id	sid	tid	token	lemma	pos
ajst-28825	1	1	academic	academic	ADJ
ajst-28825	1	2	journal	journal	NOUN
ajst-28825	1	3	of	of	ADP
ajst-28825	1	4	science	science	NOUN
ajst-28825	1	5	and	and	CCONJ
ajst-28825	1	6	technology	technology	NOUN
ajst-28825	1	7	issn	issn	NOUN
ajst-28825	1	8	:	:	PUNCT
ajst-28825	1	9	2771	2771	NUM
ajst-28825	1	10	-	-	SYM
ajst-28825	1	11	3032	3032	NUM
ajst-28825	1	12	|	|	NOUN
ajst-28825	1	13	vol	vol	NOUN
ajst-28825	1	14	.	.	PROPN
ajst-28825	1	15	13	13	NUM
ajst-28825	1	16	,	,	PUNCT
ajst-28825	1	17	no	no	INTJ
ajst-28825	1	18	.	.	NOUN
ajst-28825	1	19	3	3	NUM
ajst-28825	1	20	,	,	PUNCT
ajst-28825	1	21	2024	2024	NUM
ajst-28825	1	22	343	343	NUM
ajst-28825	1	23	rt‐detr‐based	rt‐detr‐base	VERB
ajst-28825	1	24	signal	signal	NOUN
ajst-28825	1	25	modulation	modulation	NOUN
ajst-28825	1	26	recognition	recognition	NOUN
ajst-28825	1	27	with	with	ADP
ajst-28825	1	28	aifi‐dattention	aifi‐dattention	NOUN
ajst-28825	1	29	minghao	minghao	PROPN
ajst-28825	1	30	cao1	cao1	PROPN
ajst-28825	1	31	,	,	PUNCT
ajst-28825	1	32	peng	peng	PROPN
ajst-28825	1	33	chu1	chu1	PROPN
ajst-28825	1	34	,	,	PUNCT
ajst-28825	1	35	*	*	PROPN
ajst-28825	1	36	,	,	PUNCT
ajst-28825	1	37	hongjie	hongjie	ADJ
ajst-28825	1	38	guo2	guo2	PROPN
ajst-28825	1	39	,	,	PUNCT
ajst-28825	1	40	wei	wei	PROPN
ajst-28825	1	41	xing3	xing3	PUNCT
ajst-28825	2	1	1school	1school	NUM
ajst-28825	2	2	of	of	ADP
ajst-28825	2	3	electronic	electronic	ADJ
ajst-28825	2	4	information	information	NOUN
ajst-28825	2	5	,	,	PUNCT
ajst-28825	2	6	xijing	xijing	PROPN
ajst-28825	2	7	university	university	PROPN
ajst-28825	2	8	,	,	PUNCT
ajst-28825	2	9	xian	xian	PROPN
ajst-28825	2	10	,	,	PUNCT
ajst-28825	2	11	china	china	PROPN
ajst-28825	2	12	2xi’anvocational	2xi’anvocational	PROPN
ajst-28825	2	13	university	university	PROPN
ajst-28825	2	14	of	of	ADP
ajst-28825	2	15	information	information	NOUN
ajst-28825	2	16	,	,	PUNCT
ajst-28825	2	17	xian	xian	PROPN
ajst-28825	2	18	,	,	PUNCT
ajst-28825	2	19	china	china	PROPN
ajst-28825	2	20	3unit	3unit	NUM
ajst-28825	2	21	93117	93117	NUM
ajst-28825	2	22	,	,	PUNCT
ajst-28825	2	23	nanjing	nanjing	PROPN
ajst-28825	2	24	,	,	PUNCT
ajst-28825	2	25	china	china	PROPN
ajst-28825	2	26	*	*	PUNCT
ajst-28825	2	27	corresponding	correspond	VERB
ajst-28825	2	28	author	author	NOUN
ajst-28825	2	29	abstract	abstract	NOUN
ajst-28825	2	30	:	:	PUNCT
ajst-28825	2	31	in	in	ADP
ajst-28825	2	32	response	response	NOUN
ajst-28825	2	33	to	to	ADP
ajst-28825	2	34	the	the	DET
ajst-28825	2	35	issues	issue	NOUN
ajst-28825	2	36	of	of	ADP
ajst-28825	2	37	high	high	ADJ
ajst-28825	2	38	computational	computational	ADJ
ajst-28825	2	39	complexity	complexity	NOUN
ajst-28825	2	40	,	,	PUNCT
ajst-28825	2	41	low	low	ADJ
ajst-28825	2	42	accuracy	accuracy	NOUN
ajst-28825	2	43	,	,	PUNCT
ajst-28825	2	44	and	and	CCONJ
ajst-28825	2	45	cumbersome	cumbersome	ADJ
ajst-28825	2	46	manual	manual	ADJ
ajst-28825	2	47	feature	feature	NOUN
ajst-28825	2	48	extraction	extraction	NOUN
ajst-28825	2	49	steps	step	NOUN
ajst-28825	2	50	in	in	ADP
ajst-28825	2	51	traditional	traditional	ADJ
ajst-28825	2	52	machine	machine	NOUN
ajst-28825	2	53	learning	learn	VERB
ajst-28825	2	54	algorithms	algorithm	NOUN
ajst-28825	2	55	for	for	ADP
ajst-28825	2	56	communication	communication	NOUN
ajst-28825	2	57	signal	signal	NOUN
ajst-28825	2	58	modulation	modulation	NOUN
ajst-28825	2	59	recognition	recognition	NOUN
ajst-28825	2	60	,	,	PUNCT
ajst-28825	2	61	a	a	DET
ajst-28825	2	62	communication	communication	NOUN
ajst-28825	2	63	signal	signal	NOUN
ajst-28825	2	64	modulation	modulation	NOUN
ajst-28825	2	65	recognition	recognition	NOUN
ajst-28825	2	66	model	model	NOUN
ajst-28825	2	67	based	base	VERB
ajst-28825	2	68	on	on	ADP
ajst-28825	2	69	deep	deep	ADJ
ajst-28825	2	70	learning	learning	NOUN
ajst-28825	2	71	is	be	AUX
ajst-28825	2	72	proposed	propose	VERB
ajst-28825	2	73	.	.	PUNCT
ajst-28825	3	1	this	this	DET
ajst-28825	3	2	model	model	NOUN
ajst-28825	3	3	can	can	AUX
ajst-28825	3	4	directly	directly	ADV
ajst-28825	3	5	recognize	recognize	VERB
ajst-28825	3	6	the	the	DET
ajst-28825	3	7	category	category	NOUN
ajst-28825	3	8	of	of	ADP
ajst-28825	3	9	communication	communication	NOUN
ajst-28825	3	10	signals	signal	NOUN
ajst-28825	3	11	after	after	ADP
ajst-28825	3	12	sampling	sample	VERB
ajst-28825	3	13	and	and	CCONJ
ajst-28825	3	14	is	be	AUX
ajst-28825	3	15	characterized	characterize	VERB
ajst-28825	3	16	by	by	ADP
ajst-28825	3	17	high	high	ADJ
ajst-28825	3	18	recognition	recognition	NOUN
ajst-28825	3	19	accuracy	accuracy	NOUN
ajst-28825	3	20	,	,	PUNCT
ajst-28825	3	21	strong	strong	ADJ
ajst-28825	3	22	generalization	generalization	NOUN
ajst-28825	3	23	capability	capability	NOUN
ajst-28825	3	24	,	,	PUNCT
ajst-28825	3	25	good	good	ADJ
ajst-28825	3	26	noise	noise	NOUN
ajst-28825	3	27	resistance	resistance	NOUN
ajst-28825	3	28	,	,	PUNCT
ajst-28825	3	29	and	and	CCONJ
ajst-28825	3	30	a	a	DET
ajst-28825	3	31	simplified	simplified	ADJ
ajst-28825	3	32	processing	processing	NOUN
ajst-28825	3	33	flow	flow	NOUN
ajst-28825	3	34	.	.	PUNCT
ajst-28825	4	1	it	it	PRON
ajst-28825	4	2	effectively	effectively	ADV
ajst-28825	4	3	addresses	address	VERB
ajst-28825	4	4	the	the	DET
ajst-28825	4	5	limitations	limitation	NOUN
ajst-28825	4	6	of	of	ADP
ajst-28825	4	7	traditional	traditional	ADJ
ajst-28825	4	8	algorithms	algorithm	NOUN
ajst-28825	4	9	in	in	ADP
ajst-28825	4	10	automatic	automatic	ADJ
ajst-28825	4	11	feature	feature	NOUN
ajst-28825	4	12	extraction	extraction	NOUN
ajst-28825	4	13	.	.	PUNCT
ajst-28825	5	1	through	through	ADP
ajst-28825	5	2	extensive	extensive	ADJ
ajst-28825	5	3	experiments	experiment	NOUN
ajst-28825	5	4	and	and	CCONJ
ajst-28825	5	5	accurate	accurate	ADJ
ajst-28825	5	6	analysis	analysis	NOUN
ajst-28825	5	7	of	of	ADP
ajst-28825	5	8	communication	communication	NOUN
ajst-28825	5	9	signal	signal	NOUN
ajst-28825	5	10	features	feature	NOUN
ajst-28825	5	11	,	,	PUNCT
ajst-28825	5	12	an	an	DET
ajst-28825	5	13	end	end	NOUN
ajst-28825	5	14	-	-	PUNCT
ajst-28825	5	15	to	to	ADP
ajst-28825	5	16	-	-	PUNCT
ajst-28825	5	17	end	end	NOUN
ajst-28825	5	18	model	model	NOUN
ajst-28825	5	19	based	base	VERB
ajst-28825	5	20	on	on	ADP
ajst-28825	5	21	the	the	DET
ajst-28825	5	22	transformer	transformer	NOUN
ajst-28825	5	23	rt	rt	PROPN
ajst-28825	5	24	-	-	PUNCT
ajst-28825	5	25	detr	detr	NOUN
ajst-28825	5	26	model	model	NOUN
ajst-28825	5	27	is	be	AUX
ajst-28825	5	28	adopted	adopt	VERB
ajst-28825	5	29	,	,	PUNCT
ajst-28825	5	30	achieving	achieve	VERB
ajst-28825	5	31	high	high	ADJ
ajst-28825	5	32	recognition	recognition	NOUN
ajst-28825	5	33	accuracy	accuracy	NOUN
ajst-28825	5	34	.	.	PUNCT
ajst-28825	6	1	keywords	keyword	NOUN
ajst-28825	6	2	:	:	PUNCT
ajst-28825	6	3	feature	feature	NOUN
ajst-28825	6	4	extraction	extraction	NOUN
ajst-28825	6	5	networks	network	NOUN
ajst-28825	6	6	.	.	PUNCT
ajst-28825	7	1	signal	signal	ADJ
ajst-28825	7	2	modulation	modulation	NOUN
ajst-28825	7	3	recognition	recognition	NOUN
ajst-28825	7	4	.	.	PUNCT
ajst-28825	8	1	deep	deep	ADJ
ajst-28825	8	2	learning	learning	NOUN
ajst-28825	8	3	.	.	PUNCT
ajst-28825	9	1	1	1	X
ajst-28825	9	2	.	.	X
ajst-28825	9	3	introduction	introduction	NOUN
ajst-28825	9	4	with	with	ADP
ajst-28825	9	5	the	the	DET
ajst-28825	9	6	rapid	rapid	ADJ
ajst-28825	9	7	development	development	NOUN
ajst-28825	9	8	of	of	ADP
ajst-28825	9	9	wireless	wireless	ADJ
ajst-28825	9	10	communication	communication	NOUN
ajst-28825	9	11	and	and	CCONJ
ajst-28825	9	12	signal	signal	NOUN
ajst-28825	9	13	processing	processing	NOUN
ajst-28825	9	14	technologies	technology	NOUN
ajst-28825	9	15	,	,	PUNCT
ajst-28825	9	16	signal	signal	ADJ
ajst-28825	9	17	modulation	modulation	NOUN
ajst-28825	9	18	recognition	recognition	NOUN
ajst-28825	9	19	plays	play	VERB
ajst-28825	9	20	a	a	DET
ajst-28825	9	21	crucial	crucial	ADJ
ajst-28825	9	22	role	role	NOUN
ajst-28825	9	23	in	in	ADP
ajst-28825	9	24	fields	field	NOUN
ajst-28825	9	25	such	such	ADJ
ajst-28825	9	26	as	as	ADP
ajst-28825	9	27	military	military	ADJ
ajst-28825	9	28	communication	communication	NOUN
ajst-28825	9	29	,	,	PUNCT
ajst-28825	9	30	radio	radio	NOUN
ajst-28825	9	31	monitoring	monitoring	NOUN
ajst-28825	9	32	,	,	PUNCT
ajst-28825	9	33	and	and	CCONJ
ajst-28825	9	34	cognitive	cognitive	ADJ
ajst-28825	9	35	radio	radio	NOUN
ajst-28825	9	36	.	.	PUNCT
ajst-28825	10	1	the	the	DET
ajst-28825	10	2	task	task	NOUN
ajst-28825	10	3	of	of	ADP
ajst-28825	10	4	signal	signal	ADJ
ajst-28825	10	5	modulation	modulation	NOUN
ajst-28825	10	6	recognition	recognition	NOUN
ajst-28825	10	7	is	be	AUX
ajst-28825	10	8	to	to	PART
ajst-28825	10	9	extract	extract	VERB
ajst-28825	10	10	modulation	modulation	NOUN
ajst-28825	10	11	information	information	NOUN
ajst-28825	10	12	from	from	ADP
ajst-28825	10	13	the	the	DET
ajst-28825	10	14	received	receive	VERB
ajst-28825	10	15	signals	signal	NOUN
ajst-28825	10	16	to	to	PART
ajst-28825	10	17	determine	determine	VERB
ajst-28825	10	18	the	the	DET
ajst-28825	10	19	signal	signal	NOUN
ajst-28825	10	20	's	's	PART
ajst-28825	10	21	modulation	modulation	NOUN
ajst-28825	10	22	type	type	NOUN
ajst-28825	10	23	.	.	PUNCT
ajst-28825	11	1	traditional	traditional	ADJ
ajst-28825	11	2	signal	signal	NOUN
ajst-28825	11	3	modulation	modulation	NOUN
ajst-28825	11	4	recognition	recognition	NOUN
ajst-28825	11	5	methods	method	NOUN
ajst-28825	11	6	often	often	ADV
ajst-28825	11	7	rely	rely	VERB
ajst-28825	11	8	on	on	ADP
ajst-28825	11	9	manually	manually	ADV
ajst-28825	11	10	designed	design	VERB
ajst-28825	11	11	feature	feature	NOUN
ajst-28825	11	12	extraction	extraction	NOUN
ajst-28825	11	13	and	and	CCONJ
ajst-28825	11	14	classification	classification	NOUN
ajst-28825	11	15	algorithms	algorithm	NOUN
ajst-28825	11	16	,	,	PUNCT
ajst-28825	11	17	but	but	CCONJ
ajst-28825	11	18	these	these	DET
ajst-28825	11	19	methods	method	NOUN
ajst-28825	11	20	struggle	struggle	VERB
ajst-28825	11	21	to	to	PART
ajst-28825	11	22	cope	cope	VERB
ajst-28825	11	23	in	in	ADP
ajst-28825	11	24	complex	complex	ADJ
ajst-28825	11	25	electromagnetic	electromagnetic	ADJ
ajst-28825	11	26	environments	environment	NOUN
ajst-28825	11	27	and	and	CCONJ
ajst-28825	11	28	have	have	VERB
ajst-28825	11	29	limited	limit	VERB
ajst-28825	11	30	recognition	recognition	NOUN
ajst-28825	11	31	accuracy[1	accuracy[1	PUNCT
ajst-28825	11	32	]	]	PUNCT
ajst-28825	11	33	.	.	PUNCT
ajst-28825	12	1	in	in	ADP
ajst-28825	12	2	recent	recent	ADJ
ajst-28825	12	3	years	year	NOUN
ajst-28825	12	4	,	,	PUNCT
ajst-28825	12	5	with	with	ADP
ajst-28825	12	6	the	the	DET
ajst-28825	12	7	rise	rise	NOUN
ajst-28825	12	8	of	of	ADP
ajst-28825	12	9	deep	deep	ADJ
ajst-28825	12	10	learning	learning	NOUN
ajst-28825	12	11	,	,	PUNCT
ajst-28825	12	12	researchers	researcher	NOUN
ajst-28825	12	13	have	have	AUX
ajst-28825	12	14	begun	begin	VERB
ajst-28825	12	15	exploring	explore	VERB
ajst-28825	12	16	the	the	DET
ajst-28825	12	17	use	use	NOUN
ajst-28825	12	18	of	of	ADP
ajst-28825	12	19	deep	deep	ADJ
ajst-28825	12	20	neural	neural	ADJ
ajst-28825	12	21	networks	network	NOUN
ajst-28825	12	22	for	for	ADP
ajst-28825	12	23	signal	signal	ADJ
ajst-28825	12	24	modulation	modulation	NOUN
ajst-28825	12	25	recognition	recognition	NOUN
ajst-28825	12	26	.	.	PUNCT
ajst-28825	13	1	deep	deep	ADJ
ajst-28825	13	2	learning	learning	NOUN
ajst-28825	13	3	methods	method	NOUN
ajst-28825	13	4	can	can	AUX
ajst-28825	13	5	automatically	automatically	ADV
ajst-28825	13	6	learn	learn	VERB
ajst-28825	13	7	and	and	CCONJ
ajst-28825	13	8	extract	extract	VERB
ajst-28825	13	9	features	feature	NOUN
ajst-28825	13	10	from	from	ADP
ajst-28825	13	11	signals	signal	NOUN
ajst-28825	13	12	,	,	PUNCT
ajst-28825	13	13	demonstrating	demonstrate	VERB
ajst-28825	13	14	greater	great	ADJ
ajst-28825	13	15	robustness	robustness	NOUN
ajst-28825	13	16	and	and	CCONJ
ajst-28825	13	17	higher	high	ADJ
ajst-28825	13	18	recognition	recognition	NOUN
ajst-28825	13	19	accuracy	accuracy	NOUN
ajst-28825	13	20	in	in	ADP
ajst-28825	13	21	complex	complex	ADJ
ajst-28825	13	22	environments	environment	NOUN
ajst-28825	13	23	.	.	PUNCT
ajst-28825	14	1	however	however	ADV
ajst-28825	14	2	,	,	PUNCT
ajst-28825	14	3	existing	exist	VERB
ajst-28825	14	4	deep	deep	ADJ
ajst-28825	14	5	learning	learning	NOUN
ajst-28825	14	6	-	-	PUNCT
ajst-28825	14	7	based	base	VERB
ajst-28825	14	8	modulation	modulation	NOUN
ajst-28825	14	9	recognition	recognition	NOUN
ajst-28825	14	10	methods	method	NOUN
ajst-28825	14	11	still	still	ADV
ajst-28825	14	12	face	face	VERB
ajst-28825	14	13	trade	trade	NOUN
ajst-28825	14	14	-	-	PUNCT
ajst-28825	14	15	offs	off	NOUN
ajst-28825	14	16	between	between	ADP
ajst-28825	14	17	model	model	NOUN
ajst-28825	14	18	complexity	complexity	NOUN
ajst-28825	14	19	,	,	PUNCT
ajst-28825	14	20	computational	computational	ADJ
ajst-28825	14	21	efficiency	efficiency	NOUN
ajst-28825	14	22	,	,	PUNCT
ajst-28825	14	23	and	and	CCONJ
ajst-28825	14	24	recognition	recognition	NOUN
ajst-28825	14	25	accuracy	accuracy	NOUN
ajst-28825	14	26	.	.	PUNCT
ajst-28825	15	1	therefore	therefore	ADV
ajst-28825	15	2	,	,	PUNCT
ajst-28825	15	3	designing	design	VERB
ajst-28825	15	4	a	a	DET
ajst-28825	15	5	deep	deep	ADJ
ajst-28825	15	6	learning	learning	NOUN
ajst-28825	15	7	model	model	NOUN
ajst-28825	15	8	that	that	PRON
ajst-28825	15	9	can	can	AUX
ajst-28825	15	10	maintain	maintain	VERB
ajst-28825	15	11	high	high	ADJ
ajst-28825	15	12	recognition	recognition	NOUN
ajst-28825	15	13	accuracy	accuracy	NOUN
ajst-28825	15	14	while	while	SCONJ
ajst-28825	15	15	reducing	reduce	VERB
ajst-28825	15	16	computational	computational	ADJ
ajst-28825	15	17	complexity	complexity	NOUN
ajst-28825	15	18	has	have	AUX
ajst-28825	15	19	become	become	VERB
ajst-28825	15	20	an	an	DET
ajst-28825	15	21	important	important	ADJ
ajst-28825	15	22	research	research	NOUN
ajst-28825	15	23	direction	direction	NOUN
ajst-28825	15	24	.	.	PUNCT
ajst-28825	16	1	in	in	ADP
ajst-28825	16	2	this	this	DET
ajst-28825	16	3	study	study	NOUN
ajst-28825	16	4	,	,	PUNCT
ajst-28825	16	5	we	we	PRON
ajst-28825	16	6	employ	employ	VERB
ajst-28825	16	7	a	a	DET
ajst-28825	16	8	novel	novel	ADJ
ajst-28825	16	9	rt	rt	NOUN
ajst-28825	16	10	-	-	PUNCT
ajst-28825	16	11	detr	detr	NOUN
ajst-28825	16	12	model	model	NOUN
ajst-28825	16	13	for	for	ADP
ajst-28825	16	14	signal	signal	ADJ
ajst-28825	16	15	modulation	modulation	NOUN
ajst-28825	16	16	recognition	recognition	NOUN
ajst-28825	16	17	.	.	PUNCT
ajst-28825	17	1	the	the	DET
ajst-28825	17	2	rt	rt	PROPN
ajst-28825	17	3	-	-	PUNCT
ajst-28825	17	4	detr	detr	NOUN
ajst-28825	17	5	model	model	NOUN
ajst-28825	17	6	,	,	PUNCT
ajst-28825	17	7	as	as	ADP
ajst-28825	17	8	a	a	DET
ajst-28825	17	9	lightweight	lightweight	ADJ
ajst-28825	17	10	object	object	NOUN
ajst-28825	17	11	detection	detection	NOUN
ajst-28825	17	12	model	model	NOUN
ajst-28825	17	13	,	,	PUNCT
ajst-28825	17	14	offers	offer	VERB
ajst-28825	17	15	high	high	ADJ
ajst-28825	17	16	computational	computational	ADJ
ajst-28825	17	17	efficiency	efficiency	NOUN
ajst-28825	17	18	and	and	CCONJ
ajst-28825	17	19	strong	strong	ADJ
ajst-28825	17	20	generalization	generalization	NOUN
ajst-28825	17	21	ability	ability	NOUN
ajst-28825	17	22	.	.	PUNCT
ajst-28825	18	1	to	to	PART
ajst-28825	18	2	further	far	ADV
ajst-28825	18	3	enhance	enhance	VERB
ajst-28825	18	4	the	the	DET
ajst-28825	18	5	model	model	NOUN
ajst-28825	18	6	's	's	PART
ajst-28825	18	7	performance	performance	NOUN
ajst-28825	18	8	,	,	PUNCT
ajst-28825	18	9	we	we	PRON
ajst-28825	18	10	selected	select	VERB
ajst-28825	18	11	resnet-18	resnet-18	PROPN
ajst-28825	18	12	(	(	PUNCT
ajst-28825	18	13	r18	r18	NOUN
ajst-28825	18	14	)	)	PUNCT
ajst-28825	18	15	as	as	ADP
ajst-28825	18	16	the	the	DET
ajst-28825	18	17	backbone	backbone	NOUN
ajst-28825	18	18	and	and	CCONJ
ajst-28825	18	19	integrated	integrate	VERB
ajst-28825	18	20	the	the	DET
ajst-28825	18	21	aifi	aifi	NOUN
ajst-28825	18	22	-	-	PUNCT
ajst-28825	18	23	dattention	dattention	NOUN
ajst-28825	18	24	module	module	NOUN
ajst-28825	18	25	.	.	PUNCT
ajst-28825	19	1	these	these	DET
ajst-28825	19	2	improvements	improvement	NOUN
ajst-28825	19	3	aim	aim	VERB
ajst-28825	19	4	to	to	PART
ajst-28825	19	5	enhance	enhance	VERB
ajst-28825	19	6	the	the	DET
ajst-28825	19	7	model	model	NOUN
ajst-28825	19	8	's	's	PART
ajst-28825	19	9	feature	feature	NOUN
ajst-28825	19	10	extraction	extraction	NOUN
ajst-28825	19	11	capabilities	capability	NOUN
ajst-28825	19	12	and	and	CCONJ
ajst-28825	19	13	optimize	optimize	VERB
ajst-28825	19	14	feature	feature	NOUN
ajst-28825	19	15	fusion	fusion	NOUN
