id	sid	tid	token	lemma	pos
ijassa-93	1	1	advances	advance	NOUN
ijassa-93	1	2	in	in	ADP
ijassa-93	1	3	systems	system	NOUN
ijassa-93	1	4	science	science	NOUN
ijassa-93	1	5	and	and	CCONJ
ijassa-93	1	6	applications	application	NOUN
ijassa-93	1	7	(	(	PUNCT
ijassa-93	1	8	2012	2012	NUM
ijassa-93	1	9	)	)	PUNCT
ijassa-93	1	10	vol.12	vol.12	NOUN
ijassa-93	1	11	no.1	no.1	VERB
ijassa-93	1	12	67	67	NUM
ijassa-93	1	13	-	-	SYM
ijassa-93	1	14	75	75	NUM
ijassa-93	1	15	the	the	DET
ijassa-93	1	16	research	research	NOUN
ijassa-93	1	17	on	on	ADP
ijassa-93	1	18	effective	effective	ADJ
ijassa-93	1	19	video	video	NOUN
ijassa-93	1	20	scene	scene	NOUN
ijassa-93	1	21	character	character	NOUN
ijassa-93	1	22	extraction	extraction	NOUN
ijassa-93	1	23	algorithm	algorithm	NOUN
ijassa-93	1	24	in	in	ADP
ijassa-93	1	25	natural	natural	ADJ
ijassa-93	1	26	scene	scene	NOUN
ijassa-93	1	27	images	image	VERB
ijassa-93	1	28	jinghua	jinghua	NOUN
ijassa-93	2	1	hu1	hu1	NOUN
ijassa-93	3	1	and	and	CCONJ
ijassa-93	3	2	huafeng	huafeng	PROPN
ijassa-93	3	3	kong2	kong2	PROPN
ijassa-93	3	4	1wuhan	1wuhan	NUM
ijassa-93	3	5	university	university	PROPN
ijassa-93	3	6	of	of	ADP
ijassa-93	3	7	technology	technology	NOUN
ijassa-93	3	8	,	,	PUNCT
ijassa-93	3	9	wuhan	wuhan	PROPN
ijassa-93	3	10	430070	430070	NUM
ijassa-93	4	1	2key	2key	NUM
ijassa-93	4	2	laboratory	laboratory	NOUN
ijassa-93	4	3	of	of	ADP
ijassa-93	4	4	information	information	NOUN
ijassa-93	4	5	network	network	NOUN
ijassa-93	4	6	security	security	NOUN
ijassa-93	4	7	,	,	PUNCT
ijassa-93	4	8	the	the	DET
ijassa-93	4	9	ministry	ministry	PROPN
ijassa-93	4	10	of	of	ADP
ijassa-93	4	11	public	public	ADJ
ijassa-93	4	12	security	security	NOUN
ijassa-93	4	13	,	,	PUNCT
ijassa-93	4	14	the	the	DET
ijassa-93	4	15	third	third	ADJ
ijassa-93	4	16	research	research	PROPN
ijassa-93	4	17	institute	institute	PROPN
ijassa-93	4	18	of	of	ADP
ijassa-93	4	19	ministry	ministry	PROPN
ijassa-93	4	20	of	of	ADP
ijassa-93	4	21	public	public	ADJ
ijassa-93	4	22	security	security	NOUN
ijassa-93	4	23	,	,	PUNCT
ijassa-93	4	24	shanghai	shanghai	PROPN
ijassa-93	4	25	201204	201204	NUM
ijassa-93	4	26	abstract	abstract	ADJ
ijassa-93	4	27	pcb	pcb	NOUN
ijassa-93	4	28	technology	technology	NOUN
ijassa-93	4	29	is	be	AUX
ijassa-93	4	30	particularly	particularly	ADV
ijassa-93	4	31	important	important	ADJ
ijassa-93	4	32	in	in	ADP
ijassa-93	4	33	electronic	electronic	ADJ
ijassa-93	4	34	industry	industry	NOUN
ijassa-93	4	35	.	.	PUNCT
ijassa-93	5	1	however	however	ADV
ijassa-93	5	2	,	,	PUNCT
ijassa-93	5	3	the	the	DET
ijassa-93	5	4	increasing	increasing	NOUN
ijassa-93	5	5	of	of	ADP
ijassa-93	5	6	technology	technology	NOUN
ijassa-93	5	7	complexity	complexity	NOUN
ijassa-93	5	8	makes	make	VERB
ijassa-93	5	9	it	it	PRON
ijassa-93	5	10	difficult	difficult	ADJ
ijassa-93	5	11	to	to	PART
ijassa-93	5	12	inspect	inspect	VERB
ijassa-93	5	13	the	the	DET
ijassa-93	5	14	quality	quality	NOUN
ijassa-93	5	15	of	of	ADP
ijassa-93	5	16	pcb	pcb	NOUN
ijassa-93	5	17	.	.	PUNCT
ijassa-93	6	1	aoi	aoi	PROPN
ijassa-93	6	2	system	system	NOUN
ijassa-93	6	3	greatly	greatly	ADV
ijassa-93	6	4	improves	improve	VERB
ijassa-93	6	5	the	the	DET
ijassa-93	6	6	pcb	pcb	NOUN
ijassa-93	6	7	inspection	inspection	NOUN
ijassa-93	6	8	efficiency	efficiency	NOUN
ijassa-93	6	9	,	,	PUNCT
ijassa-93	6	10	and	and	CCONJ
ijassa-93	6	11	it	it	PRON
ijassa-93	6	12	has	have	AUX
ijassa-93	6	13	become	become	VERB
ijassa-93	6	14	one	one	NUM
ijassa-93	6	15	of	of	ADP
ijassa-93	6	16	the	the	DET
ijassa-93	6	17	most	most	ADV
ijassa-93	6	18	instructive	instructive	ADJ
ijassa-93	6	19	subjects	subject	NOUN
ijassa-93	6	20	that	that	SCONJ
ijassa-93	6	21	how	how	SCONJ
ijassa-93	6	22	to	to	PART
ijassa-93	6	23	identify	identify	VERB
ijassa-93	6	24	the	the	DET
ijassa-93	6	25	text	text	NOUN
ijassa-93	6	26	information	information	NOUN
ijassa-93	6	27	in	in	ADP
ijassa-93	6	28	the	the	DET
ijassa-93	6	29	chip	chip	NOUN
ijassa-93	6	30	image	image	NOUN
ijassa-93	6	31	rapidly	rapidly	ADV
ijassa-93	6	32	and	and	CCONJ
ijassa-93	6	33	accurately	accurately	ADV
ijassa-93	6	34	.	.	PUNCT
ijassa-93	7	1	since	since	SCONJ
ijassa-93	7	2	the	the	DET
ijassa-93	7	3	chip	chip	NOUN
ijassa-93	7	4	images	image	NOUN
ijassa-93	7	5	have	have	VERB
ijassa-93	7	6	complex	complex	ADJ
ijassa-93	7	7	natural	natural	ADJ
ijassa-93	7	8	background	background	NOUN
ijassa-93	7	9	,	,	PUNCT
ijassa-93	7	10	and	and	CCONJ
ijassa-93	7	11	are	be	AUX
ijassa-93	7	12	greatly	greatly	ADV
ijassa-93	7	13	affected	affect	VERB
ijassa-93	7	14	by	by	ADP
ijassa-93	7	15	light	light	ADJ
ijassa-93	7	16	,	,	PUNCT
ijassa-93	7	17	shadow	shadow	NOUN
ijassa-93	7	18	,	,	PUNCT
ijassa-93	7	19	noise	noise	NOUN
ijassa-93	7	20	,	,	PUNCT
ijassa-93	7	21	font	font	NOUN
ijassa-93	7	22	and	and	CCONJ
ijassa-93	7	23	size	size	NOUN
ijassa-93	7	24	,	,	PUNCT
ijassa-93	7	25	texture	texture	NOUN
ijassa-93	7	26	,	,	PUNCT
ijassa-93	7	27	color	color	NOUN
ijassa-93	7	28	,	,	PUNCT
ijassa-93	7	29	position	position	NOUN
ijassa-93	7	30	and	and	CCONJ
ijassa-93	7	31	arrangement	arrangement	NOUN
ijassa-93	7	32	.	.	PUNCT
ijassa-93	8	1	,	,	PUNCT
ijassa-93	8	2	it	it	PRON
ijassa-93	8	3	is	be	AUX
ijassa-93	8	4	often	often	ADV
ijassa-93	8	5	difficult	difficult	ADJ
ijassa-93	8	6	to	to	PART
ijassa-93	8	7	inspect	inspect	VERB
ijassa-93	8	8	,	,	PUNCT
ijassa-93	8	9	extract	extract	NOUN
ijassa-93	8	10	and	and	CCONJ
ijassa-93	8	11	identity	identity	NOUN
ijassa-93	8	12	the	the	DET
ijassa-93	8	13	text	text	NOUN
ijassa-93	8	14	.	.	PUNCT
ijassa-93	9	1	this	this	DET
ijassa-93	9	2	paper	paper	NOUN
ijassa-93	9	3	presents	present	VERB
ijassa-93	9	4	an	an	DET
ijassa-93	9	5	effective	effective	ADJ
ijassa-93	9	6	algorithm	algorithm	NOUN
ijassa-93	9	7	of	of	ADP
ijassa-93	9	8	scene	scene	NOUN
ijassa-93	9	9	character	character	NOUN
ijassa-93	9	10	segmentation	segmentation	NOUN
ijassa-93	9	11	and	and	CCONJ
ijassa-93	9	12	recognition	recognition	NOUN
ijassa-93	9	13	in	in	ADP
ijassa-93	9	14	natural	natural	ADJ
ijassa-93	9	15	scene	scene	NOUN
ijassa-93	9	16	images	image	NOUN
ijassa-93	9	17	.	.	PUNCT
ijassa-93	10	1	the	the	DET
ijassa-93	10	2	algorithm	algorithm	NOUN
ijassa-93	10	3	segments	segment	VERB
ijassa-93	10	4	lines	line	NOUN
ijassa-93	10	5	of	of	ADP
ijassa-93	10	6	character	character	NOUN
ijassa-93	10	7	,	,	PUNCT
ijassa-93	10	8	then	then	ADV
ijassa-93	10	9	segments	segment	VERB
ijassa-93	10	10	every	every	DET
ijassa-93	10	11	line	line	NOUN
ijassa-93	10	12	of	of	ADP
ijassa-93	10	13	character	character	NOUN
ijassa-93	10	14	into	into	ADP
ijassa-93	10	15	individual	individual	ADJ
ijassa-93	10	16	words	word	NOUN
ijassa-93	10	17	for	for	ADP
ijassa-93	10	18	further	further	ADJ
ijassa-93	10	19	processing	processing	NOUN
ijassa-93	10	20	,	,	PUNCT
ijassa-93	10	21	such	such	ADJ
ijassa-93	10	22	as	as	ADP
ijassa-93	10	23	feature	feature	NOUN
ijassa-93	10	24	extraction	extraction	NOUN
ijassa-93	10	25	and	and	CCONJ
ijassa-93	10	26	character	character	NOUN
ijassa-93	10	27	recognition	recognition	NOUN
ijassa-93	10	28	according	accord	VERB
ijassa-93	10	29	to	to	ADP
ijassa-93	10	30	the	the	DET
ijassa-93	10	31	known	know	VERB
ijassa-93	10	32	features	feature	NOUN
ijassa-93	10	33	of	of	ADP
ijassa-93	10	34	character	character	NOUN
ijassa-93	10	35	.	.	PUNCT
ijassa-93	11	1	after	after	ADP
ijassa-93	11	2	the	the	DET
ijassa-93	11	3	segmentation	segmentation	NOUN
ijassa-93	11	4	of	of	ADP
ijassa-93	11	5	character	character	NOUN
ijassa-93	11	6	,	,	PUNCT
ijassa-93	11	7	we	we	PRON
ijassa-93	11	8	use	use	VERB
ijassa-93	11	9	an	an	DET
ijassa-93	11	10	advanced	advanced	ADJ
ijassa-93	11	11	bp	bp	NOUN
ijassa-93	11	12	algorithm	algorithm	NOUN
ijassa-93	11	13	to	to	PART
ijassa-93	11	14	recognize	recognize	VERB
ijassa-93	11	15	the	the	DET
ijassa-93	11	16	character	character	NOUN
ijassa-93	11	17	.	.	PUNCT
ijassa-93	12	1	it	it	PRON
ijassa-93	12	2	improves	improve	VERB
ijassa-93	12	3	bp	bp	PROPN
ijassa-93	12	4	mainly	mainly	ADV
ijassa-93	12	5	through	through	ADP
ijassa-93	12	6	restructuring	restructure	VERB
ijassa-93	12	7	gradient	gradient	NOUN
ijassa-93	12	8	in	in	ADP
ijassa-93	12	9	the	the	DET
ijassa-93	12	10	sigmoid	sigmoid	NOUN
ijassa-93	12	11	.	.	PUNCT
ijassa-93	13	1	the	the	DET
ijassa-93	13	2	experiment	experiment	NOUN
ijassa-93	13	3	shows	show	VERB
ijassa-93	13	4	a	a	DET
ijassa-93	13	5	significantly	significantly	ADV
ijassa-93	13	6	improvement	improvement	NOUN
ijassa-93	13	7	in	in	ADP
ijassa-93	13	8	pc	pc	NOUN
ijassa-93	13	9	:	:	PUNCT
ijassa-93	13	10	the	the	DET
ijassa-93	13	11	recognition	recognition	NOUN
ijassa-93	13	12	accuracy	accuracy	NOUN
ijassa-93	13	13	rate	rate	NOUN
ijassa-93	13	14	of	of	ADP
ijassa-93	13	15	this	this	DET
ijassa-93	13	16	system	system	NOUN
ijassa-93	13	17	is	be	AUX
ijassa-93	13	18	above	above	ADP
ijassa-93	13	19	96	96	NUM
ijassa-93	13	20	%	%	NOUN
ijassa-93	13	21	and	and	CCONJ
ijassa-93	13	22	the	the	DET
ijassa-93	13	23	response	response	NOUN
ijassa-93	13	24	time	time	NOUN
ijassa-93	13	25	is	be	AUX
ijassa-93	13	26	5ms/100	5ms/100	NUM
ijassa-93	13	27	words	word	NOUN
ijassa-93	13	28	.	.	PUNCT
ijassa-93	14	1	keywords	keyword	NOUN
ijassa-93	14	2	scenes	scene	VERB
ijassa-93	14	3	character	character	NOUN
ijassa-93	14	4	segmentation	segmentation	NOUN
ijassa-93	14	5	,	,	PUNCT
ijassa-93	14	6	feature	feature	NOUN
ijassa-93	14	7	extraction	extraction	NOUN
ijassa-93	14	8	,	,	PUNCT
ijassa-93	14	9	bp	bp	PROPN
ijassa-93	14	10	algorithm	algorithm	PROPN
ijassa-93	14	11	,	,	PUNCT
ijassa-93	14	12	and	and	CCONJ
ijassa-93	14	13	image	image	NOUN
ijassa-93	14	14	processing	process	VERB
ijassa-93	14	15	1	1	NUM
ijassa-93	14	16	introduction	introduction	NOUN
ijassa-93	14	17	as	as	ADP
ijassa-93	14	18	the	the	DET
ijassa-93	14	19	optical	optical	ADJ
ijassa-93	14	20	character	character	NOUN
ijassa-93	14	21	recognition	recognition	NOUN
ijassa-93	14	22	(	(	PUNCT
ijassa-93	14	23	ocr	ocr	NOUN
ijassa-93	14	24	)	)	PUNCT
ijassa-93	14	25	technology	technology	NOUN
ijassa-93	14	26	comes	come	VERB
ijassa-93	14	27	in	in	ADP
ijassa-93	14	28	vogue	vogue	NOUN
ijassa-93	14	29	,	,	PUNCT
ijassa-93	14	30	many	many	ADJ
ijassa-93	14	31	scholars	scholar	NOUN
ijassa-93	14	32	begin	begin	VERB
ijassa-93	14	33	to	to	PART
ijassa-93	14	34	research	research	VERB
ijassa-93	14	35	on	on	ADP
ijassa-93	14	36	the	the	DET
ijassa-93	14	37	character	character	NOUN
ijassa-93	14	38	extraction	extraction	NOUN
ijassa-93	14	39	from	from	ADP
ijassa-93	14	40	the	the	DET
ijassa-93	14	41	document	document	NOUN
ijassa-93	14	42	images	image	NOUN
ijassa-93	14	43	.	.	PUNCT
ijassa-93	15	1	till	till	SCONJ
ijassa-93	15	2	the	the	DET
ijassa-93	15	3	1990s	1990s	NUM
ijassa-93	15	4	,	,	PUNCT
ijassa-93	15	5	during	during	ADP
ijassa-93	15	6	the	the	DET
ijassa-93	15	7	rapid	rapid	ADJ
ijassa-93	15	8	development	development	NOUN
ijassa-93	15	9	of	of	ADP
ijassa-93	15	10	computer	computer	NOUN
ijassa-93	15	11	technology	technology	NOUN
ijassa-93	15	12	and	and	CCONJ
ijassa-93	15	13	multimedia	multimedia	NOUN
ijassa-93	15	14	technology	technology	NOUN
ijassa-93	15	15	,	,	PUNCT
ijassa-93	15	16	the	the	DET
ijassa-93	15	17	content	content	NOUN
ijassa-93	15	18	-	-	PUNCT
ijassa-93	15	19	based	base	VERB
ijassa-93	15	20	multimedia	multimedia	NOUN
ijassa-93	15	21	retrieval	retrieval	NOUN
ijassa-93	15	22	has	have	AUX
ijassa-93	15	23	become	become	VERB
ijassa-93	15	24	a	a	DET
ijassa-93	15	25	research	research	NOUN
ijassa-93	15	26	hotspot	hotspot	NOUN
ijassa-93	15	27	.	.	PUNCT
ijassa-93	16	1	then	then	ADV
ijassa-93	16	2	,	,	PUNCT
ijassa-93	16	3	the	the	DET
ijassa-93	16	4	character	character	NOUN
ijassa-93	16	5	extraction	extraction	NOUN
ijassa-93	16	6	in	in	ADP
ijassa-93	16	7	the	the	DET
ijassa-93	16	8	natural	natural	ADJ
ijassa-93	16	9	scene	scene	NOUN
ijassa-93	16	10	images	image	NOUN
ijassa-93	16	11	has	have	AUX
ijassa-93	16	12	again	again	ADV
ijassa-93	16	13	gradually	gradually	ADV
ijassa-93	16	14	become	become	VERB
ijassa-93	16	15	one	one	NUM
ijassa-93	16	16	of	of	ADP
ijassa-93	16	17	the	the	DET
ijassa-93	16	18	research	research	NOUN
ijassa-93	16	19	hotspots	hotspot	NOUN
ijassa-93	16	20	.	.	PUNCT
ijassa-93	17	1	usually	usually	ADV
ijassa-93	17	2	,	,	PUNCT
ijassa-93	17	3	the	the	DET
ijassa-93	17	4	natural	natural	ADJ
ijassa-93	17	5	scene	scene	NOUN
ijassa-93	17	6	image	image	NOUN
ijassa-93	17	7	character	character	NOUN
ijassa-93	17	8	has	have	VERB
ijassa-93	17	9	substantial	substantial	ADJ
ijassa-93	17	10	changes	change	NOUN
ijassa-93	17	11	in	in	ADP
ijassa-93	17	12	font	font	NOUN
ijassa-93	17	13	,	,	PUNCT
ijassa-93	17	14	size	size	NOUN
ijassa-93	17	15	,	,	PUNCT
ijassa-93	17	16	color	color	NOUN
ijassa-93	17	17	,	,	PUNCT
ijassa-93	17	18	alignment	alignment	NOUN
ijassa-93	17	19	and	and	CCONJ
ijassa-93	17	20	arrangement	arrangement	NOUN
ijassa-93	17	21	;	;	PUNCT
ijassa-93	17	22	also	also	ADV
ijassa-93	17	23	with	with	ADP
ijassa-93	17	24	the	the	DET
ijassa-93	17	25	complex	complex	ADJ
ijassa-93	17	26	character	character	NOUN
ijassa-93	17	27	background	background	NOUN
ijassa-93	17	28	,	,	PUNCT
ijassa-93	17	29	low	low	ADJ
ijassa-93	17	30	resolution	resolution	NOUN
ijassa-93	17	31	and	and	CCONJ
ijassa-93	17	32	high	high	ADJ
ijassa-93	17	33	noise	noise	NOUN
ijassa-93	17	34	.	.	PUNCT
ijassa-93	18	1	furthermore	furthermore	ADV
ijassa-93	18	2	,	,	PUNCT
ijassa-93	18	3	many	many	ADJ
ijassa-93	18	4	systems	system	NOUN
ijassa-93	18	5	also	also	ADV
ijassa-93	18	6	require	require	VERB
ijassa-93	18	7	the	the	DET
ijassa-93	18	8	algorithm	algorithm	NOUN
ijassa-93	18	9	has	have	VERB
ijassa-93	18	10	higher	high	ADJ
ijassa-93	18	11	processing	processing	NOUN
ijassa-93	18	12	speed	speed	NOUN
ijassa-93	18	13	in	in	ADP
ijassa-93	18	14	the	the	DET
ijassa-93	18	15	application	application	NOUN
ijassa-93	18	16	.	.	PUNCT
ijassa-93	19	1	all	all	PRON
ijassa-93	19	2	of	of	ADP
ijassa-93	19	3	these	these	PRON
ijassa-93	19	4	causing	cause	VERB
ijassa-93	19	5	difficulties	difficulty	NOUN
ijassa-93	19	6	to	to	PART
ijassa-93	19	7	effectively	effectively	ADV
ijassa-93	19	8	extract	extract	VERB
ijassa-93	19	9	the	the	DET
ijassa-93	19	10	characters	character	NOUN
ijassa-93	19	11	from	from	ADP
ijassa-93	19	12	the	the	DET
ijassa-93	19	13	natural	natural	ADJ
ijassa-93	19	14	scene	scene	NOUN
ijassa-93	19	15	images	image	NOUN
ijassa-93	19	16	,	,	PUNCT
ijassa-93	19	17	especially	especially	ADV
ijassa-93	19	18	for	for	ADP
ijassa-93	19	19	the	the	DET
ijassa-93	19	20	natural	natural	ADJ
ijassa-93	19	21	68	68	NUM
ijassa-93	19	22	jinghua	jinghua	NOUN
ijassa-93	19	23	hu	hu	PROPN
ijassa-93	19	24	:	:	PUNCT
ijassa-93	19	25	the	the	DET
ijassa-93	19	26	research	research	NOUN
ijassa-93	19	27	on	on	ADP
ijassa-93	19	28	effective	effective	ADJ
ijassa-93	19	29	video	video	NOUN
ijassa-93	19	30	scene	scene	NOUN
ijassa-93	19	31	character	character	NOUN
ijassa-93	19	32	extraction	extraction	NOUN
ijassa-93	19	33	algorithm	algorithm	NOUN
