123	3	al	al	PROPN
alkej-870	123	4	.	.	PUNCT
alkej-870	124	1	[	[	X
alkej-870	124	2	24	24	NUM
alkej-870	124	3	]	]	PUNCT
alkej-870	124	4	proposed	propose	VERB
alkej-870	124	5	a	a	DET
alkej-870	124	6	method	method	NOUN
alkej-870	124	7	to	to	PART
alkej-870	124	8	separate	separate	VERB
alkej-870	124	9	speech	speech	NOUN
alkej-870	124	10	signals	signal	NOUN
alkej-870	124	11	of	of	ADP
alkej-870	124	12	multiple	multiple	ADJ
alkej-870	124	13	speakers	speaker	NOUN
alkej-870	124	14	in	in	ADP
alkej-870	124	15	an	an	DET
alkej-870	124	16	audiovisual	audiovisual	ADJ
alkej-870	124	17	recording	recording	NOUN
alkej-870	124	18	using	use	VERB
alkej-870	124	19	deep	deep	ADJ
alkej-870	124	20	neural	neural	ADJ
alkej-870	124	21	networks	network	NOUN
alkej-870	124	22	.	.	PUNCT
alkej-870	125	1	the	the	DET
alkej-870	125	2	method	method	NOUN
alkej-870	125	3	jointly	jointly	ADV
alkej-870	125	4	learns	learn	VERB
alkej-870	125	5	audio	audio	ADJ
alkej-870	125	6	and	and	CCONJ
alkej-870	125	7	visual	visual	ADJ
alkej-870	125	8	features	feature	NOUN
alkej-870	125	9	of	of	ADP
alkej-870	125	10	the	the	DET
alkej-870	125	11	speech	speech	NOUN
alkej-870	125	12	signals	signal	NOUN
alkej-870	125	13	and	and	CCONJ
alkej-870	125	14	performs	perform	VERB
alkej-870	125	15	separation	separation	NOUN
alkej-870	125	16	in	in	ADP
alkej-870	125	17	the	the	DET
alkej-870	125	18	time	time	NOUN
alkej-870	125	19	domain	domain	NOUN
alkej-870	125	20	.	.	PUNCT
alkej-870	126	1	visual	visual	ADJ
alkej-870	126	2	features	feature	NOUN
alkej-870	126	3	are	be	AUX
alkej-870	126	4	extracted	extract	VERB
alkej-870	126	5	using	use	VERB
alkej-870	126	6	cnn	cnn	PROPN
alkej-870	126	7	and	and	CCONJ
alkej-870	126	8	temporal	temporal	ADJ
alkej-870	126	9	dynamics	dynamic	NOUN
alkej-870	126	10	are	be	AUX
alkej-870	126	11	modeled	model	VERB
alkej-870	126	12	using	use	VERB
alkej-870	126	13	lstm	lstm	NOUN
alkej-870	126	14	.	.	PUNCT
alkej-870	127	1	a	a	DET
alkej-870	127	2	deep	deep	ADJ
alkej-870	127	3	clustering	clustering	ADJ
alkej-870	127	4	network	network	NOUN
alkej-870	127	5	is	be	AUX
alkej-870	127	6	used	use	VERB
alkej-870	127	7	to	to	PART
alkej-870	127	8	fuse	fuse	VERB
alkej-870	127	9	audio	audio	ADJ
alkej-870	127	10	and	and	CCONJ
alkej-870	127	11	visual	visual	ADJ
alkej-870	127	12	features	feature	NOUN
alkej-870	127	13	.	.	PUNCT
alkej-870	128	1	the	the	DET
alkej-870	128	2	method	method	NOUN
alkej-870	128	3	preserves	preserve	VERB
alkej-870	128	4	the	the	DET
alkej-870	128	5	temporal	temporal	ADJ
alkej-870	128	6	characteristics	characteristic	NOUN
alkej-870	128	7	of	of	ADP
alkej-870	128	8	speech	speech	NOUN
alkej-870	128	9	signals	signal	NOUN
alkej-870	128	10	,	,	PUNCT
alkej-870	128	11	but	but	CCONJ
alkej-870	128	12	nooralhuda	nooralhuda	NOUN
alkej-870	128	13	mudhafar	mudhafar	PROPN
alkej-870	128	14	sleman	sleman	PROPN
alkej-870	128	15	al	al	PROPN
alkej-870	128	16	-	-	PUNCT
alkej-870	128	17	khwarizmi	khwarizmi	PROPN
alkej-870	128	18	engineering	engineering	NOUN
alkej-870	128	19	journal	journal	PROPN
alkej-870	128	20	,	,	PUNCT
alkej-870	128	21	vol	vol	NOUN
alkej-870	128	22	.	.	PROPN
alkej-870	128	23	19	19	NUM
alkej-870	128	24	,	,	PUNCT
alkej-870	128	25	no	no	INTJ
alkej-870	128	26	.	.	NOUN
alkej-870	128	27	4	4	NUM
alkej-870	128	28	,	,	PUNCT
alkej-870	128	29	p.p	p.p	PROPN
alkej-870	128	30	.	.	NOUN
alkej-870	128	31	4255	4255	NUM
alkej-870	128	32	(	(	PUNCT
alkej-870	128	33	2023	2023	NUM
alkej-870	128	34	)	)	PUNCT
alkej-870	128	35	46	46	NUM
alkej-870	128	36	may	may	AUX
alkej-870	128	37	not	not	PART
alkej-870	128	38	work	work	VERB
alkej-870	128	39	well	well	ADV
alkej-870	128	40	in	in	ADP
alkej-870	128	41	cases	case	NOUN
alkej-870	128	42	with	with	ADP
alkej-870	128	43	similar	similar	ADJ
alkej-870	128	44	speakers	speaker	NOUN
alkej-870	128	45	or	or	CCONJ
alkej-870	128	46	occlusions	occlusion	NOUN
alkej-870	128	47	in	in	ADP
alkej-870	128	48	the	the	DET
alkej-870	128	49	visual	visual	ADJ
alkej-870	128	50	data	datum	NOUN
alkej-870	128	51	and	and	CCONJ
alkej-870	128	52	may	may	AUX
alkej-870	128	53	have	have	VERB
alkej-870	128	54	long	long	ADJ
alkej-870	128	55	processing	processing	NOUN
alkej-870	128	56	times	time	NOUN
alkej-870	128	57	.	.	PUNCT
alkej-870	129	1	gogate	gogate	PROPN
alkej-870	129	2	et	et	PROPN
alkej-870	129	3	al	al	PROPN
alkej-870	129	4	.	.	PUNCT
alkej-870	130	1	[	[	X
alkej-870	130	2	25	25	NUM
alkej-870	130	3	]	]	PUNCT
alkej-870	130	4	proposed	propose	VERB
alkej-870	130	5	that	that	SCONJ
alkej-870	130	6	the	the	DET
alkej-870	130	7	dnn	dnn	PROPN
alkej-870	130	8	model	model	NOUN
alkej-870	130	9	combines	combine	VERB
alkej-870	130	10	binaural	binaural	ADJ
alkej-870	130	11	cues	cue	NOUN
alkej-870	130	12	and	and	CCONJ
alkej-870	130	13	visual	visual	ADJ
alkej-870	130	14	input	input	NOUN
alkej-870	130	15	to	to	PART
alkej-870	130	16	improve	improve	VERB
alkej-870	130	17	the	the	DET
alkej-870	130	18	speech	speech	NOUN
alkej-870	130	19	quality	quality	NOUN
alkej-870	130	20	of	of	ADP
alkej-870	130	21	unknown	unknown	ADJ
alkej-870	130	22	speakers	speaker	NOUN
alkej-870	130	23	in	in	ADP
alkej-870	130	24	a	a	DET
alkej-870	130	25	noisy	noisy	ADJ
alkej-870	130	26	environment	environment	NOUN
alkej-870	130	27	.	.	PUNCT
alkej-870	131	1	multi	multi	ADJ
alkej-870	131	2	-	-	ADJ
alkej-870	131	3	stream	stream	ADJ
alkej-870	131	4	cnn	cnn	PROPN
alkej-870	131	5	considers	consider	VERB
alkej-870	131	6	the	the	DET
alkej-870	131	7	temporal	temporal	ADJ
alkej-870	131	8	and	and	CCONJ
alkej-870	131	9	spatial	spatial	ADJ
alkej-870	131	10	dynamics	dynamic	NOUN
alkej-870	131	11	of	of	ADP
alkej-870	131	12	binaural	binaural	ADJ
alkej-870	131	13	audio	audio	ADJ
alkej-870	131	14	and	and	CCONJ
alkej-870	131	15	visual	visual	ADJ
alkej-870	131	16	lip	lip	NOUN
alkej-870	131	17	images	image	NOUN
alkej-870	131	18	to	to	PART
alkej-870	131	19	estimate	estimate	VERB
alkej-870	131	20	spectral	spectral	ADJ
alkej-870	131	21	masks	mask	NOUN
alkej-870	131	22	for	for	ADP
alkej-870	131	23	each	each	DET
alkej-870	131	24	channel	channel	NOUN
alkej-870	131	25	.	.	PUNCT
alkej-870	132	1	binaural	binaural	ADJ
alkej-870	132	2	audio	audio	ADJ
alkej-870	132	3	signals	signal	NOUN
alkej-870	132	4	preserve	preserve	VERB
alkej-870	132	5	spatial	spatial	ADJ
alkej-870	132	6	information	information	NOUN
alkej-870	132	7	,	,	PUNCT
alkej-870	132	8	improving	improve	VERB
alkej-870	132	9	speech	speech	NOUN
alkej-870	132	10	separation	separation	NOUN
alkej-870	132	11	.	.	PUNCT
alkej-870	133	1	disadvantages	disadvantage	NOUN
alkej-870	133	2	include	include	VERB
alkej-870	133	3	the	the	DET
alkej-870	133	4	complexity	complexity	NOUN
alkej-870	133	5	of	of	ADP
alkej-870	133	6	training	training	NOUN
alkej-870	133	7	due	due	ADJ
alkej-870	133	8	to	to	ADP
alkej-870	133	9	supervised	supervised	ADJ
alkej-870	133	10	and	and	CCONJ
alkej-870	133	11	unsupervised	unsupervised	ADJ
alkej-870	133	12	learning	learning	NOUN
alkej-870	133	13	methods	method	NOUN
alkej-870	133	14	.	.	PUNCT
alkej-870	134	1	nguyen	nguyen	PROPN
alkej-870	134	2	et	et	PROPN
alkej-870	134	3	al	al	PROPN
alkej-870	134	4	.	.	PUNCT
alkej-870	135	1	[	[	X
alkej-870	135	2	26	26	NUM
alkej-870	135	3	]	]	PUNCT
alkej-870	135	4	proposed	propose	VERB
alkej-870	135	5	a	a	DET
alkej-870	135	6	method	method	NOUN
alkej-870	135	7	to	to	PART
alkej-870	135	8	improve	improve	VERB
alkej-870	135	9	audio	audio	ADJ
alkej-870	135	10	-	-	ADJ
alkej-870	135	11	visual	visual	ADJ
alkej-870	135	12	sound	sound	NOUN
alkej-870	135	13	source	source	NOUN
alkej-870	135	14	separation	separation	NOUN
alkej-870	135	15	using	use	VERB
alkej-870	135	16	object	object	NOUN
alkej-870	135	17	-	-	PUNCT
alkej-870	135	18	level	level	NOUN
alkej-870	135	19	priors	prior	NOUN
alkej-870	135	20	.	.	PUNCT
alkej-870	136	1	they	they	PRON
alkej-870	136	2	used	use	VERB
alkej-870	136	3	a	a	DET
alkej-870	136	4	deep	deep	ADJ
alkej-870	136	5	neural	neural	ADJ
alkej-870	136	6	network	network	NOUN
alkej-870	136	7	with	with	ADP
alkej-870	136	8	convolutional	convolutional	ADJ
alkej-870	136	9	and	and	CCONJ
alkej-870	136	10	deconvolutional	deconvolutional	ADJ
alkej-870	136	11	layers	layer	NOUN
alkej-870	136	12	to	to	PART
alkej-870	136	13	learn	learn	VERB
alkej-870	136	14	audio	audio	ADJ
alkej-870	136	15	and	and	CCONJ
alkej-870	136	16	visual	visual	ADJ
alkej-870	136	17	features	feature	NOUN
alkej-870	136	18	and	and	CCONJ
alkej-870	136	19	estimate	estimate	VERB
alkej-870	136	20	masks	mask	NOUN
alkej-870	136	21	for	for	ADP
alkej-870	136	22	each	each	DET
alkej-870	136	23	sound	sound	ADJ
alkej-870	136	24	source	source	NOUN
alkej-870	136	25	.	.	PUNCT
alkej-870	137	1	the	the	DET
alkej-870	137	2	method	method	NOUN
alkej-870	137	3	involves	involve	VERB
alkej-870	137	4	two	two	NUM
alkej-870	137	5	stages	stage	NOUN
alkej-870	137	6	:	:	PUNCT
alkej-870	137	7	object	object	NOUN
alkej-870	137	8	-	-	PUNCT
alkej-870	137	9	level	level	NOUN
alkej-870	137	10	localization	localization	NOUN
alkej-870	137	11	and	and	CCONJ
alkej-870	137	12	object	object	NOUN
alkej-870	137	13	-	-	PUNCT
alkej-870	137	14	aware	aware	ADJ
alkej-870	137	15	separation	separation	NOUN
alkej-870	137	16	.	.	PUNCT
alkej-870	138	1	in	in	ADP
alkej-870	138	2	the	the	DET
alkej-870	138	3	first	first	ADJ
alkej-870	138	4	stage	stage	NOUN
alkej-870	138	5	,	,	PUNCT
alkej-870	138	6	an	an	DET
alkej-870	138	7	object	object	NOUN
alkej-870	138	8	detection	detection	NOUN
alkej-870	138	9	model	model	NOUN
alkej-870	138	10	is	be	AUX
alkej-870	138	11	used	use	VERB
alkej-870	138	12	to	to	PART
alkej-870	138	13	localize	localize	VERB
alkej-870	138	14	the	the	DET
alkej-870	138	15	sound	sound	NOUN
alkej-870	138	16	sources	source	NOUN
alkej-870	138	17	in	in	ADP
alkej-870	138	18	the	the	DET
alkej-870	138	19	video	video	NOUN
alkej-870	138	20	frames	frame	NOUN
alkej-870	138	21	.	.	PUNCT
alkej-870	139	1	the	the	DET
alkej-870	139	2	audio	audio	ADJ
alkej-870	139	3	signals	signal	NOUN
alkej-870	139	4	are	be	AUX
alkej-870	139	5	separated	separate	VERB
alkej-870	139	6	in	in	ADP
alkej-870	139	7	the	the	DET
alkej-870	139	8	second	second	ADJ
alkej-870	139	9	stage	stage	NOUN
alkej-870	139	10	based	base	VERB
alkej-870	139	11	on	on	ADP
alkej-870	139	12	the	the	DET
alkej-870	139	13	estimated	estimate	VERB
alkej-870	139	14	object	object	NOUN
alkej-870	139	15	locations	location	NOUN
alkej-870	139	16	and	and	CCONJ
alkej-870	139	17	their	their	PRON
alkej-870	139	18	corresponding	corresponding	ADJ
alkej-870	139	19	masks	mask	NOUN
alkej-870	139	20	.	.	PUNCT
alkej-870	140	1	the	the	DET
alkej-870	140	2	advantage	advantage	NOUN
alkej-870	140	3	is	be	AUX
alkej-870	140	4	that	that	SCONJ
alkej-870	140	5	it	it	PRON
alkej-870	140	6	can	can	AUX
alkej-870	140	7	better	well	ADV
alkej-870	140	8	exploit	exploit	VERB
alkej-870	140	9	object	object	NOUN
alkej-870	140	10	-	-	PUNCT
alkej-870	140	11	level	level	NOUN
alkej-870	140	12	information	information	NOUN
alkej-870	140	13	and	and	CCONJ
alkej-870	140	14	improve	improve	VERB
alkej-870	140	15	separation	separation	NOUN
alkej-870	140	16	performance	performance	NOUN
alkej-870	140	17	.	.	PUNCT
alkej-870	141	1	still	still	ADV
alkej-870	141	2	,	,	PUNCT
alkej-870	141	3	the	the	DET
alkej-870	141	4	method	method	NOUN
alkej-870	141	5	relies	rely	VERB
alkej-870	141	6	on	on	ADP
alkej-870	141	7	an	an	DET
alkej-870	141	8	object	object	NOUN
alkej-870	141	9	detection	detection	NOUN
alkej-870	141	10	model	model	NOUN
alkej-870	141	11	,	,	PUNCT
alkej-870	141	12	which	which	PRON
alkej-870	141	13	can	can	AUX
alkej-870	141	14	be	be	AUX
alkej-870	141	15	challenging	challenge	VERB
alkej-870	141	16	in	in	ADP
alkej-870	141	17	complex	complex	ADJ
alkej-870	141	18	scenes	scene	NOUN
alkej-870	141	19	with	with	ADP
alkej-870	141	20	multiple	multiple	ADJ
alkej-870	141	21	objects	object	NOUN
alkej-870	141	22	or	or	CCONJ
alkej-870	141	23	occlusions	occlusion	NOUN
alkej-870	141	24	.	.	PUNCT
alkej-870	142	1	li	li	PROPN
alkej-870	142	2	et	et	PROPN
alkej-870	142	3	al	al	PROPN
alkej-870	142	4	.	.	PUNCT
alkej-870	143	1	[	[	X
alkej-870	143	2	27	27	NUM
alkej-870	143	3	]	]	PUNCT
alkej-870	143	4	presented	present	VERB
alkej-870	143	5	a	a	DET
alkej-870	143	6	model	model	NOUN
alkej-870	143	7	for	for	ADP
alkej-870	143	8	extracting	extract	VERB
alkej-870	143	9	speech	speech	NOUN
alkej-870	143	10	signals	signal	NOUN
alkej-870	143	11	from	from	ADP
alkej-870	143	12	mixed	mixed	ADJ
alkej-870	143	13	sound	sound	NOUN
alkej-870	143	14	sources	source	NOUN
alkej-870	143	15	in	in	ADP
alkej-870	143	16	a	a	DET
alkej-870	143	17	single	single	ADJ
alkej-870	143	18	-	-	PUNCT
alkej-870	143	19	channel	channel	NOUN
alkej-870	143	20	recording	recording	NOUN
alkej-870	143	21	using	use	VERB
alkej-870	143	22	both	both	CCONJ
alkej-870	143	23	audio	audio	ADJ
alkej-870	143	24	and	and	CCONJ
alkej-870	143	25	visual	visual	ADJ
alkej-870	143	26	data	datum	NOUN
alkej-870	143	27	.	.	PUNCT
alkej-870	144	1	it	it	PRON
alkej-870	144	2	makes	make	VERB
alkej-870	144	3	use	use	NOUN
alkej-870	144	4	of	of	ADP
alkej-870	144	5	an	an	DET
alkej-870	144	6	attention	attention	NOUN
alkej-870	144	7	mechanism	mechanism	NOUN
alkej-870	144	8	and	and	CCONJ
alkej-870	144	9	is	be	AUX
alkej-870	144	10	composed	compose	VERB
alkej-870	144	11	of	of	ADP
alkej-870	144	12	fully	fully	ADV
alkej-870	144	13	connected	connected	ADJ
alkej-870	144	14	layers	layer	NOUN
alkej-870	144	15	,	,	PUNCT
alkej-870	144	16	bidirectional	bidirectional	ADJ
alkej-870	144	17	lstm	lstm	NOUN
alkej-870	144	18	,	,	PUNCT
alkej-870	144	19	and	and	CCONJ
alkej-870	144	20	cnn	cnn	PROPN
alkej-870	144	21	.	.	PUNCT
alkej-870	145	1	for	for	ADP
alkej-870	145	2	best	good	ADJ
alkej-870	145	3	results	result	NOUN
alkej-870	145	4	in	in	ADP
alkej-870	145	5	terms	term	NOUN
alkej-870	145	6	of	of	ADP
alkej-870	145	7	performance	performance	NOUN
alkej-870	145	8	,	,	PUNCT
alkej-870	145	9	it	it	PRON
alkej-870	145	10	may	may	AUX
alkej-870	145	11	require	require	VERB
alkej-870	145	12	a	a	DET
alkej-870	145	13	large	large	ADJ
alkej-870	145	14	amount	amount	NOUN
alkej-870	145	15	of	of	ADP
alkej-870	145	16	training	training	NOUN
alkej-870	145	17	data	datum	NOUN
alkej-870	145	18	and	and	CCONJ
alkej-870	145	19	be	be	AUX
alkej-870	145	20	computationally	computationally	ADV
alkej-870	145	21	costly	costly	ADJ
alkej-870	145	22	.	.	PUNCT
alkej-870	146	1	gu	gu	NOUN
alkej-870	146	2	et	et	PROPN
alkej-870	146	3	al	al	PROPN
alkej-870	146	4	.	.	PUNCT
alkej-870	147	1	[	[	X
alkej-870	147	2	28	28	NUM
alkej-870	147	3	]	]	PUNCT
alkej-870	147	4	presented	present	VERB
alkej-870	147	5	a	a	DET
alkej-870	147	6	deep	deep	ADJ
alkej-870	147	7	learning	learning	NOUN
alkej-870	147	8	method	method	NOUN
alkej-870	147	9	that	that	PRON
alkej-870	147	10	combines	combine	VERB
alkej-870	147	11	audio	audio	ADJ
alkej-870	147	12	and	and	CCONJ
alkej-870	147	13	visual	visual	ADJ
alkej-870	147	14	data	datum	NOUN
alkej-870	147	15	to	to	PART
alkej-870	147	16	distinguish	distinguish	VERB
alkej-870	147	17	desirable	desirable	ADJ
alkej-870	147	18	speech	speech	NOUN
alkej-870	147	19	from	from	ADP
alkej-870	147	20	distracting	distract	VERB
alkej-870	147	21	sources	source	NOUN
alkej-870	147	22	.	.	PUNCT
alkej-870	148	1	in	in	ADP
alkej-870	148	2	order	order	NOUN
alkej-870	148	3	to	to	PART
alkej-870	148	4	improve	improve	VERB
alkej-870	148	5	separation	separation	NOUN
alkej-870	148	6	performance	performance	NOUN
alkej-870	148	7	,	,	PUNCT
alkej-870	148	8	the	the	DET
alkej-870	148	9	model	model	NOUN
alkej-870	148	10	integrates	integrate	VERB
alkej-870	148	11	an	an	DET
alkej-870	148	12	optical	optical	ADJ
alkej-870	148	13	encoder	encoder	NOUN
alkej-870	148	14	network	network	NOUN
alkej-870	148	15	for	for	ADP
alkej-870	148	16	extracting	extract	VERB
alkej-870	148	17	visual	visual	ADJ
alkej-870	148	18	information	information	NOUN
alkej-870	148	19	,	,	PUNCT
alkej-870	148	20	a	a	DET
alkej-870	148	21	cnn	cnn	NOUN
alkej-870	148	22	for	for	ADP
alkej-870	148	23	processing	process	VERB
alkej-870	148	24	audio	audio	NOUN
alkej-870	148	25	features	feature	NOUN
alkej-870	148	26	,	,	PUNCT
alkej-870	148	27	and	and	CCONJ
alkej-870	148	28	a	a	DET
alkej-870	148	29	multimodal	multimodal	ADJ
alkej-870	148	30	fusion	fusion	NOUN
alkej-870	148	31	network	network	NOUN
alkej-870	148	32	.	.	PUNCT
alkej-870	149	1	utilizing	utilize	VERB
alkej-870	149	2	both	both	CCONJ
alkej-870	149	3	aural	aural	ADJ
alkej-870	149	4	and	and	CCONJ
alkej-870	149	5	visual	visual	ADJ
alkej-870	149	6	modalities	modality	NOUN
alkej-870	149	7	,	,	PUNCT
alkej-870	149	8	the	the	DET
alkej-870	149	9	method	method	NOUN
alkej-870	149	10	gains	gain	VERB
alkej-870	149	11	a	a	DET
alkej-870	149	12	more	more	ADV
alkej-870	149	13	comprehensive	comprehensive	ADJ
alkej-870	149	14	representation	representation	NOUN
alkej-870	149	15	of	of	ADP
alkej-870	149	16	the	the	DET
alkej-870	149	17	input	input	NOUN
alkej-870	149	18	signals	signal	NOUN
alkej-870	149	19	.	.	PUNCT
alkej-870	150	1	multiple	multiple	ADJ
alkej-870	150	2	microphones	microphone	NOUN
alkej-870	150	3	can	can	AUX
alkej-870	150	4	also	also	ADV
alkej-870	150	5	be	be	AUX
alkej-870	150	6	used	use	VERB
alkej-870	150	7	by	by	ADP
alkej-870	150	8	the	the	DET
alkej-870	150	9	model	model	NOUN
alkej-870	150	10	to	to	PART
alkej-870	150	11	acquire	acquire	VERB
alkej-870	150	12	spatial	spatial	ADJ
alkej-870	150	13	information	information	NOUN
alkej-870	150	14	and	and	CCONJ
alkej-870	150	15	improve	improve	VERB
alkej-870	150	16	separation	separation	NOUN
alkej-870	150	17	accuracy	accuracy	NOUN
alkej-870	150	18	.	.	PUNCT
alkej-870	151	1	it	it	PRON
alkej-870	151	2	is	be	AUX
alkej-870	151	3	important	important	ADJ
alkej-870	151	4	to	to	PART
alkej-870	151	5	keep	keep	VERB
alkej-870	151	6	in	in	ADP
alkej-870	151	7	mind	mind	NOUN
alkej-870	151	8	,	,	PUNCT
alkej-870	151	9	nevertheless	nevertheless	ADV
alkej-870	151	10	,	,	PUNCT
alkej-870	151	11	that	that	SCONJ
alkej-870	151	12	this	this	DET
alkej-870	151	13	method	method	NOUN
alkej-870	151	14	might	might	AUX
alkej-870	151	15	need	need	VERB
alkej-870	151	16	a	a	DET
alkej-870	151	17	significant	significant	ADJ
alkej-870	151	18	quantity	quantity	NOUN
alkej-870	151	19	of	of	ADP
alkej-870	151	20	training	training	NOUN
alkej-870	151	21	data	datum	NOUN
alkej-870	151	22	and	and	CCONJ
alkej-870	151	23	that	that	SCONJ
alkej-870	151	24	using	use	VERB
alkej-870	151	25	several	several	ADJ
alkej-870	151	26	modalities	modality	NOUN
alkej-870	151	27	and	and	CCONJ
alkej-870	151	28	channels	channel	NOUN
alkej-870	151	29	could	could	AUX
alkej-870	151	30	result	result	VERB
alkej-870	151	31	in	in	ADP
alkej-870	151	32	higher	high	ADJ
alkej-870	151	33	computational	computational	ADJ
alkej-870	151	34	complexity	complexity	NOUN
alkej-870	151	35	.	.	PUNCT
alkej-870	152	1	gan	gan	VERB
alkej-870	152	2	et	et	PROPN
alkej-870	152	3	al	al	PROPN
alkej-870	152	4	.	.	PUNCT
alkej-870	153	1	[	[	X
alkej-870	153	2	29	29	NUM
alkej-870	153	3	]	]	PUNCT
alkej-870	153	4	proposed	propose	VERB
alkej-870	153	5	an	an	DET
alkej-870	153	6	audio	audio	ADJ
alkej-870	153	7	-	-	ADJ
alkej-870	153	8	visual	visual	ADJ
alkej-870	153	9	source	source	NOUN
alkej-870	153	10	separation	separation	NOUN
alkej-870	153	11	approach	approach	NOUN
alkej-870	153	12	using	use	VERB
alkej-870	153	13	a	a	DET
alkej-870	153	14	deep	deep	ADJ
alkej-870	153	15	neural	neural	ADJ
alkej-870	153	16	network	network	NOUN
alkej-870	153	17	that	that	PRON
alkej-870	153	18	leverages	leverage	VERB
alkej-870	153	19	a	a	DET
alkej-870	153	20	musician	musician	NOUN
alkej-870	153	21	's	's	PART
alkej-870	153	22	body	body	NOUN
alkej-870	153	23	movements	movement	NOUN
alkej-870	153	24	to	to	ADP
alkej-870	153	25	separate	separate	ADJ
alkej-870	153	26	instruments	instrument	NOUN
alkej-870	153	27	.	.	PUNCT
alkej-870	154	1	two	two	NUM
alkej-870	154	2	-	-	PUNCT
alkej-870	154	3	part	part	NOUN
alkej-870	154	4	model	model	NOUN
alkej-870	154	5	:	:	PUNCT
alkej-870	154	6	video	video	NOUN
alkej-870	154	7	analysis	analysis	NOUN
alkej-870	154	8	network	network	NOUN
alkej-870	154	9	and	and	CCONJ
alkej-870	154	10	visual	visual	ADJ
alkej-870	154	11	-	-	PUNCT
alkej-870	154	12	audio	audio	NOUN
alkej-870	154	13	separation	separation	NOUN
alkej-870	154	14	network	network	NOUN
alkej-870	154	15	.	.	PUNCT
alkej-870	155	1	video	video	NOUN
alkej-870	155	2	analysis	analysis	NOUN
alkej-870	155	3	extracts	extract	NOUN
alkej-870	155	4	keypoint	keypoint	VERB
alkej-870	155	5	coordinates	coordinate	NOUN
alkej-870	155	6	and	and	CCONJ
alkej-870	155	7	global	global	ADJ
alkej-870	155	8	context	context	NOUN
alkej-870	155	9	using	use	VERB
alkej-870	155	10	contextaware	contextaware	NOUN
alkej-870	155	11	graph	graph	NOUN
alkej-870	155	12	cnn	cnn	PROPN
alkej-870	155	13	to	to	PART
alkej-870	155	14	generate	generate	VERB
alkej-870	155	15	latent	latent	NOUN
alkej-870	155	16	representation	representation	NOUN
alkej-870	155	17	.	.	PUNCT
alkej-870	156	1	visual	visual	ADJ
alkej-870	156	2	-	-	PUNCT
alkej-870	156	3	audio	audio	NOUN
alkej-870	156	4	separation	separation	NOUN
alkej-870	156	5	uses	use	VERB
alkej-870	156	6	visual	visual	ADJ
alkej-870	156	7	features	feature	NOUN
alkej-870	156	8	to	to	PART
alkej-870	156	9	separate	separate	VERB
alkej-870	156	10	audio	audio	NOUN
alkej-870	156	11	.	.	PUNCT
alkej-870	156	12	does	do	AUX
alkej-870	156	13	not	not	PART
alkej-870	156	14	require	require	VERB
alkej-870	156	15	prior	prior	ADJ
alkej-870	156	16	knowledge	knowledge	NOUN
alkej-870	156	17	or	or	CCONJ
alkej-870	156	18	training	training	NOUN
alkej-870	156	19	data	datum	NOUN
alkej-870	156	20	,	,	PUNCT
alkej-870	156	21	but	but	CCONJ
alkej-870	156	22	assumes	assume	VERB
alkej-870	156	23	each	each	DET
alkej-870	156	24	instrument	instrument	NOUN
alkej-870	156	25	is	be	AUX
alkej-870	156	26	associated	associate	VERB
alkej-870	156	27	with	with	ADP
alkej-870	156	28	specific	specific	ADJ
alkej-870	156	29	body	body	NOUN
alkej-870	156	30	movements	movement	NOUN
alkej-870	156	31	.	.	PUNCT
alkej-870	157	1	the	the	DET
alkej-870	157	2	method	method	NOUN
alkej-870	157	3	requires	require	VERB
alkej-870	157	4	a	a	DET
alkej-870	157	5	motion	motion	NOUN
alkej-870	157	6	capture	capture	NOUN
alkej-870	157	7	system	system	NOUN
alkej-870	157	8	,	,	PUNCT
alkej-870	157	9	but	but	CCONJ
alkej-870	157	10	it	it	PRON
alkej-870	157	11	is	be	AUX
alkej-870	157	12	not	not	PART
alkej-870	157	13	practical	practical	ADJ
alkej-870	157	14	in	in	ADP
alkej-870	157	15	real	real	ADJ
alkej-870	157	16	-	-	PUNCT
alkej-870	157	17	world	world	NOUN
alkej-870	157	18	scenarios	scenario	NOUN
alkej-870	157	19	.	.	PUNCT
alkej-870	158	1	zhu	zhu	PROPN
alkej-870	158	2	et	et	PROPN
alkej-870	158	3	al	al	PROPN
alkej-870	158	4	.	.	PUNCT
alkej-870	159	1	[	[	X
alkej-870	159	2	30	30	NUM
alkej-870	159	3	]	]	PUNCT
alkej-870	159	4	proposed	propose	VERB
alkej-870	159	5	an	an	DET
alkej-870	159	6	audio	audio	ADJ
alkej-870	159	7	-	-	ADJ
alkej-870	159	8	visual	visual	ADJ
alkej-870	159	9	source	source	NOUN
alkej-870	159	10	separation	separation	NOUN
alkej-870	159	11	method	method	NOUN
alkej-870	159	12	consisting	consist	VERB
alkej-870	159	13	of	of	ADP
alkej-870	159	14	two	two	NUM
alkej-870	159	15	stages	stage	NOUN
alkej-870	159	16	:	:	PUNCT
alkej-870	159	17	visual	visual	ADJ
alkej-870	159	18	feature	feature	NOUN
alkej-870	159	19	extraction	extraction	NOUN
alkej-870	159	20	and	and	CCONJ
alkej-870	159	21	sound	sound	NOUN
alkej-870	159	22	source	source	NOUN
alkej-870	159	23	separation	separation	NOUN
alkej-870	159	24	.	.	PUNCT
alkej-870	160	1	a	a	DET
alkej-870	160	2	pre	pre	ADJ
alkej-870	160	3	-	-	ADJ
alkej-870	160	4	trained	train	VERB
alkej-870	160	5	cnn	cnn	PROPN
alkej-870	160	6	extracts	extract	NOUN
alkej-870	160	7	visual	visual	ADJ
alkej-870	160	8	features	feature	NOUN
alkej-870	160	9	from	from	ADP
alkej-870	160	10	the	the	DET
alkej-870	160	11	input	input	NOUN
alkej-870	160	12	video	video	NOUN
alkej-870	160	13	frames	frame	NOUN
alkej-870	160	14	,	,	PUNCT
alkej-870	160	15	which	which	PRON
alkej-870	160	16	are	be	AUX
alkej-870	160	17	used	use	VERB
alkej-870	160	18	to	to	PART
alkej-870	160	19	compute	compute	VERB
alkej-870	160	20	attention	attention	NOUN
alkej-870	160	21	maps	map	NOUN
alkej-870	160	22	for	for	ADP
alkej-870	160	23	each	each	DET
alkej-870	160	24	sound	sound	ADJ
alkej-870	160	25	source	source	NOUN
alkej-870	160	26	.	.	PUNCT
alkej-870	161	1	a	a	DET
alkej-870	161	2	cascaded	cascade	VERB
alkej-870	161	3	opponent	opponent	NOUN
alkej-870	161	4	filter	filter	NOUN
alkej-870	161	5	network	network	NOUN
alkej-870	161	6	(	(	PUNCT
alkej-870	161	7	cofn	cofn	NOUN
alkej-870	161	8	)	)	PUNCT
alkej-870	161	9	is	be	AUX
alkej-870	161	10	used	use	VERB
alkej-870	161	11	for	for	ADP
alkej-870	161	12	sound	sound	ADJ
alkej-870	161	13	source	source	NOUN
alkej-870	161	14	separation	separation	NOUN
alkej-870	161	15	,	,	PUNCT
alkej-870	161	16	with	with	ADP
alkej-870	161	17	attention	attention	NOUN
alkej-870	161	18	maps	map	NOUN
alkej-870	161	19	guiding	guide	VERB
alkej-870	161	20	the	the	DET
alkej-870	161	21	separation	separation	NOUN
alkej-870	161	22	process	process	NOUN
alkej-870	161	23	.	.	PUNCT
alkej-870	162	1	the	the	DET
alkej-870	162	2	approach	approach	NOUN
alkej-870	162	3	improves	improve	VERB
alkej-870	162	4	separation	separation	NOUN
alkej-870	162	5	performance	performance	NOUN
alkej-870	162	6	,	,	PUNCT
alkej-870	162	7	but	but	CCONJ
alkej-870	162	8	requires	require	VERB
alkej-870	162	9	a	a	DET
alkej-870	162	10	pre	pre	ADJ
alkej-870	162	11	-	-	ADJ
alkej-870	162	12	trained	train	VERB
alkej-870	162	13	cnn	cnn	NOUN
alkej-870	162	14	and	and	CCONJ
alkej-870	162	15	significant	significant	ADJ
alkej-870	162	16	computational	computational	ADJ
alkej-870	162	17	resources	resource	NOUN
alkej-870	162	18	for	for	ADP
alkej-870	162	19	training	training	NOUN
alkej-870	162	20	.	.	PUNCT
alkej-870	163	1	tan	tan	INTJ
alkej-870	163	2	et	et	PROPN
alkej-870	163	3	al	al	PROPN
alkej-870	163	4	.	.	PUNCT
alkej-870	164	1	[	[	X
alkej-870	164	2	31	31	NUM
alkej-870	164	3	]	]	PUNCT
alkej-870	164	4	proposed	propose	VERB
alkej-870	164	5	a	a	DET
alkej-870	164	6	two	two	NUM
alkej-870	164	7	-	-	PUNCT
alkej-870	164	8	stage	stage	NOUN
alkej-870	164	9	strategy	strategy	NOUN
alkej-870	164	10	for	for	ADP
alkej-870	164	11	speech	speech	NOUN
alkej-870	164	12	enhancement	enhancement	NOUN
alkej-870	164	13	that	that	PRON
alkej-870	164	14	included	include	VERB
alkej-870	164	15	a	a	DET
alkej-870	164	16	separation	separation	NOUN
alkej-870	164	17	module	module	NOUN
alkej-870	164	18	for	for	ADP
alkej-870	164	19	noise	noise	NOUN
alkej-870	164	20	reduction	reduction	NOUN
alkej-870	164	21	and	and	CCONJ
alkej-870	164	22	a	a	DET
alkej-870	164	23	dereverberation	dereverberation	NOUN
alkej-870	164	24	module	module	NOUN
alkej-870	164	25	for	for	ADP
alkej-870	164	26	room	room	NOUN
alkej-870	164	27	reverberation	reverberation	NOUN
alkej-870	164	28	suppression	suppression	NOUN
alkej-870	164	29	.	.	PUNCT
alkej-870	165	1	a	a	DET
alkej-870	165	2	cnn	cnn	PROPN
alkej-870	165	3	is	be	AUX
alkej-870	165	4	used	use	VERB
alkej-870	165	5	in	in	ADP
alkej-870	165	6	the	the	DET
alkej-870	165	7	architecture	architecture	NOUN
alkej-870	165	8	for	for	ADP
alkej-870	165	9	audio	audio	ADJ
alkej-870	165	10	-	-	ADJ
alkej-870	165	11	visual	visual	ADJ
alkej-870	165	12	feature	feature	NOUN
alkej-870	165	13	extraction	extraction	NOUN
alkej-870	165	14	,	,	PUNCT
alkej-870	165	15	while	while	SCONJ
alkej-870	165	16	a	a	DET
alkej-870	165	17	blstm	blstm	NOUN
alkej-870	165	18	is	be	AUX
alkej-870	165	19	used	use	VERB
alkej-870	165	20	for	for	ADP
alkej-870	165	21	dereverberation	dereverberation	NOUN
alkej-870	165	22	.	.	PUNCT
alkej-870	166	1	the	the	DET
alkej-870	166	2	approach	approach	NOUN
alkej-870	166	3	has	have	VERB
alkej-870	166	4	the	the	DET
alkej-870	166	5	benefit	benefit	NOUN
alkej-870	166	6	of	of	ADP
alkej-870	166	7	being	be	AUX
alkej-870	166	8	able	able	ADJ
alkej-870	166	9	to	to	PART
alkej-870	166	10	handle	handle	VERB
alkej-870	166	11	weakly	weakly	ADJ
alkej-870	166	12	labeled	label	VERB
alkej-870	166	13	data	datum	NOUN
alkej-870	166	14	and	and	CCONJ
alkej-870	166	15	uses	use	VERB
alkej-870	166	16	a	a	DET
alkej-870	166	17	multimodal	multimodal	ADJ
alkej-870	166	18	fusion	fusion	NOUN
alkej-870	166	19	layer	layer	NOUN
alkej-870	166	20	for	for	ADP
alkej-870	166	21	feature	feature	NOUN
alkej-870	166	22	fusion	fusion	NOUN
alkej-870	166	23	.	.	PUNCT
alkej-870	167	1	however	however	ADV
alkej-870	167	2	,	,	PUNCT
alkej-870	167	3	accuracy	accuracy	NOUN
alkej-870	167	4	depends	depend	VERB
alkej-870	167	5	on	on	ADP
alkej-870	167	6	the	the	DET
alkej-870	167	7	strength	strength	NOUN
alkej-870	167	8	of	of	ADP
alkej-870	167	9	audio	audio	ADJ
alkej-870	167	10	-	-	ADJ
alkej-870	167	11	visual	visual	ADJ
alkej-870	167	12	connection	connection	NOUN
alkej-870	167	13	,	,	PUNCT
alkej-870	167	14	which	which	PRON
alkej-870	167	15	in	in	ADP
alkej-870	167	16	complicated	complicated	ADJ
alkej-870	167	17	scenarios	scenario	NOUN
alkej-870	167	18	may	may	AUX
alkej-870	167	19	not	not	PART
alkej-870	167	20	always	always	ADV
alkej-870	167	21	be	be	AUX
alkej-870	167	22	dependable	dependable	ADJ
alkej-870	167	23	.	.	PUNCT
alkej-870	168	1	qu	qu	PROPN
alkej-870	168	2	et	et	PROPN
alkej-870	168	3	al	al	PROPN
alkej-870	168	4	.	.	PUNCT
alkej-870	169	1	[	[	X
alkej-870	169	2	32	32	NUM
alkej-870	169	3	]	]	PUNCT
alkej-870	169	4	proposed	propose	VERB
alkej-870	169	5	a	a	DET
alkej-870	169	6	method	method	NOUN
alkej-870	169	7	for	for	ADP
alkej-870	169	8	improving	improve	VERB
alkej-870	169	9	speech	speech	NOUN
alkej-870	169	10	separation	separation	NOUN
alkej-870	169	11	that	that	PRON
alkej-870	169	12	combines	combine	VERB
alkej-870	169	13	auditory	auditory	ADJ
alkej-870	169	14	and	and	CCONJ
alkej-870	169	15	visual	visual	ADJ
alkej-870	169	16	modalities	modality	NOUN
alkej-870	169	17	.	.	PUNCT
alkej-870	170	1	a	a	DET
alkej-870	170	2	cnn	cnn	PROPN
alkej-870	170	3	is	be	AUX
alkej-870	170	4	used	use	VERB
alkej-870	170	5	to	to	PART
alkej-870	170	6	retrieve	retrieve	VERB
alkej-870	170	7	features	feature	NOUN
alkej-870	170	8	from	from	ADP
alkej-870	170	9	both	both	DET
alkej-870	170	10	modalities	modality	NOUN
alkej-870	170	11	,	,	PUNCT
alkej-870	170	12	which	which	PRON
alkej-870	170	13	are	be	AUX
alkej-870	170	14	subsequently	subsequently	ADV
alkej-870	170	15	integrated	integrate	VERB
alkej-870	170	16	into	into	ADP
alkej-870	170	17	a	a	DET
alkej-870	170	18	shared	share	VERB
alkej-870	170	19	embedding	embed	VERB