ajst-28825	19	16	efficiency	efficiency	NOUN
ajst-28825	19	17	,	,	PUNCT
ajst-28825	19	18	thereby	thereby	ADV
ajst-28825	19	19	improving	improve	VERB
ajst-28825	19	20	recognition	recognition	NOUN
ajst-28825	19	21	accuracy	accuracy	NOUN
ajst-28825	19	22	while	while	SCONJ
ajst-28825	19	23	keeping	keep	VERB
ajst-28825	19	24	the	the	DET
ajst-28825	19	25	model	model	NOUN
ajst-28825	19	26	lightweight	lightweight	NOUN
ajst-28825	19	27	.	.	PUNCT
ajst-28825	20	1	the	the	DET
ajst-28825	20	2	main	main	ADJ
ajst-28825	20	3	contributions	contribution	NOUN
ajst-28825	20	4	of	of	ADP
ajst-28825	20	5	this	this	DET
ajst-28825	20	6	study	study	NOUN
ajst-28825	20	7	include	include	VERB
ajst-28825	20	8	:	:	PUNCT
ajst-28825	20	9	we	we	PRON
ajst-28825	20	10	propose	propose	VERB
ajst-28825	20	11	a	a	DET
ajst-28825	20	12	lightweight	lightweight	ADJ
ajst-28825	20	13	signal	signal	NOUN
ajst-28825	20	14	modulation	modulation	NOUN
ajst-28825	20	15	recognition	recognition	NOUN
ajst-28825	20	16	method	method	NOUN
ajst-28825	20	17	based	base	VERB
ajst-28825	20	18	on	on	ADP
ajst-28825	20	19	rt	rt	PROPN
ajst-28825	20	20	-	-	PUNCT
ajst-28825	20	21	detr	detr	NOUN
ajst-28825	20	22	and	and	CCONJ
ajst-28825	20	23	improve	improve	VERB
ajst-28825	20	24	the	the	DET
ajst-28825	20	25	model	model	NOUN
ajst-28825	20	26	’s	’s	PART
ajst-28825	20	27	feature	feature	NOUN
ajst-28825	20	28	extraction	extraction	NOUN
ajst-28825	20	29	and	and	CCONJ
ajst-28825	20	30	fusion	fusion	NOUN
ajst-28825	20	31	capabilities	capability	NOUN
ajst-28825	20	32	through	through	ADP
ajst-28825	20	33	the	the	DET
ajst-28825	20	34	aifidattention	aifidattention	NOUN
ajst-28825	20	35	module	module	NOUN
ajst-28825	20	36	.	.	PUNCT
ajst-28825	21	1	we	we	PRON
ajst-28825	21	2	designed	design	VERB
ajst-28825	21	3	a	a	DET
ajst-28825	21	4	comprehensive	comprehensive	ADJ
ajst-28825	21	5	experimental	experimental	ADJ
ajst-28825	21	6	plan	plan	NOUN
ajst-28825	21	7	to	to	PART
ajst-28825	21	8	validate	validate	VERB
ajst-28825	21	9	the	the	DET
ajst-28825	21	10	effectiveness	effectiveness	NOUN
ajst-28825	21	11	of	of	ADP
ajst-28825	21	12	the	the	DET
ajst-28825	21	13	proposed	propose	VERB
ajst-28825	21	14	method	method	NOUN
ajst-28825	21	15	in	in	ADP
ajst-28825	21	16	recognizing	recognize	VERB
ajst-28825	21	17	different	different	ADJ
ajst-28825	21	18	modulation	modulation	NOUN
ajst-28825	21	19	types	type	NOUN
ajst-28825	21	20	and	and	CCONJ
ajst-28825	21	21	conducted	conduct	VERB
ajst-28825	21	22	a	a	DET
ajst-28825	21	23	detailed	detailed	ADJ
ajst-28825	21	24	comparative	comparative	ADJ
ajst-28825	21	25	analysis	analysis	NOUN
ajst-28825	21	26	with	with	ADP
ajst-28825	21	27	traditional	traditional	ADJ
ajst-28825	21	28	methods	method	NOUN
ajst-28825	21	29	.	.	PUNCT
ajst-28825	22	1	the	the	DET
ajst-28825	22	2	structure	structure	NOUN
ajst-28825	22	3	of	of	ADP
ajst-28825	22	4	this	this	DET
ajst-28825	22	5	paper	paper	NOUN
ajst-28825	22	6	is	be	AUX
ajst-28825	22	7	as	as	SCONJ
ajst-28825	22	8	follows	follow	VERB
ajst-28825	22	9	:	:	PUNCT
ajst-28825	22	10	section	section	NOUN
ajst-28825	22	11	2	2	NUM
ajst-28825	22	12	introduces	introduce	NOUN
ajst-28825	22	13	related	relate	VERB
ajst-28825	22	14	work	work	NOUN
ajst-28825	22	15	,	,	PUNCT
ajst-28825	22	16	section	section	NOUN
ajst-28825	22	17	3	3	NUM
ajst-28825	22	18	provides	provide	VERB
ajst-28825	22	19	a	a	DET
ajst-28825	22	20	detailed	detailed	ADJ
ajst-28825	22	21	description	description	NOUN
ajst-28825	22	22	of	of	ADP
ajst-28825	22	23	our	our	PRON
ajst-28825	22	24	proposed	propose	VERB
ajst-28825	22	25	method	method	NOUN
ajst-28825	22	26	,	,	PUNCT
ajst-28825	22	27	and	and	CCONJ
ajst-28825	22	28	section	section	NOUN
ajst-28825	22	29	4	4	NUM
ajst-28825	22	30	presents	present	VERB
ajst-28825	22	31	experimental	experimental	ADJ
ajst-28825	22	32	results	result	NOUN
ajst-28825	22	33	and	and	CCONJ
ajst-28825	22	34	analysis	analysis	NOUN
ajst-28825	22	35	.	.	PUNCT
ajst-28825	23	1	2	2	X
ajst-28825	23	2	.	.	X
ajst-28825	23	3	related	relate	VERB
ajst-28825	23	4	research	research	NOUN
ajst-28825	23	5	2.1	2.1	NUM
ajst-28825	23	6	.	.	PUNCT
ajst-28825	24	1	traditional	traditional	ADJ
ajst-28825	24	2	signal	signal	NOUN
ajst-28825	24	3	modulation	modulation	NOUN
ajst-28825	24	4	recognition	recognition	NOUN
ajst-28825	24	5	methods	method	NOUN
ajst-28825	24	6	signal	signal	VERB
ajst-28825	24	7	modulation	modulation	NOUN
ajst-28825	24	8	recognition	recognition	NOUN
ajst-28825	24	9	plays	play	VERB
ajst-28825	24	10	a	a	DET
ajst-28825	24	11	crucial	crucial	ADJ
ajst-28825	24	12	role	role	NOUN
ajst-28825	24	13	in	in	ADP
ajst-28825	24	14	wireless	wireless	ADJ
ajst-28825	24	15	communication	communication	NOUN
ajst-28825	24	16	.	.	PUNCT
ajst-28825	25	1	early	early	ADJ
ajst-28825	25	2	research	research	NOUN
ajst-28825	25	3	primarily	primarily	ADV
ajst-28825	25	4	relied	rely	VERB
ajst-28825	25	5	on	on	ADP
ajst-28825	25	6	traditional	traditional	ADJ
ajst-28825	25	7	handcrafted	handcraft	VERB
ajst-28825	25	8	feature	feature	NOUN
ajst-28825	25	9	extraction	extraction	NOUN
ajst-28825	25	10	and	and	CCONJ
ajst-28825	25	11	classification	classification	NOUN
ajst-28825	25	12	algorithms	algorithm	NOUN
ajst-28825	25	13	.	.	PUNCT
ajst-28825	26	1	common	common	ADJ
ajst-28825	26	2	methods	method	NOUN
ajst-28825	26	3	include	include	VERB
ajst-28825	26	4	recognition	recognition	NOUN
ajst-28825	26	5	based	base	VERB
ajst-28825	26	6	on	on	ADP
ajst-28825	26	7	statistical	statistical	ADJ
ajst-28825	26	8	features	feature	NOUN
ajst-28825	26	9	,	,	PUNCT
ajst-28825	26	10	instantaneous	instantaneous	ADJ
ajst-28825	26	11	feature	feature	NOUN
ajst-28825	26	12	analysis	analysis	NOUN
ajst-28825	26	13	,	,	PUNCT
ajst-28825	26	14	and	and	CCONJ
ajst-28825	26	15	cyclostationary	cyclostationary	ADJ
ajst-28825	26	16	feature	feature	NOUN
ajst-28825	26	17	extraction	extraction	NOUN
ajst-28825	26	18	.	.	PUNCT
ajst-28825	27	1	these	these	DET
ajst-28825	27	2	methods	method	NOUN
ajst-28825	27	3	typically	typically	ADV
ajst-28825	27	4	extract	extract	VERB
ajst-28825	27	5	frequency	frequency	NOUN
ajst-28825	27	6	-	-	PUNCT
ajst-28825	27	7	domain	domain	NOUN
ajst-28825	27	8	,	,	PUNCT
ajst-28825	27	9	time	time	NOUN
ajst-28825	27	10	-	-	PUNCT
ajst-28825	27	11	domain	domain	NOUN
ajst-28825	27	12	,	,	PUNCT
ajst-28825	27	13	or	or	CCONJ
ajst-28825	27	14	instantaneous	instantaneous	ADJ
ajst-28825	27	15	features	feature	NOUN
ajst-28825	27	16	of	of	ADP
ajst-28825	27	17	the	the	DET
ajst-28825	27	18	signal	signal	NOUN
ajst-28825	27	19	,	,	PUNCT
ajst-28825	27	20	such	such	ADJ
ajst-28825	27	21	as	as	ADP
ajst-28825	27	22	amplitude	amplitude	NOUN
ajst-28825	27	23	,	,	PUNCT
ajst-28825	27	24	phase	phase	NOUN
ajst-28825	27	25	,	,	PUNCT
ajst-28825	27	26	frequency	frequency	NOUN
ajst-28825	27	27	,	,	PUNCT
ajst-28825	27	28	and	and	CCONJ
ajst-28825	27	29	instantaneous	instantaneous	ADJ
ajst-28825	27	30	power	power	NOUN
ajst-28825	27	31	,	,	PUNCT
ajst-28825	27	32	to	to	PART
ajst-28825	27	33	distinguish	distinguish	VERB
ajst-28825	27	34	modulation	modulation	NOUN
ajst-28825	27	35	types	type	NOUN
ajst-28825	27	36	.	.	PUNCT
ajst-28825	28	1	for	for	ADP
ajst-28825	28	2	instance	instance	NOUN
ajst-28825	28	3	,	,	PUNCT
ajst-28825	28	4	methods	method	NOUN
ajst-28825	28	5	based	base	VERB
ajst-28825	28	6	on	on	ADP
ajst-28825	28	7	higher	high	ADJ
ajst-28825	28	8	-	-	PUNCT
ajst-28825	28	9	order	order	NOUN
ajst-28825	28	10	cumulants	cumulant	NOUN
ajst-28825	28	11	(	(	PUNCT
ajst-28825	28	12	hoc	hoc	X
ajst-28825	28	13	)	)	PUNCT
ajst-28825	28	14	have	have	AUX
ajst-28825	28	15	performed	perform	VERB
ajst-28825	28	16	well	well	ADV
ajst-28825	28	17	in	in	ADP
ajst-28825	28	18	modulation	modulation	NOUN
ajst-28825	28	19	recognition	recognition	NOUN
ajst-28825	28	20	,	,	PUNCT
ajst-28825	28	21	especially	especially	ADV
ajst-28825	28	22	in	in	ADP
ajst-28825	28	23	high	high	ADJ
ajst-28825	28	24	signal	signal	NOUN
ajst-28825	28	25	-	-	PUNCT
ajst-28825	28	26	to	to	ADP
ajst-28825	28	27	-	-	PUNCT
ajst-28825	28	28	noise	noise	NOUN
ajst-28825	28	29	ratio	ratio	NOUN
ajst-28825	28	30	(	(	PUNCT
ajst-28825	28	31	snr	snr	NOUN
ajst-28825	28	32	)	)	PUNCT
ajst-28825	28	33	scenarios	scenario	NOUN
ajst-28825	28	34	.	.	PUNCT
ajst-28825	29	1	however	however	ADV
ajst-28825	29	2	,	,	PUNCT
ajst-28825	29	3	these	these	DET
ajst-28825	29	4	methods	method	NOUN
ajst-28825	29	5	heavily	heavily	ADV
ajst-28825	29	6	rely	rely	VERB
ajst-28825	29	7	on	on	ADP
ajst-28825	29	8	the	the	DET
ajst-28825	29	9	precision	precision	NOUN
ajst-28825	29	10	and	and	CCONJ
ajst-28825	29	11	robustness	robustness	NOUN
ajst-28825	29	12	of	of	ADP
ajst-28825	29	13	feature	feature	NOUN
ajst-28825	29	14	extraction	extraction	NOUN
ajst-28825	29	15	,	,	PUNCT
ajst-28825	29	16	and	and	CCONJ
ajst-28825	29	17	their	their	PRON
ajst-28825	29	18	recognition	recognition	NOUN
ajst-28825	29	19	performance	performance	NOUN
ajst-28825	29	20	tends	tend	VERB
ajst-28825	29	21	to	to	PART
ajst-28825	29	22	degrade	degrade	VERB
ajst-28825	29	23	in	in	ADP
ajst-28825	29	24	complex	complex	ADJ
ajst-28825	29	25	and	and	CCONJ
ajst-28825	29	26	dynamic	dynamic	ADJ
ajst-28825	29	27	electromagnetic	electromagnetic	ADJ
ajst-28825	29	28	environments	environment	NOUN
ajst-28825	29	29	.	.	PUNCT
ajst-28825	30	1	machine	machine	NOUN
ajst-28825	30	2	learning	learning	NOUN
ajst-28825	30	3	methods	method	NOUN
ajst-28825	30	4	,	,	PUNCT
ajst-28825	30	5	such	such	ADJ
ajst-28825	30	6	as	as	ADP
ajst-28825	30	7	support	support	NOUN
ajst-28825	30	8	vector	vector	NOUN
ajst-28825	30	9	machines	machine	NOUN
ajst-28825	30	10	(	(	PUNCT
ajst-28825	30	11	svm	svm	PROPN
ajst-28825	30	12	)	)	PUNCT
ajst-28825	30	13	and	and	CCONJ
ajst-28825	30	14	decision	decision	NOUN
ajst-28825	30	15	trees	tree	NOUN
ajst-28825	30	16	,	,	PUNCT
ajst-28825	30	17	have	have	AUX
ajst-28825	30	18	also	also	ADV
ajst-28825	30	19	been	be	AUX
ajst-28825	30	20	widely	widely	ADV
ajst-28825	30	21	applied	apply	VERB
ajst-28825	30	22	to	to	ADP
ajst-28825	30	23	modulation	modulation	NOUN
ajst-28825	30	24	recognition	recognition	NOUN
ajst-28825	30	25	.	.	PUNCT
ajst-28825	31	1	these	these	DET
ajst-28825	31	2	methods	method	NOUN
ajst-28825	31	3	classify	classify	VERB
ajst-28825	31	4	signals	signal	NOUN
ajst-28825	31	5	by	by	ADP
ajst-28825	31	6	inputting	inputte	VERB
ajst-28825	31	7	handcrafted	handcraft	VERB
ajst-28825	31	8	features	feature	NOUN
ajst-28825	31	9	into	into	ADP
ajst-28825	31	10	the	the	DET
ajst-28825	31	11	classifiers	classifier	NOUN
ajst-28825	31	12	.	.	PUNCT
ajst-28825	32	1	however	however	ADV
ajst-28825	32	2	,	,	PUNCT
ajst-28825	32	3	handcrafted	handcraft	VERB
ajst-28825	32	4	feature	feature	NOUN
ajst-28825	32	5	extraction	extraction	NOUN
ajst-28825	32	6	has	have	VERB
ajst-28825	32	7	significant	significant	ADJ
ajst-28825	32	8	limitations	limitation	NOUN
ajst-28825	32	9	,	,	PUNCT
ajst-28825	32	10	especially	especially	ADV
ajst-28825	32	11	when	when	SCONJ
ajst-28825	32	12	the	the	DET
ajst-28825	32	13	signals	signal	NOUN
ajst-28825	32	14	are	be	AUX
ajst-28825	32	15	affected	affect	VERB
ajst-28825	32	16	by	by	ADP
ajst-28825	32	17	noise	noise	NOUN
ajst-28825	32	18	or	or	CCONJ
ajst-28825	32	19	multipath	multipath	NOUN
ajst-28825	32	20	effects	effect	NOUN
ajst-28825	32	21	,	,	PUNCT
ajst-28825	32	22	causing	cause	VERB
ajst-28825	32	23	the	the	DET
ajst-28825	32	24	performance	performance	NOUN
ajst-28825	32	25	of	of	ADP
ajst-28825	32	26	traditional	traditional	ADJ
ajst-28825	32	27	methods	method	NOUN
ajst-28825	32	28	to	to	PART
ajst-28825	32	29	deteriorate	deteriorate	VERB
ajst-28825	32	30	significantly	significantly	ADV
ajst-28825	32	31	.	.	PUNCT
ajst-28825	33	1	344	344	NUM
ajst-28825	33	2	figure	figure	NOUN
ajst-28825	33	3	1	1	NUM
ajst-28825	33	4	.	.	PUNCT
ajst-28825	33	5	principle	principle	NOUN
ajst-28825	33	6	of	of	ADP
ajst-28825	33	7	svm	svm	PROPN
ajst-28825	33	8	.	.	PROPN
ajst-28825	33	9	2.2	2.2	NUM
ajst-28825	33	10	.	.	PUNCT
ajst-28825	34	1	deep	deep	ADJ
ajst-28825	34	2	learning	learning	NOUN
ajst-28825	34	3	-	-	PUNCT
ajst-28825	34	4	based	base	VERB
ajst-28825	34	5	signal	signal	NOUN
ajst-28825	34	6	modulation	modulation	NOUN
ajst-28825	34	7	recognition	recognition	NOUN
ajst-28825	34	8	with	with	ADP
ajst-28825	34	9	the	the	DET
ajst-28825	34	10	rapid	rapid	ADJ
ajst-28825	34	11	development	development	NOUN
ajst-28825	34	12	of	of	ADP
ajst-28825	34	13	deep	deep	ADJ
ajst-28825	34	14	learning	learning	NOUN
ajst-28825	34	15	,	,	PUNCT
ajst-28825	34	16	researchers	researcher	NOUN
ajst-28825	34	17	have	have	AUX
ajst-28825	34	18	begun	begin	VERB
ajst-28825	34	19	exploring	explore	VERB
ajst-28825	34	20	the	the	DET
ajst-28825	34	21	use	use	NOUN
ajst-28825	34	22	of	of	ADP
ajst-28825	34	23	deep	deep	ADJ
ajst-28825	34	24	neural	neural	ADJ
ajst-28825	34	25	networks	network	NOUN
ajst-28825	34	26	to	to	PART
ajst-28825	34	27	automatically	automatically	ADV
ajst-28825	34	28	extract	extract	VERB
ajst-28825	34	29	features	feature	NOUN
ajst-28825	34	30	from	from	ADP
ajst-28825	34	31	signals	signal	NOUN
ajst-28825	34	32	,	,	PUNCT
ajst-28825	34	33	addressing	address	VERB
ajst-28825	34	34	the	the	DET
ajst-28825	34	35	shortcomings	shortcoming	NOUN
ajst-28825	34	36	of	of	ADP
ajst-28825	34	37	traditional	traditional	ADJ
ajst-28825	34	38	handcrafted	handcraft	VERB
ajst-28825	34	39	feature	feature	NOUN
ajst-28825	34	40	extraction	extraction	NOUN
ajst-28825	34	41	.	.	PUNCT
ajst-28825	35	1	convolutional	convolutional	ADJ
ajst-28825	35	2	neural	neural	ADJ
ajst-28825	35	3	networks	network	NOUN
ajst-28825	35	4	(	(	PUNCT
ajst-28825	35	5	cnns	cnns	PROPN
ajst-28825	35	6	)	)	PUNCT
ajst-28825	35	7	,	,	PUNCT
ajst-28825	35	8	known	know	VERB
ajst-28825	35	9	for	for	ADP
ajst-28825	35	10	their	their	PRON
ajst-28825	35	11	excellent	excellent	ADJ
ajst-28825	35	12	performance	performance	NOUN
ajst-28825	35	13	in	in	ADP
ajst-28825	35	14	image	image	NOUN
ajst-28825	35	15	processing	processing	NOUN
ajst-28825	35	16	,	,	PUNCT
ajst-28825	35	17	were	be	AUX
ajst-28825	35	18	introduced	introduce	VERB
ajst-28825	35	19	into	into	ADP
ajst-28825	35	20	signal	signal	ADJ
ajst-28825	35	21	modulation	modulation	NOUN
ajst-28825	35	22	recognition[2	recognition[2	PROPN
ajst-28825	35	23	]	]	PUNCT
ajst-28825	35	24	.	.	PUNCT
ajst-28825	36	1	o'shea	o'shea	PROPN
ajst-28825	36	2	et	et	PROPN
ajst-28825	36	3	al	al	PROPN
ajst-28825	36	4	.	.	PROPN
ajst-28825	37	1	first	first	PROPN
ajst-28825	37	2	proposed	propose	VERB
ajst-28825	37	3	using	use	VERB
ajst-28825	37	4	cnns	cnn	NOUN
ajst-28825	37	5	to	to	PART
ajst-28825	37	6	directly	directly	ADV
ajst-28825	37	7	learn	learn	VERB
ajst-28825	37	8	features	feature	NOUN
ajst-28825	37	9	from	from	ADP
ajst-28825	37	10	the	the	DET
ajst-28825	37	11	i	i	PROPN
ajst-28825	37	12	/	/	SYM
ajst-28825	37	13	q	q	NOUN
ajst-28825	37	14	data	datum	NOUN
ajst-28825	37	15	of	of	ADP
ajst-28825	37	16	signals	signal	NOUN
ajst-28825	37	17	and	and	CCONJ
ajst-28825	37	18	perform	perform	VERB
ajst-28825	37	19	modulation	modulation	NOUN
ajst-28825	37	20	classification	classification	NOUN
ajst-28825	37	21	.	.	PUNCT
ajst-28825	38	1	this	this	DET
ajst-28825	38	2	method	method	NOUN
ajst-28825	38	3	achieved	achieve	VERB
ajst-28825	38	4	remarkable	remarkable	ADJ
ajst-28825	38	5	results	result	NOUN
ajst-28825	38	6	on	on	ADP
ajst-28825	38	7	multiple	multiple	ADJ
ajst-28825	38	8	public	public	ADJ
ajst-28825	38	9	datasets	dataset	NOUN
ajst-28825	38	10	.	.	PUNCT
ajst-28825	39	1	subsequently	subsequently	ADV
ajst-28825	39	2	,	,	PUNCT
ajst-28825	39	3	recurrent	recurrent	ADJ
ajst-28825	39	4	neural	neural	ADJ
ajst-28825	39	5	networks	network	NOUN
ajst-28825	39	6	(	(	PUNCT
ajst-28825	39	7	rnns	rnns	PROPN
ajst-28825	39	8	)	)	PUNCT
ajst-28825	39	9	and	and	CCONJ
ajst-28825	39	10	their	their	PRON
ajst-28825	39	11	variants	variant	NOUN
ajst-28825	39	12	,	,	PUNCT
ajst-28825	39	13	such	such	ADJ
ajst-28825	39	14	as	as	ADP
ajst-28825	39	15	long	long	ADJ
ajst-28825	39	16	short	short	ADJ
ajst-28825	39	17	-	-	PUNCT
ajst-28825	39	18	term	term	NOUN
ajst-28825	39	19	memory	memory	NOUN
ajst-28825	39	20	networks	network	NOUN
ajst-28825	39	21	(	(	PUNCT
ajst-28825	39	22	lstms	lstms	ADJ
ajst-28825	39	23	)	)	PUNCT
ajst-28825	39	24	,	,	PUNCT
ajst-28825	39	25	were	be	AUX
ajst-28825	39	26	introduced	introduce	VERB
ajst-28825	39	27	to	to	PART
ajst-28825	39	28	capture	capture	VERB
ajst-28825	39	29	temporal	temporal	ADJ
ajst-28825	39	30	information	information	NOUN
ajst-28825	39	31	in	in	ADP
ajst-28825	39	32	signals[9	signals[9	NOUN