ijassa-93	19	34	...	...	PUNCT
ijassa-93	19	35	scene	scene	NOUN
ijassa-93	19	36	images	image	NOUN
ijassa-93	19	37	based	base	VERB
ijassa-93	19	38	on	on	ADP
ijassa-93	19	39	the	the	DET
ijassa-93	19	40	videos	video	NOUN
ijassa-93	19	41	.	.	PUNCT
ijassa-93	20	1	many	many	ADJ
ijassa-93	20	2	domestic	domestic	ADJ
ijassa-93	20	3	and	and	CCONJ
ijassa-93	20	4	foreign	foreign	ADJ
ijassa-93	20	5	scholars	scholar	NOUN
ijassa-93	20	6	have	have	AUX
ijassa-93	20	7	made	make	VERB
ijassa-93	20	8	beneficial	beneficial	ADJ
ijassa-93	20	9	explorations	exploration	NOUN
ijassa-93	20	10	and	and	CCONJ
ijassa-93	20	11	attempts	attempt	NOUN
ijassa-93	20	12	in	in	ADP
ijassa-93	20	13	this	this	DET
ijassa-93	20	14	field	field	NOUN
ijassa-93	20	15	.	.	PUNCT
ijassa-93	21	1	y.zhong[1	y.zhong[1	PROPN
ijassa-93	21	2	]	]	X
ijassa-93	21	3	firstly	firstly	ADV
ijassa-93	21	4	put	put	VERB
ijassa-93	21	5	forward	forward	ADV
ijassa-93	21	6	the	the	DET
ijassa-93	21	7	resolution	resolution	NOUN
ijassa-93	21	8	that	that	PRON
ijassa-93	21	9	position	position	VERB
ijassa-93	21	10	the	the	DET
ijassa-93	21	11	characters	character	NOUN
ijassa-93	21	12	in	in	ADP
ijassa-93	21	13	the	the	DET
ijassa-93	21	14	complex	complex	ADJ
ijassa-93	21	15	images	image	NOUN
ijassa-93	21	16	.	.	PUNCT
ijassa-93	22	1	however	however	ADV
ijassa-93	22	2	,	,	PUNCT
ijassa-93	22	3	this	this	DET
ijassa-93	22	4	resolution	resolution	NOUN
ijassa-93	22	5	mainly	mainly	ADV
ijassa-93	22	6	aims	aim	VERB
ijassa-93	22	7	at	at	ADP
ijassa-93	22	8	the	the	DET
ijassa-93	22	9	character	character	NOUN
ijassa-93	22	10	positioning	positioning	NOUN
ijassa-93	22	11	in	in	ADP
ijassa-93	22	12	the	the	DET
ijassa-93	22	13	image	image	NOUN
ijassa-93	22	14	scanning	scanning	NOUN
ijassa-93	22	15	for	for	ADP
ijassa-93	22	16	the	the	DET
ijassa-93	22	17	colorful	colorful	ADJ
ijassa-93	22	18	disc	disc	NOUN
ijassa-93	22	19	covers	cover	NOUN
ijassa-93	22	20	,	,	PUNCT
ijassa-93	22	21	and	and	CCONJ
ijassa-93	22	22	fails	fail	VERB
ijassa-93	22	23	to	to	PART
ijassa-93	22	24	be	be	AUX
ijassa-93	22	25	directly	directly	ADV
ijassa-93	22	26	applied	apply	VERB
ijassa-93	22	27	into	into	ADP
ijassa-93	22	28	the	the	DET
ijassa-93	22	29	natural	natural	ADJ
ijassa-93	22	30	scene	scene	NOUN
ijassa-93	22	31	images	image	NOUN
ijassa-93	22	32	.	.	PUNCT
ijassa-93	23	1	a.k.jai[2	a.k.jai[2	PROPN
ijassa-93	23	2	]	]	PUNCT
ijassa-93	23	3	and	and	CCONJ
ijassa-93	23	4	others	other	NOUN
ijassa-93	23	5	proposed	propose	VERB
ijassa-93	23	6	a	a	DET
ijassa-93	23	7	kind	kind	NOUN
ijassa-93	23	8	of	of	ADP
ijassa-93	23	9	character	character	NOUN
ijassa-93	23	10	positioning	positioning	NOUN
ijassa-93	23	11	method	method	NOUN
ijassa-93	23	12	,	,	PUNCT
ijassa-93	23	13	which	which	PRON
ijassa-93	23	14	applies	apply	VERB
ijassa-93	23	15	to	to	ADP
ijassa-93	23	16	the	the	DET
ijassa-93	23	17	newspaper	newspaper	NOUN
ijassa-93	23	18	,	,	PUNCT
ijassa-93	23	19	web	web	NOUN
ijassa-93	23	20	pages	page	NOUN
ijassa-93	23	21	and	and	CCONJ
ijassa-93	23	22	general	general	ADJ
ijassa-93	23	23	images	image	NOUN
ijassa-93	23	24	and	and	CCONJ
ijassa-93	23	25	video	video	NOUN
ijassa-93	23	26	frames	frame	NOUN
ijassa-93	23	27	,	,	PUNCT
ijassa-93	23	28	but	but	CCONJ
ijassa-93	23	29	not	not	PART
ijassa-93	23	30	ideal	ideal	ADJ
ijassa-93	23	31	to	to	PART
ijassa-93	23	32	identify	identify	VERB
ijassa-93	23	33	the	the	DET
ijassa-93	23	34	small	small	ADJ
ijassa-93	23	35	character	character	NOUN
ijassa-93	23	36	fonts	font	NOUN
ijassa-93	23	37	.	.	PUNCT
ijassa-93	24	1	m.a.smith[3	m.a.smith[3	NOUN
ijassa-93	24	2	]	]	PUNCT
ijassa-93	24	3	and	and	CCONJ
ijassa-93	24	4	others	other	NOUN
ijassa-93	24	5	developed	develop	VERB
ijassa-93	24	6	a	a	DET
ijassa-93	24	7	kind	kind	NOUN
ijassa-93	24	8	of	of	ADP
ijassa-93	24	9	method	method	NOUN
ijassa-93	24	10	detecting	detect	VERB
ijassa-93	24	11	the	the	DET
ijassa-93	24	12	characters	character	NOUN
ijassa-93	24	13	on	on	ADP
ijassa-93	24	14	the	the	DET
ijassa-93	24	15	images	image	NOUN
ijassa-93	24	16	.	.	PUNCT
ijassa-93	25	1	however	however	ADV
ijassa-93	25	2	,	,	PUNCT
ijassa-93	25	3	because	because	SCONJ
ijassa-93	25	4	of	of	ADP
ijassa-93	25	5	the	the	DET
ijassa-93	25	6	limitation	limitation	NOUN
ijassa-93	25	7	in	in	ADP
ijassa-93	25	8	size	size	NOUN
ijassa-93	25	9	,	,	PUNCT
ijassa-93	25	10	this	this	DET
ijassa-93	25	11	method	method	NOUN
ijassa-93	25	12	can	can	AUX
ijassa-93	25	13	only	only	ADV
ijassa-93	25	14	detect	detect	VERB
ijassa-93	25	15	the	the	DET
ijassa-93	25	16	characters	character	NOUN
ijassa-93	25	17	within	within	ADP
ijassa-93	25	18	the	the	DET
ijassa-93	25	19	specific	specific	ADJ
ijassa-93	25	20	font	font	NOUN
ijassa-93	25	21	scope	scope	NOUN
ijassa-93	25	22	.	.	PUNCT
ijassa-93	26	1	it	it	PRON
ijassa-93	26	2	can	can	AUX
ijassa-93	26	3	not	not	PART
ijassa-93	26	4	use	use	VERB
ijassa-93	26	5	the	the	DET
ijassa-93	26	6	feature	feature	NOUN
ijassa-93	26	7	that	that	PRON
ijassa-93	26	8	the	the	DET
ijassa-93	26	9	same	same	ADJ
ijassa-93	26	10	characters	character	NOUN
ijassa-93	26	11	will	will	AUX
ijassa-93	26	12	appear	appear	VERB
ijassa-93	26	13	in	in	ADP
ijassa-93	26	14	multi	multi	ADJ
ijassa-93	26	15	frame	frame	NOUN
ijassa-93	26	16	to	to	PART
ijassa-93	26	17	further	far	ADV
ijassa-93	26	18	enhance	enhance	VERB
ijassa-93	26	19	the	the	DET
ijassa-93	26	20	character	character	NOUN
ijassa-93	26	21	detection	detection	NOUN
ijassa-93	26	22	performance	performance	NOUN
ijassa-93	26	23	.	.	PUNCT
ijassa-93	27	1	also	also	ADV
ijassa-93	27	2	,	,	PUNCT
ijassa-93	27	3	the	the	DET
ijassa-93	27	4	word	word	NOUN
ijassa-93	27	5	segmentation	segmentation	NOUN
ijassa-93	27	6	prepared	prepare	VERB
ijassa-93	27	7	for	for	ADP
ijassa-93	27	8	the	the	DET
ijassa-93	27	9	ocr	ocr	NOUN
ijassa-93	27	10	is	be	AUX
ijassa-93	27	11	not	not	PART
ijassa-93	27	12	done	do	VERB
ijassa-93	27	13	.	.	PUNCT
ijassa-93	28	1	sato[4	sato[4	NOUN
ijassa-93	28	2	]	]	PUNCT
ijassa-93	28	3	and	and	CCONJ
ijassa-93	28	4	others	other	NOUN
ijassa-93	28	5	developed	develop	VERB
ijassa-93	28	6	a	a	DET
ijassa-93	28	7	character	character	NOUN
ijassa-93	28	8	segmentation	segmentation	NOUN
ijassa-93	28	9	and	and	CCONJ
ijassa-93	28	10	recognition	recognition	NOUN
ijassa-93	28	11	system	system	NOUN
ijassa-93	28	12	aiming	aim	VERB
ijassa-93	28	13	at	at	ADP
ijassa-93	28	14	the	the	DET
ijassa-93	28	15	static	static	ADJ
ijassa-93	28	16	low	low	ADJ
ijassa-93	28	17	resolution	resolution	NOUN
ijassa-93	28	18	headlines	headline	NOUN
ijassa-93	28	19	.	.	PUNCT
ijassa-93	29	1	the	the	DET
ijassa-93	29	2	system	system	NOUN
ijassa-93	29	3	will	will	AUX
ijassa-93	29	4	take	take	VERB
ijassa-93	29	5	advantage	advantage	NOUN
ijassa-93	29	6	of	of	ADP
ijassa-93	29	7	the	the	DET
ijassa-93	29	8	method	method	NOUN
ijassa-93	29	9	mentioned	mention	VERB
ijassa-93	29	10	in	in	ADP
ijassa-93	29	11	literature[3	literature[3	PROPN
ijassa-93	29	12	]	]	PUNCT
ijassa-93	29	13	to	to	PART
ijassa-93	29	14	recognize	recognize	VERB
ijassa-93	29	15	the	the	DET
ijassa-93	29	16	character	character	NOUN
ijassa-93	29	17	.	.	PUNCT
ijassa-93	30	1	then	then	ADV
ijassa-93	30	2	magnify	magnify	VERB
ijassa-93	30	3	as	as	ADP
ijassa-93	30	4	4	4	NUM
ijassa-93	30	5	times	time	NOUN
ijassa-93	30	6	as	as	ADP
ijassa-93	30	7	the	the	DET
ijassa-93	30	8	recognized	recognize	VERB
ijassa-93	30	9	characters	character	NOUN
ijassa-93	30	10	,	,	PUNCT
ijassa-93	30	11	which	which	PRON
ijassa-93	30	12	will	will	AUX
ijassa-93	30	13	use	use	VERB
ijassa-93	30	14	the	the	DET
ijassa-93	30	15	smallest	small	ADJ
ijassa-93	30	16	time	time	NOUN
ijassa-93	30	17	-	-	PUNCT
ijassa-93	30	18	based	base	VERB
ijassa-93	30	19	image	image	NOUN
ijassa-93	30	20	pixed	pixe	VERB
ijassa-93	30	21	value	value	NOUN
ijassa-93	30	22	to	to	PART
ijassa-93	30	23	conformify	conformify	VERB
ijassa-93	30	24	the	the	DET
ijassa-93	30	25	headline	headline	NOUN
ijassa-93	30	26	.	.	PUNCT
ijassa-93	31	1	such	such	ADJ
ijassa-93	31	2	system	system	NOUN
ijassa-93	31	3	has	have	VERB
ijassa-93	31	4	a	a	DET
ijassa-93	31	5	good	good	ADJ
ijassa-93	31	6	effect	effect	NOUN
ijassa-93	31	7	on	on	ADP
ijassa-93	31	8	news	news	NOUN
ijassa-93	31	9	program	program	NOUN
ijassa-93	31	10	.	.	PUNCT
ijassa-93	32	1	r.lienhart[5	r.lienhart[5	NOUN
ijassa-93	32	2	-	-	PUNCT
ijassa-93	32	3	6	6	NUM
ijassa-93	32	4	]	]	PUNCT
ijassa-93	32	5	and	and	CCONJ
ijassa-93	32	6	others	other	NOUN
ijassa-93	32	7	have	have	AUX
ijassa-93	32	8	developed	develop	VERB
ijassa-93	32	9	systems	system	NOUN
ijassa-93	32	10	in	in	ADP
ijassa-93	32	11	character	character	NOUN
ijassa-93	32	12	detecting	detecting	NOUN
ijassa-93	32	13	,	,	PUNCT
ijassa-93	32	14	identifying	identify	VERB
ijassa-93	32	15	,	,	PUNCT
ijassa-93	32	16	recognizing	recognize	VERB
ijassa-93	32	17	successively	successively	ADV
ijassa-93	32	18	.	.	PUNCT
ijassa-93	33	1	the	the	DET
ijassa-93	33	2	color	color	NOUN
ijassa-93	33	3	-	-	PUNCT
ijassa-93	33	4	based	base	VERB
ijassa-93	33	5	algorithm	algorithm	NOUN
ijassa-93	33	6	system	system	NOUN
ijassa-93	33	7	of	of	ADP
ijassa-93	33	8	inter	inter	ADJ
ijassa-93	33	9	infiltration	infiltration	NOUN
ijassa-93	33	10	segmentation	segmentation	NOUN
ijassa-93	33	11	adopted	adopt	VERB
ijassa-93	33	12	in	in	ADP
ijassa-93	33	13	literature[5	literature[5	PROPN
ijassa-93	33	14	]	]	PUNCT
ijassa-93	33	15	,	,	PUNCT
ijassa-93	33	16	only	only	ADV
ijassa-93	33	17	deals	deal	NOUN
ijassa-93	33	18	with	with	ADP
ijassa-93	33	19	image	image	NOUN
ijassa-93	33	20	segmentation	segmentation	NOUN
ijassa-93	33	21	and	and	CCONJ
ijassa-93	33	22	recognition	recognition	NOUN
ijassa-93	33	23	in	in	ADP
ijassa-93	33	24	single	single	ADJ
ijassa-93	33	25	frame	frame	NOUN
ijassa-93	33	26	,	,	PUNCT
ijassa-93	33	27	without	without	ADP
ijassa-93	33	28	consideration	consideration	NOUN
ijassa-93	33	29	in	in	ADP
ijassa-93	33	30	successive	successive	ADJ
ijassa-93	33	31	images	image	NOUN
ijassa-93	33	32	in	in	ADP
ijassa-93	33	33	multi	multi	ADJ
ijassa-93	33	34	frame	frame	NOUN
ijassa-93	33	35	.	.	PUNCT
ijassa-93	34	1	however	however	ADV
ijassa-93	34	2	,	,	PUNCT
ijassa-93	34	3	the	the	DET
ijassa-93	34	4	system	system	NOUN
ijassa-93	34	5	in	in	ADP
ijassa-93	34	6	literature	literature	NOUN
ijassa-93	34	7	[	[	X
ijassa-93	34	8	6	6	NUM
ijassa-93	34	9	]	]	PUNCT
ijassa-93	34	10	takes	take	VERB
ijassa-93	34	11	further	further	ADJ
ijassa-93	34	12	consideration	consideration	NOUN
ijassa-93	34	13	on	on	ADP
ijassa-93	34	14	texture	texture	ADJ
ijassa-93	34	15	feature	feature	NOUN
ijassa-93	34	16	in	in	ADP
ijassa-93	34	17	title	title	NOUN
ijassa-93	34	18	character	character	NOUN
ijassa-93	34	19	,	,	PUNCT
ijassa-93	34	20	and	and	CCONJ
ijassa-93	34	21	tracking	tracking	NOUN
ijassa-93	34	22	and	and	CCONJ
ijassa-93	34	23	conforming	conforming	NOUN
ijassa-93	34	24	in	in	ADP
ijassa-93	34	25	successive	successive	ADJ
ijassa-93	34	26	images	image	NOUN
ijassa-93	34	27	in	in	ADP
ijassa-93	34	28	multi	multi	ADJ
ijassa-93	34	29	frame	frame	NOUN
ijassa-93	34	30	,	,	PUNCT
ijassa-93	34	31	which	which	PRON
ijassa-93	34	32	makes	make	VERB
ijassa-93	34	33	it	it	PRON
ijassa-93	34	34	a	a	DET
ijassa-93	34	35	better	well	ADJ
ijassa-93	34	36	recognition	recognition	NOUN
ijassa-93	34	37	in	in	ADP
ijassa-93	34	38	ocr	ocr	PROPN
ijassa-93	34	39	.	.	PUNCT
ijassa-93	35	1	character	character	NOUN
ijassa-93	35	2	recognition	recognition	NOUN
ijassa-93	35	3	has	have	VERB
ijassa-93	35	4	a	a	DET
ijassa-93	35	5	close	close	ADJ
ijassa-93	35	6	connection	connection	NOUN
ijassa-93	35	7	with	with	ADP
ijassa-93	35	8	character	character	NOUN
ijassa-93	35	9	segmentation	segmentation	NOUN
ijassa-93	35	10	which	which	PRON
ijassa-93	35	11	is	be	AUX
ijassa-93	35	12	one	one	NUM
ijassa-93	35	13	of	of	ADP
ijassa-93	35	14	critical	critical	ADJ
ijassa-93	35	15	factors	factor	NOUN
ijassa-93	35	16	in	in	ADP
ijassa-93	35	17	character	character	NOUN
ijassa-93	35	18	recognition[7	recognition[7	NOUN
ijassa-93	35	19	]	]	PUNCT
ijassa-93	35	20	.	.	PUNCT
ijassa-93	36	1	since	since	SCONJ
ijassa-93	36	2	the	the	DET
ijassa-93	36	3	threshold	threshold	NOUN
ijassa-93	36	4	process	process	NOUN
ijassa-93	36	5	in	in	ADP
ijassa-93	36	6	natural	natural	ADJ
ijassa-93	36	7	scene	scene	NOUN
ijassa-93	36	8	image	image	NOUN
ijassa-93	36	9	tends	tend	VERB
ijassa-93	36	10	to	to	PART
ijassa-93	36	11	bring	bring	VERB
ijassa-93	36	12	about	about	ADP
ijassa-93	36	13	noise	noise	NOUN
ijassa-93	36	14	and	and	CCONJ
ijassa-93	36	15	low	low	ADJ
ijassa-93	36	16	quality	quality	NOUN
ijassa-93	36	17	problems	problem	NOUN
ijassa-93	36	18	,	,	PUNCT
ijassa-93	36	19	and	and	CCONJ
ijassa-93	36	20	probably	probably	ADV
ijassa-93	36	21	results	result	VERB
ijassa-93	36	22	in	in	ADP
ijassa-93	36	23	character	character	NOUN
ijassa-93	36	24	conglutination	conglutination	NOUN
ijassa-93	36	25	.	.	PUNCT
ijassa-93	37	1	such	such	ADJ
ijassa-93	37	2	traditional	traditional	ADJ
ijassa-93	37	3	character	character	NOUN
ijassa-93	37	4	segmentation	segmentation	NOUN
ijassa-93	37	5	will	will	AUX
ijassa-93	37	6	not	not	PART
ijassa-93	37	7	be	be	AUX
ijassa-93	37	8	satisfied	satisfied	ADJ
ijassa-93	37	9	with	with	ADP
ijassa-93	37	10	people	people	NOUN
ijassa-93	37	11	.	.	PUNCT
ijassa-93	38	1	in	in	ADP
ijassa-93	38	2	order	order	NOUN
ijassa-93	38	3	to	to	PART
ijassa-93	38	4	identify	identify	VERB
ijassa-93	38	5	characters	character	NOUN
ijassa-93	38	6	in	in	ADP
ijassa-93	38	7	the	the	DET
ijassa-93	38	8	natural	natural	ADJ
ijassa-93	38	9	scenes	scene	NOUN
ijassa-93	38	10	fast	fast	VERB
ijassa-93	38	11	and	and	CCONJ
ijassa-93	38	12	accurately	accurately	ADV
ijassa-93	38	13	,	,	PUNCT
ijassa-93	38	14	localizing	localize	VERB
ijassa-93	38	15	the	the	DET
ijassa-93	38	16	chunks	chunk	NOUN
ijassa-93	38	17	should	should	AUX
ijassa-93	38	18	be	be	AUX
ijassa-93	38	19	taken	take	VERB
ijassa-93	38	20	as	as	ADP
ijassa-93	38	21	first	first	ADJ
ijassa-93	38	22	priority	priority	NOUN
ijassa-93	38	23	,	,	PUNCT
ijassa-93	38	24	and	and	CCONJ
ijassa-93	38	25	later	later	ADV
ijassa-93	38	26	extracting	extract	VERB
ijassa-93	38	27	smaller	small	ADJ
ijassa-93	38	28	chunks	chunk	NOUN
ijassa-93	38	29	.	.	PUNCT
ijassa-93	39	1	therefore	therefore	ADV
ijassa-93	39	2	,	,	PUNCT
ijassa-93	39	3	the	the	DET
ijassa-93	39	4	small	small	ADJ
ijassa-93	39	5	chunk	chunk	NOUN
ijassa-93	39	6	could	could	AUX
ijassa-93	39	7	be	be	AUX
ijassa-93	39	8	used	use	VERB
ijassa-93	39	9	respectively	respectively	ADV
ijassa-93	39	10	to	to	PART
ijassa-93	39	11	sharpen	sharpen	VERB
ijassa-93	39	12	the	the	DET
ijassa-93	39	13	noise	noise	NOUN
ijassa-93	39	14	tremendously	tremendously	ADV
ijassa-93	39	15	,	,	PUNCT
ijassa-93	39	16	without	without	ADP
ijassa-93	39	17	the	the	DET
ijassa-93	39	18	complex	complex	ADJ
ijassa-93	39	19	impact	impact	NOUN
ijassa-93	39	20	from	from	ADP
ijassa-93	39	21	other	other	ADJ
ijassa-93	39	22	parts	part	NOUN
ijassa-93	39	23	in	in	ADP
ijassa-93	39	24	images.as	images.as	PRON