alkej-870	170	20	space	space	NOUN
alkej-870	170	21	with	with	ADP
alkej-870	170	22	the	the	DET
alkej-870	170	23	help	help	NOUN
alkej-870	170	24	of	of	ADP
alkej-870	170	25	a	a	DET
alkej-870	170	26	multi	multi	ADJ
alkej-870	170	27	-	-	ADJ
alkej-870	170	28	modal	modal	ADJ
alkej-870	170	29	fusion	fusion	NOUN
alkej-870	170	30	network	network	NOUN
alkej-870	170	31	.	.	PUNCT
alkej-870	171	1	then	then	ADV
alkej-870	171	2	,	,	PUNCT
alkej-870	171	3	using	use	VERB
alkej-870	171	4	the	the	DET
alkej-870	171	5	joint	joint	ADJ
alkej-870	171	6	embedding	embed	VERB
alkej-870	171	7	space	space	NOUN
alkej-870	171	8	as	as	ADP
alkej-870	171	9	a	a	DET
alkej-870	171	10	reference	reference	NOUN
alkej-870	171	11	,	,	PUNCT
alkej-870	171	12	a	a	DET
alkej-870	171	13	target	target	NOUN
alkej-870	171	14	speech	speech	NOUN
alkej-870	171	15	separation	separation	NOUN
alkej-870	171	16	network	network	NOUN
alkej-870	171	17	is	be	AUX
alkej-870	171	18	trained	train	VERB
alkej-870	171	19	to	to	PART
alkej-870	171	20	extract	extract	VERB
alkej-870	171	21	the	the	DET
alkej-870	171	22	target	target	NOUN
alkej-870	171	23	speaker	speaker	NOUN
alkej-870	171	24	's	's	PART
alkej-870	171	25	speech	speech	NOUN
alkej-870	171	26	from	from	ADP
alkej-870	171	27	the	the	DET
alkej-870	171	28	mixture	mixture	NOUN
alkej-870	171	29	.	.	PUNCT
alkej-870	172	1	the	the	DET
alkej-870	172	2	approach	approach	NOUN
alkej-870	172	3	is	be	AUX
alkej-870	172	4	adaptable	adaptable	ADJ
alkej-870	172	5	and	and	CCONJ
alkej-870	172	6	nooralhuda	nooralhuda	NOUN
alkej-870	172	7	mudhafar	mudhafar	PROPN
alkej-870	172	8	sleman	sleman	PROPN
alkej-870	172	9	al	al	PROPN
alkej-870	172	10	-	-	PUNCT
alkej-870	172	11	khwarizmi	khwarizmi	PROPN
alkej-870	172	12	engineering	engineering	NOUN
alkej-870	172	13	journal	journal	PROPN
alkej-870	172	14	,	,	PUNCT
alkej-870	172	15	vol	vol	NOUN
alkej-870	172	16	.	.	PROPN
alkej-870	172	17	19	19	NUM
alkej-870	172	18	,	,	PUNCT
alkej-870	172	19	no	no	INTJ
alkej-870	172	20	.	.	NOUN
alkej-870	172	21	4	4	NUM
alkej-870	172	22	,	,	PUNCT
alkej-870	172	23	p.p	p.p	PROPN
alkej-870	172	24	.	.	NOUN
alkej-870	172	25	4255	4255	NUM
alkej-870	172	26	(	(	PUNCT
alkej-870	172	27	2023	2023	NUM
alkej-870	172	28	)	)	PUNCT
alkej-870	172	29	47	47	NUM
alkej-870	172	30	may	may	AUX
alkej-870	172	31	be	be	AUX
alkej-870	172	32	applied	apply	VERB
alkej-870	172	33	with	with	ADP
alkej-870	172	34	a	a	DET
alkej-870	172	35	variety	variety	NOUN
alkej-870	172	36	of	of	ADP
alkej-870	172	37	input	input	NOUN
alkej-870	172	38	modalities	modality	NOUN
alkej-870	172	39	;	;	PUNCT
alkej-870	172	40	however	however	ADV
alkej-870	172	41	,	,	PUNCT
alkej-870	172	42	it	it	PRON
alkej-870	172	43	may	may	AUX
alkej-870	172	44	need	need	VERB
alkej-870	172	45	high	high	ADJ
alkej-870	172	46	-	-	PUNCT
alkej-870	172	47	quality	quality	NOUN
alkej-870	172	48	video	video	NOUN
alkej-870	172	49	input	input	NOUN
alkej-870	172	50	,	,	PUNCT
alkej-870	172	51	which	which	PRON
alkej-870	172	52	is	be	AUX
alkej-870	172	53	not	not	PART
alkej-870	172	54	always	always	ADV
alkej-870	172	55	feasible	feasible	ADJ
alkej-870	172	56	in	in	ADP
alkej-870	172	57	real	real	ADJ
alkej-870	172	58	-	-	PUNCT
alkej-870	172	59	world	world	NOUN
alkej-870	172	60	scenarios	scenario	NOUN
alkej-870	172	61	.	.	PUNCT
alkej-870	173	1	rahman	rahman	PROPN
alkej-870	173	2	et	et	PROPN
alkej-870	173	3	al	al	PROPN
alkej-870	173	4	.	.	PUNCT
alkej-870	174	1	[	[	X
alkej-870	174	2	33	33	NUM
alkej-870	174	3	]	]	PUNCT
alkej-870	174	4	proposed	propose	VERB
alkej-870	174	5	a	a	DET
alkej-870	174	6	weakly	weakly	ADV
alkej-870	174	7	supervised	supervised	ADJ
alkej-870	174	8	audio	audio	ADJ
alkej-870	174	9	-	-	ADJ
alkej-870	174	10	visual	visual	ADJ
alkej-870	174	11	sound	sound	NOUN
alkej-870	174	12	source	source	NOUN
alkej-870	174	13	detection	detection	NOUN
alkej-870	174	14	and	and	CCONJ
alkej-870	174	15	separation	separation	NOUN
alkej-870	174	16	method	method	NOUN
alkej-870	174	17	using	use	VERB
alkej-870	174	18	video	video	NOUN
alkej-870	174	19	-	-	PUNCT
alkej-870	174	20	level	level	NOUN
alkej-870	174	21	labels	label	NOUN
alkej-870	174	22	for	for	ADP
alkej-870	174	23	joint	joint	ADJ
alkej-870	174	24	learning	learning	NOUN
alkej-870	174	25	of	of	ADP
alkej-870	174	26	visual	visual	ADJ
alkej-870	174	27	and	and	CCONJ
alkej-870	174	28	auditory	auditory	ADJ
alkej-870	174	29	segments	segment	NOUN
alkej-870	174	30	.	.	PUNCT
alkej-870	175	1	it	it	PRON
alkej-870	175	2	consists	consist	VERB
alkej-870	175	3	of	of	ADP
alkej-870	175	4	a	a	DET
alkej-870	175	5	video	video	NOUN
alkej-870	175	6	frame	frame	NOUN
alkej-870	175	7	semantic	semantic	ADJ
alkej-870	175	8	segmentation	segmentation	NOUN
alkej-870	175	9	path	path	NOUN
alkej-870	175	10	and	and	CCONJ
alkej-870	175	11	a	a	DET
alkej-870	175	12	spectrogram	spectrogram	NOUN
alkej-870	175	13	mask	mask	NOUN
alkej-870	175	14	prediction	prediction	NOUN
alkej-870	175	15	path	path	NOUN
alkej-870	175	16	implemented	implement	VERB
alkej-870	175	17	using	use	VERB
alkej-870	175	18	an	an	DET
alkej-870	175	19	attention	attention	NOUN
alkej-870	175	20	u	u	ADJ
alkej-870	175	21	-	-	ADJ
alkej-870	175	22	net	net	ADJ
alkej-870	175	23	architecture	architecture	NOUN
alkej-870	175	24	.	.	PUNCT
alkej-870	176	1	the	the	DET
alkej-870	176	2	final	final	ADJ
alkej-870	176	3	mask	mask	NOUN
alkej-870	176	4	is	be	AUX
alkej-870	176	5	constructed	construct	VERB
alkej-870	176	6	using	use	VERB
alkej-870	176	7	multi	multi	ADJ
alkej-870	176	8	-	-	ADJ
alkej-870	176	9	modal	modal	ADJ
alkej-870	176	10	audio	audio	ADJ
alkej-870	176	11	-	-	ADJ
alkej-870	176	12	visual	visual	ADJ
alkej-870	176	13	features	feature	NOUN
alkej-870	176	14	.	.	PUNCT
alkej-870	177	1	it	it	PRON
alkej-870	177	2	can	can	AUX
alkej-870	177	3	handle	handle	VERB
alkej-870	177	4	multiple	multiple	ADJ
alkej-870	177	5	sound	sound	NOUN
alkej-870	177	6	sources	source	NOUN
alkej-870	177	7	and	and	CCONJ
alkej-870	177	8	does	do	AUX
alkej-870	177	9	not	not	PART
alkej-870	177	10	require	require	VERB
alkej-870	177	11	precise	precise	ADJ
alkej-870	177	12	temporal	temporal	ADJ
alkej-870	177	13	annotations	annotation	NOUN
alkej-870	177	14	.	.	PUNCT
alkej-870	178	1	it	it	PRON
alkej-870	178	2	has	have	VERB
alkej-870	178	3	a	a	DET
alkej-870	178	4	lower	low	ADJ
alkej-870	178	5	separation	separation	NOUN
alkej-870	178	6	performance	performance	NOUN
alkej-870	178	7	than	than	ADP
alkej-870	178	8	fully	fully	ADV
alkej-870	178	9	-	-	PUNCT
alkej-870	178	10	supervised	supervise	VERB
alkej-870	178	11	methods	method	NOUN
alkej-870	178	12	,	,	PUNCT
alkej-870	178	13	and	and	CCONJ
alkej-870	178	14	the	the	DET
alkej-870	178	15	quality	quality	NOUN
alkej-870	178	16	of	of	ADP
alkej-870	178	17	visual	visual	ADJ
alkej-870	178	18	features	feature	NOUN
alkej-870	178	19	can	can	AUX
alkej-870	178	20	affect	affect	VERB
alkej-870	178	21	the	the	DET
alkej-870	178	22	separation	separation	NOUN
alkej-870	178	23	performance	performance	NOUN
alkej-870	178	24	.	.	PUNCT
alkej-870	179	1	gao	gao	PROPN
alkej-870	179	2	et	et	PROPN
alkej-870	179	3	al	al	PROPN
alkej-870	179	4	.	.	PUNCT
alkej-870	180	1	[	[	X
alkej-870	180	2	34	34	NUM
alkej-870	180	3	]	]	PUNCT
alkej-870	180	4	proposed	propose	VERB
alkej-870	180	5	a	a	DET
alkej-870	180	6	deep	deep	ADJ
alkej-870	180	7	learning	learning	NOUN
alkej-870	180	8	approach	approach	NOUN
alkej-870	180	9	for	for	ADP
alkej-870	180	10	audio	audio	ADJ
alkej-870	180	11	-	-	ADJ
alkej-870	180	12	visual	visual	ADJ
alkej-870	180	13	speech	speech	NOUN
alkej-870	180	14	separation	separation	NOUN
alkej-870	180	15	.	.	PUNCT
alkej-870	181	1	it	it	PRON
alkej-870	181	2	consists	consist	VERB
alkej-870	181	3	of	of	ADP
alkej-870	181	4	two	two	NUM
alkej-870	181	5	networks	network	NOUN
alkej-870	181	6	:	:	PUNCT
alkej-870	181	7	an	an	DET
alkej-870	181	8	audio	audio	ADJ
alkej-870	181	9	network	network	NOUN
alkej-870	181	10	and	and	CCONJ
alkej-870	181	11	a	a	DET
alkej-870	181	12	visual	visual	ADJ
alkej-870	181	13	network	network	NOUN
alkej-870	181	14	,	,	PUNCT
alkej-870	181	15	both	both	PRON
alkej-870	181	16	are	be	AUX
alkej-870	181	17	cnns	cnns	ADJ
alkej-870	181	18	.	.	PUNCT
alkej-870	182	1	the	the	DET
alkej-870	182	2	networks	network	NOUN
alkej-870	182	3	are	be	AUX
alkej-870	182	4	jointly	jointly	ADV
alkej-870	182	5	trained	train	VERB
alkej-870	182	6	using	use	VERB
alkej-870	182	7	a	a	DET
alkej-870	182	8	cross	cross	ADJ
alkej-870	182	9	-	-	ADJ
alkej-870	182	10	modal	modal	ADJ
alkej-870	182	11	consistency	consistency	NOUN
alkej-870	182	12	loss	loss	NOUN
alkej-870	182	13	.	.	PUNCT
alkej-870	183	1	the	the	DET
alkej-870	183	2	approach	approach	NOUN
alkej-870	183	3	demonstrates	demonstrate	VERB
alkej-870	183	4	good	good	ADJ
alkej-870	183	5	performance	performance	NOUN
alkej-870	183	6	even	even	ADV
alkej-870	183	7	in	in	ADP
alkej-870	183	8	challenging	challenging	ADJ
alkej-870	183	9	scenarios	scenario	NOUN
alkej-870	183	10	but	but	CCONJ
alkej-870	183	11	requires	require	VERB
alkej-870	183	12	paired	pair	VERB
alkej-870	183	13	audio	audio	NOUN
alkej-870	183	14	and	and	CCONJ
alkej-870	183	15	video	video	NOUN
alkej-870	183	16	data	datum	NOUN
alkej-870	183	17	during	during	ADP
alkej-870	183	18	training	training	NOUN
alkej-870	183	19	.	.	PUNCT
alkej-870	184	1	however	however	ADV
alkej-870	184	2	,	,	PUNCT
alkej-870	184	3	its	its	PRON
alkej-870	184	4	performance	performance	NOUN
alkej-870	184	5	may	may	AUX
alkej-870	184	6	be	be	AUX
alkej-870	184	7	limited	limit	VERB
alkej-870	184	8	in	in	ADP
alkej-870	184	9	cases	case	NOUN
alkej-870	184	10	of	of	ADP
alkej-870	184	11	significant	significant	ADJ
alkej-870	184	12	overlap	overlap	NOUN
alkej-870	184	13	between	between	ADP
alkej-870	184	14	speakers	speaker	NOUN
alkej-870	184	15	or	or	CCONJ
alkej-870	184	16	when	when	SCONJ
alkej-870	184	17	the	the	DET
alkej-870	184	18	speakers	speaker	NOUN
alkej-870	184	19	are	be	AUX
alkej-870	184	20	not	not	PART
alkej-870	184	21	visible	visible	ADJ
alkej-870	184	22	in	in	ADP
alkej-870	184	23	the	the	DET
alkej-870	184	24	video	video	NOUN
alkej-870	184	25	.	.	PUNCT
alkej-870	185	1	majumder	majumder	NOUN
alkej-870	185	2	et	et	PROPN
alkej-870	185	3	al	al	PROPN
alkej-870	185	4	.	.	PUNCT
alkej-870	186	1	[	[	X
alkej-870	186	2	35	35	NUM
alkej-870	186	3	]	]	PUNCT
alkej-870	186	4	proposed	propose	VERB
alkej-870	186	5	an	an	DET
alkej-870	186	6	audio	audio	ADJ
alkej-870	186	7	-	-	ADJ
alkej-870	186	8	visual	visual	ADJ
alkej-870	186	9	source	source	NOUN
alkej-870	186	10	separation	separation	NOUN
alkej-870	186	11	approach	approach	NOUN
alkej-870	186	12	using	use	VERB
alkej-870	186	13	a	a	DET
alkej-870	186	14	robot	robot	NOUN
alkej-870	186	15	equipped	equip	VERB
alkej-870	186	16	with	with	ADP
alkej-870	186	17	a	a	DET
alkej-870	186	18	camera	camera	NOUN
alkej-870	186	19	and	and	CCONJ
alkej-870	186	20	microphones	microphone	NOUN
alkej-870	186	21	.	.	PUNCT
alkej-870	187	1	a	a	DET
alkej-870	187	2	deep	deep	ADJ
alkej-870	187	3	learning	learning	NOUN
alkej-870	187	4	model	model	NOUN
alkej-870	187	5	consisting	consist	VERB
alkej-870	187	6	of	of	ADP
alkej-870	187	7	cnn	cnn	PROPN
alkej-870	187	8	predicts	predict	VERB
alkej-870	187	9	each	each	DET
alkej-870	187	10	sound	sound	ADJ
alkej-870	187	11	source	source	NOUN
alkej-870	187	12	's	's	PART
alkej-870	187	13	arrival	arrival	NOUN
alkej-870	187	14	direction	direction	NOUN
alkej-870	187	15	,	,	PUNCT
alkej-870	187	16	and	and	CCONJ
alkej-870	187	17	the	the	DET
alkej-870	187	18	robot	robot	NOUN
alkej-870	187	19	moves	move	VERB
alkej-870	187	20	its	its	PRON
alkej-870	187	21	head	head	NOUN
alkej-870	187	22	to	to	PART
alkej-870	187	23	focus	focus	VERB
alkej-870	187	24	on	on	ADP
alkej-870	187	25	the	the	DET
alkej-870	187	26	desired	desire	VERB
alkej-870	187	27	sources	source	NOUN
alkej-870	187	28	.	.	PUNCT
alkej-870	188	1	audio	audio	ADJ
alkej-870	188	2	signals	signal	NOUN
alkej-870	188	3	are	be	AUX
alkej-870	188	4	then	then	ADV
alkej-870	188	5	separated	separate	VERB
alkej-870	188	6	using	use	VERB
alkej-870	188	7	a	a	DET
alkej-870	188	8	deep	deep	ADJ
alkej-870	188	9	neural	neural	ADJ
alkej-870	188	10	network	network	NOUN
alkej-870	188	11	.	.	PUNCT
alkej-870	189	1	the	the	DET
alkej-870	189	2	approach	approach	NOUN
alkej-870	189	3	allows	allow	VERB
alkej-870	189	4	for	for	ADP
alkej-870	189	5	real	real	ADJ
alkej-870	189	6	-	-	PUNCT
alkej-870	189	7	time	time	NOUN
alkej-870	189	8	audio	audio	ADJ
alkej-870	189	9	-	-	ADJ
alkej-870	189	10	visual	visual	ADJ
alkej-870	189	11	source	source	NOUN
alkej-870	189	12	separation	separation	NOUN
alkej-870	189	13	in	in	ADP
alkej-870	189	14	dynamic	dynamic	ADJ
alkej-870	189	15	environments	environment	NOUN
alkej-870	189	16	but	but	CCONJ
alkej-870	189	17	may	may	AUX
alkej-870	189	18	be	be	AUX
alkej-870	189	19	disruptive	disruptive	ADJ
alkej-870	189	20	in	in	ADP
alkej-870	189	21	quiet	quiet	ADJ
alkej-870	189	22	or	or	CCONJ
alkej-870	189	23	crowded	crowded	ADJ
alkej-870	189	24	spaces	space	NOUN
alkej-870	189	25	.	.	PUNCT
alkej-870	190	1	liu	liu	PROPN
alkej-870	190	2	et	et	PROPN
alkej-870	190	3	al	al	PROPN
alkej-870	191	1	[	[	X
alkej-870	191	2	36	36	NUM
alkej-870	191	3	]	]	PUNCT
alkej-870	191	4	proposed	propose	VERB
alkej-870	191	5	a	a	DET
alkej-870	191	6	two	two	NUM
alkej-870	191	7	-	-	PUNCT
alkej-870	191	8	stage	stage	NOUN
alkej-870	191	9	feature	feature	NOUN
alkej-870	191	10	fusion	fusion	NOUN
alkej-870	191	11	approach	approach	NOUN
alkej-870	191	12	for	for	ADP
alkej-870	191	13	audio	audio	ADJ
alkej-870	191	14	-	-	ADJ
alkej-870	191	15	visual	visual	ADJ
alkej-870	191	16	speech	speech	NOUN
alkej-870	191	17	separation	separation	NOUN
alkej-870	191	18	.	.	PUNCT
alkej-870	192	1	stage	stage	NOUN
alkej-870	192	2	one	one	PRON
alkej-870	192	3	uses	use	VERB
alkej-870	192	4	a	a	DET
alkej-870	192	5	cnn	cnn	NOUN
alkej-870	192	6	to	to	PART
alkej-870	192	7	extract	extract	VERB
alkej-870	192	8	features	feature	NOUN
alkej-870	192	9	from	from	ADP
alkej-870	192	10	both	both	DET
alkej-870	192	11	audio	audio	ADJ
alkej-870	192	12	and	and	CCONJ
alkej-870	192	13	visual	visual	ADJ
alkej-870	192	14	modalities	modality	NOUN
alkej-870	192	15	,	,	PUNCT
alkej-870	192	16	fused	fuse	VERB
alkej-870	192	17	using	use	VERB
alkej-870	192	18	a	a	DET
alkej-870	192	19	gated	gate	VERB
alkej-870	192	20	fusion	fusion	NOUN
alkej-870	192	21	mechanism	mechanism	NOUN
alkej-870	192	22	.	.	PUNCT
alkej-870	193	1	stage	stage	NOUN
alkej-870	193	2	two	two	NUM
alkej-870	193	3	uses	use	VERB
alkej-870	193	4	an	an	DET
alkej-870	193	5	rnn	rnn	NOUN
alkej-870	193	6	to	to	PART
alkej-870	193	7	separate	separate	VERB
alkej-870	193	8	mixed	mixed	ADJ
alkej-870	193	9	speech	speech	NOUN
alkej-870	193	10	signals	signal	NOUN
alkej-870	193	11	into	into	ADP
alkej-870	193	12	individual	individual	ADJ
alkej-870	193	13	sources	source	NOUN
alkej-870	193	14	.	.	PUNCT
alkej-870	194	1	advantages	advantage	NOUN
alkej-870	194	2	include	include	VERB
alkej-870	194	3	handling	handle	VERB
alkej-870	194	4	complex	complex	ADJ
alkej-870	194	5	scenes	scene	NOUN
alkej-870	194	6	and	and	CCONJ
alkej-870	194	7	improved	improved	ADJ
alkej-870	194	8	performance	performance	NOUN
alkej-870	194	9	.	.	PUNCT
alkej-870	195	1	disadvantages	disadvantage	NOUN
alkej-870	195	2	include	include	VERB
alkej-870	195	3	dependence	dependence	NOUN
alkej-870	195	4	on	on	ADP
alkej-870	195	5	synchronization	synchronization	NOUN
alkej-870	195	6	and	and	CCONJ
alkej-870	195	7	increased	increase	VERB
alkej-870	195	8	computational	computational	ADJ
alkej-870	195	9	complexity	complexity	NOUN
alkej-870	195	10	.	.	PUNCT
alkej-870	196	1	tian	tian	PROPN
alkej-870	196	2	et	et	PROPN
alkej-870	196	3	al	al	PROPN
alkej-870	196	4	.	.	PUNCT
alkej-870	197	1	[	[	X
alkej-870	197	2	37	37	NUM
alkej-870	197	3	]	]	PUNCT
alkej-870	197	4	proposed	propose	VERB
alkej-870	197	5	an	an	DET
alkej-870	197	6	audio	audio	ADJ
alkej-870	197	7	-	-	ADJ
alkej-870	197	8	visual	visual	ADJ
alkej-870	197	9	sound	sound	NOUN
alkej-870	197	10	source	source	NOUN
alkej-870	197	11	separation	separation	NOUN
alkej-870	197	12	approach	approach	NOUN
alkej-870	197	13	that	that	PRON
alkej-870	197	14	uses	use	VERB
alkej-870	197	15	visual	visual	ADJ
alkej-870	197	16	information	information	NOUN
alkej-870	197	17	to	to	ADP
alkej-870	197	18	separate	separate	ADJ
alkej-870	197	19	target	target	NOUN
alkej-870	197	20	sound	sound	NOUN
alkej-870	197	21	sources	source	NOUN
alkej-870	197	22	.	.	PUNCT
alkej-870	198	1	object	object	NOUN
alkej-870	198	2	candidates	candidate	NOUN
alkej-870	198	3	are	be	AUX
alkej-870	198	4	obtained	obtain	VERB
alkej-870	198	5	using	use	VERB
alkej-870	198	6	faster	fast	ADJ
alkej-870	198	7	rcnn	rcnn	NOUN
alkej-870	198	8	,	,	PUNCT
alkej-870	198	9	and	and	CCONJ
alkej-870	198	10	a	a	DET
alkej-870	198	11	cyclic	cyclic	ADJ
alkej-870	198	12	co	co	ADJ
alkej-870	198	13	-	-	ADJ
alkej-870	198	14	learning	learning	ADJ
alkej-870	198	15	framework	framework	NOUN
alkej-870	198	16	jointly	jointly	ADV
alkej-870	198	17	optimizes	optimize	VERB
alkej-870	198	18	sounding	sound	VERB
alkej-870	198	19	object	object	NOUN
alkej-870	198	20	visual	visual	ADJ
alkej-870	198	21	grounding	grounding	NOUN
alkej-870	198	22	(	(	PUNCT
alkej-870	198	23	sovg	sovg	PROPN
alkej-870	198	24	)	)	PUNCT
alkej-870	198	25	and	and	CCONJ
alkej-870	198	26	sound	sound	NOUN
alkej-870	198	27	separation	separation	NOUN
alkej-870	198	28	(	(	PUNCT
alkej-870	198	29	ss	ss	NOUN
alkej-870	198	30	)	)	PUNCT
alkej-870	198	31	tasks	task	NOUN
alkej-870	198	32	.	.	PUNCT
alkej-870	199	1	the	the	DET
alkej-870	199	2	sovg	sovg	PROPN
alkej-870	199	3	module	module	NOUN
alkej-870	199	4	uses	use	VERB
alkej-870	199	5	a	a	DET
alkej-870	199	6	pre	pre	ADJ
alkej-870	199	7	-	-	ADJ
alkej-870	199	8	trained	train	VERB
alkej-870	199	9	object	object	NOUN
alkej-870	199	10	detection	detection	NOUN
alkej-870	199	11	network	network	NOUN
alkej-870	199	12	to	to	PART
alkej-870	199	13	locate	locate	VERB
alkej-870	199	14	the	the	DET
alkej-870	199	15	target	target	NOUN
alkej-870	199	16	sounding	sound	VERB
alkej-870	199	17	object	object	NOUN
alkej-870	199	18	and	and	CCONJ
alkej-870	199	19	project	project	NOUN
alkej-870	199	20	visual	visual	ADJ
alkej-870	199	21	features	feature	VERB
alkej-870	199	22	onto	onto	ADP
alkej-870	199	23	a	a	DET
alkej-870	199	24	shared	share	VERB
alkej-870	199	25	embedding	embed	VERB
alkej-870	199	26	space	space	NOUN
alkej-870	199	27	with	with	ADP
alkej-870	199	28	audio	audio	ADJ
alkej-870	199	29	features	feature	NOUN
alkej-870	199	30	.	.	PUNCT
alkej-870	200	1	the	the	DET
alkej-870	200	2	ss	ss	PROPN
alkej-870	200	3	module	module	NOUN
alkej-870	200	4	separates	separate	VERB
alkej-870	200	5	sound	sound	ADJ
alkej-870	200	6	sources	source	NOUN
alkej-870	200	7	based	base	VERB
alkej-870	200	8	on	on	ADP
alkej-870	200	9	the	the	DET
alkej-870	200	10	shared	shared	ADJ
alkej-870	200	11	embedding	embed	VERB
alkej-870	200	12	space	space	NOUN
alkej-870	200	13	,	,	PUNCT
alkej-870	200	14	and	and	CCONJ
alkej-870	200	15	the	the	DET
alkej-870	200	16	estimated	estimate	VERB
alkej-870	200	17	source	source	NOUN
alkej-870	200	18	is	be	AUX
alkej-870	200	19	used	use	VERB
alkej-870	200	20	to	to	PART
alkej-870	200	21	update	update	VERB
alkej-870	200	22	the	the	DET
alkej-870	200	23	sovg	sovg	PROPN
alkej-870	200	24	module	module	NOUN
alkej-870	200	25	for	for	ADP
alkej-870	200	26	better	well	ADJ
alkej-870	200	27	visual	visual	ADJ
alkej-870	200	28	grounding	grounding	NOUN
alkej-870	200	29	.	.	PUNCT
alkej-870	201	1	the	the	DET
alkej-870	201	2	approach	approach	NOUN
alkej-870	201	3	can	can	AUX
alkej-870	201	4	handle	handle	VERB
alkej-870	201	5	multiple	multiple	ADJ
alkej-870	201	6	sound	sound	NOUN
alkej-870	201	7	sources	source	NOUN
alkej-870	201	8	but	but	CCONJ
alkej-870	201	9	may	may	AUX
alkej-870	201	10	be	be	AUX
alkej-870	201	11	limited	limit	VERB
alkej-870	201	12	by	by	ADP
alkej-870	201	13	the	the	DET
alkej-870	201	14	object	object	NOUN
alkej-870	201	15	detection	detection	NOUN
alkej-870	201	16	network	network	NOUN
alkej-870	201	17	's	's	PART
alkej-870	201	18	accuracy	accuracy	NOUN
alkej-870	201	19	and	and	CCONJ
alkej-870	201	20	visual	visual	ADJ
alkej-870	201	21	information	information	NOUN
alkej-870	201	22	quality	quality	NOUN
alkej-870	201	23	.	.	PUNCT
alkej-870	202	1	makishima	makishima	PROPN
alkej-870	202	2	et	et	PROPN
alkej-870	202	3	al	al	PROPN
alkej-870	202	4	.	.	PUNCT
alkej-870	203	1	[	[	X
alkej-870	203	2	38	38	NUM
alkej-870	203	3	]	]	PUNCT
alkej-870	203	4	proposed	propose	VERB
alkej-870	203	5	a	a	DET
alkej-870	203	6	deep	deep	ADJ
alkej-870	203	7	learning	learning	NOUN
alkej-870	203	8	method	method	NOUN
alkej-870	203	9	that	that	PRON
alkej-870	203	10	utilizes	utilize	VERB
alkej-870	203	11	both	both	DET
alkej-870	203	12	auditory	auditory	ADJ
alkej-870	203	13	and	and	CCONJ
alkej-870	203	14	visual	visual	ADJ
alkej-870	203	15	information	information	NOUN
alkej-870	203	16	to	to	ADP
alkej-870	203	17	separate	separate	ADJ
alkej-870	203	18	speech	speech	NOUN
alkej-870	203	19	signals	signal	NOUN
alkej-870	203	20	from	from	ADP
alkej-870	203	21	an	an	DET
alkej-870	203	22	audio	audio	ADJ
alkej-870	203	23	-	-	ADJ
alkej-870	203	24	visual	visual	ADJ
alkej-870	203	25	mixture	mixture	NOUN
alkej-870	203	26	.	.	PUNCT
alkej-870	204	1	it	it	PRON
alkej-870	204	2	consists	consist	VERB
alkej-870	204	3	of	of	ADP
alkej-870	204	4	two	two	NUM
alkej-870	204	5	subnetworks	subnetwork	NOUN
alkej-870	204	6	,	,	PUNCT
alkej-870	204	7	a	a	DET
alkej-870	204	8	visual	visual	ADJ
alkej-870	204	9	network	network	NOUN
alkej-870	204	10	and	and	CCONJ
alkej-870	204	11	an	an	DET
alkej-870	204	12	auditory	auditory	ADJ
alkej-870	204	13	network	network	NOUN
alkej-870	204	14	,	,	PUNCT
alkej-870	204	15	which	which	PRON
alkej-870	204	16	generate	generate	VERB
alkej-870	204	17	embeddings	embedding	NOUN
alkej-870	204	18	that	that	PRON
alkej-870	204	19	are	be	AUX
alkej-870	204	20	concatenated	concatenate	VERB
alkej-870	204	21	and	and	CCONJ
alkej-870	204	22	passed	pass	VERB
alkej-870	204	23	through	through	ADP
alkej-870	204	24	a	a	DET
alkej-870	204	25	decoder	decoder	NOUN
alkej-870	204	26	network	network	NOUN
alkej-870	204	27	.	.	PUNCT
alkej-870	205	1	the	the	DET
alkej-870	205	2	model	model	NOUN
alkej-870	205	3	introduces	introduce	VERB
alkej-870	205	4	a	a	DET
alkej-870	205	5	cross	cross	ADJ
alkej-870	205	6	-	-	ADJ
alkej-870	205	7	modal	modal	ADJ
alkej-870	205	8	correspondence	correspondence	NOUN
alkej-870	205	9	loss	loss	NOUN
alkej-870	205	10	function	function	NOUN
alkej-870	205	11	to	to	PART
alkej-870	205	12	align	align	VERB
alkej-870	205	13	auditory	auditory	ADJ
alkej-870	205	14	and	and	CCONJ
alkej-870	205	15	visual	visual	ADJ
alkej-870	205	16	information	information	NOUN
alkej-870	205	17	and	and	CCONJ
alkej-870	205	18	improve	improve	VERB
alkej-870	205	19	performance	performance	NOUN
alkej-870	205	20	.	.	PUNCT
alkej-870	206	1	one	one	NUM
alkej-870	206	2	disadvantage	disadvantage	NOUN
alkej-870	206	3	of	of	ADP
alkej-870	206	4	the	the	DET
alkej-870	206	5	approach	approach	NOUN
alkej-870	206	6	is	be	AUX
alkej-870	206	7	its	its	PRON
alkej-870	206	8	high	high	ADJ
alkej-870	206	9	dependency	dependency	NOUN
alkej-870	206	10	on	on	ADP
alkej-870	206	11	visual	visual	ADJ
alkej-870	206	12	data	datum	NOUN
alkej-870	206	13	,	,	PUNCT
alkej-870	206	14	making	make	VERB
alkej-870	206	15	it	it	PRON
alkej-870	206	16	unreliable	unreliable	ADJ
alkej-870	206	17	when	when	SCONJ
alkej-870	206	18	it	it	PRON
alkej-870	206	19	is	be	AUX
alkej-870	206	20	limited	limited	ADJ
alkej-870	206	21	.	.	PUNCT
alkej-870	207	1	nguyen	nguyen	NOUN
alkej-870	207	2	et	et	PROPN
alkej-870	207	3	al	al	PROPN
alkej-870	207	4	.	.	PUNCT
alkej-870	208	1	[	[	X
alkej-870	208	2	39	39	NUM
alkej-870	208	3	]	]	PUNCT
alkej-870	208	4	proposed	propose	VERB
alkej-870	208	5	an	an	DET
alkej-870	208	6	unsupervised	unsupervised	ADJ
alkej-870	208	7	technique	technique	NOUN
alkej-870	208	8	based	base	VERB
alkej-870	208	9	on	on	ADP
alkej-870	208	10	audio	audio	ADJ
alkej-870	208	11	-	-	ADJ
alkej-870	208	12	visual	visual	ADJ
alkej-870	208	13	generative	generative	ADJ
alkej-870	208	14	modeling	modeling	NOUN
alkej-870	208	15	for	for	ADP
alkej-870	208	16	the	the	DET
alkej-870	208	17	purpose	purpose	NOUN
alkej-870	208	18	of	of	ADP
alkej-870	208	19	speech	speech	NOUN
alkej-870	208	20	separation	separation	NOUN
alkej-870	208	21	.	.	PUNCT
alkej-870	209	1	they	they	PRON
alkej-870	209	2	used	use	VERB
alkej-870	209	3	a	a	DET
alkej-870	209	4	variational	variational	ADJ
alkej-870	209	5	auto	auto	NOUN
alkej-870	209	6	-	-	PUNCT
alkej-870	209	7	encoder	encoder	NOUN
alkej-870	209	8	(	(	PUNCT
alkej-870	209	9	vae	vae	PROPN
alkej-870	209	10	)	)	PUNCT
alkej-870	209	11	to	to	PART
alkej-870	209	12	learn	learn	VERB
alkej-870	209	13	a	a	DET
alkej-870	209	14	variable	variable	ADJ
alkej-870	209	15	generative	generative	ADJ
alkej-870	209	16	model	model	NOUN
alkej-870	209	17	from	from	ADP
alkej-870	209	18	clean	clean	ADJ
alkej-870	209	19	speech	speech	NOUN
alkej-870	209	20	.	.	PUNCT
alkej-870	210	1	the	the	DET
alkej-870	210	2	model	model	NOUN
alkej-870	210	3	uses	use	VERB
alkej-870	210	4	visual	visual	ADJ
alkej-870	210	5	information	information	NOUN
alkej-870	210	6	(	(	PUNCT
alkej-870	210	7	lip	lip	NOUN
alkej-870	210	8	movements	movement	NOUN
alkej-870	210	9	)	)	PUNCT
alkej-870	210	10	associated	associate	VERB
alkej-870	210	11	with	with	ADP
alkej-870	210	12	each	each	DET
alkej-870	210	13	speaker	speaker	NOUN
alkej-870	210	14	to	to	PART
alkej-870	210	15	separate	separate	VERB
alkej-870	210	16	the	the	DET
alkej-870	210	17	audio	audio	ADJ
alkej-870	210	18	streams	stream	NOUN
alkej-870	210	19	through	through	ADP
alkej-870	210	20	a	a	DET
alkej-870	210	21	visual	visual	ADJ
alkej-870	210	22	network	network	NOUN
alkej-870	210	23	.	.	PUNCT
alkej-870	211	1	the	the	DET
alkej-870	211	2	system	system	NOUN
alkej-870	211	3	does	do	AUX
alkej-870	211	4	not	not	PART
alkej-870	211	5	require	require	VERB
alkej-870	211	6	the	the	DET
alkej-870	211	7	clear	clear	ADJ
alkej-870	211	8	identification	identification	NOUN
alkej-870	211	9	of	of	ADP
alkej-870	211	10	acoustic	acoustic	ADJ
alkej-870	211	11	signals	signal	NOUN
alkej-870	211	12	,	,	PUNCT
alkej-870	211	13	but	but	CCONJ
alkej-870	211	14	its	its	PRON
alkej-870	211	15	performance	performance	NOUN
alkej-870	211	16	is	be	AUX
alkej-870	211	17	highly	highly	ADV
alkej-870	211	18	dependent	dependent	ADJ
alkej-870	211	19	on	on	ADP
alkej-870	211	20	visual	visual	ADJ
alkej-870	211	21	information	information	NOUN
alkej-870	211	22	and	and	CCONJ
alkej-870	211	23	has	have	VERB
alkej-870	211	24	a	a	DET
alkej-870	211	25	high	high	ADJ
alkej-870	211	26	computational	computational	ADJ
alkej-870	211	27	cost	cost	NOUN
alkej-870	211	28	.	.	PUNCT
alkej-870	212	1	lee	lee	PROPN
alkej-870	212	2	et	et	PROPN
alkej-870	212	3	al	al	PROPN
alkej-870	212	4	.	.	PUNCT
alkej-870	213	1	[	[	X
alkej-870	213	2	40	40	NUM
alkej-870	213	3	]	]	PUNCT
alkej-870	213	4	proposed	propose	VERB
alkej-870	213	5	a	a	DET
alkej-870	213	6	deep	deep	ADJ
alkej-870	213	7	learning	learning	NOUN
alkej-870	213	8	modality	modality	NOUN
alkej-870	213	9	that	that	PRON
alkej-870	213	10	uses	use	VERB
alkej-870	213	11	cnns	cnn	NOUN
alkej-870	213	12	to	to	PART
alkej-870	213	13	extract	extract	VERB
alkej-870	213	14	audio	audio	ADJ
alkej-870	213	15	and	and	CCONJ
alkej-870	213	16	visual	visual	ADJ
alkej-870	213	17	information	information	NOUN
alkej-870	213	18	.	.	PUNCT
alkej-870	214	1	the	the	DET
alkej-870	214	2	outputs	output	NOUN
alkej-870	214	3	of	of	ADP
alkej-870	214	4	cnns	cnn	NOUN
alkej-870	214	5	are	be	AUX
alkej-870	214	6	then	then	ADV
alkej-870	214	7	integrated	integrate	VERB
alkej-870	214	8	into	into	ADP
alkej-870	214	9	a	a	DET
alkej-870	214	10	joint	joint	ADJ
alkej-870	214	11	representation	representation	NOUN
alkej-870	214	12	by	by	ADP
alkej-870	214	13	using	use	VERB
alkej-870	214	14	a	a	DET
alkej-870	214	15	cross	cross	ADJ
alkej-870	214	16	-	-	ADJ
alkej-870	214	17	modal	modal	ADJ
alkej-870	214	18	affinity	affinity	NOUN
alkej-870	214	19	function	function	NOUN
alkej-870	214	20	.	.	PUNCT
alkej-870	215	1	the	the	DET
alkej-870	215	2	target	target	NOUN
alkej-870	215	3	speech	speech	NOUN
alkej-870	215	4	signal	signal	NOUN
alkej-870	215	5	is	be	AUX
alkej-870	215	6	reconstructed	reconstruct	VERB
alkej-870	215	7	using	use	VERB
alkej-870	215	8	the	the	DET
alkej-870	215	9	combined	combined	ADJ
alkej-870	215	10	representation	representation	NOUN
alkej-870	215	11	.	.	PUNCT
alkej-870	216	1	the	the	DET
alkej-870	216	2	approach	approach	NOUN
alkej-870	216	3	relies	rely	VERB
alkej-870	216	4	purely	purely	ADV
alkej-870	216	5	on	on	ADP
alkej-870	216	6	the	the	DET
alkej-870	216	7	connection	connection	NOUN
alkej-870	216	8	between	between	ADP
alkej-870	216	9	the	the	DET
alkej-870	216	10	audio	audio	ADJ
alkej-870	216	11	and	and	CCONJ
alkej-870	216	12	visual	visual	ADJ
alkej-870	216	13	signals	signal	NOUN
alkej-870	216	14	and	and	CCONJ
alkej-870	216	15	does	do	AUX
alkej-870	216	16	not	not	PART
alkej-870	216	17	require	require	VERB
alkej-870	216	18	prior	prior	ADJ
alkej-870	216	19	information	information	NOUN
alkej-870	216	20	of	of	ADP
alkej-870	216	21	the	the	DET
alkej-870	216	22	speakers	speaker	NOUN
alkej-870	216	23	'	'	PART
alkej-870	216	24	speech	speech	NOUN
alkej-870	216	25	patterns	pattern	NOUN
alkej-870	216	26	or	or	CCONJ
alkej-870	216	27	microphone	microphone	NOUN
alkej-870	216	28	placements	placement	NOUN
alkej-870	216	29	.	.	PUNCT
alkej-870	217	1	one	one	NUM
alkej-870	217	2	disadvantage	disadvantage	NOUN
alkej-870	217	3	of	of	ADP
alkej-870	217	4	the	the	DET
alkej-870	217	5	approach	approach	NOUN
alkej-870	217	6	is	be	AUX
alkej-870	217	7	its	its	PRON
alkej-870	217	8	high	high	ADJ
alkej-870	217	9	dependency	dependency	NOUN
alkej-870	217	10	on	on	ADP
alkej-870	217	11	the	the	DET
alkej-870	217	12	strong	strong	ADJ
alkej-870	217	13	coupling	coupling	NOUN
alkej-870	217	14	between	between	ADP
alkej-870	217	15	audio	audio	ADJ
alkej-870	217	16	and	and	CCONJ
alkej-870	217	17	visual	visual	ADJ
alkej-870	217	18	data	datum	NOUN
alkej-870	217	19	input	input	NOUN
alkej-870	217	20	.	.	PUNCT
alkej-870	218	1	zhu	zhu	PROPN
alkej-870	218	2	et	et	PROPN
alkej-870	218	3	al	al	PROPN
alkej-870	218	4	.	.	PUNCT
alkej-870	219	1	[	[	X
alkej-870	219	2	41	41	NUM
alkej-870	219	3	]	]	PUNCT
alkej-870	219	4	proposed	propose	VERB
alkej-870	219	5	an	an	DET
alkej-870	219	6	object	object	NOUN