ajst-28825	39	33	]	]	PUNCT
ajst-28825	39	34	.	.	PUNCT
ajst-28825	40	1	by	by	ADP
ajst-28825	40	2	modeling	model	VERB
ajst-28825	40	3	the	the	DET
ajst-28825	40	4	temporal	temporal	ADJ
ajst-28825	40	5	dependencies	dependency	NOUN
ajst-28825	40	6	of	of	ADP
ajst-28825	40	7	the	the	DET
ajst-28825	40	8	signals	signal	NOUN
ajst-28825	40	9	,	,	PUNCT
ajst-28825	40	10	these	these	DET
ajst-28825	40	11	methods	method	NOUN
ajst-28825	40	12	have	have	AUX
ajst-28825	40	13	demonstrated	demonstrate	VERB
ajst-28825	40	14	strong	strong	ADJ
ajst-28825	40	15	robustness	robustness	NOUN
ajst-28825	40	16	in	in	ADP
ajst-28825	40	17	low	low	ADJ
ajst-28825	40	18	snr	snr	NOUN
ajst-28825	40	19	environments	environment	NOUN
ajst-28825	40	20	.	.	PUNCT
ajst-28825	41	1	however	however	ADV
ajst-28825	41	2	,	,	PUNCT
ajst-28825	41	3	rnn	rnn	NOUN
ajst-28825	41	4	-	-	PUNCT
ajst-28825	41	5	based	base	VERB
ajst-28825	41	6	models	model	NOUN
ajst-28825	41	7	are	be	AUX
ajst-28825	41	8	time	time	NOUN
ajst-28825	41	9	-	-	PUNCT
ajst-28825	41	10	consuming	consume	VERB
ajst-28825	41	11	to	to	PART
ajst-28825	41	12	train	train	VERB
ajst-28825	41	13	and	and	CCONJ
ajst-28825	41	14	are	be	AUX
ajst-28825	41	15	prone	prone	ADJ
ajst-28825	41	16	to	to	ADP
ajst-28825	41	17	gradient	gradient	VERB
ajst-28825	41	18	vanishing	vanishing	NOUN
ajst-28825	41	19	or	or	CCONJ
ajst-28825	41	20	exploding	explode	VERB
ajst-28825	41	21	issues	issue	NOUN
ajst-28825	41	22	.	.	PUNCT
ajst-28825	42	1	in	in	ADP
ajst-28825	42	2	recent	recent	ADJ
ajst-28825	42	3	years	year	NOUN
ajst-28825	42	4	,	,	PUNCT
ajst-28825	42	5	hybrid	hybrid	ADJ
ajst-28825	42	6	architectures	architecture	NOUN
ajst-28825	42	7	combining	combine	VERB
ajst-28825	42	8	cnns	cnn	NOUN
ajst-28825	42	9	and	and	CCONJ
ajst-28825	42	10	rnns	rnn	NOUN
ajst-28825	42	11	,	,	PUNCT
ajst-28825	42	12	such	such	ADJ
ajst-28825	42	13	as	as	ADP
ajst-28825	42	14	cnn	cnn	PROPN
ajst-28825	42	15	-	-	PUNCT
ajst-28825	42	16	rnn	rnn	PROPN
ajst-28825	42	17	and	and	CCONJ
ajst-28825	42	18	cnn	cnn	PROPN
ajst-28825	42	19	-	-	PUNCT
ajst-28825	42	20	lstm	lstm	ADJ
ajst-28825	42	21	models	model	NOUN
ajst-28825	42	22	,	,	PUNCT
ajst-28825	42	23	have	have	AUX
ajst-28825	42	24	gained	gain	VERB
ajst-28825	42	25	attention	attention	NOUN
ajst-28825	42	26	.	.	PUNCT
ajst-28825	43	1	these	these	DET
ajst-28825	43	2	models	model	NOUN
ajst-28825	43	3	improve	improve	VERB
ajst-28825	43	4	modulation	modulation	NOUN
ajst-28825	43	5	recognition	recognition	NOUN
ajst-28825	43	6	accuracy	accuracy	NOUN
ajst-28825	43	7	by	by	ADP
ajst-28825	43	8	fusing	fuse	VERB
ajst-28825	43	9	spatial	spatial	ADJ
ajst-28825	43	10	and	and	CCONJ
ajst-28825	43	11	temporal	temporal	ADJ
ajst-28825	43	12	features	feature	NOUN
ajst-28825	43	13	.	.	PUNCT
ajst-28825	44	1	although	although	SCONJ
ajst-28825	44	2	these	these	DET
ajst-28825	44	3	deep	deep	ADJ
ajst-28825	44	4	learning	learning	NOUN
ajst-28825	44	5	methods	method	NOUN
ajst-28825	44	6	outperform	outperform	VERB
ajst-28825	44	7	traditional	traditional	ADJ
ajst-28825	44	8	methods	method	NOUN
ajst-28825	44	9	,	,	PUNCT
ajst-28825	44	10	they	they	PRON
ajst-28825	44	11	often	often	ADV
ajst-28825	44	12	require	require	VERB
ajst-28825	44	13	substantial	substantial	ADJ
ajst-28825	44	14	computational	computational	ADJ
ajst-28825	44	15	resources	resource	NOUN
ajst-28825	44	16	and	and	CCONJ
ajst-28825	44	17	training	training	NOUN
ajst-28825	44	18	time	time	NOUN
ajst-28825	44	19	,	,	PUNCT
ajst-28825	44	20	posing	pose	VERB
ajst-28825	44	21	challenges	challenge	NOUN
ajst-28825	44	22	for	for	ADP
ajst-28825	44	23	real	real	ADJ
ajst-28825	44	24	-	-	PUNCT
ajst-28825	44	25	time	time	NOUN
ajst-28825	44	26	applications	application	NOUN
ajst-28825	44	27	.	.	PUNCT
ajst-28825	45	1	2.3	2.3	NUM
ajst-28825	45	2	.	.	PUNCT
ajst-28825	46	1	lightweight	lightweight	ADJ
ajst-28825	46	2	networks	network	NOUN
ajst-28825	46	3	and	and	CCONJ
ajst-28825	46	4	transformer	transformer	NOUN
ajst-28825	46	5	applications	application	NOUN
ajst-28825	46	6	to	to	PART
ajst-28825	46	7	address	address	VERB
ajst-28825	46	8	the	the	DET
ajst-28825	46	9	high	high	ADJ
ajst-28825	46	10	computational	computational	ADJ
ajst-28825	46	11	complexity	complexity	NOUN
ajst-28825	46	12	of	of	ADP
ajst-28825	46	13	deep	deep	ADJ
ajst-28825	46	14	learning	learning	NOUN
ajst-28825	46	15	models	model	NOUN
ajst-28825	46	16	,	,	PUNCT
ajst-28825	46	17	lightweight	lightweight	ADJ
ajst-28825	46	18	network	network	NOUN
ajst-28825	46	19	structures	structure	NOUN
ajst-28825	46	20	have	have	AUX
ajst-28825	46	21	been	be	AUX
ajst-28825	46	22	widely	widely	ADV
ajst-28825	46	23	researched	research	VERB
ajst-28825	46	24	.	.	PUNCT
ajst-28825	47	1	lightweight	lightweight	PROPN
ajst-28825	47	2	cnn	cnn	PROPN
ajst-28825	47	3	models	model	NOUN
ajst-28825	47	4	,	,	PUNCT
ajst-28825	47	5	such	such	ADJ
ajst-28825	47	6	as	as	ADP
ajst-28825	47	7	mobilenet	mobilenet	NOUN
ajst-28825	47	8	and	and	CCONJ
ajst-28825	47	9	squeezenet	squeezenet	NOUN
ajst-28825	47	10	,	,	PUNCT
ajst-28825	47	11	reduce	reduce	VERB
ajst-28825	47	12	the	the	DET
ajst-28825	47	13	number	number	NOUN
ajst-28825	47	14	of	of	ADP
ajst-28825	47	15	parameters	parameter	NOUN
ajst-28825	47	16	and	and	CCONJ
ajst-28825	47	17	computation	computation	NOUN
ajst-28825	47	18	,	,	PUNCT
ajst-28825	47	19	enabling	enable	VERB
ajst-28825	47	20	efficient	efficient	ADJ
ajst-28825	47	21	performance	performance	NOUN
ajst-28825	47	22	on	on	ADP
ajst-28825	47	23	resourceconstrained	resourceconstraine	VERB
ajst-28825	47	24	platforms	platform	NOUN
ajst-28825	47	25	like	like	ADP
ajst-28825	47	26	mobile	mobile	ADJ
ajst-28825	47	27	devices	device	NOUN
ajst-28825	47	28	.	.	PUNCT
ajst-28825	48	1	while	while	SCONJ
ajst-28825	48	2	these	these	DET
ajst-28825	48	3	models	model	NOUN
ajst-28825	48	4	perform	perform	VERB
ajst-28825	48	5	well	well	ADV
ajst-28825	48	6	in	in	ADP
ajst-28825	48	7	specific	specific	ADJ
ajst-28825	48	8	application	application	NOUN
ajst-28825	48	9	scenarios	scenario	NOUN
ajst-28825	48	10	,	,	PUNCT
ajst-28825	48	11	they	they	PRON
ajst-28825	48	12	still	still	ADV
ajst-28825	48	13	struggle	struggle	VERB
ajst-28825	48	14	with	with	ADP
ajst-28825	48	15	accuracy	accuracy	NOUN
ajst-28825	48	16	when	when	SCONJ
ajst-28825	48	17	handling	handle	VERB
ajst-28825	48	18	complex	complex	ADJ
ajst-28825	48	19	signal	signal	ADJ
ajst-28825	48	20	modulation	modulation	NOUN
ajst-28825	48	21	recognition	recognition	NOUN
ajst-28825	48	22	tasks	task	NOUN
ajst-28825	48	23	.	.	PUNCT
ajst-28825	49	1	the	the	DET
ajst-28825	49	2	transformer	transformer	NOUN
ajst-28825	49	3	structure	structure	NOUN
ajst-28825	49	4	,	,	PUNCT
ajst-28825	49	5	known	know	VERB
ajst-28825	49	6	for	for	ADP
ajst-28825	49	7	its	its	PRON
ajst-28825	49	8	success	success	NOUN
ajst-28825	49	9	in	in	ADP
ajst-28825	49	10	natural	natural	ADJ
ajst-28825	49	11	language	language	NOUN
ajst-28825	49	12	processing	processing	NOUN
ajst-28825	49	13	,	,	PUNCT
ajst-28825	49	14	has	have	AUX
ajst-28825	49	15	garnered	garner	VERB
ajst-28825	49	16	attention	attention	NOUN
ajst-28825	49	17	and	and	CCONJ
ajst-28825	49	18	has	have	AUX
ajst-28825	49	19	been	be	AUX
ajst-28825	49	20	gradually	gradually	ADV
ajst-28825	49	21	introduced	introduce	VERB
ajst-28825	49	22	into	into	ADP
ajst-28825	49	23	visual	visual	ADJ
ajst-28825	49	24	tasks	task	NOUN
ajst-28825	49	25	.	.	PUNCT
ajst-28825	50	1	its	its	PRON
ajst-28825	50	2	self	self	NOUN
ajst-28825	50	3	-	-	PUNCT
ajst-28825	50	4	attention	attention	NOUN
ajst-28825	50	5	mechanism	mechanism	NOUN
ajst-28825	50	6	can	can	AUX
ajst-28825	50	7	capture	capture	VERB
ajst-28825	50	8	global	global	ADJ
ajst-28825	50	9	features	feature	NOUN
ajst-28825	50	10	without	without	ADP
ajst-28825	50	11	relying	rely	VERB
ajst-28825	50	12	on	on	ADP
ajst-28825	50	13	temporal	temporal	ADJ
ajst-28825	50	14	sequences	sequence	NOUN
ajst-28825	50	15	,	,	PUNCT
ajst-28825	50	16	providing	provide	VERB
ajst-28825	50	17	new	new	ADJ
ajst-28825	50	18	approaches	approach	NOUN
ajst-28825	50	19	to	to	PART
ajst-28825	50	20	feature	feature	NOUN
ajst-28825	50	21	extraction	extraction	NOUN
ajst-28825	50	22	.	.	PUNCT
ajst-28825	51	1	rt	rt	PROPN
ajst-28825	51	2	-	-	PUNCT
ajst-28825	51	3	detr	detr	NOUN
ajst-28825	51	4	,	,	PUNCT
ajst-28825	51	5	a	a	DET
ajst-28825	51	6	lightweight	lightweight	ADJ
ajst-28825	51	7	object	object	NOUN
ajst-28825	51	8	detection	detection	NOUN
ajst-28825	51	9	model	model	NOUN
ajst-28825	51	10	that	that	PRON
ajst-28825	51	11	integrates	integrate	VERB
ajst-28825	51	12	the	the	DET
ajst-28825	51	13	advantages	advantage	NOUN
ajst-28825	51	14	of	of	ADP
ajst-28825	51	15	the	the	DET
ajst-28825	51	16	transformer	transformer	NOUN
ajst-28825	51	17	,	,	PUNCT
ajst-28825	51	18	can	can	AUX
ajst-28825	51	19	extract	extract	VERB
ajst-28825	51	20	rich	rich	ADJ
ajst-28825	51	21	feature	feature	NOUN
ajst-28825	51	22	information	information	NOUN
ajst-28825	51	23	while	while	SCONJ
ajst-28825	51	24	maintaining	maintain	VERB
ajst-28825	51	25	computational	computational	ADJ
ajst-28825	51	26	efficiency[3	efficiency[3	NOUN
ajst-28825	51	27	]	]	PUNCT
ajst-28825	51	28	,	,	PUNCT
ajst-28825	51	29	making	make	VERB
ajst-28825	51	30	it	it	PRON
ajst-28825	51	31	highly	highly	ADV
ajst-28825	51	32	promising	promising	ADJ
ajst-28825	51	33	for	for	ADP
ajst-28825	51	34	signal	signal	ADJ
ajst-28825	51	35	modulation	modulation	NOUN
ajst-28825	51	36	recognition	recognition	NOUN
ajst-28825	51	37	tasks	task	NOUN
ajst-28825	51	38	.	.	PUNCT
ajst-28825	52	1	2.4	2.4	NUM
ajst-28825	52	2	.	.	PUNCT
ajst-28825	52	3	model	model	NOUN
ajst-28825	52	4	enhancement	enhancement	NOUN
ajst-28825	52	5	and	and	CCONJ
ajst-28825	52	6	feature	feature	NOUN
ajst-28825	52	7	fusion	fusion	NOUN
ajst-28825	52	8	techniques	technique	NOUN
ajst-28825	52	9	to	to	PART
ajst-28825	52	10	further	far	ADV
ajst-28825	52	11	enhance	enhance	VERB
ajst-28825	52	12	the	the	DET
ajst-28825	52	13	performance	performance	NOUN
ajst-28825	52	14	of	of	ADP
ajst-28825	52	15	deep	deep	ADJ
ajst-28825	52	16	learning	learning	NOUN
ajst-28825	52	17	models	model	NOUN
ajst-28825	52	18	in	in	ADP
ajst-28825	52	19	signal	signal	ADJ
ajst-28825	52	20	modulation	modulation	NOUN
ajst-28825	52	21	recognition	recognition	NOUN
ajst-28825	52	22	,	,	PUNCT
ajst-28825	52	23	researchers	researcher	NOUN
ajst-28825	52	24	have	have	AUX
ajst-28825	52	25	proposed	propose	VERB
ajst-28825	52	26	various	various	ADJ
ajst-28825	52	27	enhancement	enhancement	NOUN
ajst-28825	52	28	and	and	CCONJ
ajst-28825	52	29	optimization	optimization	NOUN
ajst-28825	52	30	strategies	strategy	NOUN
ajst-28825	52	31	.	.	PUNCT
ajst-28825	53	1	attention	attention	NOUN
ajst-28825	53	2	mechanisms	mechanism	NOUN
ajst-28825	53	3	have	have	AUX
ajst-28825	53	4	been	be	AUX
ajst-28825	53	5	widely	widely	ADV
ajst-28825	53	6	applied	apply	VERB
ajst-28825	53	7	in	in	ADP
ajst-28825	53	8	deep	deep	ADJ
ajst-28825	53	9	learning	learning	NOUN
ajst-28825	53	10	,	,	PUNCT
ajst-28825	53	11	particularly	particularly	ADV
ajst-28825	53	12	when	when	SCONJ
ajst-28825	53	13	dealing	deal	VERB
ajst-28825	53	14	with	with	ADP
ajst-28825	53	15	signals	signal	NOUN
ajst-28825	53	16	with	with	ADP
ajst-28825	53	17	complex	complex	ADJ
ajst-28825	53	18	patterns	pattern	NOUN
ajst-28825	53	19	,	,	PUNCT
ajst-28825	53	20	as	as	SCONJ
ajst-28825	53	21	they	they	PRON
ajst-28825	53	22	effectively	effectively	ADV
ajst-28825	53	23	enhance	enhance	VERB
ajst-28825	53	24	the	the	DET
ajst-28825	53	25	expression	expression	NOUN
ajst-28825	53	26	of	of	ADP
ajst-28825	53	27	important	important	ADJ
ajst-28825	53	28	features	feature	NOUN
ajst-28825	53	29	.	.	PUNCT
ajst-28825	54	1	aifi	aifi	NOUN
ajst-28825	54	2	-	-	PUNCT
ajst-28825	54	3	dattention	dattention	NOUN
ajst-28825	54	4	,	,	PUNCT
ajst-28825	54	5	a	a	DET
ajst-28825	54	6	dynamic	dynamic	ADJ
ajst-28825	54	7	attention	attention	NOUN
ajst-28825	54	8	mechanism	mechanism	NOUN
ajst-28825	54	9	,	,	PUNCT
ajst-28825	54	10	can	can	AUX
ajst-28825	54	11	adaptively	adaptively	ADV
ajst-28825	54	12	adjust	adjust	VERB
ajst-28825	54	13	attention	attention	NOUN
ajst-28825	54	14	weights	weight	NOUN
ajst-28825	54	15	according	accord	VERB
ajst-28825	54	16	to	to	ADP
ajst-28825	54	17	the	the	DET
ajst-28825	54	18	characteristics	characteristic	NOUN
ajst-28825	54	19	of	of	ADP
ajst-28825	54	20	the	the	DET
ajst-28825	54	21	input	input	NOUN
ajst-28825	54	22	signal	signal	NOUN
ajst-28825	54	23	,	,	PUNCT
ajst-28825	54	24	improving	improve	VERB
ajst-28825	54	25	the	the	DET
ajst-28825	54	26	model	model	NOUN
ajst-28825	54	27	's	's	PART
ajst-28825	54	28	feature	feature	NOUN
ajst-28825	54	29	extraction	extraction	NOUN
ajst-28825	54	30	capabilities[4	capabilities[4	NOUN
ajst-28825	54	31	]	]	PUNCT
ajst-28825	54	32	.	.	PUNCT
ajst-28825	55	1	moreover	moreover	ADV
ajst-28825	55	2	,	,	PUNCT
ajst-28825	55	3	feature	feature	NOUN
ajst-28825	55	4	fusion	fusion	NOUN
ajst-28825	55	5	techniques	technique	NOUN
ajst-28825	55	6	have	have	AUX
ajst-28825	55	7	been	be	AUX
ajst-28825	55	8	applied	apply	VERB
ajst-28825	55	9	in	in	ADP
ajst-28825	55	10	multimodal	multimodal	NOUN
ajst-28825	55	11	signal	signal	NOUN
ajst-28825	55	12	processing	processing	NOUN
ajst-28825	55	13	and	and	CCONJ
ajst-28825	55	14	complex	complex	ADJ
ajst-28825	55	15	scene	scene	NOUN
ajst-28825	55	16	recognition	recognition	NOUN
ajst-28825	55	17	.	.	PUNCT
ajst-28825	56	1	the	the	DET
ajst-28825	56	2	aifi	aifi	NOUN
ajst-28825	56	3	-	-	PUNCT
ajst-28825	56	4	efficientadditive	efficientadditive	ADJ
ajst-28825	56	5	module	module	NOUN
ajst-28825	56	6	improves	improve	VERB
ajst-28825	56	7	traditional	traditional	ADJ
ajst-28825	56	8	additive	additive	ADJ
ajst-28825	56	9	feature	feature	NOUN
ajst-28825	56	10	fusion	fusion	NOUN
ajst-28825	56	11	methods	method	NOUN
ajst-28825	56	12	,	,	PUNCT
ajst-28825	56	13	enhancing	enhance	VERB
ajst-28825	56	14	the	the	DET
ajst-28825	56	15	synergy	synergy	NOUN
ajst-28825	56	16	between	between	ADP
ajst-28825	56	17	features	feature	NOUN
ajst-28825	56	18	while	while	SCONJ
ajst-28825	56	19	maintaining	maintain	VERB
ajst-28825	56	20	computational	computational	ADJ
ajst-28825	56	21	efficiency	efficiency	NOUN
ajst-28825	56	22	,	,	PUNCT
ajst-28825	56	23	thereby	thereby	ADV
ajst-28825	56	24	improving	improve	VERB
ajst-28825	56	25	modulation	modulation	NOUN
ajst-28825	56	26	recognition	recognition	NOUN
ajst-28825	56	27	accuracy	accuracy	NOUN
ajst-28825	56	28	.	.	PUNCT
ajst-28825	57	1	2.5	2.5	NUM
ajst-28825	57	2	.	.	PUNCT
ajst-28825	58	1	contributions	contribution	NOUN
ajst-28825	58	2	of	of	ADP
ajst-28825	58	3	this	this	DET
ajst-28825	58	4	research	research	NOUN
ajst-28825	58	5	building	building	NOUN
ajst-28825	58	6	upon	upon	SCONJ
ajst-28825	58	7	the	the	DET
ajst-28825	58	8	aforementioned	aforementioned	ADJ
ajst-28825	58	9	research	research	NOUN
ajst-28825	58	10	,	,	PUNCT
ajst-28825	58	11	this	this	DET
ajst-28825	58	12	study	study	NOUN
ajst-28825	58	13	proposes	propose	VERB
ajst-28825	58	14	a	a	DET
ajst-28825	58	15	signal	signal	ADJ
ajst-28825	58	16	modulation	modulation	NOUN
ajst-28825	58	17	recognition	recognition	NOUN
ajst-28825	58	18	method	method	NOUN
ajst-28825	58	19	based	base	VERB
ajst-28825	58	20	on	on	ADP
ajst-28825	58	21	the	the	DET
ajst-28825	58	22	rt	rt	PROPN
ajst-28825	58	23	-	-	PUNCT
ajst-28825	58	24	detr	detr	NOUN
ajst-28825	58	25	model	model	NOUN
ajst-28825	58	26	.	.	PUNCT
ajst-28825	59	1	by	by	ADP
ajst-28825	59	2	combining	combine	VERB
ajst-28825	59	3	resnet-18	resnet-18	PROPN
ajst-28825	59	4	(	(	PUNCT
ajst-28825	59	5	r18	r18	NOUN
ajst-28825	59	6	)	)	PUNCT
ajst-28825	59	7	as	as	ADP
ajst-28825	59	8	the	the	DET
ajst-28825	59	9	backbone	backbone	NOUN
ajst-28825	59	10	and	and	CCONJ
ajst-28825	59	11	introducing	introduce	VERB
ajst-28825	59	12	the	the	DET
ajst-28825	59	13	aifi	aifi	NOUN
ajst-28825	59	14	-	-	PUNCT
ajst-28825	59	15	dattention	dattention	NOUN
ajst-28825	59	16	module	module	NOUN
ajst-28825	59	17	,	,	PUNCT
ajst-28825	59	18	our	our	PRON
ajst-28825	59	19	model	model	NOUN
ajst-28825	59	20	significantly	significantly	ADV
ajst-28825	59	21	improves	improve	VERB
ajst-28825	59	22	the	the	DET
ajst-28825	59	23	accuracy	accuracy	NOUN
ajst-28825	59	24	and	and	CCONJ
ajst-28825	59	25	efficiency	efficiency	NOUN
ajst-28825	59	26	of	of	ADP
ajst-28825	59	27	signal	signal	ADJ
ajst-28825	59	28	modulation	modulation	NOUN
ajst-28825	59	29	recognition	recognition	NOUN
ajst-28825	59	30	while	while	SCONJ
ajst-28825	59	31	maintaining	maintain	VERB
ajst-28825	59	32	a	a	DET
ajst-28825	59	33	lightweight	lightweight	ADJ