ijassa-93	39	25	a	a	DET
ijassa-93	39	26	result	result	NOUN
ijassa-93	39	27	,	,	PUNCT
ijassa-93	39	28	this	this	DET
ijassa-93	39	29	paper	paper	NOUN
ijassa-93	39	30	presents	present	VERB
ijassa-93	39	31	a	a	DET
ijassa-93	39	32	algorithm	algorithm	NOUN
ijassa-93	39	33	of	of	ADP
ijassa-93	39	34	scene	scene	NOUN
ijassa-93	39	35	character	character	NOUN
ijassa-93	39	36	segmentation	segmentation	NOUN
ijassa-93	39	37	and	and	CCONJ
ijassa-93	39	38	recognition	recognition	NOUN
ijassa-93	39	39	in	in	ADP
ijassa-93	39	40	natural	natural	ADJ
ijassa-93	39	41	scene	scene	NOUN
ijassa-93	39	42	images	image	NOUN
ijassa-93	39	43	,	,	PUNCT
ijassa-93	39	44	based	base	VERB
ijassa-93	39	45	on	on	ADP
ijassa-93	39	46	feature	feature	NOUN
ijassa-93	39	47	feedback	feedback	NOUN
ijassa-93	39	48	,	,	PUNCT
ijassa-93	39	49	which	which	PRON
ijassa-93	39	50	will	will	AUX
ijassa-93	39	51	segment	segment	VERB
ijassa-93	39	52	chinese	chinese	PROPN
ijassa-93	39	53	,	,	PUNCT
ijassa-93	39	54	english	english	ADJ
ijassa-93	39	55	and	and	CCONJ
ijassa-93	39	56	adherent	adherent	ADJ
ijassa-93	39	57	character	character	NOUN
ijassa-93	39	58	.	.	PUNCT
ijassa-93	40	1	after	after	ADP
ijassa-93	40	2	the	the	DET
ijassa-93	40	3	segmentation	segmentation	NOUN
ijassa-93	40	4	of	of	ADP
ijassa-93	40	5	character	character	NOUN
ijassa-93	40	6	,	,	PUNCT
ijassa-93	40	7	an	an	DET
ijassa-93	40	8	improved	improved	ADJ
ijassa-93	40	9	bp	bp	PROPN
ijassa-93	40	10	algorithm	algorithm	NOUN
ijassa-93	40	11	is	be	AUX
ijassa-93	40	12	afterwards	afterwards	ADV
ijassa-93	40	13	used	use	VERB
ijassa-93	40	14	to	to	PART
ijassa-93	40	15	recognize	recognize	VERB
ijassa-93	40	16	the	the	DET
ijassa-93	40	17	character	character	NOUN
ijassa-93	40	18	.	.	PUNCT
ijassa-93	41	1	advances	advance	NOUN
ijassa-93	41	2	in	in	ADP
ijassa-93	41	3	systems	system	NOUN
ijassa-93	41	4	science	science	NOUN
ijassa-93	41	5	and	and	CCONJ
ijassa-93	41	6	applications	application	NOUN
ijassa-93	41	7	(	(	PUNCT
ijassa-93	41	8	2012	2012	NUM
ijassa-93	41	9	)	)	PUNCT
ijassa-93	41	10	,	,	PUNCT
ijassa-93	41	11	vol.12	vol.12	NOUN
ijassa-93	41	12	,	,	PUNCT
ijassa-93	41	13	no.1	no.1	X
ijassa-93	41	14	69	69	NUM
ijassa-93	41	15	2	2	NUM
ijassa-93	41	16	feature	feature	NOUN
ijassa-93	41	17	feedback	feedback	NOUN
ijassa-93	41	18	based	base	VERB
ijassa-93	41	19	algorithm	algorithm	NOUN
ijassa-93	41	20	of	of	ADP
ijassa-93	41	21	character	character	NOUN
ijassa-93	41	22	segmentation	segmentation	NOUN
ijassa-93	41	23	&	&	CCONJ
ijassa-93	41	24	recognition	recognition	NOUN
ijassa-93	41	25	between	between	ADP
ijassa-93	41	26	chinese	chinese	PROPN
ijassa-93	41	27	,	,	PUNCT
ijassa-93	41	28	english	english	ADJ
ijassa-93	41	29	and	and	CCONJ
ijassa-93	41	30	adherent	adherent	ADJ
ijassa-93	41	31	character	character	NOUN
ijassa-93	41	32	several	several	ADJ
ijassa-93	41	33	traditional	traditional	ADJ
ijassa-93	41	34	methods	method	NOUN
ijassa-93	41	35	in	in	ADP
ijassa-93	41	36	character	character	NOUN
ijassa-93	41	37	segmentation	segmentation	NOUN
ijassa-93	41	38	&	&	CCONJ
ijassa-93	41	39	recognition[8	recognition[8	PROPN
ijassa-93	41	40	]	]	X
ijassa-93	41	41	1⃝	1⃝	NUM
ijassa-93	41	42	imagebased	imagebase	VERB
ijassa-93	41	43	analysis	analysis	NOUN
ijassa-93	41	44	.	.	PUNCT
ijassa-93	42	1	search	search	NOUN
ijassa-93	42	2	for	for	ADP
ijassa-93	42	3	rational	rational	ADJ
ijassa-93	42	4	segment	segment	NOUN
ijassa-93	42	5	point	point	NOUN
ijassa-93	42	6	between	between	ADP
ijassa-93	42	7	characters	character	NOUN
ijassa-93	42	8	,	,	PUNCT
ijassa-93	42	9	chiefly	chiefly	ADV
ijassa-93	42	10	adopt	adopt	VERB
ijassa-93	42	11	static	static	ADJ
ijassa-93	42	12	projection	projection	NOUN
ijassa-93	42	13	analysis	analysis	NOUN
ijassa-93	42	14	;	;	PUNCT
ijassa-93	42	15	2⃝	2⃝	NUM
ijassa-93	42	16	recognitionbased	recognitionbase	VERB
ijassa-93	42	17	.	.	PUNCT
ijassa-93	43	1	select	select	VERB
ijassa-93	43	2	various	various	ADJ
ijassa-93	43	3	kinds	kind	NOUN
ijassa-93	43	4	of	of	ADP
ijassa-93	43	5	current	current	ADJ
ijassa-93	43	6	segmentations	segmentation	NOUN
ijassa-93	43	7	via	via	ADP
ijassa-93	43	8	identifying	identify	VERB
ijassa-93	43	9	ability	ability	NOUN
ijassa-93	43	10	before	before	ADP
ijassa-93	43	11	actual	actual	ADJ
ijassa-93	43	12	segmentation	segmentation	NOUN
ijassa-93	43	13	.	.	PUNCT
ijassa-93	44	1	3⃝	3⃝	NUM
ijassa-93	44	2	synthetical	synthetical	NOUN
ijassa-93	44	3	based	base	VERB
ijassa-93	44	4	on	on	ADP
ijassa-93	44	5	image	image	NOUN
ijassa-93	44	6	analysis	analysis	NOUN
ijassa-93	44	7	and	and	CCONJ
ijassa-93	44	8	recognition	recognition	NOUN
ijassa-93	44	9	.	.	PUNCT
ijassa-93	45	1	reduce	reduce	VERB
ijassa-93	45	2	and	and	CCONJ
ijassa-93	45	3	filter	filter	NOUN
ijassa-93	45	4	vertical	vertical	ADJ
ijassa-93	45	5	segmentation	segmentation	NOUN
ijassa-93	45	6	hypothesis	hypothesis	NOUN
ijassa-93	45	7	by	by	ADP
ijassa-93	45	8	image	image	NOUN
ijassa-93	45	9	analysis	analysis	NOUN
ijassa-93	45	10	.	.	PUNCT
ijassa-93	46	1	4⃝	4⃝	NUM
ijassa-93	46	2	unity	unity	NOUN
ijassa-93	46	3	recognition	recognition	NOUN
ijassa-93	46	4	.	.	PUNCT
ijassa-93	47	1	use	use	VERB
ijassa-93	47	2	the	the	DET
ijassa-93	47	3	whole	whole	ADJ
ijassa-93	47	4	word	word	NOUN
ijassa-93	47	5	as	as	ADP
ijassa-93	47	6	recognizing	recognize	VERB
ijassa-93	47	7	object	object	NOUN
ijassa-93	47	8	,	,	PUNCT
ijassa-93	47	9	according	accord	VERB
ijassa-93	47	10	to	to	ADP
ijassa-93	47	11	the	the	DET
ijassa-93	47	12	features	feature	NOUN
ijassa-93	47	13	of	of	ADP
ijassa-93	47	14	the	the	DET
ijassa-93	47	15	whole	whole	ADJ
ijassa-93	47	16	word	word	NOUN
ijassa-93	47	17	,	,	PUNCT
ijassa-93	47	18	to	to	PART
ijassa-93	47	19	avoid	avoid	VERB
ijassa-93	47	20	segmentation	segmentation	NOUN
ijassa-93	47	21	harm	harm	NOUN
ijassa-93	47	22	on	on	ADP
ijassa-93	47	23	character	character	NOUN
ijassa-93	47	24	.	.	PUNCT
ijassa-93	48	1	however	however	ADV
ijassa-93	48	2	,	,	PUNCT
ijassa-93	48	3	in	in	ADP
ijassa-93	48	4	the	the	DET
ijassa-93	48	5	case	case	NOUN
ijassa-93	48	6	of	of	ADP
ijassa-93	48	7	word	word	NOUN
ijassa-93	48	8	conglutination	conglutination	NOUN
ijassa-93	48	9	,	,	PUNCT
ijassa-93	48	10	those	those	DET
ijassa-93	48	11	methods	method	NOUN
ijassa-93	48	12	can	can	AUX
ijassa-93	48	13	not	not	PART
ijassa-93	48	14	segment	segment	VERB
ijassa-93	48	15	adherent	adherent	ADJ
ijassa-93	48	16	characters	character	NOUN
ijassa-93	48	17	accurately	accurately	ADV
ijassa-93	48	18	.	.	PUNCT
ijassa-93	49	1	this	this	DET
ijassa-93	49	2	paper	paper	NOUN
ijassa-93	49	3	deals	deal	VERB
ijassa-93	49	4	with	with	ADP
ijassa-93	49	5	this	this	DET
ijassa-93	49	6	kind	kind	NOUN
ijassa-93	49	7	of	of	ADP
ijassa-93	49	8	situation	situation	NOUN
ijassa-93	49	9	by	by	ADP
ijassa-93	49	10	an	an	DET
ijassa-93	49	11	algorithm	algorithm	NOUN
ijassa-93	49	12	of	of	ADP
ijassa-93	49	13	feature	feature	NOUN
ijassa-93	49	14	feedbackbased	feedbackbase	VERB
ijassa-93	49	15	character	character	NOUN
ijassa-93	49	16	segmentation	segmentation	NOUN
ijassa-93	49	17	and	and	CCONJ
ijassa-93	49	18	recognition	recognition	NOUN
ijassa-93	49	19	between	between	ADP
ijassa-93	49	20	chinese	chinese	PROPN
ijassa-93	49	21	,	,	PUNCT
ijassa-93	49	22	english	english	ADJ
ijassa-93	49	23	and	and	CCONJ
ijassa-93	49	24	adherent	adherent	ADJ
ijassa-93	49	25	characters	character	NOUN
ijassa-93	49	26	.	.	PUNCT
ijassa-93	50	1	first	first	ADJ
ijassa-93	50	2	segment	segment	VERB
ijassa-93	50	3	each	each	DET
ijassa-93	50	4	line	line	NOUN
ijassa-93	50	5	of	of	ADP
ijassa-93	50	6	character	character	NOUN
ijassa-93	50	7	,	,	PUNCT
ijassa-93	50	8	and	and	CCONJ
ijassa-93	50	9	then	then	ADV
ijassa-93	50	10	search	search	VERB
ijassa-93	50	11	and	and	CCONJ
ijassa-93	50	12	extract	extract	VERB
ijassa-93	50	13	the	the	DET
ijassa-93	50	14	adherent	adherent	ADJ
ijassa-93	50	15	character	character	NOUN
ijassa-93	50	16	in	in	ADP
ijassa-93	50	17	each	each	DET
ijassa-93	50	18	line	line	NOUN
ijassa-93	50	19	,	,	PUNCT
ijassa-93	50	20	finally	finally	ADV
ijassa-93	50	21	transfer	transfer	VERB
ijassa-93	50	22	small	small	ADJ
ijassa-93	50	23	segmenting	segmenting	NOUN
ijassa-93	50	24	module	module	NOUN
ijassa-93	50	25	to	to	PART
ijassa-93	50	26	discover	discover	VERB
ijassa-93	50	27	the	the	DET
ijassa-93	50	28	most	most	ADV
ijassa-93	50	29	reliable	reliable	ADJ
ijassa-93	50	30	segment	segment	NOUN
ijassa-93	50	31	point	point	NOUN
ijassa-93	50	32	,	,	PUNCT
ijassa-93	50	33	so	so	SCONJ
ijassa-93	50	34	as	as	SCONJ
ijassa-93	50	35	to	to	PART
ijassa-93	50	36	achieve	achieve	VERB
ijassa-93	50	37	the	the	DET
ijassa-93	50	38	segmentation	segmentation	NOUN
ijassa-93	50	39	purpose.key	purpose.key	ADJ
ijassa-93	50	40	points	point	NOUN
ijassa-93	50	41	as	as	SCONJ
ijassa-93	50	42	follows	follow	VERB
ijassa-93	50	43	:	:	PUNCT
ijassa-93	50	44	1.identify	1.identify	NUM
ijassa-93	50	45	the	the	DET
ijassa-93	50	46	category	category	NOUN
ijassa-93	50	47	of	of	ADP
ijassa-93	50	48	adherent	adherent	ADJ
ijassa-93	50	49	characters	character	NOUN
ijassa-93	50	50	after	after	ADP
ijassa-93	50	51	initial	initial	ADJ
ijassa-93	50	52	identification	identification	NOUN
ijassa-93	50	53	,	,	PUNCT
ijassa-93	50	54	the	the	DET
ijassa-93	50	55	recognition	recognition	NOUN
ijassa-93	50	56	and	and	CCONJ
ijassa-93	50	57	length	length	NOUN
ijassa-93	50	58	of	of	ADP
ijassa-93	50	59	each	each	DET
ijassa-93	50	60	image	image	NOUN
ijassa-93	50	61	as	as	SCONJ
ijassa-93	50	62	follows	follow	VERB
ijassa-93	50	63	:	:	PUNCT
ijassa-93	50	64	1⃝	1⃝	NUM
ijassa-93	50	65	suppose	suppose	VERB
ijassa-93	50	66	the	the	DET
ijassa-93	50	67	length	length	NOUN
ijassa-93	50	68	of	of	ADP
ijassa-93	50	69	each	each	DET
ijassa-93	50	70	character	character	NOUN
ijassa-93	50	71	image	image	NOUN
ijassa-93	50	72	is	be	AUX
ijassa-93	50	73	w	w	NOUN
ijassa-93	50	74	,	,	PUNCT
ijassa-93	50	75	and	and	CCONJ
ijassa-93	50	76	the	the	DET
ijassa-93	50	77	average	average	ADJ
ijassa-93	50	78	length	length	NOUN
ijassa-93	50	79	of	of	ADP
ijassa-93	50	80	each	each	DET
ijassa-93	50	81	character	character	NOUN
ijassa-93	50	82	image	image	NOUN
ijassa-93	50	83	is	be	AUX
ijassa-93	50	84	w.	w.	NOUN
ijassa-93	50	85	when	when	SCONJ
ijassa-93	50	86	w	w	NOUN
ijassa-93	50	87	is	be	AUX
ijassa-93	50	88	much	much	ADV
ijassa-93	50	89	wider	wide	ADJ
ijassa-93	50	90	than	than	ADP
ijassa-93	50	91	wv	wv	PROPN
ijassa-93	50	92	,	,	PUNCT
ijassa-93	50	93	and	and	CCONJ
ijassa-93	50	94	pick	pick	VERB
ijassa-93	50	95	up	up	ADP
ijassa-93	50	96	a	a	DET
ijassa-93	50	97	threshold	threshold	NOUN
ijassa-93	50	98	via	via	ADP
ijassa-93	50	99	experiment	experiment	NOUN
ijassa-93	50	100	,	,	PUNCT
ijassa-93	50	101	we	we	PRON
ijassa-93	50	102	could	could	AUX
ijassa-93	50	103	identify	identify	VERB
ijassa-93	50	104	whether	whether	SCONJ
ijassa-93	50	105	it	it	PRON
ijassa-93	50	106	is	be	AUX
ijassa-93	50	107	adherent	adherent	ADJ
ijassa-93	50	108	character	character	NOUN
ijassa-93	50	109	or	or	CCONJ
ijassa-93	50	110	not	not	PART
ijassa-93	50	111	.	.	PUNCT
ijassa-93	51	1	2⃝	2⃝	NUM
ijassa-93	51	2	after	after	ADP
ijassa-93	51	3	identifying	identify	VERB
ijassa-93	51	4	adherent	adherent	ADJ
ijassa-93	51	5	character	character	NOUN
ijassa-93	51	6	image	image	NOUN
ijassa-93	51	7	,	,	PUNCT
ijassa-93	51	8	according	accord	VERB
ijassa-93	51	9	to	to	ADP
ijassa-93	51	10	english	english	ADJ
ijassa-93	51	11	character	character	NOUN
ijassa-93	51	12	image	image	NOUN
ijassa-93	51	13	length	length	NOUN
ijassa-93	51	14	is	be	AUX
ijassa-93	51	15	wider	wide	ADJ
ijassa-93	51	16	than	than	ADP
ijassa-93	51	17	the	the	DET
ijassa-93	51	18	chinese	chinese	ADJ
ijassa-93	51	19	character	character	NOUN
ijassa-93	51	20	image	image	NOUN
ijassa-93	51	21	length	length	NOUN
ijassa-93	51	22	,	,	PUNCT
ijassa-93	51	23	and	and	CCONJ
ijassa-93	51	24	the	the	DET
ijassa-93	51	25	separation	separation	NOUN
ijassa-93	51	26	distance	distance	NOUN
ijassa-93	51	27	is	be	AUX
ijassa-93	51	28	small	small	ADJ
ijassa-93	51	29	between	between	ADP
ijassa-93	51	30	the	the	DET
ijassa-93	51	31	adjoining	adjoining	ADJ
ijassa-93	51	32	characters	character	NOUN
ijassa-93	51	33	,	,	PUNCT
ijassa-93	51	34	we	we	PRON
ijassa-93	51	35	could	could	AUX
ijassa-93	51	36	identify	identify	VERB
ijassa-93	51	37	whether	whether	SCONJ
ijassa-93	51	38	the	the	DET
ijassa-93	51	39	adherent	adherent	ADJ
ijassa-93	51	40	character	character	NOUN
ijassa-93	51	41	image	image	NOUN
ijassa-93	51	42	is	be	AUX
ijassa-93	51	43	chinese	chinese	ADJ
ijassa-93	51	44	or	or	CCONJ
ijassa-93	51	45	english	english	ADJ
ijassa-93	51	46	;	;	PUNCT
ijassa-93	51	47	according	accord	VERB
ijassa-93	51	48	to	to	ADP
ijassa-93	51	49	average	average	ADJ
ijassa-93	51	50	english	english	ADJ
ijassa-93	51	51	adherent	adherent	NOUN
ijassa-93	51	52	character	character	NOUN
ijassa-93	51	53	length	length	NOUN
ijassa-93	51	54	is	be	AUX
ijassa-93	51	55	shorter	short	ADJ
ijassa-93	51	56	than	than	ADP
ijassa-93	51	57	the	the	DET
ijassa-93	51	58	chinese	chinese	ADJ
ijassa-93	51	59	one	one	NOUN
ijassa-93	51	60	,	,	PUNCT
ijassa-93	51	61	we	we	PRON
ijassa-93	51	62	could	could	AUX
ijassa-93	51	63	identify	identify	VERB
ijassa-93	51	64	those	those	DET
ijassa-93	51	65	character	character	NOUN
ijassa-93	51	66	images	image	NOUN
ijassa-93	51	67	with	with	ADP
ijassa-93	51	68	wrong	wrong	ADJ
ijassa-93	51	69	structure	structure	NOUN
ijassa-93	51	70	as	as	ADP
ijassa-93	51	71	adherent	adherent	ADJ
ijassa-93	51	72	character	character	NOUN
ijassa-93	51	73	image	image	NOUN
ijassa-93	51	74	.	.	PUNCT
ijassa-93	52	1	2.identify	2.identify	NUM
ijassa-93	52	2	the	the	DET
ijassa-93	52	3	category	category	NOUN
ijassa-93	52	4	of	of	ADP
ijassa-93	52	5	adherent	adherent	ADJ
ijassa-93	52	6	characters	character	NOUN
ijassa-93	52	7	for	for	ADP
ijassa-93	52	8	the	the	DET
ijassa-93	52	9	adherent	adherent	ADJ
ijassa-93	52	10	chinese	chinese	ADJ
ijassa-93	52	11	character	character	NOUN
ijassa-93	52	12	,	,	PUNCT
ijassa-93	52	13	we	we	PRON
ijassa-93	52	14	can	can	AUX
ijassa-93	52	15	segment	segment	VERB
ijassa-93	52	16	the	the	DET
ijassa-93	52	17	chinese	chinese	ADJ
ijassa-93	52	18	character	character	NOUN
ijassa-93	52	19	by	by	ADP
ijassa-93	52	20	judging	judge	VERB
ijassa-93	52	21	the	the	DET
ijassa-93	52	22	potential	potential	ADJ
ijassa-93	52	23	character	character	NOUN
ijassa-93	52	24	image	image	NOUN
ijassa-93	52	25	length	length	NOUN
ijassa-93	52	26	.	.	PUNCT
ijassa-93	53	1	we	we	PRON
ijassa-93	53	2	use	use	VERB
ijassa-93	53	3	identifying	identify	VERB
ijassa-93	53	4	module	module	NOUN
ijassa-93	53	5	to	to	PART
ijassa-93	53	6	judge	judge	VERB
ijassa-93	53	7	the	the	DET
ijassa-93	53	8	border	border	NOUN
ijassa-93	53	9	to	to	PART
ijassa-93	53	10	segment	segment	VERB
ijassa-93	53	11	the	the	DET
ijassa-93	53	12	character	character	NOUN
ijassa-93	53	13	.	.	PUNCT
ijassa-93	54	1	in	in	ADP
ijassa-93	54	2	this	this	DET
ijassa-93	54	3	process	process	NOUN
ijassa-93	54	4	,	,	PUNCT