alkej-870	219	7	categorybased	categorybase	VERB
alkej-870	219	8	technique	technique	NOUN
alkej-870	219	9	for	for	ADP
alkej-870	219	10	visual	visual	ADJ
alkej-870	219	11	sound	sound	NOUN
alkej-870	219	12	source	source	NOUN
alkej-870	219	13	separation	separation	NOUN
alkej-870	219	14	.	.	PUNCT
alkej-870	220	1	at	at	ADP
alkej-870	220	2	first	first	ADV
alkej-870	220	3	,	,	PUNCT
alkej-870	220	4	each	each	DET
alkej-870	220	5	pixel	pixel	NOUN
alkej-870	220	6	in	in	ADP
alkej-870	220	7	the	the	DET
alkej-870	220	8	video	video	NOUN
alkej-870	220	9	recording	recording	NOUN
alkej-870	220	10	is	be	AUX
alkej-870	220	11	categorized	categorize	VERB
alkej-870	220	12	into	into	ADP
alkej-870	220	13	a	a	DET
alkej-870	220	14	specific	specific	ADJ
alkej-870	220	15	item	item	NOUN
alkej-870	220	16	nooralhuda	nooralhuda	NOUN
alkej-870	220	17	mudhafar	mudhafar	PROPN
alkej-870	220	18	sleman	sleman	PROPN
alkej-870	220	19	al	al	PROPN
alkej-870	220	20	-	-	PUNCT
alkej-870	220	21	khwarizmi	khwarizmi	PROPN
alkej-870	220	22	engineering	engineering	NOUN
alkej-870	220	23	journal	journal	PROPN
alkej-870	220	24	,	,	PUNCT
alkej-870	220	25	vol	vol	NOUN
alkej-870	220	26	.	.	PROPN
alkej-870	220	27	19	19	NUM
alkej-870	220	28	,	,	PUNCT
alkej-870	220	29	no	no	INTJ
alkej-870	220	30	.	.	NOUN
alkej-870	220	31	4	4	NUM
alkej-870	220	32	,	,	PUNCT
alkej-870	220	33	p.p	p.p	PROPN
alkej-870	220	34	.	.	NOUN
alkej-870	220	35	4255	4255	NUM
alkej-870	220	36	(	(	PUNCT
alkej-870	220	37	2023	2023	NUM
alkej-870	220	38	)	)	PUNCT
alkej-870	220	39	48	48	NUM
alkej-870	220	40	category	category	NOUN
alkej-870	220	41	by	by	ADP
alkej-870	220	42	using	use	VERB
alkej-870	220	43	a	a	DET
alkej-870	220	44	pre	pre	ADJ
alkej-870	220	45	-	-	ADJ
alkej-870	220	46	trained	train	VERB
alkej-870	220	47	object	object	NOUN
alkej-870	220	48	identification	identification	NOUN
alkej-870	220	49	network	network	NOUN
alkej-870	220	50	.	.	PUNCT
alkej-870	221	1	the	the	DET
alkej-870	221	2	audio	audio	ADJ
alkej-870	221	3	stream	stream	NOUN
alkej-870	221	4	and	and	CCONJ
alkej-870	221	5	item	item	NOUN
alkej-870	221	6	category	category	NOUN
alkej-870	221	7	map	map	NOUN
alkej-870	221	8	are	be	AUX
alkej-870	221	9	then	then	ADV
alkej-870	221	10	sent	send	VERB
alkej-870	221	11	into	into	ADP
alkej-870	221	12	a	a	DET
alkej-870	221	13	neural	neural	ADJ
alkej-870	221	14	network	network	NOUN
alkej-870	221	15	,	,	PUNCT
alkej-870	221	16	which	which	PRON
alkej-870	221	17	is	be	AUX
alkej-870	221	18	trained	train	VERB
alkej-870	221	19	to	to	PART
alkej-870	221	20	perform	perform	VERB
alkej-870	221	21	the	the	DET
alkej-870	221	22	separation	separation	NOUN
alkej-870	221	23	.	.	PUNCT
alkej-870	222	1	the	the	DET
alkej-870	222	2	approach	approach	NOUN
alkej-870	222	3	can	can	AUX
alkej-870	222	4	increase	increase	VERB
alkej-870	222	5	separation	separation	NOUN
alkej-870	222	6	performance	performance	NOUN
alkej-870	222	7	by	by	ADP
alkej-870	222	8	utilizing	utilize	VERB
alkej-870	222	9	object	object	NOUN
alkej-870	222	10	category	category	NOUN
alkej-870	222	11	information	information	NOUN
alkej-870	222	12	but	but	CCONJ
alkej-870	222	13	it	it	PRON
alkej-870	222	14	depends	depend	VERB
alkej-870	222	15	on	on	ADP
alkej-870	222	16	its	its	PRON
alkej-870	222	17	availability	availability	NOUN
alkej-870	222	18	and	and	CCONJ
alkej-870	222	19	can	can	AUX
alkej-870	222	20	only	only	ADV
alkej-870	222	21	conduct	conduct	VERB
alkej-870	222	22	single	single	ADJ
alkej-870	222	23	-	-	PUNCT
alkej-870	222	24	frame	frame	NOUN
alkej-870	222	25	separation	separation	NOUN
alkej-870	222	26	without	without	ADP
alkej-870	222	27	taking	take	VERB
alkej-870	222	28	temporal	temporal	ADJ
alkej-870	222	29	information	information	NOUN
alkej-870	222	30	into	into	ADP
alkej-870	222	31	account	account	NOUN
alkej-870	222	32	.	.	PUNCT
alkej-870	223	1	gu	gu	NOUN
alkej-870	223	2	et	et	PROPN
alkej-870	223	3	al	al	PROPN
alkej-870	223	4	.	.	PUNCT
alkej-870	224	1	[	[	X
alkej-870	224	2	42	42	NUM
alkej-870	224	3	]	]	PUNCT
alkej-870	224	4	proposed	propose	VERB
alkej-870	224	5	a	a	DET
alkej-870	224	6	deep	deep	ADJ
alkej-870	224	7	learning	learning	NOUN
alkej-870	224	8	method	method	NOUN
alkej-870	224	9	to	to	PART
alkej-870	224	10	estimate	estimate	VERB
alkej-870	224	11	beamforming	beamforme	VERB
alkej-870	224	12	filters	filter	NOUN
alkej-870	224	13	for	for	ADP
alkej-870	224	14	time	time	NOUN
alkej-870	224	15	and	and	CCONJ
alkej-870	224	16	frequency	frequency	NOUN
alkej-870	224	17	domains	domain	NOUN
alkej-870	224	18	.	.	PUNCT
alkej-870	225	1	the	the	DET
alkej-870	225	2	method	method	NOUN
alkej-870	225	3	consists	consist	VERB
alkej-870	225	4	of	of	ADP
alkej-870	225	5	a	a	DET
alkej-870	225	6	frequency	frequency	NOUN
alkej-870	225	7	-	-	PUNCT
alkej-870	225	8	domain	domain	NOUN
alkej-870	225	9	beamforming	beamforming	NOUN
alkej-870	225	10	network	network	NOUN
alkej-870	225	11	and	and	CCONJ
alkej-870	225	12	a	a	DET
alkej-870	225	13	time	time	NOUN
alkej-870	225	14	-	-	PUNCT
alkej-870	225	15	domain	domain	NOUN
alkej-870	225	16	beamforming	beamforming	NOUN
alkej-870	225	17	network	network	NOUN
alkej-870	225	18	,	,	PUNCT
alkej-870	225	19	which	which	PRON
alkej-870	225	20	enhance	enhance	VERB
alkej-870	225	21	the	the	DET
alkej-870	225	22	target	target	NOUN
alkej-870	225	23	speech	speech	NOUN
alkej-870	225	24	and	and	CCONJ
alkej-870	225	25	suppress	suppress	VERB
alkej-870	225	26	interfering	interfere	VERB
alkej-870	225	27	sources	source	NOUN
alkej-870	225	28	.	.	PUNCT
alkej-870	226	1	the	the	DET
alkej-870	226	2	approach	approach	NOUN
alkej-870	226	3	is	be	AUX
alkej-870	226	4	more	more	ADV
alkej-870	226	5	flexible	flexible	ADJ
alkej-870	226	6	and	and	CCONJ
alkej-870	226	7	adaptable	adaptable	ADJ
alkej-870	226	8	to	to	ADP
alkej-870	226	9	different	different	ADJ
alkej-870	226	10	situations	situation	NOUN
alkej-870	226	11	than	than	ADP
alkej-870	226	12	traditional	traditional	ADJ
alkej-870	226	13	beamforming	beamforme	VERB
alkej-870	226	14	methods	method	NOUN
alkej-870	226	15	but	but	CCONJ
alkej-870	226	16	requires	require	VERB
alkej-870	226	17	a	a	DET
alkej-870	226	18	two	two	NUM
alkej-870	226	19	-	-	PUNCT
alkej-870	226	20	stage	stage	NOUN
alkej-870	226	21	approach	approach	NOUN
alkej-870	226	22	,	,	PUNCT
alkej-870	226	23	which	which	PRON
alkej-870	226	24	may	may	AUX
alkej-870	226	25	increase	increase	VERB
alkej-870	226	26	computational	computational	ADJ
alkej-870	226	27	complexity	complexity	NOUN
alkej-870	226	28	and	and	CCONJ
alkej-870	226	29	training	training	NOUN
alkej-870	226	30	time	time	NOUN
alkej-870	226	31	.	.	PUNCT
alkej-870	227	1	oya	oya	PROPN
alkej-870	227	2	et	et	PROPN
alkej-870	227	3	al	al	PROPN
alkej-870	227	4	.	.	PUNCT
alkej-870	228	1	[	[	X
alkej-870	228	2	43	43	NUM
alkej-870	228	3	]	]	PUNCT
alkej-870	228	4	proposed	propose	VERB
alkej-870	228	5	a	a	DET
alkej-870	228	6	method	method	NOUN
alkej-870	228	7	for	for	ADP
alkej-870	228	8	audiovisual	audiovisual	ADJ
alkej-870	228	9	source	source	NOUN
alkej-870	228	10	separation	separation	NOUN
alkej-870	228	11	that	that	PRON
alkej-870	228	12	uses	use	VERB
alkej-870	228	13	bounding	bounding	NOUN
alkej-870	228	14	boxes	box	NOUN
alkej-870	228	15	as	as	ADP
alkej-870	228	16	supervision	supervision	NOUN
alkej-870	228	17	.	.	PUNCT
alkej-870	229	1	the	the	DET
alkej-870	229	2	method	method	NOUN
alkej-870	229	3	has	have	VERB
alkej-870	229	4	two	two	NUM
alkej-870	229	5	stages	stage	NOUN
alkej-870	229	6	:	:	PUNCT
alkej-870	229	7	object	object	VERB
alkej-870	229	8	detection	detection	NOUN
alkej-870	229	9	to	to	PART
alkej-870	229	10	obtain	obtain	VERB
alkej-870	229	11	bounding	bounding	NOUN
alkej-870	229	12	boxes	box	NOUN
alkej-870	229	13	,	,	PUNCT
alkej-870	229	14	and	and	CCONJ
alkej-870	229	15	a	a	DET
alkej-870	229	16	neural	neural	ADJ
alkej-870	229	17	network	network	NOUN
alkej-870	229	18	for	for	ADP
alkej-870	229	19	separation	separation	NOUN
alkej-870	229	20	.	.	PUNCT
alkej-870	230	1	it	it	PRON
alkej-870	230	2	does	do	AUX
alkej-870	230	3	not	not	PART
alkej-870	230	4	require	require	VERB
alkej-870	230	5	manual	manual	ADJ
alkej-870	230	6	annotation	annotation	NOUN
alkej-870	230	7	but	but	CCONJ
alkej-870	230	8	relies	rely	VERB
alkej-870	230	9	on	on	ADP
alkej-870	230	10	the	the	DET
alkej-870	230	11	accuracy	accuracy	NOUN
alkej-870	230	12	of	of	ADP
alkej-870	230	13	the	the	DET
alkej-870	230	14	detection	detection	NOUN
alkej-870	230	15	model	model	NOUN
alkej-870	230	16	.	.	PUNCT
alkej-870	231	1	zhu	zhu	PROPN
alkej-870	231	2	at	at	ADP
alkej-870	231	3	al	al	PROPN
alkej-870	231	4	.	.	PUNCT
alkej-870	232	1	[	[	X
alkej-870	232	2	44	44	NUM
alkej-870	232	3	]	]	PUNCT
alkej-870	232	4	proposed	propose	VERB
alkej-870	232	5	a	a	DET
alkej-870	232	6	method	method	NOUN
alkej-870	232	7	for	for	ADP
alkej-870	232	8	audiovisual	audiovisual	ADJ
alkej-870	232	9	sound	sound	NOUN
alkej-870	232	10	source	source	NOUN
alkej-870	232	11	separation	separation	NOUN
alkej-870	232	12	and	and	CCONJ
alkej-870	232	13	localization	localization	NOUN
alkej-870	232	14	using	use	VERB
alkej-870	232	15	self	self	NOUN
alkej-870	232	16	-	-	PUNCT
alkej-870	232	17	supervised	supervise	VERB
alkej-870	232	18	motion	motion	NOUN
alkej-870	232	19	representations	representation	NOUN
alkej-870	232	20	.	.	PUNCT
alkej-870	233	1	learns	learn	VERB
alkej-870	233	2	motion	motion	NOUN
alkej-870	233	3	features	feature	NOUN
alkej-870	233	4	from	from	ADP
alkej-870	233	5	video	video	NOUN
alkej-870	233	6	data	datum	NOUN
alkej-870	233	7	without	without	ADP
alkej-870	233	8	explicit	explicit	ADJ
alkej-870	233	9	annotations	annotation	NOUN
alkej-870	233	10	and	and	CCONJ
alkej-870	233	11	incorporates	incorporate	VERB
alkej-870	233	12	them	they	PRON
alkej-870	233	13	into	into	ADP
alkej-870	233	14	a	a	DET
alkej-870	233	15	deep	deep	ADJ
alkej-870	233	16	neural	neural	ADJ
alkej-870	233	17	network	network	NOUN
alkej-870	233	18	for	for	ADP
alkej-870	233	19	separation	separation	NOUN
alkej-870	233	20	and	and	CCONJ
alkej-870	233	21	localization	localization	NOUN
alkej-870	233	22	.	.	PUNCT
alkej-870	234	1	the	the	DET
alkej-870	234	2	visual	visual	ADJ
alkej-870	234	3	motion	motion	NOUN
alkej-870	234	4	pattern	pattern	NOUN
alkej-870	234	5	of	of	ADP
alkej-870	234	6	sound	sound	ADJ
alkej-870	234	7	sources	source	NOUN
alkej-870	234	8	is	be	AUX
alkej-870	234	9	used	use	VERB
alkej-870	234	10	to	to	PART
alkej-870	234	11	estimate	estimate	VERB
alkej-870	234	12	location	location	NOUN
alkej-870	234	13	.	.	PUNCT
alkej-870	235	1	learns	learn	VERB
alkej-870	235	2	motion	motion	NOUN
alkej-870	235	3	features	feature	NOUN
alkej-870	235	4	with	with	ADP
alkej-870	235	5	cnn	cnn	PROPN
alkej-870	235	6	to	to	PART
alkej-870	235	7	predict	predict	VERB
alkej-870	235	8	future	future	ADJ
alkej-870	235	9	frames	frame	NOUN
alkej-870	235	10	from	from	ADP
alkej-870	235	11	past	past	ADJ
alkej-870	235	12	frames	frame	NOUN
alkej-870	235	13	.	.	PUNCT
alkej-870	236	1	it	it	PRON
alkej-870	236	2	is	be	AUX
alkej-870	236	3	scalable	scalable	ADJ
alkej-870	236	4	,	,	PUNCT
alkej-870	236	5	efficient	efficient	ADJ
alkej-870	236	6	,	,	PUNCT
alkej-870	236	7	and	and	CCONJ
alkej-870	236	8	eliminates	eliminate	VERB
alkej-870	236	9	the	the	DET
alkej-870	236	10	need	need	NOUN
alkej-870	236	11	for	for	ADP
alkej-870	236	12	manual	manual	ADJ
alkej-870	236	13	feature	feature	NOUN
alkej-870	236	14	engineering	engineering	NOUN
alkej-870	236	15	.	.	PUNCT
alkej-870	237	1	performance	performance	NOUN
alkej-870	237	2	depends	depend	VERB
alkej-870	237	3	on	on	ADP
alkej-870	237	4	video	video	NOUN
alkej-870	237	5	quality	quality	NOUN
alkej-870	237	6	and	and	CCONJ
alkej-870	237	7	resolution	resolution	NOUN
alkej-870	237	8	,	,	PUNCT
alkej-870	237	9	which	which	PRON
alkej-870	237	10	may	may	AUX
alkej-870	237	11	be	be	AUX
alkej-870	237	12	limited	limit	VERB
alkej-870	237	13	in	in	ADP
alkej-870	237	14	real	real	ADJ
alkej-870	237	15	-	-	PUNCT
alkej-870	237	16	world	world	NOUN
alkej-870	237	17	applications	application	NOUN
alkej-870	237	18	with	with	ADP
alkej-870	237	19	low	low	ADJ
alkej-870	237	20	-	-	PUNCT
alkej-870	237	21	resolution	resolution	NOUN
alkej-870	237	22	or	or	CCONJ
alkej-870	237	23	noisy	noisy	ADJ
alkej-870	237	24	videos	video	NOUN
alkej-870	237	25	.	.	PUNCT
alkej-870	238	1	pham	pham	PROPN
alkej-870	238	2	et	et	PROPN
alkej-870	238	3	al	al	PROPN
alkej-870	238	4	.	.	PUNCT
alkej-870	239	1	[	[	X
alkej-870	239	2	45	45	NUM
alkej-870	239	3	]	]	PUNCT
alkej-870	239	4	proposed	propose	VERB
alkej-870	239	5	a	a	DET
alkej-870	239	6	novel	novel	ADJ
alkej-870	239	7	approach	approach	NOUN
alkej-870	239	8	for	for	ADP
alkej-870	239	9	training	train	VERB
alkej-870	239	10	a	a	DET
alkej-870	239	11	cross	cross	ADJ
alkej-870	239	12	-	-	ADJ
alkej-870	239	13	modal	modal	ADJ
alkej-870	239	14	retrieval	retrieval	NOUN
alkej-870	239	15	framework	framework	NOUN
alkej-870	239	16	using	use	VERB
alkej-870	239	17	video	video	NOUN
alkej-870	239	18	data	datum	NOUN
alkej-870	239	19	.	.	PUNCT
alkej-870	240	1	the	the	DET
alkej-870	240	2	framework	framework	NOUN
alkej-870	240	3	is	be	AUX
alkej-870	240	4	refined	refine	VERB
alkej-870	240	5	through	through	ADP
alkej-870	240	6	three	three	NUM
alkej-870	240	7	loss	loss	NOUN
alkej-870	240	8	functions	function	NOUN
alkej-870	240	9	:	:	PUNCT
alkej-870	240	10	separation	separation	NOUN
alkej-870	240	11	loss	loss	NOUN
alkej-870	240	12	,	,	PUNCT
alkej-870	240	13	objectconsistency	objectconsistency	NOUN
alkej-870	240	14	loss	loss	NOUN
alkej-870	240	15	,	,	PUNCT
alkej-870	240	16	and	and	CCONJ
alkej-870	240	17	cross	cross	ADJ
alkej-870	240	18	-	-	ADJ
alkej-870	240	19	modal	modal	ADJ
alkej-870	240	20	loss	loss	NOUN
alkej-870	240	21	.	.	PUNCT
alkej-870	241	1	by	by	ADP
alkej-870	241	2	incorporating	incorporate	VERB
alkej-870	241	3	visual	visual	ADJ
alkej-870	241	4	guidance	guidance	NOUN
alkej-870	241	5	,	,	PUNCT
alkej-870	241	6	this	this	DET
alkej-870	241	7	method	method	NOUN
alkej-870	241	8	enhances	enhance	VERB
alkej-870	241	9	source	source	NOUN
alkej-870	241	10	separation	separation	NOUN
alkej-870	241	11	performance	performance	NOUN
alkej-870	241	12	,	,	PUNCT
alkej-870	241	13	especially	especially	ADV
alkej-870	241	14	in	in	ADP
alkej-870	241	15	situations	situation	NOUN
alkej-870	241	16	where	where	SCONJ
alkej-870	241	17	the	the	DET
alkej-870	241	18	audio	audio	ADJ
alkej-870	241	19	signal	signal	NOUN
alkej-870	241	20	is	be	AUX
alkej-870	241	21	degraded	degrade	VERB
alkej-870	241	22	or	or	CCONJ
alkej-870	241	23	the	the	DET
alkej-870	241	24	sources	source	NOUN
alkej-870	241	25	are	be	AUX
alkej-870	241	26	closely	closely	ADV
alkej-870	241	27	positioned	position	VERB
alkej-870	241	28	.	.	PUNCT
alkej-870	242	1	however	however	ADV
alkej-870	242	2	,	,	PUNCT
alkej-870	242	3	it	it	PRON
alkej-870	242	4	is	be	AUX
alkej-870	242	5	crucial	crucial	ADJ
alkej-870	242	6	to	to	PART
alkej-870	242	7	remember	remember	VERB
alkej-870	242	8	that	that	SCONJ
alkej-870	242	9	this	this	DET
alkej-870	242	10	method	method	NOUN
alkej-870	242	11	depends	depend	VERB
alkej-870	242	12	on	on	ADP
alkej-870	242	13	auditory	auditory	ADJ
alkej-870	242	14	and	and	CCONJ
alkej-870	242	15	visual	visual	ADJ
alkej-870	242	16	data	datum	NOUN
alkej-870	242	17	,	,	PUNCT
alkej-870	242	18	which	which	PRON
alkej-870	242	19	might	might	AUX
alkej-870	242	20	not	not	PART
alkej-870	242	21	always	always	ADV
alkej-870	242	22	be	be	AUX
alkej-870	242	23	available	available	ADJ
alkej-870	242	24	or	or	CCONJ
alkej-870	242	25	practical	practical	ADJ
alkej-870	242	26	in	in	ADP
alkej-870	242	27	real	real	ADJ
alkej-870	242	28	-	-	PUNCT
alkej-870	242	29	world	world	NOUN
alkej-870	242	30	situations	situation	NOUN
alkej-870	242	31	.	.	PUNCT
alkej-870	243	1	other	other	ADJ
alkej-870	243	2	authors	author	NOUN
alkej-870	243	3	created	create	VERB
alkej-870	243	4	a	a	DET
alkej-870	243	5	new	new	ADJ
alkej-870	243	6	network	network	NOUN
alkej-870	243	7	architecture	architecture	NOUN
alkej-870	243	8	for	for	ADP
alkej-870	243	9	audio	audio	ADJ
alkej-870	243	10	-	-	ADJ
alkej-870	243	11	visual	visual	ADJ
alkej-870	243	12	source	source	NOUN
alkej-870	243	13	separation	separation	NOUN
alkej-870	243	14	that	that	PRON
alkej-870	243	15	is	be	AUX
alkej-870	243	16	also	also	ADV
alkej-870	243	17	based	base	VERB
alkej-870	243	18	on	on	ADP
alkej-870	243	19	deep	deep	ADJ
alkej-870	243	20	learning	learning	NOUN
alkej-870	243	21	:	:	PUNCT
alkej-870	243	22	xu	xu	PROPN
alkej-870	243	23	et	et	PROPN
alkej-870	243	24	al	al	PROPN
alkej-870	243	25	.	.	PUNCT
alkej-870	244	1	[	[	X
alkej-870	244	2	46	46	NUM
alkej-870	244	3	]	]	PUNCT
alkej-870	244	4	proposed	propose	VERB
alkej-870	244	5	a	a	DET
alkej-870	244	6	method	method	NOUN
alkej-870	244	7	using	use	VERB
alkej-870	244	8	a	a	DET
alkej-870	244	9	neural	neural	ADJ
alkej-870	244	10	network	network	NOUN
alkej-870	244	11	called	call	VERB
alkej-870	244	12	minus	minus	PROPN
alkej-870	244	13	-	-	PUNCT
alkej-870	244	14	plus	plus	CCONJ
alkej-870	244	15	net	net	NOUN
alkej-870	244	16	(	(	PUNCT
alkej-870	244	17	mpn	mpn	NOUN
alkej-870	244	18	)	)	PUNCT
alkej-870	244	19	to	to	PART
alkej-870	244	20	separate	separate	VERB
alkej-870	244	21	sound	sound	NOUN
alkej-870	244	22	sources	source	NOUN
alkej-870	244	23	from	from	ADP
alkej-870	244	24	mixture	mixture	NOUN
alkej-870	244	25	signals	signal	NOUN
alkej-870	244	26	using	use	VERB
alkej-870	244	27	audio	audio	ADJ
alkej-870	244	28	and	and	CCONJ
alkej-870	244	29	visual	visual	ADJ
alkej-870	244	30	input	input	NOUN
alkej-870	244	31	.	.	PUNCT
alkej-870	245	1	the	the	DET
alkej-870	245	2	mpn	mpn	NOUN
alkej-870	245	3	recursively	recursively	ADV
alkej-870	245	4	separates	separate	VERB
alkej-870	245	5	the	the	DET
alkej-870	245	6	sources	source	NOUN
alkej-870	245	7	in	in	ADP
alkej-870	245	8	a	a	DET
alkej-870	245	9	coarse	coarse	NOUN
alkej-870	245	10	-	-	PUNCT
alkej-870	245	11	to	to	ADP
alkej-870	245	12	-	-	PUNCT
alkej-870	245	13	fine	fine	ADJ
alkej-870	245	14	manner	manner	NOUN
alkej-870	245	15	and	and	CCONJ
alkej-870	245	16	can	can	AUX
alkej-870	245	17	handle	handle	VERB
alkej-870	245	18	complex	complex	ADJ
alkej-870	245	19	mixtures	mixture	NOUN
alkej-870	245	20	of	of	ADP
alkej-870	245	21	sources	source	NOUN
alkej-870	245	22	.	.	PUNCT
alkej-870	246	1	the	the	DET
alkej-870	246	2	approach	approach	NOUN
alkej-870	246	3	has	have	VERB
alkej-870	246	4	advantages	advantage	NOUN
alkej-870	246	5	in	in	ADP
alkej-870	246	6	improved	improved	ADJ
alkej-870	246	7	performance	performance	NOUN
alkej-870	246	8	and	and	CCONJ
alkej-870	246	9	the	the	DET
alkej-870	246	10	separation	separation	NOUN
alkej-870	246	11	of	of	ADP
alkej-870	246	12	multiple	multiple	ADJ
alkej-870	246	13	sources	source	NOUN
alkej-870	246	14	,	,	PUNCT
alkej-870	246	15	but	but	CCONJ
alkej-870	246	16	relies	rely	VERB
alkej-870	246	17	on	on	ADP
alkej-870	246	18	pre	pre	ADJ
alkej-870	246	19	-	-	NOUN
alkej-870	246	20	segmentation	segmentation	NOUN
alkej-870	246	21	and	and	CCONJ
alkej-870	246	22	has	have	AUX
alkej-870	246	23	longer	long	ADV
alkej-870	246	24	processing	processing	NOUN
alkej-870	246	25	times	time	NOUN
alkej-870	246	26	.	.	PUNCT
alkej-870	247	1	zhao	zhao	PROPN
alkej-870	247	2	et	et	PROPN
alkej-870	247	3	al	al	PROPN
alkej-870	247	4	.	.	PUNCT
alkej-870	248	1	[	[	X
alkej-870	248	2	47	47	NUM
alkej-870	248	3	]	]	PUNCT
alkej-870	248	4	proposed	propose	VERB
alkej-870	248	5	a	a	DET
alkej-870	248	6	deep	deep	ADJ
alkej-870	248	7	learningbased	learningbase	VERB
alkej-870	248	8	method	method	NOUN
alkej-870	248	9	for	for	ADP
alkej-870	248	10	audio	audio	ADJ
alkej-870	248	11	-	-	ADJ
alkej-870	248	12	visual	visual	ADJ
alkej-870	248	13	sound	sound	NOUN
alkej-870	248	14	separation	separation	NOUN
alkej-870	248	15	using	use	VERB
alkej-870	248	16	visual	visual	ADJ
alkej-870	248	17	motion	motion	NOUN
alkej-870	248	18	features	feature	NOUN
alkej-870	248	19	.	.	PUNCT
alkej-870	249	1	the	the	DET
alkej-870	249	2	approach	approach	NOUN
alkej-870	249	3	involves	involve	VERB
alkej-870	249	4	two	two	NUM
alkej-870	249	5	stages	stage	NOUN
alkej-870	249	6	:	:	PUNCT
alkej-870	249	7	training	train	VERB
alkej-870	249	8	a	a	DET
alkej-870	249	9	motion	motion	NOUN
alkej-870	249	10	-	-	PUNCT
alkej-870	249	11	to	to	ADP
alkej-870	249	12	-	-	PUNCT
alkej-870	249	13	sound	sound	NOUN
alkej-870	249	14	network	network	NOUN
alkej-870	249	15	(	(	PUNCT
alkej-870	249	16	mtsn	mtsn	PROPN
alkej-870	249	17	)	)	PUNCT
alkej-870	249	18	to	to	PART
alkej-870	249	19	map	map	VERB
alkej-870	249	20	motion	motion	NOUN
alkej-870	249	21	features	feature	NOUN
alkej-870	249	22	to	to	PART
alkej-870	249	23	sound	sound	VERB
alkej-870	249	24	sources	source	NOUN
alkej-870	249	25	,	,	PUNCT
alkej-870	249	26	then	then	ADV
alkej-870	249	27	refining	refine	VERB
alkej-870	249	28	the	the	DET
alkej-870	249	29	estimated	estimate	VERB
alkej-870	249	30	sources	source	NOUN
alkej-870	249	31	using	use	VERB
alkej-870	249	32	a	a	DET
alkej-870	249	33	sound	sound	ADJ
alkej-870	249	34	refinement	refinement	NOUN
alkej-870	249	35	network	network	NOUN
alkej-870	249	36	(	(	PUNCT
alkej-870	249	37	srn	srn	PROPN
alkej-870	249	38	)	)	PUNCT
alkej-870	249	39	with	with	ADP
alkej-870	249	40	spectrograms	spectrogram	NOUN
alkej-870	249	41	and	and	CCONJ
alkej-870	249	42	original	original	ADJ
alkej-870	249	43	audio	audio	NOUN
alkej-870	249	44	.	.	PUNCT
alkej-870	250	1	the	the	DET
alkej-870	250	2	method	method	NOUN
alkej-870	250	3	is	be	AUX
alkej-870	250	4	flexible	flexible	ADJ
alkej-870	250	5	and	and	CCONJ
alkej-870	250	6	adaptable	adaptable	ADJ
alkej-870	250	7	but	but	CCONJ
alkej-870	250	8	may	may	AUX
alkej-870	250	9	not	not	PART
alkej-870	250	10	work	work	VERB
alkej-870	250	11	well	well	ADV
alkej-870	250	12	in	in	ADP
alkej-870	250	13	low	low	ADJ
alkej-870	250	14	-	-	PUNCT
alkej-870	250	15	light	light	ADJ
alkej-870	250	16	or	or	CCONJ
alkej-870	250	17	noisy	noisy	ADJ
alkej-870	250	18	environments	environment	NOUN
alkej-870	250	19	and	and	CCONJ
alkej-870	250	20	can	can	AUX
alkej-870	250	21	struggle	struggle	VERB
alkej-870	250	22	with	with	ADP
alkej-870	250	23	visually	visually	ADV
alkej-870	250	24	similar	similar	ADJ
alkej-870	250	25	or	or	CCONJ
alkej-870	250	26	closely	closely	ADV
alkej-870	250	27	located	locate	VERB
alkej-870	250	28	objects	object	NOUN
alkej-870	250	29	.	.	PUNCT
alkej-870	251	1	lu	lu	VERB
alkej-870	251	2	et	et	PROPN
alkej-870	251	3	al	al	PROPN
alkej-870	251	4	.	.	PUNCT
alkej-870	252	1	[	[	X
alkej-870	252	2	48	48	NUM
alkej-870	252	3	]	]	PUNCT
alkej-870	252	4	proposed	propose	VERB
alkej-870	252	5	an	an	DET
alkej-870	252	6	audio	audio	ADJ
alkej-870	252	7	-	-	ADJ
alkej-870	252	8	visual	visual	ADJ
alkej-870	252	9	method	method	NOUN
alkej-870	252	10	to	to	PART
alkej-870	252	11	separate	separate	VERB
alkej-870	252	12	speech	speech	NOUN
alkej-870	252	13	signals	signal	NOUN
alkej-870	252	14	from	from	ADP
alkej-870	252	15	a	a	DET
alkej-870	252	16	mixture	mixture	NOUN
alkej-870	252	17	of	of	ADP
alkej-870	252	18	multiple	multiple	ADJ
alkej-870	252	19	speakers	speaker	NOUN
alkej-870	252	20	using	use	VERB
alkej-870	252	21	a	a	DET
alkej-870	252	22	deep	deep	ADJ
alkej-870	252	23	neural	neural	ADJ
alkej-870	252	24	network	network	NOUN
alkej-870	252	25	architecture	architecture	NOUN
alkej-870	252	26	called	call	VERB
alkej-870	252	27	avdc	avdc	NOUN
alkej-870	252	28	.	.	PUNCT
alkej-870	253	1	the	the	DET
alkej-870	253	2	avdc	avdc	ADJ
alkej-870	253	3	network	network	NOUN
alkej-870	253	4	clusters	cluster	NOUN
alkej-870	253	5	audio	audio	ADJ
alkej-870	253	6	and	and	CCONJ
alkej-870	253	7	visual	visual	ADJ
alkej-870	253	8	features	feature	NOUN
alkej-870	253	9	using	use	VERB
alkej-870	253	10	convolutional	convolutional	ADJ
alkej-870	253	11	and	and	CCONJ
alkej-870	253	12	recurrent	recurrent	ADJ
alkej-870	253	13	layers	layer	NOUN
alkej-870	253	14	,	,	PUNCT
alkej-870	253	15	then	then	ADV
alkej-870	253	16	applies	apply	VERB
alkej-870	253	17	a	a	DET
alkej-870	253	18	clustering	clustering	ADJ
alkej-870	253	19	algorithm	algorithm	NOUN
alkej-870	253	20	to	to	ADP
alkej-870	253	21	separate	separate	ADJ
alkej-870	253	22	speech	speech	NOUN
alkej-870	253	23	signals	signal	NOUN
alkej-870	253	24	.	.	PUNCT
alkej-870	254	1	the	the	DET
alkej-870	254	2	method	method	NOUN
alkej-870	254	3	works	work	VERB
alkej-870	254	4	with	with	ADP
alkej-870	254	5	one	one	NUM
alkej-870	254	6	modality	modality	NOUN
alkej-870	254	7	and	and	CCONJ
alkej-870	254	8	is	be	AUX
alkej-870	254	9	computationally	computationally	ADV
alkej-870	254	10	expensive	expensive	ADJ
alkej-870	254	11	,	,	PUNCT
alkej-870	254	12	but	but	CCONJ
alkej-870	254	13	can	can	AUX
alkej-870	254	14	separate	separate	VERB
alkej-870	254	15	speech	speech	NOUN
alkej-870	254	16	signals	signal	NOUN
alkej-870	254	17	in	in	ADP
alkej-870	254	18	challenging	challenge	VERB
alkej-870	254	19	scenarios	scenario	NOUN
alkej-870	254	20	.	.	PUNCT
alkej-870	255	1	gao	gao	PROPN
alkej-870	255	2	et	et	PROPN
alkej-870	255	3	al	al	PROPN
alkej-870	255	4	.	.	PUNCT
alkej-870	256	1	[	[	X
alkej-870	256	2	49	49	NUM
alkej-870	256	3	]	]	PUNCT
alkej-870	256	4	proposed	propose	VERB
alkej-870	256	5	a	a	DET
alkej-870	256	6	method	method	NOUN
alkej-870	256	7	for	for	ADP
alkej-870	256	8	audiovisual	audiovisual	ADJ
alkej-870	256	9	source	source	NOUN
alkej-870	256	10	separation	separation	NOUN
alkej-870	256	11	using	use	VERB
alkej-870	256	12	a	a	DET
alkej-870	256	13	co	co	NOUN
alkej-870	256	14	-	-	NOUN
alkej-870	256	15	separation	separation	NOUN
alkej-870	256	16	network	network	NOUN
alkej-870	256	17	,	,	PUNCT
alkej-870	256	18	which	which	PRON
alkej-870	256	19	learns	learn	VERB
alkej-870	256	20	to	to	ADP
alkej-870	256	21	separate	separate	ADJ
alkej-870	256	22	sounds	sound	NOUN
alkej-870	256	23	from	from	ADP
alkej-870	256	24	different	different	ADJ
alkej-870	256	25	visual	visual	ADJ
alkej-870	256	26	objects	object	NOUN
alkej-870	256	27	by	by	ADP
alkej-870	256	28	jointly	jointly	ADV
alkej-870	256	29	modeling	model	VERB
alkej-870	256	30	audio	audio	ADJ
alkej-870	256	31	and	and	CCONJ
alkej-870	256	32	visual	visual	ADJ
alkej-870	256	33	features	feature	NOUN
alkej-870	256	34	.	.	PUNCT
alkej-870	257	1	the	the	DET
alkej-870	257	2	approach	approach	NOUN
alkej-870	257	3	is	be	AUX
alkej-870	257	4	real	real	ADJ
alkej-870	257	5	-	-	PUNCT
alkej-870	257	6	time	time	NOUN
alkej-870	257	7	and	and	CCONJ
alkej-870	257	8	generalizable	generalizable	ADJ
alkej-870	257	9	to	to	ADP
alkej-870	257	10	different	different	ADJ
alkej-870	257	11	scenarios	scenario	NOUN
alkej-870	257	12	without	without	ADP
alkej-870	257	13	prior	prior	ADJ
alkej-870	257	14	knowledge	knowledge	NOUN
alkej-870	257	15	of	of	ADP
alkej-870	257	16	sound	sound	ADJ
alkej-870	257	17	sources	source	NOUN
alkej-870	257	18	.	.	PUNCT
alkej-870	258	1	still	still	ADV
alkej-870	258	2	,	,	PUNCT
alkej-870	258	3	it	it	PRON
alkej-870	258	4	requires	require	VERB
alkej-870	258	5	synchronized	synchronize	VERB
alkej-870	258	6	audio	audio	ADJ
alkej-870	258	7	and	and	CCONJ
alkej-870	258	8	video	video	NOUN
alkej-870	258	9	inputs	input	NOUN
alkej-870	258	10	and	and	CCONJ
alkej-870	258	11	may	may	AUX
alkej-870	258	12	degrade	degrade	VERB
alkej-870	258	13	in	in	ADP
alkej-870	258	14	the	the	DET
alkej-870	258	15	presence	presence	NOUN
alkej-870	258	16	of	of	ADP
alkej-870	258	17	occlusions	occlusion	NOUN
alkej-870	258	18	or	or	CCONJ
alkej-870	258	19	incomplete	incomplete	ADJ
alkej-870	258	20	visual	visual	ADJ
alkej-870	258	21	information	information	NOUN
alkej-870	258	22	.	.	PUNCT
alkej-870	259	1	table	table	NOUN
alkej-870	259	2	2	2	NUM
alkej-870	259	3	illustrates	illustrate	VERB
alkej-870	259	4	deep	deep	ADJ
alkej-870	259	5	-	-	PUNCT
alkej-870	259	6	learning	learn	VERB
alkej-870	259	7	features	feature	NOUN
alkej-870	259	8	for	for	ADP
alkej-870	259	9	avss	avss	NOUN
alkej-870	259	10	in	in	ADP
alkej-870	259	11	terms	term	NOUN
alkej-870	259	12	of	of	ADP
alkej-870	259	13	complexity	complexity	NOUN
alkej-870	259	14	,	,	PUNCT
alkej-870	259	15	data	datum	NOUN
alkej-870	259	16	availability	availability	NOUN
alkej-870	259	17	,	,	PUNCT
alkej-870	259	18	and	and	CCONJ
alkej-870	259	19	other	other	ADJ
alkej-870	259	20	features	feature	NOUN
alkej-870	259	21	.	.	PUNCT
alkej-870	260	1	nooralhuda	nooralhuda	PROPN
alkej-870	260	2	mudhafar	mudhafar	PROPN
alkej-870	260	3	sleman	sleman	PROPN
alkej-870	260	4	al	al	PROPN
alkej-870	260	5	-	-	PUNCT
alkej-870	260	6	khwarizmi	khwarizmi	PROPN
alkej-870	260	7	engineering	engineering	NOUN
alkej-870	260	8	journal	journal	PROPN
alkej-870	260	9	,	,	PUNCT
alkej-870	260	10	vol	vol	NOUN
alkej-870	260	11	.	.	PROPN
alkej-870	260	12	19	19	NUM
alkej-870	260	13	,	,	PUNCT
alkej-870	260	14	no	no	INTJ
alkej-870	260	15	.	.	NOUN
alkej-870	260	16	4	4	NUM
alkej-870	260	17	,	,	PUNCT
alkej-870	260	18	p.p	p.p	PROPN
alkej-870	260	19	.	.	NOUN
alkej-870	260	20	4255	4255	NUM
alkej-870	260	21	(	(	PUNCT
alkej-870	260	22	2023	2023	NUM
alkej-870	260	23	)	)	PUNCT
alkej-870	260	24	49	49	NUM
alkej-870	260	25	table	table	NOUN
alkej-870	260	26	2	2	NUM
alkej-870	260	27	.	.	X
alkej-870	261	1	deep	deep	ADJ
alkej-870	261	2	learning	learning	NOUN
alkej-870	261	3	features	feature	NOUN
alkej-870	261	4	for	for	ADP
alkej-870	261	5	avss	avss	NOUN
alkej-870	261	6	feature	feature	NOUN
alkej-870	261	7	deep	deep	ADJ
alkej-870	261	8	learning	learning	NOUN
alkej-870	261	9	-	-	PUNCT
alkej-870	261	10	based	base	VERB
alkej-870	261	11	avss	avss	NOUN
alkej-870	261	12	type	type	NOUN
alkej-870	261	13	of	of	ADP
alkej-870	261	14	model	model	NOUN
alkej-870	261	15	neural	neural	ADJ
alkej-870	261	16	networks	network	NOUN
alkej-870	261	17	(	(	PUNCT
alkej-870	261	18	cnns	cnns	PROPN
alkej-870	261	19	,	,	PUNCT
alkej-870	261	20	rnns	rnn	NOUN
alkej-870	261	21	,	,	PUNCT
alkej-870	261	22	gans	gan	NOUN
alkej-870	261	23	)	)	PUNCT
alkej-870	261	24	training	training	NOUN
alkej-870	261	25	data	datum	NOUN
alkej-870	261	26	availability	availability	NOUN
alkej-870	261	27	require	require	VERB
alkej-870	261	28	large	large	ADJ
alkej-870	261	29	amounts	amount	NOUN
alkej-870	261	30	of	of	ADP
alkej-870	261	31	labeled	label	VERB
alkej-870	261	32	data	data	NOUN
alkej-870	261	33	model	model	NOUN
alkej-870	261	34	complexity	complexity	NOUN
alkej-870	261	35	complex	complex	ADJ
alkej-870	261	36	models	model	NOUN
alkej-870	261	37	with	with	ADP
alkej-870	261	38	high	high	ADJ
alkej-870	261	39	computational	computational	ADJ
alkej-870	261	40	requirements	requirement	NOUN
alkej-870	261	41	performance	performance	NOUN
alkej-870	261	42	can	can	AUX