ajst-28825	59	34	structure	structure	NOUN
ajst-28825	59	35	.	.	PUNCT
ajst-28825	60	1	we	we	PRON
ajst-28825	60	2	conducted	conduct	VERB
ajst-28825	60	3	comprehensive	comprehensive	ADJ
ajst-28825	60	4	testing	testing	NOUN
ajst-28825	60	5	of	of	ADP
ajst-28825	60	6	the	the	DET
ajst-28825	60	7	model	model	NOUN
ajst-28825	60	8	in	in	ADP
ajst-28825	60	9	various	various	ADJ
ajst-28825	60	10	complex	complex	ADJ
ajst-28825	60	11	channel	channel	NOUN
ajst-28825	60	12	environments	environment	NOUN
ajst-28825	60	13	,	,	PUNCT
ajst-28825	60	14	and	and	CCONJ
ajst-28825	60	15	the	the	DET
ajst-28825	60	16	experimental	experimental	ADJ
ajst-28825	60	17	results	result	NOUN
ajst-28825	60	18	indicate	indicate	VERB
ajst-28825	60	19	that	that	SCONJ
ajst-28825	60	20	our	our	PRON
ajst-28825	60	21	method	method	NOUN
ajst-28825	60	22	outperforms	outperform	VERB
ajst-28825	60	23	traditional	traditional	ADJ
ajst-28825	60	24	methods	method	NOUN
ajst-28825	60	25	in	in	ADP
ajst-28825	60	26	terms	term	NOUN
ajst-28825	60	27	of	of	ADP
ajst-28825	60	28	recognition	recognition	NOUN
ajst-28825	60	29	accuracy	accuracy	NOUN
ajst-28825	60	30	and	and	CCONJ
ajst-28825	60	31	computational	computational	ADJ
ajst-28825	60	32	efficiency	efficiency	NOUN
ajst-28825	60	33	.	.	PUNCT
ajst-28825	61	1	figure	figure	NOUN
ajst-28825	61	2	2	2	NUM
ajst-28825	61	3	.	.	PUNCT
ajst-28825	61	4	overview	overview	NOUN
ajst-28825	61	5	of	of	ADP
ajst-28825	61	6	rt	rt	NOUN
ajst-28825	61	7	-	-	PUNCT
ajst-28825	61	8	detra	detra	PROPN
ajst-28825	61	9	345	345	NUM
ajst-28825	61	10	3	3	NUM
ajst-28825	61	11	.	.	PUNCT
ajst-28825	61	12	model	model	NOUN
ajst-28825	61	13	structure	structure	NOUN
ajst-28825	61	14	and	and	CCONJ
ajst-28825	61	15	dataset	dataset	VERB
ajst-28825	61	16	3.1	3.1	NUM
ajst-28825	61	17	.	.	PUNCT
ajst-28825	61	18	model	model	NOUN
ajst-28825	61	19	overview	overview	NOUN
ajst-28825	61	20	in	in	ADP
ajst-28825	61	21	this	this	DET
ajst-28825	61	22	study	study	NOUN
ajst-28825	61	23	,	,	PUNCT
ajst-28825	61	24	we	we	PRON
ajst-28825	61	25	employ	employ	VERB
ajst-28825	61	26	the	the	DET
ajst-28825	61	27	rt	rt	PROPN
ajst-28825	61	28	-	-	PUNCT
ajst-28825	61	29	detr	detr	NOUN
ajst-28825	61	30	model	model	NOUN
ajst-28825	61	31	as	as	ADP
ajst-28825	61	32	the	the	DET
ajst-28825	61	33	foundation	foundation	NOUN
ajst-28825	61	34	for	for	ADP
ajst-28825	61	35	signal	signal	ADJ
ajst-28825	61	36	modulation	modulation	NOUN
ajst-28825	61	37	recognition	recognition	NOUN
ajst-28825	61	38	.	.	PUNCT
ajst-28825	62	1	rt	rt	PROPN
ajst-28825	62	2	-	-	PUNCT
ajst-28825	62	3	detr	detr	NOUN
ajst-28825	62	4	is	be	AUX
ajst-28825	62	5	a	a	DET
ajst-28825	62	6	lightweight	lightweight	ADJ
ajst-28825	62	7	object	object	NOUN
ajst-28825	62	8	detection	detection	NOUN
ajst-28825	62	9	model	model	NOUN
ajst-28825	62	10	,	,	PUNCT
ajst-28825	62	11	and	and	CCONJ
ajst-28825	62	12	its	its	PRON
ajst-28825	62	13	transformerbased	transformerbase	VERB
ajst-28825	62	14	architecture	architecture	NOUN
ajst-28825	62	15	effectively	effectively	ADV
ajst-28825	62	16	captures	capture	VERB
ajst-28825	62	17	the	the	DET
ajst-28825	62	18	global	global	ADJ
ajst-28825	62	19	features	feature	NOUN
ajst-28825	62	20	of	of	ADP
ajst-28825	62	21	input	input	NOUN
ajst-28825	62	22	signals	signal	NOUN
ajst-28825	62	23	,	,	PUNCT
ajst-28825	62	24	making	make	VERB
ajst-28825	62	25	it	it	PRON
ajst-28825	62	26	well	well	ADV
ajst-28825	62	27	-	-	PUNCT
ajst-28825	62	28	suited	suit	VERB
ajst-28825	62	29	for	for	ADP
ajst-28825	62	30	the	the	DET
ajst-28825	62	31	task	task	NOUN
ajst-28825	62	32	of	of	ADP
ajst-28825	62	33	complex	complex	ADJ
ajst-28825	62	34	signal	signal	ADJ
ajst-28825	62	35	modulation	modulation	NOUN
ajst-28825	62	36	recognition	recognition	NOUN
ajst-28825	62	37	.	.	PUNCT
ajst-28825	63	1	to	to	PART
ajst-28825	63	2	further	far	ADV
ajst-28825	63	3	enhance	enhance	VERB
ajst-28825	63	4	model	model	NOUN
ajst-28825	63	5	performance	performance	NOUN
ajst-28825	63	6	,	,	PUNCT
ajst-28825	63	7	we	we	PRON
ajst-28825	63	8	made	make	VERB
ajst-28825	63	9	the	the	DET
ajst-28825	63	10	following	follow	VERB
ajst-28825	63	11	improvements	improvement	NOUN
ajst-28825	63	12	based	base	VERB
ajst-28825	63	13	on	on	ADP
ajst-28825	63	14	rt	rt	PROPN
ajst-28825	63	15	-	-	PUNCT
ajst-28825	63	16	detr	detr	NOUN
ajst-28825	63	17	:	:	PUNCT
ajst-28825	63	18	backbone	backbone	NOUN
ajst-28825	63	19	selection	selection	NOUN
ajst-28825	63	20	:	:	PUNCT
ajst-28825	63	21	we	we	PRON
ajst-28825	63	22	chose	choose	VERB
ajst-28825	63	23	resnet-18	resnet-18	PROPN
ajst-28825	63	24	(	(	PUNCT
ajst-28825	63	25	r18	r18	NOUN
ajst-28825	63	26	)	)	PUNCT
ajst-28825	63	27	as	as	ADP
ajst-28825	63	28	the	the	DET
ajst-28825	63	29	backbone	backbone	NOUN
ajst-28825	63	30	of	of	ADP
ajst-28825	63	31	the	the	DET
ajst-28825	63	32	model	model	NOUN
ajst-28825	63	33	.	.	PUNCT
ajst-28825	64	1	resnet-18	resnet-18	PROPN
ajst-28825	64	2	is	be	AUX
ajst-28825	64	3	a	a	DET
ajst-28825	64	4	lightweight	lightweight	ADJ
ajst-28825	64	5	convolutional	convolutional	ADJ
ajst-28825	64	6	neural	neural	ADJ
ajst-28825	64	7	network	network	NOUN
ajst-28825	64	8	architecture	architecture	NOUN
ajst-28825	64	9	that	that	PRON
ajst-28825	64	10	uses	use	VERB
ajst-28825	64	11	residual	residual	ADJ
ajst-28825	64	12	connections	connection	NOUN
ajst-28825	64	13	to	to	PART
ajst-28825	64	14	mitigate	mitigate	VERB
ajst-28825	64	15	the	the	DET
ajst-28825	64	16	vanishing	vanish	VERB
ajst-28825	64	17	gradient	gradient	NOUN
ajst-28825	64	18	problem	problem	NOUN
ajst-28825	64	19	.	.	PUNCT
ajst-28825	65	1	this	this	PRON
ajst-28825	65	2	enables	enable	VERB
ajst-28825	65	3	effective	effective	ADJ
ajst-28825	65	4	extraction	extraction	NOUN
ajst-28825	65	5	of	of	ADP
ajst-28825	65	6	deep	deep	ADJ
ajst-28825	65	7	features	feature	NOUN
ajst-28825	65	8	while	while	SCONJ
ajst-28825	65	9	keeping	keep	VERB
ajst-28825	65	10	the	the	DET
ajst-28825	65	11	model	model	NOUN
ajst-28825	65	12	lightweight	lightweight	NOUN
ajst-28825	65	13	.	.	PUNCT
ajst-28825	66	1	dynamic	dynamic	ADJ
ajst-28825	66	2	attention	attention	NOUN
ajst-28825	66	3	mechanism	mechanism	NOUN
ajst-28825	66	4	(	(	PUNCT
ajst-28825	66	5	aifi	aifi	NOUN
ajst-28825	66	6	-	-	PUNCT
ajst-28825	66	7	dattention	dattention	NOUN
ajst-28825	66	8	):	):	PUNCT
ajst-28825	66	9	to	to	PART
ajst-28825	66	10	improve	improve	VERB
ajst-28825	66	11	the	the	DET
ajst-28825	66	12	model	model	NOUN
ajst-28825	66	13	's	's	PART
ajst-28825	66	14	sensitivity	sensitivity	NOUN
ajst-28825	66	15	to	to	ADP
ajst-28825	66	16	complex	complex	ADJ
ajst-28825	66	17	signal	signal	NOUN
ajst-28825	66	18	features	feature	NOUN
ajst-28825	66	19	,	,	PUNCT
ajst-28825	66	20	we	we	PRON
ajst-28825	66	21	integrated	integrate	VERB
ajst-28825	66	22	the	the	DET
ajst-28825	66	23	aifi	aifi	NOUN
ajst-28825	66	24	-	-	PUNCT
ajst-28825	66	25	dattention	dattention	NOUN
ajst-28825	66	26	module	module	NOUN
ajst-28825	66	27	.	.	PUNCT
ajst-28825	67	1	this	this	DET
ajst-28825	67	2	module	module	NOUN
ajst-28825	67	3	,	,	PUNCT
ajst-28825	67	4	based	base	VERB
ajst-28825	67	5	on	on	ADP
ajst-28825	67	6	dynamic	dynamic	ADJ
ajst-28825	67	7	attention	attention	NOUN
ajst-28825	67	8	mechanisms	mechanism	NOUN
ajst-28825	67	9	,	,	PUNCT
ajst-28825	67	10	adaptively	adaptively	ADV
ajst-28825	67	11	adjusts	adjust	VERB
ajst-28825	67	12	the	the	DET
ajst-28825	67	13	attention	attention	NOUN
ajst-28825	67	14	weights	weight	NOUN
ajst-28825	67	15	of	of	ADP
ajst-28825	67	16	different	different	ADJ
ajst-28825	67	17	feature	feature	NOUN
ajst-28825	67	18	maps	map	NOUN
ajst-28825	67	19	,	,	PUNCT
ajst-28825	67	20	thereby	thereby	ADV
ajst-28825	67	21	improving	improve	VERB
ajst-28825	67	22	feature	feature	NOUN
ajst-28825	67	23	extraction	extraction	NOUN
ajst-28825	67	24	precision	precision	NOUN
ajst-28825	67	25	and	and	CCONJ
ajst-28825	67	26	robustness	robustness	NOUN
ajst-28825	67	27	.	.	PUNCT
ajst-28825	68	1	3.2	3.2	NUM
ajst-28825	68	2	.	.	PUNCT
ajst-28825	68	3	data	datum	NOUN
ajst-28825	68	4	preprocessing	preprocesse	VERB
ajst-28825	68	5	before	before	ADP
ajst-28825	68	6	performing	perform	VERB
ajst-28825	68	7	signal	signal	NOUN
ajst-28825	68	8	modulation	modulation	NOUN
ajst-28825	68	9	recognition	recognition	NOUN
ajst-28825	68	10	,	,	PUNCT
ajst-28825	68	11	we	we	PRON
ajst-28825	68	12	preprocess	preprocess	VERB
ajst-28825	68	13	the	the	DET
ajst-28825	68	14	input	input	NOUN
ajst-28825	68	15	signal	signal	NOUN
ajst-28825	68	16	data	datum	NOUN
ajst-28825	68	17	using	use	VERB
ajst-28825	68	18	the	the	DET
ajst-28825	68	19	following	following	ADJ
ajst-28825	68	20	methods	method	NOUN
ajst-28825	68	21	:	:	PUNCT
ajst-28825	68	22	normalization	normalization	NOUN
ajst-28825	68	23	:	:	PUNCT
ajst-28825	68	24	all	all	DET
ajst-28825	68	25	input	input	NOUN
ajst-28825	68	26	signal	signal	NOUN
ajst-28825	68	27	data	datum	NOUN
ajst-28825	68	28	is	be	AUX
ajst-28825	68	29	normalized	normalize	VERB
ajst-28825	68	30	to	to	ADP
ajst-28825	68	31	the	the	DET
ajst-28825	68	32	[	[	X
ajst-28825	68	33	0,1	0,1	NUM
ajst-28825	68	34	]	]	PUNCT
ajst-28825	68	35	range	range	NOUN
ajst-28825	68	36	to	to	PART
ajst-28825	68	37	eliminate	eliminate	VERB
ajst-28825	68	38	amplitude	amplitude	NOUN
ajst-28825	68	39	differences	difference	NOUN
ajst-28825	68	40	between	between	ADP
ajst-28825	68	41	signals	signal	NOUN
ajst-28825	68	42	,	,	PUNCT
ajst-28825	68	43	reducing	reduce	VERB
ajst-28825	68	44	their	their	PRON
ajst-28825	68	45	impact	impact	NOUN
ajst-28825	68	46	on	on	ADP
ajst-28825	68	47	the	the	DET
ajst-28825	68	48	model[10	model[10	NOUN
ajst-28825	68	49	]	]	PUNCT
ajst-28825	68	50	.	.	PUNCT
ajst-28825	69	1	signal	signal	ADJ
ajst-28825	69	2	framing	framing	NOUN
ajst-28825	69	3	and	and	CCONJ
ajst-28825	69	4	overlapping	overlap	VERB
ajst-28825	69	5	:	:	PUNCT
ajst-28825	69	6	to	to	PART
ajst-28825	69	7	capture	capture	VERB
ajst-28825	69	8	richer	rich	ADJ
ajst-28825	69	9	temporal	temporal	ADJ
ajst-28825	69	10	information	information	NOUN
ajst-28825	69	11	,	,	PUNCT
ajst-28825	69	12	the	the	DET
ajst-28825	69	13	input	input	NOUN
ajst-28825	69	14	signal	signal	NOUN
ajst-28825	69	15	is	be	AUX
ajst-28825	69	16	divided	divide	VERB
ajst-28825	69	17	into	into	ADP
ajst-28825	69	18	multiple	multiple	ADJ
ajst-28825	69	19	frames	frame	NOUN
ajst-28825	69	20	,	,	PUNCT
ajst-28825	69	21	with	with	ADP
ajst-28825	69	22	a	a	DET
ajst-28825	69	23	certain	certain	ADJ
ajst-28825	69	24	overlap	overlap	NOUN
ajst-28825	69	25	between	between	ADP
ajst-28825	69	26	frames	frame	NOUN
ajst-28825	69	27	.	.	PUNCT
ajst-28825	70	1	each	each	DET
ajst-28825	70	2	frame	frame	NOUN
ajst-28825	70	3	of	of	ADP
ajst-28825	70	4	the	the	DET
ajst-28825	70	5	signal	signal	NOUN
ajst-28825	70	6	is	be	AUX
ajst-28825	70	7	treated	treat	VERB
ajst-28825	70	8	as	as	ADP
ajst-28825	70	9	an	an	DET
ajst-28825	70	10	independent	independent	ADJ
ajst-28825	70	11	input	input	NOUN
ajst-28825	70	12	sample	sample	NOUN
ajst-28825	70	13	.	.	PUNCT
ajst-28825	71	1	data	datum	NOUN
ajst-28825	71	2	augmentation	augmentation	NOUN
ajst-28825	71	3	:	:	PUNCT
ajst-28825	71	4	to	to	PART
ajst-28825	71	5	increase	increase	VERB
ajst-28825	71	6	the	the	DET
ajst-28825	71	7	diversity	diversity	NOUN
ajst-28825	71	8	of	of	ADP
ajst-28825	71	9	the	the	DET
ajst-28825	71	10	training	training	NOUN
ajst-28825	71	11	data	datum	NOUN
ajst-28825	71	12	,	,	PUNCT
ajst-28825	71	13	we	we	PRON
ajst-28825	71	14	applied	apply	VERB
ajst-28825	71	15	data	datum	NOUN
ajst-28825	71	16	augmentation	augmentation	NOUN
ajst-28825	71	17	techniques	technique	NOUN
ajst-28825	71	18	such	such	ADJ
ajst-28825	71	19	as	as	ADP
ajst-28825	71	20	random	random	ADJ
ajst-28825	71	21	noise	noise	NOUN
ajst-28825	71	22	addition	addition	NOUN
ajst-28825	71	23	and	and	CCONJ
ajst-28825	71	24	frequency	frequency	NOUN
ajst-28825	71	25	shifting[11	shifting[11	NOUN
ajst-28825	71	26	]	]	PUNCT
ajst-28825	71	27	.	.	PUNCT
ajst-28825	72	1	these	these	DET
ajst-28825	72	2	augmentations	augmentation	NOUN
ajst-28825	72	3	help	help	VERB
ajst-28825	72	4	the	the	DET
ajst-28825	72	5	model	model	NOUN
ajst-28825	72	6	generalize	generalize	VERB
ajst-28825	72	7	better	well	ADV
ajst-28825	72	8	to	to	ADP
ajst-28825	72	9	different	different	ADJ
ajst-28825	72	10	channel	channel	NOUN
ajst-28825	72	11	environments	environment	NOUN
ajst-28825	72	12	.	.	PUNCT
ajst-28825	73	1	3.3	3.3	NUM
ajst-28825	73	2	.	.	PUNCT
ajst-28825	74	1	network	network	NOUN
ajst-28825	74	2	architecture	architecture	NOUN
ajst-28825	74	3	design	design	VERB
ajst-28825	74	4	the	the	DET
ajst-28825	74	5	overall	overall	ADJ
ajst-28825	74	6	structure	structure	NOUN
ajst-28825	74	7	of	of	ADP
ajst-28825	74	8	the	the	DET
ajst-28825	74	9	rt	rt	PROPN
ajst-28825	74	10	-	-	PUNCT
ajst-28825	74	11	detr	detr	NOUN
ajst-28825	74	12	model	model	NOUN
ajst-28825	74	13	is	be	AUX
ajst-28825	74	14	composed	compose	VERB
ajst-28825	74	15	of	of	ADP
ajst-28825	74	16	the	the	DET
ajst-28825	74	17	following	follow	VERB
ajst-28825	74	18	parts	part	NOUN
ajst-28825	74	19	:	:	PUNCT
ajst-28825	74	20	backbone	backbone	NOUN
ajst-28825	74	21	network	network	NOUN
ajst-28825	74	22	:	:	PUNCT
ajst-28825	74	23	resnet-18	resnet-18	PROPN
ajst-28825	74	24	serves	serve	VERB
ajst-28825	74	25	as	as	ADP
ajst-28825	74	26	the	the	DET
ajst-28825	74	27	backbone	backbone	NOUN
ajst-28825	74	28	,	,	PUNCT
ajst-28825	74	29	extracting	extract	VERB
ajst-28825	74	30	primary	primary	ADJ
ajst-28825	74	31	features	feature	NOUN
ajst-28825	74	32	from	from	ADP
ajst-28825	74	33	the	the	DET
ajst-28825	74	34	input	input	NOUN
ajst-28825	74	35	signal[6	signal[6	PROPN
ajst-28825	74	36	]	]	PUNCT
ajst-28825	74	37	.	.	PUNCT
ajst-28825	75	1	the	the	DET
ajst-28825	75	2	network	network	NOUN
ajst-28825	75	3	consists	consist	VERB
ajst-28825	75	4	of	of	ADP
ajst-28825	75	5	several	several	ADJ
ajst-28825	75	6	convolutional	convolutional	ADJ
ajst-28825	75	7	layers	layer	NOUN
ajst-28825	75	8	and	and	CCONJ
ajst-28825	75	9	residual	residual	ADJ
ajst-28825	75	10	blocks	block	NOUN
ajst-28825	75	11	,	,	PUNCT
ajst-28825	75	12	with	with	ADP
ajst-28825	75	13	each	each	DET
ajst-28825	75	14	residual	residual	ADJ
ajst-28825	75	15	block	block	NOUN
ajst-28825	75	16	using	use	VERB
ajst-28825	75	17	skip	skip	ADJ
ajst-28825	75	18	connections	connection	NOUN
ajst-28825	75	19	to	to	PART
ajst-28825	75	20	enhance	enhance	VERB
ajst-28825	75	21	gradient	gradient	ADJ
ajst-28825	75	22	flow	flow	NOUN
ajst-28825	75	23	,	,	PUNCT
ajst-28825	75	24	ensuring	ensure	VERB
ajst-28825	75	25	stable	stable	ADJ
ajst-28825	75	26	training	training	NOUN
ajst-28825	75	27	of	of	ADP
ajst-28825	75	28	the	the	DET
ajst-28825	75	29	deep	deep	ADJ
ajst-28825	75	30	network	network	NOUN
ajst-28825	75	31	.	.	PUNCT
ajst-28825	76	1	transformer	transformer	NOUN
ajst-28825	76	2	encoder	encoder	NOUN
ajst-28825	76	3	:	:	PUNCT
ajst-28825	76	4	after	after	ADP
ajst-28825	76	5	extracting	extract	VERB
ajst-28825	76	6	primary	primary	ADJ
ajst-28825	76	7	features	feature	NOUN
ajst-28825	76	8	,	,	PUNCT
ajst-28825	76	9	the	the	DET
ajst-28825	76	10	feature	feature	NOUN
ajst-28825	76	11	maps	map	NOUN
ajst-28825	76	12	are	be	AUX
ajst-28825	76	13	input	input	VERB
ajst-28825	76	14	into	into	ADP
ajst-28825	76	15	the	the	DET
ajst-28825	76	16	transformer	transformer	NOUN
ajst-28825	76	17	encoder	encoder	NOUN
ajst-28825	76	18	.	.	PUNCT
ajst-28825	77	1	the	the	DET
ajst-28825	77	2	self	self	NOUN
ajst-28825	77	3	-	-	PUNCT
ajst-28825	77	4	attention	attention	NOUN
ajst-28825	77	5	mechanism	mechanism	NOUN
ajst-28825	77	6	of	of	ADP
ajst-28825	77	7	the	the	DET
ajst-28825	77	8	transformer	transformer	NOUN
ajst-28825	77	9	allows	allow	VERB
ajst-28825	77	10	it	it	PRON
ajst-28825	77	11	to	to	PART
ajst-28825	77	12	effectively	effectively	ADV
ajst-28825	77	13	capture	capture	VERB
ajst-28825	77	14	global	global	ADJ
ajst-28825	77	15	information	information	NOUN
ajst-28825	77	16	in	in	ADP
ajst-28825	77	17	the	the	DET
ajst-28825	77	18	signal	signal	NOUN
ajst-28825	77	19	,	,	PUNCT
ajst-28825	77	20	enhancing	enhance	VERB
ajst-28825	77	21	recognition	recognition	NOUN
ajst-28825	77	22	performance[13	performance[13	NOUN
ajst-28825	77	23	]	]	X
ajst-28825	77	24	.	.	PUNCT
ajst-28825	78	1	aifi	aifi	NOUN
ajst-28825	78	2	-	-	PUNCT
ajst-28825	78	3	dattention	dattention	NOUN
ajst-28825	78	4	module	module	NOUN
ajst-28825	78	5	:	:	PUNCT
ajst-28825	78	6	the	the	DET
ajst-28825	78	7	aifi	aifi	NOUN
ajst-28825	78	8	-	-	PUNCT
ajst-28825	78	9	dattention	dattention	NOUN
ajst-28825	78	10	module	module	NOUN
ajst-28825	78	11	is	be	AUX