ijassa-93	54	5	we	we	PRON
ijassa-93	54	6	confirm	confirm	VERB
ijassa-93	54	7	the	the	DET
ijassa-93	54	8	most	most	ADV
ijassa-93	54	9	reliable	reliable	ADJ
ijassa-93	54	10	segmenting	segmenting	NOUN
ijassa-93	54	11	point	point	NOUN
ijassa-93	54	12	by	by	ADP
ijassa-93	54	13	right	right	NOUN
ijassa-93	54	14	-	-	PUNCT
ijassa-93	54	15	to	to	AUX
ijassa-93	54	16	-	-	PUNCT
ijassa-93	54	17	left	left	ADJ
ijassa-93	54	18	and	and	CCONJ
ijassa-93	54	19	left	leave	VERB
ijassa-93	54	20	-	-	PUNCT
ijassa-93	54	21	to	to	ADP
ijassa-93	54	22	-	-	PUNCT
ijassa-93	54	23	right	right	NOUN
ijassa-93	54	24	searching	search	VERB
ijassa-93	54	25	methods.algorithm	methods.algorithm	PROPN
ijassa-93	54	26	as	as	SCONJ
ijassa-93	54	27	follows	follow	VERB
ijassa-93	54	28	:	:	PUNCT
ijassa-93	54	29	step1	step1	PROPN
ijassa-93	54	30	,	,	PUNCT
ijassa-93	54	31	record	record	VERB
ijassa-93	54	32	the	the	DET
ijassa-93	54	33	value	value	NOUN
ijassa-93	54	34	of	of	ADP
ijassa-93	54	35	left	left	ADJ
ijassa-93	54	36	border	border	NOUN
ijassa-93	54	37	bl	bl	PROPN
ijassa-93	54	38	,	,	PUNCT
ijassa-93	54	39	right	right	ADJ
ijassa-93	54	40	border	border	NOUN
ijassa-93	54	41	br	br	NOUN
ijassa-93	54	42	and	and	CCONJ
ijassa-93	54	43	line	line	NOUN
ijassa-93	54	44	height	height	NOUN
ijassa-93	54	45	hl	hl	PROPN
ijassa-93	54	46	.	.	PUNCT
ijassa-93	55	1	step2	step2	PROPN
ijassa-93	55	2	,	,	PUNCT
ijassa-93	55	3	cobfirm	cobfirm	ADJ
ijassa-93	55	4	x0	x0	PROPN
ijassa-93	55	5	value	value	NOUN
ijassa-93	55	6	(	(	PUNCT
ijassa-93	55	7	x0	x0	PROPN
ijassa-93	55	8	=	=	PUNCT
ijassa-93	55	9	hl)from	hl)from	ADP
ijassa-93	55	10	the	the	DET
ijassa-93	55	11	left	left	ADJ
ijassa-93	55	12	border	border	NOUN
ijassa-93	55	13	,	,	PUNCT
ijassa-93	55	14	choose	choose	VERB
ijassa-93	55	15	one	one	NUM
ijassa-93	55	16	threshold	threshold	NOUN
ijassa-93	55	17	value	value	NOUN
ijassa-93	55	18	t0	t0	PROPN
ijassa-93	55	19	,	,	PUNCT
ijassa-93	55	20	the	the	DET
ijassa-93	55	21	segmenting	segmenting	NOUN
ijassa-93	55	22	point	point	NOUN
ijassa-93	55	23	should	should	AUX
ijassa-93	55	24	be	be	AUX
ijassa-93	55	25	picked	pick	VERB
ijassa-93	55	26	up	up	ADP
ijassa-93	55	27	from	from	ADP
ijassa-93	55	28	(	(	PUNCT
ijassa-93	55	29	x0	x0	PROPN
ijassa-93	55	30	-	-	PUNCT
ijassa-93	55	31	t0	t0	PROPN
ijassa-93	55	32	,	,	PUNCT
ijassa-93	55	33	x0+t0	x0+t0	PROPN
ijassa-93	55	34	)	)	PUNCT
ijassa-93	55	35	,	,	PUNCT
ijassa-93	55	36	modulate	modulate	VERB
ijassa-93	55	37	it	it	PRON
ijassa-93	55	38	according	accord	VERB
ijassa-93	55	39	to	to	ADP
ijassa-93	55	40	steplong	steplong	PROPN
ijassa-93	55	41	.	.	PUNCT
ijassa-93	56	1	70	70	NUM
ijassa-93	56	2	jinghua	jinghua	PROPN
ijassa-93	56	3	hu	hu	PROPN
ijassa-93	56	4	:	:	PUNCT
ijassa-93	56	5	the	the	DET
ijassa-93	56	6	research	research	NOUN
ijassa-93	56	7	on	on	ADP
ijassa-93	56	8	effective	effective	ADJ
ijassa-93	56	9	video	video	NOUN
ijassa-93	56	10	scene	scene	NOUN
ijassa-93	56	11	character	character	NOUN
ijassa-93	56	12	extraction	extraction	NOUN
ijassa-93	56	13	algorithm	algorithm	NOUN
ijassa-93	56	14	...	...	PUNCT
ijassa-93	57	1	step3	step3	NOUN
ijassa-93	57	2	,	,	PUNCT
ijassa-93	57	3	transfer	transfer	VERB
ijassa-93	57	4	identifying	identifying	NOUN
ijassa-93	57	5	module	module	NOUN
ijassa-93	57	6	,	,	PUNCT
ijassa-93	57	7	and	and	CCONJ
ijassa-93	57	8	judge	judge	VERB
ijassa-93	57	9	whether	whether	SCONJ
ijassa-93	57	10	the	the	DET
ijassa-93	57	11	result	result	NOUN
ijassa-93	57	12	is	be	AUX
ijassa-93	57	13	in	in	ADP
ijassa-93	57	14	the	the	DET
ijassa-93	57	15	range	range	NOUN
ijassa-93	57	16	or	or	CCONJ
ijassa-93	57	17	not	not	PART
ijassa-93	57	18	.	.	PUNCT
ijassa-93	58	1	if	if	SCONJ
ijassa-93	58	2	not	not	PART
ijassa-93	58	3	,	,	PUNCT
ijassa-93	58	4	do	do	AUX
ijassa-93	58	5	step2	step2	PROPN
ijassa-93	58	6	;	;	PUNCT
ijassa-93	58	7	if	if	SCONJ
ijassa-93	58	8	is	be	AUX
ijassa-93	58	9	,	,	PUNCT
ijassa-93	58	10	do	do	AUX
ijassa-93	58	11	step	step	VERB
ijassa-93	58	12	4	4	NUM
ijassa-93	58	13	.	.	PUNCT
ijassa-93	58	14	step	step	NOUN
ijassa-93	58	15	4	4	NUM
ijassa-93	58	16	,	,	PUNCT
ijassa-93	58	17	save	save	VERB
ijassa-93	58	18	segmenting	segmenting	NOUN
ijassa-93	58	19	result	result	PROPN
ijassa-93	58	20	br1	br1	PROPN
ijassa-93	58	21	,	,	PUNCT
ijassa-93	58	22	take	take	VERB
ijassa-93	58	23	br1	br1	PROPN
ijassa-93	58	24	as	as	ADP
ijassa-93	58	25	the	the	DET
ijassa-93	58	26	left	left	ADJ
ijassa-93	58	27	border	border	NOUN
ijassa-93	58	28	,	,	PUNCT
ijassa-93	58	29	and	and	CCONJ
ijassa-93	58	30	then	then	ADV
ijassa-93	58	31	do	do	VERB
ijassa-93	58	32	step2	step2	PROPN
ijassa-93	58	33	and	and	CCONJ
ijassa-93	58	34	the	the	DET
ijassa-93	58	35	following	follow	VERB
ijassa-93	58	36	steps	step	NOUN
ijassa-93	58	37	successively	successively	ADV
ijassa-93	58	38	,	,	PUNCT
ijassa-93	58	39	till	till	SCONJ
ijassa-93	58	40	the	the	DET
ijassa-93	58	41	right	right	ADJ
ijassa-93	58	42	border	border	NOUN
ijassa-93	58	43	br	br	NOUN
ijassa-93	58	44	.	.	PUNCT
ijassa-93	59	1	to	to	PART
ijassa-93	59	2	compare	compare	VERB
ijassa-93	59	3	the	the	DET
ijassa-93	59	4	validity	validity	NOUN
ijassa-93	59	5	of	of	ADP
ijassa-93	59	6	the	the	DET
ijassa-93	59	7	final	final	ADJ
ijassa-93	59	8	segmenting	segmenting	NOUN
ijassa-93	59	9	result	result	NOUN
ijassa-93	59	10	,	,	PUNCT
ijassa-93	59	11	do	do	VERB
ijassa-93	59	12	all	all	DET
ijassa-93	59	13	steps	step	NOUN
ijassa-93	59	14	in	in	ADP
ijassa-93	59	15	turns	turn	NOUN
ijassa-93	59	16	from	from	ADP
ijassa-93	59	17	the	the	DET
ijassa-93	59	18	right	right	ADJ
ijassa-93	59	19	border	border	NOUN
ijassa-93	59	20	,	,	PUNCT
ijassa-93	59	21	and	and	CCONJ
ijassa-93	59	22	save	save	VERB
ijassa-93	59	23	the	the	DET
ijassa-93	59	24	best	good	ADJ
ijassa-93	59	25	segmenting	segmenting	NOUN
ijassa-93	59	26	result	result	NOUN
ijassa-93	59	27	.	.	PUNCT
ijassa-93	60	1	3.segment	3.segment	NUM
ijassa-93	60	2	adherent	adherent	ADJ
ijassa-93	60	3	english	english	ADJ
ijassa-93	60	4	number	number	NOUN
ijassa-93	60	5	for	for	ADP
ijassa-93	60	6	the	the	DET
ijassa-93	60	7	adherent	adherent	ADJ
ijassa-93	60	8	english	english	ADJ
ijassa-93	60	9	number	number	NOUN
ijassa-93	60	10	,	,	PUNCT
ijassa-93	60	11	first	first	ADV
ijassa-93	60	12	use	use	VERB
ijassa-93	60	13	border	border	NOUN
ijassa-93	60	14	searching	search	VERB
ijassa-93	60	15	algorithm	algorithm	NOUN
ijassa-93	60	16	to	to	PART
ijassa-93	60	17	segment	segment	VERB
ijassa-93	60	18	these	these	DET
ijassa-93	60	19	adherent	adherent	ADJ
ijassa-93	60	20	character	character	NOUN
ijassa-93	60	21	with	with	ADP
ijassa-93	60	22	gaps	gap	NOUN
ijassa-93	60	23	in	in	ADP
ijassa-93	60	24	between	between	ADP
ijassa-93	60	25	,	,	PUNCT
ijassa-93	60	26	which	which	PRON
ijassa-93	60	27	can	can	AUX
ijassa-93	60	28	not	not	PART
ijassa-93	60	29	be	be	AUX
ijassa-93	60	30	segmented	segment	VERB
ijassa-93	60	31	by	by	ADP
ijassa-93	60	32	projection	projection	NOUN
ijassa-93	60	33	method	method	NOUN
ijassa-93	60	34	.	.	PUNCT
ijassa-93	61	1	then	then	ADV
ijassa-93	61	2	suppose	suppose	VERB
ijassa-93	61	3	the	the	DET
ijassa-93	61	4	potential	potential	ADJ
ijassa-93	61	5	height	height	NOUN
ijassa-93	61	6	of	of	ADP
ijassa-93	61	7	each	each	DET
ijassa-93	61	8	adherent	adherent	NOUN
ijassa-93	61	9	character	character	NOUN
ijassa-93	61	10	string	string	NOUN
ijassa-93	61	11	is	be	AUX
ijassa-93	61	12	x	x	PRON
ijassa-93	61	13	,	,	PUNCT
ijassa-93	61	14	according	accord	VERB
ijassa-93	61	15	to	to	ADP
ijassa-93	61	16	x	x	PRON
ijassa-93	61	17	,	,	PUNCT
ijassa-93	61	18	transfer	transfer	VERB
ijassa-93	61	19	the	the	DET
ijassa-93	61	20	identifying	identify	VERB
ijassa-93	61	21	module	module	NOUN
ijassa-93	61	22	to	to	PART
ijassa-93	61	23	segment	segment	VERB
ijassa-93	61	24	the	the	DET
ijassa-93	61	25	adherent	adherent	NOUN
ijassa-93	61	26	character	character	NOUN
ijassa-93	61	27	string	string	NOUN
ijassa-93	61	28	so	so	SCONJ
ijassa-93	61	29	as	as	SCONJ
ijassa-93	61	30	to	to	PART
ijassa-93	61	31	locate	locate	VERB
ijassa-93	61	32	the	the	DET
ijassa-93	61	33	segmenting	segmenting	NOUN
ijassa-93	61	34	point	point	NOUN
ijassa-93	61	35	.	.	PUNCT
ijassa-93	62	1	border	border	NOUN
ijassa-93	62	2	searching	searching	NOUN
ijassa-93	62	3	algorithm	algorithm	NOUN
ijassa-93	62	4	as	as	SCONJ
ijassa-93	62	5	follows	follow	VERB
ijassa-93	62	6	:	:	PUNCT
ijassa-93	62	7	step1	step1	PROPN
ijassa-93	62	8	,	,	PUNCT
ijassa-93	62	9	take	take	VERB
ijassa-93	62	10	note	note	NOUN
ijassa-93	62	11	of	of	ADP
ijassa-93	62	12	left	left	ADJ
ijassa-93	62	13	rect	rect	PROPN
ijassa-93	62	14	urlli	urlli	PROPN
ijassa-93	62	15	,	,	PUNCT
ijassa-93	62	16	tiand	tiand	VERB
ijassa-93	62	17	right	right	ADJ
ijassa-93	62	18	rect	rect	PROPN
ijassa-93	62	19	urlri	urlri	NOUN
ijassa-93	62	20	,	,	PUNCT
ijassa-93	62	21	bi	bi	NOUN
ijassa-93	62	22	,	,	PUNCT
ijassa-93	62	23	and	and	CCONJ
ijassa-93	62	24	pick	pick	VERB
ijassa-93	62	25	up	up	ADP
ijassa-93	62	26	a	a	DET
ijassa-93	62	27	threshold	threshold	NOUN
ijassa-93	62	28	value	value	NOUN
ijassa-93	62	29	t	t	PROPN
ijassa-93	62	30	,	,	PUNCT
ijassa-93	62	31	find	find	VERB
ijassa-93	62	32	a	a	DET
ijassa-93	62	33	p	p	NOUN
ijassa-93	62	34	point	point	NOUN
ijassa-93	62	35	whose	whose	DET
ijassa-93	62	36	gray	gray	ADJ
ijassa-93	62	37	-	-	PUNCT
ijassa-93	62	38	scale	scale	NOUN
ijassa-93	62	39	value	value	NOUN
ijassa-93	62	40	is	be	AUX
ijassa-93	62	41	1	1	NUM
ijassa-93	62	42	,	,	PUNCT
ijassa-93	62	43	and	and	CCONJ
ijassa-93	62	44	take	take	VERB
ijassa-93	62	45	it	it	PRON
ijassa-93	62	46	as	as	ADP
ijassa-93	62	47	initial	initial	ADJ
ijassa-93	62	48	point	point	NOUN
ijassa-93	62	49	of	of	ADP
ijassa-93	62	50	the	the	DET
ijassa-93	62	51	contour	contour	NOUN
ijassa-93	62	52	line	line	NOUN
ijassa-93	62	53	.	.	PUNCT
ijassa-93	63	1	step2	step2	PROPN
ijassa-93	63	2	,	,	PUNCT
ijassa-93	63	3	start	start	VERB
ijassa-93	63	4	with	with	ADP
ijassa-93	63	5	the	the	DET
ijassa-93	63	6	fund	fund	NOUN
ijassa-93	63	7	point	point	NOUN
ijassa-93	63	8	and	and	CCONJ
ijassa-93	63	9	continue	continue	VERB
ijassa-93	63	10	with	with	ADP
ijassa-93	63	11	the	the	DET
ijassa-93	63	12	searching	search	VERB
ijassa-93	63	13	job	job	NOUN
ijassa-93	63	14	.	.	PUNCT
ijassa-93	64	1	if	if	SCONJ
ijassa-93	64	2	the	the	DET
ijassa-93	64	3	gray	gray	ADJ
ijassa-93	64	4	-	-	PUNCT
ijassa-93	64	5	scale	scale	NOUN
ijassa-93	64	6	value	value	NOUN
ijassa-93	64	7	is	be	AUX
ijassa-93	64	8	1	1	NUM
ijassa-93	64	9	,	,	PUNCT
ijassa-93	64	10	please	please	INTJ
ijassa-93	64	11	carry	carry	VERB
ijassa-93	64	12	on	on	ADP
ijassa-93	64	13	to	to	ADP
ijassa-93	64	14	the	the	DET
ijassa-93	64	15	left	left	NOUN
ijassa-93	64	16	;	;	PUNCT
ijassa-93	64	17	if	if	SCONJ
ijassa-93	64	18	the	the	DET
ijassa-93	64	19	gray	gray	ADJ
ijassa-93	64	20	-	-	PUNCT
ijassa-93	64	21	scale	scale	NOUN
ijassa-93	64	22	value	value	NOUN
ijassa-93	64	23	is	be	AUX
ijassa-93	64	24	0	0	NUM
ijassa-93	64	25	,	,	PUNCT
ijassa-93	64	26	please	please	INTJ
ijassa-93	64	27	carry	carry	VERB
ijassa-93	64	28	on	on	ADP
ijassa-93	64	29	from	from	ADP
ijassa-93	64	30	the	the	DET
ijassa-93	64	31	right	right	ADJ
ijassa-93	64	32	turn	turn	NOUN
ijassa-93	64	33	.	.	PUNCT
ijassa-93	65	1	once	once	ADV
ijassa-93	65	2	meet	meet	VERB
ijassa-93	65	3	the	the	DET
ijassa-93	65	4	point	point	NOUN
ijassa-93	65	5	whose	whose	DET
ijassa-93	65	6	gray	gray	ADJ
ijassa-93	65	7	-	-	PUNCT
ijassa-93	65	8	scale	scale	NOUN
ijassa-93	65	9	value	value	NOUN
ijassa-93	65	10	is	be	AUX
ijassa-93	65	11	1	1	NUM
ijassa-93	65	12	,	,	PUNCT
ijassa-93	65	13	these	these	DET
ijassa-93	65	14	points	point	NOUN
ijassa-93	65	15	would	would	AUX
ijassa-93	65	16	be	be	AUX
ijassa-93	65	17	contour	contour	NOUN
ijassa-93	65	18	point	point	NOUN
ijassa-93	65	19	.	.	PUNCT
ijassa-93	66	1	step3	step3	PROPN
ijassa-93	66	2	,	,	PUNCT
ijassa-93	66	3	repeat	repeat	NOUN
ijassa-93	66	4	step	step	NOUN
ijassa-93	66	5	2	2	NUM
ijassa-93	66	6	till	till	SCONJ
ijassa-93	66	7	the	the	DET
ijassa-93	66	8	contour	contour	NOUN
ijassa-93	66	9	line	line	NOUN
ijassa-93	66	10	point	point	NOUN
ijassa-93	66	11	coincides	coincide	VERB
ijassa-93	66	12	with	with	ADP
ijassa-93	66	13	the	the	DET
ijassa-93	66	14	initial	initial	ADJ
ijassa-93	66	15	point	point	NOUN
ijassa-93	66	16	;	;	PUNCT
ijassa-93	66	17	step4	step4	PROPN
ijassa-93	66	18	,	,	PUNCT
ijassa-93	66	19	confirm	confirm	VERB
ijassa-93	66	20	the	the	DET
ijassa-93	66	21	right	right	ADJ
ijassa-93	66	22	border	border	NOUN
ijassa-93	66	23	via	via	ADP
ijassa-93	66	24	contour	contour	NOUN
ijassa-93	66	25	line	line	NOUN
ijassa-93	66	26	searching	searching	NOUN
ijassa-93	66	27	.	.	PUNCT
ijassa-93	67	1	if	if	SCONJ
ijassa-93	67	2	br1	br1	PROPN
ijassa-93	67	3	coincides	coincide	VERB
ijassa-93	67	4	with	with	ADP
ijassa-93	67	5	the	the	DET
ijassa-93	67	6	right	right	ADJ
ijassa-93	67	7	border	border	NOUN
ijassa-93	67	8	,	,	PUNCT
ijassa-93	67	9	there	there	PRON
ijassa-93	67	10	is	be	VERB
ijassa-93	67	11	no	no	DET
ijassa-93	67	12	adherent	adherent	ADJ
ijassa-93	67	13	character	character	NOUN
ijassa-93	67	14	;	;	PUNCT
ijassa-93	67	15	if	if	SCONJ
ijassa-93	67	16	not	not	PART
ijassa-93	67	17	,	,	PUNCT
ijassa-93	67	18	for	for	ADP
ijassa-93	67	19	identified	identify	VERB
ijassa-93	67	20	character	character	NOUN
ijassa-93	67	21	we	we	PRON
ijassa-93	67	22	can	can	AUX
ijassa-93	67	23	pick	pick	VERB
ijassa-93	67	24	up	up	ADP
ijassa-93	67	25	a	a	DET
ijassa-93	67	26	suitable	suitable	ADJ
ijassa-93	67	27	threshold	threshold	NOUN
ijassa-93	67	28	via	via	ADP
ijassa-93	67	29	experiment	experiment	NOUN
ijassa-93	67	30	,	,	PUNCT
ijassa-93	67	31	and	and	CCONJ
ijassa-93	67	32	change	change	VERB
ijassa-93	67	33	it	it	PRON
ijassa-93	67	34	to	to	ADP
ijassa-93	67	35	the	the	DET
ijassa-93	67	36	left	left	NOUN
ijassa-93	67	37	or	or	CCONJ
ijassa-93	67	38	eight	eight	NUM
ijassa-93	67	39	.	.	PUNCT
ijassa-93	68	1	then	then	ADV
ijassa-93	68	2	use	use	VERB
ijassa-93	68	3	it	it	PRON
ijassa-93	68	4	as	as	ADP
ijassa-93	68	5	the	the	DET
ijassa-93	68	6	segmenting	segmenting	NOUN
ijassa-93	68	7	position	position	NOUN
ijassa-93	68	8	and	and	CCONJ
ijassa-93	68	9	save	save	VERB
ijassa-93	68	10	the	the	DET
ijassa-93	68	11	result.if	result.if	NOUN
ijassa-93	68	12	character	character	NOUN
ijassa-93	68	13	width	width	VERB
ijassa-93	68	14	to	to	ADP
ijassa-93	68	15	height	height	NOUN
ijassa-93	68	16	ratio	ratio	NOUN
ijassa-93	68	17	is	be	AUX
ijassa-93	68	18	large	large	ADJ
ijassa-93	68	19	,	,	PUNCT
ijassa-93	68	20	the	the	DET
ijassa-93	68	21	character	character	NOUN
ijassa-93	68	22	is	be	AUX
ijassa-93	68	23	adherent	adherent	ADJ
ijassa-93	68	24	;	;	PUNCT
ijassa-93	68	25	if	if	SCONJ
ijassa-93	68	26	the	the	DET
ijassa-93	68	27	ratio	ratio	NOUN
ijassa-93	68	28	is	be	AUX
ijassa-93	68	29	relatively	relatively	ADV
ijassa-93	68	30	small	small	ADJ
ijassa-93	68	31	,	,	PUNCT
ijassa-93	68	32	we	we	PRON
ijassa-93	68	33	could	could	AUX
ijassa-93	68	34	transfer	transfer	VERB
ijassa-93	68	35	the	the	DET