alkej-870	261	43	achieve	achieve	VERB
alkej-870	261	44	state	state	NOUN
alkej-870	261	45	-	-	PUNCT
alkej-870	261	46	of	of	ADP
alkej-870	261	47	-	-	PUNCT
alkej-870	261	48	the	the	DET
alkej-870	261	49	-	-	PUNCT
alkej-870	261	50	art	art	NOUN
alkej-870	261	51	performance	performance	NOUN
alkej-870	261	52	on	on	ADP
alkej-870	261	53	many	many	ADJ
alkej-870	261	54	tasks	task	NOUN
alkej-870	261	55	robustness	robustness	NOUN
alkej-870	261	56	to	to	ADP
alkej-870	261	57	noise	noise	NOUN
alkej-870	261	58	can	can	AUX
alkej-870	261	59	handle	handle	VERB
alkej-870	261	60	complex	complex	ADJ
alkej-870	261	61	,	,	PUNCT
alkej-870	261	62	noisy	noisy	ADJ
alkej-870	261	63	audio	audio	ADJ
alkej-870	261	64	-	-	ADJ
alkej-870	261	65	visual	visual	ADJ
alkej-870	261	66	scenes	scene	NOUN
alkej-870	261	67	generalization	generalization	NOUN
alkej-870	261	68	can	can	AUX
alkej-870	261	69	generalize	generalize	VERB
alkej-870	261	70	well	well	ADV
alkej-870	261	71	to	to	ADP
alkej-870	261	72	unseen	unseen	ADJ
alkej-870	261	73	data	datum	NOUN
alkej-870	261	74	training	training	NOUN
alkej-870	261	75	time	time	NOUN
alkej-870	261	76	longer	long	ADV
alkej-870	261	77	training	train	VERB
alkej-870	261	78	times	time	NOUN
alkej-870	261	79	due	due	ADP
alkej-870	261	80	to	to	ADP
alkej-870	261	81	the	the	DET
alkej-870	261	82	complexity	complexity	NOUN
alkej-870	261	83	of	of	ADP
alkej-870	261	84	the	the	DET
alkej-870	261	85	models	model	NOUN
alkej-870	261	86	4	4	NUM
alkej-870	261	87	.	.	PUNCT
alkej-870	261	88	datasets	dataset	VERB
alkej-870	261	89	an	an	DET
alkej-870	261	90	audio	audio	ADJ
alkej-870	261	91	-	-	ADJ
alkej-870	261	92	visual	visual	ADJ
alkej-870	261	93	dataset	dataset	NOUN
alkej-870	261	94	is	be	AUX
alkej-870	261	95	a	a	DET
alkej-870	261	96	collection	collection	NOUN
alkej-870	261	97	of	of	ADP
alkej-870	261	98	data	datum	NOUN
alkej-870	261	99	that	that	PRON
alkej-870	261	100	includes	include	VERB
alkej-870	261	101	both	both	CCONJ
alkej-870	261	102	audio	audio	ADJ
alkej-870	261	103	and	and	CCONJ
alkej-870	261	104	visual	visual	ADJ
alkej-870	261	105	information	information	NOUN
alkej-870	261	106	.	.	PUNCT
alkej-870	262	1	this	this	DET
alkej-870	262	2	dataset	dataset	NOUN
alkej-870	262	3	type	type	NOUN
alkej-870	262	4	is	be	AUX
alkej-870	262	5	often	often	ADV
alkej-870	262	6	used	use	VERB
alkej-870	262	7	in	in	ADP
alkej-870	262	8	machine	machine	NOUN
alkej-870	262	9	learning	learning	NOUN
alkej-870	262	10	and	and	CCONJ
alkej-870	262	11	computer	computer	NOUN
alkej-870	262	12	vision	vision	NOUN
alkej-870	262	13	applications	application	NOUN
alkej-870	262	14	where	where	SCONJ
alkej-870	262	15	audio	audio	ADJ
alkej-870	262	16	and	and	CCONJ
alkej-870	262	17	visual	visual	ADJ
alkej-870	262	18	features	feature	NOUN
alkej-870	262	19	are	be	AUX
alkej-870	262	20	needed	need	VERB
alkej-870	262	21	to	to	PART
alkej-870	262	22	train	train	VERB
alkej-870	262	23	a	a	DET
alkej-870	262	24	model	model	NOUN
alkej-870	262	25	.	.	PUNCT
alkej-870	263	1	audio	audio	ADJ
alkej-870	263	2	-	-	ADJ
alkej-870	263	3	visual	visual	ADJ
alkej-870	263	4	datasets	dataset	NOUN
alkej-870	263	5	can	can	AUX
alkej-870	263	6	include	include	VERB
alkej-870	263	7	video	video	NOUN
alkej-870	263	8	recordings	recording	NOUN
alkej-870	263	9	with	with	ADP
alkej-870	263	10	accompanying	accompany	VERB
alkej-870	263	11	audio	audio	NOUN
alkej-870	263	12	tracks	track	NOUN
alkej-870	263	13	or	or	CCONJ
alkej-870	263	14	separate	separate	ADJ
alkej-870	263	15	audio	audio	ADJ
alkej-870	263	16	and	and	CCONJ
alkej-870	263	17	visual	visual	ADJ
alkej-870	263	18	files	file	NOUN
alkej-870	263	19	that	that	PRON
alkej-870	263	20	are	be	AUX
alkej-870	263	21	synchronized	synchronize	VERB
alkej-870	263	22	to	to	ADP
alkej-870	263	23	each	each	DET
alkej-870	263	24	other	other	ADJ
alkej-870	263	25	.	.	PUNCT
alkej-870	264	1	it	it	PRON
alkej-870	264	2	is	be	AUX
alkej-870	264	3	important	important	ADJ
alkej-870	264	4	to	to	PART
alkej-870	264	5	follow	follow	VERB
alkej-870	264	6	certain	certain	ADJ
alkej-870	264	7	techniques	technique	NOUN
alkej-870	264	8	and	and	CCONJ
alkej-870	264	9	protocols	protocol	NOUN
alkej-870	264	10	to	to	PART
alkej-870	264	11	create	create	VERB
alkej-870	264	12	a	a	DET
alkej-870	264	13	dataset	dataset	NOUN
alkej-870	264	14	for	for	ADP
alkej-870	264	15	audio	audio	ADJ
alkej-870	264	16	-	-	ADJ
alkej-870	264	17	visual	visual	ADJ
alkej-870	264	18	source	source	NOUN
alkej-870	264	19	separation	separation	NOUN
alkej-870	264	20	.	.	PUNCT
alkej-870	265	1	this	this	PRON
alkej-870	265	2	includes	include	VERB
alkej-870	265	3	gathering	gather	VERB
alkej-870	265	4	a	a	DET
alkej-870	265	5	diverse	diverse	ADJ
alkej-870	265	6	range	range	NOUN
alkej-870	265	7	of	of	ADP
alkej-870	265	8	data	datum	NOUN
alkej-870	265	9	,	,	PUNCT
alkej-870	265	10	accurately	accurately	ADV
alkej-870	265	11	labeling	label	VERB
alkej-870	265	12	the	the	DET
alkej-870	265	13	audio	audio	ADJ
alkej-870	265	14	and	and	CCONJ
alkej-870	265	15	visual	visual	ADJ
alkej-870	265	16	components	component	NOUN
alkej-870	265	17	,	,	PUNCT
alkej-870	265	18	preprocessing	preprocesse	VERB
alkej-870	265	19	the	the	DET
alkej-870	265	20	signals	signal	NOUN
alkej-870	265	21	,	,	PUNCT
alkej-870	265	22	dividing	divide	VERB
alkej-870	265	23	the	the	DET
alkej-870	265	24	dataset	dataset	NOUN
alkej-870	265	25	into	into	ADP
alkej-870	265	26	appropriate	appropriate	ADJ
alkej-870	265	27	subsets	subset	NOUN
alkej-870	265	28	,	,	PUNCT
alkej-870	265	29	maintaining	maintain	VERB
alkej-870	265	30	a	a	DET
alkej-870	265	31	balanced	balanced	ADJ
alkej-870	265	32	distribution	distribution	NOUN
alkej-870	265	33	of	of	ADP
alkej-870	265	34	samples	sample	NOUN
alkej-870	265	35	,	,	PUNCT
alkej-870	265	36	and	and	CCONJ
alkej-870	265	37	applying	apply	VERB
alkej-870	265	38	data	datum	NOUN
alkej-870	265	39	augmentation	augmentation	NOUN
alkej-870	265	40	techniques	technique	NOUN
alkej-870	265	41	.	.	PUNCT
alkej-870	266	1	implementing	implement	VERB
alkej-870	266	2	these	these	DET
alkej-870	266	3	steps	step	NOUN
alkej-870	266	4	makes	make	VERB
alkej-870	266	5	the	the	DET
alkej-870	266	6	dataset	dataset	NOUN
alkej-870	266	7	more	more	ADV
alkej-870	266	8	representative	representative	ADJ
alkej-870	266	9	,	,	PUNCT
alkej-870	266	10	reliable	reliable	ADJ
alkej-870	266	11	,	,	PUNCT
alkej-870	266	12	and	and	CCONJ
alkej-870	266	13	suitable	suitable	ADJ
alkej-870	266	14	for	for	ADP
alkej-870	266	15	training	training	NOUN
alkej-870	266	16	and	and	CCONJ
alkej-870	266	17	evaluating	evaluate	VERB
alkej-870	266	18	audio	audio	ADJ
alkej-870	266	19	-	-	ADJ
alkej-870	266	20	visual	visual	ADJ
alkej-870	266	21	source	source	NOUN
alkej-870	266	22	separation	separation	NOUN
alkej-870	266	23	models	model	NOUN
alkej-870	266	24	.	.	PUNCT
alkej-870	267	1	table	table	NOUN
alkej-870	267	2	3	3	NUM
alkej-870	267	3	shown	show	VERB
alkej-870	267	4	below	below	ADP
alkej-870	267	5	illustrates	illustrate	VERB
alkej-870	267	6	some	some	PRON
alkej-870	267	7	of	of	ADP
alkej-870	267	8	the	the	DET
alkej-870	267	9	commonly	commonly	ADV
alkej-870	267	10	known	know	VERB
alkej-870	267	11	datasets	dataset	NOUN
alkej-870	267	12	.	.	PUNCT
alkej-870	268	1	table	table	NOUN
alkej-870	268	2	3	3	NUM
alkej-870	268	3	.	.	PUNCT
alkej-870	268	4	audio	audio	ADJ
alkej-870	268	5	-	-	ADJ
alkej-870	268	6	visual	visual	ADJ
alkej-870	268	7	datasets	dataset	NOUN
alkej-870	268	8	dataset	dataset	VERB
alkej-870	268	9	description	description	NOUN
alkej-870	268	10	grid	grid	NOUN
alkej-870	268	11	audio	audio	ADJ
alkej-870	268	12	-	-	ADJ
alkej-870	268	13	visual	visual	ADJ
alkej-870	268	14	speech	speech	NOUN
alkej-870	268	15	recognition	recognition	NOUN
alkej-870	268	16	dataset	dataset	VERB
alkej-870	268	17	consisting	consist	VERB
alkej-870	268	18	of	of	ADP
alkej-870	268	19	33	33	NUM
alkej-870	268	20	speakers	speaker	NOUN
alkej-870	268	21	uttering	utter	VERB
alkej-870	268	22	1000	1000	NUM
alkej-870	268	23	phrases	phrase	NOUN
alkej-870	268	24	.	.	PUNCT
alkej-870	269	1	contains	contain	VERB
alkej-870	269	2	video	video	NOUN
alkej-870	269	3	,	,	PUNCT
alkej-870	269	4	audio	audio	NOUN
alkej-870	269	5	,	,	PUNCT
alkej-870	269	6	and	and	CCONJ
alkej-870	269	7	aligned	align	VERB
alkej-870	269	8	phonemic	phonemic	ADJ
alkej-870	269	9	transcriptions	transcription	NOUN
alkej-870	270	1	[	[	X
alkej-870	270	2	50	50	NUM
alkej-870	270	3	]	]	PUNCT
alkej-870	270	4	.	.	PUNCT
alkej-870	271	1	lombard	lombard	PROPN
alkej-870	271	2	grid	grid	PROPN
alkej-870	271	3	audiovisual	audiovisual	ADJ
alkej-870	271	4	lombard	lombard	PROPN
alkej-870	271	5	speech	speech	NOUN
alkej-870	271	6	corpus	corpus	NOUN
alkej-870	271	7	,	,	PUNCT
alkej-870	271	8	freely	freely	ADV
alkej-870	271	9	available	available	ADJ
alkej-870	271	10	for	for	ADP
alkej-870	271	11	download	download	NOUN
alkej-870	271	12	.	.	PUNCT
alkej-870	272	1	it	it	PRON
alkej-870	272	2	contains	contain	VERB
alkej-870	272	3	5400	5400	NUM
alkej-870	272	4	utterances	utterance	NOUN
alkej-870	272	5	(	(	PUNCT
alkej-870	272	6	2700	2700	NUM
alkej-870	272	7	lombard	lombard	NOUN
alkej-870	272	8	and	and	CCONJ
alkej-870	272	9	2700	2700	NUM
alkej-870	272	10	plain	plain	ADJ
alkej-870	272	11	reference	reference	NOUN
alkej-870	272	12	utterances	utterance	NOUN
alkej-870	272	13	)	)	PUNCT
alkej-870	272	14	,	,	PUNCT
alkej-870	272	15	produced	produce	VERB
alkej-870	272	16	by	by	ADP
alkej-870	272	17	54	54	NUM
alkej-870	272	18	talkers	talker	NOUN
alkej-870	272	19	[	[	X
alkej-870	272	20	51	51	NUM
alkej-870	272	21	]	]	PUNCT
alkej-870	272	22	.	.	PUNCT
alkej-870	273	1	tcd	tcd	PROPN
alkej-870	273	2	timit	timit	PROPN
alkej-870	273	3	designed	design	VERB
alkej-870	273	4	for	for	ADP
alkej-870	273	5	continuous	continuous	ADJ
alkej-870	273	6	audiovisual	audiovisual	ADJ
alkej-870	273	7	speech	speech	NOUN
alkej-870	273	8	recognition	recognition	NOUN
alkej-870	273	9	research	research	NOUN
alkej-870	273	10	.	.	PUNCT
alkej-870	274	1	consists	consist	VERB
alkej-870	274	2	of	of	ADP
alkej-870	274	3	high	high	ADJ
alkej-870	274	4	-	-	PUNCT
alkej-870	274	5	quality	quality	NOUN
alkej-870	274	6	audio	audio	NOUN
alkej-870	274	7	and	and	CCONJ
alkej-870	274	8	video	video	NOUN
alkej-870	274	9	footage	footage	NOUN
alkej-870	274	10	of	of	ADP
alkej-870	274	11	62	62	NUM
alkej-870	274	12	speakers	speaker	NOUN
alkej-870	274	13	reading	read	VERB
alkej-870	274	14	a	a	DET
alkej-870	274	15	total	total	NOUN
alkej-870	274	16	of	of	ADP
alkej-870	274	17	6913	6913	NUM
alkej-870	274	18	phonetically	phonetically	ADV
alkej-870	274	19	rich	rich	ADJ
alkej-870	274	20	sentences	sentence	NOUN
alkej-870	274	21	[	[	X
alkej-870	274	22	52	52	NUM
alkej-870	274	23	]	]	PUNCT
alkej-870	274	24	.	.	PUNCT
alkej-870	275	1	ouluvs	ouluvs	NOUN
alkej-870	275	2	the	the	DET
alkej-870	275	3	ouluvs	ouluvs	NOUN
alkej-870	275	4	database	database	NOUN
alkej-870	275	5	includes	include	VERB
alkej-870	275	6	video	video	NOUN
alkej-870	275	7	and	and	CCONJ
alkej-870	275	8	audio	audio	NOUN
alkej-870	275	9	data	datum	NOUN
alkej-870	275	10	of	of	ADP
alkej-870	275	11	20	20	NUM
alkej-870	275	12	subjects	subject	NOUN
alkej-870	275	13	consisting	consist	VERB
alkej-870	275	14	of	of	ADP
alkej-870	275	15	10	10	NUM
alkej-870	275	16	phrases	phrase	NOUN
alkej-870	275	17	[	[	X
alkej-870	275	18	53	53	NUM
alkej-870	275	19	]	]	PUNCT
alkej-870	275	20	.	.	PUNCT
alkej-870	276	1	ouluvs2	ouluvs2	NOUN
alkej-870	277	1	a	a	DET
alkej-870	277	2	multi	multi	ADJ
alkej-870	277	3	-	-	ADJ
alkej-870	277	4	view	view	ADJ
alkej-870	277	5	audiovisual	audiovisual	ADJ
alkej-870	277	6	database	database	NOUN
alkej-870	277	7	for	for	ADP
alkej-870	277	8	non	non	ADJ
alkej-870	277	9	-	-	ADJ
alkej-870	277	10	rigid	rigid	ADJ
alkej-870	277	11	mouth	mouth	NOUN
alkej-870	277	12	motion	motion	NOUN
alkej-870	277	13	analysis	analysis	NOUN
alkej-870	277	14	.	.	PUNCT
alkej-870	278	1	it	it	PRON
alkej-870	278	2	includes	include	VERB
alkej-870	278	3	more	more	ADJ
alkej-870	278	4	than	than	ADP
alkej-870	278	5	50	50	NUM
alkej-870	278	6	speakers	speaker	NOUN
alkej-870	278	7	uttering	utter	VERB
alkej-870	278	8	three	three	NUM
alkej-870	278	9	types	type	NOUN
alkej-870	278	10	of	of	ADP
alkej-870	278	11	utterances	utterance	NOUN
alkej-870	278	12	[	[	X
alkej-870	278	13	54	54	NUM
alkej-870	278	14	]	]	PUNCT
alkej-870	278	15	.	.	PUNCT
alkej-870	279	1	voxceleb	voxceleb	PROPN
alkej-870	279	2	an	an	DET
alkej-870	279	3	audio	audio	ADJ
alkej-870	279	4	-	-	ADJ
alkej-870	279	5	visual	visual	ADJ
alkej-870	279	6	dataset	dataset	NOUN
alkej-870	279	7	consisting	consist	VERB
alkej-870	279	8	of	of	ADP
alkej-870	279	9	short	short	ADJ
alkej-870	279	10	clips	clip	NOUN
alkej-870	279	11	of	of	ADP
alkej-870	279	12	human	human	ADJ
alkej-870	279	13	speech	speech	NOUN
alkej-870	279	14	,	,	PUNCT
alkej-870	279	15	extracted	extract	VERB
alkej-870	279	16	from	from	ADP
alkej-870	279	17	interview	interview	NOUN
alkej-870	279	18	videos	video	NOUN
alkej-870	279	19	uploaded	upload	VERB
alkej-870	279	20	to	to	ADP
alkej-870	279	21	youtube	youtube	NOUN
alkej-870	279	22	[	[	X
alkej-870	279	23	55	55	NUM
alkej-870	279	24	]	]	PUNCT
alkej-870	279	25	.	.	PUNCT
alkej-870	280	1	voxceleb2	voxceleb2	PROPN
alkej-870	280	2	contains	contain	VERB
alkej-870	280	3	over	over	ADP
alkej-870	280	4	1	1	NUM
alkej-870	280	5	million	million	NUM
alkej-870	280	6	utterances	utterance	NOUN
alkej-870	280	7	for	for	ADP
alkej-870	280	8	6,112	6,112	NUM
alkej-870	280	9	celebrities	celebrity	NOUN
alkej-870	280	10	,	,	PUNCT
alkej-870	280	11	extracted	extract	VERB
alkej-870	280	12	from	from	ADP
alkej-870	280	13	videos	video	NOUN
alkej-870	280	14	uploaded	upload	VERB
alkej-870	280	15	to	to	ADP
alkej-870	280	16	youtube	youtube	NOUN
alkej-870	280	17	[	[	X
alkej-870	280	18	56	56	NUM
alkej-870	280	19	]	]	PUNCT
alkej-870	280	20	.	.	PUNCT
alkej-870	281	1	lrw	lrw	PROPN
alkej-870	281	2	consists	consist	VERB
alkej-870	281	3	of	of	ADP
alkej-870	281	4	up	up	ADP
alkej-870	281	5	to	to	PART
alkej-870	281	6	1000	1000	NUM
alkej-870	281	7	utterances	utterance	NOUN
alkej-870	281	8	of	of	ADP
alkej-870	281	9	500	500	NUM
alkej-870	281	10	different	different	ADJ
alkej-870	281	11	words	word	NOUN
alkej-870	281	12	,	,	PUNCT
alkej-870	281	13	spoken	speak	VERB
alkej-870	281	14	by	by	ADP
alkej-870	281	15	hundreds	hundred	NOUN
alkej-870	281	16	of	of	ADP
alkej-870	281	17	different	different	ADJ
alkej-870	281	18	speakers	speaker	NOUN
alkej-870	282	1	[	[	X
alkej-870	282	2	57	57	NUM
alkej-870	282	3	]	]	PUNCT
alkej-870	282	4	.	.	PUNCT
alkej-870	283	1	lrs	lrs	PROPN
alkej-870	283	2	consists	consist	VERB
alkej-870	283	3	of	of	ADP
alkej-870	283	4	thousands	thousand	NOUN
alkej-870	283	5	of	of	ADP
alkej-870	283	6	spoken	spoken	ADJ
alkej-870	283	7	sentences	sentence	NOUN
alkej-870	283	8	from	from	ADP
alkej-870	283	9	bbc	bbc	PROPN
alkej-870	283	10	television	television	NOUN
alkej-870	283	11	.	.	PUNCT
alkej-870	284	1	each	each	DET
alkej-870	284	2	sentence	sentence	NOUN
alkej-870	284	3	is	be	AUX
alkej-870	284	4	up	up	ADP
alkej-870	284	5	to	to	PART
alkej-870	284	6	100	100	NUM
alkej-870	284	7	characters	character	NOUN
alkej-870	284	8	in	in	ADP
alkej-870	284	9	length	length	NOUN
alkej-870	284	10	[	[	X
alkej-870	284	11	58	58	NUM
alkej-870	284	12	]	]	PUNCT
alkej-870	284	13	.	.	PUNCT
alkej-870	285	1	lrs3	lrs3	PROPN
alkej-870	285	2	consists	consist	VERB
alkej-870	285	3	of	of	ADP
alkej-870	285	4	thousands	thousand	NOUN
alkej-870	285	5	of	of	ADP
alkej-870	285	6	spoken	spoken	ADJ
alkej-870	285	7	sentences	sentence	NOUN
alkej-870	285	8	from	from	ADP
alkej-870	285	9	ted	ted	ADJ
alkej-870	285	10	and	and	CCONJ
alkej-870	285	11	tedx	tedx	ADJ
alkej-870	285	12	videos	video	NOUN
alkej-870	285	13	[	[	X
alkej-870	285	14	59	59	NUM
alkej-870	285	15	]	]	PUNCT
alkej-870	285	16	.	.	PUNCT
alkej-870	286	1	avaactivespeaker	avaactivespeaker	PROPN
alkej-870	286	2	contains	contain	VERB
alkej-870	286	3	about	about	ADV
alkej-870	286	4	38.5	38.5	NUM
alkej-870	286	5	hours	hour	NOUN
alkej-870	286	6	of	of	ADP
alkej-870	286	7	face	face	NOUN
alkej-870	286	8	tracks	track	NOUN
alkej-870	286	9	,	,	PUNCT
alkej-870	286	10	and	and	CCONJ
alkej-870	286	11	the	the	DET
alkej-870	286	12	corresponding	corresponding	ADJ
alkej-870	286	13	audio	audio	NOUN
alkej-870	286	14	[	[	X
alkej-870	286	15	60	60	NUM
alkej-870	286	16	]	]	PUNCT
alkej-870	286	17	.	.	PUNCT
alkej-870	287	1	avspeech	avspeech	ADJ
alkej-870	287	2	large	large	ADJ
alkej-870	287	3	-	-	PUNCT
alkej-870	287	4	scale	scale	NOUN
alkej-870	287	5	audio	audio	ADJ
alkej-870	287	6	-	-	ADJ
alkej-870	287	7	visual	visual	ADJ
alkej-870	287	8	dataset	dataset	NOUN
alkej-870	287	9	comprising	comprising	NOUN
alkej-870	287	10	speech	speech	NOUN
alkej-870	287	11	video	video	NOUN
alkej-870	287	12	clips	clip	NOUN
alkej-870	287	13	with	with	ADP
alkej-870	287	14	no	no	DET
alkej-870	287	15	interfering	interfere	VERB
alkej-870	287	16	background	background	NOUN
alkej-870	287	17	noises	noise	NOUN
alkej-870	287	18	contains	contain	VERB
alkej-870	287	19	roughly	roughly	ADV
alkej-870	287	20	4700	4700	NUM
alkej-870	287	21	hours	hour	NOUN
alkej-870	287	22	of	of	ADP
alkej-870	287	23	video	video	NOUN
alkej-870	287	24	segments	segment	NOUN
alkej-870	287	25	[	[	X
alkej-870	287	26	17	17	NUM
alkej-870	287	27	]	]	SYM
alkej-870	287	28	.	.	PUNCT
alkej-870	288	1	5	5	X
alkej-870	288	2	.	.	X
alkej-870	288	3	conclusions	conclusion	NOUN
alkej-870	288	4	the	the	DET
alkej-870	288	5	rapid	rapid	ADJ
alkej-870	288	6	advancement	advancement	NOUN
alkej-870	288	7	of	of	ADP
alkej-870	288	8	deep	deep	ADJ
alkej-870	288	9	learning	learning	NOUN
alkej-870	288	10	approaches	approach	NOUN
alkej-870	288	11	has	have	AUX
alkej-870	288	12	revolutionized	revolutionize	VERB
alkej-870	288	13	the	the	DET
alkej-870	288	14	field	field	NOUN
alkej-870	288	15	of	of	ADP
alkej-870	288	16	audiovisual	audiovisual	ADJ
alkej-870	288	17	source	source	NOUN
alkej-870	288	18	separation	separation	NOUN
alkej-870	288	19	,	,	PUNCT
alkej-870	288	20	enabling	enable	VERB
alkej-870	288	21	significant	significant	ADJ
alkej-870	288	22	progress	progress	NOUN
alkej-870	288	23	in	in	ADP
alkej-870	288	24	separating	separate	VERB
alkej-870	288	25	speech	speech	NOUN
alkej-870	288	26	and	and	CCONJ
alkej-870	288	27	other	other	ADJ
alkej-870	288	28	audio	audio	ADJ
alkej-870	288	29	sources	source	NOUN
alkej-870	288	30	from	from	ADP
alkej-870	288	31	complex	complex	ADJ
alkej-870	288	32	audio	audio	ADJ
alkej-870	288	33	-	-	ADJ
alkej-870	288	34	visual	visual	ADJ
alkej-870	288	35	mixtures	mixture	NOUN
alkej-870	288	36	.	.	PUNCT
alkej-870	289	1	dnns	dnn	NOUN
alkej-870	289	2	,	,	PUNCT
alkej-870	289	3	cnns	cnn	NOUN
alkej-870	289	4	,	,	PUNCT
alkej-870	289	5	lstm	lstm	ADJ
alkej-870	289	6	,	,	PUNCT
alkej-870	289	7	and	and	CCONJ
alkej-870	289	8	blstm	blstm	NOUN
alkej-870	289	9	have	have	AUX
alkej-870	289	10	emerged	emerge	VERB
alkej-870	289	11	as	as	ADP
alkej-870	289	12	powerful	powerful	ADJ
alkej-870	289	13	tools	tool	NOUN
alkej-870	289	14	to	to	PART
alkej-870	289	15	address	address	VERB
alkej-870	289	16	the	the	DET
alkej-870	289	17	challenges	challenge	NOUN
alkej-870	289	18	posed	pose	VERB
alkej-870	289	19	by	by	ADP
alkej-870	289	20	audio	audio	ADJ
alkej-870	289	21	-	-	ADJ
alkej-870	289	22	visual	visual	ADJ
alkej-870	289	23	sources	source	NOUN
alkej-870	289	24	'	'	PART
alkej-870	289	25	multi	multi	ADJ
alkej-870	289	26	-	-	NOUN
alkej-870	289	27	modality	modality	NOUN
alkej-870	289	28	and	and	CCONJ
alkej-870	289	29	dynamic	dynamic	ADJ
alkej-870	289	30	nature	nature	NOUN
alkej-870	289	31	.	.	PUNCT
alkej-870	290	1	researchers	researcher	NOUN
alkej-870	290	2	should	should	AUX
alkej-870	290	3	consider	consider	VERB
alkej-870	290	4	the	the	DET
alkej-870	290	5	task	task	NOUN
alkej-870	290	6	requirements	requirement	NOUN
alkej-870	290	7	,	,	PUNCT
alkej-870	290	8	available	available	ADJ
alkej-870	290	9	data	datum	NOUN
alkej-870	290	10	,	,	PUNCT
alkej-870	290	11	model	model	NOUN
alkej-870	290	12	nooralhuda	nooralhuda	PROPN
alkej-870	290	13	mudhafar	mudhafar	PROPN
alkej-870	290	14	sleman	sleman	PROPN
alkej-870	290	15	al	al	PROPN
alkej-870	290	16	-	-	PUNCT
alkej-870	290	17	khwarizmi	khwarizmi	PROPN
alkej-870	290	18	engineering	engineering	NOUN
alkej-870	290	19	journal	journal	PROPN
alkej-870	290	20	,	,	PUNCT
alkej-870	290	21	vol	vol	NOUN
alkej-870	290	22	.	.	PROPN
alkej-870	291	1	19	19	NUM
alkej-870	291	2	,	,	PUNCT
alkej-870	291	3	no	no	INTJ
alkej-870	291	4	.	.	NOUN
alkej-870	291	5	4	4	NUM
alkej-870	291	6	,	,	PUNCT
alkej-870	291	7	p.p	p.p	PROPN
alkej-870	291	8	.	.	NOUN
alkej-870	291	9	4255	4255	NUM
alkej-870	291	10	(	(	PUNCT
alkej-870	291	11	2023	2023	NUM
alkej-870	291	12	)	)	PUNCT
alkej-870	291	13	50	50	NUM
alkej-870	291	14	complexity	complexity	NOUN
alkej-870	291	15	,	,	PUNCT
alkej-870	291	16	computational	computational	ADJ
alkej-870	291	17	resources	resource	NOUN
alkej-870	291	18	,	,	PUNCT
alkej-870	291	19	interpretability	interpretability	NOUN
alkej-870	291	20	needs	need	NOUN
alkej-870	291	21	,	,	PUNCT
alkej-870	291	22	previous	previous	ADJ
alkej-870	291	23	research	research	NOUN
alkej-870	291	24	,	,	PUNCT
alkej-870	291	25	and	and	CCONJ
alkej-870	291	26	time	time	NOUN
alkej-870	291	27	/	/	SYM
alkej-870	291	28	resource	resource	NOUN
alkej-870	291	29	constraints	constraint	NOUN
alkej-870	291	30	when	when	SCONJ
alkej-870	291	31	choosing	choose	VERB
alkej-870	291	32	a	a	DET
alkej-870	291	33	method	method	NOUN
alkej-870	291	34	.	.	PUNCT
alkej-870	292	1	through	through	ADP
alkej-870	292	2	careful	careful	ADJ
alkej-870	292	3	evaluation	evaluation	NOUN
alkej-870	292	4	of	of	ADP
alkej-870	292	5	these	these	DET
alkej-870	292	6	factors	factor	NOUN
alkej-870	292	7	,	,	PUNCT
alkej-870	292	8	researchers	researcher	NOUN
alkej-870	292	9	can	can	AUX
alkej-870	292	10	make	make	VERB
alkej-870	292	11	informed	informed	ADJ
alkej-870	292	12	decisions	decision	NOUN
alkej-870	292	13	and	and	CCONJ
alkej-870	292	14	choose	choose	VERB
alkej-870	292	15	a	a	DET
alkej-870	292	16	suitable	suitable	ADJ
alkej-870	292	17	method	method	NOUN
alkej-870	292	18	.	.	PUNCT
alkej-870	293	1	it	it	PRON
alkej-870	293	2	is	be	AUX
alkej-870	293	3	important	important	ADJ
alkej-870	293	4	to	to	PART
alkej-870	293	5	note	note	VERB
alkej-870	293	6	that	that	DET
alkej-870	293	7	blstm	blstm	NOUN
alkej-870	293	8	alone	alone	ADV
alkej-870	293	9	may	may	AUX
alkej-870	293	10	not	not	PART
alkej-870	293	11	be	be	AUX
alkej-870	293	12	sufficient	sufficient	ADJ
alkej-870	293	13	,	,	PUNCT
alkej-870	293	14	but	but	CCONJ
alkej-870	293	15	it	it	PRON
alkej-870	293	16	plays	play	VERB
alkej-870	293	17	a	a	DET
alkej-870	293	18	crucial	crucial	ADJ
alkej-870	293	19	role	role	NOUN
alkej-870	293	20	as	as	ADP
alkej-870	293	21	a	a	DET
alkej-870	293	22	key	key	ADJ
alkej-870	293	23	component	component	NOUN
alkej-870	293	24	in	in	ADP
alkej-870	293	25	many	many	ADJ
alkej-870	293	26	state	state	NOUN
alkej-870	293	27	-	-	PUNCT
alkej-870	293	28	of	of	ADP
alkej-870	293	29	-	-	PUNCT
alkej-870	293	30	the	the	DET
alkej-870	293	31	-	-	PUNCT
alkej-870	293	32	art	art	NOUN
alkej-870	293	33	architectures	architecture	NOUN
alkej-870	293	34	.	.	PUNCT
alkej-870	294	1	the	the	DET
alkej-870	294	2	combination	combination	NOUN
alkej-870	294	3	of	of	ADP
alkej-870	294	4	blstm	blstm	NOUN
alkej-870	294	5	with	with	ADP
alkej-870	294	6	other	other	ADJ
alkej-870	294	7	modules	module	NOUN
alkej-870	294	8	like	like	ADP
alkej-870	294	9	cnns	cnn	NOUN
alkej-870	294	10	and	and	CCONJ
alkej-870	294	11	dnns	dnn	NOUN
alkej-870	294	12	has	have	AUX
alkej-870	294	13	demonstrated	demonstrate	VERB
alkej-870	294	14	promising	promise	VERB
alkej-870	294	15	outcomes	outcome	NOUN
alkej-870	294	16	in	in	ADP
alkej-870	294	17	enhancing	enhance	VERB
alkej-870	294	18	separation	separation	NOUN
alkej-870	294	19	performance	performance	NOUN
alkej-870	294	20	.	.	PUNCT
alkej-870	295	1	thus	thus	ADV
alkej-870	295	2	,	,	PUNCT
alkej-870	295	3	the	the	DET
alkej-870	295	4	synergistic	synergistic	ADJ
alkej-870	295	5	utilization	utilization	NOUN
alkej-870	295	6	of	of	ADP
alkej-870	295	7	these	these	DET
alkej-870	295	8	techniques	technique	NOUN
alkej-870	295	9	within	within	ADP
alkej-870	295	10	an	an	DET
alkej-870	295	11	integrated	integrate	VERB
alkej-870	295	12	system	system	NOUN
alkej-870	295	13	yields	yield	VERB
alkej-870	295	14	the	the	DET
alkej-870	295	15	most	most	ADV
alkej-870	295	16	promising	promising	ADJ
alkej-870	295	17	results	result	NOUN
alkej-870	295	18	,	,	PUNCT
alkej-870	295	19	rather	rather	ADV
alkej-870	295	20	than	than	ADP
alkej-870	295	21	relying	rely	VERB
alkej-870	295	22	on	on	ADP
alkej-870	295	23	any	any	DET
alkej-870	295	24	single	single	ADJ
alkej-870	295	25	technique	technique	NOUN
alkej-870	295	26	in	in	ADP
alkej-870	295	27	isolation	isolation	NOUN
alkej-870	295	28	.	.	PUNCT
alkej-870	296	1	moreover	moreover	ADV
alkej-870	296	2	,	,	PUNCT
alkej-870	296	3	incorporating	incorporate	VERB
alkej-870	296	4	these	these	DET
alkej-870	296	5	models	model	NOUN
alkej-870	296	6	with	with	ADP
alkej-870	296	7	additional	additional	ADJ
alkej-870	296	8	deep	deep	ADJ
alkej-870	296	9	learning	learning	NOUN
alkej-870	296	10	techniques	technique	NOUN
alkej-870	296	11	,	,	PUNCT
alkej-870	296	12	such	such	ADJ
alkej-870	296	13	as	as	ADP
alkej-870	296	14	attention	attention	NOUN
alkej-870	296	15	mechanisms	mechanism	NOUN
alkej-870	296	16	and	and	CCONJ
alkej-870	296	17	generative	generative	ADJ
alkej-870	296	18	adversarial	adversarial	ADJ
alkej-870	296	19	networks	network	NOUN
alkej-870	296	20	,	,	PUNCT
alkej-870	296	21	has	have	AUX
alkej-870	296	22	remarkably	remarkably	ADV
alkej-870	296	23	improved	improve	VERB
alkej-870	296	24	the	the	DET
alkej-870	296	25	separation	separation	NOUN
alkej-870	296	26	process	process	NOUN
alkej-870	296	27	.	.	PUNCT
alkej-870	297	1	large	large	ADJ
alkej-870	297	2	-	-	PUNCT
alkej-870	297	3	scale	scale	NOUN
alkej-870	297	4	datasets	dataset	NOUN
alkej-870	297	5	and	and	CCONJ
alkej-870	297	6	the	the	DET
alkej-870	297	7	availability	availability	NOUN
alkej-870	297	8	of	of	ADP
alkej-870	297	9	high	high	ADJ
alkej-870	297	10	-	-	PUNCT
alkej-870	297	11	performance	performance	NOUN
alkej-870	297	12	computing	computing	NOUN
alkej-870	297	13	resources	resource	NOUN
alkej-870	297	14	have	have	AUX
alkej-870	297	15	facilitated	facilitate	VERB
alkej-870	297	16	the	the	DET
alkej-870	297	17	training	training	NOUN
alkej-870	297	18	of	of	ADP
alkej-870	297	19	deep	deep	ADJ
alkej-870	297	20	models	model	NOUN
alkej-870	297	21	and	and	CCONJ
alkej-870	297	22	the	the	DET
alkej-870	297	23	development	development	NOUN
alkej-870	297	24	of	of	ADP
alkej-870	297	25	more	more	ADJ
alkej-870	297	26	complex	complex	ADJ
alkej-870	297	27	architectures	architecture	NOUN
alkej-870	297	28	that	that	PRON
alkej-870	297	29	can	can	AUX
alkej-870	297	30	handle	handle	VERB
alkej-870	297	31	the	the	DET
alkej-870	297	32	variability	variability	NOUN
alkej-870	297	33	of	of	ADP
alkej-870	297	34	acoustical	acoustical	ADJ
alkej-870	297	35	conditions	condition	NOUN
alkej-870	297	36	.	.	PUNCT
alkej-870	298	1	despite	despite	SCONJ
alkej-870	298	2	these	these	DET
alkej-870	298	3	advancements	advancement	NOUN
alkej-870	298	4	,	,	PUNCT
alkej-870	298	5	significant	significant	ADJ
alkej-870	298	6	challenges	challenge	NOUN
alkej-870	298	7	remain	remain	VERB
alkej-870	298	8	,	,	PUNCT
alkej-870	298	9	including	include	VERB
alkej-870	298	10	handling	handle	VERB
alkej-870	298	11	variable	variable	ADJ
alkej-870	298	12	speech	speech	NOUN
alkej-870	298	13	overlap	overlap	NOUN
alkej-870	298	14	and	and	CCONJ
alkej-870	298	15	room	room	NOUN
alkej-870	298	16	acoustics	acoustic	NOUN
alkej-870	298	17	.	.	PUNCT
alkej-870	299	1	there	there	PRON
alkej-870	299	2	is	be	VERB
alkej-870	299	3	a	a	DET
alkej-870	299	4	need	need	NOUN
alkej-870	299	5	for	for	ADP
alkej-870	299	6	robust	robust	ADJ
alkej-870	299	7	models	model	NOUN
alkej-870	299	8	that	that	PRON
alkej-870	299	9	can	can	AUX
alkej-870	299	10	generalize	generalize	VERB
alkej-870	299	11	across	across	ADP
alkej-870	299	12	different	different	ADJ
alkej-870	299	13	acoustic	acoustic	ADJ
alkej-870	299	14	environments	environment	NOUN
alkej-870	299	15	and	and	CCONJ
alkej-870	299	16	effectively	effectively	ADV
alkej-870	299	17	handle	handle	VERB
alkej-870	299	18	various	various	ADJ
alkej-870	299	19	types	type	NOUN
alkej-870	299	20	of	of	ADP
alkej-870	299	21	audio	audio	ADJ
alkej-870	299	22	-	-	ADJ
alkej-870	299	23	visual	visual	ADJ
alkej-870	299	24	mixtures	mixture	NOUN
alkej-870	299	25	.	.	PUNCT
alkej-870	300	1	in	in	ADP
alkej-870	300	2	conclusion	conclusion	NOUN
alkej-870	300	3	,	,	PUNCT
alkej-870	300	4	the	the	DET
alkej-870	300	5	use	use	NOUN
alkej-870	300	6	of	of	ADP
alkej-870	300	7	deep	deep	ADJ
alkej-870	300	8	learning	learning	NOUN
alkej-870	300	9	approaches	approach	NOUN
alkej-870	300	10	in	in	ADP
alkej-870	300	11	audio	audio	ADJ
alkej-870	300	12	-	-	ADJ
alkej-870	300	13	visual	visual	ADJ
alkej-870	300	14	source	source	NOUN
alkej-870	300	15	separation	separation	NOUN
alkej-870	300	16	has	have	AUX
alkej-870	300	17	enabled	enable	VERB
alkej-870	300	18	significant	significant	ADJ
alkej-870	300	19	progress	progress	NOUN
alkej-870	300	20	in	in	ADP
alkej-870	300	21	the	the	DET
alkej-870	300	22	field	field	NOUN
alkej-870	300	23	and	and	CCONJ
alkej-870	300	24	holds	hold	VERB
alkej-870	300	25	great	great	ADJ
alkej-870	300	26	promise	promise	NOUN
alkej-870	300	27	for	for	ADP
alkej-870	300	28	further	further	ADJ
alkej-870	300	29	advancements	advancement	NOUN
alkej-870	300	30	in	in	ADP
alkej-870	300	31	the	the	DET
alkej-870	300	32	future	future	NOUN
alkej-870	300	33	.	.	PUNCT
alkej-870	301	1	continued	continue	VERB
alkej-870	301	2	research	research	NOUN
alkej-870	301	3	and	and	CCONJ
alkej-870	301	4	development	development	NOUN
alkej-870	301	5	of	of	ADP
alkej-870	301	6	these	these	DET
alkej-870	301	7	techniques	technique	NOUN
alkej-870	301	8	will	will	AUX
alkej-870	301	9	be	be	AUX
alkej-870	301	10	essential	essential	ADJ
alkej-870	301	11	to	to	PART
alkej-870	301	12	address	address	VERB
alkej-870	301	13	the	the	DET
alkej-870	301	14	remaining	remain	VERB
alkej-870	301	15	challenges	challenge	NOUN
alkej-870	301	16	and	and	CCONJ
alkej-870	301	17	facilitate	facilitate	VERB
alkej-870	301	18	the	the	DET
alkej-870	301	19	adoption	adoption	NOUN
alkej-870	301	20	of	of	ADP
alkej-870	301	21	audio	audio	ADJ
alkej-870	301	22	-	-	ADJ
alkej-870	301	23	visual	visual	ADJ