ajst-28825	78	12	integrated	integrate	VERB
ajst-28825	78	13	into	into	ADP
ajst-28825	78	14	the	the	DET
ajst-28825	78	15	network	network	NOUN
ajst-28825	78	16	to	to	PART
ajst-28825	78	17	apply	apply	VERB
ajst-28825	78	18	weighted	weighted	ADJ
ajst-28825	78	19	processing	processing	NOUN
ajst-28825	78	20	to	to	ADP
ajst-28825	78	21	the	the	DET
ajst-28825	78	22	feature	feature	NOUN
ajst-28825	78	23	maps	map	NOUN
ajst-28825	78	24	output	output	NOUN
ajst-28825	78	25	by	by	ADP
ajst-28825	78	26	the	the	DET
ajst-28825	78	27	transformer	transformer	NOUN
ajst-28825	78	28	encoder	encoder	NOUN
ajst-28825	78	29	.	.	PUNCT
ajst-28825	79	1	its	its	PRON
ajst-28825	79	2	dynamic	dynamic	ADJ
ajst-28825	79	3	attention	attention	NOUN
ajst-28825	79	4	mechanism	mechanism	NOUN
ajst-28825	79	5	allows	allow	VERB
ajst-28825	79	6	the	the	DET
ajst-28825	79	7	network	network	NOUN
ajst-28825	79	8	to	to	PART
ajst-28825	79	9	adaptively	adaptively	ADV
ajst-28825	79	10	adjust	adjust	VERB
ajst-28825	79	11	attention	attention	NOUN
ajst-28825	79	12	distribution	distribution	NOUN
ajst-28825	79	13	,	,	PUNCT
ajst-28825	79	14	better	well	ADV
ajst-28825	79	15	capturing	capture	VERB
ajst-28825	79	16	key	key	ADJ
ajst-28825	79	17	features	feature	NOUN
ajst-28825	79	18	classifier	classifier	NOUN
ajst-28825	79	19	:	:	PUNCT
ajst-28825	79	20	the	the	DET
ajst-28825	79	21	final	final	ADJ
ajst-28825	79	22	feature	feature	NOUN
ajst-28825	79	23	map	map	NOUN
ajst-28825	79	24	is	be	AUX
ajst-28825	79	25	input	input	VERB
ajst-28825	79	26	into	into	ADP
ajst-28825	79	27	a	a	DET
ajst-28825	79	28	fully	fully	ADV
ajst-28825	79	29	connected	connect	VERB
ajst-28825	79	30	layer	layer	NOUN
ajst-28825	79	31	for	for	ADP
ajst-28825	79	32	classification	classification	NOUN
ajst-28825	79	33	.	.	PUNCT
ajst-28825	80	1	this	this	DET
ajst-28825	80	2	layer	layer	NOUN
ajst-28825	80	3	outputs	output	VERB
ajst-28825	80	4	the	the	DET
ajst-28825	80	5	probability	probability	NOUN
ajst-28825	80	6	distribution	distribution	NOUN
ajst-28825	80	7	corresponding	correspond	VERB
ajst-28825	80	8	to	to	ADP
ajst-28825	80	9	various	various	ADJ
ajst-28825	80	10	modulation	modulation	NOUN
ajst-28825	80	11	signals	signal	NOUN
ajst-28825	80	12	,	,	PUNCT
ajst-28825	80	13	with	with	ADP
ajst-28825	80	14	the	the	DET
ajst-28825	80	15	predicted	predict	VERB
ajst-28825	80	16	modulation	modulation	NOUN
ajst-28825	80	17	type	type	NOUN
ajst-28825	80	18	being	be	AUX
ajst-28825	80	19	the	the	DET
ajst-28825	80	20	one	one	NOUN
ajst-28825	80	21	with	with	ADP
ajst-28825	80	22	the	the	DET
ajst-28825	80	23	highest	high	ADJ
ajst-28825	80	24	probability	probability	NOUN
ajst-28825	80	25	.	.	PUNCT
ajst-28825	81	1	3.4	3.4	NUM
ajst-28825	81	2	.	.	PUNCT
ajst-28825	81	3	model	model	NOUN
ajst-28825	81	4	training	training	NOUN
ajst-28825	81	5	and	and	CCONJ
ajst-28825	81	6	optimization	optimization	NOUN
ajst-28825	81	7	we	we	PRON
ajst-28825	81	8	use	use	VERB
ajst-28825	81	9	the	the	DET
ajst-28825	81	10	cross	cross	ADJ
ajst-28825	81	11	-	-	ADJ
ajst-28825	81	12	entropy	entropy	ADJ
ajst-28825	81	13	loss	loss	NOUN
ajst-28825	81	14	function	function	NOUN
ajst-28825	81	15	as	as	ADP
ajst-28825	81	16	the	the	DET
ajst-28825	81	17	optimization	optimization	NOUN
ajst-28825	81	18	objective	objective	NOUN
ajst-28825	81	19	of	of	ADP
ajst-28825	81	20	the	the	DET
ajst-28825	81	21	model	model	NOUN
ajst-28825	81	22	.	.	PUNCT
ajst-28825	82	1	the	the	DET
ajst-28825	82	2	model	model	NOUN
ajst-28825	82	3	training	training	NOUN
ajst-28825	82	4	process	process	NOUN
ajst-28825	82	5	uses	use	VERB
ajst-28825	82	6	the	the	DET
ajst-28825	82	7	adam	adam	PROPN
ajst-28825	82	8	optimizer	optimizer	NOUN
ajst-28825	82	9	with	with	ADP
ajst-28825	82	10	the	the	DET
ajst-28825	82	11	following	follow	VERB
ajst-28825	82	12	hyperparameters	hyperparameter	NOUN
ajst-28825	82	13	:	:	PUNCT
ajst-28825	82	14	learning	learn	VERB
ajst-28825	82	15	rate	rate	NOUN
ajst-28825	82	16	:	:	PUNCT
ajst-28825	82	17	the	the	DET
ajst-28825	82	18	initial	initial	ADJ
ajst-28825	82	19	learning	learning	NOUN
ajst-28825	82	20	rate	rate	NOUN
ajst-28825	82	21	is	be	AUX
ajst-28825	82	22	set	set	VERB
ajst-28825	82	23	to	to	ADP
ajst-28825	82	24	0.001	0.001	NUM
ajst-28825	82	25	,	,	PUNCT
ajst-28825	82	26	and	and	CCONJ
ajst-28825	82	27	a	a	DET
ajst-28825	82	28	cosine	cosine	NOUN
ajst-28825	82	29	annealing	anneal	VERB
ajst-28825	82	30	strategy	strategy	NOUN
ajst-28825	82	31	is	be	AUX
ajst-28825	82	32	used	use	VERB
ajst-28825	82	33	to	to	PART
ajst-28825	82	34	dynamically	dynamically	ADV
ajst-28825	82	35	adjust	adjust	VERB
ajst-28825	82	36	the	the	DET
ajst-28825	82	37	learning	learning	NOUN
ajst-28825	82	38	rate	rate	NOUN
ajst-28825	82	39	to	to	PART
ajst-28825	82	40	ensure	ensure	VERB
ajst-28825	82	41	stability	stability	NOUN
ajst-28825	82	42	and	and	CCONJ
ajst-28825	82	43	fast	fast	ADJ
ajst-28825	82	44	convergence	convergence	NOUN
ajst-28825	82	45	during	during	ADP
ajst-28825	82	46	training	training	NOUN
ajst-28825	82	47	.	.	PUNCT
ajst-28825	83	1	batch	batch	NOUN
ajst-28825	83	2	size	size	NOUN
ajst-28825	83	3	:	:	PUNCT
ajst-28825	83	4	the	the	DET
ajst-28825	83	5	batch	batch	NOUN
ajst-28825	83	6	size	size	NOUN
ajst-28825	83	7	is	be	AUX
ajst-28825	83	8	set	set	VERB
ajst-28825	83	9	to	to	ADP
ajst-28825	83	10	32	32	NUM
ajst-28825	83	11	,	,	PUNCT
ajst-28825	83	12	striking	strike	VERB
ajst-28825	83	13	a	a	DET
ajst-28825	83	14	good	good	ADJ
ajst-28825	83	15	balance	balance	NOUN
ajst-28825	83	16	between	between	ADP
ajst-28825	83	17	training	training	NOUN
ajst-28825	83	18	speed	speed	NOUN
ajst-28825	83	19	and	and	CCONJ
ajst-28825	83	20	model	model	NOUN
ajst-28825	83	21	performance	performance	NOUN
ajst-28825	83	22	.	.	PUNCT
ajst-28825	84	1	epochs	epoch	NOUN
ajst-28825	84	2	:	:	PUNCT
ajst-28825	84	3	the	the	DET
ajst-28825	84	4	model	model	NOUN
ajst-28825	84	5	was	be	AUX
ajst-28825	84	6	trained	train	VERB
ajst-28825	84	7	for	for	ADP
ajst-28825	84	8	100	100	NUM
ajst-28825	84	9	epochs	epoch	NOUN
ajst-28825	84	10	.	.	PUNCT
ajst-28825	85	1	after	after	ADP
ajst-28825	85	2	each	each	DET
ajst-28825	85	3	epoch	epoch	NOUN
ajst-28825	85	4	,	,	PUNCT
ajst-28825	85	5	the	the	DET
ajst-28825	85	6	model	model	NOUN
ajst-28825	85	7	's	's	PART
ajst-28825	85	8	performance	performance	NOUN
ajst-28825	85	9	was	be	AUX
ajst-28825	85	10	evaluated	evaluate	VERB
ajst-28825	85	11	on	on	ADP
ajst-28825	85	12	the	the	DET
ajst-28825	85	13	validation	validation	NOUN
ajst-28825	85	14	set	set	NOUN
ajst-28825	85	15	,	,	PUNCT
ajst-28825	85	16	and	and	CCONJ
ajst-28825	85	17	the	the	DET
ajst-28825	85	18	model	model	NOUN
ajst-28825	85	19	with	with	ADP
ajst-28825	85	20	the	the	DET
ajst-28825	85	21	highest	high	ADJ
ajst-28825	85	22	validation	validation	NOUN
ajst-28825	85	23	accuracy	accuracy	NOUN
ajst-28825	85	24	was	be	AUX
ajst-28825	85	25	saved	save	VERB
ajst-28825	85	26	as	as	ADP
ajst-28825	85	27	the	the	DET
ajst-28825	85	28	final	final	ADJ
ajst-28825	85	29	model	model	NOUN
ajst-28825	85	30	.	.	PUNCT
ajst-28825	86	1	weight	weight	NOUN
ajst-28825	86	2	decay	decay	NOUN
ajst-28825	86	3	:	:	PUNCT
ajst-28825	86	4	to	to	PART
ajst-28825	86	5	prevent	prevent	VERB
ajst-28825	86	6	overfitting	overfitting	NOUN
ajst-28825	86	7	,	,	PUNCT
ajst-28825	86	8	a	a	DET
ajst-28825	86	9	weight	weight	NOUN
ajst-28825	86	10	decay	decay	NOUN
ajst-28825	86	11	strategy	strategy	NOUN
ajst-28825	86	12	with	with	ADP
ajst-28825	86	13	a	a	DET
ajst-28825	86	14	decay	decay	NOUN
ajst-28825	86	15	rate	rate	NOUN
ajst-28825	86	16	of	of	ADP
ajst-28825	86	17	0.0001	0.0001	NUM
ajst-28825	86	18	was	be	AUX
ajst-28825	86	19	employed	employ	VERB
ajst-28825	86	20	.	.	PUNCT
ajst-28825	87	1	gradient	gradient	ADJ
ajst-28825	87	2	clipping	clipping	NOUN
ajst-28825	87	3	:	:	PUNCT
ajst-28825	87	4	to	to	PART
ajst-28825	87	5	avoid	avoid	VERB
ajst-28825	87	6	the	the	DET
ajst-28825	87	7	gradient	gradient	ADJ
ajst-28825	87	8	explosion	explosion	NOUN
ajst-28825	87	9	problem	problem	NOUN
ajst-28825	87	10	,	,	PUNCT
ajst-28825	87	11	a	a	DET
ajst-28825	87	12	gradient	gradient	NOUN
ajst-28825	87	13	clipping	clip	VERB
ajst-28825	87	14	strategy	strategy	NOUN
ajst-28825	87	15	was	be	AUX
ajst-28825	87	16	used	use	VERB
ajst-28825	87	17	with	with	ADP
ajst-28825	87	18	a	a	DET
ajst-28825	87	19	threshold	threshold	NOUN
ajst-28825	87	20	set	set	VERB
ajst-28825	87	21	at	at	ADP
ajst-28825	87	22	1.0	1.0	NUM
ajst-28825	87	23	.	.	PUNCT
ajst-28825	88	1	additionally	additionally	ADV
ajst-28825	88	2	,	,	PUNCT
ajst-28825	88	3	we	we	PRON
ajst-28825	88	4	applied	apply	VERB
ajst-28825	88	5	an	an	DET
ajst-28825	88	6	early	early	ADJ
ajst-28825	88	7	stopping	stopping	NOUN
ajst-28825	88	8	strategy	strategy	NOUN
ajst-28825	88	9	during	during	ADP
ajst-28825	88	10	training	training	NOUN
ajst-28825	88	11	.	.	PUNCT
ajst-28825	89	1	if	if	SCONJ
ajst-28825	89	2	the	the	DET
ajst-28825	89	3	validation	validation	NOUN
ajst-28825	89	4	loss	loss	NOUN
ajst-28825	89	5	did	do	AUX
ajst-28825	89	6	not	not	PART
ajst-28825	89	7	decrease	decrease	VERB
ajst-28825	89	8	for	for	ADP
ajst-28825	89	9	10	10	NUM
ajst-28825	89	10	consecutive	consecutive	ADJ
ajst-28825	89	11	epochs	epoch	NOUN
ajst-28825	89	12	,	,	PUNCT
ajst-28825	89	13	the	the	DET
ajst-28825	89	14	training	training	NOUN
ajst-28825	89	15	process	process	NOUN
ajst-28825	89	16	was	be	AUX
ajst-28825	89	17	terminated	terminate	VERB
ajst-28825	89	18	early[12	early[12	NOUN
ajst-28825	89	19	]	]	X
ajst-28825	89	20	.	.	PUNCT
ajst-28825	90	1	3.5	3.5	NUM
ajst-28825	90	2	.	.	PUNCT
ajst-28825	91	1	experimental	experimental	ADJ
ajst-28825	91	2	setup	setup	NOUN
ajst-28825	91	3	to	to	PART
ajst-28825	91	4	validate	validate	VERB
ajst-28825	91	5	the	the	DET
ajst-28825	91	6	effectiveness	effectiveness	NOUN
ajst-28825	91	7	of	of	ADP
ajst-28825	91	8	the	the	DET
ajst-28825	91	9	proposed	propose	VERB
ajst-28825	91	10	model	model	NOUN
ajst-28825	91	11	,	,	PUNCT
ajst-28825	91	12	we	we	PRON
ajst-28825	91	13	designed	design	VERB
ajst-28825	91	14	a	a	DET
ajst-28825	91	15	series	series	NOUN
ajst-28825	91	16	of	of	ADP
ajst-28825	91	17	experiments	experiment	NOUN
ajst-28825	91	18	.	.	PUNCT
ajst-28825	92	1	these	these	DET
ajst-28825	92	2	experiments	experiment	NOUN
ajst-28825	92	3	were	be	AUX
ajst-28825	92	4	conducted	conduct	VERB
ajst-28825	92	5	on	on	ADP
ajst-28825	92	6	multiple	multiple	ADJ
ajst-28825	92	7	public	public	ADJ
ajst-28825	92	8	datasets	dataset	NOUN
ajst-28825	92	9	,	,	PUNCT
ajst-28825	92	10	covering	cover	VERB
ajst-28825	92	11	different	different	ADJ
ajst-28825	92	12	signal	signal	NOUN
ajst-28825	92	13	-	-	PUNCT
ajst-28825	92	14	to	to	ADP
ajst-28825	92	15	-	-	PUNCT
ajst-28825	92	16	noise	noise	NOUN
ajst-28825	92	17	ratios	ratio	NOUN
ajst-28825	92	18	(	(	PUNCT
ajst-28825	92	19	snr	snr	PROPN
ajst-28825	92	20	)	)	PUNCT
ajst-28825	92	21	and	and	CCONJ
ajst-28825	92	22	various	various	ADJ
ajst-28825	92	23	modulation	modulation	NOUN
ajst-28825	92	24	types	type	NOUN
ajst-28825	92	25	.	.	PUNCT
ajst-28825	93	1	we	we	PRON
ajst-28825	93	2	compared	compare	VERB
ajst-28825	93	3	the	the	DET
ajst-28825	93	4	performance	performance	NOUN
ajst-28825	93	5	of	of	ADP
ajst-28825	93	6	models	model	NOUN
ajst-28825	93	7	using	use	VERB
ajst-28825	93	8	different	different	ADJ
ajst-28825	93	9	backbones	backbone	NOUN
ajst-28825	93	10	(	(	PUNCT
ajst-28825	93	11	such	such	ADJ
ajst-28825	93	12	as	as	ADP
ajst-28825	93	13	resnet-18	resnet-18	PROPN
ajst-28825	93	14	)	)	PUNCT
ajst-28825	93	15	and	and	CCONJ
ajst-28825	93	16	evaluated	evaluate	VERB
ajst-28825	93	17	the	the	DET
ajst-28825	93	18	models	model	NOUN
ajst-28825	93	19	with	with	ADP
ajst-28825	93	20	and	and	CCONJ
ajst-28825	93	21	without	without	ADP
ajst-28825	93	22	the	the	DET
ajst-28825	93	23	aifi	aifi	NOUN
ajst-28825	93	24	-	-	PUNCT
ajst-28825	93	25	dattention	dattention	NOUN
ajst-28825	93	26	.	.	PUNCT
ajst-28825	94	1	the	the	DET
ajst-28825	94	2	experimental	experimental	ADJ
ajst-28825	94	3	results	result	NOUN
ajst-28825	94	4	were	be	AUX
ajst-28825	94	5	evaluated	evaluate	VERB
ajst-28825	94	6	using	use	VERB
ajst-28825	94	7	metrics	metric	NOUN
ajst-28825	94	8	such	such	ADJ
ajst-28825	94	9	as	as	ADP
ajst-28825	94	10	accuracy	accuracy	NOUN
ajst-28825	94	11	,	,	PUNCT
ajst-28825	94	12	precision	precision	NOUN
ajst-28825	94	13	,	,	PUNCT
ajst-28825	94	14	and	and	CCONJ
ajst-28825	94	15	recall	recall	NOUN
ajst-28825	94	16	.	.	PUNCT
ajst-28825	95	1	3.6	3.6	NUM
ajst-28825	95	2	.	.	PUNCT
ajst-28825	96	1	dataset	dataset	NOUN
ajst-28825	96	2	generation	generation	NOUN
ajst-28825	96	3	the	the	DET
ajst-28825	96	4	dataset	dataset	NOUN
ajst-28825	96	5	for	for	ADP
ajst-28825	96	6	this	this	DET
ajst-28825	96	7	study	study	NOUN
ajst-28825	96	8	was	be	AUX
ajst-28825	96	9	generated	generate	VERB
ajst-28825	96	10	through	through	ADP
ajst-28825	96	11	matlab	matlab	PROPN
ajst-28825	96	12	simulations	simulation	NOUN
ajst-28825	96	13	,	,	PUNCT
ajst-28825	96	14	with	with	ADP
ajst-28825	96	15	the	the	DET
ajst-28825	96	16	signal	signal	NOUN
ajst-28825	96	17	-	-	PUNCT
ajst-28825	96	18	to	to	ADP
ajst-28825	96	19	-	-	PUNCT
ajst-28825	96	20	noise	noise	NOUN
ajst-28825	96	21	ratio	ratio	NOUN
ajst-28825	96	22	(	(	PUNCT
ajst-28825	96	23	snr	snr	PROPN
ajst-28825	96	24	)	)	PUNCT
ajst-28825	96	25	for	for	ADP
ajst-28825	96	26	all	all	DET
ajst-28825	96	27	signals	signal	NOUN
ajst-28825	96	28	set	set	VERB
ajst-28825	96	29	at	at	ADP
ajst-28825	96	30	30	30	NUM
ajst-28825	96	31	db	db	NOUN
ajst-28825	96	32	.	.	PUNCT
ajst-28825	97	1	this	this	DET
ajst-28825	97	2	snr	snr	NOUN
ajst-28825	97	3	level	level	NOUN
ajst-28825	97	4	represents	represent	VERB
ajst-28825	97	5	a	a	DET
ajst-28825	97	6	low	low	ADJ
ajst-28825	97	7	-	-	PUNCT
ajst-28825	97	8	noise	noise	NOUN
ajst-28825	97	9	environment	environment	NOUN
ajst-28825	97	10	,	,	PUNCT
ajst-28825	97	11	suitable	suitable	ADJ
ajst-28825	97	12	for	for	ADP
ajst-28825	97	13	testing	test	VERB
ajst-28825	97	14	the	the	DET
ajst-28825	97	15	model	model	NOUN
ajst-28825	97	16	's	's	PART
ajst-28825	97	17	recognition	recognition	NOUN
ajst-28825	97	18	ability	ability	NOUN
ajst-28825	97	19	under	under	ADP
ajst-28825	97	20	ideal	ideal	ADJ
ajst-28825	97	21	conditions	condition	NOUN
ajst-28825	97	22	.	.	PUNCT
ajst-28825	98	1	the	the	DET
ajst-28825	98	2	dataset	dataset	NOUN
ajst-28825	98	3	generation	generation	NOUN
ajst-28825	98	4	process	process	NOUN
ajst-28825	98	5	is	be	AUX
ajst-28825	98	6	as	as	SCONJ
ajst-28825	98	7	follows	follow	VERB
ajst-28825	98	8	:	:	PUNCT
ajst-28825	98	9	modulation	modulation	NOUN
ajst-28825	98	10	types	type	NOUN
ajst-28825	98	11	:	:	PUNCT
ajst-28825	98	12	the	the	DET
ajst-28825	98	13	dataset	dataset	NOUN
ajst-28825	98	14	includes	include	VERB
ajst-28825	98	15	multiple	multiple	ADJ
ajst-28825	98	16	common	common	ADJ
ajst-28825	98	17	digital	digital	ADJ
ajst-28825	98	18	modulation	modulation	NOUN
ajst-28825	98	19	types	type	NOUN
ajst-28825	98	20	,	,	PUNCT
ajst-28825	98	21	including	include	VERB
ajst-28825	98	22	bpsk	bpsk	NOUN
ajst-28825	98	23	,	,	PUNCT
ajst-28825	98	24	qpsk	qpsk	NOUN
ajst-28825	98	25	,	,	PUNCT
ajst-28825	98	26	8psk	8psk	NUM
ajst-28825	98	27	,	,	PUNCT
ajst-28825	98	28	16qam	16qam	NOUN
ajst-28825	98	29	,	,	PUNCT
ajst-28825	98	30	and	and	CCONJ
ajst-28825	99	1	64qam[5	64qam[5	NUM
ajst-28825	99	2	]	]	PUNCT
ajst-28825	99	3	.	.	PUNCT
ajst-28825	99	4	signal	signal	PROPN
ajst-28825	99	5	length	length	NOUN
ajst-28825	99	6	and	and	CCONJ
ajst-28825	99	7	sampling	sample	VERB
ajst-28825	99	8	rate	rate	NOUN
ajst-28825	99	9	:	:	PUNCT
ajst-28825	99	10	each	each	DET
ajst-28825	99	11	signal	signal	NOUN
ajst-28825	99	12	length	length	NOUN
ajst-28825	99	13	is	be	AUX
ajst-28825	99	14	set	set	VERB
ajst-28825	99	15	to	to	ADP
ajst-28825	99	16	1024	1024	NUM
ajst-28825	99	17	sample	sample	NOUN
ajst-28825	99	18	points	point	NOUN
ajst-28825	99	19	,	,	PUNCT
ajst-28825	99	20	with	with	ADP
ajst-28825	99	21	a	a	DET
ajst-28825	99	22	sampling	sample	VERB
ajst-28825	99	23	rate	rate	NOUN
ajst-28825	99	24	of	of	ADP
ajst-28825	99	25	1	1	NUM
ajst-28825	99	26	mhz	mhz	NOUN
ajst-28825	99	27	.	.	PUNCT
ajst-28825	100	1	the	the	DET
ajst-28825	100	2	generated	generate	VERB
ajst-28825	100	3	signals	signal	NOUN
ajst-28825	100	4	are	be	AUX
ajst-28825	100	5	stored	store	VERB
ajst-28825	100	6	in	in	ADP
ajst-28825	100	7	complex	complex	ADJ
ajst-28825	100	8	i	i	PROPN
ajst-28825	100	9	/	/	SYM
ajst-28825	100	10	q	q	NOUN
ajst-28825	100	11	data	data	NOUN
ajst-28825	100	12	format	format	NOUN
ajst-28825	100	13	for	for	ADP
ajst-28825	100	14	subsequent	subsequent	ADJ
ajst-28825	100	15	processing	processing	NOUN