ijassa-93	68	36	identifying	identify	VERB
ijassa-93	68	37	module	module	NOUN
ijassa-93	68	38	to	to	PART
ijassa-93	68	39	check	check	VERB
ijassa-93	68	40	it	it	PRON
ijassa-93	68	41	is	be	AUX
ijassa-93	68	42	or	or	CCONJ
ijassa-93	68	43	not	not	PART
ijassa-93	68	44	,	,	PUNCT
ijassa-93	68	45	after	after	SCONJ
ijassa-93	68	46	it	it	PRON
ijassa-93	68	47	being	be	AUX
ijassa-93	68	48	segmented	segment	VERB
ijassa-93	68	49	.	.	PUNCT
ijassa-93	69	1	if	if	SCONJ
ijassa-93	69	2	it	it	PRON
ijassa-93	69	3	is	be	AUX
ijassa-93	69	4	adherent	adherent	ADJ
ijassa-93	69	5	we	we	PRON
ijassa-93	69	6	can	can	AUX
ijassa-93	69	7	calculate	calculate	VERB
ijassa-93	69	8	its	its	PRON
ijassa-93	69	9	potential	potential	ADJ
ijassa-93	69	10	height	height	NOUN
ijassa-93	69	11	value	value	NOUN
ijassa-93	69	12	,	,	PUNCT
ijassa-93	69	13	according	accord	VERB
ijassa-93	69	14	to	to	ADP
ijassa-93	69	15	whose	whose	DET
ijassa-93	69	16	height	height	NOUN
ijassa-93	69	17	we	we	PRON
ijassa-93	69	18	adjust	adjust	VERB
ijassa-93	69	19	a	a	DET
ijassa-93	69	20	threshold	threshold	NOUN
ijassa-93	69	21	value	value	NOUN
ijassa-93	69	22	and	and	CCONJ
ijassa-93	69	23	use	use	VERB
ijassa-93	69	24	it	it	PRON
ijassa-93	69	25	as	as	ADP
ijassa-93	69	26	its	its	PRON
ijassa-93	69	27	possible	possible	ADJ
ijassa-93	69	28	width	width	NOUN
ijassa-93	69	29	.	.	PUNCT
ijassa-93	70	1	then	then	ADV
ijassa-93	70	2	we	we	PRON
ijassa-93	70	3	take	take	VERB
ijassa-93	70	4	advantage	advantage	NOUN
ijassa-93	70	5	of	of	ADP
ijassa-93	70	6	identifying	identify	VERB
ijassa-93	70	7	module	module	NOUN
ijassa-93	70	8	to	to	PART
ijassa-93	70	9	segment	segment	VERB
ijassa-93	70	10	the	the	DET
ijassa-93	70	11	adherent	adherent	ADJ
ijassa-93	70	12	character	character	NOUN
ijassa-93	70	13	and	and	CCONJ
ijassa-93	70	14	confirm	confirm	VERB
ijassa-93	70	15	the	the	DET
ijassa-93	70	16	segmenting	segment	VERB
ijassa-93	70	17	position	position	NOUN
ijassa-93	70	18	via	via	ADP
ijassa-93	70	19	identified	identify	VERB
ijassa-93	70	20	result	result	NOUN
ijassa-93	70	21	.	.	PUNCT
ijassa-93	71	1	save	save	VERB
ijassa-93	71	2	the	the	DET
ijassa-93	71	3	correct	correct	ADJ
ijassa-93	71	4	final	final	ADJ
ijassa-93	71	5	segmenting	segmenting	NOUN
ijassa-93	71	6	result	result	NOUN
ijassa-93	71	7	.	.	PUNCT
ijassa-93	72	1	the	the	DET
ijassa-93	72	2	english	english	ADJ
ijassa-93	72	3	character	character	NOUN
ijassa-93	72	4	baseline	baseline	NOUN
ijassa-93	72	5	feature	feature	NOUN
ijassa-93	72	6	as	as	SCONJ
ijassa-93	72	7	follows	follow	VERB
ijassa-93	72	8	:	:	PUNCT
ijassa-93	72	9	for	for	ADP
ijassa-93	72	10	the	the	DET
ijassa-93	72	11	adherent	adherent	ADJ
ijassa-93	72	12	english	english	ADJ
ijassa-93	72	13	number	number	NOUN
ijassa-93	72	14	,	,	PUNCT
ijassa-93	72	15	if	if	SCONJ
ijassa-93	72	16	the	the	DET
ijassa-93	72	17	line	line	NOUN
ijassa-93	72	18	height	height	NOUN
ijassa-93	72	19	is	be	AUX
ijassa-93	72	20	close	close	ADJ
ijassa-93	72	21	to	to	ADP
ijassa-93	72	22	5/6	5/6	NUM
ijassa-93	72	23	of	of	ADP
ijassa-93	72	24	chinese	chinese	ADJ
ijassa-93	72	25	fig.1	fig.1	PROPN
ijassa-93	72	26	english	english	PROPN
ijassa-93	72	27	character	character	NOUN
ijassa-93	72	28	baseline	baseline	NOUN
ijassa-93	72	29	feature	feature	NOUN
ijassa-93	72	30	line	line	NOUN
ijassa-93	72	31	height	height	NOUN
ijassa-93	72	32	,	,	PUNCT
ijassa-93	72	33	there	there	PRON
ijassa-93	72	34	are	be	VERB
ijassa-93	72	35	up	up	ADV
ijassa-93	72	36	-	-	PUNCT
ijassa-93	72	37	protruding	protrude	VERB
ijassa-93	72	38	letters	letter	NOUN
ijassa-93	72	39	(	(	PUNCT
ijassa-93	72	40	e.g.	e.g.	ADV
ijassa-93	72	41	t	t	PROPN
ijassa-93	72	42	)	)	PUNCT
ijassa-93	72	43	as	as	ADV
ijassa-93	72	44	well	well	ADV
ijassa-93	72	45	as	as	ADP
ijassa-93	72	46	down	down	ADV
ijassa-93	72	47	-	-	PUNCT
ijassa-93	72	48	protruding	protrude	VERB
ijassa-93	72	49	letters	letter	NOUN
ijassa-93	72	50	(	(	PUNCT
ijassa-93	72	51	e.g.	e.g.	ADV
ijassa-93	72	52	p	p	ADJ
ijassa-93	72	53	)	)	PUNCT
ijassa-93	72	54	.	.	PUNCT
ijassa-93	73	1	we	we	PRON
ijassa-93	73	2	confirm	confirm	VERB
ijassa-93	73	3	the	the	DET
ijassa-93	73	4	baseline	baseline	NOUN
ijassa-93	73	5	approximately	approximately	ADV
ijassa-93	73	6	as	as	ADV
ijassa-93	73	7	much	much	ADJ
ijassa-93	73	8	as	as	ADP
ijassa-93	73	9	1/4	1/4	NUM
ijassa-93	73	10	as	as	ADP
ijassa-93	73	11	adheradvances	adheradvance	NOUN
ijassa-93	73	12	in	in	ADP
ijassa-93	73	13	systems	system	NOUN
ijassa-93	73	14	science	science	NOUN
ijassa-93	73	15	and	and	CCONJ
ijassa-93	73	16	applications	application	NOUN
ijassa-93	73	17	(	(	PUNCT
ijassa-93	73	18	2012	2012	NUM
ijassa-93	73	19	)	)	PUNCT
ijassa-93	73	20	,	,	PUNCT
ijassa-93	73	21	vol.12	vol.12	NOUN
ijassa-93	73	22	,	,	PUNCT
ijassa-93	73	23	no.1	no.1	X
ijassa-93	73	24	71	71	NUM
ijassa-93	73	25	ent	ent	NOUN
ijassa-93	73	26	character	character	NOUN
ijassa-93	73	27	height	height	NOUN
ijassa-93	73	28	,	,	PUNCT
ijassa-93	73	29	and	and	CCONJ
ijassa-93	73	30	height	height	NOUN
ijassa-93	73	31	x	x	VERB
ijassa-93	73	32	is	be	AUX
ijassa-93	73	33	approximately	approximately	ADV
ijassa-93	73	34	as	as	ADV
ijassa-93	73	35	much	much	ADJ
ijassa-93	73	36	as	as	ADP
ijassa-93	73	37	1/2	1/2	NUM
ijassa-93	73	38	of	of	ADP
ijassa-93	73	39	adherent	adherent	ADJ
ijassa-93	73	40	character	character	NOUN
ijassa-93	73	41	height	height	NOUN
ijassa-93	73	42	;	;	PUNCT
ijassa-93	73	43	if	if	SCONJ
ijassa-93	73	44	the	the	DET
ijassa-93	73	45	line	line	NOUN
ijassa-93	73	46	height	height	NOUN
ijassa-93	73	47	is	be	AUX
ijassa-93	73	48	close	close	ADJ
ijassa-93	73	49	to	to	ADP
ijassa-93	73	50	2/3	2/3	NUM
ijassa-93	73	51	of	of	ADP
ijassa-93	73	52	chinese	chinese	ADJ
ijassa-93	73	53	line	line	NOUN
ijassa-93	73	54	height	height	NOUN
ijassa-93	73	55	,	,	PUNCT
ijassa-93	73	56	there	there	PRON
ijassa-93	73	57	are	be	VERB
ijassa-93	73	58	up	up	ADV
ijassa-93	73	59	-	-	PUNCT
ijassa-93	73	60	protruding	protrude	VERB
ijassa-93	73	61	letter	letter	NOUN
ijassa-93	73	62	or	or	CCONJ
ijassa-93	73	63	down	down	ADV
ijassa-93	73	64	-	-	PUNCT
ijassa-93	73	65	protruding	protrude	VERB
ijassa-93	73	66	letter	letter	NOUN
ijassa-93	73	67	.	.	PUNCT
ijassa-93	74	1	we	we	PRON
ijassa-93	74	2	confirm	confirm	VERB
ijassa-93	74	3	the	the	DET
ijassa-93	74	4	baseline	baseline	NOUN
ijassa-93	74	5	approximately	approximately	ADV
ijassa-93	74	6	as	as	ADV
ijassa-93	74	7	much	much	ADJ
ijassa-93	74	8	as	as	ADP
ijassa-93	74	9	1/3	1/3	NUM
ijassa-93	74	10	as	as	ADP
ijassa-93	74	11	adherent	adherent	ADJ
ijassa-93	74	12	character	character	NOUN
ijassa-93	74	13	height	height	NOUN
ijassa-93	74	14	,	,	PUNCT
ijassa-93	74	15	and	and	CCONJ
ijassa-93	74	16	height	height	NOUN
ijassa-93	74	17	x	x	VERB
ijassa-93	74	18	is	be	AUX
ijassa-93	74	19	approximately	approximately	ADV
ijassa-93	74	20	as	as	ADV
ijassa-93	74	21	much	much	ADJ
ijassa-93	74	22	as	as	ADP
ijassa-93	74	23	2/3	2/3	NUM
ijassa-93	74	24	of	of	ADP
ijassa-93	74	25	adherent	adherent	ADJ
ijassa-93	74	26	character	character	NOUN
ijassa-93	74	27	height	height	NOUN
ijassa-93	74	28	;	;	PUNCT
ijassa-93	74	29	if	if	SCONJ
ijassa-93	74	30	the	the	DET
ijassa-93	74	31	line	line	NOUN
ijassa-93	74	32	height	height	NOUN
ijassa-93	74	33	is	be	AUX
ijassa-93	74	34	close	close	ADJ
ijassa-93	74	35	to	to	ADP
ijassa-93	74	36	1/3	1/3	NUM
ijassa-93	74	37	of	of	ADP
ijassa-93	74	38	chinese	chinese	ADJ
ijassa-93	74	39	line	line	NOUN
ijassa-93	74	40	height	height	NOUN
ijassa-93	74	41	,	,	PUNCT
ijassa-93	74	42	there	there	PRON
ijassa-93	74	43	are	be	VERB
ijassa-93	74	44	neither	neither	PRON
ijassa-93	74	45	upprotruding	upprotrude	VERB
ijassa-93	74	46	letter	letter	NOUN
ijassa-93	74	47	nor	nor	CCONJ
ijassa-93	74	48	down	down	ADV
ijassa-93	74	49	-	-	PUNCT
ijassa-93	74	50	protruding	protrude	VERB
ijassa-93	74	51	letter	letter	NOUN
ijassa-93	74	52	.	.	PUNCT
ijassa-93	75	1	we	we	PRON
ijassa-93	75	2	confirm	confirm	VERB
ijassa-93	75	3	the	the	DET
ijassa-93	75	4	baseline	baseline	NOUN
ijassa-93	75	5	approximately	approximately	ADV
ijassa-93	75	6	as	as	ADV
ijassa-93	75	7	much	much	ADJ
ijassa-93	75	8	as	as	ADP
ijassa-93	75	9	1/2	1/2	NUM
ijassa-93	75	10	as	as	ADP
ijassa-93	75	11	adherent	adherent	ADJ
ijassa-93	75	12	character	character	NOUN
ijassa-93	75	13	height	height	NOUN
ijassa-93	75	14	,	,	PUNCT
ijassa-93	75	15	and	and	CCONJ
ijassa-93	75	16	height	height	NOUN
ijassa-93	75	17	x	x	VERB
ijassa-93	75	18	is	be	AUX
ijassa-93	75	19	approximately	approximately	ADV
ijassa-93	75	20	as	as	ADV
ijassa-93	75	21	much	much	ADJ
ijassa-93	75	22	as	as	ADP
ijassa-93	75	23	adherent	adherent	ADJ
ijassa-93	75	24	character	character	NOUN
ijassa-93	75	25	height	height	NOUN
ijassa-93	75	26	.	.	PUNCT
ijassa-93	76	1	and	and	CCONJ
ijassa-93	76	2	we	we	PRON
ijassa-93	76	3	can	can	AUX
ijassa-93	76	4	adjust	adjust	VERB
ijassa-93	76	5	the	the	DET
ijassa-93	76	6	potential	potential	ADJ
ijassa-93	76	7	segmenting	segmenting	NOUN
ijassa-93	76	8	position	position	NOUN
ijassa-93	76	9	to	to	PART
ijassa-93	76	10	meet	meet	VERB
ijassa-93	76	11	height	height	NOUN
ijassa-93	76	12	x	x	PUNCT
ijassa-93	76	13	via	via	ADP
ijassa-93	76	14	threshold	threshold	NOUN
ijassa-93	76	15	-	-	PUNCT
ijassa-93	76	16	adjusting	adjust	VERB
ijassa-93	76	17	.	.	PUNCT
ijassa-93	77	1	and	and	CCONJ
ijassa-93	77	2	the	the	DET
ijassa-93	77	3	same	same	ADJ
ijassa-93	77	4	method	method	NOUN
ijassa-93	77	5	could	could	AUX
ijassa-93	77	6	be	be	AUX
ijassa-93	77	7	used	use	VERB
ijassa-93	77	8	in	in	ADP
ijassa-93	77	9	english	english	ADJ
ijassa-93	77	10	number	number	NOUN
ijassa-93	77	11	adhesion	adhesion	NOUN
ijassa-93	77	12	and	and	CCONJ
ijassa-93	77	13	chinese	chinese	ADJ
ijassa-93	77	14	adhesion	adhesion	NOUN
ijassa-93	77	15	.	.	PUNCT
ijassa-93	78	1	feature	feature	NOUN
ijassa-93	78	2	feedback	feedback	NOUN
ijassa-93	78	3	based	base	VERB
ijassa-93	78	4	algorithm	algorithm	NOUN
ijassa-93	78	5	of	of	ADP
ijassa-93	78	6	character	character	NOUN
ijassa-93	78	7	segmentation	segmentation	NOUN
ijassa-93	78	8	&	&	CCONJ
ijassa-93	78	9	recognition	recognition	NOUN
ijassa-93	78	10	between	between	ADP
ijassa-93	78	11	chinese	chinese	PROPN
ijassa-93	78	12	,	,	PUNCT
ijassa-93	78	13	english	english	ADJ
ijassa-93	78	14	and	and	CCONJ
ijassa-93	78	15	adherent	adherent	NOUN
ijassa-93	78	16	character	character	NOUN
ijassa-93	78	17	identify	identify	VERB
ijassa-93	78	18	result	result	NOUN
ijassa-93	78	19	as	as	ADP
ijassa-93	78	20	fig.2	fig.2	PROPN
ijassa-93	78	21	and	and	CCONJ
ijassa-93	78	22	fig.3	fig.3	PROPN
ijassa-93	78	23	.	.	PUNCT
ijassa-93	79	1	fig.2	fig.2	PROPN
ijassa-93	79	2	english	english	ADJ
ijassa-93	79	3	character	character	NOUN
ijassa-93	79	4	segmenting	segment	VERB
ijassa-93	79	5	result	result	NOUN
ijassa-93	79	6	fig.3	fig.3	ADJ
ijassa-93	79	7	chinese	chinese	ADJ
ijassa-93	79	8	character	character	NOUN
ijassa-93	79	9	segmenting	segment	VERB
ijassa-93	79	10	result	result	NOUN
ijassa-93	79	11	3	3	NUM
ijassa-93	79	12	character	character	NOUN
ijassa-93	79	13	size	size	NOUN
ijassa-93	79	14	normalization	normalization	NOUN
ijassa-93	79	15	after	after	ADP
ijassa-93	79	16	being	be	AUX
ijassa-93	79	17	segmented	segment	VERB
ijassa-93	79	18	,	,	PUNCT
ijassa-93	79	19	each	each	DET
ijassa-93	79	20	character	character	NOUN
ijassa-93	79	21	is	be	AUX
ijassa-93	79	22	in	in	ADP
ijassa-93	79	23	different	different	ADJ
ijassa-93	79	24	size	size	NOUN
ijassa-93	79	25	;	;	PUNCT
ijassa-93	79	26	therefore	therefore	ADV
ijassa-93	79	27	the	the	DET
ijassa-93	79	28	segmented	segment	VERB
ijassa-93	79	29	character	character	NOUN
ijassa-93	79	30	region	region	NOUN
ijassa-93	79	31	is	be	AUX
ijassa-93	79	32	different.in	different.in	PROPN
ijassa-93	79	33	order	order	NOUN
ijassa-93	79	34	to	to	PART
ijassa-93	79	35	extract	extract	VERB
ijassa-93	79	36	the	the	DET
ijassa-93	79	37	character	character	NOUN
ijassa-93	79	38	feature	feature	NOUN
ijassa-93	79	39	,	,	PUNCT
ijassa-93	79	40	we	we	PRON
ijassa-93	79	41	need	need	VERB
ijassa-93	79	42	to	to	PART
ijassa-93	79	43	normalize	normalize	VERB
ijassa-93	79	44	the	the	DET
ijassa-93	79	45	segmented	segment	VERB
ijassa-93	79	46	character	character	NOUN
ijassa-93	79	47	region	region	NOUN
ijassa-93	79	48	.	.	PUNCT
ijassa-93	80	1	suppose	suppose	VERB
ijassa-93	80	2	height	height	NOUN
ijassa-93	80	3	and	and	CCONJ
ijassa-93	80	4	width	width	VERB
ijassa-93	80	5	to	to	PART
ijassa-93	80	6	represent	represent	VERB
ijassa-93	80	7	respectively	respectively	ADV
ijassa-93	80	8	normalized	normalize	VERB
ijassa-93	80	9	character	character	NOUN
ijassa-93	80	10	vertical	vertical	ADJ
ijassa-93	80	11	and	and	CCONJ
ijassa-93	80	12	horizontal	horizontal	ADJ
ijassa-93	80	13	length	length	NOUN
ijassa-93	80	14	.	.	PUNCT
ijassa-93	81	1	and	and	CCONJ
ijassa-93	81	2	h	h	NOUN
ijassa-93	81	3	and	and	CCONJ
ijassa-93	81	4	w	w	PROPN
ijassa-93	81	5	respectively	respectively	ADV
ijassa-93	81	6	represent	represent	VERB
ijassa-93	81	7	height	height	NOUN
ijassa-93	81	8	and	and	CCONJ
ijassa-93	81	9	width	width	NOUN
ijassa-93	81	10	of	of	ADP
ijassa-93	81	11	the	the	DET
ijassa-93	81	12	character	character	NOUN
ijassa-93	81	13	rectangle	rectangle	NOUN
ijassa-93	81	14	.	.	PUNCT
ijassa-93	82	1	therefore	therefore	ADV
ijassa-93	82	2	,	,	PUNCT
ijassa-93	82	3	the	the	DET
ijassa-93	82	4	ratio	ratio	NOUN
ijassa-93	82	5	of	of	ADP
ijassa-93	82	6	vertical	vertical	ADJ
ijassa-93	82	7	to	to	ADP
ijassa-93	82	8	horizontal	horizontal	ADJ
ijassa-93	82	9	hr	hr	PROPN
ijassa-93	82	10	and	and	CCONJ
ijassa-93	82	11	wr	wr	PROPN
ijassa-93	82	12	would	would	AUX
ijassa-93	82	13	be	be	AUX
ijassa-93	82	14	:	:	PUNCT
ijassa-93	82	15	hr	hr	NOUN
ijassa-93	82	16	=	=	PUNCT
ijassa-93	82	17	height	height	NOUN
ijassa-93	82	18	h	h	NOUN
ijassa-93	82	19	,	,	PUNCT
ijassa-93	82	20	wr	wr	NOUN
ijassa-93	82	21	=	=	PUNCT
ijassa-93	82	22	width	width	VERB
ijassa-93	82	23	w	w	PROPN
ijassa-93	82	24	72	72	NUM
ijassa-93	82	25	jinghua	jinghua	NOUN
ijassa-93	82	26	hu	hu	PROPN
ijassa-93	82	27	:	:	PUNCT
ijassa-93	82	28	the	the	DET
ijassa-93	82	29	research	research	NOUN
ijassa-93	82	30	on	on	ADP
ijassa-93	82	31	effective	effective	ADJ
ijassa-93	82	32	video	video	NOUN
ijassa-93	82	33	scene	scene	NOUN
ijassa-93	82	34	character	character	NOUN
ijassa-93	82	35	extraction	extraction	NOUN
ijassa-93	82	36	algorithm	algorithm	NOUN
ijassa-93	82	37	...	...	PUNCT
ijassa-93	83	1	then	then	ADV
ijassa-93	83	2	we	we	PRON
ijassa-93	83	3	make	make	VERB
ijassa-93	83	4	coordinate	coordinate	NOUN
ijassa-93	83	5	mapping	mapping	NOUN
ijassa-93	83	6	,	,	PUNCT
ijassa-93	83	7	and	and	CCONJ
ijassa-93	83	8	we	we	PRON
ijassa-93	83	9	suppose	suppose	VERB
ijassa-93	83	10	coordinate	coordinate	NOUN
ijassa-93	83	11	(	(	PUNCT
ijassa-93	83	12	i	i	PRON
ijassa-93	83	13	new	new	ADJ
ijassa-93	83	14	,	,	PUNCT
ijassa-93	83	15	j	j	PROPN
ijassa-93	83	16	new	new	ADJ
ijassa-93	83	17	)	)	PUNCT
ijassa-93	83	18	is	be	AUX
ijassa-93	83	19	the	the	DET
ijassa-93	83	20	mapped	map	VERB