alkej-870	301	24	source	source	NOUN
alkej-870	301	25	separation	separation	NOUN
alkej-870	301	26	in	in	ADP
alkej-870	301	27	practical	practical	ADJ
alkej-870	301	28	applications	application	NOUN
alkej-870	301	29	.	.	PUNCT
alkej-870	302	1	acknowledgments	acknowledgment	NOUN
alkej-870	302	2	this	this	DET
alkej-870	302	3	work	work	NOUN
alkej-870	302	4	is	be	AUX
alkej-870	302	5	supported	support	VERB
alkej-870	302	6	by	by	ADP
alkej-870	302	7	the	the	DET
alkej-870	302	8	information	information	NOUN
alkej-870	302	9	and	and	CCONJ
alkej-870	302	10	communications	communication	NOUN
alkej-870	302	11	engineering	engineering	NOUN
alkej-870	302	12	department	department	PROPN
alkej-870	302	13	,	,	PUNCT
alkej-870	302	14	alkhawarizmi	alkhawarizmi	PROPN
alkej-870	302	15	engineering	engineering	PROPN
alkej-870	302	16	college	college	PROPN
alkej-870	302	17	,	,	PUNCT
alkej-870	302	18	university	university	NOUN
alkej-870	302	19	of	of	ADP
alkej-870	302	20	baghdad	baghdad	PROPN
alkej-870	302	21	.	.	PUNCT
alkej-870	303	1	references	reference	NOUN
alkej-870	303	2	[	[	X
alkej-870	303	3	1	1	NUM
alkej-870	303	4	]	]	PUNCT
alkej-870	303	5	a.	a.	PROPN
alkej-870	303	6	al	al	PROPN
alkej-870	303	7	-	-	PUNCT
alkej-870	303	8	tmeme	tmeme	PROPN
alkej-870	303	9	,	,	PUNCT
alkej-870	303	10	w.	w.	PROPN
alkej-870	303	11	l.	l.	PROPN
alkej-870	303	12	woo	woo	PROPN
alkej-870	303	13	,	,	PUNCT
alkej-870	303	14	s.	s.	PROPN
alkej-870	303	15	s.	s.	PROPN
alkej-870	303	16	dlay	dlay	PROPN
alkej-870	303	17	and	and	CCONJ
alkej-870	303	18	b.	b.	PROPN
alkej-870	303	19	gao	gao	PROPN
alkej-870	303	20	,	,	PUNCT
alkej-870	303	21	"	"	PUNCT
alkej-870	303	22	underdetermined	underdetermined	ADJ
alkej-870	303	23	convolutive	convolutive	ADJ
alkej-870	303	24	source	source	NOUN
alkej-870	303	25	separation	separation	NOUN
alkej-870	303	26	using	use	VERB
alkej-870	303	27	gem	gem	NOUN
alkej-870	303	28	-	-	PUNCT
alkej-870	303	29	mu	mu	NOUN
alkej-870	303	30	with	with	ADP
alkej-870	303	31	variational	variational	ADJ
alkej-870	303	32	approximated	approximate	VERB
alkej-870	303	33	optimum	optimum	ADJ
alkej-870	303	34	model	model	NOUN
alkej-870	303	35	order	order	NOUN
alkej-870	303	36	nmf2d	nmf2d	ADP
alkej-870	303	37	,	,	PUNCT
alkej-870	303	38	"	"	PUNCT
alkej-870	303	39	in	in	ADP
alkej-870	303	40	ieee	ieee	NOUN
alkej-870	303	41	/	/	SYM
alkej-870	303	42	acm	acm	PROPN
alkej-870	303	43	transactions	transaction	NOUN
alkej-870	303	44	on	on	ADP
alkej-870	303	45	audio	audio	NOUN
alkej-870	303	46	,	,	PUNCT
alkej-870	303	47	speech	speech	NOUN
alkej-870	303	48	,	,	PUNCT
alkej-870	303	49	and	and	CCONJ
alkej-870	303	50	language	language	NOUN
alkej-870	303	51	processing	processing	NOUN
alkej-870	303	52	,	,	PUNCT
alkej-870	303	53	vol	vol	NOUN
alkej-870	303	54	.	.	PROPN
alkej-870	304	1	25	25	NUM
alkej-870	304	2	,	,	PUNCT
alkej-870	304	3	no	no	INTJ
alkej-870	304	4	.	.	NOUN
alkej-870	304	5	1	1	NUM
alkej-870	304	6	,	,	PUNCT
alkej-870	304	7	pp	pp	ADJ
alkej-870	304	8	.	.	PUNCT
alkej-870	305	1	35	35	NUM
alkej-870	305	2	-	-	SYM
alkej-870	305	3	49	49	NUM
alkej-870	305	4	,	,	PUNCT
alkej-870	305	5	jan	jan	PROPN
alkej-870	305	6	.	.	PROPN
alkej-870	305	7	2017	2017	NUM
alkej-870	305	8	.	.	PUNCT
alkej-870	306	1	http://dx.doi.org/10.1109/taslp.2016.2620	http://dx.doi.org/10.1109/taslp.2016.2620	NOUN
alkej-870	306	2	600	600	NUM
alkej-870	306	3	.	.	PUNCT
alkej-870	307	1	[	[	X
alkej-870	307	2	2	2	NUM
alkej-870	307	3	]	]	X
alkej-870	307	4	woo	woo	NOUN
alkej-870	307	5	,	,	PUNCT
alkej-870	307	6	w.l	w.l	PROPN
alkej-870	307	7	.	.	PROPN
alkej-870	307	8	;	;	PUNCT
alkej-870	307	9	dlay	dlay	PROPN
alkej-870	307	10	,	,	PUNCT
alkej-870	307	11	s.s	s.s	PROPN
alkej-870	307	12	.	.	PROPN
alkej-870	307	13	;	;	PUNCT
alkej-870	307	14	al	al	PROPN
alkej-870	307	15	-	-	PUNCT
alkej-870	307	16	tmeme	tmeme	PROPN
alkej-870	307	17	,	,	PUNCT
alkej-870	307	18	a.	a.	NOUN
alkej-870	307	19	;	;	PUNCT
alkej-870	307	20	gao	gao	PROPN
alkej-870	307	21	,	,	PUNCT
alkej-870	307	22	b.	b.	PROPN
alkej-870	307	23	"	"	PUNCT
alkej-870	307	24	reverberant	reverberant	ADJ
alkej-870	307	25	signal	signal	NOUN
alkej-870	307	26	separation	separation	NOUN
alkej-870	307	27	using	use	VERB
alkej-870	307	28	optimized	optimize	VERB
alkej-870	307	29	complex	complex	ADJ
alkej-870	307	30	sparse	sparse	ADJ
alkej-870	307	31	nonnegative	nonnegative	ADJ
alkej-870	307	32	tensor	tensor	NOUN
alkej-870	307	33	deconvolution	deconvolution	NOUN
alkej-870	307	34	on	on	ADP
alkej-870	307	35	spectral	spectral	ADJ
alkej-870	307	36	covariance	covariance	NOUN
alkej-870	307	37	matrix	matrix	NOUN
alkej-870	307	38	"	"	PUNCT
alkej-870	307	39	.	.	PUNCT
alkej-870	308	1	digit	digit	PROPN
alkej-870	308	2	.	.	PUNCT
alkej-870	308	3	signal	signal	NOUN
alkej-870	308	4	process	process	NOUN
alkej-870	308	5	.	.	PUNCT
alkej-870	309	1	2018	2018	NUM
alkej-870	309	2	,	,	PUNCT
alkej-870	309	3	83	83	NUM
alkej-870	309	4	,	,	PUNCT
alkej-870	309	5	9	9	NUM
alkej-870	309	6	–	–	SYM
alkej-870	309	7	23	23	NUM
alkej-870	309	8	.	.	PUNCT
alkej-870	310	1	http://dx.doi.org/10.1016/j.dsp.2018.07.018	http://dx.doi.org/10.1016/j.dsp.2018.07.018	NOUN
alkej-870	311	1	[	[	X
alkej-870	311	2	3	3	X
alkej-870	311	3	]	]	X
alkej-870	311	4	al	al	PROPN
alkej-870	311	5	-	-	PUNCT
alkej-870	311	6	tmeme	tmeme	PROPN
alkej-870	311	7	,	,	PUNCT
alkej-870	311	8	a.	a.	NOUN
alkej-870	311	9	;	;	PUNCT
alkej-870	311	10	woo	woo	PROPN
alkej-870	311	11	,	,	PUNCT
alkej-870	311	12	w.l	w.l	PROPN
alkej-870	311	13	.	.	PROPN
alkej-870	311	14	;	;	PUNCT
alkej-870	311	15	dlay	dlay	PROPN
alkej-870	311	16	,	,	PUNCT
alkej-870	311	17	s.	s.	PROPN
alkej-870	311	18	;	;	PUNCT
alkej-870	311	19	gao	gao	PROPN
alkej-870	311	20	,	,	PUNCT
alkej-870	311	21	b.	b.	PROPN
alkej-870	312	1	"	"	PUNCT
alkej-870	312	2	single	single	ADJ
alkej-870	312	3	channel	channel	NOUN
alkej-870	312	4	informed	inform	VERB
alkej-870	312	5	signal	signal	NOUN
alkej-870	312	6	separation	separation	NOUN
alkej-870	312	7	using	use	VERB
alkej-870	312	8	artificial	artificial	ADJ
alkej-870	312	9	-	-	PUNCT
alkej-870	312	10	stereophonic	stereophonic	ADJ
alkej-870	312	11	mixtures	mixture	NOUN
alkej-870	312	12	and	and	CCONJ
alkej-870	312	13	exemplar	exemplar	NOUN
alkej-870	312	14	-	-	PUNCT
alkej-870	312	15	guided	guide	VERB
alkej-870	312	16	matrix	matrix	NOUN
alkej-870	312	17	factor	factor	NOUN
alkej-870	312	18	deconvolution	deconvolution	NOUN
alkej-870	312	19	"	"	PUNCT
alkej-870	312	20	.	.	PUNCT
alkej-870	313	1	int	int	NOUN
alkej-870	313	2	.	.	PUNCT
alkej-870	314	1	j.	j.	PROPN
alkej-870	314	2	adapt	adapt	PROPN
alkej-870	314	3	.	.	PUNCT
alkej-870	315	1	control	control	PROPN
alkej-870	315	2	.	.	PUNCT
alkej-870	316	1	signal	signal	NOUN
alkej-870	316	2	process	process	NOUN
alkej-870	316	3	.	.	PUNCT
alkej-870	317	1	2018	2018	NUM
alkej-870	317	2	,	,	PUNCT
alkej-870	317	3	32	32	NUM
alkej-870	317	4	,	,	PUNCT
alkej-870	317	5	1259–1281	1259–1281	NUM
alkej-870	317	6	.	.	PUNCT
alkej-870	318	1	http://dx.doi.org/10.1002/acs.2912	http://dx.doi.org/10.1002/acs.2912	PROPN
alkej-870	318	2	.	.	PUNCT
alkej-870	319	1	[	[	X
alkej-870	319	2	4	4	X
alkej-870	319	3	]	]	PUNCT
alkej-870	319	4	ahmed	ahmed	PROPN
alkej-870	319	5	al	al	PROPN
alkej-870	319	6	-	-	PUNCT
alkej-870	319	7	tmeme	tmeme	PROPN
alkej-870	319	8	,	,	PUNCT
alkej-870	319	9	w.l	w.l	PROPN
alkej-870	319	10	.	.	PROPN
alkej-870	319	11	woo	woo	PROPN
alkej-870	319	12	,	,	PUNCT
alkej-870	319	13	s.s	s.s	PROPN
alkej-870	319	14	.	.	PROPN
alkej-870	319	15	dlay	dlay	PROPN
alkej-870	319	16	,	,	PUNCT
alkej-870	319	17	and	and	CCONJ
alkej-870	319	18	b.	b.	PROPN
alkej-870	319	19	gao	gao	PROPN
alkej-870	319	20	,	,	PUNCT
alkej-870	319	21	"	"	PUNCT
alkej-870	319	22	underdetermined	underdetermined	ADJ
alkej-870	319	23	reverberant	reverberant	ADJ
alkej-870	319	24	acoustic	acoustic	ADJ
alkej-870	319	25	source	source	NOUN
alkej-870	319	26	separation	separation	NOUN
alkej-870	319	27	using	use	VERB
alkej-870	319	28	weighted	weight	VERB
alkej-870	319	29	full	full	ADJ
alkej-870	319	30	-	-	PUNCT
alkej-870	319	31	rank	rank	NOUN
alkej-870	319	32	nonnegative	nonnegative	ADJ
alkej-870	319	33	tensor	tensor	NOUN
alkej-870	319	34	models	model	NOUN
alkej-870	319	35	,	,	PUNCT
alkej-870	319	36	"	"	PUNCT
alkej-870	319	37	j.	j.	PROPN
alkej-870	319	38	acoust	acoust	PROPN
alkej-870	319	39	.	.	PUNCT
alkej-870	320	1	soc	soc	PROPN
alkej-870	320	2	.	.	PUNCT
alkej-870	321	1	am	be	AUX
alkej-870	321	2	,	,	PUNCT
alkej-870	321	3	138	138	NUM
alkej-870	321	4	,	,	PUNCT
alkej-870	321	5	3411	3411	NUM
alkej-870	321	6	,	,	PUNCT
alkej-870	321	7	2015	2015	NUM
alkej-870	321	8	.	.	PUNCT
alkej-870	322	1	http://dx.doi.org/10.1121/1.4923156	http://dx.doi.org/10.1121/1.4923156	PROPN
alkej-870	322	2	.	.	PUNCT
alkej-870	323	1	[	[	X
alkej-870	323	2	5	5	NUM
alkej-870	323	3	]	]	X
alkej-870	323	4	amer	amer	PROPN
alkej-870	323	5	,	,	PUNCT
alkej-870	323	6	r.	r.	PROPN
alkej-870	323	7	,	,	PUNCT
alkej-870	323	8	and	and	CCONJ
alkej-870	323	9	al	al	PROPN
alkej-870	323	10	tmeme	tmeme	PROPN
alkej-870	323	11	,	,	PUNCT
alkej-870	323	12	a.	a.	NOUN
alkej-870	323	13	"	"	PUNCT
alkej-870	323	14	hybrid	hybrid	ADJ
alkej-870	323	15	deep	deep	ADJ
alkej-870	323	16	learning	learning	NOUN
alkej-870	323	17	model	model	NOUN
alkej-870	323	18	for	for	ADP
alkej-870	323	19	singing	singe	VERB
alkej-870	323	20	voice	voice	NOUN
alkej-870	323	21	separation	separation	NOUN
alkej-870	323	22	"	"	PUNCT
alkej-870	323	23	.	.	PUNCT
alkej-870	324	1	mendel	mendel	PROPN
alkej-870	324	2	27	27	NUM
alkej-870	324	3	,	,	PUNCT
alkej-870	324	4	2	2	NUM
alkej-870	324	5	(	(	PUNCT
alkej-870	324	6	2021	2021	NUM
alkej-870	324	7	)	)	PUNCT
alkej-870	324	8	,	,	PUNCT
alkej-870	324	9	44–50	44–50	NUM
alkej-870	324	10	.	.	PUNCT
alkej-870	325	1	http://dx.doi.org/10.13164/mendel.2021.2.04	http://dx.doi.org/10.13164/mendel.2021.2.04	NOUN
alkej-870	325	2	4	4	X
alkej-870	325	3	.	.	PUNCT
alkej-870	326	1	[	[	X
alkej-870	326	2	6	6	NUM
alkej-870	326	3	]	]	X
alkej-870	326	4	mahmood	mahmood	NOUN
alkej-870	326	5	,	,	PUNCT
alkej-870	326	6	israa	israa	ADJ
alkej-870	326	7	n.	n.	PROPN
alkej-870	326	8	and	and	CCONJ
alkej-870	326	9	hasanen	hasanen	PROPN
alkej-870	326	10	s.	s.	PROPN
alkej-870	326	11	abdullah	abdullah	PROPN
alkej-870	326	12	,	,	PUNCT
alkej-870	326	13	"	"	PUNCT
alkej-870	326	14	telecom	telecom	NOUN
alkej-870	326	15	churn	churn	NOUN
alkej-870	326	16	prediction	prediction	NOUN
alkej-870	326	17	based	base	VERB
alkej-870	326	18	on	on	ADP
alkej-870	326	19	deep	deep	ADJ
alkej-870	326	20	learning	learning	NOUN
alkej-870	326	21	approach	approach	NOUN
alkej-870	326	22	"	"	PUNCT
alkej-870	326	23	(	(	PUNCT
alkej-870	326	24	2022	2022	NUM
alkej-870	326	25	)	)	PUNCT
alkej-870	326	26	63(6	63(6	PROPN
alkej-870	326	27	)	)	PUNCT
alkej-870	326	28	iraqi	iraqi	ADJ
alkej-870	326	29	journal	journal	NOUN
alkej-870	326	30	of	of	ADP
alkej-870	326	31	science	science	NOUN
alkej-870	326	32	.	.	PUNCT
alkej-870	327	1	http://dx.doi.org/10.24996/ijs.2022.63.6.32	http://dx.doi.org/10.24996/ijs.2022.63.6.32	ADJ
alkej-870	327	2	.	.	PUNCT
alkej-870	328	1	[	[	X
alkej-870	328	2	7	7	NUM
alkej-870	328	3	]	]	X
alkej-870	328	4	jameel	jameel	X
alkej-870	328	5	,	,	PUNCT
alkej-870	328	6	humam	humam	PROPN
alkej-870	328	7	khaled	khale	VERB
alkej-870	328	8	and	and	CCONJ
alkej-870	328	9	ban	ban	VERB
alkej-870	328	10	nadeem	nadeem	ADJ
alkej-870	328	11	dhannoon	dhannoon	NOUN
alkej-870	328	12	,	,	PUNCT
alkej-870	328	13	"	"	PUNCT
alkej-870	328	14	gait	gait	ADJ
alkej-870	328	15	recognition	recognition	NOUN
alkej-870	328	16	based	base	VERB
alkej-870	328	17	on	on	ADP
alkej-870	328	18	deep	deep	ADJ
alkej-870	328	19	learning	learning	NOUN
alkej-870	328	20	"	"	PUNCT
alkej-870	328	21	(	(	PUNCT
alkej-870	328	22	2022	2022	NUM
alkej-870	328	23	)	)	PUNCT
alkej-870	328	24	63(1	63(1	NUM
alkej-870	328	25	)	)	PUNCT
alkej-870	328	26	iraqi	iraqi	ADJ
alkej-870	328	27	journal	journal	NOUN
alkej-870	328	28	of	of	ADP
alkej-870	328	29	science	science	NOUN
alkej-870	328	30	.	.	PUNCT
alkej-870	329	1	http://dx.doi.org/10.24996/ijs.2022.63.1.36	http://dx.doi.org/10.24996/ijs.2022.63.1.36	NOUN
alkej-870	329	2	.	.	PUNCT
alkej-870	330	1	[	[	X
alkej-870	330	2	8	8	NUM
alkej-870	330	3	]	]	X
alkej-870	330	4	al	al	PROPN
alkej-870	330	5	-	-	PUNCT
alkej-870	330	6	akkam	akkam	PROPN
alkej-870	330	7	,	,	PUNCT
alkej-870	330	8	reem	reem	PROPN
alkej-870	330	9	mohammed	mohammed	PROPN
alkej-870	330	10	jasim	jasim	PROPN
alkej-870	330	11	and	and	CCONJ
alkej-870	330	12	mohammed	mohammed	PROPN
alkej-870	330	13	sahib	sahib	PROPN
alkej-870	330	14	mahdi	mahdi	PROPN
alkej-870	330	15	altaei	altaei	PROPN
alkej-870	330	16	,	,	PUNCT
alkej-870	330	17	"	"	PUNCT
alkej-870	330	18	plants	plant	VERB
alkej-870	330	19	leaf	leaf	NOUN
alkej-870	330	20	diseases	disease	NOUN
alkej-870	330	21	detection	detection	NOUN
alkej-870	330	22	using	use	VERB
alkej-870	330	23	deep	deep	ADJ
alkej-870	330	24	learning	learning	NOUN
alkej-870	330	25	"	"	PUNCT
alkej-870	330	26	(	(	PUNCT
alkej-870	330	27	2022	2022	NUM
alkej-870	330	28	)	)	PUNCT
alkej-870	330	29	63(2	63(2	NOUN
alkej-870	330	30	)	)	PUNCT
alkej-870	330	31	iraqi	iraqi	ADJ
alkej-870	330	32	journal	journal	NOUN
alkej-870	330	33	of	of	ADP
alkej-870	330	34	science	science	NOUN
alkej-870	330	35	.	.	PUNCT
alkej-870	331	1	http://dx.doi.org/10.24996/ijs.2022.63.2.34	http://dx.doi.org/10.24996/ijs.2022.63.2.34	NOUN
alkej-870	331	2	.	.	PUNCT
alkej-870	332	1	[	[	X
alkej-870	332	2	9	9	NUM
alkej-870	332	3	]	]	SYM
alkej-870	332	4	hussein	hussein	PROPN
alkej-870	332	5	,	,	PUNCT
alkej-870	332	6	noor	noor	PROPN
alkej-870	332	7	alhuda	alhuda	PROPN
alkej-870	332	8	khalid	khalid	PROPN
alkej-870	332	9	and	and	CCONJ
alkej-870	332	10	basad	basad	ADJ
alkej-870	332	11	alsarray	alsarray	NOUN
alkej-870	332	12	,	,	PUNCT
alkej-870	332	13	"	"	PUNCT
alkej-870	332	14	deep	deep	ADJ
alkej-870	332	15	learning	learning	NOUN
alkej-870	332	16	and	and	CCONJ
alkej-870	332	17	machine	machine	NOUN
alkej-870	332	18	learning	learn	VERB
alkej-870	332	19	via	via	ADP
alkej-870	332	20	a	a	DET
alkej-870	332	21	genetic	genetic	ADJ
alkej-870	332	22	algorithm	algorithm	NOUN
alkej-870	332	23	to	to	PART
alkej-870	332	24	classify	classify	VERB
alkej-870	332	25	breast	breast	NOUN
alkej-870	332	26	cancer	cancer	NOUN
alkej-870	332	27	dna	dna	PROPN
alkej-870	332	28	data	data	PROPN
alkej-870	332	29	"	"	PUNCT
alkej-870	332	30	(	(	PUNCT
alkej-870	332	31	2022	2022	NUM
alkej-870	332	32	)	)	PUNCT
alkej-870	332	33	63(7	63(7	NOUN
alkej-870	332	34	)	)	PUNCT
alkej-870	332	35	iraqi	iraqi	ADJ
alkej-870	332	36	journal	journal	NOUN
alkej-870	332	37	of	of	ADP
alkej-870	332	38	science	science	NOUN
alkej-870	332	39	.	.	PUNCT
alkej-870	333	1	http://dx.doi.org/10.24996/ijs.2022.63.7.36	http://dx.doi.org/10.24996/ijs.2022.63.7.36	NOUN
alkej-870	333	2	.	.	PUNCT
alkej-870	334	1	[	[	X
alkej-870	334	2	10	10	NUM
alkej-870	334	3	]	]	X
alkej-870	334	4	n.	n.	PROPN
alkej-870	334	5	takahashi	takahashi	PROPN
alkej-870	334	6	,	,	PUNCT
alkej-870	334	7	m.	m.	PROPN
alkej-870	334	8	k.	k.	PROPN
alkej-870	334	9	singh	singh	PROPN
alkej-870	334	10	,	,	PUNCT
alkej-870	334	11	s.	s.	PROPN
alkej-870	334	12	basak	basak	PROPN
alkej-870	334	13	,	,	PUNCT
alkej-870	334	14	p.	p.	PROPN
alkej-870	334	15	sudarsanam	sudarsanam	PROPN
alkej-870	334	16	,	,	PUNCT
alkej-870	334	17	s.	s.	PROPN
alkej-870	334	18	ganapathy	ganapathy	PROPN
alkej-870	334	19	and	and	CCONJ
alkej-870	334	20	y.	y.	PROPN
alkej-870	334	21	mitsufuji	mitsufuji	PROPN
alkej-870	334	22	,	,	PUNCT
alkej-870	334	23	"	"	PUNCT
alkej-870	334	24	improving	improve	VERB
alkej-870	334	25	voice	voice	NOUN
alkej-870	334	26	separation	separation	NOUN
alkej-870	334	27	by	by	ADP
alkej-870	334	28	incorporating	incorporate	VERB
alkej-870	334	29	end	end	NOUN
alkej-870	334	30	-	-	PUNCT
alkej-870	334	31	to	to	ADP
alkej-870	334	32	-	-	PUNCT
alkej-870	334	33	end	end	NOUN
alkej-870	334	34	speech	speech	NOUN
alkej-870	334	35	http://dx.doi.org/10.1109/taslp.2016.2620600	http://dx.doi.org/10.1109/taslp.2016.2620600	PROPN
alkej-870	334	36	http://dx.doi.org/10.1109/taslp.2016.2620600	http://dx.doi.org/10.1109/taslp.2016.2620600	PROPN
alkej-870	334	37	http://dx.doi.org/10.1016/j.dsp.2018.07.018	http://dx.doi.org/10.1016/j.dsp.2018.07.018	NOUN
alkej-870	334	38	http://dx.doi.org/10.1002/acs.2912	http://dx.doi.org/10.1002/acs.2912	VERB
alkej-870	334	39	http://dx.doi.org/10.1121/1.4923156	http://dx.doi.org/10.1121/1.4923156	PROPN
alkej-870	334	40	http://dx.doi.org/10.13164/mendel.2021.2.044	http://dx.doi.org/10.13164/mendel.2021.2.044	PROPN
alkej-870	334	41	http://dx.doi.org/10.13164/mendel.2021.2.044	http://dx.doi.org/10.13164/mendel.2021.2.044	X
alkej-870	334	42	http://dx.doi.org/10.24996/ijs.2022.63.6.32	http://dx.doi.org/10.24996/ijs.2022.63.6.32	ADV
alkej-870	334	43	http://dx.doi.org/10.24996/ijs.2022.63.1.36	http://dx.doi.org/10.24996/ijs.2022.63.1.36	VERB
alkej-870	334	44	http://dx.doi.org/10.24996/ijs.2022.63.2.34	http://dx.doi.org/10.24996/ijs.2022.63.2.34	NOUN
alkej-870	334	45	http://dx.doi.org/10.24996/ijs.2022.63.7.36	http://dx.doi.org/10.24996/ijs.2022.63.7.36	NOUN
alkej-870	334	46	nooralhuda	nooralhuda	NOUN
alkej-870	334	47	mudhafar	mudhafar	PROPN
alkej-870	334	48	sleman	sleman	PROPN
alkej-870	334	49	al	al	PROPN
alkej-870	334	50	-	-	PUNCT
alkej-870	334	51	khwarizmi	khwarizmi	PROPN
alkej-870	334	52	engineering	engineering	NOUN
alkej-870	334	53	journal	journal	PROPN
alkej-870	334	54	,	,	PUNCT
alkej-870	334	55	vol	vol	NOUN
alkej-870	334	56	.	.	PROPN
alkej-870	334	57	19	19	NUM
alkej-870	334	58	,	,	PUNCT
alkej-870	334	59	no	no	INTJ
alkej-870	334	60	.	.	NOUN
alkej-870	334	61	4	4	NUM
alkej-870	334	62	,	,	PUNCT
alkej-870	334	63	p.p	p.p	PROPN
alkej-870	334	64	.	.	NOUN
alkej-870	334	65	4255	4255	NUM
alkej-870	334	66	(	(	PUNCT
alkej-870	334	67	2023	2023	NUM
alkej-870	334	68	)	)	PUNCT
alkej-870	334	69	51	51	NUM
alkej-870	334	70	recognition	recognition	NOUN
alkej-870	334	71	,	,	PUNCT
alkej-870	334	72	"	"	PUNCT
alkej-870	334	73	ieee	ieee	PROPN
alkej-870	334	74	international	international	ADJ
alkej-870	334	75	conference	conference	NOUN
alkej-870	334	76	on	on	ADP
alkej-870	334	77	acoustics	acoustic	NOUN
alkej-870	334	78	,	,	PUNCT
alkej-870	334	79	speech	speech	NOUN
alkej-870	334	80	and	and	CCONJ
alkej-870	334	81	signal	signal	NOUN
alkej-870	334	82	processing	processing	NOUN
alkej-870	334	83	(	(	PUNCT
alkej-870	334	84	icassp	icassp	PROPN
alkej-870	334	85	)	)	PUNCT
alkej-870	334	86	,	,	PUNCT
alkej-870	334	87	barcelona	barcelona	PROPN
alkej-870	334	88	,	,	PUNCT
alkej-870	334	89	spain	spain	PROPN
alkej-870	334	90	,	,	PUNCT
alkej-870	334	91	2020	2020	NUM
alkej-870	334	92	,	,	PUNCT
alkej-870	334	93	pp	pp	ADP
alkej-870	334	94	.	.	PUNCT
alkej-870	335	1	41	41	NUM
alkej-870	335	2	-	-	SYM
alkej-870	335	3	45	45	NUM
alkej-870	335	4	,	,	PUNCT
alkej-870	335	5	http://dx.doi.org/10.1109/icassp40776.2	http://dx.doi.org/10.1109/icassp40776.2	PROPN
alkej-870	335	6	020.9053845	020.9053845	PROPN
alkej-870	335	7	.	.	PUNCT
alkej-870	336	1	[	[	X
alkej-870	336	2	11	11	NUM
alkej-870	336	3	]	]	X
alkej-870	336	4	s.-w	s.-w	PROPN
alkej-870	336	5	.	.	PUNCT
alkej-870	337	1	chung	chung	PROPN
alkej-870	337	2	,	,	PUNCT
alkej-870	337	3	s.	s.	PROPN
alkej-870	337	4	choe	choe	PROPN
alkej-870	337	5	,	,	PUNCT
alkej-870	337	6	j.	j.	PROPN
alkej-870	337	7	s.	s.	PROPN
alkej-870	337	8	chung	chung	PROPN
alkej-870	337	9	,	,	PUNCT
alkej-870	337	10	and	and	CCONJ
alkej-870	337	11	h.-g	h.-g	PROPN
alkej-870	337	12	.	.	PUNCT
alkej-870	338	1	kang	kang	PROPN
alkej-870	338	2	,	,	PUNCT
alkej-870	338	3	"	"	PUNCT
alkej-870	338	4	facefilter	facefilter	NOUN
alkej-870	338	5	:	:	PUNCT
alkej-870	338	6	audio	audio	ADJ
alkej-870	338	7	-	-	ADJ
alkej-870	338	8	visual	visual	ADJ
alkej-870	338	9	speech	speech	NOUN
alkej-870	338	10	separation	separation	NOUN
alkej-870	338	11	using	use	VERB
alkej-870	338	12	still	still	ADV
alkej-870	338	13	images	image	NOUN
alkej-870	338	14	,	,	PUNCT
alkej-870	338	15	"	"	PUNCT
alkej-870	338	16	proc	proc	NOUN
alkej-870	338	17	.	.	PUNCT
alkej-870	338	18	of	of	ADP
alkej-870	338	19	interspeech	interspeech	NOUN
alkej-870	338	20	,	,	PUNCT
alkej-870	338	21	2020	2020	NUM
alkej-870	338	22	.	.	PUNCT
alkej-870	339	1	http://dx.doi.org/10.21437/interspeech.202	http://dx.doi.org/10.21437/interspeech.202	PROPN
alkej-870	339	2	0	0	NUM
alkej-870	339	3	-	-	SYM
alkej-870	339	4	1065	1065	NUM
alkej-870	339	5	.	.	PUNCT
alkej-870	340	1	[	[	X
alkej-870	340	2	12	12	NUM
alkej-870	340	3	]	]	X
alkej-870	340	4	r.	r.	PROPN
alkej-870	340	5	gu	gu	PROPN
alkej-870	340	6	,	,	PUNCT
alkej-870	340	7	s.-x	s.-x	NOUN
alkej-870	340	8	.	.	PUNCT
alkej-870	341	1	zhang	zhang	PROPN
alkej-870	341	2	,	,	PUNCT
alkej-870	341	3	y.	y.	PROPN
alkej-870	341	4	xu	xu	PROPN
alkej-870	341	5	,	,	PUNCT
alkej-870	341	6	l.	l.	PROPN
alkej-870	341	7	chen	chen	PROPN
alkej-870	341	8	,	,	PUNCT
alkej-870	341	9	y.	y.	PROPN
alkej-870	341	10	zou	zou	PROPN
alkej-870	341	11	,	,	PUNCT
alkej-870	341	12	and	and	CCONJ
alkej-870	341	13	d.	d.	PROPN
alkej-870	341	14	yu	yu	PROPN
alkej-870	341	15	,	,	PUNCT
alkej-870	341	16	"	"	PUNCT
alkej-870	341	17	multi	multi	ADJ
alkej-870	341	18	-	-	ADJ
alkej-870	341	19	modal	modal	ADJ
alkej-870	341	20	multichannel	multichannel	NOUN
alkej-870	341	21	target	target	NOUN
alkej-870	341	22	speech	speech	NOUN
alkej-870	341	23	separation	separation	NOUN
alkej-870	341	24	,	,	PUNCT
alkej-870	341	25	"	"	PUNCT
alkej-870	341	26	ieee	ieee	NOUN
alkej-870	341	27	journal	journal	NOUN
alkej-870	341	28	of	of	ADP
alkej-870	341	29	selected	select	VERB
alkej-870	341	30	topics	topic	NOUN
alkej-870	341	31	in	in	ADP
alkej-870	341	32	signal	signal	ADJ
alkej-870	341	33	processing,2020	processing,2020	NOUN
alkej-870	341	34	.	.	PUNCT
alkej-870	342	1	http://dx.doi.org/10.1109/jstsp.2020.2980	http://dx.doi.org/10.1109/jstsp.2020.2980	ADJ
alkej-870	342	2	956	956	NUM
alkej-870	342	3	.	.	PUNCT
alkej-870	343	1	[	[	X
alkej-870	343	2	13	13	NUM
alkej-870	343	3	]	]	PUNCT
alkej-870	343	4	z.	z.	PROPN
alkej-870	343	5	zhang	zhang	PROPN
alkej-870	343	6	,	,	PUNCT
alkej-870	343	7	y.	y.	PROPN
alkej-870	343	8	xu	xu	PROPN
alkej-870	343	9	,	,	PUNCT
alkej-870	343	10	m.	m.	PROPN
alkej-870	343	11	yu	yu	PROPN
alkej-870	343	12	,	,	PUNCT
alkej-870	343	13	s.-x	s.-x	NOUN
alkej-870	343	14	.	.	PUNCT
alkej-870	344	1	zhang	zhang	PROPN
alkej-870	344	2	,	,	PUNCT
alkej-870	344	3	l.	l.	PROPN
alkej-870	344	4	chen	chen	PROPN
alkej-870	344	5	,	,	PUNCT
alkej-870	344	6	and	and	CCONJ
alkej-870	344	7	d.	d.	PROPN
alkej-870	344	8	yu	yu	PROPN
alkej-870	344	9	,	,	PUNCT
alkej-870	344	10	"	"	PUNCT
alkej-870	344	11	adlmvdr	adlmvdr	ADV
alkej-870	344	12	:	:	PUNCT
alkej-870	344	13	all	all	DET
alkej-870	344	14	deep	deep	ADJ
alkej-870	344	15	learning	learning	NOUN
alkej-870	344	16	mvdr	mvdr	AUX
alkej-870	344	17	beamformer	beamformer	VERB
alkej-870	344	18	for	for	ADP
alkej-870	344	19	target	target	NOUN
alkej-870	344	20	speech	speech	NOUN
alkej-870	344	21	separation	separation	NOUN
alkej-870	344	22	,	,	PUNCT
alkej-870	344	23	"	"	PUNCT
alkej-870	344	24	icassp	icassp	PROPN
alkej-870	344	25	,	,	PUNCT
alkej-870	344	26	pp	pp	ADV
alkej-870	344	27	.	.	PUNCT
alkej-870	345	1	6089	6089	NUM
alkej-870	345	2	–	–	PUNCT
alkej-870	345	3	6093	6093	NUM
alkej-870	345	4	,	,	PUNCT
alkej-870	345	5	2021	2021	NUM
alkej-870	345	6	.	.	PUNCT
alkej-870	346	1	http://dx.doi.org/10.1109/icassp39728.20	http://dx.doi.org/10.1109/icassp39728.20	NOUN
alkej-870	346	2	21.9413594	21.9413594	NUM
alkej-870	346	3	.	.	PUNCT
alkej-870	347	1	[	[	X
alkej-870	347	2	14	14	NUM
alkej-870	347	3	]	]	X
alkej-870	347	4	g.	g.	PROPN
alkej-870	347	5	li	li	PROPN
alkej-870	347	6	,	,	PUNCT
alkej-870	347	7	j.	j.	PROPN
alkej-870	347	8	yu	yu	PROPN
alkej-870	347	9	,	,	PUNCT
alkej-870	347	10	j.	j.	PROPN
alkej-870	347	11	deng	deng	PROPN
alkej-870	347	12	,	,	PUNCT
alkej-870	347	13	x.	x.	PROPN
alkej-870	347	14	liu	liu	PROPN
alkej-870	347	15	and	and	CCONJ
alkej-870	347	16	h.	h.	PROPN
alkej-870	347	17	meng	meng	PROPN
alkej-870	347	18	,	,	PUNCT
alkej-870	347	19	"	"	PUNCT
alkej-870	347	20	audio	audio	ADJ
alkej-870	347	21	-	-	ADJ
alkej-870	347	22	visual	visual	ADJ
alkej-870	347	23	multi	multi	ADJ
alkej-870	347	24	-	-	ADJ
alkej-870	347	25	channel	channel	NOUN
alkej-870	347	26	speech	speech	NOUN
alkej-870	347	27	separation	separation	NOUN
alkej-870	347	28	,	,	PUNCT
alkej-870	347	29	dereverberation	dereverberation	NOUN
alkej-870	347	30	and	and	CCONJ
alkej-870	347	31	recognition	recognition	NOUN
alkej-870	347	32	,	,	PUNCT
alkej-870	347	33	"	"	PUNCT
alkej-870	347	34	ieee	ieee	PROPN
alkej-870	347	35	international	international	ADJ
alkej-870	347	36	conference	conference	NOUN
alkej-870	347	37	on	on	ADP
alkej-870	347	38	acoustics	acoustic	NOUN
alkej-870	347	39	,	,	PUNCT
alkej-870	347	40	speech	speech	NOUN
alkej-870	347	41	and	and	CCONJ
alkej-870	347	42	signal	signal	NOUN
alkej-870	347	43	processing	processing	NOUN
alkej-870	347	44	(	(	PUNCT
alkej-870	347	45	icassp	icassp	PROPN
alkej-870	347	46	)	)	PUNCT
alkej-870	347	47	,	,	PUNCT
alkej-870	347	48	singapore	singapore	PROPN
alkej-870	347	49	,	,	PUNCT
alkej-870	347	50	2022	2022	NUM
alkej-870	347	51	,	,	PUNCT
alkej-870	347	52	pp	pp	ADJ
alkej-870	347	53	.	.	PUNCT
alkej-870	347	54	6042	6042	NUM
alkej-870	347	55	-	-	SYM
alkej-870	347	56	6046	6046	NUM
alkej-870	347	57	,	,	PUNCT
alkej-870	347	58	http://dx.doi.org/10.1109/icassp43922.20	http://dx.doi.org/10.1109/icassp43922.20	X
alkej-870	347	59	22.9747237	22.9747237	NUM
alkej-870	347	60	.	.	PUNCT
alkej-870	348	1	[	[	X
alkej-870	348	2	15	15	NUM
alkej-870	348	3	]	]	X
alkej-870	348	4	j.	j.	PROPN
alkej-870	348	5	ong	ong	PROPN
alkej-870	348	6	,	,	PUNCT
alkej-870	348	7	b.	b.	PROPN
alkej-870	348	8	t.	t.	PROPN
alkej-870	348	9	vo	vo	PROPN
alkej-870	348	10	,	,	PUNCT
alkej-870	348	11	s.	s.	PROPN
alkej-870	348	12	nordholm	nordholm	PROPN
alkej-870	348	13	,	,	PUNCT
alkej-870	348	14	b.	b.	PROPN
alkej-870	348	15	-n	-n	PROPN
alkej-870	348	16	.	.	PROPN
alkej-870	349	1	vo	vo	PROPN
alkej-870	349	2	,	,	PUNCT
alkej-870	349	3	d.	d.	PROPN
alkej-870	349	4	moratuwage	moratuwage	NOUN
alkej-870	349	5	and	and	CCONJ
alkej-870	349	6	c.	c.	NOUN
alkej-870	349	7	shim	shim	NOUN
alkej-870	349	8	,	,	PUNCT
alkej-870	349	9	"	"	PUNCT
alkej-870	349	10	audiovisual	audiovisual	ADJ
alkej-870	349	11	based	base	VERB
alkej-870	349	12	online	online	ADJ
alkej-870	349	13	multi	multi	ADJ
alkej-870	349	14	-	-	ADJ
alkej-870	349	15	source	source	NOUN
alkej-870	349	16	separation	separation	NOUN
alkej-870	349	17	,	,	PUNCT
alkej-870	349	18	"	"	PUNCT
alkej-870	349	19	in	in	ADP
alkej-870	349	20	ieee	ieee	NOUN
alkej-870	349	21	/	/	SYM
alkej-870	349	22	acm	acm	PROPN
alkej-870	349	23	transactions	transaction	NOUN
alkej-870	349	24	on	on	ADP
alkej-870	349	25	audio	audio	NOUN
alkej-870	349	26	,	,	PUNCT
alkej-870	349	27	speech	speech	NOUN
alkej-870	349	28	,	,	PUNCT
alkej-870	349	29	and	and	CCONJ
alkej-870	349	30	language	language	NOUN
alkej-870	349	31	processing	processing	NOUN
alkej-870	349	32	,	,	PUNCT
alkej-870	349	33	vol	vol	NOUN
alkej-870	349	34	.	.	PROPN
alkej-870	349	35	30	30	NUM
alkej-870	349	36	,	,	PUNCT
alkej-870	349	37	pp	pp	ADJ
alkej-870	349	38	.	.	PUNCT
alkej-870	350	1	1219	1219	NUM
alkej-870	350	2	-	-	SYM
alkej-870	350	3	1234	1234	NUM
alkej-870	350	4	,	,	PUNCT
alkej-870	350	5	2022	2022	NUM
alkej-870	350	6	.	.	PUNCT
alkej-870	351	1	http://dx.doi.org/10.1109/taslp.2022.315	http://dx.doi.org/10.1109/taslp.2022.315	NOUN
alkej-870	351	2	6758	6758	NUM
alkej-870	351	3	.	.	PUNCT
alkej-870	352	1	[	[	X
alkej-870	352	2	16	16	NUM
alkej-870	352	3	]	]	PUNCT
alkej-870	352	4	t.	t.	PROPN
alkej-870	352	5	afouras	afouras	PROPN
alkej-870	352	6	,	,	PUNCT
alkej-870	352	7	j.	j.	PROPN
alkej-870	352	8	s.	s.	PROPN
alkej-870	352	9	chung	chung	PROPN
alkej-870	352	10	,	,	PUNCT
alkej-870	352	11	and	and	CCONJ
alkej-870	352	12	a.	a.	NOUN
alkej-870	352	13	zisserman	zisserman	PROPN
alkej-870	352	14	,	,	PUNCT
alkej-870	352	15	"	"	PUNCT
alkej-870	352	16	the	the	DET
alkej-870	352	17	conversation	conversation	NOUN
alkej-870	352	18	:	:	PUNCT
alkej-870	352	19	deep	deep	ADJ
alkej-870	352	20	audio	audio	ADJ
alkej-870	352	21	-	-	ADJ
alkej-870	352	22	visual	visual	ADJ
alkej-870	352	23	speech	speech	NOUN
alkej-870	352	24	enhancement,"2018	enhancement,"2018	PROPN
alkej-870	352	25	.	.	PUNCT
alkej-870	353	1	http://dx.doi.org/10.21437/interspeech.2018	http://dx.doi.org/10.21437/interspeech.2018	NOUN
alkej-870	354	1	-1400	-1400	INTJ
alkej-870	354	2	.	.	PUNCT
alkej-870	355	1	[	[	X
alkej-870	355	2	17	17	NUM
alkej-870	355	3	]	]	PUNCT
alkej-870	355	4	a.	a.	NOUN
alkej-870	355	5	ephrat	ephrat	PROPN
alkej-870	355	6	,	,	PUNCT
alkej-870	355	7	i.	i.	PROPN
alkej-870	355	8	mosseri	mosseri	PROPN
alkej-870	355	9	,	,	PUNCT
alkej-870	355	10	o.	o.	PROPN
alkej-870	355	11	lang	lang	PROPN
alkej-870	355	12	,	,	PUNCT
alkej-870	355	13	t.	t.	PROPN
alkej-870	355	14	dekel	dekel	PROPN
alkej-870	355	15	,	,	PUNCT
alkej-870	355	16	k.	k.	PROPN
alkej-870	355	17	wilson	wilson	PROPN
alkej-870	355	18	,	,	PUNCT
alkej-870	355	19	a.	a.	NOUN
alkej-870	355	20	hassidim	hassidim	PROPN
alkej-870	355	21	,	,	PUNCT
alkej-870	355	22	w.	w.	PROPN
alkej-870	355	23	t.	t.	PROPN
alkej-870	355	24	freeman	freeman	PROPN
alkej-870	355	25	,	,	PUNCT
alkej-870	355	26	and	and	CCONJ
alkej-870	355	27	m.	m.	PROPN
alkej-870	355	28	rubinstein	rubinstein	PROPN
alkej-870	355	29	,	,	PUNCT
alkej-870	355	30	"	"	PUNCT
alkej-870	355	31	looking	look	VERB
alkej-870	355	32	to	to	PART
alkej-870	355	33	listen	listen	VERB
alkej-870	355	34	at	at	ADP
alkej-870	355	35	the	the	DET
alkej-870	355	36	cocktail	cocktail	NOUN
alkej-870	355	37	party	party	NOUN
alkej-870	355	38	:	:	PUNCT
alkej-870	355	39	a	a	DET
alkej-870	355	40	speaker	speaker	NOUN
alkej-870	355	41	-	-	PUNCT
alkej-870	355	42	independent	independent	ADJ
alkej-870	355	43	audiovisual	audiovisual	ADJ
alkej-870	355	44	model	model	NOUN
alkej-870	355	45	for	for	ADP
alkej-870	355	46	speech	speech	NOUN
alkej-870	355	47	separation	separation	NOUN
alkej-870	355	48	,	,	PUNCT
alkej-870	355	49	"	"	PUNCT
alkej-870	355	50	acm	acm	PROPN
alkej-870	355	51	trans	trans	PROPN
alkej-870	355	52	.	.	PUNCT
alkej-870	355	53	graph	graph	NOUN
alkej-870	355	54	.	.	PUNCT
alkej-870	355	55	,	,	PUNCT
alkej-870	355	56	pp	pp	ADJ
alkej-870	355	57	.	.	PUNCT
alkej-870	356	1	112:1–112:11	112:1–112:11	NUM
alkej-870	356	2	,	,	PUNCT
alkej-870	356	3	2018	2018	NUM
alkej-870	356	4	.	.	PUNCT
alkej-870	357	1	http://dx.doi.org/10.1145/3197517.3201357	http://dx.doi.org/10.1145/3197517.3201357	ADJ
alkej-870	357	2	.	.	PUNCT