ajst-28825	100	16	and	and	CCONJ
ajst-28825	100	17	feature	feature	NOUN
ajst-28825	100	18	extraction	extraction	NOUN
ajst-28825	100	19	.	.	PUNCT
ajst-28825	101	1	data	datum	NOUN
ajst-28825	101	2	augmentation	augmentation	NOUN
ajst-28825	101	3	:	:	PUNCT
ajst-28825	101	4	although	although	SCONJ
ajst-28825	101	5	all	all	DET
ajst-28825	101	6	signals	signal	NOUN
ajst-28825	101	7	have	have	VERB
ajst-28825	101	8	an	an	DET
ajst-28825	101	9	snr	snr	NOUN
ajst-28825	101	10	set	set	VERB
ajst-28825	101	11	at	at	ADP
ajst-28825	101	12	30	30	NUM
ajst-28825	101	13	db	db	NOUN
ajst-28825	101	14	,	,	PUNCT
ajst-28825	101	15	other	other	ADJ
ajst-28825	101	16	simulation	simulation	NOUN
ajst-28825	101	17	conditions	condition	NOUN
ajst-28825	101	18	such	such	ADJ
ajst-28825	101	19	as	as	ADP
ajst-28825	101	20	frequency	frequency	NOUN
ajst-28825	101	21	offset	offset	NOUN
ajst-28825	101	22	and	and	CCONJ
ajst-28825	101	23	phase	phase	NOUN
ajst-28825	101	24	shift	shift	NOUN
ajst-28825	101	25	were	be	AUX
ajst-28825	101	26	introduced	introduce	VERB
ajst-28825	101	27	through	through	ADP
ajst-28825	101	28	data	datum	NOUN
ajst-28825	101	29	augmentation	augmentation	NOUN
ajst-28825	101	30	,	,	PUNCT
ajst-28825	101	31	increasing	increase	VERB
ajst-28825	101	32	the	the	DET
ajst-28825	101	33	diversity	diversity	NOUN
ajst-28825	101	34	and	and	CCONJ
ajst-28825	101	35	complexity	complexity	NOUN
ajst-28825	101	36	of	of	ADP
ajst-28825	101	37	the	the	DET
ajst-28825	101	38	data	datum	NOUN
ajst-28825	101	39	.	.	PUNCT
ajst-28825	102	1	dataset	dataset	ADJ
ajst-28825	102	2	splitting	splitting	NOUN
ajst-28825	102	3	:	:	PUNCT
ajst-28825	102	4	the	the	DET
ajst-28825	102	5	generated	generate	VERB
ajst-28825	102	6	signal	signal	NOUN
ajst-28825	102	7	data	datum	NOUN
ajst-28825	102	8	is	be	AUX
ajst-28825	102	9	split	split	VERB
ajst-28825	102	10	into	into	ADP
ajst-28825	102	11	training	training	NOUN
ajst-28825	102	12	,	,	PUNCT
ajst-28825	102	13	validation	validation	NOUN
ajst-28825	102	14	,	,	PUNCT
ajst-28825	102	15	and	and	CCONJ
ajst-28825	102	16	test	test	NOUN
ajst-28825	102	17	sets	set	NOUN
ajst-28825	102	18	in	in	ADP
ajst-28825	102	19	a	a	DET
ajst-28825	102	20	70	70	NUM
ajst-28825	102	21	%	%	NOUN
ajst-28825	102	22	,	,	PUNCT
ajst-28825	102	23	15	15	NUM
ajst-28825	102	24	%	%	NOUN
ajst-28825	102	25	,	,	PUNCT
ajst-28825	102	26	and	and	CCONJ
ajst-28825	102	27	15	15	NUM
ajst-28825	102	28	%	%	NOUN
ajst-28825	102	29	ratio	ratio	NOUN
ajst-28825	102	30	,	,	PUNCT
ajst-28825	102	31	ensuring	ensure	VERB
ajst-28825	102	32	an	an	DET
ajst-28825	102	33	even	even	ADJ
ajst-28825	102	34	distribution	distribution	NOUN
ajst-28825	102	35	of	of	ADP
ajst-28825	102	36	modulation	modulation	NOUN
ajst-28825	102	37	types	type	NOUN
ajst-28825	102	38	across	across	ADP
ajst-28825	102	39	different	different	ADJ
ajst-28825	102	40	datasets	dataset	NOUN
ajst-28825	102	41	.	.	PUNCT
ajst-28825	103	1	346	346	NUM
ajst-28825	103	2	4	4	NUM
ajst-28825	103	3	.	.	PUNCT
ajst-28825	103	4	experimental	experimental	ADJ
ajst-28825	103	5	process	process	NOUN
ajst-28825	103	6	and	and	CCONJ
ajst-28825	103	7	results	result	VERB
ajst-28825	103	8	analysis	analysis	NOUN
ajst-28825	103	9	4.1	4.1	NUM
ajst-28825	103	10	.	.	PUNCT
ajst-28825	104	1	training	training	NOUN
ajst-28825	104	2	process	process	NOUN
ajst-28825	104	3	in	in	ADP
ajst-28825	104	4	the	the	DET
ajst-28825	104	5	training	training	NOUN
ajst-28825	104	6	process	process	NOUN
ajst-28825	104	7	,	,	PUNCT
ajst-28825	104	8	the	the	DET
ajst-28825	104	9	data	data	NOUN
ajst-28825	104	10	features	feature	NOUN
ajst-28825	104	11	of	of	ADP
ajst-28825	104	12	communication	communication	NOUN
ajst-28825	104	13	signals	signal	NOUN
ajst-28825	104	14	are	be	AUX
ajst-28825	104	15	first	first	ADV
ajst-28825	104	16	extracted	extract	VERB
ajst-28825	104	17	using	use	VERB
ajst-28825	104	18	convolution	convolution	NOUN
ajst-28825	104	19	windows	window	NOUN
ajst-28825	104	20	in	in	ADP
ajst-28825	104	21	the	the	DET
ajst-28825	104	22	convolutional	convolutional	ADJ
ajst-28825	104	23	neural	neural	ADJ
ajst-28825	104	24	network	network	NOUN
ajst-28825	104	25	(	(	PUNCT
ajst-28825	104	26	cnn	cnn	PROPN
ajst-28825	104	27	)	)	PUNCT
ajst-28825	104	28	.	.	PUNCT
ajst-28825	105	1	the	the	DET
ajst-28825	105	2	categorical	categorical	ADJ
ajst-28825	105	3	crossentropy	crossentropy	NOUN
ajst-28825	105	4	loss	loss	NOUN
ajst-28825	105	5	function	function	NOUN
ajst-28825	105	6	is	be	AUX
ajst-28825	105	7	employed	employ	VERB
ajst-28825	105	8	,	,	PUNCT
ajst-28825	105	9	and	and	CCONJ
ajst-28825	105	10	the	the	DET
ajst-28825	105	11	loss	loss	NOUN
ajst-28825	105	12	function	function	NOUN
ajst-28825	105	13	𝐿	𝐿	PROPN
ajst-28825	105	14	is	be	AUX
ajst-28825	105	15	defined	define	VERB
ajst-28825	105	16	as	as	SCONJ
ajst-28825	105	17	follows	follow	VERB
ajst-28825	105	18	𝐿	𝐿	PROPN
ajst-28825	105	19	∑	∑	PROPN
ajst-28825	105	20	𝑡	𝑡	PROPN
ajst-28825	105	21	,	,	PUNCT
ajst-28825	105	22	log	log	NOUN
ajst-28825	105	23	𝑝	𝑝	NOUN
ajst-28825	105	24	,	,	PUNCT
ajst-28825	105	25	(	(	PUNCT
ajst-28825	105	26	1	1	X
ajst-28825	105	27	)	)	PUNCT
ajst-28825	105	28	where	where	SCONJ
ajst-28825	105	29	t	t	PROPN
ajst-28825	105	30	represents	represent	VERB
ajst-28825	105	31	the	the	DET
ajst-28825	105	32	true	true	ADJ
ajst-28825	105	33	labels	label	NOUN
ajst-28825	105	34	,	,	PUNCT
ajst-28825	105	35	i	i	PRON
ajst-28825	105	36	represents	represent	VERB
ajst-28825	105	37	the	the	DET
ajst-28825	105	38	input	input	NOUN
ajst-28825	105	39	data	datum	NOUN
ajst-28825	105	40	,	,	PUNCT
ajst-28825	105	41	j	j	PROPN
ajst-28825	105	42	represents	represent	VERB
ajst-28825	105	43	the	the	DET
ajst-28825	105	44	categories	category	NOUN
ajst-28825	105	45	,	,	PUNCT
ajst-28825	105	46	and	and	CCONJ
ajst-28825	105	47	p	p	NOUN
ajst-28825	105	48	represents	represent	VERB
ajst-28825	105	49	the	the	DET
ajst-28825	105	50	predicted	predict	VERB
ajst-28825	105	51	results	result	NOUN
ajst-28825	105	52	.	.	PUNCT
ajst-28825	106	1	this	this	DET
ajst-28825	106	2	loss	loss	NOUN
ajst-28825	106	3	function	function	NOUN
ajst-28825	106	4	is	be	AUX
ajst-28825	106	5	commonly	commonly	ADV
ajst-28825	106	6	used	use	VERB
ajst-28825	106	7	in	in	ADP
ajst-28825	106	8	multi	multi	ADJ
ajst-28825	106	9	-	-	ADJ
ajst-28825	106	10	class	class	ADJ
ajst-28825	106	11	classification	classification	NOUN
ajst-28825	106	12	tasks	task	NOUN
ajst-28825	106	13	,	,	PUNCT
ajst-28825	106	14	such	such	ADJ
ajst-28825	106	15	as	as	ADP
ajst-28825	106	16	when	when	SCONJ
ajst-28825	106	17	using	use	VERB
ajst-28825	106	18	the	the	DET
ajst-28825	106	19	softmax	softmax	NOUN
ajst-28825	106	20	function	function	NOUN
ajst-28825	106	21	as	as	ADP
ajst-28825	106	22	the	the	DET
ajst-28825	106	23	final	final	ADJ
ajst-28825	106	24	output	output	NOUN
ajst-28825	106	25	.	.	PUNCT
ajst-28825	107	1	the	the	DET
ajst-28825	107	2	optimizer	optimizer	NOUN
ajst-28825	107	3	continuously	continuously	ADV
ajst-28825	107	4	reduces	reduce	VERB
ajst-28825	107	5	the	the	DET
ajst-28825	107	6	loss	loss	NOUN
ajst-28825	107	7	function	function	NOUN
ajst-28825	107	8	to	to	PART
ajst-28825	107	9	update	update	VERB
ajst-28825	107	10	the	the	DET
ajst-28825	107	11	parameters	parameter	NOUN
ajst-28825	107	12	of	of	ADP
ajst-28825	107	13	the	the	DET
ajst-28825	107	14	hidden	hidden	ADJ
ajst-28825	107	15	layers	layer	NOUN
ajst-28825	107	16	.	.	PUNCT
ajst-28825	108	1	in	in	ADP
ajst-28825	108	2	this	this	DET
ajst-28825	108	3	case	case	NOUN
ajst-28825	108	4	,	,	PUNCT
ajst-28825	108	5	we	we	PRON
ajst-28825	108	6	selected	select	VERB
ajst-28825	108	7	the	the	DET
ajst-28825	108	8	adam	adam	PROPN
ajst-28825	108	9	optimizer	optimizer	NOUN
ajst-28825	108	10	,	,	PUNCT
ajst-28825	108	11	which	which	PRON
ajst-28825	108	12	is	be	AUX
ajst-28825	108	13	currently	currently	ADV
ajst-28825	108	14	one	one	NUM
ajst-28825	108	15	of	of	ADP
ajst-28825	108	16	the	the	DET
ajst-28825	108	17	most	most	ADV
ajst-28825	108	18	widely	widely	ADV
ajst-28825	108	19	used	use	VERB
ajst-28825	108	20	optimizers[7	optimizers[7	ADP
ajst-28825	108	21	]	]	PUNCT
ajst-28825	108	22	.	.	PUNCT
ajst-28825	109	1	according	accord	VERB
ajst-28825	109	2	to	to	ADP
ajst-28825	109	3	the	the	DET
ajst-28825	109	4	literature[16	literature[16	PROPN
ajst-28825	109	5	]	]	PUNCT
ajst-28825	109	6	,	,	PUNCT
ajst-28825	109	7	this	this	DET
ajst-28825	109	8	optimizer	optimizer	NOUN
ajst-28825	109	9	offers	offer	VERB
ajst-28825	109	10	advantages	advantage	NOUN
ajst-28825	109	11	such	such	ADJ
ajst-28825	109	12	as	as	ADP
ajst-28825	109	13	low	low	ADJ
ajst-28825	109	14	memory	memory	NOUN
ajst-28825	109	15	requirements	requirement	NOUN
ajst-28825	109	16	,	,	PUNCT
ajst-28825	109	17	simple	simple	ADJ
ajst-28825	109	18	implementation	implementation	NOUN
ajst-28825	109	19	,	,	PUNCT
ajst-28825	109	20	and	and	CCONJ
ajst-28825	109	21	high	high	ADJ
ajst-28825	109	22	computational	computational	ADJ
ajst-28825	109	23	efficiency	efficiency	NOUN
ajst-28825	109	24	.	.	PUNCT
ajst-28825	110	1	the	the	DET
ajst-28825	110	2	learning	learning	NOUN
ajst-28825	110	3	rate	rate	NOUN
ajst-28825	110	4	for	for	ADP
ajst-28825	110	5	the	the	DET
ajst-28825	110	6	optimizer	optimizer	NOUN
ajst-28825	110	7	was	be	AUX
ajst-28825	110	8	set	set	VERB
ajst-28825	110	9	to	to	ADP
ajst-28825	110	10	a	a	DET
ajst-28825	110	11	fixed	fix	VERB
ajst-28825	110	12	value	value	NOUN
ajst-28825	110	13	of	of	ADP
ajst-28825	110	14	0.001	0.001	NUM
ajst-28825	110	15	,	,	PUNCT
ajst-28825	110	16	and	and	CCONJ
ajst-28825	110	17	a	a	DET
ajst-28825	110	18	dropout	dropout	NOUN
ajst-28825	110	19	rate	rate	NOUN
ajst-28825	110	20	of	of	ADP
ajst-28825	110	21	0.5	0.5	NUM
ajst-28825	110	22	was	be	AUX
ajst-28825	110	23	used	use	VERB
ajst-28825	110	24	to	to	PART
ajst-28825	110	25	randomly	randomly	VERB
ajst-28825	110	26	deactivate	deactivate	VERB
ajst-28825	110	27	neurons	neuron	NOUN
ajst-28825	110	28	,	,	PUNCT
ajst-28825	110	29	preventing	prevent	VERB
ajst-28825	110	30	overfitting	overfitte	VERB
ajst-28825	110	31	due	due	ADP
ajst-28825	110	32	to	to	ADP
ajst-28825	110	33	the	the	DET
ajst-28825	110	34	large	large	ADJ
ajst-28825	110	35	number	number	NOUN
ajst-28825	110	36	of	of	ADP
ajst-28825	110	37	parameters	parameter	NOUN
ajst-28825	110	38	in	in	ADP
ajst-28825	110	39	the	the	DET
ajst-28825	110	40	fully	fully	ADV
ajst-28825	110	41	connected	connect	VERB
ajst-28825	110	42	layer[14	layer[14	PROPN
ajst-28825	110	43	]	]	PUNCT
ajst-28825	110	44	.	.	PUNCT
ajst-28825	111	1	the	the	DET
ajst-28825	111	2	confusion	confusion	NOUN
ajst-28825	111	3	matrix	matrix	NOUN
ajst-28825	111	4	in	in	ADP
ajst-28825	111	5	figure	figure	NOUN
ajst-28825	111	6	3	3	NUM
ajst-28825	111	7	shows	show	VERB
ajst-28825	111	8	the	the	DET
ajst-28825	111	9	classification	classification	NOUN
ajst-28825	111	10	performance	performance	NOUN
ajst-28825	111	11	at	at	ADP
ajst-28825	111	12	a	a	DET
ajst-28825	111	13	signal	signal	NOUN
ajst-28825	111	14	-	-	PUNCT
ajst-28825	111	15	to	to	ADP
ajst-28825	111	16	-	-	PUNCT
ajst-28825	111	17	noise	noise	NOUN
ajst-28825	111	18	ratio	ratio	NOUN
ajst-28825	111	19	(	(	PUNCT
ajst-28825	111	20	snr	snr	PROPN
ajst-28825	111	21	)	)	PUNCT
ajst-28825	111	22	of	of	ADP
ajst-28825	111	23	30db	30db	NOUN
ajst-28825	111	24	.	.	PUNCT
ajst-28825	112	1	the	the	DET
ajst-28825	112	2	horizontal	horizontal	ADJ
ajst-28825	112	3	axis	axis	NOUN
ajst-28825	112	4	represents	represent	VERB
ajst-28825	112	5	the	the	DET
ajst-28825	112	6	nine	nine	NUM
ajst-28825	112	7	predicted	predict	VERB
ajst-28825	112	8	modulation	modulation	NOUN
ajst-28825	112	9	categories	category	NOUN
ajst-28825	112	10	,	,	PUNCT
ajst-28825	112	11	while	while	SCONJ
ajst-28825	112	12	the	the	DET
ajst-28825	112	13	vertical	vertical	ADJ
ajst-28825	112	14	axis	axis	NOUN
ajst-28825	112	15	represents	represent	VERB
ajst-28825	112	16	the	the	DET
ajst-28825	112	17	actual	actual	ADJ
ajst-28825	112	18	modulation	modulation	NOUN
ajst-28825	112	19	categories	category	NOUN
ajst-28825	112	20	.	.	PUNCT
ajst-28825	113	1	this	this	DET
ajst-28825	113	2	confusion	confusion	NOUN
ajst-28825	113	3	matrix	matrix	NOUN
ajst-28825	113	4	provides	provide	VERB
ajst-28825	113	5	a	a	DET
ajst-28825	113	6	visual	visual	ADJ
ajst-28825	113	7	representation	representation	NOUN
ajst-28825	113	8	of	of	ADP
ajst-28825	113	9	how	how	SCONJ
ajst-28825	113	10	well	well	ADV
ajst-28825	113	11	the	the	DET
ajst-28825	113	12	model	model	NOUN
ajst-28825	113	13	correctly	correctly	ADV
ajst-28825	113	14	classifies	classify	VERB
ajst-28825	113	15	each	each	DET
ajst-28825	113	16	modulation	modulation	NOUN
ajst-28825	113	17	type	type	VERB
ajst-28825	113	18	under	under	ADP
ajst-28825	113	19	the	the	DET
ajst-28825	113	20	given	give	VERB
ajst-28825	113	21	conditions[8	conditions[8	NOUN
ajst-28825	113	22	]	]	X
ajst-28825	113	23	.	.	PUNCT
ajst-28825	114	1	this	this	DET
ajst-28825	114	2	process	process	NOUN
ajst-28825	114	3	,	,	PUNCT
ajst-28825	114	4	combining	combine	VERB
ajst-28825	114	5	the	the	DET
ajst-28825	114	6	loss	loss	NOUN
ajst-28825	114	7	function	function	NOUN
ajst-28825	114	8	,	,	PUNCT
ajst-28825	114	9	adam	adam	PROPN
ajst-28825	114	10	optimizer	optimizer	NOUN
ajst-28825	114	11	,	,	PUNCT
ajst-28825	114	12	and	and	CCONJ
ajst-28825	114	13	dropout	dropout	NOUN
ajst-28825	114	14	techniques	technique	NOUN
ajst-28825	114	15	,	,	PUNCT
ajst-28825	114	16	helped	help	VERB
ajst-28825	114	17	achieve	achieve	VERB
ajst-28825	114	18	better	well	ADJ
ajst-28825	114	19	generalization	generalization	NOUN
ajst-28825	114	20	and	and	CCONJ
ajst-28825	114	21	avoided	avoid	VERB
ajst-28825	114	22	overfitting	overfitting	NOUN
ajst-28825	114	23	during	during	ADP
ajst-28825	114	24	training	training	NOUN
ajst-28825	114	25	.	.	PUNCT
ajst-28825	115	1	figure	figure	NOUN
ajst-28825	115	2	3	3	NUM
ajst-28825	115	3	.	.	PUNCT
ajst-28825	116	1	the	the	DET
ajst-28825	116	2	confusion	confusion	NOUN
ajst-28825	116	3	matrix	matrix	NOUN
ajst-28825	116	4	at	at	ADP
ajst-28825	116	5	a	a	DET
ajst-28825	116	6	signal	signal	NOUN
ajst-28825	116	7	-	-	PUNCT
ajst-28825	116	8	to	to	ADP
ajst-28825	116	9	-	-	PUNCT
ajst-28825	116	10	noise	noise	NOUN
ajst-28825	116	11	ratio	ratio	NOUN
ajst-28825	116	12	(	(	PUNCT
ajst-28825	116	13	snr	snr	PROPN
ajst-28825	116	14	)	)	PUNCT
ajst-28825	116	15	of	of	ADP
ajst-28825	116	16	30	30	NUM
ajst-28825	116	17	db	db	NOUN
ajst-28825	116	18	.	.	PROPN
ajst-28825	116	19	4.2	4.2	NUM
ajst-28825	116	20	.	.	PUNCT
ajst-28825	116	21	results	result	VERB
ajst-28825	116	22	analysis	analysis	NOUN
ajst-28825	116	23	the	the	DET
ajst-28825	116	24	neural	neural	ADJ
ajst-28825	116	25	network	network	NOUN
ajst-28825	116	26	was	be	AUX
ajst-28825	116	27	trained	train	VERB
ajst-28825	116	28	for	for	ADP
ajst-28825	116	29	100	100	NUM
ajst-28825	116	30	epochs	epoch	NOUN
ajst-28825	116	31	using	use	VERB
ajst-28825	116	32	the	the	DET
ajst-28825	116	33	training	training	NOUN
ajst-28825	116	34	data	datum	NOUN
ajst-28825	116	35	,	,	PUNCT
ajst-28825	116	36	while	while	SCONJ
ajst-28825	116	37	the	the	DET
ajst-28825	116	38	test	test	NOUN
ajst-28825	116	39	data	datum	NOUN
ajst-28825	116	40	was	be	AUX
ajst-28825	116	41	used	use	VERB
ajst-28825	116	42	to	to	PART
ajst-28825	116	43	evaluate	evaluate	VERB
ajst-28825	116	44	the	the	DET
ajst-28825	116	45	model	model	NOUN
ajst-28825	116	46	after	after	ADP
ajst-28825	116	47	each	each	DET
ajst-28825	116	48	epoch	epoch	NOUN
ajst-28825	116	49	.	.	PUNCT
ajst-28825	117	1	figure	figure	VERB
ajst-28825	117	2	4	4	NUM
ajst-28825	117	3	shows	show	VERB
ajst-28825	117	4	the	the	DET
ajst-28825	117	5	loss	loss	NOUN
ajst-28825	117	6	reduction	reduction	NOUN
ajst-28825	117	7	curve	curve	NOUN
ajst-28825	117	8	,	,	PUNCT
ajst-28825	117	9	where	where	SCONJ
ajst-28825	117	10	the	the	DET
ajst-28825	117	11	loss	loss	NOUN
ajst-28825	117	12	decreases	decrease	VERB
ajst-28825	117	13	to	to	ADP
ajst-28825	117	14	around	around	ADP
ajst-28825	117	15	0.2	0.2	NUM
ajst-28825	117	16	at	at	ADP
ajst-28825	117	17	its	its	PRON
ajst-28825	117	18	lowest	low	ADJ
ajst-28825	117	19	point	point	NOUN
ajst-28825	117	20	.	.	PUNCT
ajst-28825	118	1	figure	figure	NOUN
ajst-28825	118	2	5	5	NUM
ajst-28825	118	3	illustrates	illustrate	VERB
ajst-28825	118	4	the	the	DET
ajst-28825	118	5	training	training	NOUN
ajst-28825	118	6	accuracy	accuracy	NOUN
ajst-28825	118	7	and	and	CCONJ
ajst-28825	118	8	test	test	NOUN
ajst-28825	118	9	accuracy	accuracy	NOUN
ajst-28825	118	10	of	of	ADP
ajst-28825	118	11	the	the	DET
ajst-28825	118	12	model	model	NOUN
ajst-28825	118	13	.	.	PUNCT
ajst-28825	119	1	the	the	DET