ijassa-93	83	21	coordinate	coordinate	NOUN
ijassa-93	83	22	(	(	PUNCT
ijassa-93	83	23	i	i	PROPN
ijassa-93	83	24	,	,	PUNCT
ijassa-93	83	25	j	j	PROPN
ijassa-93	83	26	)	)	PUNCT
ijassa-93	83	27	after	after	SCONJ
ijassa-93	83	28	it	it	PRON
ijassa-93	83	29	has	have	AUX
ijassa-93	83	30	been	be	AUX
ijassa-93	83	31	normalized	normalize	VERB
ijassa-93	83	32	.	.	PUNCT
ijassa-93	84	1	similarly	similarly	ADV
ijassa-93	84	2	we	we	PRON
ijassa-93	84	3	map	map	VERB
ijassa-93	84	4	the	the	DET
ijassa-93	84	5	following	following	ADJ
ijassa-93	84	6	coordinates	coordinate	NOUN
ijassa-93	84	7	:	:	PUNCT
ijassa-93	84	8	i	i	PRON
ijassa-93	84	9	new	new	ADJ
ijassa-93	84	10	=	=	SYM
ijassa-93	84	11	top+	top+	PROPN
ijassa-93	84	12	(	(	PUNCT
ijassa-93	84	13	i−	i−	PROPN
ijassa-93	84	14	top)/hr	top)/hr	PROPN
ijassa-93	84	15	,	,	PUNCT
ijassa-93	84	16	j	j	PROPN
ijassa-93	84	17	new	new	ADJ
ijassa-93	84	18	=	=	X
ijassa-93	84	19	left+	left+	X
ijassa-93	84	20	(	(	PUNCT
ijassa-93	84	21	j	j	PROPN
ijassa-93	84	22	−	−	PROPN
ijassa-93	84	23	left)/wr	left)/wr	PROPN
ijassa-93	84	24	note	note	NOUN
ijassa-93	84	25	:	:	PUNCT
ijassa-93	84	26	top	top	ADJ
ijassa-93	84	27	is	be	AUX
ijassa-93	84	28	the	the	DET
ijassa-93	84	29	vertical	vertical	ADJ
ijassa-93	84	30	coordinate	coordinate	NOUN
ijassa-93	84	31	of	of	ADP
ijassa-93	84	32	the	the	DET
ijassa-93	84	33	point	point	NOUN
ijassa-93	84	34	a	a	PRON
ijassa-93	84	35	in	in	ADP
ijassa-93	84	36	the	the	DET
ijassa-93	84	37	mornalized	mornalize	VERB
ijassa-93	84	38	character	character	NOUN
ijassa-93	84	39	region	region	NOUN
ijassa-93	84	40	,	,	PUNCT
ijassa-93	84	41	and	and	CCONJ
ijassa-93	84	42	left	leave	VERB
ijassa-93	84	43	is	be	AUX
ijassa-93	84	44	the	the	DET
ijassa-93	84	45	horizontal	horizontal	ADJ
ijassa-93	84	46	coordinate	coordinate	NOUN
ijassa-93	84	47	of	of	ADP
ijassa-93	84	48	the	the	DET
ijassa-93	84	49	point	point	NOUN
ijassa-93	84	50	a	a	PRON
ijassa-93	84	51	in	in	ADP
ijassa-93	84	52	the	the	DET
ijassa-93	84	53	mornalized	mornalize	VERB
ijassa-93	84	54	character	character	NOUN
ijassa-93	84	55	region	region	NOUN
ijassa-93	84	56	.	.	PUNCT
ijassa-93	85	1	the	the	DET
ijassa-93	85	2	character	character	NOUN
ijassa-93	85	3	size	size	NOUN
ijassa-93	85	4	normalized	normalize	VERB
ijassa-93	85	5	result	result	NOUN
ijassa-93	85	6	is	be	AUX
ijassa-93	85	7	shown	show	VERB
ijassa-93	85	8	in	in	ADP
ijassa-93	85	9	fig.4	fig.4	PROPN
ijassa-93	85	10	.	.	PROPN
ijassa-93	85	11	assign	assign	VERB
ijassa-93	85	12	the	the	DET
ijassa-93	85	13	gray	gray	ADJ
ijassa-93	85	14	-	-	PUNCT
ijassa-93	85	15	scale	scale	NOUN
ijassa-93	85	16	value	value	NOUN
ijassa-93	85	17	of	of	ADP
ijassa-93	85	18	the	the	DET
ijassa-93	85	19	character	character	NOUN
ijassa-93	85	20	coordinate	coordinate	NOUN
ijassa-93	85	21	(	(	PUNCT
ijassa-93	85	22	i	i	PROPN
ijassa-93	85	23	,	,	PUNCT
ijassa-93	85	24	j	j	PROPN
ijassa-93	85	25	)	)	PUNCT
ijassa-93	85	26	before	before	ADP
ijassa-93	85	27	being	be	AUX
ijassa-93	85	28	normalized	normalize	VERB
ijassa-93	85	29	to	to	ADP
ijassa-93	85	30	the	the	DET
ijassa-93	85	31	normalized	normalize	VERB
ijassa-93	85	32	character	character	NOUN
ijassa-93	85	33	coordinate	coordinate	NOUN
ijassa-93	85	34	(	(	PUNCT
ijassa-93	85	35	i	i	PRON
ijassa-93	85	36	new	new	ADJ
ijassa-93	85	37	,	,	PUNCT
ijassa-93	85	38	j	j	PROPN
ijassa-93	85	39	new	new	ADJ
ijassa-93	85	40	)	)	PUNCT
ijassa-93	85	41	to	to	PART
ijassa-93	85	42	accomplish	accomplish	VERB
ijassa-93	85	43	the	the	DET
ijassa-93	85	44	character	character	NOUN
ijassa-93	85	45	normalization	normalization	NOUN
ijassa-93	85	46	.	.	PUNCT
ijassa-93	86	1	(	(	PUNCT
ijassa-93	86	2	a	a	X
ijassa-93	86	3	)	)	PUNCT
ijassa-93	86	4	before	before	ADP
ijassa-93	86	5	normalization	normalization	NOUN
ijassa-93	86	6	(	(	PUNCT
ijassa-93	86	7	b	b	NOUN
ijassa-93	86	8	)	)	PUNCT
ijassa-93	86	9	after	after	ADP
ijassa-93	86	10	normalization	normalization	NOUN
ijassa-93	86	11	fig.4	fig.4	PRON
ijassa-93	86	12	contrasts	contrast	VERB
ijassa-93	86	13	two	two	NUM
ijassa-93	86	14	normalization	normalization	NOUN
ijassa-93	86	15	results	result	VERB
ijassa-93	86	16	4	4	NUM
ijassa-93	86	17	improvement	improvement	NOUN
ijassa-93	86	18	of	of	ADP
ijassa-93	86	19	bp	bp	PROPN
ijassa-93	86	20	algorithm	algorithm	NOUN
ijassa-93	86	21	base	base	NOUN
ijassa-93	86	22	on	on	ADP
ijassa-93	86	23	the	the	DET
ijassa-93	86	24	traditional	traditional	ADJ
ijassa-93	86	25	bp	bp	PROPN
ijassa-93	86	26	algorithm	algorithm	NOUN
ijassa-93	86	27	,	,	PUNCT
ijassa-93	86	28	this	this	DET
ijassa-93	86	29	paper	paper	NOUN
ijassa-93	86	30	presents	present	VERB
ijassa-93	86	31	an	an	DET
ijassa-93	86	32	improved	improved	ADJ
ijassa-93	86	33	algorithm	algorithm	NOUN
ijassa-93	86	34	which	which	PRON
ijassa-93	86	35	improves	improve	VERB
ijassa-93	86	36	bp	bp	PROPN
ijassa-93	86	37	mainly	mainly	ADV
ijassa-93	86	38	through	through	ADP
ijassa-93	86	39	restructuring	restructure	VERB
ijassa-93	86	40	gradient	gradient	NOUN
ijassa-93	86	41	in	in	ADP
ijassa-93	86	42	the	the	DET
ijassa-93	86	43	sigmoid	sigmoid	NOUN
ijassa-93	86	44	.	.	PUNCT
ijassa-93	87	1	if	if	SCONJ
ijassa-93	87	2	the	the	DET
ijassa-93	87	3	value	value	NOUN
ijassa-93	87	4	is	be	AUX
ijassa-93	87	5	supposed	suppose	VERB
ijassa-93	87	6	too	too	ADV
ijassa-93	87	7	much	much	ADJ
ijassa-93	87	8	,	,	PUNCT
ijassa-93	87	9	the	the	DET
ijassa-93	87	10	input	input	NOUN
ijassa-93	87	11	0	0	NUM
ijassa-93	87	12	or	or	CCONJ
ijassa-93	87	13	1	1	NUM
ijassa-93	87	14	of	of	ADP
ijassa-93	87	15	each	each	DET
ijassa-93	87	16	layer	layer	NOUN
ijassa-93	87	17	would	would	AUX
ijassa-93	87	18	be	be	AUX
ijassa-93	87	19	scattered	scatter	VERB
ijassa-93	87	20	,	,	PUNCT
ijassa-93	87	21	and	and	CCONJ
ijassa-93	87	22	the	the	DET
ijassa-93	87	23	study	study	NOUN
ijassa-93	87	24	result	result	NOUN
ijassa-93	87	25	is	be	AUX
ijassa-93	87	26	worse	bad	ADJ
ijassa-93	87	27	.	.	PUNCT
ijassa-93	88	1	otherwise	otherwise	ADV
ijassa-93	88	2	,	,	PUNCT
ijassa-93	88	3	if	if	SCONJ
ijassa-93	88	4	too	too	ADV
ijassa-93	88	5	little	little	ADJ
ijassa-93	88	6	,	,	PUNCT
ijassa-93	88	7	the	the	DET
ijassa-93	88	8	system	system	NOUN
ijassa-93	88	9	linearity	linearity	NOUN
ijassa-93	88	10	is	be	AUX
ijassa-93	88	11	enhanced	enhance	VERB
ijassa-93	88	12	whereas	whereas	SCONJ
ijassa-93	88	13	nonlinearity	nonlinearity	NOUN
ijassa-93	88	14	is	be	AUX
ijassa-93	88	15	weakened	weaken	VERB
ijassa-93	88	16	.	.	PUNCT
ijassa-93	89	1	therefore	therefore	ADV
ijassa-93	89	2	,	,	PUNCT
ijassa-93	89	3	the	the	DET
ijassa-93	89	4	best	good	ADJ
ijassa-93	89	5	value	value	NOUN
ijassa-93	89	6	is	be	AUX
ijassa-93	89	7	between	between	ADP
ijassa-93	89	8	the	the	DET
ijassa-93	89	9	above	above	ADJ
ijassa-93	89	10	both	both	PRON
ijassa-93	89	11	.	.	PUNCT
ijassa-93	90	1	the	the	DET
ijassa-93	90	2	most	most	ADV
ijassa-93	90	3	important	important	ADJ
ijassa-93	90	4	point	point	NOUN
ijassa-93	90	5	is	be	AUX
ijassa-93	90	6	to	to	PART
ijassa-93	90	7	amend	amend	VERB
ijassa-93	90	8	the	the	DET
ijassa-93	90	9	gradient	gradient	NOUN
ijassa-93	90	10	of	of	ADP
ijassa-93	90	11	each	each	DET
ijassa-93	90	12	allnodes	allnode	NOUN
ijassa-93	90	13	in	in	ADP
ijassa-93	90	14	the	the	DET
ijassa-93	90	15	sigmoid	sigmoid	NOUN
ijassa-93	90	16	so	so	SCONJ
ijassa-93	90	17	as	as	SCONJ
ijassa-93	90	18	to	to	PART
ijassa-93	90	19	find	find	VERB
ijassa-93	90	20	a	a	DET
ijassa-93	90	21	best	good	ADJ
ijassa-93	90	22	value	value	NOUN
ijassa-93	90	23	.	.	PUNCT
ijassa-93	91	1	in	in	ADP
ijassa-93	91	2	the	the	DET
ijassa-93	91	3	improved	improved	ADJ
ijassa-93	91	4	bp	bp	PROPN
ijassa-93	91	5	algorithm	algorithm	NOUN
ijassa-93	91	6	,	,	PUNCT
ijassa-93	91	7	sigmoid	sigmoid	NOUN
ijassa-93	91	8	is	be	AUX
ijassa-93	91	9	turned	turn	VERB
ijassa-93	91	10	into	into	ADP
ijassa-93	91	11	binary	binary	ADJ
ijassa-93	91	12	driving	driving	NOUN
ijassa-93	91	13	function	function	NOUN
ijassa-93	91	14	on	on	ADP
ijassa-93	91	15	the	the	DET
ijassa-93	91	16	connection	connection	NOUN
ijassa-93	91	17	of	of	ADP
ijassa-93	91	18	gradient	gradient	NOUN
ijassa-93	91	19	and	and	CCONJ
ijassa-93	91	20	the	the	DET
ijassa-93	91	21	potential	potential	ADJ
ijassa-93	91	22	value	value	NOUN
ijassa-93	91	23	of	of	ADP
ijassa-93	91	24	the	the	DET
ijassa-93	91	25	junction	junction	NOUN
ijassa-93	91	26	point	point	NOUN
ijassa-93	91	27	:	:	PUNCT
ijassa-93	91	28	φ(α	φ(α	ADJ
ijassa-93	91	29	,	,	PUNCT
ijassa-93	91	30	υ	υ	NOUN
ijassa-93	91	31	)	)	PUNCT
ijassa-93	91	32	=	=	SYM
ijassa-93	91	33	1	1	NUM
ijassa-93	91	34	1	1	NUM
ijassa-93	91	35	+	+	CCONJ
ijassa-93	91	36	e−∞	e−∞	NUM
ijassa-93	91	37	we	we	PRON
ijassa-93	91	38	can	can	AUX
ijassa-93	91	39	easily	easily	ADV
ijassa-93	91	40	deduce	deduce	VERB
ijassa-93	91	41	that	that	SCONJ
ijassa-93	91	42	after	after	ADP
ijassa-93	91	43	all	all	DET
ijassa-93	91	44	the	the	DET
ijassa-93	91	45	amendment	amendment	NOUN
ijassa-93	91	46	,	,	PUNCT
ijassa-93	91	47	adjusting	adjust	VERB
ijassa-93	91	48	method	method	NOUN
ijassa-93	91	49	of	of	ADP
ijassa-93	91	50	the	the	DET
ijassa-93	91	51	weight	weight	NOUN
ijassa-93	91	52	and	and	CCONJ
ijassa-93	91	53	threshold	threshold	NOUN
ijassa-93	91	54	is	be	AUX
ijassa-93	91	55	the	the	DET
ijassa-93	91	56	same	same	ADJ
ijassa-93	91	57	with	with	ADP
ijassa-93	91	58	bp	bp	PROPN
ijassa-93	91	59	algorithm	algorithm	PROPN
ijassa-93	91	60	.	.	PUNCT
ijassa-93	92	1	suppose	suppose	VERB
ijassa-93	92	2	the	the	DET
ijassa-93	92	3	actual	actual	ADJ
ijassa-93	92	4	output	output	NOUN
ijassa-93	92	5	vector	vector	NOUN
ijassa-93	92	6	o	o	NOUN
ijassa-93	92	7	=	=	X
ijassa-93	92	8	{	{	PUNCT
ijassa-93	92	9	o0o1	o0o1	PROPN
ijassa-93	92	10	,	,	PUNCT
ijassa-93	92	11	.	.	PUNCT
ijassa-93	92	12	.	.	PUNCT
ijassa-93	93	1	.	.	PUNCT
ijassa-93	94	1	,	,	PUNCT
ijassa-93	94	2	ol	ol	ADJ
ijassa-93	94	3	}	}	PUNCT
ijassa-93	94	4	,	,	PUNCT
ijassa-93	94	5	teacher	teacher	NOUN
ijassa-93	94	6	vector	vector	NOUN
ijassa-93	94	7	t	t	PROPN
ijassa-93	94	8	=	=	SYM
ijassa-93	94	9	{	{	PUNCT
ijassa-93	94	10	t0	t0	PROPN
ijassa-93	94	11	,	,	PUNCT
ijassa-93	94	12	t1	t1	PROPN
ijassa-93	94	13	,	,	PUNCT
ijassa-93	94	14	.	.	PUNCT
ijassa-93	94	15	.	.	PUNCT
ijassa-93	95	1	.	.	PUNCT
ijassa-93	96	1	,	,	PUNCT
ijassa-93	96	2	tl	tl	PROPN
ijassa-93	96	3	}	}	PUNCT
ijassa-93	96	4	,	,	PUNCT
ijassa-93	96	5	mean	mean	ADJ
ijassa-93	96	6	square	square	ADJ
ijassa-93	96	7	error	error	NOUN
ijassa-93	96	8	e	e	NOUN
ijassa-93	96	9	:	:	PUNCT
ijassa-93	96	10	e	e	X
ijassa-93	96	11	=	=	SYM
ijassa-93	96	12	1	1	NUM
ijassa-93	96	13	2	2	NUM
ijassa-93	96	14	∑	∑	PROPN
ijassa-93	96	15	k	k	PROPN
ijassa-93	96	16	(	(	PUNCT
ijassa-93	96	17	tk	tk	PROPN
ijassa-93	96	18	−ok)2	−ok)2	PROPN
ijassa-93	96	19	advances	advance	VERB
ijassa-93	96	20	in	in	ADP
ijassa-93	96	21	systems	system	NOUN
ijassa-93	96	22	science	science	NOUN
ijassa-93	96	23	and	and	CCONJ
ijassa-93	96	24	applications	application	NOUN
ijassa-93	96	25	(	(	PUNCT
ijassa-93	96	26	2012	2012	NUM
ijassa-93	96	27	)	)	PUNCT
ijassa-93	96	28	,	,	PUNCT
ijassa-93	96	29	vol.12	vol.12	NOUN
ijassa-93	96	30	,	,	PUNCT
ijassa-93	96	31	no.1	no.1	X
ijassa-93	96	32	73	73	NUM
ijassa-93	96	33	gradient	gradient	NOUN
ijassa-93	96	34	α	α	PROPN
ijassa-93	96	35	is	be	AUX
ijassa-93	96	36	changed	change	VERB
ijassa-93	96	37	according	accord	VERB
ijassa-93	96	38	to	to	ADP
ijassa-93	96	39	e	e	PROPN
ijassa-93	96	40	′	′	NUM
ijassa-93	96	41	negative	negative	ADJ
ijassa-93	96	42	gradient	gradient	NOUN
ijassa-93	96	43	change	change	NOUN
ijassa-93	96	44	,	,	PUNCT
ijassa-93	96	45	where	where	SCONJ
ijassa-93	96	46	mean	mean	ADJ
ijassa-93	96	47	square	square	ADJ
ijassa-93	96	48	error	error	NOUN
ijassa-93	96	49	e	e	NOUN
ijassa-93	96	50	reduces	reduce	VERB
ijassa-93	96	51	the	the	DET
ijassa-93	96	52	least	least	ADJ
ijassa-93	96	53	.	.	PUNCT
ijassa-93	97	1	hence	hence	ADV
ijassa-93	97	2	the	the	DET
ijassa-93	97	3	improved	improved	ADJ
ijassa-93	97	4	gradient	gradient	ADJ
ijassa-93	97	5	algorithm	algorithm	NOUN
ijassa-93	97	6	would	would	AUX
ijassa-93	97	7	be	be	AUX
ijassa-93	97	8	:	:	PUNCT
ijassa-93	98	1	∆α∞	∆α∞	PROPN
ijassa-93	98	2	=	=	PUNCT
ijassa-93	98	3	∂e	∂e	PROPN
ijassa-93	98	4	∂α	∂α	PROPN
ijassa-93	98	5	the	the	DET
ijassa-93	98	6	node	node	ADJ
ijassa-93	98	7	point	point	NOUN
ijassa-93	98	8	of	of	ADP
ijassa-93	98	9	the	the	DET
ijassa-93	98	10	output	output	NOUN
ijassa-93	98	11	layer	layer	NOUN
ijassa-93	98	12	is	be	AUX
ijassa-93	98	13	:	:	PUNCT
ijassa-93	98	14	ok	ok	ADJ
ijassa-93	98	15	=	=	SYM
ijassa-93	98	16	1	1	NUM
ijassa-93	98	17	e−αk(µk−θk	e−αk(µk−θk	PROPN
ijassa-93	98	18	)	)	PUNCT
ijassa-93	98	19	,	,	PUNCT
ijassa-93	98	20	then	then	ADV
ijassa-93	98	21	∆αk	∆αk	VERB
ijassa-93	99	1	=	=	SYM
ijassa-93	99	2	−η	−η	NOUN
ijassa-93	99	3	∂e	∂e	PROPN
ijassa-93	99	4	∂αk	∂αk	PROPN
ijassa-93	99	5	=	=	PROPN
ijassa-93	99	6	−η	−η	NOUN
ijassa-93	99	7	∂e	∂e	PROPN
ijassa-93	99	8	ok	ok	INTJ
ijassa-93	99	9	·	·	PUNCT
ijassa-93	99	10	∂ok	∂ok	PROPN
ijassa-93	99	11	∂αk	∂αk	PROPN
ijassa-93	99	12	=	=	PUNCT
ijassa-93	99	13	−η(tk	−η(tk	ADP
ijassa-93	99	14	−ok	−ok	NOUN
ijassa-93	99	15	)	)	PUNCT
ijassa-93	99	16	∂	∂	NOUN
ijassa-93	99	17	∂αk	∂αk	NOUN
ijassa-93	99	18	1	1	NUM
ijassa-93	99	19	1	1	NUM
ijassa-93	99	20	+	+	NUM
ijassa-93	99	21	e−αk(µk−θk	e−αk(µk−θk	PROPN
ijassa-93	99	22	)	)	PUNCT
ijassa-93	99	23	=	=	PUNCT
ijassa-93	99	24	η(tk	η(tk	PROPN
ijassa-93	99	25	−ok)(µk	−ok)(µk	ADV
ijassa-93	99	26	−	−	PROPN
ijassa-93	99	27	θk)ok(1−ok	θk)ok(1−ok	NOUN
ijassa-93	99	28	)	)	PUNCT
ijassa-93	99	29	η	η	PROPN
ijassa-93	99	30	is	be	AUX
ijassa-93	99	31	studying	study	VERB
ijassa-93	99	32	step	step	NOUN
ijassa-93	99	33	length	length	NOUN
ijassa-93	99	34	,	,	PUNCT
ijassa-93	99	35	okis	oki	NOUN
ijassa-93	99	36	the	the	DET
ijassa-93	99	37	actual	actual	ADJ
ijassa-93	99	38	output	output	NOUN
ijassa-93	99	39	of	of	ADP
ijassa-93	99	40	output	output	NOUN
ijassa-93	99	41	layer	layer	NOUN
ijassa-93	99	42	neuron	neuron	PROPN
ijassa-93	100	1	k	k	NOUN
ijassa-93	100	2	,	,	PUNCT
ijassa-93	100	3	tkis	tkis	ADJ
ijassa-93	100	4	teacher	teacher	NOUN
ijassa-93	100	5	signal,µkis	signal,µki	VERB
ijassa-93	100	6	the	the	DET
ijassa-93	100	7	input	input	NOUN
ijassa-93	100	8	signal	signal	NOUN
ijassa-93	100	9	linear	linear	ADJ
ijassa-93	100	10	combination	combination	NOUN
ijassa-93	100	11	of	of	ADP
ijassa-93	100	12	the	the	DET
ijassa-93	100	13	output	output	NOUN
ijassa-93	100	14	layer	layer	NOUN
ijassa-93	100	15	node	node	NOUN
ijassa-93	100	16	,	,	PUNCT
ijassa-93	100	17	then	then	ADV
ijassa-93	100	18	µk	µk	INTJ
ijassa-93	100	19	=	=	PUNCT
ijassa-93	100	20	∑	∑	PUNCT
ijassa-93	100	21	k	k	PROPN
ijassa-93	100	22	hjwkj	hjwkj	PROPN
ijassa-93	100	23	,	,	PUNCT
ijassa-93	100	24	wkj	wkj	PROPN
ijassa-93	100	25	is	be	AUX
ijassa-93	100	26	weight;θkis	weight;θkis	PROPN
ijassa-93	100	27	threshold;hj	threshold;hj	PROPN
ijassa-93	100	28	is	be	AUX
ijassa-93	100	29	the	the	DET
ijassa-93	100	30	input	input	NOUN
ijassa-93	100	31	signal	signal	NOUN
ijassa-93	100	32	of	of	ADP