alkej-870	358	1	[	[	X
alkej-870	358	2	18	18	NUM
alkej-870	358	3	]	]	X
alkej-870	358	4	jen	jen	PROPN
alkej-870	358	5	-	-	PUNCT
alkej-870	358	6	cheng	cheng	PROPN
alkej-870	358	7	hou	hou	PROPN
alkej-870	358	8	,	,	PUNCT
alkej-870	358	9	syu	syu	PROPN
alkej-870	358	10	-	-	PUNCT
alkej-870	358	11	siang	siang	PROPN
alkej-870	358	12	wang	wang	PROPN
alkej-870	358	13	,	,	PUNCT
alkej-870	358	14	yinghui	yinghui	PROPN
alkej-870	358	15	lai	lai	PROPN
alkej-870	358	16	,	,	PUNCT
alkej-870	358	17	yu	yu	PROPN
alkej-870	358	18	tsao	tsao	PROPN
alkej-870	358	19	,	,	PUNCT
alkej-870	358	20	hsiu	hsiu	PROPN
alkej-870	358	21	-	-	PUNCT
alkej-870	358	22	wen	wen	PROPN
alkej-870	358	23	chang	chang	PROPN
alkej-870	358	24	,	,	PUNCT
alkej-870	358	25	and	and	CCONJ
alkej-870	358	26	hsin	hsin	PROPN
alkej-870	358	27	-	-	PUNCT
alkej-870	358	28	min	min	PROPN
alkej-870	358	29	wang	wang	PROPN
alkej-870	358	30	,	,	PUNCT
alkej-870	358	31	"	"	PUNCT
alkej-870	358	32	audio	audio	ADJ
alkej-870	358	33	-	-	ADJ
alkej-870	358	34	visual	visual	ADJ
alkej-870	358	35	speech	speech	NOUN
alkej-870	358	36	enhancement	enhancement	NOUN
alkej-870	358	37	using	use	VERB
alkej-870	358	38	multimodal	multimodal	NOUN
alkej-870	358	39	deep	deep	ADJ
alkej-870	358	40	convolutional	convolutional	ADJ
alkej-870	358	41	neural	neural	ADJ
alkej-870	358	42	networks	network	NOUN
alkej-870	358	43	,	,	PUNCT
alkej-870	358	44	"	"	PUNCT
alkej-870	358	45	ieee	ieee	NOUN
alkej-870	358	46	transactions	transaction	NOUN
alkej-870	358	47	on	on	ADP
alkej-870	358	48	emerging	emerge	VERB
alkej-870	358	49	topics	topic	NOUN
alkej-870	358	50	in	in	ADP
alkej-870	358	51	computational	computational	ADJ
alkej-870	358	52	intelligence	intelligence	NOUN
alkej-870	358	53	,	,	PUNCT
alkej-870	358	54	vol	vol	NOUN
alkej-870	358	55	.	.	PROPN
alkej-870	358	56	2	2	NUM
alkej-870	358	57	,	,	PUNCT
alkej-870	358	58	no	no	INTJ
alkej-870	358	59	.	.	NOUN
alkej-870	358	60	2	2	NUM
alkej-870	358	61	,	,	PUNCT
alkej-870	358	62	pp	pp	ADJ
alkej-870	358	63	.	.	PUNCT
alkej-870	359	1	117–128	117–128	NUM
alkej-870	359	2	,	,	PUNCT
alkej-870	359	3	2018	2018	NUM
alkej-870	359	4	.	.	PUNCT
alkej-870	360	1	http://dx.doi.org/10.1109/tetci.2017.2784	http://dx.doi.org/10.1109/tetci.2017.2784	PROPN
alkej-870	360	2	878	878	NUM
alkej-870	360	3	.	.	PUNCT
alkej-870	361	1	[	[	X
alkej-870	361	2	19	19	NUM
alkej-870	361	3	]	]	PUNCT
alkej-870	361	4	r.	r.	PROPN
alkej-870	361	5	lu	lu	PROPN
alkej-870	361	6	,	,	PUNCT
alkej-870	361	7	z.	z.	PROPN
alkej-870	361	8	duan	duan	PROPN
alkej-870	361	9	,	,	PUNCT
alkej-870	361	10	and	and	CCONJ
alkej-870	361	11	c.	c.	PROPN
alkej-870	361	12	zhang	zhang	PROPN
alkej-870	361	13	,	,	PUNCT
alkej-870	361	14	"	"	PUNCT
alkej-870	361	15	listen	listen	VERB
alkej-870	361	16	and	and	CCONJ
alkej-870	361	17	look	look	VERB
alkej-870	361	18	:	:	PUNCT
alkej-870	361	19	audio	audio	ADJ
alkej-870	361	20	–	–	PUNCT
alkej-870	361	21	visual	visual	ADJ
alkej-870	361	22	matching	matching	NOUN
alkej-870	361	23	assisted	assist	VERB
alkej-870	361	24	speech	speech	NOUN
alkej-870	361	25	source	source	NOUN
alkej-870	361	26	separation	separation	NOUN
alkej-870	361	27	,	,	PUNCT
alkej-870	361	28	"	"	PUNCT
alkej-870	361	29	ieee	ieee	NOUN
alkej-870	361	30	signal	signal	NOUN
alkej-870	361	31	processing	processing	NOUN
alkej-870	361	32	letters	letter	NOUN
alkej-870	361	33	,	,	PUNCT
alkej-870	361	34	vol	vol	NOUN
alkej-870	361	35	.	.	PROPN
alkej-870	361	36	25	25	NUM
alkej-870	361	37	,	,	PUNCT
alkej-870	361	38	no	no	INTJ
alkej-870	361	39	.	.	NOUN
alkej-870	361	40	9	9	NUM
alkej-870	361	41	,	,	PUNCT
alkej-870	361	42	pp	pp	ADJ
alkej-870	361	43	.	.	PUNCT
alkej-870	362	1	1315	1315	NUM
alkej-870	362	2	–	–	PUNCT
alkej-870	362	3	1319,2018	1319,2018	NUM
alkej-870	362	4	.	.	PUNCT
alkej-870	363	1	http://dx.doi.org/10.1109/lsp.2018.285356	http://dx.doi.org/10.1109/lsp.2018.285356	NOUN
alkej-870	363	2	6	6	NUM
alkej-870	363	3	.	.	PUNCT
alkej-870	364	1	[	[	X
alkej-870	364	2	20	20	NUM
alkej-870	364	3	]	]	PUNCT
alkej-870	364	4	h.	h.	PROPN
alkej-870	364	5	zhao	zhao	PROPN
alkej-870	364	6	,	,	PUNCT
alkej-870	364	7	c.	c.	PROPN
alkej-870	364	8	gan	gan	PROPN
alkej-870	364	9	,	,	PUNCT
alkej-870	364	10	a.	a.	NOUN
alkej-870	364	11	rouditchenko	rouditchenko	PROPN
alkej-870	364	12	,	,	PUNCT
alkej-870	364	13	c.	c.	PROPN
alkej-870	364	14	vondrick	vondrick	PROPN
alkej-870	364	15	,	,	PUNCT
alkej-870	364	16	j.	j.	PROPN
alkej-870	364	17	mcdermott	mcdermott	PROPN
alkej-870	364	18	,	,	PUNCT
alkej-870	364	19	and	and	CCONJ
alkej-870	364	20	a.	a.	NOUN
alkej-870	364	21	torralba	torralba	PROPN
alkej-870	364	22	,	,	PUNCT
alkej-870	364	23	"	"	PUNCT
alkej-870	364	24	the	the	DET
alkej-870	364	25	sound	sound	NOUN
alkej-870	364	26	of	of	ADP
alkej-870	364	27	pixels	pixel	NOUN
alkej-870	364	28	,	,	PUNCT
alkej-870	364	29	"	"	PUNCT
alkej-870	364	30	in	in	ADP
alkej-870	364	31	proc	proc	NOUN
alkej-870	364	32	.	.	PUNCT
alkej-870	364	33	of	of	ADP
alkej-870	364	34	eccv	eccv	ADJ
alkej-870	364	35	,	,	PUNCT
alkej-870	364	36	2018	2018	NUM
alkej-870	364	37	.	.	PUNCT
alkej-870	365	1	http://dx.doi.org/10.1007/978-3-03001246-5_35	http://dx.doi.org/10.1007/978-3-03001246-5_35	NOUN
alkej-870	365	2	.	.	PUNCT
alkej-870	366	1	[	[	X
alkej-870	366	2	21	21	NUM
alkej-870	366	3	]	]	PUNCT
alkej-870	366	4	a.	a.	NOUN
alkej-870	366	5	gabbay	gabbay	PROPN
alkej-870	366	6	,	,	PUNCT
alkej-870	366	7	a.	a.	NOUN
alkej-870	366	8	ephrat	ephrat	PROPN
alkej-870	366	9	,	,	PUNCT
alkej-870	366	10	t.	t.	PROPN
alkej-870	366	11	halperin	halperin	PROPN
alkej-870	366	12	,	,	PUNCT
alkej-870	366	13	and	and	CCONJ
alkej-870	366	14	s.	s.	PROPN
alkej-870	366	15	peleg	peleg	PROPN
alkej-870	366	16	,	,	PUNCT
alkej-870	366	17	"	"	PUNCT
alkej-870	366	18	seeing	see	VERB
alkej-870	366	19	through	through	ADP
alkej-870	366	20	noise	noise	NOUN
alkej-870	366	21	:	:	PUNCT
alkej-870	366	22	visually	visually	ADV
alkej-870	366	23	driven	drive	VERB
alkej-870	366	24	speaker	speaker	NOUN
alkej-870	366	25	separation	separation	NOUN
alkej-870	366	26	and	and	CCONJ
alkej-870	366	27	enhancement	enhancement	NOUN
alkej-870	366	28	,	,	PUNCT
alkej-870	366	29	"	"	PUNCT
alkej-870	366	30	in	in	ADP
alkej-870	366	31	proc	proc	NOUN
alkej-870	366	32	.	.	PUNCT
alkej-870	366	33	of	of	ADP
alkej-870	366	34	icassp	icassp	PROPN
alkej-870	366	35	,	,	PUNCT
alkej-870	366	36	2018	2018	NUM
alkej-870	366	37	.	.	PUNCT
alkej-870	367	1	http://dx.doi.org/10.1109/icassp.2018.84	http://dx.doi.org/10.1109/icassp.2018.84	NUM
alkej-870	367	2	62527	62527	NUM
alkej-870	367	3	.	.	PUNCT
alkej-870	368	1	[	[	X
alkej-870	368	2	22	22	NUM
alkej-870	368	3	]	]	PUNCT
alkej-870	368	4	m.	m.	NOUN
alkej-870	368	5	gogate	gogate	NOUN
alkej-870	368	6	,	,	PUNCT
alkej-870	368	7	a.	a.	PROPN
alkej-870	368	8	adeel	adeel	PROPN
alkej-870	368	9	,	,	PUNCT
alkej-870	368	10	r.	r.	PROPN
alkej-870	368	11	marxer	marxer	PROPN
alkej-870	368	12	,	,	PUNCT
alkej-870	368	13	j.	j.	PROPN
alkej-870	368	14	barker	barker	PROPN
alkej-870	368	15	,	,	PUNCT
alkej-870	368	16	and	and	CCONJ
alkej-870	368	17	a.	a.	NOUN
alkej-870	368	18	hussain	hussain	PROPN
alkej-870	368	19	,	,	PUNCT
alkej-870	368	20	"	"	PUNCT
alkej-870	368	21	dnn	dnn	PROPN
alkej-870	368	22	driven	drive	VERB
alkej-870	368	23	speaker	speaker	NOUN
alkej-870	368	24	independent	independent	ADJ
alkej-870	368	25	audio	audio	ADJ
alkej-870	368	26	-	-	ADJ
alkej-870	368	27	visual	visual	ADJ
alkej-870	368	28	mask	mask	NOUN
alkej-870	368	29	estimation	estimation	NOUN
alkej-870	368	30	for	for	ADP
alkej-870	368	31	speech	speech	NOUN
alkej-870	368	32	separation	separation	NOUN
alkej-870	368	33	,	,	PUNCT
alkej-870	368	34	"	"	PUNCT
alkej-870	368	35	in	in	ADP
alkej-870	368	36	proc	proc	NOUN
alkej-870	368	37	.	.	PUNCT
alkej-870	369	1	of	of	ADP
alkej-870	369	2	interspeech	interspeech	NOUN
alkej-870	369	3	,	,	PUNCT
alkej-870	369	4	2018	2018	NUM
alkej-870	369	5	.	.	PUNCT
alkej-870	370	1	http://dx.doi.org/10.21437/interspeech.2018	http://dx.doi.org/10.21437/interspeech.2018	NOUN
alkej-870	370	2	-2516	-2516	X
alkej-870	370	3	.	.	PUNCT
alkej-870	371	1	[	[	X
alkej-870	371	2	23	23	NUM
alkej-870	371	3	]	]	X
alkej-870	371	4	g.	g.	PROPN
alkej-870	371	5	morrone	morrone	PROPN
alkej-870	371	6	,	,	PUNCT
alkej-870	371	7	s.	s.	PROPN
alkej-870	371	8	bergamaschi	bergamaschi	PROPN
alkej-870	371	9	,	,	PUNCT
alkej-870	371	10	l.	l.	PROPN
alkej-870	371	11	pasa	pasa	PROPN
alkej-870	371	12	,	,	PUNCT
alkej-870	371	13	l.	l.	PROPN
alkej-870	371	14	fadiga	fadiga	ADV
alkej-870	371	15	,	,	PUNCT
alkej-870	371	16	v.	v.	ADP
alkej-870	371	17	tikhanoff	tikhanoff	NOUN
alkej-870	371	18	,	,	PUNCT
alkej-870	371	19	and	and	CCONJ
alkej-870	371	20	l.	l.	PROPN
alkej-870	371	21	badino	badino	PROPN
alkej-870	371	22	,	,	PUNCT
alkej-870	371	23	"	"	PUNCT
alkej-870	371	24	face	face	VERB
alkej-870	371	25	landmark	landmark	NOUN
alkej-870	371	26	-	-	PUNCT
alkej-870	371	27	based	base	VERB
alkej-870	371	28	speaker	speaker	NOUN
alkej-870	371	29	-	-	PUNCT
alkej-870	371	30	independent	independent	ADJ
alkej-870	371	31	audio	audio	ADJ
alkej-870	371	32	-	-	ADJ
alkej-870	371	33	visual	visual	ADJ
alkej-870	371	34	speech	speech	NOUN
alkej-870	371	35	enhancement	enhancement	NOUN
alkej-870	371	36	in	in	ADP
alkej-870	371	37	multitalker	multitalker	NOUN
alkej-870	371	38	environments	environment	NOUN
alkej-870	371	39	,	,	PUNCT
alkej-870	371	40	"	"	PUNCT
alkej-870	371	41	in	in	ADP
alkej-870	371	42	ieee	ieee	PROPN
alkej-870	371	43	international	international	ADJ
alkej-870	371	44	conference	conference	NOUN
alkej-870	371	45	on	on	ADP
alkej-870	371	46	acoustics	acoustic	NOUN
alkej-870	371	47	,	,	PUNCT
alkej-870	371	48	speech	speech	NOUN
alkej-870	371	49	and	and	CCONJ
alkej-870	371	50	signal	signal	NOUN
alkej-870	371	51	processing	processing	NOUN
alkej-870	371	52	(	(	PUNCT
alkej-870	371	53	icassp	icassp	PROPN
alkej-870	371	54	)	)	PUNCT
alkej-870	371	55	,	,	PUNCT
alkej-870	371	56	2019	2019	NUM
alkej-870	371	57	,	,	PUNCT
alkej-870	371	58	pp	pp	ADJ
alkej-870	371	59	.	.	PUNCT
alkej-870	372	1	6900–6904	6900–6904	NUM
alkej-870	372	2	.	.	PUNCT
alkej-870	373	1	http://dx.doi.org/10.1109/icassp.2019.86	http://dx.doi.org/10.1109/icassp.2019.86	NUM
alkej-870	373	2	82061	82061	NUM
alkej-870	373	3	.	.	PUNCT
alkej-870	374	1	[	[	X
alkej-870	374	2	24	24	NUM
alkej-870	374	3	]	]	PUNCT
alkej-870	374	4	j.	j.	PROPN
alkej-870	374	5	wu	wu	PROPN
alkej-870	374	6	,	,	PUNCT
alkej-870	374	7	y.	y.	PROPN
alkej-870	374	8	xu	xu	PROPN
alkej-870	374	9	,	,	PUNCT
alkej-870	374	10	s.	s.	PROPN
alkej-870	374	11	zhang	zhang	PROPN
alkej-870	374	12	,	,	PUNCT
alkej-870	374	13	l.	l.	PROPN
alkej-870	374	14	chen	chen	PROPN
alkej-870	374	15	,	,	PUNCT
alkej-870	374	16	m.	m.	PROPN
alkej-870	374	17	yu	yu	PROPN
alkej-870	374	18	,	,	PUNCT
alkej-870	374	19	l.	l.	PROPN
alkej-870	374	20	xie	xie	PROPN
alkej-870	374	21	,	,	PUNCT
alkej-870	374	22	and	and	CCONJ
alkej-870	374	23	d.	d.	PROPN
alkej-870	374	24	yu	yu	PROPN
alkej-870	374	25	,	,	PUNCT
alkej-870	374	26	"	"	PUNCT
alkej-870	374	27	time	time	NOUN
alkej-870	374	28	domain	domain	NOUN
alkej-870	374	29	audio	audio	ADJ
alkej-870	374	30	visual	visual	ADJ
alkej-870	374	31	speech	speech	NOUN
alkej-870	374	32	separation	separation	NOUN
alkej-870	374	33	,	,	PUNCT
alkej-870	374	34	"	"	PUNCT
alkej-870	374	35	in	in	ADP
alkej-870	374	36	proc	proc	NOUN
alkej-870	374	37	.	.	PUNCT
alkej-870	375	1	ieee	ieee	PROPN
alkej-870	375	2	autom	autom	PROPN
alkej-870	375	3	.	.	PUNCT
alkej-870	376	1	speech	speech	NOUN
alkej-870	376	2	recognit	recognit	VERB
alkej-870	376	3	.	.	PUNCT
alkej-870	377	1	understanding	understand	VERB
alkej-870	377	2	workshop	workshop	NOUN
alkej-870	377	3	,	,	PUNCT
alkej-870	377	4	2019	2019	NUM
alkej-870	377	5	,	,	PUNCT
alkej-870	377	6	pp	pp	ADJ
alkej-870	377	7	.	.	PUNCT
alkej-870	378	1	667–673	667–673	X
alkej-870	378	2	.	.	NOUN
alkej-870	378	3	http://dx.doi.org/10.1109/asru46091.201	http://dx.doi.org/10.1109/asru46091.201	PROPN
alkej-870	378	4	9.9003983	9.9003983	NUM
alkej-870	378	5	.	.	PUNCT
alkej-870	379	1	[	[	X
alkej-870	379	2	25	25	NUM
alkej-870	379	3	]	]	PUNCT
alkej-870	379	4	mandar	mandar	PROPN
alkej-870	379	5	gogate	gogate	PROPN
alkej-870	379	6	et	et	PROPN
alkej-870	379	7	al	al	PROPN
alkej-870	379	8	.	.	PUNCT
alkej-870	380	1	"	"	PUNCT
alkej-870	380	2	deep	deep	ADJ
alkej-870	380	3	neural	neural	ADJ
alkej-870	380	4	network	network	NOUN
alkej-870	380	5	driven	drive	VERB
alkej-870	380	6	binaural	binaural	ADJ
alkej-870	380	7	audio	audio	ADJ
alkej-870	380	8	visual	visual	ADJ
alkej-870	380	9	http://dx.doi.org/10.1109/icassp40776.2020.9053845	http://dx.doi.org/10.1109/icassp40776.2020.9053845	PROPN
alkej-870	380	10	http://dx.doi.org/10.1109/icassp40776.2020.9053845	http://dx.doi.org/10.1109/icassp40776.2020.9053845	PROPN
alkej-870	380	11	http://dx.doi.org/10.21437/interspeech.2020-1065	http://dx.doi.org/10.21437/interspeech.2020-1065	PROPN
alkej-870	380	12	http://dx.doi.org/10.21437/interspeech.2020-1065	http://dx.doi.org/10.21437/interspeech.2020-1065	PROPN
alkej-870	381	1	http://dx.doi.org/10.1109/jstsp.2020.2980956	http://dx.doi.org/10.1109/jstsp.2020.2980956	PROPN
alkej-870	381	2	http://dx.doi.org/10.1109/jstsp.2020.2980956	http://dx.doi.org/10.1109/jstsp.2020.2980956	PROPN
alkej-870	381	3	http://dx.doi.org/10.1109/icassp39728.2021.9413594	http://dx.doi.org/10.1109/icassp39728.2021.9413594	PROPN
alkej-870	381	4	http://dx.doi.org/10.1109/icassp39728.2021.9413594	http://dx.doi.org/10.1109/icassp39728.2021.9413594	PROPN
alkej-870	381	5	http://dx.doi.org/10.1109/icassp43922.2022.9747237	http://dx.doi.org/10.1109/icassp43922.2022.9747237	PROPN
alkej-870	381	6	http://dx.doi.org/10.1109/icassp43922.2022.9747237	http://dx.doi.org/10.1109/icassp43922.2022.9747237	PROPN
alkej-870	381	7	http://dx.doi.org/10.1109/taslp.2022.3156758	http://dx.doi.org/10.1109/taslp.2022.3156758	PROPN
alkej-870	381	8	http://dx.doi.org/10.1109/taslp.2022.3156758	http://dx.doi.org/10.1109/taslp.2022.3156758	PROPN
alkej-870	381	9	http://dx.doi.org/10.21437/interspeech.2018-1400	http://dx.doi.org/10.21437/interspeech.2018-1400	PROPN
alkej-870	381	10	http://dx.doi.org/10.21437/interspeech.2018-1400	http://dx.doi.org/10.21437/interspeech.2018-1400	PROPN
alkej-870	382	1	http://dx.doi.org/10.1145/3197517.3201357	http://dx.doi.org/10.1145/3197517.3201357	PROPN
alkej-870	382	2	http://dx.doi.org/10.1145/3197517.3201357	http://dx.doi.org/10.1145/3197517.3201357	PROPN
alkej-870	382	3	http://dx.doi.org/10.1109/tetci.2017.2784878	http://dx.doi.org/10.1109/tetci.2017.2784878	ADJ
alkej-870	382	4	http://dx.doi.org/10.1109/tetci.2017.2784878	http://dx.doi.org/10.1109/tetci.2017.2784878	ADJ
alkej-870	382	5	http://dx.doi.org/10.1109/lsp.2018.2853566	http://dx.doi.org/10.1109/lsp.2018.2853566	INTJ
alkej-870	382	6	http://dx.doi.org/10.1109/lsp.2018.2853566	http://dx.doi.org/10.1109/lsp.2018.2853566	INTJ
alkej-870	382	7	http://dx.doi.org/10.1007/978-3-030-01246-5_35	http://dx.doi.org/10.1007/978-3-030-01246-5_35	PROPN
alkej-870	382	8	http://dx.doi.org/10.1007/978-3-030-01246-5_35	http://dx.doi.org/10.1007/978-3-030-01246-5_35	PROPN
alkej-870	382	9	http://dx.doi.org/10.1109/icassp.2018.8462527	http://dx.doi.org/10.1109/icassp.2018.8462527	PROPN
alkej-870	382	10	http://dx.doi.org/10.1109/icassp.2018.8462527	http://dx.doi.org/10.1109/icassp.2018.8462527	PROPN
alkej-870	382	11	http://dx.doi.org/10.21437/interspeech.2018-2516	http://dx.doi.org/10.21437/interspeech.2018-2516	VERB
alkej-870	382	12	http://dx.doi.org/10.21437/interspeech.2018-2516	http://dx.doi.org/10.21437/interspeech.2018-2516	PROPN
alkej-870	382	13	http://dx.doi.org/10.1109/icassp.2019.8682061	http://dx.doi.org/10.1109/icassp.2019.8682061	AUX
alkej-870	382	14	http://dx.doi.org/10.1109/icassp.2019.8682061	http://dx.doi.org/10.1109/icassp.2019.8682061	VERB
alkej-870	382	15	http://dx.doi.org/10.1109/asru46091.2019.9003983	http://dx.doi.org/10.1109/asru46091.2019.9003983	PROPN
alkej-870	382	16	http://dx.doi.org/10.1109/asru46091.2019.9003983	http://dx.doi.org/10.1109/asru46091.2019.9003983	PROPN
alkej-870	382	17	nooralhuda	nooralhuda	PROPN
alkej-870	382	18	mudhafar	mudhafar	PROPN
alkej-870	382	19	sleman	sleman	PROPN
alkej-870	382	20	al	al	PROPN
alkej-870	382	21	-	-	PUNCT
alkej-870	382	22	khwarizmi	khwarizmi	PROPN
alkej-870	382	23	engineering	engineering	NOUN
alkej-870	382	24	journal	journal	PROPN
alkej-870	382	25	,	,	PUNCT
alkej-870	382	26	vol	vol	NOUN
alkej-870	382	27	.	.	PROPN
alkej-870	382	28	19	19	NUM
alkej-870	382	29	,	,	PUNCT
alkej-870	382	30	no	no	INTJ
alkej-870	382	31	.	.	NOUN
alkej-870	382	32	4	4	NUM
alkej-870	382	33	,	,	PUNCT
alkej-870	382	34	p.p	p.p	PROPN
alkej-870	382	35	.	.	NOUN
alkej-870	382	36	4255	4255	NUM
alkej-870	382	37	(	(	PUNCT
alkej-870	382	38	2023	2023	NUM
alkej-870	382	39	)	)	PUNCT
alkej-870	382	40	52	52	NUM
alkej-870	382	41	speech	speech	NOUN
alkej-870	382	42	separation	separation	NOUN
alkej-870	382	43	"	"	PUNCT
alkej-870	382	44	.	.	PUNCT
alkej-870	383	1	in	in	ADP
alkej-870	383	2	:	:	PUNCT
alkej-870	383	3	international	international	ADJ
alkej-870	383	4	joint	joint	ADJ
alkej-870	383	5	conference	conference	NOUN
alkej-870	383	6	on	on	ADP
alkej-870	383	7	neural	neural	ADJ
alkej-870	383	8	networks	network	NOUN
alkej-870	383	9	(	(	PUNCT
alkej-870	383	10	ijcnn	ijcnn	PROPN
alkej-870	383	11	)	)	PUNCT
alkej-870	383	12	.	.	PUNCT
alkej-870	384	1	ieee	ieee	PROPN
alkej-870	384	2	.	.	PUNCT
alkej-870	385	1	2020	2020	NUM
alkej-870	385	2	,	,	PUNCT
alkej-870	385	3	pp	pp	ADV
alkej-870	385	4	.	.	PUNCT
alkej-870	386	1	1–7	1–7	X
alkej-870	386	2	.	.	PUNCT
alkej-870	387	1	http://dx.doi.org/10.1109/ijcnn48605.202	http://dx.doi.org/10.1109/ijcnn48605.202	PROPN
alkej-870	387	2	0.9207517	0.9207517	NUM
alkej-870	387	3	.	.	PUNCT
alkej-870	388	1	[	[	X
alkej-870	388	2	26	26	NUM
alkej-870	388	3	]	]	PUNCT
alkej-870	388	4	q.	q.	PROPN
alkej-870	388	5	nguyen	nguyen	PROPN
alkej-870	388	6	,	,	PUNCT
alkej-870	388	7	j.	j.	PROPN
alkej-870	388	8	richter	richter	PROPN
alkej-870	388	9	,	,	PUNCT
alkej-870	388	10	m.	m.	NOUN
alkej-870	388	11	lauri	lauri	PROPN
alkej-870	388	12	,	,	PUNCT
alkej-870	388	13	t.	t.	PROPN
alkej-870	388	14	gerkmann	gerkmann	PROPN
alkej-870	388	15	and	and	CCONJ
alkej-870	388	16	s.	s.	PROPN
alkej-870	388	17	frintrop	frintrop	PROPN
alkej-870	388	18	,	,	PUNCT
alkej-870	388	19	"	"	PUNCT
alkej-870	388	20	improving	improve	VERB
alkej-870	388	21	mix	mix	NOUN
alkej-870	388	22	-	-	PUNCT
alkej-870	388	23	and	and	CCONJ
alkej-870	388	24	-	-	PUNCT
alkej-870	388	25	separate	separate	ADJ
alkej-870	388	26	training	training	NOUN
alkej-870	388	27	in	in	ADP
alkej-870	388	28	audio	audio	ADJ
alkej-870	388	29	-	-	ADJ
alkej-870	388	30	visual	visual	ADJ
alkej-870	388	31	sound	sound	NOUN
alkej-870	388	32	source	source	NOUN
alkej-870	388	33	separation	separation	NOUN
alkej-870	388	34	with	with	ADP
alkej-870	388	35	an	an	DET
alkej-870	388	36	object	object	NOUN
alkej-870	388	37	prior	prior	ADV
alkej-870	388	38	,	,	PUNCT
alkej-870	388	39	"	"	PUNCT
alkej-870	388	40	2020	2020	NUM
alkej-870	388	41	(	(	PUNCT
alkej-870	388	42	icpr	icpr	PROPN
alkej-870	388	43	)	)	PUNCT
alkej-870	388	44	.	.	PUNCT
alkej-870	389	1	http://dx.doi.org/10.1109/icpr48806.2021	http://dx.doi.org/10.1109/icpr48806.2021	PROPN
alkej-870	389	2	.	.	PUNCT
alkej-870	390	1	9412174	9412174	NUM
alkej-870	390	2	.	.	PUNCT
alkej-870	391	1	[	[	X
alkej-870	391	2	27	27	NUM
alkej-870	391	3	]	]	X
alkej-870	391	4	c.	c.	PROPN
alkej-870	391	5	li	li	PROPN
alkej-870	391	6	and	and	CCONJ
alkej-870	391	7	y.	y.	PROPN
alkej-870	391	8	qian	qian	PROPN
alkej-870	391	9	,	,	PUNCT
alkej-870	391	10	"	"	PUNCT
alkej-870	391	11	deep	deep	ADJ
alkej-870	391	12	audio	audio	ADJ
alkej-870	391	13	-	-	ADJ
alkej-870	391	14	visual	visual	ADJ
alkej-870	391	15	speech	speech	NOUN
alkej-870	391	16	separation	separation	NOUN
alkej-870	391	17	with	with	ADP
alkej-870	391	18	attention	attention	NOUN
alkej-870	391	19	mechanism	mechanism	NOUN
alkej-870	391	20	,	,	PUNCT
alkej-870	391	21	"	"	PUNCT
alkej-870	391	22	ieee	ieee	PROPN
alkej-870	391	23	international	international	ADJ
alkej-870	391	24	conference	conference	NOUN
alkej-870	391	25	on	on	ADP
alkej-870	391	26	acoustics	acoustic	NOUN
alkej-870	391	27	,	,	PUNCT
alkej-870	391	28	speech	speech	NOUN
alkej-870	391	29	and	and	CCONJ
alkej-870	391	30	signal	signal	NOUN
alkej-870	391	31	processing	processing	NOUN
alkej-870	391	32	(	(	PUNCT
alkej-870	391	33	icassp	icassp	PROPN
alkej-870	391	34	)	)	PUNCT
alkej-870	391	35	,	,	PUNCT
alkej-870	391	36	barcelona	barcelona	PROPN
alkej-870	391	37	,	,	PUNCT
alkej-870	391	38	spain	spain	PROPN
alkej-870	391	39	,	,	PUNCT
alkej-870	391	40	2020	2020	NUM
alkej-870	391	41	,	,	PUNCT
alkej-870	391	42	pp	pp	ADV
alkej-870	391	43	.	.	PUNCT
alkej-870	391	44	7314	7314	NUM
alkej-870	391	45	-	-	SYM
alkej-870	391	46	7318	7318	NUM
alkej-870	391	47	,	,	PUNCT
alkej-870	391	48	http://dx.doi.org/10.1109/icassp40776.20	http://dx.doi.org/10.1109/icassp40776.20	PROPN
alkej-870	391	49	20.9054180	20.9054180	NUM
alkej-870	391	50	.	.	PUNCT
alkej-870	392	1	[	[	X
alkej-870	392	2	28	28	NUM
alkej-870	392	3	]	]	X
alkej-870	392	4	r.	r.	PROPN
alkej-870	392	5	gu	gu	PROPN
alkej-870	392	6	et	et	PROPN
alkej-870	392	7	al	al	PROPN
alkej-870	392	8	.	.	PROPN
alkej-870	392	9	,	,	PUNCT
alkej-870	392	10	"	"	PUNCT
alkej-870	392	11	multi	multi	ADJ
alkej-870	392	12	-	-	ADJ
alkej-870	392	13	modal	modal	ADJ
alkej-870	392	14	multi	multi	ADJ
alkej-870	392	15	-	-	ADJ
alkej-870	392	16	channel	channel	ADJ
alkej-870	392	17	target	target	NOUN
alkej-870	392	18	speech	speech	NOUN
alkej-870	392	19	separation	separation	NOUN
alkej-870	392	20	,	,	PUNCT
alkej-870	392	21	"	"	PUNCT
alkej-870	392	22	ieee	ieee	PROPN
alkej-870	392	23	j	j	PROPN
alkej-870	392	24	-	-	NOUN
alkej-870	392	25	stsp	stsp	NOUN
alkej-870	392	26	,	,	PUNCT
alkej-870	392	27	2020	2020	NUM
alkej-870	392	28	.	.	PUNCT
alkej-870	393	1	http://dx.doi.org/10.1109/jstsp.2020.2980	http://dx.doi.org/10.1109/jstsp.2020.2980	INTJ
alkej-870	393	2	956	956	NUM
alkej-870	393	3	.	.	PUNCT
alkej-870	394	1	[	[	X
alkej-870	394	2	29	29	NUM
alkej-870	394	3	]	]	X
alkej-870	394	4	c.	c.	PROPN
alkej-870	394	5	gan	gan	PROPN
alkej-870	394	6	,	,	PUNCT
alkej-870	394	7	d.	d.	PROPN
alkej-870	394	8	huang	huang	PROPN
alkej-870	394	9	,	,	PUNCT
alkej-870	394	10	h.	h.	PROPN
alkej-870	394	11	zhao	zhao	PROPN
alkej-870	394	12	,	,	PUNCT
alkej-870	394	13	j.	j.	PROPN
alkej-870	394	14	b.	b.	PROPN
alkej-870	394	15	tenenbaum	tenenbaum	PROPN
alkej-870	394	16	,	,	PUNCT
alkej-870	394	17	and	and	CCONJ
alkej-870	394	18	a.	a.	NOUN
alkej-870	394	19	torralba	torralba	PROPN
alkej-870	394	20	.	.	PUNCT
alkej-870	395	1	"	"	PUNCT
alkej-870	395	2	music	music	NOUN
alkej-870	395	3	gesture	gesture	NOUN
alkej-870	395	4	for	for	ADP
alkej-870	395	5	visual	visual	ADJ
alkej-870	395	6	sound	sound	NOUN
alkej-870	395	7	separation	separation	NOUN
alkej-870	395	8	"	"	PUNCT
alkej-870	395	9	.	.	PUNCT
alkej-870	396	1	in	in	ADP
alkej-870	396	2	ieee	ieee	NOUN
alkej-870	396	3	/	/	SYM
alkej-870	396	4	cvf	cvf	NOUN
alkej-870	396	5	conference	conference	NOUN
alkej-870	396	6	on	on	ADP
alkej-870	396	7	computer	computer	NOUN
alkej-870	396	8	vision	vision	NOUN
alkej-870	396	9	and	and	CCONJ
alkej-870	396	10	pattern	pattern	NOUN
alkej-870	396	11	recognition	recognition	NOUN
alkej-870	396	12	(	(	PUNCT
alkej-870	396	13	cvpr	cvpr	NOUN
alkej-870	396	14	)	)	PUNCT
alkej-870	396	15	,	,	PUNCT
alkej-870	396	16	pages	page	NOUN
alkej-870	396	17	10475–10484	10475–10484	NUM
alkej-870	396	18	,	,	PUNCT
alkej-870	396	19	2020	2020	NUM
alkej-870	396	20	.	.	PUNCT
alkej-870	397	1	http://dx.doi.org/10.1109/cvpr42600.2020	http://dx.doi.org/10.1109/cvpr42600.2020	PROPN
alkej-870	397	2	.01049	.01049	PROPN
alkej-870	397	3	.	.	PUNCT
alkej-870	398	1	[	[	X
alkej-870	398	2	30	30	NUM
alkej-870	398	3	]	]	PUNCT
alkej-870	398	4	lingyu	lingyu	NOUN
alkej-870	398	5	zhu	zhu	PROPN
alkej-870	398	6	and	and	CCONJ
alkej-870	398	7	esa	esa	PROPN
alkej-870	398	8	rahtu	rahtu	PROPN
alkej-870	398	9	.	.	PUNCT
alkej-870	398	10	,	,	PUNCT
alkej-870	398	11	"	"	PUNCT
alkej-870	398	12	visually	visually	ADV
alkej-870	398	13	guided	guide	VERB
alkej-870	398	14	sound	sound	NOUN
alkej-870	398	15	source	source	NOUN
alkej-870	398	16	separation	separation	NOUN
alkej-870	398	17	using	use	VERB
alkej-870	398	18	cascaded	cascade	VERB
alkej-870	398	19	opponent	opponent	NOUN
alkej-870	398	20	filter	filter	NOUN
alkej-870	398	21	network	network	NOUN
alkej-870	398	22	"	"	PUNCT
alkej-870	398	23	proc	proc	NOUN
alkej-870	398	24	.	.	PUNCT
alkej-870	399	1	of	of	ADP
alkej-870	399	2	accv	accv	PROPN
alkej-870	399	3	,	,	PUNCT
alkej-870	399	4	2020	2020	NUM
alkej-870	399	5	.	.	PUNCT
alkej-870	399	6	http://dx.doi.org/10.1007/978-3-030-695446_25	http://dx.doi.org/10.1007/978-3-030-695446_25	X
alkej-870	399	7	.	.	PUNCT
alkej-870	400	1	[	[	X
alkej-870	400	2	31	31	NUM
alkej-870	400	3	]	]	PUNCT
alkej-870	400	4	k.	k.	PROPN
alkej-870	400	5	tan	tan	PROPN
alkej-870	400	6	,	,	PUNCT
alkej-870	400	7	y.	y.	PROPN
alkej-870	400	8	xu	xu	PROPN
alkej-870	400	9	,	,	PUNCT
alkej-870	400	10	s.-x	s.-x	NOUN
alkej-870	400	11	.	.	PUNCT
alkej-870	401	1	zhang	zhang	PROPN
alkej-870	401	2	,	,	PUNCT
alkej-870	401	3	m.	m.	PROPN
alkej-870	401	4	yu	yu	PROPN
alkej-870	401	5	,	,	PUNCT
alkej-870	401	6	and	and	CCONJ
alkej-870	401	7	d.	d.	PROPN
alkej-870	401	8	yu	yu	PROPN
alkej-870	401	9	,	,	PUNCT
alkej-870	401	10	"	"	PUNCT
alkej-870	401	11	audio	audio	ADJ
alkej-870	401	12	-	-	ADJ
alkej-870	401	13	visual	visual	ADJ
alkej-870	401	14	speech	speech	NOUN
alkej-870	401	15	separation	separation	NOUN
alkej-870	401	16	and	and	CCONJ
alkej-870	401	17	dereverberation	dereverberation	NOUN
alkej-870	401	18	with	with	ADP
alkej-870	401	19	a	a	DET
alkej-870	401	20	two	two	NUM
alkej-870	401	21	-	-	PUNCT
alkej-870	401	22	stage	stage	NOUN
alkej-870	401	23	multimodal	multimodal	NOUN
alkej-870	401	24	network	network	NOUN
alkej-870	401	25	,	,	PUNCT
alkej-870	401	26	"	"	PUNCT
alkej-870	401	27	ieee	ieee	NOUN
alkej-870	401	28	journal	journal	NOUN
alkej-870	401	29	of	of	ADP
alkej-870	401	30	selected	select	VERB
alkej-870	401	31	topics	topic	NOUN
alkej-870	401	32	in	in	ADP
alkej-870	401	33	signal	signal	ADJ
alkej-870	401	34	processing	processing	NOUN
alkej-870	401	35	,	,	PUNCT
alkej-870	401	36	vol	vol	NOUN
alkej-870	401	37	.	.	PROPN
alkej-870	401	38	14	14	NUM
alkej-870	401	39	,	,	PUNCT
alkej-870	401	40	no	no	INTJ
alkej-870	401	41	.	.	NOUN
alkej-870	401	42	3	3	NUM
alkej-870	401	43	,	,	PUNCT
alkej-870	401	44	pp	pp	ADJ
alkej-870	401	45	.	.	PUNCT
alkej-870	402	1	542–553	542–553	NUM
alkej-870	402	2	,	,	PUNCT
alkej-870	402	3	2020	2020	NUM
alkej-870	402	4	.	.	PUNCT
alkej-870	403	1	http://dx.doi.org/10.1109/jstsp.2020.2987	http://dx.doi.org/10.1109/jstsp.2020.2987	PROPN
alkej-870	403	2	209	209	NUM
alkej-870	403	3	.	.	PUNCT
alkej-870	404	1	[	[	X
alkej-870	404	2	32	32	NUM
alkej-870	404	3	]	]	PUNCT
alkej-870	404	4	l.	l.	PROPN
alkej-870	404	5	qu	qu	PROPN
alkej-870	404	6	,	,	PUNCT
alkej-870	404	7	c.	c.	PROPN
alkej-870	404	8	weber	weber	PROPN
alkej-870	404	9	,	,	PUNCT
alkej-870	404	10	and	and	CCONJ
alkej-870	404	11	s.	s.	PROPN
alkej-870	404	12	wermter	wermter	PROPN
alkej-870	404	13	,	,	PUNCT
alkej-870	404	14	"	"	PUNCT
alkej-870	404	15	multimodal	multimodal	ADJ
alkej-870	404	16	target	target	NOUN
alkej-870	404	17	speech	speech	NOUN
alkej-870	404	18	separation	separation	NOUN
alkej-870	404	19	with	with	ADP
alkej-870	404	20	voice	voice	NOUN
alkej-870	404	21	and	and	CCONJ
alkej-870	404	22	face	face	NOUN
alkej-870	404	23	references	reference	NOUN
alkej-870	404	24	,	,	PUNCT
alkej-870	404	25	"	"	PUNCT
alkej-870	404	26	proc	proc	NOUN
alkej-870	404	27	.	.	PUNCT
alkej-870	404	28	of	of	ADP
alkej-870	404	29	interspeech	interspeech	NOUN
alkej-870	404	30	,	,	PUNCT
alkej-870	404	31	2020	2020	NUM
alkej-870	404	32	.	.	PUNCT
alkej-870	405	1	http://dx.doi.org/10.21437/interspeech.2020	http://dx.doi.org/10.21437/interspeech.2020	X
alkej-870	405	2	-1697	-1697	PUNCT
alkej-870	405	3	.	.	PUNCT
alkej-870	406	1	[	[	X
alkej-870	406	2	33	33	NUM
alkej-870	406	3	]	]	PUNCT
alkej-870	406	4	t.	t.	PROPN
alkej-870	406	5	rahman	rahman	PROPN
alkej-870	406	6	and	and	CCONJ
alkej-870	406	7	l.	l.	PROPN
alkej-870	406	8	sigal	sigal	PROPN
alkej-870	406	9	,	,	PUNCT
alkej-870	406	10	"	"	PUNCT
alkej-870	406	11	weaklysupervised	weaklysupervise	VERB
alkej-870	406	12	audio	audio	ADJ
alkej-870	406	13	-	-	ADJ
alkej-870	406	14	visual	visual	ADJ
alkej-870	406	15	sound	sound	NOUN
alkej-870	406	16	source	source	NOUN
alkej-870	406	17	detection	detection	NOUN
alkej-870	406	18	and	and	CCONJ
alkej-870	406	19	separation,"ieee	separation,"ieee	VERB
alkej-870	406	20	international	international	ADJ
alkej-870	406	21	conference	conference	NOUN
alkej-870	406	22	on	on	ADP
alkej-870	406	23	multimedia	multimedia	NOUN
alkej-870	406	24	and	and	CCONJ
alkej-870	406	25	expo	expo	NOUN
alkej-870	406	26	(	(	PUNCT
alkej-870	406	27	icme	icme	PROPN
alkej-870	406	28	)	)	PUNCT
alkej-870	406	29	,	,	PUNCT
alkej-870	406	30	shenzhen	shenzhen	PROPN
alkej-870	406	31	,	,	PUNCT
alkej-870	406	32	china	china	PROPN
alkej-870	406	33	,	,	PUNCT
alkej-870	406	34	2021	2021	NUM
alkej-870	406	35	,	,	PUNCT
alkej-870	406	36	pp	pp	ADJ
alkej-870	406	37	.	.	PUNCT
alkej-870	407	1	1	1	NUM
alkej-870	407	2	-	-	SYM
alkej-870	407	3	6	6	NUM
alkej-870	407	4	.	.	PUNCT
alkej-870	407	5	http://dx.doi.org/10.1109/icme51207.2021	http://dx.doi.org/10.1109/icme51207.2021	PROPN
alkej-870	407	6	.9428196	.9428196	PROPN
alkej-870	407	7	.	.	PUNCT
alkej-870	408	1	[	[	X
alkej-870	408	2	34	34	NUM
alkej-870	408	3	]	]	X
alkej-870	408	4	r.	r.	PROPN
alkej-870	408	5	gao	gao	PROPN
alkej-870	408	6	and	and	CCONJ
alkej-870	408	7	k.	k.	PROPN
alkej-870	408	8	grauman	grauman	NOUN
alkej-870	408	9	.	.	PUNCT
alkej-870	409	1	"	"	PUNCT
alkej-870	409	2	visualvoice	visualvoice	NOUN
alkej-870	409	3	:	:	PUNCT
alkej-870	409	4	audio	audio	ADJ
alkej-870	409	5	-	-	ADJ
alkej-870	409	6	visual	visual	ADJ
alkej-870	409	7	speech	speech	NOUN
alkej-870	409	8	separation	separation	NOUN
alkej-870	409	9	with	with	ADP
alkej-870	409	10	crossmodal	crossmodal	ADJ
alkej-870	409	11	consistency	consistency	NOUN
alkej-870	409	12	"	"	PUNCT
alkej-870	409	13	.	.	PUNCT
alkej-870	410	1	in	in	ADP
alkej-870	410	2	cvpr	cvpr	NOUN
alkej-870	410	3	,	,	PUNCT
alkej-870	410	4	2021	2021	NUM
alkej-870	410	5	.	.	PUNCT
alkej-870	410	6	8	8	NUM
alkej-870	410	7	,	,	PUNCT
alkej-870	410	8	45	45	NUM
alkej-870	410	9	.	.	PUNCT