ajst-28825	119	2	training	training	NOUN
ajst-28825	119	3	accuracy	accuracy	NOUN
ajst-28825	119	4	of	of	ADP
ajst-28825	119	5	the	the	DET
ajst-28825	119	6	deep	deep	ADJ
ajst-28825	119	7	neural	neural	ADJ
ajst-28825	119	8	network	network	NOUN
ajst-28825	119	9	reaches	reach	VERB
ajst-28825	119	10	a	a	DET
ajst-28825	119	11	maximum	maximum	NOUN
ajst-28825	119	12	of	of	ADP
ajst-28825	119	13	71.7	71.7	NUM
ajst-28825	119	14	%	%	NOUN
ajst-28825	119	15	,	,	PUNCT
ajst-28825	119	16	and	and	CCONJ
ajst-28825	119	17	the	the	DET
ajst-28825	119	18	test	test	NOUN
ajst-28825	119	19	accuracy	accuracy	NOUN
ajst-28825	119	20	reaches	reach	VERB
ajst-28825	119	21	a	a	DET
ajst-28825	119	22	maximum	maximum	NOUN
ajst-28825	119	23	of	of	ADP
ajst-28825	119	24	72	72	NUM
ajst-28825	119	25	%	%	NOUN
ajst-28825	119	26	.	.	PUNCT
ajst-28825	120	1	the	the	DET
ajst-28825	120	2	test	test	NOUN
ajst-28825	120	3	accuracy	accuracy	NOUN
ajst-28825	120	4	increases	increase	VERB
ajst-28825	120	5	alongside	alongside	ADP
ajst-28825	120	6	the	the	DET
ajst-28825	120	7	training	training	NOUN
ajst-28825	120	8	accuracy	accuracy	NOUN
ajst-28825	120	9	until	until	SCONJ
ajst-28825	120	10	both	both	DET
ajst-28825	120	11	curves	curve	NOUN
ajst-28825	120	12	stabilize	stabilize	VERB
ajst-28825	120	13	,	,	PUNCT
ajst-28825	120	14	with	with	ADP
ajst-28825	120	15	no	no	DET
ajst-28825	120	16	signs	sign	NOUN
ajst-28825	120	17	of	of	ADP
ajst-28825	120	18	extreme	extreme	ADJ
ajst-28825	120	19	divergence	divergence	NOUN
ajst-28825	120	20	.	.	PUNCT
ajst-28825	121	1	there	there	PRON
ajst-28825	121	2	were	be	VERB
ajst-28825	121	3	no	no	DET
ajst-28825	121	4	issues	issue	NOUN
ajst-28825	121	5	with	with	ADP
ajst-28825	121	6	overfitting	overfitte	VERB
ajst-28825	121	7	or	or	CCONJ
ajst-28825	121	8	underfitting	underfitting	NOUN
ajst-28825	121	9	,	,	PUNCT
ajst-28825	121	10	and	and	CCONJ
ajst-28825	121	11	the	the	DET
ajst-28825	121	12	neural	neural	ADJ
ajst-28825	121	13	network	network	NOUN
ajst-28825	121	14	successfully	successfully	ADV
ajst-28825	121	15	learned	learn	VERB
ajst-28825	121	16	the	the	DET
ajst-28825	121	17	features	feature	NOUN
ajst-28825	121	18	of	of	ADP
ajst-28825	121	19	the	the	DET
ajst-28825	121	20	training	training	NOUN
ajst-28825	121	21	data	datum	NOUN
ajst-28825	121	22	.	.	PUNCT
ajst-28825	122	1	the	the	DET
ajst-28825	122	2	test	test	NOUN
ajst-28825	122	3	results	result	NOUN
ajst-28825	122	4	indicate	indicate	VERB
ajst-28825	122	5	that	that	SCONJ
ajst-28825	122	6	the	the	DET
ajst-28825	122	7	trained	train	VERB
ajst-28825	122	8	model	model	NOUN
ajst-28825	122	9	exhibits	exhibit	VERB
ajst-28825	122	10	strong	strong	ADJ
ajst-28825	122	11	generalization	generalization	NOUN
ajst-28825	122	12	capabilities	capability	NOUN
ajst-28825	122	13	.	.	PUNCT
ajst-28825	123	1	figure	figure	VERB
ajst-28825	123	2	4	4	NUM
ajst-28825	123	3	.	.	PUNCT
ajst-28825	124	1	the	the	DET
ajst-28825	124	2	variation	variation	NOUN
ajst-28825	124	3	curve	curve	NOUN
ajst-28825	124	4	of	of	ADP
ajst-28825	124	5	the	the	DET
ajst-28825	124	6	classification	classification	NOUN
ajst-28825	124	7	loss	loss	NOUN
ajst-28825	124	8	.	.	PUNCT
ajst-28825	125	1	347	347	NUM
ajst-28825	125	2	figure	figure	NOUN
ajst-28825	125	3	5	5	NUM
ajst-28825	125	4	.	.	PUNCT
ajst-28825	126	1	the	the	DET
ajst-28825	126	2	accuracy	accuracy	NOUN
ajst-28825	126	3	of	of	ADP
ajst-28825	126	4	the	the	DET
ajst-28825	126	5	model	model	NOUN
ajst-28825	126	6	during	during	ADP
ajst-28825	126	7	training	training	NOUN
ajst-28825	126	8	.	.	PUNCT
ajst-28825	127	1	5	5	X
ajst-28825	127	2	.	.	X
ajst-28825	127	3	conclusions	conclusion	NOUN
ajst-28825	127	4	this	this	DET
ajst-28825	127	5	paper	paper	NOUN
ajst-28825	127	6	conducted	conduct	VERB
ajst-28825	127	7	extensive	extensive	ADJ
ajst-28825	127	8	experiments	experiment	NOUN
ajst-28825	127	9	to	to	PART
ajst-28825	127	10	test	test	VERB
ajst-28825	127	11	various	various	ADJ
ajst-28825	127	12	hyperparameters	hyperparameter	NOUN
ajst-28825	127	13	,	,	PUNCT
ajst-28825	127	14	and	and	CCONJ
ajst-28825	127	15	while	while	SCONJ
ajst-28825	127	16	continuously	continuously	ADV
ajst-28825	127	17	optimizing	optimize	VERB
ajst-28825	127	18	the	the	DET
ajst-28825	127	19	network	network	NOUN
ajst-28825	127	20	structure	structure	NOUN
ajst-28825	127	21	,	,	PUNCT
ajst-28825	127	22	a	a	DET
ajst-28825	127	23	new	new	ADJ
ajst-28825	127	24	deep	deep	ADJ
ajst-28825	127	25	neural	neural	ADJ
ajst-28825	127	26	network	network	NOUN
ajst-28825	127	27	model	model	NOUN
ajst-28825	127	28	was	be	AUX
ajst-28825	127	29	redesigned	redesign	VERB
ajst-28825	127	30	.	.	PUNCT
ajst-28825	128	1	the	the	DET
ajst-28825	128	2	experiments	experiment	NOUN
ajst-28825	128	3	have	have	AUX
ajst-28825	128	4	demonstrated	demonstrate	VERB
ajst-28825	128	5	the	the	DET
ajst-28825	128	6	effectiveness	effectiveness	NOUN
ajst-28825	128	7	of	of	ADP
ajst-28825	128	8	this	this	DET
ajst-28825	128	9	approach	approach	NOUN
ajst-28825	128	10	.	.	PUNCT
ajst-28825	129	1	compared	compare	VERB
ajst-28825	129	2	to	to	ADP
ajst-28825	129	3	traditional	traditional	ADJ
ajst-28825	129	4	methods	method	NOUN
ajst-28825	129	5	,	,	PUNCT
ajst-28825	129	6	there	there	PRON
ajst-28825	129	7	has	have	AUX
ajst-28825	129	8	been	be	AUX
ajst-28825	129	9	a	a	DET
ajst-28825	129	10	significant	significant	ADJ
ajst-28825	129	11	improvement	improvement	NOUN
ajst-28825	129	12	in	in	ADP
ajst-28825	129	13	recognizing	recognize	VERB
ajst-28825	129	14	various	various	ADJ
ajst-28825	129	15	modulation	modulation	NOUN
ajst-28825	129	16	types	type	NOUN
ajst-28825	129	17	.	.	PUNCT
ajst-28825	130	1	most	most	ADV
ajst-28825	130	2	importantly	importantly	ADV
ajst-28825	130	3	,	,	PUNCT
ajst-28825	130	4	this	this	DET
ajst-28825	130	5	approach	approach	NOUN
ajst-28825	130	6	solves	solve	VERB
ajst-28825	130	7	the	the	DET
ajst-28825	130	8	problem	problem	NOUN
ajst-28825	130	9	of	of	ADP
ajst-28825	130	10	end	end	NOUN
ajst-28825	130	11	-	-	PUNCT
ajst-28825	130	12	to	to	ADP
ajst-28825	130	13	-	-	PUNCT
ajst-28825	130	14	end	end	NOUN
ajst-28825	130	15	signal	signal	NOUN
ajst-28825	130	16	recognition	recognition	NOUN
ajst-28825	130	17	and	and	CCONJ
ajst-28825	130	18	eliminates	eliminate	VERB
ajst-28825	130	19	the	the	DET
ajst-28825	130	20	cumbersome	cumbersome	ADJ
ajst-28825	130	21	process	process	NOUN
ajst-28825	130	22	of	of	ADP
ajst-28825	130	23	manual	manual	ADJ
ajst-28825	130	24	feature	feature	NOUN
ajst-28825	130	25	extraction	extraction	NOUN
ajst-28825	130	26	.	.	PUNCT
ajst-28825	131	1	the	the	DET
ajst-28825	131	2	test	test	NOUN
ajst-28825	131	3	accuracy	accuracy	NOUN
ajst-28825	131	4	is	be	AUX
ajst-28825	131	5	very	very	ADV
ajst-28825	131	6	high	high	ADJ
ajst-28825	131	7	,	,	PUNCT
ajst-28825	131	8	and	and	CCONJ
ajst-28825	131	9	the	the	DET
ajst-28825	131	10	model	model	NOUN
ajst-28825	131	11	exhibits	exhibit	VERB
ajst-28825	131	12	good	good	ADJ
ajst-28825	131	13	generalization	generalization	NOUN
ajst-28825	131	14	capability	capability	NOUN
ajst-28825	131	15	.	.	PUNCT
ajst-28825	132	1	it	it	PRON
ajst-28825	132	2	is	be	AUX
ajst-28825	132	3	believed	believe	VERB
ajst-28825	132	4	that	that	SCONJ
ajst-28825	132	5	with	with	ADP
ajst-28825	132	6	continued	continued	ADJ
ajst-28825	132	7	research	research	NOUN
ajst-28825	132	8	by	by	ADP
ajst-28825	132	9	more	more	ADJ
ajst-28825	132	10	scholars	scholar	NOUN
ajst-28825	132	11	and	and	CCONJ
ajst-28825	132	12	the	the	DET
ajst-28825	132	13	ongoing	ongoing	ADJ
ajst-28825	132	14	development	development	NOUN
ajst-28825	132	15	of	of	ADP
ajst-28825	132	16	deep	deep	ADJ
ajst-28825	132	17	learning	learning	NOUN
ajst-28825	132	18	,	,	PUNCT
ajst-28825	132	19	there	there	PRON
ajst-28825	132	20	will	will	AUX
ajst-28825	132	21	undoubtedly	undoubtedly	ADV
ajst-28825	132	22	be	be	AUX
ajst-28825	132	23	further	further	ADJ
ajst-28825	132	24	breakthroughs	breakthrough	NOUN
ajst-28825	132	25	in	in	ADP
ajst-28825	132	26	modulation	modulation	NOUN
ajst-28825	132	27	recognition	recognition	NOUN
ajst-28825	132	28	methods	method	NOUN
ajst-28825	132	29	for	for	ADP
ajst-28825	132	30	signal	signal	NOUN
ajst-28825	132	31	and	and	CCONJ
ajst-28825	132	32	information	information	NOUN
ajst-28825	132	33	processing	processing	NOUN
ajst-28825	132	34	,	,	PUNCT
ajst-28825	132	35	and	and	CCONJ
ajst-28825	132	36	accuracy	accuracy	NOUN
ajst-28825	132	37	will	will	AUX
ajst-28825	132	38	continue	continue	VERB
ajst-28825	132	39	to	to	PART
ajst-28825	132	40	improve	improve	VERB
ajst-28825	132	41	.	.	PUNCT
ajst-28825	133	1	however	however	ADV
ajst-28825	133	2	,	,	PUNCT
ajst-28825	133	3	there	there	PRON
ajst-28825	133	4	are	be	VERB
ajst-28825	133	5	still	still	ADV
ajst-28825	133	6	many	many	ADJ
ajst-28825	133	7	areas	area	NOUN
ajst-28825	133	8	for	for	ADP
ajst-28825	133	9	improvement	improvement	NOUN
ajst-28825	133	10	in	in	ADP
ajst-28825	133	11	this	this	DET
ajst-28825	133	12	experiment	experiment	NOUN
ajst-28825	133	13	,	,	PUNCT
ajst-28825	133	14	such	such	ADJ
ajst-28825	133	15	as	as	ADP
ajst-28825	133	16	the	the	DET
ajst-28825	133	17	limited	limited	ADJ
ajst-28825	133	18	amount	amount	NOUN
ajst-28825	133	19	of	of	ADP
ajst-28825	133	20	data	datum	NOUN
ajst-28825	133	21	,	,	PUNCT
ajst-28825	133	22	as	as	ADV
ajst-28825	133	23	well	well	ADV
ajst-28825	133	24	as	as	ADP
ajst-28825	133	25	room	room	NOUN
ajst-28825	133	26	for	for	ADP
ajst-28825	133	27	further	further	ADJ
ajst-28825	133	28	enhancement	enhancement	NOUN
ajst-28825	133	29	in	in	ADP
ajst-28825	133	30	network	network	NOUN
ajst-28825	133	31	structure	structure	NOUN
ajst-28825	133	32	and	and	CCONJ
ajst-28825	133	33	hyperparameters	hyperparameter	NOUN
ajst-28825	133	34	.	.	PUNCT
ajst-28825	134	1	future	future	ADJ
ajst-28825	134	2	work	work	NOUN
ajst-28825	134	3	will	will	AUX
ajst-28825	134	4	involve	involve	VERB
ajst-28825	134	5	continuous	continuous	ADJ
ajst-28825	134	6	attempts	attempt	NOUN
ajst-28825	134	7	and	and	CCONJ
ajst-28825	134	8	refinements	refinement	NOUN
ajst-28825	134	9	.	.	PUNCT
ajst-28825	135	1	references	reference	NOUN
ajst-28825	135	2	[	[	X
ajst-28825	135	3	1	1	NUM
ajst-28825	135	4	]	]	X
ajst-28825	135	5	maganioti	maganioti	NOUN
ajst-28825	135	6	,	,	PUNCT
ajst-28825	135	7	a.e	a.e	PROPN
ajst-28825	135	8	.	.	PROPN
ajst-28825	135	9	,	,	PUNCT
ajst-28825	135	10	chrissanthi	chrissanthi	PROPN
ajst-28825	135	11	,	,	PUNCT
ajst-28825	135	12	h.d	h.d	PROPN
ajst-28825	135	13	.	.	PROPN
ajst-28825	135	14	,	,	PUNCT
ajst-28825	135	15	charalabos	charalabos	PROPN
ajst-28825	135	16	,	,	PUNCT
ajst-28825	135	17	p.c	p.c	PROPN
ajst-28825	135	18	.	.	PROPN
ajst-28825	135	19	,	,	PUNCT
ajst-28825	135	20	andreas	andreas	PROPN
ajst-28825	135	21	,	,	PUNCT
ajst-28825	135	22	r.d	r.d	PROPN
ajst-28825	135	23	.	.	PROPN
ajst-28825	135	24	,	,	PUNCT
ajst-28825	135	25	george	george	PROPN
ajst-28825	135	26	,	,	PUNCT
ajst-28825	135	27	p.n	p.n	PROPN
ajst-28825	135	28	.	.	PROPN
ajst-28825	135	29	and	and	CCONJ
ajst-28825	135	30	christos	christos	PROPN
ajst-28825	135	31	,	,	PUNCT
ajst-28825	135	32	c.n	c.n	PROPN
ajst-28825	135	33	.	.	PROPN
ajst-28825	135	34	cointegration	cointegration	NOUN
ajst-28825	135	35	of	of	ADP
ajst-28825	135	36	eventrelated	eventrelated	ADJ
ajst-28825	135	37	potential	potential	ADJ
ajst-28825	135	38	(	(	PUNCT
ajst-28825	135	39	erp	erp	NOUN
ajst-28825	135	40	)	)	PUNCT
ajst-28825	135	41	signals	signal	NOUN
ajst-28825	135	42	in	in	ADP
ajst-28825	135	43	experiments	experiment	NOUN
ajst-28825	135	44	with	with	ADP
ajst-28825	135	45	different	different	ADJ
ajst-28825	135	46	electromagnetic	electromagnetic	ADJ
ajst-28825	135	47	field	field	NOUN
ajst-28825	135	48	(	(	PUNCT
ajst-28825	135	49	emf	emf	NOUN
ajst-28825	135	50	)	)	PUNCT
ajst-28825	135	51	conditions	condition	NOUN
ajst-28825	135	52	.	.	PUNCT
ajst-28825	136	1	health	health	NOUN
ajst-28825	136	2	,	,	PUNCT
ajst-28825	136	3	(	(	PUNCT
ajst-28825	136	4	2010	2010	NUM
ajst-28825	136	5	)	)	PUNCT
ajst-28825	136	6	2	2	NUM
ajst-28825	136	7	,	,	PUNCT
ajst-28825	136	8	400	400	NUM
ajst-28825	136	9	-	-	SYM
ajst-28825	136	10	406	406	NUM
ajst-28825	136	11	.	.	PUNCT
ajst-28825	137	1	[	[	X
ajst-28825	137	2	2	2	NUM
ajst-28825	137	3	]	]	PUNCT
ajst-28825	137	4	bootorabi	bootorabi	NOUN
ajst-28825	137	5	,	,	PUNCT
ajst-28825	137	6	f.	f.	PROPN
ajst-28825	137	7	,	,	PUNCT
ajst-28825	137	8	haapasalo	haapasalo	PROPN
ajst-28825	137	9	,	,	PUNCT
ajst-28825	137	10	j.	j.	PROPN
ajst-28825	137	11	,	,	PUNCT
ajst-28825	137	12	smith	smith	PROPN
ajst-28825	137	13	,	,	PUNCT
ajst-28825	137	14	e.	e.	PROPN
ajst-28825	137	15	,	,	PUNCT
ajst-28825	137	16	haapasalo	haapasalo	PROPN
ajst-28825	137	17	,	,	PUNCT
ajst-28825	137	18	h.	h.	PROPN
ajst-28825	137	19	and	and	CCONJ
ajst-28825	137	20	parkkila	parkkila	PROPN
ajst-28825	137	21	,	,	PUNCT
ajst-28825	137	22	s.	s.	PROPN
ajst-28825	137	23	carbonic	carbonic	PROPN
ajst-28825	137	24	anhydrase	anhydrase	NOUN
ajst-28825	137	25	vii	vii	PROPN
ajst-28825	137	26	—	—	PUNCT
ajst-28825	137	27	a	a	DET
ajst-28825	137	28	potential	potential	ADJ
ajst-28825	137	29	prognostic	prognostic	ADJ
ajst-28825	137	30	marker	marker	NOUN
ajst-28825	137	31	in	in	ADP
ajst-28825	137	32	gliomas	glioma	NOUN
ajst-28825	137	33	.	.	PUNCT
ajst-28825	138	1	health	health	NOUN
ajst-28825	138	2	,	,	PUNCT
ajst-28825	138	3	(	(	PUNCT
ajst-28825	138	4	2011	2011	NUM
ajst-28825	138	5	)	)	PUNCT
ajst-28825	138	6	3	3	NUM
ajst-28825	138	7	,	,	PUNCT
ajst-28825	138	8	6	6	NUM
ajst-28825	138	9	-	-	SYM
ajst-28825	138	10	12	12	NUM
ajst-28825	138	11	.	.	PUNCT
ajst-28825	139	1	[	[	X
ajst-28825	139	2	3	3	NUM
ajst-28825	139	3	]	]	X
ajst-28825	139	4	shafik	shafik	NOUN
ajst-28825	139	5	,	,	PUNCT
ajst-28825	139	6	r.	r.	PROPN
ajst-28825	139	7	a.	a.	PROPN
ajst-28825	139	8	,	,	PUNCT
ajst-28825	139	9	rahman	rahman	PROPN
ajst-28825	139	10	,	,	PUNCT
ajst-28825	139	11	s.	s.	PROPN
ajst-28825	139	12	,	,	PUNCT
ajst-28825	139	13	&	&	CCONJ
ajst-28825	139	14	supangkat	supangkat	PROPN
ajst-28825	139	15	,	,	PUNCT
ajst-28825	139	16	s.	s.	PROPN
ajst-28825	139	17	h.	h.	PROPN
ajst-28825	139	18	automatic	automatic	ADJ
ajst-28825	139	19	modulation	modulation	NOUN
ajst-28825	139	20	classification	classification	NOUN
ajst-28825	139	21	for	for	ADP
ajst-28825	139	22	cognitive	cognitive	ADJ
ajst-28825	139	23	radios	radio	NOUN
ajst-28825	139	24	using	use	VERB
ajst-28825	139	25	cyclostationary	cyclostationary	ADJ
ajst-28825	139	26	features	feature	NOUN
ajst-28825	139	27	.	.	PUNCT
ajst-28825	140	1	ieee	ieee	NOUN
ajst-28825	140	2	communications	communication	NOUN
ajst-28825	140	3	surveys	survey	NOUN
ajst-28825	140	4	&	&	CCONJ
ajst-28825	140	5	tutorials	tutorial	NOUN
ajst-28825	140	6	,	,	PUNCT
ajst-28825	140	7	(	(	PUNCT
ajst-28825	140	8	2012	2012	NUM
ajst-28825	140	9	)	)	PUNCT
ajst-28825	140	10	14(1	14(1	NUM
ajst-28825	140	11	)	)	PUNCT
ajst-28825	140	12	,	,	PUNCT
ajst-28825	140	13	105	105	NUM
ajst-28825	140	14	-	-	SYM
ajst-28825	140	15	117	117	NUM
ajst-28825	140	16	.	.	PUNCT
ajst-28825	141	1	[	[	X
ajst-28825	141	2	4	4	NUM
ajst-28825	141	3	]	]	X
ajst-28825	141	4	o'shea	o'shea	PROPN
ajst-28825	141	5	,	,	PUNCT
ajst-28825	141	6	t.	t.	PROPN
ajst-28825	141	7	j.	j.	PROPN
ajst-28825	141	8	,	,	PUNCT
ajst-28825	141	9	&	&	CCONJ
ajst-28825	141	10	west	west	PROPN
ajst-28825	141	11	,	,	PUNCT
ajst-28825	141	12	n.	n.	NOUN
ajst-28825	141	13	radio	radio	NOUN
ajst-28825	141	14	machine	machine	NOUN
ajst-28825	141	15	learning	learn	VERB
ajst-28825	141	16	dataset	dataset	ADJ
ajst-28825	141	17	generation	generation	NOUN
ajst-28825	141	18	with	with	ADP
ajst-28825	141	19	gnu	gnu	NOUN
ajst-28825	141	20	radio	radio	NOUN
ajst-28825	141	21	.	.	PUNCT
ajst-28825	142	1	proceedings	proceeding	NOUN
ajst-28825	142	2	of	of	ADP
ajst-28825	142	3	the	the	DET
ajst-28825	142	4	gnu	gnu	PROPN
ajst-28825	142	5	radio	radio	NOUN
ajst-28825	142	6	conference	conference	NOUN
ajst-28825	142	7	,	,	PUNCT
ajst-28825	142	8	(	(	PUNCT
ajst-28825	142	9	2017	2017	NUM
ajst-28825	142	10	)	)	PUNCT
ajst-28825	142	11	1(1	1(1	NUM
ajst-28825	142	12	)	)	PUNCT
ajst-28825	142	13	,	,	PUNCT
ajst-28825	142	14	1	1	NUM
ajst-28825	142	15	-	-	SYM
ajst-28825	142	16	9	9	NUM
ajst-28825	142	17	.	.	PUNCT
ajst-28825	143	1	[	[	X
ajst-28825	143	2	5	5	NUM
ajst-28825	143	3	]	]	X
ajst-28825	143	4	carion	carion	NOUN
ajst-28825	143	5	,	,	PUNCT
ajst-28825	143	6	n.	n.	NOUN
ajst-28825	143	7	,	,	PUNCT
ajst-28825	143	8	massa	massa	PROPN
ajst-28825	143	9	,	,	PUNCT
ajst-28825	143	10	f.	f.	PROPN
ajst-28825	143	11	,	,	PUNCT
ajst-28825	143	12	synnaeve	synnaeve	NOUN
ajst-28825	143	13	,	,	PUNCT
ajst-28825	143	14	g.	g.	PROPN
ajst-28825	143	15	,	,	PUNCT