ijassa-93	100	33	output	output	NOUN
ijassa-93	100	34	layer	layer	NOUN
ijassa-93	100	35	nodepoint	nodepoint	NOUN
ijassa-93	100	36	,	,	PUNCT
ijassa-93	100	37	that	that	PRON
ijassa-93	100	38	is	be	AUX
ijassa-93	100	39	the	the	DET
ijassa-93	100	40	output	output	NOUN
ijassa-93	100	41	of	of	ADP
ijassa-93	100	42	intermediate	intermediate	ADJ
ijassa-93	100	43	layer	layer	NOUN
ijassa-93	100	44	neuron	neuron	NOUN
ijassa-93	100	45	.	.	PUNCT
ijassa-93	101	1	for	for	ADP
ijassa-93	101	2	the	the	DET
ijassa-93	101	3	node	node	ADJ
ijassa-93	101	4	point	point	NOUN
ijassa-93	101	5	of	of	ADP
ijassa-93	101	6	∆αj	∆αj	PRON
ijassa-93	101	7	=	=	SYM
ijassa-93	101	8	−η	−η	NOUN
ijassa-93	101	9	∂e	∂e	PROPN
ijassa-93	101	10	∂αj	∂αj	PROPN
ijassa-93	101	11	=	=	SYM
ijassa-93	101	12	−η	−η	PROPN
ijassa-93	101	13	∂e	∂e	PROPN
ijassa-93	101	14	∂hj	∂hj	PROPN
ijassa-93	101	15	·	·	PUNCT
ijassa-93	101	16	∂hj	∂hj	PROPN
ijassa-93	101	17	∂αj	∂αj	ADJ
ijassa-93	101	18	hidden	hide	VERB
ijassa-93	101	19	layer	layer	NOUN
ijassa-93	101	20	∂hj	∂hj	PROPN
ijassa-93	101	21	∂αj	∂αj	PROPN
ijassa-93	101	22	=	=	SYM
ijassa-93	101	23	∂	∂	NUM
ijassa-93	101	24	αj	αj	NOUN
ijassa-93	101	25	·	·	SYM
ijassa-93	101	26	1	1	NUM
ijassa-93	101	27	1	1	NUM
ijassa-93	101	28	+	+	CCONJ
ijassa-93	101	29	e−αj(µj−θj	e−αj(µj−θj	ADJ
ijassa-93	101	30	)	)	PUNCT
ijassa-93	101	31	=	=	SYM
ijassa-93	102	1	(	(	PUNCT
ijassa-93	102	2	µj	µj	INTJ
ijassa-93	102	3	−	−	PROPN
ijassa-93	102	4	θj)hj(1−hj	θj)hj(1−hj	PROPN
ijassa-93	102	5	)	)	PUNCT
ijassa-93	102	6	∂e	∂e	PROPN
ijassa-93	102	7	∂hj	∂hj	PROPN
ijassa-93	102	8	=	=	PUNCT
ijassa-93	102	9	∑	∑	PUNCT
ijassa-93	102	10	k	k	PROPN
ijassa-93	102	11	∂e	∂e	PROPN
ijassa-93	102	12	∂µk	∂µk	PROPN
ijassa-93	102	13	·	·	PUNCT
ijassa-93	102	14	∂µk	∂µk	PROPN
ijassa-93	102	15	hj	hj	PROPN
ijassa-93	102	16	=	=	SYM
ijassa-93	102	17	∑	∑	PUNCT
ijassa-93	102	18	k	k	PROPN
ijassa-93	102	19	∂e	∂e	PROPN
ijassa-93	102	20	∂µk	∂µk	PROPN
ijassa-93	102	21	·	·	PUNCT
ijassa-93	102	22	wkj	wkj	PROPN
ijassa-93	102	23	∂e	∂e	PROPN
ijassa-93	102	24	∂µk	∂µk	PROPN
ijassa-93	102	25	=	=	SYM
ijassa-93	102	26	∂e	∂e	PROPN
ijassa-93	102	27	∂ok	∂ok	PROPN
ijassa-93	102	28	·	·	PUNCT
ijassa-93	103	1	ok	ok	INTJ
ijassa-93	103	2	µj	µj	PROPN
ijassa-93	103	3	=	=	SYM
ijassa-93	103	4	−(tk	−(tk	PROPN
ijassa-93	103	5	−ok)αkok(1−ok	−ok)αkok(1−ok	NOUN
ijassa-93	103	6	)	)	PUNCT
ijassa-93	103	7	substitute	substitute	NOUN
ijassa-93	103	8	in	in	ADP
ijassa-93	103	9	turns	turn	NOUN
ijassa-93	103	10	would	would	AUX
ijassa-93	103	11	be	be	AUX
ijassa-93	103	12	:	:	PUNCT
ijassa-93	103	13	∆αj	∆αj	ADV
ijassa-93	103	14	=	=	PUNCT
ijassa-93	103	15	η(µj	η(µj	VERB
ijassa-93	103	16	−	−	PROPN
ijassa-93	103	17	θj)hj(1−hj	θj)hj(1−hj	PROPN
ijassa-93	103	18	)	)	PUNCT
ijassa-93	103	19	∑	∑	PROPN
ijassa-93	104	1	k	k	PROPN
ijassa-93	104	2	(	(	PUNCT
ijassa-93	104	3	tk	tk	PROPN
ijassa-93	104	4	−ok)αkok(1−ok)wkj	−ok)αkok(1−ok)wkj	PROPN
ijassa-93	104	5	using	use	VERB
ijassa-93	104	6	driving	drive	VERB
ijassa-93	104	7	function	function	NOUN
ijassa-93	104	8	via	via	ADP
ijassa-93	104	9	the	the	DET
ijassa-93	104	10	improved	improved	ADJ
ijassa-93	104	11	gradient	gradient	NOUN
ijassa-93	104	12	,	,	PUNCT
ijassa-93	104	13	we	we	PRON
ijassa-93	104	14	can	can	AUX
ijassa-93	104	15	adjust	adjust	VERB
ijassa-93	104	16	the	the	DET
ijassa-93	104	17	gradient	gradient	NOUN
ijassa-93	104	18	of	of	ADP
ijassa-93	104	19	sigmoid	sigmoid	NOUN
ijassa-93	104	20	to	to	ADP
ijassa-93	104	21	its	its	PRON
ijassa-93	104	22	best	good	ADJ
ijassa-93	104	23	in	in	ADP
ijassa-93	104	24	the	the	DET
ijassa-93	104	25	process	process	NOUN
ijassa-93	104	26	.	.	PUNCT
ijassa-93	105	1	to	to	PART
ijassa-93	105	2	study	study	VERB
ijassa-93	105	3	the	the	DET
ijassa-93	105	4	photo	photo	NOUN
ijassa-93	105	5	sample	sample	NOUN
ijassa-93	105	6	,	,	PUNCT
ijassa-93	105	7	we	we	PRON
ijassa-93	105	8	memorize	memorize	VERB
ijassa-93	105	9	the	the	DET
ijassa-93	105	10	character	character	NOUN
ijassa-93	105	11	feature	feature	NOUN
ijassa-93	105	12	in	in	ADP
ijassa-93	105	13	the	the	DET
ijassa-93	105	14	sample	sample	NOUN
ijassa-93	105	15	and	and	CCONJ
ijassa-93	105	16	find	find	VERB
ijassa-93	105	17	the	the	DET
ijassa-93	105	18	weight	weight	NOUN
ijassa-93	105	19	and	and	CCONJ
ijassa-93	105	20	threshold	threshold	NOUN
ijassa-93	105	21	,	,	PUNCT
ijassa-93	105	22	and	and	CCONJ
ijassa-93	105	23	then	then	ADV
ijassa-93	105	24	we	we	PRON
ijassa-93	105	25	establish	establish	VERB
ijassa-93	105	26	and	and	CCONJ
ijassa-93	105	27	improve	improve	VERB
ijassa-93	105	28	bp	bp	PROPN
ijassa-93	105	29	neural	neural	ADJ
ijassa-93	105	30	network	network	NOUN
ijassa-93	105	31	.	.	PUNCT
ijassa-93	106	1	then	then	ADV
ijassa-93	106	2	we	we	PRON
ijassa-93	106	3	can	can	AUX
ijassa-93	106	4	input	input	VERB
ijassa-93	106	5	a	a	DET
ijassa-93	106	6	photo	photo	NOUN
ijassa-93	106	7	,	,	PUNCT
ijassa-93	106	8	using	use	VERB
ijassa-93	106	9	bp	bp	PROPN
ijassa-93	106	10	neural	neural	ADJ
ijassa-93	106	11	network	network	NOUN
ijassa-93	106	12	,	,	PUNCT
ijassa-93	106	13	to	to	PART
ijassa-93	106	14	find	find	VERB
ijassa-93	106	15	a	a	DET
ijassa-93	106	16	closest	close	ADJ
ijassa-93	106	17	character	character	NOUN
ijassa-93	106	18	to	to	ADP
ijassa-93	106	19	the	the	DET
ijassa-93	106	20	input	input	NOUN
ijassa-93	106	21	chinese	chinese	ADJ
ijassa-93	106	22	character	character	NOUN
ijassa-93	106	23	in	in	ADP
ijassa-93	106	24	the	the	DET
ijassa-93	106	25	sample	sample	NOUN
ijassa-93	106	26	.	.	PUNCT
ijassa-93	107	1	the	the	DET
ijassa-93	107	2	steps	step	NOUN
ijassa-93	107	3	as	as	SCONJ
ijassa-93	107	4	follows	follow	VERB
ijassa-93	107	5	:	:	PUNCT
ijassa-93	107	6	(	(	PUNCT
ijassa-93	107	7	1)network	1)network	NUM
ijassa-93	107	8	topologymost	topologymost	NOUN
ijassa-93	107	9	characters	character	NOUN
ijassa-93	107	10	to	to	PART
ijassa-93	107	11	be	be	AUX
ijassa-93	107	12	identified	identify	VERB
ijassa-93	107	13	are	be	AUX
ijassa-93	107	14	chinese	chinese	ADJ
ijassa-93	107	15	characters	character	NOUN
ijassa-93	107	16	.	.	PUNCT
ijassa-93	108	1	because	because	SCONJ
ijassa-93	108	2	the	the	DET
ijassa-93	108	3	chinese	chinese	ADJ
ijassa-93	108	4	character	character	NOUN
ijassa-93	108	5	structure	structure	NOUN
ijassa-93	108	6	is	be	AUX
ijassa-93	108	7	quadrel	quadrel	PROPN
ijassa-93	108	8	,	,	PUNCT
ijassa-93	108	9	we	we	PRON
ijassa-93	108	10	divide	divide	VERB
ijassa-93	108	11	them	they	PRON
ijassa-93	108	12	into	into	ADP
ijassa-93	108	13	26×26	26×26	NUM
ijassa-93	108	14	unit	unit	NOUN
ijassa-93	108	15	.	.	PUNCT
ijassa-93	109	1	each	each	DET
ijassa-93	109	2	unit	unit	NOUN
ijassa-93	109	3	corresponds	correspond	VERB
ijassa-93	109	4	to	to	ADP
ijassa-93	109	5	one	one	NUM
ijassa-93	109	6	input	input	NOUN
ijassa-93	109	7	;	;	PUNCT
ijassa-93	109	8	hence	hence	ADV
ijassa-93	109	9	the	the	DET
ijassa-93	109	10	output	output	NOUN
ijassa-93	109	11	nodes	node	NOUN
ijassa-93	109	12	in	in	ADP
ijassa-93	109	13	the	the	DET
ijassa-93	109	14	neural	neural	ADJ
ijassa-93	109	15	74	74	NUM
ijassa-93	109	16	jinghua	jinghua	PROPN
ijassa-93	109	17	hu	hu	PROPN
ijassa-93	109	18	:	:	PUNCT
ijassa-93	109	19	the	the	DET
ijassa-93	109	20	research	research	NOUN
ijassa-93	109	21	on	on	ADP
ijassa-93	109	22	effective	effective	ADJ
ijassa-93	109	23	video	video	NOUN
ijassa-93	109	24	scene	scene	NOUN
ijassa-93	109	25	character	character	NOUN
ijassa-93	109	26	extraction	extraction	NOUN
ijassa-93	109	27	algorithm	algorithm	NOUN
ijassa-93	109	28	...	...	PUNCT
ijassa-93	110	1	network	network	NOUN
ijassa-93	110	2	would	would	AUX
ijassa-93	110	3	be	be	AUX
ijassa-93	110	4	26×26=676	26×26=676	NUM
ijassa-93	110	5	.	.	PUNCT
ijassa-93	111	1	the	the	DET
ijassa-93	111	2	numbers	number	NOUN
ijassa-93	111	3	of	of	ADP
ijassa-93	111	4	the	the	DET
ijassa-93	111	5	output	output	NOUN
ijassa-93	111	6	nodes	node	NOUN
ijassa-93	111	7	in	in	ADP
ijassa-93	111	8	the	the	DET
ijassa-93	111	9	neural	neural	ADJ
ijassa-93	111	10	network	network	NOUN
ijassa-93	111	11	is	be	AUX
ijassa-93	111	12	the	the	DET
ijassa-93	111	13	same	same	ADJ
ijassa-93	111	14	with	with	SCONJ
ijassa-93	111	15	the	the	DET
ijassa-93	111	16	characters	character	NOUN
ijassa-93	111	17	to	to	PART
ijassa-93	111	18	be	be	AUX
ijassa-93	111	19	identified	identify	VERB
ijassa-93	111	20	.	.	PUNCT
ijassa-93	112	1	(	(	PUNCT
ijassa-93	112	2	2)output	2)output	NUM
ijassa-93	112	3	node	node	ADJ
ijassa-93	112	4	confirmnesscalculation	confirmnesscalculation	NOUN
ijassa-93	112	5	of	of	ADP
ijassa-93	112	6	node	node	ADJ
ijassa-93	112	7	output	output	NOUN
ijassa-93	112	8	is	be	AUX
ijassa-93	112	9	equivalent	equivalent	ADJ
ijassa-93	112	10	to	to	ADP
ijassa-93	112	11	calculation	calculation	NOUN
ijassa-93	112	12	of	of	ADP
ijassa-93	112	13	the	the	DET
ijassa-93	112	14	function	function	NOUN
ijassa-93	112	15	sigmoid	sigmoid	NOUN
ijassa-93	112	16	output	output	NOUN
ijassa-93	112	17	.	.	PUNCT
ijassa-93	113	1	because	because	SCONJ
ijassa-93	113	2	of	of	ADP
ijassa-93	113	3	the	the	DET
ijassa-93	113	4	functions	function	NOUN
ijassa-93	113	5	nature	nature	NOUN
ijassa-93	113	6	,	,	PUNCT
ijassa-93	113	7	we	we	PRON
ijassa-93	113	8	would	would	AUX
ijassa-93	113	9	be	be	AUX
ijassa-93	113	10	aware	aware	ADJ
ijassa-93	113	11	of	of	ADP
ijassa-93	113	12	its	its	PRON
ijassa-93	113	13	value	value	NOUN
ijassa-93	113	14	range	range	NOUN
ijassa-93	113	15	[	[	X
ijassa-93	113	16	0,1	0,1	NUM
ijassa-93	113	17	]	]	PUNCT
ijassa-93	113	18	.	.	PUNCT
ijassa-93	114	1	the	the	PRON
ijassa-93	114	2	closer	close	ADV
ijassa-93	114	3	the	the	DET
ijassa-93	114	4	value	value	NOUN
ijassa-93	114	5	is	be	AUX
ijassa-93	114	6	to	to	ADP
ijassa-93	114	7	the	the	DET
ijassa-93	114	8	limits	limit	NOUN
ijassa-93	114	9	,	,	PUNCT
ijassa-93	114	10	the	the	PRON
ijassa-93	114	11	less	less	ADV
ijassa-93	114	12	sensitive	sensitive	ADJ
ijassa-93	114	13	the	the	DET
ijassa-93	114	14	function	function	NOUN
ijassa-93	114	15	value	value	NOUN
ijassa-93	114	16	changes	change	NOUN
ijassa-93	114	17	to	to	ADP
ijassa-93	114	18	its	its	PRON
ijassa-93	114	19	independent	independent	ADJ
ijassa-93	114	20	variable	variable	NOUN
ijassa-93	114	21	.	.	PUNCT
ijassa-93	115	1	therefore	therefore	ADV
ijassa-93	115	2	,	,	PUNCT
ijassa-93	115	3	we	we	PRON
ijassa-93	115	4	can	can	AUX
ijassa-93	115	5	enhance	enhance	VERB
ijassa-93	115	6	the	the	DET
ijassa-93	115	7	nodes	node	NOUN
ijassa-93	115	8	precision	precision	NOUN
ijassa-93	115	9	by	by	ADP
ijassa-93	115	10	its	its	PRON
ijassa-93	115	11	output	output	NOUN
ijassa-93	115	12	-	-	PUNCT
ijassa-93	115	13	to	to	ADP
ijassa-93	115	14	-	-	PUNCT
ijassa-93	115	15	input	input	NOUN
ijassa-93	115	16	sensitivity	sensitivity	NOUN
ijassa-93	115	17	.	.	PUNCT
ijassa-93	116	1	we	we	PRON
ijassa-93	116	2	suppose	suppose	VERB
ijassa-93	116	3	t	t	PROPN
ijassa-93	116	4	is	be	AUX
ijassa-93	116	5	0.9	0.9	NUM
ijassa-93	116	6	,	,	PUNCT
ijassa-93	116	7	and	and	CCONJ
ijassa-93	116	8	f	f	PROPN
ijassa-93	116	9	is	be	AUX
ijassa-93	116	10	0.1	0.1	NUM
ijassa-93	116	11	;	;	PUNCT
ijassa-93	116	12	for	for	ADP
ijassa-93	116	13	input	input	NOUN
ijassa-93	116	14	,	,	PUNCT
ijassa-93	116	15	if	if	SCONJ
ijassa-93	116	16	with	with	ADP
ijassa-93	116	17	stroke	stroke	NOUN
ijassa-93	116	18	,	,	PUNCT
ijassa-93	116	19	then	then	ADV
ijassa-93	116	20	input	input	VERB
ijassa-93	116	21	0.9	0.9	NUM
ijassa-93	116	22	,	,	PUNCT
ijassa-93	116	23	if	if	SCONJ
ijassa-93	116	24	not	not	PART
ijassa-93	116	25	,	,	PUNCT
ijassa-93	116	26	we	we	PRON
ijassa-93	116	27	input	input	VERB
ijassa-93	116	28	0.1	0.1	NUM
ijassa-93	116	29	.	.	PUNCT
ijassa-93	117	1	and	and	CCONJ
ijassa-93	117	2	we	we	PRON
ijassa-93	117	3	deal	deal	VERB
ijassa-93	117	4	with	with	ADP
ijassa-93	117	5	output	output	NOUN
ijassa-93	117	6	the	the	DET
ijassa-93	117	7	same	same	ADJ
ijassa-93	117	8	way	way	NOUN
ijassa-93	117	9	.	.	PUNCT
ijassa-93	118	1	(	(	PUNCT
ijassa-93	118	2	3)edit	3)edit	NUM
ijassa-93	118	3	of	of	ADP
ijassa-93	118	4	training	train	VERB
ijassa-93	118	5	samplethe	samplethe	DET
ijassa-93	118	6	more	more	ADJ
ijassa-93	118	7	samples	sample	NOUN
ijassa-93	118	8	the	the	DET
ijassa-93	118	9	collection	collection	NOUN
ijassa-93	118	10	owns	own	VERB
ijassa-93	118	11	,	,	PUNCT
ijassa-93	118	12	the	the	DET
ijassa-93	118	13	more	more	ADV
ijassa-93	118	14	corresponding	corresponding	ADJ
ijassa-93	118	15	sample	sample	NOUN
ijassa-93	118	16	each	each	DET
ijassa-93	118	17	chinese	chinese	ADJ
ijassa-93	118	18	character	character	NOUN
ijassa-93	118	19	has	have	VERB
ijassa-93	118	20	.	.	PUNCT
ijassa-93	119	1	also	also	ADV
ijassa-93	119	2	the	the	DET
ijassa-93	119	3	better	well	ADV
ijassa-93	119	4	identifying	identify	VERB
ijassa-93	119	5	ability	ability	NOUN
ijassa-93	119	6	the	the	DET
ijassa-93	119	7	neural	neural	ADJ
ijassa-93	119	8	network	network	NOUN
ijassa-93	119	9	is	be	AUX
ijassa-93	119	10	after	after	ADP
ijassa-93	119	11	being	be	AUX
ijassa-93	119	12	trained	train	VERB
ijassa-93	119	13	,	,	PUNCT
ijassa-93	119	14	however	however	ADV
ijassa-93	119	15	,	,	PUNCT
ijassa-93	119	16	the	the	DET
ijassa-93	119	17	more	more	ADJ
ijassa-93	119	18	time	time	NOUN
ijassa-93	119	19	the	the	DET
ijassa-93	119	20	training	training	NOUN
ijassa-93	119	21	takes	take	VERB
ijassa-93	119	22	.	.	PUNCT
ijassa-93	120	1	here	here	ADV
ijassa-93	120	2	we	we	PRON
ijassa-93	120	3	only	only	ADV
ijassa-93	120	4	take	take	VERB
ijassa-93	120	5	the	the	DET
ijassa-93	120	6	sized	sized	ADJ
ijassa-93	120	7	18	18	NUM
ijassa-93	120	8	thick	thick	ADJ
ijassa-93	120	9	song	song	NOUN
ijassa-93	120	10	ti	ti	NOUN
ijassa-93	120	11	character	character	NOUN
ijassa-93	120	12	as	as	ADP
ijassa-93	120	13	sample	sample	NOUN
ijassa-93	120	14	.	.	PUNCT
ijassa-93	121	1	there	there	PRON
ijassa-93	121	2	are	be	VERB
ijassa-93	121	3	english	english	ADJ
ijassa-93	121	4	characters	character	NOUN
ijassa-93	121	5	,	,	PUNCT
ijassa-93	121	6	numbers	number	NOUN
ijassa-93	121	7	,	,	PUNCT
ijassa-93	121	8	3,775	3,775	NUM
ijassa-93	121	9	chinese	chinese	ADJ
ijassa-93	121	10	characters	character	NOUN
ijassa-93	121	11	of	of	ADP
ijassa-93	121	12	one	one	NUM
ijassa-93	121	13	-	-	PUNCT
ijassa-93	121	14	level	level	NOUN
ijassa-93	121	15	,	,	PUNCT
ijassa-93	121	16	and	and	CCONJ
ijassa-93	121	17	basic	basic	ADJ
ijassa-93	121	18	punctuation	punctuation	NOUN
ijassa-93	121	19	in	in	ADP
ijassa-93	121	20	the	the	DET
ijassa-93	121	21	sample	sample	NOUN
ijassa-93	121	22	collection	collection	NOUN
ijassa-93	121	23	.	.	PUNCT
ijassa-93	122	1	(	(	PUNCT
ijassa-93	122	2	4)experiments	4)experiments	NUM
ijassa-93	122	3	and	and	CCONJ
ijassa-93	122	4	analysis	analysis	NOUN
ijassa-93	122	5	the	the	DET
ijassa-93	122	6	identifying	identify	VERB
ijassa-93	122	7	result	result	NOUN
ijassa-93	122	8	of	of	ADP
ijassa-93	122	9	bp	bp	PROPN
ijassa-93	122	10	-	-	PUNCT
ijassa-93	122	11	algorithm	algorithm	NOUN
ijassa-93	122	12	-	-	PUNCT
ijassa-93	122	13	based	base	VERB
ijassa-93	122	14	character	character	NOUN