alkej-870	410	10	http://dx.doi.org/10.1109/cvpr46437.2021	http://dx.doi.org/10.1109/cvpr46437.2021	PROPN
alkej-870	410	11	.01524	.01524	PROPN
alkej-870	410	12	.	.	PUNCT
alkej-870	411	1	[	[	X
alkej-870	411	2	35	35	NUM
alkej-870	411	3	]	]	X
alkej-870	411	4	majumder	majumder	NOUN
alkej-870	411	5	,	,	PUNCT
alkej-870	411	6	s.	s.	PROPN
alkej-870	411	7	,	,	PUNCT
alkej-870	411	8	al	al	PROPN
alkej-870	411	9	-	-	PUNCT
alkej-870	411	10	halah	halah	PROPN
alkej-870	411	11	,	,	PUNCT
alkej-870	411	12	z.	z.	PROPN
alkej-870	411	13	,	,	PUNCT
alkej-870	411	14	grauman	grauman	NOUN
alkej-870	411	15	,	,	PUNCT
alkej-870	411	16	k.	k.	PROPN
alkej-870	411	17	:	:	PUNCT
alkej-870	411	18	"	"	PUNCT
alkej-870	411	19	move2hear	move2hear	NOUN
alkej-870	411	20	:	:	PUNCT
alkej-870	411	21	active	active	ADJ
alkej-870	411	22	audio	audio	ADJ
alkej-870	411	23	-	-	ADJ
alkej-870	411	24	visual	visual	ADJ
alkej-870	411	25	source	source	NOUN
alkej-870	411	26	separation	separation	NOUN
alkej-870	411	27	"	"	PUNCT
alkej-870	411	28	.	.	PUNCT
alkej-870	412	1	in	in	ADP
alkej-870	412	2	:	:	PUNCT
alkej-870	412	3	iccv	iccv	NOUN
alkej-870	412	4	(	(	PUNCT
alkej-870	412	5	2021	2021	NUM
alkej-870	412	6	)	)	PUNCT
alkej-870	412	7	.	.	PUNCT
alkej-870	413	1	http://dx.doi.org/10.1109/iccv48922.2021	http://dx.doi.org/10.1109/iccv48922.2021	PROPN
alkej-870	413	2	.00034	.00034	PROPN
alkej-870	413	3	.	.	PUNCT
alkej-870	414	1	[	[	X
alkej-870	414	2	36	36	NUM
alkej-870	414	3	]	]	X
alkej-870	414	4	y.	y.	PROPN
alkej-870	414	5	liu	liu	PROPN
alkej-870	414	6	and	and	CCONJ
alkej-870	414	7	y.	y.	PROPN
alkej-870	414	8	wei	wei	PROPN
alkej-870	414	9	,	,	PUNCT
alkej-870	414	10	"	"	PUNCT
alkej-870	414	11	multi	multi	ADJ
alkej-870	414	12	-	-	ADJ
alkej-870	414	13	modal	modal	ADJ
alkej-870	414	14	speech	speech	NOUN
alkej-870	414	15	separation	separation	NOUN
alkej-870	414	16	based	base	VERB
alkej-870	414	17	on	on	ADP
alkej-870	414	18	two	two	NUM
alkej-870	414	19	-	-	PUNCT
alkej-870	414	20	stage	stage	NOUN
alkej-870	414	21	feature	feature	NOUN
alkej-870	414	22	fusion	fusion	NOUN
alkej-870	414	23	,	,	PUNCT
alkej-870	414	24	"	"	PUNCT
alkej-870	414	25	ieee	ieee	NOUN
alkej-870	414	26	6th	6th	ADJ
alkej-870	414	27	international	international	ADJ
alkej-870	414	28	conference	conference	NOUN
alkej-870	414	29	on	on	ADP
alkej-870	414	30	signal	signal	NOUN
alkej-870	414	31	and	and	CCONJ
alkej-870	414	32	image	image	NOUN
alkej-870	414	33	processing	processing	NOUN
alkej-870	414	34	(	(	PUNCT
alkej-870	414	35	icsip	icsip	NOUN
alkej-870	414	36	)	)	PUNCT
alkej-870	414	37	,	,	PUNCT
alkej-870	414	38	nanjing	nanjing	PROPN
alkej-870	414	39	,	,	PUNCT
alkej-870	414	40	china	china	PROPN
alkej-870	414	41	,	,	PUNCT
alkej-870	414	42	2021	2021	NUM
alkej-870	414	43	,	,	PUNCT
alkej-870	414	44	pp	pp	ADV
alkej-870	414	45	.	.	PUNCT
alkej-870	415	1	800	800	NUM
alkej-870	415	2	-	-	SYM
alkej-870	415	3	805	805	NUM
alkej-870	415	4	,	,	PUNCT
alkej-870	415	5	http://dx.doi.org/10.1109/icsip52628.2021	http://dx.doi.org/10.1109/icsip52628.2021	PROPN
alkej-870	415	6	.9688674	.9688674	PROPN
alkej-870	415	7	.	.	PUNCT
alkej-870	416	1	[	[	X
alkej-870	416	2	37	37	NUM
alkej-870	416	3	]	]	X
alkej-870	416	4	y.	y.	PROPN
alkej-870	416	5	tian	tian	PROPN
alkej-870	416	6	,	,	PUNCT
alkej-870	416	7	d.	d.	PROPN
alkej-870	416	8	hu	hu	PROPN
alkej-870	416	9	and	and	CCONJ
alkej-870	416	10	c.	c.	PROPN
alkej-870	416	11	xu	xu	PROPN
alkej-870	416	12	,	,	PUNCT
alkej-870	416	13	"	"	PUNCT
alkej-870	416	14	cyclic	cyclic	ADJ
alkej-870	416	15	colearning	colearning	NOUN
alkej-870	416	16	of	of	ADP
alkej-870	416	17	sounding	sound	VERB
alkej-870	416	18	object	object	NOUN
alkej-870	416	19	visual	visual	ADJ
alkej-870	416	20	grounding	grounding	NOUN
alkej-870	416	21	and	and	CCONJ
alkej-870	416	22	sound	sound	NOUN
alkej-870	416	23	separation	separation	NOUN
alkej-870	416	24	,	,	PUNCT
alkej-870	416	25	"	"	PUNCT
alkej-870	416	26	ieee	ieee	NOUN
alkej-870	416	27	/	/	SYM
alkej-870	416	28	cvf	cvf	NOUN
alkej-870	416	29	conference	conference	NOUN
alkej-870	416	30	on	on	ADP
alkej-870	416	31	computer	computer	NOUN
alkej-870	416	32	vision	vision	NOUN
alkej-870	416	33	and	and	CCONJ
alkej-870	416	34	pattern	pattern	NOUN
alkej-870	416	35	recognition	recognition	NOUN
alkej-870	416	36	(	(	PUNCT
alkej-870	416	37	cvpr	cvpr	NOUN
alkej-870	416	38	)	)	PUNCT
alkej-870	416	39	,	,	PUNCT
alkej-870	416	40	nashville	nashville	PROPN
alkej-870	416	41	,	,	PUNCT
alkej-870	416	42	tn	tn	PROPN
alkej-870	416	43	,	,	PUNCT
alkej-870	416	44	usa	usa	PROPN
alkej-870	416	45	,	,	PUNCT
alkej-870	416	46	2021	2021	NUM
alkej-870	416	47	,	,	PUNCT
alkej-870	416	48	pp	pp	ADJ
alkej-870	416	49	.	.	PUNCT
alkej-870	417	1	2744	2744	NUM
alkej-870	417	2	-	-	SYM
alkej-870	417	3	2753	2753	NUM
alkej-870	417	4	.	.	PUNCT
alkej-870	418	1	http://dx.doi.org/10.1109/cvpr46437.2021	http://dx.doi.org/10.1109/cvpr46437.2021	NOUN
alkej-870	418	2	.00277	.00277	PROPN
alkej-870	418	3	.	.	PUNCT
alkej-870	419	1	[	[	X
alkej-870	419	2	38	38	NUM
alkej-870	419	3	]	]	SYM
alkej-870	419	4	makishima	makishima	PROPN
alkej-870	419	5	,	,	PUNCT
alkej-870	419	6	n.	n.	NOUN
alkej-870	419	7	,	,	PUNCT
alkej-870	419	8	ihori	ihori	PROPN
alkej-870	419	9	,	,	PUNCT
alkej-870	419	10	m.	m.	NOUN
alkej-870	419	11	,	,	PUNCT
alkej-870	419	12	takashima	takashima	PROPN
alkej-870	419	13	,	,	PUNCT
alkej-870	419	14	a.	a.	NOUN
alkej-870	419	15	,	,	PUNCT
alkej-870	419	16	tanaka	tanaka	PROPN
alkej-870	419	17	,	,	PUNCT
alkej-870	419	18	t.	t.	PROPN
alkej-870	419	19	,	,	PUNCT
alkej-870	419	20	orihashi	orihashi	PROPN
alkej-870	419	21	,	,	PUNCT
alkej-870	419	22	s.	s.	PROPN
alkej-870	419	23	,	,	PUNCT
alkej-870	419	24	masumura	masumura	NOUN
alkej-870	419	25	,	,	PUNCT
alkej-870	419	26	r.	r.	PROPN
alkej-870	419	27	:	:	PUNCT
alkej-870	419	28	"	"	PUNCT
alkej-870	419	29	audio	audio	ADJ
alkej-870	419	30	-	-	ADJ
alkej-870	419	31	visual	visual	ADJ
alkej-870	419	32	speech	speech	NOUN
alkej-870	419	33	separation	separation	NOUN
alkej-870	419	34	using	use	VERB
alkej-870	419	35	cross	cross	ADJ
alkej-870	419	36	-	-	ADJ
alkej-870	419	37	modal	modal	ADJ
alkej-870	419	38	correspondence	correspondence	NOUN
alkej-870	419	39	loss	loss	NOUN
alkej-870	419	40	"	"	PUNCT
alkej-870	419	41	ieee	ieee	NOUN
alkej-870	419	42	international	international	ADJ
alkej-870	419	43	conference	conference	NOUN
alkej-870	419	44	on	on	ADP
alkej-870	419	45	acoustics	acoustic	NOUN
alkej-870	419	46	,	,	PUNCT
alkej-870	419	47	speech	speech	NOUN
alkej-870	419	48	and	and	CCONJ
alkej-870	419	49	signal	signal	NOUN
alkej-870	419	50	processing	processing	NOUN
alkej-870	419	51	(	(	PUNCT
alkej-870	419	52	icassp	icassp	PROPN
alkej-870	419	53	)	)	PUNCT
alkej-870	419	54	.	.	PUNCT
alkej-870	420	1	pp	pp	ADV
alkej-870	420	2	.	.	PUNCT
alkej-870	421	1	6673–6677	6673–6677	NUM
alkej-870	421	2	.	.	PUNCT
alkej-870	422	1	ieee	ieee	NOUN
alkej-870	422	2	(	(	PUNCT
alkej-870	422	3	2021	2021	NUM
alkej-870	422	4	)	)	PUNCT
alkej-870	422	5	.	.	PUNCT
alkej-870	423	1	http://dx.doi.org/10.1109/icassp39728.20	http://dx.doi.org/10.1109/icassp39728.20	NOUN
alkej-870	423	2	21.9413491	21.9413491	NUM
alkej-870	423	3	.	.	PUNCT
alkej-870	424	1	[	[	X
alkej-870	424	2	39	39	NUM
alkej-870	424	3	]	]	X
alkej-870	424	4	v.	v.	PROPN
alkej-870	424	5	-n	-n	NOUN
alkej-870	424	6	.	.	PUNCT
alkej-870	424	7	nguyen	nguyen	PROPN
alkej-870	424	8	,	,	PUNCT
alkej-870	424	9	m.	m.	NOUN
alkej-870	424	10	sadeghi	sadeghi	PROPN
alkej-870	424	11	,	,	PUNCT
alkej-870	424	12	e.	e.	PROPN
alkej-870	424	13	ricci	ricci	PROPN
alkej-870	424	14	and	and	CCONJ
alkej-870	424	15	x.	x.	PROPN
alkej-870	424	16	alameda	alameda	PROPN
alkej-870	424	17	-	-	PUNCT
alkej-870	424	18	pineda	pineda	PROPN
alkej-870	424	19	,	,	PUNCT
alkej-870	424	20	"	"	PUNCT
alkej-870	424	21	deep	deep	ADJ
alkej-870	424	22	variational	variational	ADJ
alkej-870	424	23	generative	generative	ADJ
alkej-870	424	24	models	model	NOUN
alkej-870	424	25	for	for	ADP
alkej-870	424	26	audio	audio	ADJ
alkej-870	424	27	-	-	ADJ
alkej-870	424	28	visual	visual	ADJ
alkej-870	424	29	speech	speech	NOUN
alkej-870	424	30	separation	separation	NOUN
alkej-870	424	31	,	,	PUNCT
alkej-870	424	32	"	"	PUNCT
alkej-870	424	33	ieee	ieee	NOUN
alkej-870	424	34	31st	31st	NOUN
alkej-870	424	35	international	international	ADJ
alkej-870	424	36	workshop	workshop	NOUN
alkej-870	424	37	on	on	ADP
alkej-870	424	38	machine	machine	NOUN
alkej-870	424	39	learning	learn	VERB
alkej-870	424	40	for	for	ADP
alkej-870	424	41	signal	signal	ADJ
alkej-870	424	42	processing	processing	NOUN
alkej-870	424	43	(	(	PUNCT
alkej-870	424	44	mlsp	mlsp	NOUN
alkej-870	424	45	)	)	PUNCT
alkej-870	424	46	,	,	PUNCT
alkej-870	424	47	gold	gold	NOUN
alkej-870	424	48	coast	coast	NOUN
alkej-870	424	49	,	,	PUNCT
alkej-870	424	50	australia	australia	PROPN
alkej-870	424	51	,	,	PUNCT
alkej-870	424	52	2021	2021	NUM
alkej-870	424	53	,	,	PUNCT
alkej-870	424	54	pp.1	pp.1	PROPN
alkej-870	424	55	-	-	PUNCT
alkej-870	424	56	6	6	NUM
alkej-870	424	57	http://dx.doi.org/10.1109/mlsp52302.2021	http://dx.doi.org/10.1109/mlsp52302.2021	NOUN
alkej-870	424	58	.9596406	.9596406	VERB
alkej-870	424	59	.	.	PUNCT
alkej-870	425	1	[	[	X
alkej-870	425	2	40	40	NUM
alkej-870	425	3	]	]	PUNCT
alkej-870	425	4	jiyoung	jiyoung	PROPN
alkej-870	425	5	lee	lee	PROPN
alkej-870	425	6	,	,	PUNCT
alkej-870	425	7	soo	soo	PROPN
alkej-870	425	8	-	-	PUNCT
alkej-870	425	9	whan	whan	PROPN
alkej-870	425	10	chung	chung	PROPN
alkej-870	425	11	,	,	PUNCT
alkej-870	425	12	sunok	sunok	NOUN
alkej-870	425	13	kim	kim	PROPN
alkej-870	425	14	,	,	PUNCT
alkej-870	425	15	hong	hong	PROPN
alkej-870	425	16	-	-	PUNCT
alkej-870	425	17	goo	goo	NOUN
alkej-870	425	18	kang	kang	PROPN
alkej-870	425	19	,	,	PUNCT
alkej-870	425	20	and	and	CCONJ
alkej-870	425	21	kwanghoon	kwanghoon	PROPN
alkej-870	425	22	sohn	sohn	PROPN
alkej-870	425	23	,	,	PUNCT
alkej-870	425	24	"	"	PUNCT
alkej-870	425	25	looking	look	VERB
alkej-870	425	26	into	into	ADP
alkej-870	425	27	your	your	PRON
alkej-870	425	28	speech	speech	NOUN
alkej-870	425	29	:	:	PUNCT
alkej-870	425	30	learning	learn	VERB
alkej-870	425	31	cross	cross	ADJ
alkej-870	425	32	-	-	ADJ
alkej-870	425	33	modal	modal	ADJ
alkej-870	425	34	affinity	affinity	NOUN
alkej-870	425	35	for	for	ADP
alkej-870	425	36	audio	audio	ADJ
alkej-870	425	37	-	-	ADJ
alkej-870	425	38	visual	visual	ADJ
alkej-870	425	39	speech	speech	NOUN
alkej-870	425	40	separation	separation	NOUN
alkej-870	425	41	,	,	PUNCT
alkej-870	425	42	"	"	PUNCT
alkej-870	425	43	in	in	ADP
alkej-870	425	44	proceedings	proceeding	NOUN
alkej-870	425	45	of	of	ADP
alkej-870	425	46	the	the	DET
alkej-870	425	47	ieee	ieee	NOUN
alkej-870	425	48	/	/	SYM
alkej-870	425	49	cvf	cvf	NOUN
alkej-870	425	50	conference	conference	NOUN
alkej-870	425	51	on	on	ADP
alkej-870	425	52	computer	computer	NOUN
alkej-870	425	53	vision	vision	NOUN
alkej-870	425	54	and	and	CCONJ
alkej-870	425	55	pattern	pattern	NOUN
alkej-870	425	56	recognition	recognition	NOUN
alkej-870	425	57	,	,	PUNCT
alkej-870	425	58	2021	2021	NUM
alkej-870	425	59	,	,	PUNCT
alkej-870	425	60	pp	pp	ADJ
alkej-870	425	61	.	.	PUNCT
alkej-870	425	62	1336	1336	NUM
alkej-870	425	63	–	–	PUNCT
alkej-870	425	64	1345	1345	NUM
alkej-870	425	65	.	.	PUNCT
alkej-870	426	1	http://dx.doi.org/10.1109/ijcnn48605.2020.9207517	http://dx.doi.org/10.1109/ijcnn48605.2020.9207517	PROPN
alkej-870	427	1	http://dx.doi.org/10.1109/ijcnn48605.2020.9207517	http://dx.doi.org/10.1109/ijcnn48605.2020.9207517	PROPN
alkej-870	427	2	http://dx.doi.org/10.1109/icpr48806.2021.9412174	http://dx.doi.org/10.1109/icpr48806.2021.9412174	PROPN
alkej-870	427	3	http://dx.doi.org/10.1109/icpr48806.2021.9412174	http://dx.doi.org/10.1109/icpr48806.2021.9412174	PROPN
alkej-870	427	4	http://dx.doi.org/10.1109/icassp40776.2020.9054180	http://dx.doi.org/10.1109/icassp40776.2020.9054180	PROPN
alkej-870	427	5	http://dx.doi.org/10.1109/icassp40776.2020.9054180	http://dx.doi.org/10.1109/icassp40776.2020.9054180	PROPN
alkej-870	427	6	http://dx.doi.org/10.1109/jstsp.2020.2980956	http://dx.doi.org/10.1109/jstsp.2020.2980956	PROPN
alkej-870	428	1	http://dx.doi.org/10.1109/jstsp.2020.2980956	http://dx.doi.org/10.1109/jstsp.2020.2980956	PROPN
alkej-870	429	1	http://dx.doi.org/10.1109/cvpr42600.2020.01049	http://dx.doi.org/10.1109/cvpr42600.2020.01049	PROPN
alkej-870	429	2	http://dx.doi.org/10.1109/cvpr42600.2020.01049	http://dx.doi.org/10.1109/cvpr42600.2020.01049	PROPN
alkej-870	429	3	http://dx.doi.org/10.1007/978-3-030-69544-6_25	http://dx.doi.org/10.1007/978-3-030-69544-6_25	VERB
alkej-870	430	1	http://dx.doi.org/10.1007/978-3-030-69544-6_25	http://dx.doi.org/10.1007/978-3-030-69544-6_25	PRON
alkej-870	430	2	http://dx.doi.org/10.1109/jstsp.2020.2987209	http://dx.doi.org/10.1109/jstsp.2020.2987209	NOUN
alkej-870	430	3	http://dx.doi.org/10.1109/jstsp.2020.2987209	http://dx.doi.org/10.1109/jstsp.2020.2987209	NOUN
alkej-870	431	1	http://dx.doi.org/10.21437/interspeech.2020-1697	http://dx.doi.org/10.21437/interspeech.2020-1697	INTJ
alkej-870	432	1	http://dx.doi.org/10.21437/interspeech.2020-1697	http://dx.doi.org/10.21437/interspeech.2020-1697	NOUN
alkej-870	432	2	http://dx.doi.org/10.1109/icme51207.2021.9428196	http://dx.doi.org/10.1109/icme51207.2021.9428196	ADV
alkej-870	432	3	http://dx.doi.org/10.1109/icme51207.2021.9428196	http://dx.doi.org/10.1109/icme51207.2021.9428196	ADV
alkej-870	432	4	http://dx.doi.org/10.1109/cvpr46437.2021.01524	http://dx.doi.org/10.1109/cvpr46437.2021.01524	ADJ
alkej-870	432	5	http://dx.doi.org/10.1109/cvpr46437.2021.01524	http://dx.doi.org/10.1109/cvpr46437.2021.01524	PROPN
alkej-870	432	6	http://dx.doi.org/10.1109/iccv48922.2021.00034	http://dx.doi.org/10.1109/iccv48922.2021.00034	PROPN
alkej-870	432	7	http://dx.doi.org/10.1109/iccv48922.2021.00034	http://dx.doi.org/10.1109/iccv48922.2021.00034	PROPN
alkej-870	433	1	http://dx.doi.org/10.1109/icsip52628.2021.9688674	http://dx.doi.org/10.1109/icsip52628.2021.9688674	NOUN
alkej-870	433	2	http://dx.doi.org/10.1109/icsip52628.2021.9688674	http://dx.doi.org/10.1109/icsip52628.2021.9688674	VERB
alkej-870	434	1	http://dx.doi.org/10.1109/cvpr46437.2021.00277	http://dx.doi.org/10.1109/cvpr46437.2021.00277	NOUN
alkej-870	435	1	http://dx.doi.org/10.1109/cvpr46437.2021.00277	http://dx.doi.org/10.1109/cvpr46437.2021.00277	NOUN
alkej-870	435	2	http://dx.doi.org/10.1109/icassp39728.2021.9413491	http://dx.doi.org/10.1109/icassp39728.2021.9413491	NOUN
alkej-870	435	3	http://dx.doi.org/10.1109/icassp39728.2021.9413491	http://dx.doi.org/10.1109/icassp39728.2021.9413491	PROPN
alkej-870	435	4	http://dx.doi.org/10.1109/mlsp52302.2021.9596406	http://dx.doi.org/10.1109/mlsp52302.2021.9596406	PROPN
alkej-870	435	5	http://dx.doi.org/10.1109/mlsp52302.2021.9596406	http://dx.doi.org/10.1109/mlsp52302.2021.9596406	PROPN
alkej-870	435	6	nooralhuda	nooralhuda	PROPN
alkej-870	435	7	mudhafar	mudhafar	PROPN
alkej-870	435	8	sleman	sleman	PROPN
alkej-870	435	9	al	al	PROPN
alkej-870	435	10	-	-	PUNCT
alkej-870	435	11	khwarizmi	khwarizmi	PROPN
alkej-870	435	12	engineering	engineering	NOUN
alkej-870	435	13	journal	journal	PROPN
alkej-870	435	14	,	,	PUNCT
alkej-870	435	15	vol	vol	NOUN
alkej-870	435	16	.	.	PROPN
alkej-870	435	17	19	19	NUM
alkej-870	435	18	,	,	PUNCT
alkej-870	435	19	no	no	INTJ
alkej-870	435	20	.	.	NOUN
alkej-870	435	21	4	4	NUM
alkej-870	435	22	,	,	PUNCT
alkej-870	435	23	p.p	p.p	PROPN
alkej-870	435	24	.	.	NOUN
alkej-870	435	25	4255	4255	NUM
alkej-870	435	26	(	(	PUNCT
alkej-870	435	27	2023	2023	NUM
alkej-870	435	28	)	)	PUNCT
alkej-870	435	29	53	53	NUM
alkej-870	435	30	http://dx.doi.org/10.1109/cvpr46437.2021	http://dx.doi.org/10.1109/cvpr46437.2021	NOUN
alkej-870	435	31	.00139	.00139	NOUN
alkej-870	435	32	.	.	PUNCT
alkej-870	436	1	[	[	X
alkej-870	436	2	41	41	NUM
alkej-870	436	3	]	]	PUNCT
alkej-870	436	4	lingyu	lingyu	NOUN
alkej-870	436	5	zhu	zhu	PROPN
alkej-870	436	6	and	and	CCONJ
alkej-870	436	7	esa	esa	PROPN
alkej-870	436	8	rahtu	rahtu	PROPN
alkej-870	436	9	.	.	PUNCT
alkej-870	437	1	"	"	PUNCT
alkej-870	437	2	leveraging	leverage	VERB
alkej-870	437	3	category	category	NOUN
alkej-870	437	4	information	information	NOUN
alkej-870	437	5	for	for	ADP
alkej-870	437	6	single	single	ADJ
alkej-870	437	7	-	-	PUNCT
alkej-870	437	8	frame	frame	NOUN
alkej-870	437	9	visual	visual	ADJ
alkej-870	437	10	sound	sound	NOUN
alkej-870	437	11	source	source	NOUN
alkej-870	437	12	separation	separation	NOUN
alkej-870	437	13	,	,	PUNCT
alkej-870	437	14	"	"	PUNCT
alkej-870	437	15	in	in	ADP
alkej-870	437	16	ieee	ieee	NOUN
alkej-870	437	17	2021	2021	NUM
alkej-870	437	18	9th	9th	ADJ
alkej-870	437	19	european	european	ADJ
alkej-870	437	20	workshop	workshop	NOUN
alkej-870	437	21	on	on	ADP
alkej-870	437	22	visual	visual	ADJ
alkej-870	437	23	information	information	NOUN
alkej-870	437	24	processing	processing	NOUN
alkej-870	437	25	(	(	PUNCT
alkej-870	437	26	euvip	euvip	PROPN
alkej-870	437	27	)	)	PUNCT
alkej-870	437	28	,	,	PUNCT
alkej-870	437	29	pages	page	NOUN
alkej-870	437	30	1	1	NUM
alkej-870	437	31	–	–	PUNCT
alkej-870	437	32	6	6	NUM
alkej-870	437	33	.	.	PUNCT
alkej-870	438	1	http://dx.doi.org/10.1109/euvip50544.202	http://dx.doi.org/10.1109/euvip50544.202	PROPN
alkej-870	438	2	1.9484036	1.9484036	NUM
alkej-870	438	3	.	.	PUNCT
alkej-870	439	1	[	[	X
alkej-870	439	2	42	42	NUM
alkej-870	439	3	]	]	X
alkej-870	439	4	r.	r.	PROPN
alkej-870	439	5	gu	gu	PROPN
alkej-870	439	6	,	,	PUNCT
alkej-870	439	7	s.	s.	PROPN
alkej-870	439	8	-x	-x	PROPN
alkej-870	439	9	.	.	PROPN
alkej-870	439	10	zhang	zhang	PROPN
alkej-870	439	11	,	,	PUNCT
alkej-870	439	12	y.	y.	PROPN
alkej-870	439	13	zou	zou	PROPN
alkej-870	439	14	and	and	CCONJ
alkej-870	439	15	d.	d.	PROPN
alkej-870	439	16	yu	yu	PROPN
alkej-870	439	17	,	,	PUNCT
alkej-870	439	18	"	"	PUNCT
alkej-870	439	19	towards	towards	ADP
alkej-870	439	20	unified	unify	VERB
alkej-870	439	21	all	all	DET
alkej-870	439	22	-	-	PUNCT
alkej-870	439	23	neural	neural	ADJ
alkej-870	439	24	beamforming	beamforming	NOUN
alkej-870	439	25	for	for	ADP
alkej-870	439	26	time	time	NOUN
alkej-870	439	27	and	and	CCONJ
alkej-870	439	28	frequency	frequency	NOUN
alkej-870	439	29	domain	domain	NOUN
alkej-870	439	30	speech	speech	NOUN
alkej-870	439	31	separation	separation	NOUN
alkej-870	439	32	,	,	PUNCT
alkej-870	439	33	"	"	PUNCT
alkej-870	439	34	in	in	ADP
alkej-870	439	35	ieee	ieee	NOUN
alkej-870	439	36	/	/	SYM
alkej-870	439	37	acm	acm	PROPN
alkej-870	439	38	transactions	transaction	NOUN
alkej-870	439	39	on	on	ADP
alkej-870	439	40	audio	audio	NOUN
alkej-870	439	41	,	,	PUNCT
alkej-870	439	42	speech	speech	NOUN
alkej-870	439	43	,	,	PUNCT
alkej-870	439	44	and	and	CCONJ
alkej-870	439	45	language	language	NOUN
alkej-870	439	46	processing	processing	NOUN
alkej-870	439	47	,	,	PUNCT
alkej-870	439	48	vol	vol	NOUN
alkej-870	439	49	.	.	PROPN
alkej-870	439	50	31	31	NUM
alkej-870	439	51	,	,	PUNCT
alkej-870	439	52	pp	pp	ADJ
alkej-870	439	53	.	.	PUNCT
alkej-870	440	1	849	849	NUM
alkej-870	440	2	-	-	SYM
alkej-870	440	3	862	862	NUM
alkej-870	440	4	,	,	PUNCT
alkej-870	440	5	2023	2023	NUM
alkej-870	440	6	.	.	PUNCT
alkej-870	441	1	http://dx.doi.org/10.1109/taslp.2022.322	http://dx.doi.org/10.1109/taslp.2022.322	PROPN
alkej-870	441	2	9261	9261	NUM
alkej-870	441	3	.	.	PUNCT
alkej-870	442	1	[	[	X
alkej-870	442	2	43	43	NUM
alkej-870	442	3	]	]	X
alkej-870	442	4	t.	t.	X
alkej-870	442	5	oya	oya	PROPN
alkej-870	442	6	,	,	PUNCT
alkej-870	442	7	s.	s.	PROPN
alkej-870	442	8	iwase	iwase	PROPN
alkej-870	442	9	and	and	CCONJ
alkej-870	442	10	s.	s.	PROPN
alkej-870	442	11	morishima	morishima	PROPN
alkej-870	442	12	,	,	PUNCT
alkej-870	442	13	"	"	PUNCT
alkej-870	442	14	the	the	DET
alkej-870	442	15	sound	sound	NOUN
alkej-870	442	16	of	of	ADP
alkej-870	442	17	bounding	bounding	NOUN
alkej-870	442	18	-	-	PUNCT
alkej-870	442	19	boxes	box	NOUN
alkej-870	442	20	,	,	PUNCT
alkej-870	442	21	"	"	PUNCT
alkej-870	442	22	26th	26th	ADJ
alkej-870	442	23	international	international	ADJ
alkej-870	442	24	conference	conference	NOUN
alkej-870	442	25	on	on	ADP
alkej-870	442	26	pattern	pattern	NOUN
alkej-870	442	27	recognition	recognition	NOUN
alkej-870	442	28	(	(	PUNCT
alkej-870	442	29	icpr	icpr	PROPN
alkej-870	442	30	)	)	PUNCT
alkej-870	442	31	,	,	PUNCT
alkej-870	442	32	montreal	montreal	PROPN
alkej-870	442	33	,	,	PUNCT
alkej-870	442	34	qc	qc	PROPN
alkej-870	442	35	,	,	PUNCT
alkej-870	442	36	canada	canada	PROPN
alkej-870	442	37	,	,	PUNCT
alkej-870	442	38	2022	2022	NUM
alkej-870	442	39	,	,	PUNCT
alkej-870	442	40	pp	pp	ADJ
alkej-870	442	41	.	.	PUNCT
alkej-870	443	1	9	9	NUM
alkej-870	443	2	-	-	SYM
alkej-870	443	3	15	15	NUM
alkej-870	443	4	.	.	PUNCT
alkej-870	443	5	http://dx.doi.org/10.1109/icpr56361.2022	http://dx.doi.org/10.1109/icpr56361.2022	NOUN
alkej-870	443	6	.	.	PUNCT
alkej-870	444	1	9956384	9956384	NUM
alkej-870	444	2	.	.	PUNCT
alkej-870	445	1	[	[	X
alkej-870	445	2	44	44	NUM
alkej-870	445	3	]	]	PUNCT
alkej-870	445	4	lingyu	lingyu	NOUN
alkej-870	445	5	zhu	zhu	PROPN
alkej-870	445	6	and	and	CCONJ
alkej-870	445	7	esa	esa	PROPN
alkej-870	445	8	rahtu	rahtu	PROPN
alkej-870	445	9	.	.	PUNCT
alkej-870	446	1	"	"	PUNCT
alkej-870	446	2	visually	visually	ADV
alkej-870	446	3	guided	guide	VERB
alkej-870	446	4	sound	sound	NOUN
alkej-870	446	5	source	source	NOUN
alkej-870	446	6	separation	separation	NOUN
alkej-870	446	7	and	and	CCONJ
alkej-870	446	8	localization	localization	NOUN
alkej-870	446	9	using	use	VERB
alkej-870	446	10	self	self	NOUN
alkej-870	446	11	-	-	PUNCT
alkej-870	446	12	supervised	supervise	VERB
alkej-870	446	13	motion	motion	NOUN
alkej-870	446	14	representations	representation	NOUN
alkej-870	446	15	"	"	PUNCT
alkej-870	446	16	in	in	ADP
alkej-870	446	17	proceedings	proceeding	NOUN
alkej-870	446	18	of	of	ADP
alkej-870	446	19	the	the	DET
alkej-870	446	20	ieee	ieee	NOUN
alkej-870	446	21	/	/	SYM
alkej-870	446	22	cvf	cvf	NOUN
alkej-870	446	23	winter	winter	NOUN
alkej-870	446	24	conference	conference	NOUN
alkej-870	446	25	on	on	ADP
alkej-870	446	26	applications	application	NOUN
alkej-870	446	27	of	of	ADP
alkej-870	446	28	computer	computer	NOUN
alkej-870	446	29	vision	vision	NOUN
alkej-870	446	30	,	,	PUNCT
alkej-870	446	31	pages	page	NOUN
alkej-870	446	32	1289–1299	1289–1299	NUM
alkej-870	446	33	,	,	PUNCT
alkej-870	446	34	2022	2022	NUM
alkej-870	446	35	.	.	PUNCT
alkej-870	447	1	http://dx.doi.org/10.1109/wacv51458.202	http://dx.doi.org/10.1109/wacv51458.202	PROPN
alkej-870	447	2	2.00223	2.00223	NUM
alkej-870	447	3	.	.	PUNCT
alkej-870	448	1	[	[	X
alkej-870	448	2	45	45	NUM
alkej-870	448	3	]	]	X
alkej-870	448	4	d.	d.	PROPN
alkej-870	448	5	-h	-h	PUNCT
alkej-870	448	6	.	.	PUNCT
alkej-870	448	7	pham	pham	PROPN
alkej-870	448	8	,	,	PUNCT
alkej-870	448	9	q.	q.	PROPN
alkej-870	448	10	-a	-a	PROPN
alkej-870	448	11	.	.	PUNCT
alkej-870	448	12	do	do	VERB
alkej-870	448	13	,	,	PUNCT
alkej-870	449	1	t.	t.	PROPN
alkej-870	449	2	t.	t.	PROPN
alkej-870	449	3	-h	-h	PROPN
alkej-870	449	4	.	.	PUNCT
alkej-870	449	5	duong	duong	PROPN
alkej-870	449	6	,	,	PUNCT
alkej-870	449	7	t.	t.	PROPN
alkej-870	449	8	-l	-l	PROPN
alkej-870	449	9	.	.	PUNCT
alkej-870	449	10	le	le	PROPN
alkej-870	449	11	and	and	CCONJ
alkej-870	449	12	p.	p.	NOUN
alkej-870	449	13	-l	-l	PROPN
alkej-870	449	14	.	.	PUNCT
alkej-870	450	1	nguyen	nguyen	PROPN
alkej-870	450	2	,	,	PUNCT
alkej-870	450	3	"	"	PUNCT
alkej-870	450	4	end	end	NOUN
alkej-870	450	5	-	-	PUNCT
alkej-870	450	6	to	to	ADP
alkej-870	450	7	-	-	PUNCT
alkej-870	450	8	end	end	NOUN
alkej-870	450	9	visual	visual	ADJ
alkej-870	450	10	-	-	PUNCT
alkej-870	450	11	guided	guide	VERB
alkej-870	450	12	audio	audio	NOUN
alkej-870	450	13	source	source	NOUN
alkej-870	450	14	separation	separation	NOUN
alkej-870	450	15	with	with	ADP
alkej-870	450	16	enhanced	enhanced	ADJ
alkej-870	450	17	losses	loss	NOUN
alkej-870	450	18	,	,	PUNCT
alkej-870	450	19	"	"	PUNCT
alkej-870	450	20	asia	asia	PROPN
alkej-870	450	21	-	-	PUNCT
alkej-870	450	22	pacific	pacific	PROPN
alkej-870	450	23	signal	signal	NOUN
alkej-870	450	24	and	and	CCONJ
alkej-870	450	25	information	information	NOUN
alkej-870	450	26	processing	processing	NOUN
alkej-870	450	27	association	association	NOUN
alkej-870	450	28	annual	annual	ADJ
alkej-870	450	29	summit	summit	NOUN
alkej-870	450	30	and	and	CCONJ
alkej-870	450	31	conference	conference	NOUN
alkej-870	450	32	(	(	PUNCT
alkej-870	450	33	apsipa	apsipa	PROPN
alkej-870	450	34	asc	asc	PROPN
alkej-870	450	35	)	)	PUNCT
alkej-870	450	36	,	,	PUNCT
alkej-870	450	37	chiang	chiang	PROPN
alkej-870	450	38	mai	mai	PROPN
alkej-870	450	39	,	,	PUNCT
alkej-870	450	40	thailand	thailand	PROPN
alkej-870	450	41	,	,	PUNCT
alkej-870	450	42	2022	2022	NUM
alkej-870	450	43	,	,	PUNCT
alkej-870	450	44	pp	pp	ADJ
alkej-870	450	45	.	.	PUNCT
alkej-870	451	1	2022	2022	NUM
alkej-870	451	2	-	-	SYM
alkej-870	451	3	2028	2028	NUM
alkej-870	451	4	.	.	PUNCT
alkej-870	452	1	http://dx.doi.org/10.23919/apsipaasc55	http://dx.doi.org/10.23919/apsipaasc55	PROPN
alkej-870	453	1	919.2022.9980162	919.2022.9980162	NUM
alkej-870	453	2	.	.	PUNCT
alkej-870	454	1	[	[	X
alkej-870	454	2	46	46	NUM
alkej-870	454	3	]	]	X
alkej-870	454	4	xudong	xudong	PROPN
alkej-870	454	5	xu	xu	PROPN
alkej-870	454	6	,	,	PUNCT
alkej-870	454	7	bo	bo	PROPN
alkej-870	454	8	dai	dai	PROPN
alkej-870	454	9	,	,	PUNCT
alkej-870	454	10	and	and	CCONJ
alkej-870	454	11	dahua	dahua	PROPN
alkej-870	454	12	lin	lin	PROPN
alkej-870	454	13	,	,	PUNCT
alkej-870	454	14	"	"	PUNCT
alkej-870	454	15	recursive	recursive	ADJ
alkej-870	454	16	visual	visual	ADJ
alkej-870	454	17	sound	sound	NOUN
alkej-870	454	18	separation	separation	NOUN
alkej-870	454	19	using	use	VERB
alkej-870	454	20	minus	minus	NOUN
alkej-870	454	21	-	-	PUNCT
alkej-870	454	22	plus	plus	CCONJ
alkej-870	454	23	net	net	NOUN
alkej-870	454	24	,	,	PUNCT
alkej-870	454	25	"	"	PUNCT
alkej-870	454	26	in	in	ADP
alkej-870	454	27	proceedings	proceeding	NOUN
alkej-870	454	28	of	of	ADP
alkej-870	454	29	the	the	DET
alkej-870	454	30	ieee	ieee	NOUN
alkej-870	454	31	international	international	PROPN
alkej-870	454	32	conference	conference	NOUN
alkej-870	454	33	on	on	ADP
alkej-870	454	34	computer	computer	NOUN
alkej-870	454	35	vision	vision	NOUN
alkej-870	454	36	,	,	PUNCT
alkej-870	454	37	2019	2019	NUM
alkej-870	454	38	,	,	PUNCT
alkej-870	454	39	pp	pp	ADJ
alkej-870	454	40	.	.	PUNCT
alkej-870	455	1	882–891	882–891	NUM
alkej-870	455	2	.	.	PUNCT
alkej-870	456	1	http://dx.doi.org/10.1109/iccv.2019.0009	http://dx.doi.org/10.1109/iccv.2019.0009	NOUN
alkej-870	456	2	7	7	NUM
alkej-870	456	3	.	.	PUNCT
alkej-870	457	1	[	[	X
alkej-870	457	2	47	47	NUM
alkej-870	457	3	]	]	PUNCT
alkej-870	457	4	hang	hang	PROPN
alkej-870	457	5	zhao	zhao	PROPN
alkej-870	457	6	,	,	PUNCT
alkej-870	457	7	chuang	chuang	PROPN
alkej-870	457	8	gan	gan	PROPN
alkej-870	457	9	,	,	PUNCT
alkej-870	457	10	wei	wei	PROPN
alkej-870	457	11	-	-	PROPN
alkej-870	457	12	chiu	chiu	PROPN
alkej-870	457	13	ma	ma	PROPN
alkej-870	457	14	,	,	PUNCT
alkej-870	457	15	and	and	CCONJ
alkej-870	457	16	antonio	antonio	PROPN
alkej-870	457	17	torralba	torralba	PROPN
alkej-870	457	18	,	,	PUNCT
alkej-870	457	19	"	"	PUNCT
alkej-870	457	20	the	the	DET
alkej-870	457	21	sound	sound	NOUN
alkej-870	457	22	of	of	ADP
alkej-870	457	23	motions	motion	NOUN
alkej-870	457	24	,	,	PUNCT
alkej-870	457	25	"	"	PUNCT
alkej-870	457	26	in	in	ADP
alkej-870	457	27	proceedings	proceeding	NOUN
alkej-870	457	28	of	of	ADP
alkej-870	457	29	the	the	DET
alkej-870	457	30	ieee	ieee	NOUN
alkej-870	457	31	international	international	PROPN
alkej-870	457	32	conference	conference	NOUN
alkej-870	457	33	on	on	ADP
alkej-870	457	34	computer	computer	NOUN
alkej-870	457	35	vision	vision	NOUN
alkej-870	457	36	,	,	PUNCT
alkej-870	457	37	2019	2019	NUM
alkej-870	457	38	,	,	PUNCT
alkej-870	457	39	pp	pp	ADJ
alkej-870	457	40	.	.	PUNCT
alkej-870	458	1	1735–1744	1735–1744	NUM
alkej-870	458	2	.	.	PUNCT
alkej-870	458	3	http://dx.doi.org/10.1109/iccv.2019.0018	http://dx.doi.org/10.1109/iccv.2019.0018	NOUN
alkej-870	458	4	2	2	NUM
alkej-870	458	5	.	.	PUNCT
alkej-870	459	1	[	[	X
alkej-870	459	2	48	48	NUM
alkej-870	459	3	]	]	PUNCT
alkej-870	459	4	r.	r.	PROPN
alkej-870	459	5	lu	lu	PROPN
alkej-870	459	6	,	,	PUNCT
alkej-870	459	7	z.	z.	PROPN
alkej-870	459	8	duan	duan	PROPN
alkej-870	459	9	,	,	PUNCT
alkej-870	459	10	and	and	CCONJ
alkej-870	459	11	c.	c.	PROPN
alkej-870	459	12	zhang	zhang	PROPN
alkej-870	459	13	,	,	PUNCT
alkej-870	459	14	"	"	PUNCT
alkej-870	459	15	audiovisual	audiovisual	ADJ
alkej-870	459	16	deep	deep	ADJ
alkej-870	459	17	clustering	clustering	NOUN
alkej-870	459	18	for	for	ADP
alkej-870	459	19	speech	speech	NOUN
alkej-870	459	20	separation	separation	NOUN
alkej-870	459	21	,	,	PUNCT
alkej-870	459	22	"	"	PUNCT
alkej-870	459	23	ieee	ieee	PROPN
alkej-870	459	24	acm	acm	PROPN
alkej-870	459	25	trans	trans	PROPN
alkej-870	459	26	.	.	PUNCT
alkej-870	459	27	audio	audio	PROPN
alkej-870	459	28	speech	speech	NOUN
alkej-870	459	29	lang	lang	PROPN
alkej-870	459	30	.	.	PUNCT
alkej-870	459	31	process	process	NOUN
alkej-870	459	32	.	.	PUNCT
alkej-870	459	33	,	,	PUNCT
alkej-870	459	34	vol	vol	NOUN
alkej-870	459	35	.	.	PROPN
alkej-870	459	36	27	27	NUM
alkej-870	459	37	,	,	PUNCT
alkej-870	459	38	no	no	INTJ
alkej-870	459	39	.	.	NOUN
alkej-870	459	40	11	11	NUM
alkej-870	459	41	,	,	PUNCT
alkej-870	459	42	pp	pp	ADJ
alkej-870	459	43	.	.	PUNCT
alkej-870	460	1	1697–1712	1697–1712	NUM
alkej-870	460	2	,	,	PUNCT
alkej-870	460	3	2019	2019	NUM
alkej-870	460	4	.	.	PUNCT
alkej-870	461	1	http://dx.doi.org/10.1109/taslp.2019.292	http://dx.doi.org/10.1109/taslp.2019.292	X
alkej-870	461	2	8140	8140	NUM
alkej-870	461	3	.	.	PUNCT
alkej-870	462	1	[	[	X
alkej-870	462	2	49	49	NUM
alkej-870	462	3	]	]	PUNCT
alkej-870	462	4	ruohan	ruohan	NOUN
alkej-870	462	5	gao	gao	PROPN
alkej-870	462	6	and	and	CCONJ
alkej-870	462	7	kristen	kristen	PROPN
alkej-870	462	8	grauman	grauman	NOUN
alkej-870	462	9	,	,	PUNCT
alkej-870	462	10	"	"	PUNCT
alkej-870	462	11	coseparating	coseparate	VERB
alkej-870	462	12	sounds	sound	NOUN
alkej-870	462	13	of	of	ADP
alkej-870	462	14	visual	visual	ADJ
alkej-870	462	15	objects	object	NOUN
alkej-870	462	16	,	,	PUNCT
alkej-870	462	17	"	"	PUNCT
alkej-870	462	18	in	in	ADP
alkej-870	462	19	proc	proc	NOUN
alkej-870	462	20	.	.	PUNCT
alkej-870	463	1	iccv	iccv	PROPN
alkej-870	463	2	,	,	PUNCT
alkej-870	463	3	2019	2019	NUM
alkej-870	463	4	.	.	PUNCT
alkej-870	464	1	http://dx.doi.org/10.1109/iccv.2019.0039	http://dx.doi.org/10.1109/iccv.2019.0039	NOUN
alkej-870	464	2	8	8	NUM