ajst-28825	143	16	usunier	usunier	PROPN
ajst-28825	143	17	,	,	PUNCT
ajst-28825	143	18	n.	n.	NOUN
ajst-28825	143	19	,	,	PUNCT
ajst-28825	143	20	kirillov	kirillov	PROPN
ajst-28825	143	21	,	,	PUNCT
ajst-28825	143	22	a.	a.	NOUN
ajst-28825	143	23	,	,	PUNCT
ajst-28825	143	24	&	&	CCONJ
ajst-28825	143	25	zagoruyko	zagoruyko	PROPN
ajst-28825	143	26	,	,	PUNCT
ajst-28825	143	27	s.	s.	PROPN
ajst-28825	143	28	end	end	PROPN
ajst-28825	143	29	-	-	PUNCT
ajst-28825	143	30	to	to	ADP
ajst-28825	143	31	-	-	PUNCT
ajst-28825	143	32	end	end	NOUN
ajst-28825	143	33	object	object	NOUN
ajst-28825	143	34	detection	detection	NOUN
ajst-28825	143	35	with	with	ADP
ajst-28825	143	36	transformers	transformer	NOUN
ajst-28825	143	37	.	.	PUNCT
ajst-28825	144	1	proceedings	proceeding	NOUN
ajst-28825	144	2	of	of	ADP
ajst-28825	144	3	the	the	DET
ajst-28825	144	4	european	european	PROPN
ajst-28825	144	5	conference	conference	PROPN
ajst-28825	144	6	on	on	ADP
ajst-28825	144	7	computer	computer	NOUN
ajst-28825	144	8	vision	vision	NOUN
ajst-28825	144	9	,	,	PUNCT
ajst-28825	144	10	(	(	PUNCT
ajst-28825	144	11	2020	2020	NUM
ajst-28825	144	12	)	)	PUNCT
ajst-28825	144	13	213	213	NUM
ajst-28825	144	14	-	-	SYM
ajst-28825	144	15	229	229	NUM
ajst-28825	144	16	.	.	PUNCT
ajst-28825	145	1	[	[	X
ajst-28825	145	2	6	6	NUM
ajst-28825	145	3	]	]	SYM
ajst-28825	145	4	hu	hu	PROPN
ajst-28825	145	5	,	,	PUNCT
ajst-28825	145	6	j.	j.	PROPN
ajst-28825	145	7	,	,	PUNCT
ajst-28825	145	8	shen	shen	PROPN
ajst-28825	145	9	,	,	PUNCT
ajst-28825	145	10	l.	l.	PROPN
ajst-28825	145	11	,	,	PUNCT
ajst-28825	145	12	&	&	CCONJ
ajst-28825	145	13	sun	sun	PROPN
ajst-28825	145	14	,	,	PUNCT
ajst-28825	145	15	g.	g.	PROPN
ajst-28825	145	16	squeeze	squeeze	NOUN
ajst-28825	145	17	-	-	PUNCT
ajst-28825	145	18	and	and	CCONJ
ajst-28825	145	19	-	-	PUNCT
ajst-28825	145	20	excitation	excitation	NOUN
ajst-28825	145	21	networks	network	NOUN
ajst-28825	145	22	.	.	PUNCT
ajst-28825	146	1	proceedings	proceeding	NOUN
ajst-28825	146	2	of	of	ADP
ajst-28825	146	3	the	the	DET
ajst-28825	146	4	ieee	ieee	NOUN
ajst-28825	146	5	conference	conference	NOUN
ajst-28825	146	6	on	on	ADP
ajst-28825	146	7	computer	computer	NOUN
ajst-28825	146	8	vision	vision	NOUN
ajst-28825	146	9	and	and	CCONJ
ajst-28825	146	10	pattern	pattern	NOUN
ajst-28825	146	11	recognition	recognition	NOUN
ajst-28825	146	12	,	,	PUNCT
ajst-28825	146	13	(	(	PUNCT
ajst-28825	146	14	2018	2018	NUM
ajst-28825	146	15	)	)	PUNCT
ajst-28825	146	16	7132	7132	NUM
ajst-28825	146	17	-	-	SYM
ajst-28825	146	18	7141	7141	NUM
ajst-28825	146	19	.	.	PUNCT
ajst-28825	147	1	[	[	X
ajst-28825	147	2	7	7	NUM
ajst-28825	147	3	]	]	X
ajst-28825	147	4	amini	amini	PROPN
ajst-28825	147	5	,	,	PUNCT
ajst-28825	147	6	a.	a.	NOUN
ajst-28825	147	7	,	,	PUNCT
ajst-28825	147	8	shirani	shirani	NOUN
ajst-28825	147	9	-	-	PUNCT
ajst-28825	147	10	mehr	mehr	NOUN
ajst-28825	147	11	,	,	PUNCT
ajst-28825	147	12	h.	h.	PROPN
ajst-28825	147	13	,	,	PUNCT
ajst-28825	147	14	&	&	CCONJ
ajst-28825	147	15	karbasi	karbasi	NOUN
ajst-28825	147	16	,	,	PUNCT
ajst-28825	147	17	a.	a.	PROPN
ajst-28825	147	18	digital	digital	PROPN
ajst-28825	147	19	modulation	modulation	NOUN
ajst-28825	147	20	classification	classification	NOUN
ajst-28825	147	21	:	:	PUNCT
ajst-28825	147	22	a	a	DET
ajst-28825	147	23	deep	deep	ADJ
ajst-28825	147	24	learning	learning	NOUN
ajst-28825	147	25	approach	approach	NOUN
ajst-28825	147	26	.	.	PUNCT
ajst-28825	148	1	ieee	ieee	NOUN
ajst-28825	148	2	transactions	transaction	NOUN
ajst-28825	148	3	on	on	ADP
ajst-28825	148	4	communications	communication	NOUN
ajst-28825	148	5	,	,	PUNCT
ajst-28825	148	6	(	(	PUNCT
ajst-28825	148	7	2017	2017	NUM
ajst-28825	148	8	)	)	PUNCT
ajst-28825	148	9	65(11	65(11	NUM
ajst-28825	148	10	)	)	PUNCT
ajst-28825	148	11	,	,	PUNCT
ajst-28825	148	12	4658	4658	NUM
ajst-28825	148	13	-	-	SYM
ajst-28825	148	14	4668	4668	NUM
ajst-28825	148	15	.	.	PUNCT
ajst-28825	149	1	[	[	X
ajst-28825	149	2	8	8	NUM
ajst-28825	149	3	]	]	PUNCT
ajst-28825	149	4	he	he	PRON
ajst-28825	149	5	,	,	PUNCT
ajst-28825	149	6	k.	k.	PROPN
ajst-28825	149	7	,	,	PUNCT
ajst-28825	149	8	zhang	zhang	PROPN
ajst-28825	149	9	,	,	PUNCT
ajst-28825	149	10	x.	x.	PROPN
ajst-28825	149	11	,	,	PUNCT
ajst-28825	149	12	ren	ren	PROPN
ajst-28825	149	13	,	,	PUNCT
ajst-28825	149	14	s.	s.	PROPN
ajst-28825	149	15	,	,	PUNCT
ajst-28825	149	16	&	&	CCONJ
ajst-28825	149	17	sun	sun	PROPN
ajst-28825	149	18	,	,	PUNCT
ajst-28825	149	19	j.	j.	PROPN
ajst-28825	149	20	deep	deep	ADJ
ajst-28825	149	21	residual	residual	ADJ
ajst-28825	149	22	learning	learning	NOUN
ajst-28825	149	23	for	for	ADP
ajst-28825	149	24	image	image	NOUN
ajst-28825	149	25	recognition	recognition	NOUN
ajst-28825	149	26	.	.	PUNCT
ajst-28825	150	1	proceedings	proceeding	NOUN
ajst-28825	150	2	of	of	ADP
ajst-28825	150	3	the	the	DET
ajst-28825	150	4	ieee	ieee	NOUN
ajst-28825	150	5	conference	conference	NOUN
ajst-28825	150	6	on	on	ADP
ajst-28825	150	7	computer	computer	NOUN
ajst-28825	150	8	vision	vision	NOUN
ajst-28825	150	9	and	and	CCONJ
ajst-28825	150	10	pattern	pattern	NOUN
ajst-28825	150	11	recognition	recognition	NOUN
ajst-28825	150	12	,	,	PUNCT
ajst-28825	150	13	(	(	PUNCT
ajst-28825	150	14	2016	2016	NUM
ajst-28825	150	15	)	)	PUNCT
ajst-28825	150	16	770	770	NUM
ajst-28825	150	17	-	-	SYM
ajst-28825	150	18	778	778	NUM
ajst-28825	150	19	.	.	PUNCT
ajst-28825	151	1	[	[	X
ajst-28825	151	2	9	9	NUM
ajst-28825	151	3	]	]	SYM
ajst-28825	151	4	kingma	kingma	PROPN
ajst-28825	151	5	,	,	PUNCT
ajst-28825	151	6	d.	d.	PROPN
ajst-28825	151	7	p.	p.	PROPN
ajst-28825	151	8	,	,	PUNCT
ajst-28825	151	9	&	&	CCONJ
ajst-28825	151	10	ba	ba	PROPN
ajst-28825	151	11	,	,	PUNCT
ajst-28825	151	12	j.	j.	PROPN
ajst-28825	151	13	adam	adam	PROPN
ajst-28825	151	14	:	:	PUNCT
ajst-28825	151	15	a	a	DET
ajst-28825	151	16	method	method	NOUN
ajst-28825	151	17	for	for	ADP
ajst-28825	151	18	stochastic	stochastic	ADJ
ajst-28825	151	19	optimization	optimization	NOUN
ajst-28825	151	20	.	.	PUNCT
ajst-28825	152	1	proceedings	proceeding	NOUN
ajst-28825	152	2	of	of	ADP
ajst-28825	152	3	the	the	DET
ajst-28825	152	4	international	international	ADJ
ajst-28825	152	5	conference	conference	NOUN
ajst-28825	152	6	on	on	ADP
ajst-28825	152	7	learning	learn	VERB
ajst-28825	152	8	representations(2015	representations(2015	PROPN
ajst-28825	152	9	)	)	PUNCT
ajst-28825	152	10	(	(	PUNCT
ajst-28825	152	11	iclr	iclr	NOUN
ajst-28825	152	12	)	)	PUNCT
ajst-28825	152	13	.	.	PUNCT
ajst-28825	153	1	[	[	X
ajst-28825	153	2	10	10	NUM
ajst-28825	153	3	]	]	X
ajst-28825	153	4	powers	power	NOUN
ajst-28825	153	5	,	,	PUNCT
ajst-28825	153	6	d.	d.	PROPN
ajst-28825	153	7	m.	m.	PROPN
ajst-28825	153	8	w.	w.	PROPN
ajst-28825	153	9	evaluation	evaluation	PROPN
ajst-28825	153	10	:	:	PUNCT
ajst-28825	153	11	from	from	ADP
ajst-28825	153	12	precision	precision	NOUN
ajst-28825	153	13	,	,	PUNCT
ajst-28825	153	14	recall	recall	NOUN
ajst-28825	153	15	and	and	CCONJ
ajst-28825	153	16	fmeasure	fmeasure	NOUN
ajst-28825	153	17	to	to	ADP
ajst-28825	153	18	roc	roc	PROPN
ajst-28825	153	19	,	,	PUNCT
ajst-28825	153	20	informedness	informedness	NOUN
ajst-28825	153	21	,	,	PUNCT
ajst-28825	153	22	markedness	markedness	ADJ
ajst-28825	153	23	and	and	CCONJ
ajst-28825	153	24	correlation	correlation	NOUN
ajst-28825	153	25	.	.	PUNCT
ajst-28825	154	1	journal	journal	NOUN
ajst-28825	154	2	of	of	ADP
ajst-28825	154	3	machine	machine	NOUN
ajst-28825	154	4	learning	learning	NOUN
ajst-28825	154	5	technologies	technology	NOUN
ajst-28825	154	6	,	,	PUNCT
ajst-28825	154	7	(	(	PUNCT
ajst-28825	154	8	2011	2011	NUM
ajst-28825	154	9	)	)	PUNCT
ajst-28825	154	10	2(1	2(1	NUM
ajst-28825	154	11	)	)	PUNCT
ajst-28825	154	12	,	,	PUNCT
ajst-28825	154	13	37	37	NUM
ajst-28825	154	14	-	-	SYM
ajst-28825	154	15	63	63	NUM
ajst-28825	154	16	.	.	PUNCT
ajst-28825	155	1	[	[	X
ajst-28825	155	2	11	11	NUM
ajst-28825	155	3	]	]	X
ajst-28825	155	4	hochreiter	hochreiter	PROPN
ajst-28825	155	5	,	,	PUNCT
ajst-28825	155	6	s.	s.	PROPN
ajst-28825	155	7	,	,	PUNCT
ajst-28825	155	8	&	&	CCONJ
ajst-28825	155	9	schmidhuber	schmidhuber	PROPN
ajst-28825	155	10	,	,	PUNCT
ajst-28825	155	11	j.	j.	PROPN
ajst-28825	155	12	long	long	PROPN
ajst-28825	155	13	short	short	ADJ
ajst-28825	155	14	-	-	PUNCT
ajst-28825	155	15	term	term	NOUN
ajst-28825	155	16	memory	memory	NOUN
ajst-28825	155	17	.	.	PUNCT
ajst-28825	156	1	neural	neural	ADJ
ajst-28825	156	2	computation	computation	NOUN
ajst-28825	156	3	,	,	PUNCT
ajst-28825	156	4	(	(	PUNCT
ajst-28825	156	5	1997	1997	NUM
ajst-28825	156	6	)	)	PUNCT
ajst-28825	156	7	9(8	9(8	NUM
ajst-28825	156	8	)	)	PUNCT
ajst-28825	156	9	,	,	PUNCT
ajst-28825	156	10	1735	1735	NUM
ajst-28825	156	11	-	-	SYM
ajst-28825	156	12	1780	1780	NUM
ajst-28825	156	13	.	.	PUNCT
ajst-28825	157	1	[	[	X
ajst-28825	157	2	12	12	NUM
ajst-28825	157	3	]	]	X
ajst-28825	157	4	lecun	lecun	ADJ
ajst-28825	157	5	,	,	PUNCT
ajst-28825	157	6	y.	y.	PROPN
ajst-28825	157	7	,	,	PUNCT
ajst-28825	157	8	bottou	bottou	PROPN
ajst-28825	157	9	,	,	PUNCT
ajst-28825	157	10	l.	l.	PROPN
ajst-28825	157	11	,	,	PUNCT
ajst-28825	157	12	orr	orr	PROPN
ajst-28825	157	13	,	,	PUNCT
ajst-28825	157	14	g.	g.	PROPN
ajst-28825	157	15	b.	b.	PROPN
ajst-28825	157	16	,	,	PUNCT
ajst-28825	157	17	&	&	CCONJ
ajst-28825	157	18	müller	müller	PROPN
ajst-28825	157	19	,	,	PUNCT
ajst-28825	157	20	k.-r	k.-r	PROPN
ajst-28825	157	21	.	.	PUNCT
ajst-28825	158	1	(	(	PUNCT
ajst-28825	158	2	2012	2012	NUM
ajst-28825	158	3	)	)	PUNCT
ajst-28825	158	4	.	.	PUNCT
ajst-28825	159	1	efficient	efficient	ADJ
ajst-28825	159	2	backprop	backprop	NOUN
ajst-28825	159	3	.	.	PUNCT
ajst-28825	160	1	in	in	ADP
ajst-28825	160	2	g.	g.	PROPN
ajst-28825	160	3	montavon	montavon	PROPN
ajst-28825	160	4	,	,	PUNCT
ajst-28825	160	5	g.	g.	PROPN
ajst-28825	160	6	b.	b.	PROPN
ajst-28825	160	7	orr	orr	PROPN
ajst-28825	160	8	,	,	PUNCT
ajst-28825	160	9	&	&	CCONJ
ajst-28825	160	10	k.-r	k.-r	PROPN
ajst-28825	160	11	.	.	PUNCT
ajst-28825	161	1	müller	müller	PROPN
ajst-28825	161	2	(	(	PUNCT
ajst-28825	161	3	eds	eds	PROPN
ajst-28825	161	4	.	.	PUNCT
ajst-28825	161	5	)	)	PUNCT
ajst-28825	161	6	,	,	PUNCT
ajst-28825	161	7	neural	neural	ADJ
ajst-28825	161	8	networks	network	NOUN
ajst-28825	161	9	:	:	PUNCT
ajst-28825	161	10	tricks	trick	NOUN
ajst-28825	161	11	of	of	ADP
ajst-28825	161	12	the	the	DET
ajst-28825	161	13	trade	trade	NOUN
ajst-28825	161	14	(	(	PUNCT
ajst-28825	161	15	pp	pp	ADJ
ajst-28825	161	16	.	.	PUNCT
ajst-28825	162	1	9	9	NUM
ajst-28825	162	2	-	-	SYM
ajst-28825	162	3	48	48	NUM
ajst-28825	162	4	)	)	PUNCT
ajst-28825	162	5	.	.	PUNCT
ajst-28825	163	1	springer	springer	NOUN
ajst-28825	163	2	.	.	PUNCT
ajst-28825	164	1	[	[	X
ajst-28825	164	2	13	13	NUM
ajst-28825	164	3	]	]	PUNCT
ajst-28825	164	4	shorten	shorten	NOUN
ajst-28825	164	5	,	,	PUNCT
ajst-28825	164	6	c.	c.	PROPN
ajst-28825	164	7	,	,	PUNCT
ajst-28825	164	8	&	&	CCONJ
ajst-28825	164	9	khoshgoftaar	khoshgoftaar	PROPN
ajst-28825	164	10	,	,	PUNCT
ajst-28825	164	11	t.	t.	PROPN
ajst-28825	164	12	m.	m.	NOUN
ajst-28825	164	13	a	a	DET
ajst-28825	164	14	survey	survey	NOUN
ajst-28825	164	15	on	on	ADP
ajst-28825	164	16	image	image	NOUN
ajst-28825	164	17	data	datum	NOUN
ajst-28825	164	18	augmentation	augmentation	NOUN
ajst-28825	164	19	for	for	ADP
ajst-28825	164	20	deep	deep	ADJ
ajst-28825	164	21	learning	learning	NOUN
ajst-28825	164	22	.	.	PUNCT
ajst-28825	165	1	journal	journal	NOUN
ajst-28825	165	2	of	of	ADP
ajst-28825	165	3	big	big	ADJ
ajst-28825	165	4	data	datum	NOUN
ajst-28825	165	5	,	,	PUNCT
ajst-28825	165	6	6(60	6(60	NUM
ajst-28825	165	7	)	)	PUNCT
ajst-28825	165	8	,	,	PUNCT
ajst-28825	165	9	(	(	PUNCT
ajst-28825	165	10	2019	2019	NUM
ajst-28825	165	11	)	)	PUNCT
ajst-28825	165	12	.	.	PUNCT
ajst-28825	166	1	1	1	NUM
ajst-28825	166	2	-	-	SYM
ajst-28825	166	3	48	48	NUM
ajst-28825	166	4	.	.	PUNCT
ajst-28825	167	1	[	[	X
ajst-28825	167	2	14	14	NUM
ajst-28825	167	3	]	]	X
ajst-28825	167	4	prechelt	prechelt	NOUN
ajst-28825	167	5	,	,	PUNCT
ajst-28825	167	6	l.	l.	PROPN
ajst-28825	167	7	(	(	PUNCT
ajst-28825	167	8	1998	1998	NUM
ajst-28825	167	9	)	)	PUNCT
ajst-28825	167	10	.	.	PUNCT
ajst-28825	168	1	early	early	ADJ
ajst-28825	168	2	stopping	stopping	NOUN
ajst-28825	168	3	but	but	CCONJ
ajst-28825	168	4	when	when	SCONJ
ajst-28825	168	5	?	?	PUNCT
ajst-28825	169	1	in	in	ADP
ajst-28825	169	2	g.	g.	PROPN
ajst-28825	169	3	b.	b.	PROPN
ajst-28825	169	4	orr	orr	PROPN
ajst-28825	169	5	&	&	CCONJ
ajst-28825	169	6	k.-r	k.-r	PROPN
ajst-28825	169	7	.	.	PUNCT
ajst-28825	170	1	müller	müller	PROPN
ajst-28825	170	2	(	(	PUNCT
ajst-28825	170	3	eds	eds	PROPN
ajst-28825	170	4	.	.	PUNCT
ajst-28825	170	5	)	)	PUNCT
ajst-28825	170	6	,	,	PUNCT
ajst-28825	170	7	neural	neural	ADJ
ajst-28825	170	8	networks	network	NOUN
ajst-28825	170	9	:	:	PUNCT
ajst-28825	170	10	tricks	trick	NOUN
ajst-28825	170	11	of	of	ADP
ajst-28825	170	12	the	the	DET
ajst-28825	170	13	trade	trade	NOUN
ajst-28825	170	14	(	(	PUNCT
ajst-28825	170	15	pp	pp	ADJ
ajst-28825	170	16	.	.	PUNCT
ajst-28825	171	1	55	55	NUM
ajst-28825	171	2	-	-	SYM
ajst-28825	171	3	69	69	NUM
ajst-28825	171	4	)	)	PUNCT
ajst-28825	171	5	.	.	PUNCT
ajst-28825	172	1	springer	springer	NOUN
ajst-28825	172	2	.	.	PUNCT
ajst-28825	173	1	[	[	X
ajst-28825	173	2	15	15	NUM
ajst-28825	173	3	]	]	X
ajst-28825	173	4	dosovitskiy	dosovitskiy	NOUN
ajst-28825	173	5	,	,	PUNCT
ajst-28825	173	6	a.	a.	PROPN
ajst-28825	173	7	,	,	PUNCT
ajst-28825	173	8	beyer	beyer	PROPN
ajst-28825	173	9	,	,	PUNCT
ajst-28825	173	10	l.	l.	PROPN
ajst-28825	173	11	,	,	PUNCT
ajst-28825	173	12	kolesnikov	kolesnikov	PROPN
ajst-28825	173	13	,	,	PUNCT
ajst-28825	173	14	a.	a.	NOUN
ajst-28825	173	15	,	,	PUNCT
ajst-28825	173	16	weissenborn	weissenborn	ADJ
ajst-28825	173	17	,	,	PUNCT
ajst-28825	173	18	d.	d.	PROPN
ajst-28825	173	19	,	,	PUNCT
ajst-28825	173	20	et	et	PROPN
ajst-28825	173	21	al	al	PROPN
ajst-28825	173	22	.	.	PROPN
ajst-28825	173	23	(	(	PUNCT
ajst-28825	173	24	2020	2020	NUM
ajst-28825	173	25	)	)	PUNCT
ajst-28825	173	26	.	.	PUNCT
ajst-28825	174	1	an	an	DET
ajst-28825	174	2	image	image	NOUN
ajst-28825	174	3	is	be	AUX
ajst-28825	174	4	worth	worth	ADJ
ajst-28825	174	5	16x16	16x16	NUM
ajst-28825	174	6	words	word	NOUN
ajst-28825	174	7	:	:	PUNCT
ajst-28825	174	8	transformers	transformer	NOUN
ajst-28825	174	9	for	for	ADP
ajst-28825	174	10	image	image	NOUN
ajst-28825	174	11	recognition	recognition	NOUN
ajst-28825	174	12	at	at	ADP
ajst-28825	174	13	scale	scale	NOUN
ajst-28825	174	14	.	.	PUNCT
ajst-28825	175	1	proceedings	proceeding	NOUN
ajst-28825	175	2	of	of	ADP
ajst-28825	175	3	the	the	DET
ajst-28825	175	4	international	international	ADJ
ajst-28825	175	5	conference	conference	NOUN
ajst-28825	175	6	on	on	ADP
ajst-28825	175	7	learning	learn	VERB
ajst-28825	175	8	representations	representation	NOUN
ajst-28825	175	9	(	(	PUNCT
ajst-28825	175	10	iclr	iclr	NOUN
ajst-28825	175	11	)	)	PUNCT
ajst-28825	175	12	.	.	PUNCT
ajst-28825	176	1	[	[	X
ajst-28825	176	2	16	16	NUM
ajst-28825	176	3	]	]	X
ajst-28825	176	4	srivastava	srivastava	PROPN
ajst-28825	176	5	,	,	PUNCT
ajst-28825	176	6	n.	n.	PROPN
ajst-28825	176	7	,	,	PUNCT
ajst-28825	176	8	hinton	hinton	PROPN
ajst-28825	176	9	,	,	PUNCT
ajst-28825	176	10	g.	g.	PROPN
ajst-28825	176	11	,	,	PUNCT
ajst-28825	176	12	krizhevsky	krizhevsky	PROPN
ajst-28825	176	13	,	,	PUNCT
ajst-28825	176	14	a.	a.	NOUN
ajst-28825	176	15	,	,	PUNCT
ajst-28825	176	16	sutskever	sutskever	PROPN
ajst-28825	176	17	,	,	PUNCT
ajst-28825	176	18	i.	i.	PROPN
ajst-28825	176	19	,	,	PUNCT
ajst-28825	176	20	&	&	CCONJ
ajst-28825	176	21	salakhutdinov	salakhutdinov	PROPN
ajst-28825	176	22	,	,	PUNCT
ajst-28825	176	23	r.	r.	PROPN
ajst-28825	176	24	dropout	dropout	PROPN
ajst-28825	176	25	:	:	PUNCT
ajst-28825	176	26	a	a	DET
ajst-28825	176	27	simple	simple	ADJ
ajst-28825	176	28	way	way	NOUN
ajst-28825	176	29	to	to	PART
ajst-28825	176	30	prevent	prevent	VERB
ajst-28825	176	31	neural	neural	ADJ
ajst-28825	176	32	networks	network	NOUN
ajst-28825	176	33	from	from	ADP
ajst-28825	176	34	overfitting	overfitte	VERB
ajst-28825	176	35	.	.	PUNCT
ajst-28825	177	1	journal	journal	NOUN
ajst-28825	177	2	of	of	ADP
ajst-28825	177	3	machine	machine	NOUN
ajst-28825	177	4	learning	learn	VERB
ajst-28825	177	5	research	research	NOUN
ajst-28825	177	6	,	,	PUNCT
ajst-28825	177	7	15(1	15(1	NUM
ajst-28825	177	8	)	)	PUNCT
ajst-28825	177	9	,	,	PUNCT
ajst-28825	177	10	(	(	PUNCT
ajst-28825	177	11	2014	2014	NUM
ajst-28825	177	12	)	)	PUNCT
ajst-28825	177	13	.	.	PUNCT
ajst-28825	178	1	1929	1929	NUM
ajst-28825	178	2	-	-	SYM
ajst-28825	178	3	1958	1958	NUM
ajst-28825	178	4	.	.	PUNCT