ijassa-93	122	15	identifying	identify	VERB
ijassa-93	122	16	algorithm	algorithm	NOUN
ijassa-93	122	17	is	be	AUX
ijassa-93	122	18	shown	show	VERB
ijassa-93	122	19	in	in	ADP
ijassa-93	122	20	fig.5	fig.5	PROPN
ijassa-93	122	21	.	.	PUNCT
ijassa-93	123	1	this	this	DET
ijassa-93	123	2	system	system	NOUN
ijassa-93	123	3	is	be	AUX
ijassa-93	123	4	designed	design	VERB
ijassa-93	123	5	to	to	PART
ijassa-93	123	6	deal	deal	VERB
ijassa-93	123	7	with	with	ADP
ijassa-93	123	8	images	image	NOUN
ijassa-93	123	9	including	include	VERB
ijassa-93	123	10	character	character	NOUN
ijassa-93	123	11	,	,	PUNCT
ijassa-93	123	12	whose	whose	DET
ijassa-93	123	13	recognition	recognition	NOUN
ijassa-93	123	14	rate	rate	NOUN
ijassa-93	123	15	is	be	AUX
ijassa-93	123	16	above	above	ADP
ijassa-93	123	17	96	96	NUM
ijassa-93	123	18	%	%	NOUN
ijassa-93	123	19	,	,	PUNCT
ijassa-93	123	20	and	and	CCONJ
ijassa-93	123	21	its	its	PRON
ijassa-93	123	22	response	response	NOUN
ijassa-93	123	23	time	time	NOUN
ijassa-93	123	24	is	be	AUX
ijassa-93	123	25	5ms	5ms	ADJ
ijassa-93	123	26	/	/	SYM
ijassa-93	123	27	100	100	NUM
ijassa-93	123	28	words	word	NOUN
ijassa-93	123	29	.	.	PUNCT
ijassa-93	124	1	compared	compare	VERB
ijassa-93	124	2	to	to	ADP
ijassa-93	124	3	the	the	DET
ijassa-93	124	4	traditional	traditional	ADJ
ijassa-93	124	5	recognition	recognition	NOUN
ijassa-93	124	6	method	method	NOUN
ijassa-93	124	7	it	it	PRON
ijassa-93	124	8	has	have	AUX
ijassa-93	124	9	been	be	AUX
ijassa-93	124	10	improved	improve	VERB
ijassa-93	124	11	,	,	PUNCT
ijassa-93	124	12	whose	whose	DET
ijassa-93	124	13	recognition	recognition	NOUN
ijassa-93	124	14	result	result	NOUN
ijassa-93	124	15	is	be	AUX
ijassa-93	124	16	better	well	ADJ
ijassa-93	124	17	.	.	PUNCT
ijassa-93	125	1	however	however	ADV
ijassa-93	125	2	,	,	PUNCT
ijassa-93	125	3	the	the	PRON
ijassa-93	125	4	worse	bad	ADJ
ijassa-93	125	5	the	the	DET
ijassa-93	125	6	images	image	NOUN
ijassa-93	125	7	quality	quality	NOUN
ijassa-93	125	8	will	will	AUX
ijassa-93	125	9	result	result	VERB
ijassa-93	125	10	in	in	ADP
ijassa-93	125	11	the	the	DET
ijassa-93	125	12	worse	bad	ADJ
ijassa-93	125	13	the	the	DET
ijassa-93	125	14	recognition	recognition	NOUN
ijassa-93	125	15	result	result	VERB
ijassa-93	125	16	.	.	PUNCT
ijassa-93	126	1	in	in	ADP
ijassa-93	126	2	the	the	DET
ijassa-93	126	3	image	image	NOUN
ijassa-93	126	4	treatment	treatment	NOUN
ijassa-93	126	5	phase	phase	NOUN
ijassa-93	126	6	,	,	PUNCT
ijassa-93	126	7	the	the	DET
ijassa-93	126	8	wiping	wiping	NOUN
ijassa-93	126	9	-	-	PUNCT
ijassa-93	126	10	out	out	NOUN
ijassa-93	126	11	of	of	ADP
ijassa-93	126	12	useless	useless	ADJ
ijassa-93	126	13	information	information	NOUN
ijassa-93	126	14	is	be	AUX
ijassa-93	126	15	partial	partial	ADJ
ijassa-93	126	16	,	,	PUNCT
ijassa-93	126	17	which	which	PRON
ijassa-93	126	18	has	have	VERB
ijassa-93	126	19	an	an	DET
ijassa-93	126	20	effect	effect	NOUN
ijassa-93	126	21	on	on	ADP
ijassa-93	126	22	character	character	NOUN
ijassa-93	126	23	recognition	recognition	NOUN
ijassa-93	126	24	and	and	CCONJ
ijassa-93	126	25	the	the	DET
ijassa-93	126	26	whole	whole	ADJ
ijassa-93	126	27	recognition	recognition	NOUN
ijassa-93	126	28	result	result	VERB
ijassa-93	126	29	.	.	PUNCT
ijassa-93	127	1	fig.5	fig.5	VERB
ijassa-93	127	2	results	result	NOUN
ijassa-93	127	3	of	of	ADP
ijassa-93	127	4	character	character	NOUN
ijassa-93	127	5	identifying	identify	VERB
ijassa-93	127	6	algorithm	algorithm	NOUN
ijassa-93	127	7	5	5	NUM
ijassa-93	127	8	conclusion	conclusion	NOUN
ijassa-93	127	9	according	accord	VERB
ijassa-93	127	10	to	to	ADP
ijassa-93	127	11	the	the	DET
ijassa-93	127	12	above	above	ADJ
ijassa-93	127	13	discussion	discussion	NOUN
ijassa-93	127	14	,	,	PUNCT
ijassa-93	127	15	the	the	DET
ijassa-93	127	16	identifying	identify	VERB
ijassa-93	127	17	result	result	NOUN
ijassa-93	127	18	of	of	ADP
ijassa-93	127	19	bp	bp	PROPN
ijassa-93	127	20	-	-	PUNCT
ijassa-93	127	21	algorithm	algorithm	NOUN
ijassa-93	127	22	-	-	PUNCT
ijassa-93	127	23	based	base	VERB
ijassa-93	127	24	character	character	NOUN
ijassa-93	127	25	identifying	identify	VERB
ijassa-93	127	26	algorithm	algorithm	NOUN
ijassa-93	127	27	can	can	AUX
ijassa-93	127	28	improve	improve	VERB
ijassa-93	127	29	the	the	DET
ijassa-93	127	30	efficiency	efficiency	NOUN
ijassa-93	127	31	of	of	ADP
ijassa-93	127	32	ic	ic	PRON
ijassa-93	127	33	identifying	identify	VERB
ijassa-93	127	34	in	in	ADP
ijassa-93	127	35	aoi	aoi	NOUN
ijassa-93	127	36	system	system	NOUN
ijassa-93	127	37	.	.	PUNCT
ijassa-93	128	1	advances	advance	NOUN
ijassa-93	128	2	in	in	ADP
ijassa-93	128	3	systems	system	NOUN
ijassa-93	128	4	science	science	NOUN
ijassa-93	128	5	and	and	CCONJ
ijassa-93	128	6	applications	application	NOUN
ijassa-93	128	7	(	(	PUNCT
ijassa-93	128	8	2012	2012	NUM
ijassa-93	128	9	)	)	PUNCT
ijassa-93	128	10	,	,	PUNCT
ijassa-93	128	11	vol.12	vol.12	NOUN
ijassa-93	128	12	,	,	PUNCT
ijassa-93	128	13	no.1	no.1	X
ijassa-93	128	14	75	75	NUM
ijassa-93	128	15	references	reference	NOUN
ijassa-93	128	16	[	[	X
ijassa-93	128	17	1	1	NUM
ijassa-93	128	18	]	]	PUNCT
ijassa-93	128	19	yu	yu	PROPN
ijassa-93	128	20	zhong	zhong	PROPN
ijassa-93	128	21	,	,	PUNCT
ijassa-93	128	22	k.karu	k.karu	PROPN
ijassa-93	128	23	,	,	PUNCT
ijassa-93	128	24	a.k.jain	a.k.jain	NOUN
ijassa-93	128	25	.	.	PUNCT
ijassa-93	129	1	(	(	PUNCT
ijassa-93	129	2	1995	1995	NUM
ijassa-93	129	3	)	)	PUNCT
ijassa-93	129	4	,	,	PUNCT
ijassa-93	129	5	“	"	PUNCT
ijassa-93	129	6	locating	locate	VERB
ijassa-93	129	7	text	text	NOUN
ijassa-93	129	8	in	in	ADP
ijassa-93	129	9	complex	complex	ADJ
ijassa-93	129	10	color	color	NOUN
ijassa-93	129	11	images	image	NOUN
ijassa-93	129	12	”	"	PUNCT
ijassa-93	129	13	,	,	PUNCT
ijassa-93	129	14	in	in	ADP
ijassa-93	129	15	proceedings	proceeding	NOUN
ijassa-93	129	16	of	of	ADP
ijassa-93	129	17	the	the	DET
ijassa-93	129	18	third	third	ADJ
ijassa-93	129	19	international	international	ADJ
ijassa-93	129	20	conference	conference	NOUN
ijassa-93	129	21	on	on	ADP
ijassa-93	129	22	document	document	NOUN
ijassa-93	129	23	analysis	analysis	NOUN
ijassa-93	129	24	and	and	CCONJ
ijassa-93	129	25	recognition	recognition	NOUN
ijassa-93	129	26	,	,	PUNCT
ijassa-93	129	27	vol.1	vol.1	PROPN
ijassa-93	129	28	,	,	PUNCT
ijassa-93	129	29	no.14	no.14	PROPN
ijassa-93	129	30	,	,	PUNCT
ijassa-93	129	31	pp.146	pp.146	PROPN
ijassa-93	129	32	-	-	PUNCT
ijassa-93	129	33	149	149	NUM
ijassa-93	129	34	.	.	PUNCT
ijassa-93	130	1	[	[	X
ijassa-93	130	2	2	2	NUM
ijassa-93	130	3	]	]	PUNCT
ijassa-93	130	4	a.k.jain	a.k.jain	NOUN
ijassa-93	130	5	,	,	PUNCT
ijassa-93	130	6	b.yu	b.yu	PROPN
ijassa-93	130	7	.	.	PUNCT
ijassa-93	131	1	(	(	PUNCT
ijassa-93	131	2	1998	1998	NUM
ijassa-93	131	3	)	)	PUNCT
ijassa-93	131	4	,	,	PUNCT
ijassa-93	131	5	“	"	PUNCT
ijassa-93	131	6	automatic	automatic	ADJ
ijassa-93	131	7	text	text	NOUN
ijassa-93	131	8	location	location	NOUN
ijassa-93	131	9	in	in	ADP
ijassa-93	131	10	images	image	NOUN
ijassa-93	131	11	and	and	CCONJ
ijassa-93	131	12	video	video	NOUN
ijassa-93	131	13	frames	frame	NOUN
ijassa-93	131	14	”	"	PUNCT
ijassa-93	131	15	,	,	PUNCT
ijassa-93	131	16	pattern	pattern	NOUN
ijassa-93	131	17	recognition	recognition	NOUN
ijassa-93	131	18	,	,	PUNCT
ijassa-93	131	19	vol.13	vol.13	PROPN
ijassa-93	131	20	,	,	PUNCT
ijassa-93	131	21	no.12	no.12	NOUN
ijassa-93	131	22	,	,	PUNCT
ijassa-93	131	23	pp.2005	pp.2005	PROPN
ijassa-93	131	24	-	-	PUNCT
ijassa-93	131	25	2076	2076	NUM
ijassa-93	131	26	.	.	PUNCT
ijassa-93	132	1	[	[	X
ijassa-93	132	2	3	3	NUM
ijassa-93	132	3	]	]	X
ijassa-93	132	4	m.a.smith	m.a.smith	NOUN
ijassa-93	132	5	,	,	PUNCT
ijassa-93	132	6	t.kanade	t.kanade	NOUN
ijassa-93	132	7	.	.	PUNCT
ijassa-93	133	1	(	(	PUNCT
ijassa-93	133	2	1995	1995	NUM
ijassa-93	133	3	)	)	PUNCT
ijassa-93	133	4	,	,	PUNCT
ijassa-93	133	5	“	"	PUNCT
ijassa-93	133	6	video	video	NOUN
ijassa-93	133	7	skimming	skimming	NOUN
ijassa-93	133	8	for	for	ADP
ijassa-93	133	9	quick	quick	ADJ
ijassa-93	133	10	browsing	browsing	NOUN
ijassa-93	133	11	based	base	VERB
ijassa-93	133	12	on	on	ADP
ijassa-93	133	13	audio	audio	ADJ
ijassa-93	133	14	and	and	CCONJ
ijassa-93	133	15	image	image	NOUN
ijassa-93	133	16	characterization	characterization	NOUN
ijassa-93	133	17	”	"	PUNCT
ijassa-93	133	18	,	,	PUNCT
ijassa-93	133	19	3technology	3technology	NUM
ijassa-93	133	20	report	report	NOUN
ijassa-93	133	21	cmu	cmu	NOUN
ijassa-93	133	22	-	-	PUNCT
ijassa-93	133	23	cs-95	cs-95	NOUN
ijassa-93	133	24	-	-	PUNCT
ijassa-93	133	25	186	186	NUM
ijassa-93	133	26	,	,	PUNCT
ijassa-93	133	27	pp.13	pp.13	NOUN
ijassa-93	133	28	-	-	NOUN
ijassa-93	133	29	15	15	NUM
ijassa-93	133	30	.	.	PUNCT
ijassa-93	134	1	[	[	X
ijassa-93	134	2	4	4	X
ijassa-93	134	3	]	]	PUNCT
ijassa-93	134	4	t.sato	t.sato	NOUN
ijassa-93	134	5	,	,	PUNCT
ijassa-93	134	6	t.kanade	t.kanade	NOUN
ijassa-93	134	7	.	.	PUNCT
ijassa-93	135	1	(	(	PUNCT
ijassa-93	135	2	1995	1995	NUM
ijassa-93	135	3	)	)	PUNCT
ijassa-93	135	4	,	,	PUNCT
ijassa-93	135	5	“	"	PUNCT
ijassa-93	135	6	video	video	NOUN
ijassa-93	135	7	ocr	ocr	NOUN
ijassa-93	135	8	:	:	PUNCT
ijassa-93	135	9	indexing	index	VERB
ijassa-93	135	10	digital	digital	ADJ
ijassa-93	135	11	news	news	NOUN
ijassa-93	135	12	libraries	library	NOUN
ijassa-93	135	13	by	by	ADP
ijassa-93	135	14	recognition	recognition	NOUN
ijassa-93	135	15	of	of	ADP
ijassa-93	135	16	superimposed	superimpose	VERB
ijassa-93	135	17	caption	caption	NOUN
ijassa-93	135	18	”	"	PUNCT
ijassa-93	135	19	,	,	PUNCT
ijassa-93	135	20	multimedia	multimedia	NOUN
ijassa-93	135	21	systems	system	NOUN
ijassa-93	135	22	,	,	PUNCT
ijassa-93	135	23	vo.l7	vo.l7	NOUN
ijassa-93	135	24	,	,	PUNCT
ijassa-93	135	25	no.5	no.5	PROPN
ijassa-93	135	26	,	,	PUNCT
ijassa-93	135	27	pp.385–395	pp.385–395	PROPN
ijassa-93	135	28	.	.	PUNCT
ijassa-93	136	1	[	[	X
ijassa-93	136	2	5	5	X
ijassa-93	136	3	]	]	X
ijassa-93	136	4	h.j	h.j	PROPN
ijassa-93	136	5	.	.	PROPN
ijassa-93	136	6	zhang	zhang	PROPN
ijassa-93	136	7	,	,	PUNCT
ijassa-93	136	8	c.y	c.y	PROPN
ijassa-93	136	9	.	.	PROPN
ijassa-93	136	10	low	low	PROPN
ijassa-93	136	11	,	,	PUNCT
ijassa-93	136	12	s.w	s.w	PROPN
ijassa-93	136	13	.	.	PROPN
ijassa-93	136	14	smoliar	smoliar	PROPN
ijassa-93	136	15	,	,	PUNCT
ijassa-93	136	16	and	and	CCONJ
ijassa-93	136	17	j.h	j.h	PROPN
ijassa-93	136	18	.	.	PROPN
ijassa-93	136	19	wu	wu	PROPN
ijassa-93	136	20	.	.	PUNCT
ijassa-93	137	1	(	(	PUNCT
ijassa-93	137	2	1995	1995	NUM
ijassa-93	137	3	)	)	PUNCT
ijassa-93	137	4	,	,	PUNCT
ijassa-93	137	5	“	"	PUNCT
ijassa-93	137	6	video	video	NOUN
ijassa-93	137	7	parsing	parsing	NOUN
ijassa-93	137	8	,	,	PUNCT
ijassa-93	137	9	retrieval	retrieval	NOUN
ijassa-93	137	10	and	and	CCONJ
ijassa-93	137	11	browsing	browse	VERB
ijassa-93	137	12	:	:	PUNCT
ijassa-93	137	13	an	an	DET
ijassa-93	137	14	integrated	integrated	ADJ
ijassa-93	137	15	and	and	CCONJ
ijassa-93	137	16	content	content	NOUN
ijassa-93	137	17	-	-	PUNCT
ijassa-93	137	18	based	base	VERB
ijassa-93	137	19	solution	solution	NOUN
ijassa-93	137	20	”	"	PUNCT
ijassa-93	137	21	,	,	PUNCT
ijassa-93	137	22	proc	proc	PROPN
ijassa-93	137	23	.	.	PUNCT
ijassa-93	137	24	acm	acm	PROPN
ijassa-93	137	25	multimedia	multimedia	PROPN
ijassa-93	137	26	,	,	PUNCT
ijassa-93	137	27	pp.15	pp.15	PROPN
ijassa-93	137	28	-	-	PROPN
ijassa-93	137	29	24	24	NUM
ijassa-93	137	30	.	.	PUNCT
ijassa-93	138	1	[	[	X
ijassa-93	138	2	6	6	NUM
ijassa-93	138	3	]	]	PUNCT
ijassa-93	138	4	r.lienhart	r.lienhart	NOUN
ijassa-93	138	5	,	,	PUNCT
ijassa-93	138	6	w.effelsberg	w.effelsberg	PRON
ijassa-93	138	7	.	.	PUNCT
ijassa-93	139	1	(	(	PUNCT
ijassa-93	139	2	2000	2000	NUM
ijassa-93	139	3	)	)	PUNCT
ijassa-93	139	4	,	,	PUNCT
ijassa-93	139	5	“	"	PUNCT
ijassa-93	139	6	automatic	automatic	ADJ
ijassa-93	139	7	text	text	NOUN
ijassa-93	139	8	segmentation	segmentation	NOUN
ijassa-93	139	9	and	and	CCONJ
ijassa-93	139	10	text	text	NOUN
ijassa-93	139	11	recognition	recognition	NOUN
ijassa-93	139	12	for	for	ADP
ijassa-93	139	13	video	video	NOUN
ijassa-93	139	14	indexing	indexing	NOUN
ijassa-93	139	15	”	"	PUNCT
ijassa-93	139	16	,	,	PUNCT
ijassa-93	139	17	multimedia	multimedia	NOUN
ijassa-93	139	18	systems	system	NOUN
ijassa-93	139	19	,	,	PUNCT
ijassa-93	139	20	vo.l8	vo.l8	NOUN
ijassa-93	139	21	,	,	PUNCT
ijassa-93	139	22	no.2	no.2	PROPN
ijassa-93	139	23	,	,	PUNCT
ijassa-93	139	24	pp.69	pp.69	NOUN
ijassa-93	139	25	-	-	PUNCT
ijassa-93	139	26	81	81	NUM
ijassa-93	139	27	.	.	PUNCT
ijassa-93	140	1	[	[	X
ijassa-93	140	2	7	7	X
ijassa-93	140	3	]	]	PUNCT
ijassa-93	140	4	huiping	huipe	VERB
ijassa-93	140	5	li	li	PROPN
ijassa-93	140	6	,	,	PUNCT
ijassa-93	140	7	omid	omid	PROPN
ijassa-93	140	8	kia	kia	PROPN
ijassa-93	140	9	,	,	PUNCT
ijassa-93	140	10	david	david	PROPN
ijassa-93	140	11	doermann	doermann	PROPN
ijassa-93	140	12	.	.	PUNCT
ijassa-93	141	1	(	(	PUNCT
ijassa-93	141	2	2004	2004	NUM
ijassa-93	141	3	)	)	PUNCT
ijassa-93	141	4	,	,	PUNCT
ijassa-93	141	5	“	"	PUNCT
ijassa-93	141	6	text	text	NOUN
ijassa-93	141	7	enhancement	enhancement	NOUN
ijassa-93	141	8	in	in	ADP
ijassa-93	141	9	digital	digital	ADJ
ijassa-93	141	10	videos	video	NOUN
ijassa-93	141	11	”	"	PUNCT
ijassa-93	141	12	,	,	PUNCT
ijassa-93	141	13	proc	proc	NOUN
ijassa-93	141	14	.	.	PUNCT
ijassa-93	142	1	spie99	spie99	ADJ
ijassa-93	142	2	-	-	PUNCT
ijassa-93	142	3	document	document	NOUN
ijassa-93	142	4	recognition	recognition	NOUN
ijassa-93	142	5	and	and	CCONJ
ijassa-93	142	6	retrieval	retrieval	NOUN
ijassa-93	142	7	,	,	PUNCT
ijassa-93	142	8	pp	pp	ADP
ijassa-93	142	9	3651:2	3651:2	NUM
ijassa-93	142	10	-	-	SYM
ijassa-93	142	11	9	9	NUM
ijassa-93	143	1	[	[	SYM
ijassa-93	143	2	8	8	NUM
ijassa-93	143	3	]	]	X
ijassa-93	143	4	casey	casey	NOUN
ijassa-93	143	5	r	r	NOUN
ijassa-93	143	6	g	g	NOUN
ijassa-93	143	7	,	,	PUNCT
ijassa-93	143	8	lecolinet	lecolinet	PROPN
ijassa-93	143	9	e.	e.	PROPN
ijassa-93	143	10	(	(	PUNCT
ijassa-93	143	11	1996	1996	NUM
ijassa-93	143	12	)	)	PUNCT
ijassa-93	143	13	,	,	PUNCT
ijassa-93	143	14	“	"	PUNCT
ijassa-93	143	15	a	a	DET
ijassa-93	143	16	survey	survey	NOUN
ijassa-93	143	17	of	of	ADP
ijassa-93	143	18	methods	method	NOUN
ijassa-93	143	19	and	and	CCONJ
ijassa-93	143	20	strategies	strategy	NOUN
ijassa-93	143	21	in	in	ADP
ijassa-93	143	22	character	character	NOUN
ijassa-93	143	23	segmentation	segmentation	NOUN
ijassa-93	143	24	”	"	PUNCT
ijassa-93	143	25	,	,	PUNCT
ijassa-93	143	26	ieee	ieee	NOUN
ijassa-93	143	27	transactions	transaction	NOUN
ijassa-93	143	28	on	on	ADP
ijassa-93	143	29	pattern	pattern	NOUN
ijassa-93	143	30	analysis	analysis	NOUN
ijassa-93	143	31	and	and	CCONJ
ijassa-93	143	32	machine	machine	NOUN
ijassa-93	143	33	intelligence	intelligence	NOUN
ijassa-93	143	34	,	,	PUNCT
ijassa-93	143	35	vol.18	vol.18	NOUN
ijassa-93	143	36	,	,	PUNCT
ijassa-93	143	37	no.7	no.7	PROPN
ijassa-93	143	38	,	,	PUNCT
ijassa-93	143	39	pp.690	pp.690	NOUN
ijassa-93	143	40	-	-	PUNCT
ijassa-93	143	41	706	706	NUM
ijassa-93	143	42	.	.	PUNCT