alkej-870	464	3	.	.	PUNCT
alkej-870	465	1	[	[	X
alkej-870	465	2	50	50	NUM
alkej-870	465	3	]	]	PUNCT
alkej-870	465	4	m.	m.	PROPN
alkej-870	465	5	cooke	cooke	PROPN
alkej-870	465	6	,	,	PUNCT
alkej-870	465	7	j.	j.	PROPN
alkej-870	465	8	barker	barker	PROPN
alkej-870	465	9	,	,	PUNCT
alkej-870	465	10	s.	s.	PROPN
alkej-870	465	11	cunningham	cunningham	PROPN
alkej-870	465	12	,	,	PUNCT
alkej-870	465	13	x.	x.	PROPN
alkej-870	465	14	shao	shao	PROPN
alkej-870	465	15	.	.	PUNCT
alkej-870	466	1	"	"	PUNCT
alkej-870	466	2	an	an	DET
alkej-870	466	3	audio	audio	ADJ
alkej-870	466	4	-	-	ADJ
alkej-870	466	5	visual	visual	ADJ
alkej-870	466	6	corpus	corpus	NOUN
alkej-870	466	7	for	for	ADP
alkej-870	466	8	speech	speech	NOUN
alkej-870	466	9	perception	perception	NOUN
alkej-870	466	10	and	and	CCONJ
alkej-870	466	11	automatic	automatic	ADJ
alkej-870	466	12	speech	speech	NOUN
alkej-870	466	13	recognition	recognition	NOUN
alkej-870	466	14	"	"	PUNCT
alkej-870	466	15	the	the	DET
alkej-870	466	16	journal	journal	NOUN
alkej-870	466	17	of	of	ADP
alkej-870	466	18	the	the	DET
alkej-870	466	19	acoustical	acoustical	ADJ
alkej-870	466	20	society	society	NOUN
alkej-870	466	21	of	of	ADP
alkej-870	466	22	america	america	PROPN
alkej-870	466	23	,	,	PUNCT
alkej-870	466	24	vol.120	vol.120	PROPN
alkej-870	466	25	,	,	PUNCT
alkej-870	466	26	no.5	no.5	PROPN
alkej-870	466	27	,	,	PUNCT
alkej-870	466	28	pp.2421–2424	pp.2421–2424	ADV
alkej-870	466	29	,	,	PUNCT
alkej-870	466	30	2006	2006	NUM
alkej-870	466	31	.	.	PUNCT
alkej-870	467	1	http://dx.doi.org/10.1121/1.2229005	http://dx.doi.org/10.1121/1.2229005	PROPN
alkej-870	467	2	.	.	PUNCT
alkej-870	468	1	[	[	X
alkej-870	468	2	51	51	NUM
alkej-870	468	3	]	]	X
alkej-870	468	4	n.	n.	NOUN
alkej-870	468	5	alghamdi	alghamdi	NOUN
alkej-870	468	6	,	,	PUNCT
alkej-870	468	7	s.	s.	PROPN
alkej-870	468	8	maddock	maddock	PROPN
alkej-870	468	9	,	,	PUNCT
alkej-870	468	10	r.	r.	PROPN
alkej-870	468	11	marxer	marxer	PROPN
alkej-870	468	12	,	,	PUNCT
alkej-870	468	13	j.	j.	PROPN
alkej-870	468	14	barker	barker	PROPN
alkej-870	468	15	,	,	PUNCT
alkej-870	468	16	g.	g.	PROPN
alkej-870	468	17	j.	j.	PROPN
alkej-870	468	18	brown	brown	PROPN
alkej-870	468	19	.	.	PUNCT
alkej-870	469	1	"	"	PUNCT
alkej-870	469	2	a	a	DET
alkej-870	469	3	corpus	corpus	NOUN
alkej-870	469	4	of	of	ADP
alkej-870	469	5	audiovisual	audiovisual	ADJ
alkej-870	469	6	lombard	lombard	NOUN
alkej-870	469	7	speech	speech	NOUN
alkej-870	469	8	with	with	ADP
alkej-870	469	9	frontal	frontal	ADJ
alkej-870	469	10	and	and	CCONJ
alkej-870	469	11	profile	profile	NOUN
alkej-870	469	12	views	view	NOUN
alkej-870	469	13	"	"	PUNCT
alkej-870	469	14	the	the	DET
alkej-870	469	15	journal	journal	NOUN
alkej-870	469	16	of	of	ADP
alkej-870	469	17	the	the	DET
alkej-870	469	18	acoustical	acoustical	ADJ
alkej-870	469	19	society	society	NOUN
alkej-870	469	20	of	of	ADP
alkej-870	469	21	america	america	PROPN
alkej-870	469	22	,	,	PUNCT
alkej-870	469	23	vol.143	vol.143	PROPN
alkej-870	469	24	,	,	PUNCT
alkej-870	469	25	no.6	no.6	PROPN
alkej-870	469	26	,	,	PUNCT
alkej-870	469	27	pp.el523	pp.el523	NOUN
alkej-870	469	28	–	–	PUNCT
alkej-870	469	29	el529	el529	ADV
alkej-870	469	30	,	,	PUNCT
alkej-870	469	31	2018	2018	NUM
alkej-870	469	32	.	.	PUNCT
alkej-870	470	1	http://dx.doi.org/10.1121/1.5042758	http://dx.doi.org/10.1121/1.5042758	NOUN
alkej-870	470	2	.	.	PUNCT
alkej-870	471	1	[	[	X
alkej-870	471	2	52	52	NUM
alkej-870	471	3	]	]	X
alkej-870	471	4	n.	n.	PROPN
alkej-870	471	5	harte	harte	PROPN
alkej-870	471	6	,	,	PUNCT
alkej-870	471	7	e.	e.	PROPN
alkej-870	471	8	gillen	gillen	PROPN
alkej-870	471	9	.	.	PUNCT
alkej-870	472	1	"	"	PUNCT
alkej-870	472	2	tcd	tcd	NOUN
alkej-870	472	3	-	-	PUNCT
alkej-870	472	4	timit	timit	NOUN
alkej-870	472	5	:	:	PUNCT
alkej-870	472	6	an	an	DET
alkej-870	472	7	audio	audio	ADJ
alkej-870	472	8	-	-	ADJ
alkej-870	472	9	visual	visual	ADJ
alkej-870	472	10	corpus	corpus	NOUN
alkej-870	472	11	of	of	ADP
alkej-870	472	12	continuous	continuous	ADJ
alkej-870	472	13	speech	speech	NOUN
alkej-870	472	14	"	"	PUNCT
alkej-870	472	15	.	.	PUNCT
alkej-870	473	1	ieee	ieee	NOUN
alkej-870	473	2	transactions	transaction	NOUN
alkej-870	473	3	on	on	ADP
alkej-870	473	4	multimedia	multimedia	NOUN
alkej-870	473	5	,	,	PUNCT
alkej-870	473	6	vol.17	vol.17	ADJ
alkej-870	473	7	,	,	PUNCT
alkej-870	473	8	no.5	no.5	PROPN
alkej-870	473	9	,	,	PUNCT
alkej-870	473	10	pp.603–615	pp.603–615	NOUN
alkej-870	473	11	,	,	PUNCT
alkej-870	473	12	2015	2015	NUM
alkej-870	473	13	.	.	PUNCT
alkej-870	474	1	http://dx.doi.org/10.1109/tmm.2015.2407	http://dx.doi.org/10.1109/tmm.2015.2407	VERB
alkej-870	474	2	694	694	NUM
alkej-870	474	3	.	.	PUNCT
alkej-870	475	1	[	[	X
alkej-870	475	2	53	53	NUM
alkej-870	475	3	]	]	PUNCT
alkej-870	475	4	g.	g.	PROPN
alkej-870	475	5	y.	y.	PROPN
alkej-870	475	6	zhao	zhao	PROPN
alkej-870	475	7	,	,	PUNCT
alkej-870	475	8	m.	m.	PROPN
alkej-870	475	9	barnard	barnard	PROPN
alkej-870	475	10	,	,	PUNCT
alkej-870	475	11	m.	m.	NOUN
alkej-870	475	12	pietikainen	pietikainen	PROPN
alkej-870	475	13	.	.	PUNCT
alkej-870	476	1	"	"	PUNCT
alkej-870	476	2	lipreading	lipread	VERB
alkej-870	476	3	with	with	ADP
alkej-870	476	4	local	local	ADJ
alkej-870	476	5	spatiotemporal	spatiotemporal	ADJ
alkej-870	476	6	descriptors	descriptor	NOUN
alkej-870	476	7	"	"	PUNCT
alkej-870	476	8	.	.	PUNCT
alkej-870	477	1	ieee	ieee	NOUN
alkej-870	477	2	transactions	transaction	NOUN
alkej-870	477	3	on	on	ADP
alkej-870	477	4	multimedia	multimedia	NOUN
alkej-870	477	5	,	,	PUNCT
alkej-870	477	6	vol.11	vol.11	PROPN
alkej-870	477	7	,	,	PUNCT
alkej-870	477	8	no.7	no.7	PROPN
alkej-870	477	9	,	,	PUNCT
alkej-870	477	10	pp.1254–1265	pp.1254–1265	NOUN
alkej-870	477	11	,	,	PUNCT
alkej-870	477	12	2009	2009	NUM
alkej-870	477	13	.	.	PUNCT
alkej-870	478	1	http://dx.doi.org/10.1109/tmm.2009.2030	http://dx.doi.org/10.1109/tmm.2009.2030	NOUN
alkej-870	478	2	637	637	NUM
alkej-870	478	3	.	.	PUNCT
alkej-870	479	1	[	[	X
alkej-870	479	2	54	54	NUM
alkej-870	479	3	]	]	X
alkej-870	479	4	i.	i.	PROPN
alkej-870	479	5	anina	anina	PROPN
alkej-870	479	6	,	,	PUNCT
alkej-870	479	7	z.	z.	PROPN
alkej-870	479	8	h.	h.	PROPN
alkej-870	479	9	zhou	zhou	PROPN
alkej-870	479	10	,	,	PUNCT
alkej-870	479	11	g.	g.	PROPN
alkej-870	479	12	y.	y.	PROPN
alkej-870	479	13	zhao	zhao	PROPN
alkej-870	479	14	,	,	PUNCT
alkej-870	479	15	m.	m.	NOUN
alkej-870	479	16	pietikäinen	pietikäinen	NOUN
alkej-870	479	17	.	.	PUNCT
alkej-870	480	1	"	"	PUNCT
alkej-870	480	2	ouluvs2	ouluvs2	NOUN
alkej-870	480	3	:	:	PUNCT
alkej-870	480	4	a	a	DET
alkej-870	480	5	multi	multi	ADJ
alkej-870	480	6	-	-	ADJ
alkej-870	480	7	view	view	ADJ
alkej-870	480	8	audiovisual	audiovisual	ADJ
alkej-870	480	9	database	database	NOUN
alkej-870	480	10	for	for	ADP
alkej-870	480	11	non	non	ADJ
alkej-870	480	12	-	-	ADJ
alkej-870	480	13	rigid	rigid	ADJ
alkej-870	480	14	mouth	mouth	NOUN
alkej-870	480	15	motion	motion	NOUN
alkej-870	480	16	analysis	analysis	NOUN
alkej-870	480	17	"	"	PUNCT
alkej-870	480	18	.	.	PUNCT
alkej-870	481	1	in	in	ADP
alkej-870	481	2	proceedings	proceeding	NOUN
alkej-870	481	3	of	of	ADP
alkej-870	481	4	the	the	DET
alkej-870	481	5	11th	11th	ADJ
alkej-870	481	6	ieee	ieee	NOUN
alkej-870	481	7	international	international	ADJ
alkej-870	481	8	conference	conference	NOUN
alkej-870	481	9	and	and	CCONJ
alkej-870	481	10	workshops	workshop	NOUN
alkej-870	481	11	on	on	ADP
alkej-870	481	12	automatic	automatic	ADJ
alkej-870	481	13	face	face	NOUN
alkej-870	481	14	and	and	CCONJ
alkej-870	481	15	gesture	gesture	NOUN
alkej-870	481	16	recognition	recognition	NOUN
alkej-870	481	17	,	,	PUNCT
alkej-870	481	18	ieee	ieee	PROPN
alkej-870	481	19	,	,	PUNCT
alkej-870	481	20	ljubljana	ljubljana	PROPN
alkej-870	481	21	,	,	PUNCT
alkej-870	481	22	slovenia	slovenia	PROPN
alkej-870	481	23	,	,	PUNCT
alkej-870	481	24	pp.1−5	pp.1−5	PROPN
alkej-870	481	25	,	,	PUNCT
alkej-870	481	26	2015	2015	NUM
alkej-870	481	27	.	.	PUNCT
alkej-870	482	1	http://dx.doi.org/10.1109/fg.2015.7163155	http://dx.doi.org/10.1109/fg.2015.7163155	PROPN
alkej-870	482	2	.	.	PUNCT
alkej-870	483	1	[	[	X
alkej-870	483	2	55	55	NUM
alkej-870	483	3	]	]	PUNCT
alkej-870	483	4	a.	a.	NOUN
alkej-870	483	5	nagrani	nagrani	PROPN
alkej-870	483	6	,	,	PUNCT
alkej-870	483	7	j.	j.	PROPN
alkej-870	483	8	s.	s.	PROPN
alkej-870	483	9	chung	chung	PROPN
alkej-870	483	10	,	,	PUNCT
alkej-870	483	11	a.	a.	NOUN
alkej-870	483	12	zisserman	zisserman	NOUN
alkej-870	483	13	.	.	PUNCT
alkej-870	484	1	"	"	PUNCT
alkej-870	484	2	voxceleb	voxceleb	NOUN
alkej-870	484	3	:	:	PUNCT
alkej-870	484	4	a	a	DET
alkej-870	484	5	large	large	ADJ
alkej-870	484	6	-	-	PUNCT
alkej-870	484	7	scale	scale	NOUN
alkej-870	484	8	speaker	speaker	NOUN
alkej-870	484	9	identification	identification	NOUN
alkej-870	484	10	dataset	dataset	NOUN
alkej-870	484	11	"	"	PUNCT
alkej-870	484	12	.	.	PUNCT
alkej-870	485	1	in	in	ADP
alkej-870	485	2	proceedings	proceeding	NOUN
alkej-870	485	3	of	of	ADP
alkej-870	485	4	the	the	DET
alkej-870	485	5	18th	18th	ADJ
alkej-870	485	6	annual	annual	ADJ
alkej-870	485	7	conference	conference	NOUN
alkej-870	485	8	of	of	ADP
alkej-870	485	9	the	the	DET
alkej-870	485	10	international	international	ADJ
alkej-870	485	11	speech	speech	NOUN
alkej-870	485	12	communication	communication	NOUN
alkej-870	485	13	http://dx.doi.org/10.1109/cvpr46437.2021.00139	http://dx.doi.org/10.1109/cvpr46437.2021.00139	PROPN
alkej-870	485	14	http://dx.doi.org/10.1109/cvpr46437.2021.00139	http://dx.doi.org/10.1109/cvpr46437.2021.00139	PROPN
alkej-870	485	15	http://dx.doi.org/10.1109/euvip50544.2021.9484036	http://dx.doi.org/10.1109/euvip50544.2021.9484036	PROPN
alkej-870	485	16	http://dx.doi.org/10.1109/euvip50544.2021.9484036	http://dx.doi.org/10.1109/euvip50544.2021.9484036	PROPN
alkej-870	485	17	http://dx.doi.org/10.1109/taslp.2022.3229261	http://dx.doi.org/10.1109/taslp.2022.3229261	PROPN
alkej-870	485	18	http://dx.doi.org/10.1109/taslp.2022.3229261	http://dx.doi.org/10.1109/taslp.2022.3229261	NOUN
alkej-870	485	19	http://dx.doi.org/10.1109/icpr56361.2022.9956384	http://dx.doi.org/10.1109/icpr56361.2022.9956384	PROPN
alkej-870	485	20	http://dx.doi.org/10.1109/icpr56361.2022.9956384	http://dx.doi.org/10.1109/icpr56361.2022.9956384	PROPN
alkej-870	485	21	http://dx.doi.org/10.1109/wacv51458.2022.00223	http://dx.doi.org/10.1109/wacv51458.2022.00223	PROPN
alkej-870	485	22	http://dx.doi.org/10.1109/wacv51458.2022.00223	http://dx.doi.org/10.1109/wacv51458.2022.00223	PROPN
alkej-870	485	23	http://dx.doi.org/10.23919/apsipaasc55919.2022.9980162	http://dx.doi.org/10.23919/apsipaasc55919.2022.9980162	PROPN
alkej-870	485	24	http://dx.doi.org/10.23919/apsipaasc55919.2022.9980162	http://dx.doi.org/10.23919/apsipaasc55919.2022.9980162	PROPN
alkej-870	485	25	http://dx.doi.org/10.1109/iccv.2019.00097	http://dx.doi.org/10.1109/iccv.2019.00097	PROPN
alkej-870	486	1	http://dx.doi.org/10.1109/iccv.2019.00097	http://dx.doi.org/10.1109/iccv.2019.00097	PROPN
alkej-870	487	1	http://dx.doi.org/10.1109/iccv.2019.00182	http://dx.doi.org/10.1109/iccv.2019.00182	NOUN
alkej-870	488	1	http://dx.doi.org/10.1109/iccv.2019.00182	http://dx.doi.org/10.1109/iccv.2019.00182	NOUN
alkej-870	488	2	http://dx.doi.org/10.1109/taslp.2019.2928140	http://dx.doi.org/10.1109/taslp.2019.2928140	PROPN
alkej-870	488	3	http://dx.doi.org/10.1109/taslp.2019.2928140	http://dx.doi.org/10.1109/taslp.2019.2928140	PROPN
alkej-870	488	4	http://dx.doi.org/10.1109/iccv.2019.00398	http://dx.doi.org/10.1109/iccv.2019.00398	PROPN
alkej-870	488	5	http://dx.doi.org/10.1109/iccv.2019.00398	http://dx.doi.org/10.1109/iccv.2019.00398	PROPN
alkej-870	488	6	http://dx.doi.org/10.1121/1.2229005	http://dx.doi.org/10.1121/1.2229005	PROPN
alkej-870	488	7	http://dx.doi.org/10.1121/1.5042758	http://dx.doi.org/10.1121/1.5042758	NOUN
alkej-870	489	1	http://dx.doi.org/10.1109/tmm.2015.2407694	http://dx.doi.org/10.1109/tmm.2015.2407694	PRON
alkej-870	489	2	http://dx.doi.org/10.1109/tmm.2015.2407694	http://dx.doi.org/10.1109/tmm.2015.2407694	PRON
alkej-870	489	3	http://dx.doi.org/10.1109/tmm.2009.2030637	http://dx.doi.org/10.1109/tmm.2009.2030637	PRON
alkej-870	490	1	http://dx.doi.org/10.1109/tmm.2009.2030637	http://dx.doi.org/10.1109/tmm.2009.2030637	NOUN
alkej-870	490	2	http://dx.doi.org/10.1109/fg.2015.7163155	http://dx.doi.org/10.1109/fg.2015.7163155	PROPN
alkej-870	490	3	http://dx.doi.org/10.1109/fg.2015.7163155	http://dx.doi.org/10.1109/fg.2015.7163155	PROPN
alkej-870	490	4	nooralhuda	nooralhuda	PROPN
alkej-870	490	5	mudhafar	mudhafar	PROPN
alkej-870	490	6	sleman	sleman	PROPN
alkej-870	490	7	al	al	PROPN
alkej-870	490	8	-	-	PUNCT
alkej-870	490	9	khwarizmi	khwarizmi	PROPN
alkej-870	490	10	engineering	engineering	NOUN
alkej-870	490	11	journal	journal	PROPN
alkej-870	490	12	,	,	PUNCT
alkej-870	490	13	vol	vol	NOUN
alkej-870	490	14	.	.	PROPN
alkej-870	490	15	19	19	NUM
alkej-870	490	16	,	,	PUNCT
alkej-870	490	17	no	no	INTJ
alkej-870	490	18	.	.	NOUN
alkej-870	490	19	4	4	NUM
alkej-870	490	20	,	,	PUNCT
alkej-870	490	21	p.p	p.p	PROPN
alkej-870	490	22	.	.	NOUN
alkej-870	490	23	4255	4255	NUM
alkej-870	490	24	(	(	PUNCT
alkej-870	490	25	2023	2023	NUM
alkej-870	490	26	)	)	PUNCT
alkej-870	490	27	54	54	NUM
alkej-870	490	28	association	association	NOUN
alkej-870	490	29	,	,	PUNCT
alkej-870	490	30	stockholm	stockholm	PROPN
alkej-870	490	31	,	,	PUNCT
alkej-870	490	32	sweden	sweden	PROPN
alkej-870	490	33	,	,	PUNCT
alkej-870	490	34	pp.2616−2620	pp.2616−2620	NOUN
alkej-870	490	35	,	,	PUNCT
alkej-870	490	36	2017	2017	NUM
alkej-870	490	37	.	.	PUNCT
alkej-870	491	1	http://dx.doi.org/10.21437/interspeech.2017	http://dx.doi.org/10.21437/interspeech.2017	X
alkej-870	492	1	-950	-950	X
alkej-870	492	2	.	.	PUNCT
alkej-870	493	1	[	[	X
alkej-870	493	2	56	56	NUM
alkej-870	493	3	]	]	PUNCT
alkej-870	493	4	j.	j.	PROPN
alkej-870	493	5	s.	s.	PROPN
alkej-870	493	6	chung	chung	PROPN
alkej-870	493	7	,	,	PUNCT
alkej-870	493	8	a.	a.	PROPN
alkej-870	493	9	nagrani	nagrani	PROPN
alkej-870	493	10	,	,	PUNCT
alkej-870	493	11	a.	a.	NOUN
alkej-870	493	12	zisserman	zisserman	NOUN
alkej-870	493	13	.	.	PUNCT
alkej-870	494	1	"	"	PUNCT
alkej-870	494	2	voxceleb2	voxceleb2	NOUN
alkej-870	494	3	:	:	PUNCT
alkej-870	494	4	deep	deep	ADJ
alkej-870	494	5	speaker	speaker	NOUN
alkej-870	494	6	recognition	recognition	NOUN
alkej-870	494	7	"	"	PUNCT
alkej-870	494	8	.	.	PUNCT
alkej-870	495	1	in	in	ADP
alkej-870	495	2	proceedings	proceeding	NOUN
alkej-870	495	3	of	of	ADP
alkej-870	495	4	the	the	DET
alkej-870	495	5	19th	19th	ADJ
alkej-870	495	6	annual	annual	ADJ
alkej-870	495	7	conference	conference	NOUN
alkej-870	495	8	of	of	ADP
alkej-870	495	9	the	the	DET
alkej-870	495	10	international	international	ADJ
alkej-870	495	11	speech	speech	NOUN
alkej-870	495	12	communication	communication	NOUN
alkej-870	495	13	association	association	PROPN
alkej-870	495	14	,	,	PUNCT
alkej-870	495	15	hyderabad	hyderabad	PROPN
alkej-870	495	16	,	,	PUNCT
alkej-870	495	17	india	india	PROPN
alkej-870	495	18	,	,	PUNCT
alkej-870	495	19	pp.1086−1090	pp.1086−1090	NOUN
alkej-870	495	20	,	,	PUNCT
alkej-870	495	21	2018	2018	NUM
alkej-870	495	22	.	.	PUNCT
alkej-870	496	1	http://dx.doi.org/10.21437/interspeech.2018	http://dx.doi.org/10.21437/interspeech.2018	NUM
alkej-870	496	2	-1929	-1929	NOUN
alkej-870	496	3	.	.	PUNCT
alkej-870	497	1	[	[	X
alkej-870	497	2	57	57	NUM
alkej-870	497	3	]	]	PUNCT
alkej-870	497	4	j.	j.	PROPN
alkej-870	497	5	s.	s.	PROPN
alkej-870	497	6	chung	chung	PROPN
alkej-870	497	7	,	,	PUNCT
alkej-870	497	8	a.	a.	NOUN
alkej-870	497	9	zisserman	zisserman	NOUN
alkej-870	497	10	.	.	PUNCT
alkej-870	498	1	"	"	PUNCT
alkej-870	498	2	lip	lip	NOUN
alkej-870	498	3	reading	read	VERB
alkej-870	498	4	in	in	ADP
alkej-870	498	5	the	the	DET
alkej-870	498	6	wild	wild	NOUN
alkej-870	498	7	"	"	PUNCT
alkej-870	498	8	.	.	PUNCT
alkej-870	499	1	in	in	ADP
alkej-870	499	2	proceedings	proceeding	NOUN
alkej-870	499	3	of	of	ADP
alkej-870	499	4	the	the	DET
alkej-870	499	5	13th	13th	ADJ
alkej-870	499	6	asian	asian	ADJ
alkej-870	499	7	conference	conference	NOUN
alkej-870	499	8	on	on	ADP
alkej-870	499	9	computer	computer	NOUN
alkej-870	499	10	vision	vision	NOUN
alkej-870	499	11	,	,	PUNCT
alkej-870	499	12	springer	springer	NOUN
alkej-870	499	13	,	,	PUNCT
alkej-870	499	14	taipei	taipei	PROPN
alkej-870	499	15	,	,	PUNCT
alkej-870	499	16	china	china	PROPN
alkej-870	499	17	,	,	PUNCT
alkej-870	499	18	pp.87−103	pp.87−103	PROPN
alkej-870	499	19	,	,	PUNCT
alkej-870	499	20	2017	2017	NUM
alkej-870	499	21	.	.	PUNCT
alkej-870	500	1	http://dx.doi.org/10.1007/978-3-319-541846_6	http://dx.doi.org/10.1007/978-3-319-541846_6	X
alkej-870	500	2	.	.	PUNCT
alkej-870	501	1	[	[	X
alkej-870	501	2	58	58	NUM
alkej-870	501	3	]	]	PUNCT
alkej-870	501	4	j.	j.	PROPN
alkej-870	501	5	s.	s.	PROPN
alkej-870	501	6	chung	chung	PROPN
alkej-870	501	7	,	,	PUNCT
alkej-870	501	8	a.	a.	NOUN
alkej-870	501	9	senior	senior	NOUN
alkej-870	501	10	,	,	PUNCT
alkej-870	501	11	o.	o.	PROPN
alkej-870	501	12	vinyals	vinyal	NOUN
alkej-870	501	13	,	,	PUNCT
alkej-870	501	14	a.	a.	NOUN
alkej-870	501	15	zisserman	zisserman	NOUN
alkej-870	501	16	.	.	PUNCT
alkej-870	502	1	"	"	PUNCT
alkej-870	502	2	lip	lip	NOUN
alkej-870	502	3	reading	read	VERB
alkej-870	502	4	sentences	sentence	NOUN
alkej-870	502	5	in	in	ADP
alkej-870	502	6	the	the	DET
alkej-870	502	7	wild	wild	NOUN
alkej-870	502	8	"	"	PUNCT
alkej-870	502	9	.	.	PUNCT
alkej-870	503	1	in	in	ADP
alkej-870	503	2	proceedings	proceeding	NOUN
alkej-870	503	3	of	of	ADP
alkej-870	503	4	ieee	ieee	NOUN
alkej-870	503	5	conference	conference	NOUN
alkej-870	503	6	on	on	ADP
alkej-870	503	7	computer	computer	NOUN
alkej-870	503	8	vision	vision	NOUN
alkej-870	503	9	and	and	CCONJ
alkej-870	503	10	pattern	pattern	NOUN
alkej-870	503	11	recognition	recognition	NOUN
alkej-870	503	12	,	,	PUNCT
alkej-870	503	13	ieee	ieee	PROPN
alkej-870	503	14	,	,	PUNCT
alkej-870	503	15	honolulu	honolulu	PROPN
alkej-870	503	16	,	,	PUNCT
alkej-870	503	17	usa	usa	PROPN
alkej-870	503	18	,	,	PUNCT
alkej-870	503	19	pp.3444−3453	pp.3444−3453	NOUN
alkej-870	503	20	,	,	PUNCT
alkej-870	503	21	2017	2017	NUM
alkej-870	503	22	.	.	PUNCT
alkej-870	504	1	http://dx.doi.org/10.1109/cvpr.2017.367	http://dx.doi.org/10.1109/cvpr.2017.367	X
alkej-870	504	2	.	.	PUNCT
alkej-870	505	1	[	[	X
alkej-870	505	2	59	59	NUM
alkej-870	505	3	]	]	PUNCT
alkej-870	505	4	j.	j.	PROPN
alkej-870	505	5	s.	s.	PROPN
alkej-870	505	6	chung	chung	PROPN
alkej-870	505	7	,	,	PUNCT
alkej-870	505	8	a.	a.	NOUN
alkej-870	505	9	zisserman	zisserman	NOUN
alkej-870	505	10	.	.	PUNCT
alkej-870	506	1	"	"	PUNCT
alkej-870	506	2	lip	lip	NOUN
alkej-870	506	3	reading	read	VERB
alkej-870	506	4	in	in	ADP
alkej-870	506	5	profile	profile	NOUN
alkej-870	506	6	"	"	PUNCT
alkej-870	506	7	.	.	PUNCT
alkej-870	507	1	in	in	ADP
alkej-870	507	2	proceedings	proceeding	NOUN
alkej-870	507	3	of	of	ADP
alkej-870	507	4	british	british	ADJ
alkej-870	507	5	machine	machine	NOUN
alkej-870	507	6	vision	vision	PROPN
alkej-870	507	7	conference	conference	PROPN
alkej-870	507	8	2017	2017	NUM
alkej-870	507	9	,	,	PUNCT
alkej-870	507	10	bmva	bmva	NOUN
alkej-870	507	11	press	press	PROPN
alkej-870	507	12	,	,	PUNCT
alkej-870	507	13	london	london	PROPN
alkej-870	507	14	,	,	PUNCT
alkej-870	507	15	uk	uk	PROPN
alkej-870	507	16	,	,	PUNCT
alkej-870	507	17	2017	2017	NUM
alkej-870	507	18	.	.	PUNCT
alkej-870	507	19	http://dx.doi.org/10.5244/c.31.155	http://dx.doi.org/10.5244/c.31.155	NOUN
alkej-870	507	20	.	.	PUNCT
alkej-870	508	1	[	[	X
alkej-870	508	2	60	60	NUM
alkej-870	508	3	]	]	X
alkej-870	508	4	j.	j.	PROPN
alkej-870	508	5	roth	roth	PROPN
alkej-870	508	6	et	et	PROPN
alkej-870	508	7	al	al	PROPN
alkej-870	508	8	.	.	PROPN
alkej-870	508	9	,	,	PUNCT
alkej-870	508	10	"	"	PUNCT
alkej-870	508	11	ava	ava	PROPN
alkej-870	508	12	active	active	ADJ
alkej-870	508	13	speaker	speaker	NOUN
alkej-870	508	14	:	:	PUNCT
alkej-870	508	15	an	an	DET
alkej-870	508	16	audio	audio	ADJ
alkej-870	508	17	-	-	ADJ
alkej-870	508	18	visual	visual	ADJ
alkej-870	508	19	dataset	dataset	NOUN
alkej-870	508	20	for	for	ADP
alkej-870	508	21	active	active	ADJ
alkej-870	508	22	speaker	speaker	NOUN
alkej-870	508	23	detection	detection	NOUN
alkej-870	508	24	,	,	PUNCT
alkej-870	508	25	"	"	PUNCT
alkej-870	508	26	ieee	ieee	PROPN
alkej-870	508	27	international	international	ADJ
alkej-870	508	28	conference	conference	NOUN
alkej-870	508	29	on	on	ADP
alkej-870	508	30	acoustics	acoustic	NOUN
alkej-870	508	31	,	,	PUNCT
alkej-870	508	32	speech	speech	NOUN
alkej-870	508	33	and	and	CCONJ
alkej-870	508	34	signal	signal	NOUN
alkej-870	508	35	processing	processing	NOUN
alkej-870	508	36	(	(	PUNCT
alkej-870	508	37	icassp	icassp	PROPN
alkej-870	508	38	)	)	PUNCT
alkej-870	508	39	,	,	PUNCT
alkej-870	508	40	barcelona	barcelona	PROPN
alkej-870	508	41	,	,	PUNCT
alkej-870	508	42	spain	spain	PROPN
alkej-870	508	43	,	,	PUNCT
alkej-870	508	44	2020	2020	NUM
alkej-870	508	45	,	,	PUNCT
alkej-870	508	46	pp	pp	ADJ
alkej-870	508	47	.	.	PUNCT
alkej-870	509	1	4492	4492	NUM
alkej-870	509	2	-	-	SYM
alkej-870	509	3	4496	4496	NUM
alkej-870	509	4	http://dx.doi.org/10.1109/icassp40776.20	http://dx.doi.org/10.1109/icassp40776.20	SYM
alkej-870	509	5	20.9053900	20.9053900	NUM
alkej-870	509	6	.	.	PUNCT
alkej-870	510	1	http://dx.doi.org/10.21437/interspeech.2017-950	http://dx.doi.org/10.21437/interspeech.2017-950	ADJ
alkej-870	510	2	http://dx.doi.org/10.21437/interspeech.2017-950	http://dx.doi.org/10.21437/interspeech.2017-950	PROPN
alkej-870	510	3	http://dx.doi.org/10.21437/interspeech.2018-1929	http://dx.doi.org/10.21437/interspeech.2018-1929	NOUN
alkej-870	510	4	http://dx.doi.org/10.21437/interspeech.2018-1929	http://dx.doi.org/10.21437/interspeech.2018-1929	NOUN
alkej-870	510	5	http://dx.doi.org/10.1007/978-3-319-54184-6_6	http://dx.doi.org/10.1007/978-3-319-54184-6_6	X
alkej-870	510	6	http://dx.doi.org/10.1007/978-3-319-54184-6_6	http://dx.doi.org/10.1007/978-3-319-54184-6_6	PROPN
alkej-870	510	7	http://dx.doi.org/10.1109/cvpr.2017.367	http://dx.doi.org/10.1109/cvpr.2017.367	PROPN
alkej-870	510	8	http://dx.doi.org/10.5244/c.31.155	http://dx.doi.org/10.5244/c.31.155	PROPN
alkej-870	510	9	http://dx.doi.org/10.1109/icassp40776.2020.9053900	http://dx.doi.org/10.1109/icassp40776.2020.9053900	PROPN
alkej-870	510	10	http://dx.doi.org/10.1109/icassp40776.2020.9053900	http://dx.doi.org/10.1109/icassp40776.2020.9053900	PROPN
alkej-870	510	11	(	(	PUNCT
alkej-870	510	12	2023	2023	NUM
alkej-870	510	13	)	)	PUNCT
alkej-870	510	14	42	42	NUM
alkej-870	510	15	-	-	SYM
alkej-870	510	16	55	55	NUM
alkej-870	510	17	،	،	NOUN
alkej-870	510	18	صفحة	صفحة	PROPN
alkej-870	510	19	4	4	NUM
alkej-870	510	20	،	،	NOUN
alkej-870	510	21	العدد19دجلة	العدد19دجلة	NOUN
alkej-870	510	22	الخوارزمي	الخوارزمي	VERB
alkej-870	510	23	الهندسية	الهندسية	PROPN
alkej-870	510	24	المجلم	المجلم	PROPN
alkej-870	510	25	نور	نور	PROPN
alkej-870	510	26	الهدى	الهدى	VERB
alkej-870	510	27	مظفر	مظفر	VERB
alkej-870	510	28	سليمان	سليمان	ADP
alkej-870	510	29	55	55	NUM
alkej-870	510	30	نظرة	نظرة	NOUN
alkej-870	510	31	عامة	عامة	ADJ
alkej-870	510	32	على	على	NOUN
alkej-870	510	33	فصل	فصل	NOUN
alkej-870	510	34	المصادر	المصادر	PROPN
alkej-870	510	35	السمعية	السمعية	PROPN
alkej-870	510	36	والبصرية	والبصرية	PROPN
alkej-870	510	37	باستخدام	باستخدام	NOUN
alkej-870	510	38	التعلم	التعلم	NOUN
alkej-870	510	39	العميق	العميق	VERB
alkej-870	510	40	محمد	محمد	ADJ
alkej-870	510	41	نجاح	نجاح	ADJ
alkej-870	510	42	مهدي	مهدي	PROPN
alkej-870	510	43	*	*	PUNCT
alkej-870	511	1	*	*	PUNCT
alkej-870	511	2	*	*	PUNCT
alkej-870	511	3	*	*	PUNCT
alkej-870	511	4	*	*	PUNCT
alkej-870	511	5	ميمياحمد	ميمياحمد	PROPN
alkej-870	511	6	الت	الت	PROPN
alkej-870	511	7	*	*	PUNCT
alkej-870	511	8	نورالهدى	نورالهدى	ADJ
alkej-870	511	9	مظفر	مظفر	VERB
alkej-870	511	10	سليمان	سليمان	NOUN
alkej-870	511	11	*	*	PUNCT
alkej-870	511	12	,	,	PUNCT
alkej-870	511	13	*	*	PUNCT
alkej-870	511	14	*	*	PUNCT
alkej-870	511	15	قسم	قسم	PRON
alkej-870	511	16	هندسة	هندسة	NOUN
alkej-870	511	17	المعلومات	المعلومات	VERB
alkej-870	511	18	واالتصاالت/	واالتصاالت/	NUM
alkej-870	511	19	كلية	كلية	NOUN
alkej-870	511	20	الهندسة	الهندسة	NOUN
alkej-870	511	21	الخوارزمي/	الخوارزمي/	NUM
alkej-870	511	22	جامعة	جامعة	NOUN
alkej-870	511	23	بغداد	بغداد	NOUN
alkej-870	511	24	/	/	SYM
alkej-870	511	25	كلية	كلية	PROPN
alkej-870	511	26	الحاسبات/	الحاسبات/	PROPN
alkej-870	511	27	جامعة	جامعة	NOUN
alkej-870	511	28	دبلن	دبلن	VERB
alkej-870	511	29	adaptمركز	adaptمركز	ADJ
alkej-870	511	30	*	*	PUNCT
alkej-870	511	31	*	*	PUNCT
alkej-870	511	32	*	*	PUNCT
alkej-870	511	33	noor.abd2103m@kecbu.uobaghdad.edu.iq	noor.abd2103m@kecbu.uobaghdad.edu.iq	NOUN
alkej-870	511	34	:	:	PUNCT
alkej-870	511	35	البريد	البريد	VERB
alkej-870	511	36	االلكتروني	االلكتروني	NOUN
alkej-870	511	37	*	*	PUNCT
alkej-870	511	38	asattar@kecbu.uobaghdad.edu.iq	asattar@kecbu.uobaghdad.edu.iq	PROPN
alkej-870	511	39	:	:	PUNCT
alkej-870	511	40	البريد	البريد	VERB
alkej-870	511	41	االلكتروني	االلكتروني	NOUN
alkej-870	511	42	*	*	NOUN
alkej-870	511	43	*	*	PUNCT
alkej-870	511	44	mohammed.mahdi@adaptcentre.ie	mohammed.mahdi@adaptcentre.ie	PROPN
alkej-870	511	45	:	:	PUNCT
alkej-870	511	46	البريد	البريد	VERB
alkej-870	511	47	االلكتروني	االلكتروني	NOUN
alkej-870	511	48	*	*	PUNCT
alkej-870	511	49	*	*	PUNCT
alkej-870	511	50	*	*	PUNCT
alkej-870	511	51	الخالصة	الخالصة	PROPN
alkej-870	511	52	نتمائ	نتمائ	NOUN
alkej-870	511	53	avss	avss	NOUN
alkej-870	511	54	(	(	PUNCT
alkej-870	511	55	القائممة	القائممة	PROPN
alkej-870	511	56	علمى	علمى	PROPN
alkej-870	511	57	تقنيمات	تقنيمات	PROPN
alkej-870	511	58	المتعلم	المتعلم	PROPN
alkej-870	511	59	العميمق	العميمق	PROPN
alkej-870	511	60	قمق	قمق	PROPN
alkej-870	511	61	avssأنظمة	avssأنظمة	PROPN
alkej-870	511	62	فصل	فصل	PROPN
alkej-870	511	63	المصادر	المصادر	PROPN
alkej-870	511	64	السممعية	السممعية	PROPN
alkej-870	511	65	والبصمرية	والبصمرية	PROPN
alkej-870	511	66	لىلبحث	لىلبحث	ADJ
alkej-870	511	67	المقدم	المقدم	NOUN
alkej-870	511	68	هنا	هنا	PROPN
alkej-870	511	69	لمحة	لمحة	PROPN
alkej-870	511	70	عامة	عامة	ADJ
alkej-870	511	71	عيعطي	عيعطي	PROPN
alkej-870	511	72	ا	ا	X
alkej-870	511	73	ايا	ايا	NOUN
alkej-870	511	74	وعيمو	وعيمو	NOUN
alkej-870	511	75	استثنائية	استثنائية	PROPN
alkej-870	511	76	في	في	ADP
alkej-870	511	77	عدد	عدد	PROPN
alkej-870	511	78	من	من	INTJ
alkej-870	511	79	المجاالت	المجاالت	PROPN
alkej-870	511	80	،	،	PROPN
alkej-870	511	81	بما	بما	PROPN
alkej-870	512	1	في	في	DET
alkej-870	512	2	ذلك	ذلك	PROPN
alkej-870	512	3	تقليل	تقليل	PROPN
alkej-870	512	4	مستويات	مستويات	PROPN
alkej-870	512	5	الضوضاء	الضوضاء	NOUN
alkej-870	512	6	،	،	PROPN
alkej-870	512	7	وتعزيز	وتعزيز	PROPN
alkej-870	512	8	التعرف	التعرف	PROPN
alkej-870	512	9	على	على	NOUN
alkej-870	512	10	الكالم	الكالم	PROPN
alkej-870	512	11	،	،	PROPN
alkej-870	512	12	وتحسين	وتحسين	PROPN
alkej-870	512	13	جودة	جودة	PROPN
alkej-870	512	14	الصوت	الصوت	PROPN
alkej-870	512	15	تمت	تمت	PROPN
alkej-870	512	16	مناقشة	مناقشة	PROPN
alkej-870	512	17	مز	مز	AUX
alkej-870	512	18	التمي	التمي	VERB
alkej-870	512	19	تحتمو	تحتمو	PROPN
alkej-870	512	20	tcd	tcd	PROPN
alkej-870	512	21	timit	timit	PROPN
alkej-870	512	22	مجموعمة	مجموعمة	PROPN
alkej-870	512	23	بيانمات	بيانمات	VERB
alkej-870	512	24	avssكل	avssكل	NOUN
alkej-870	512	25	نموذج	نموذج	NOUN
alkej-870	512	26	من	من	PROPN
alkej-870	512	27	نماذج	نماذج	VERB
alkej-870	512	28	التعلم	التعلم	PROPN
alkej-870	512	29	العميق	العميق	PROPN
alkej-870	512	30	خالل	خالل	PROPN
alkej-870	512	31	البحث	البحث	VERB
alkej-870	512	32	يث	يث	PROPN
alkej-870	512	33	يقوم	يقوم	PROPN
alkej-870	512	34	بمراجعة	بمراجعة	NOUN
alkej-870	512	35	مختلف	مختلف	NOUN
alkej-870	512	36	التجار	التجار	VERB
alkej-870	512	37	الحالية	الحالية	ADP
alkej-870	512	38	على	على	PROPN
alkej-870	512	39	مقاطع	مقاطع	ADJ
alkej-870	512	40	مجموعة	مجموعة	NOUN
alkej-870	512	41	كبيرة	كبيرة	NOUN
alkej-870	512	42	من	من	PROPN
alkej-870	512	43	ال	ال	PROPN
alkej-870	512	44	voxcelebعلى	voxcelebعلى	PROPN
alkej-870	512	45	تسجيالت	تسجيالت	PROPN
alkej-870	512	46	صوتية	صوتية	PROPN
alkej-870	512	47	ومرئية	ومرئية	PROPN
alkej-870	512	48	من	من	DET
alkej-870	512	49	الدرجة	الدرجة	PROPN
alkej-870	512	50	األولى	األولى	PROPN
alkej-870	512	51	تم	تم	NOUN
alkej-870	512	52	إنشاؤها	إنشاؤها	ADJ
alkej-870	512	53	خصيًصا	خصيًصا	PUNCT
alkej-870	512	54	لمهام	لمهام	PROPN
alkej-870	512	55	التعرف	التعرف	PROPN
alkej-870	512	56	على	على	NOUN
alkej-870	512	57	الكالم	الكالم	NOUN
alkej-870	512	58	(	(	PUNCT
alkej-870	512	59	ومجموعة	ومجموعة	PROPN
alkej-870	512	60	بيانات	بيانات	PROPN
alkej-870	512	61	مجموعات	مجموعات	PROPN
alkej-870	512	62	البيانات	البيانات	PROPN
alkej-870	512	63	المفيدة	المفيدة	PROPN
alkej-870	513	1	التي	التي	PROPN
alkej-870	513	2	تم	تم	PROPN
alkej-870	514	1	تلخيصها	تلخيصها	PROPN
alkej-870	514	2	في	في	ADP
alkej-870	514	3	هذا	هذا	NOUN
alkej-870	514	4	البحث	البحث	NOUN
alkej-870	514	5	و	و	PRON
alkej-870	514	6	التي	التي	PROPN
alkej-870	514	7	يمكن	يمكن	PROPN
alkej-870	514	8	اسمتخدامها	اسمتخدامها	ADV
alkej-870	514	9	الختبمار	الختبمار	ADJ
alkej-870	514	10	أنظممة	أنظممة	PROPN
alkej-870	514	11	هي	هي	PROPN
alkej-870	514	12	ا	ا	PROPN
alkej-870	514	13	دكالم	دكالم	PROPN
alkej-870	514	14	البشر	البشر	PROPN
alkej-870	514	15	(	(	PUNCT
alkej-870	514	16	ة	ة	PRON
alkej-870	514	17	للالسمعية	للالسمعية	PROPN
alkej-870	514	18	والبصرية	والبصرية	ADJ
alkej-870	514	19	المختصر	المختصر	PROPN
alkej-870	514	20	avss	avss	PROPN
alkej-870	514	21	الغرض	الغرض	PROPN
alkej-870	514	22	من	من	PROPN
alkej-870	514	23	هذه	هذه	PROPN
alkej-870	514	24	المراجعة	المراجعة	NOUN
alkej-870	514	25	في	في	ADP
alkej-870	514	26	شكلها	شكلها	NOUN
alkej-870	514	27	األساسي	األساسي	VERB
alkej-870	514	28	هو	هو	PROPN
alkej-870	514	29	تسليط	تسليط	NOUN
alkej-870	514	30	الضوء	الضوء	NOUN
alkej-870	514	31	على	على	NOUN
alkej-870	514	32	األهمية	األهمية	ADJ
alkej-870	514	33	المتزايدة	المتزايدة	ADJ
alkej-870	514	34	لـ	لـ	INTJ
alkej-870	514	35	avss	avss	NOUN
alkej-870	514	36	ات	ات	ADP
alkej-870	514	37	الصوتية	الصوتية	PROPN
alkej-870	514	38	في	في	ADP
alkej-870	515	1	تحسين	تحسين	PROPN
alkej-870	515	2	جودة	جودة	PROPN
alkej-870	515	3	اإلشار	اإلشار	ADJ
alkej-870	515	4	mailto:noor.abd2103m@kecbu.uobaghdad.edu.iq	mailto:noor.abd2103m@kecbu.uobaghdad.edu.iq	NOUN
alkej-870	515	5	mailto:asattar@kecbu.uobaghdad.edu.iq	mailto:asattar@kecbu.uobaghdad.edu.iq	PROPN
alkej-870	515	6	mailto:mohammed.mahdi@adaptcentre.ie	mailto:mohammed.mahdi@adaptcentre.ie	PROPN
alkej-870	515	7	acknowledgments	acknowledgment	NOUN