id	sid	tid	token	lemma	pos
hjic-847	1	1	hungarian	hungarian	ADJ
hjic-847	1	2	journal	journal	NOUN
hjic-847	1	3	of	of	ADP
hjic-847	1	4	industry	industry	NOUN
hjic-847	1	5	and	and	CCONJ
hjic-847	1	6	chemistry	chemistry	NOUN
hjic-847	1	7	vol	vol	NOUN
hjic-847	1	8	.	.	PUNCT
hjic-847	1	9	47(1	47(1	NUM
hjic-847	1	10	)	)	PUNCT
hjic-847	2	1	pp	pp	ADV
hjic-847	2	2	.	.	PUNCT
hjic-847	3	1	33–39	33–39	NUM
hjic-847	3	2	(	(	PUNCT
hjic-847	3	3	2019	2019	NUM
hjic-847	3	4	)	)	PUNCT
hjic-847	3	5	hjic.mk.uni-pannon.hu	hjic.mk.uni-pannon.hu	NOUN
hjic-847	3	6	doi	doi	PROPN
hjic-847	3	7	:	:	PUNCT
hjic-847	3	8	10.33927	10.33927	NUM
hjic-847	3	9	/	/	SYM
hjic-847	3	10	hjic-2019	hjic-2019	PROPN
hjic-847	3	11	-	-	PUNCT
hjic-847	3	12	06	06	NUM
hjic-847	3	13	automated	automate	VERB
hjic-847	3	14	labeling	labeling	NOUN
hjic-847	3	15	process	process	NOUN
hjic-847	3	16	for	for	ADP
hjic-847	3	17	unknown	unknown	ADJ
hjic-847	3	18	images	image	NOUN
hjic-847	3	19	in	in	ADP
hjic-847	3	20	an	an	DET
hjic-847	3	21	open	open	ADJ
hjic-847	3	22	-	-	PUNCT
hjic-847	3	23	world	world	NOUN
hjic-847	3	24	scenario	scenario	NOUN
hjic-847	3	25	dávid	dávid	PROPN
hjic-847	3	26	papp	papp	PROPN
hjic-847	3	27	*	*	PUNCT
hjic-847	3	28	1	1	NUM
hjic-847	3	29	and	and	CCONJ
hjic-847	3	30	gábor	gábor	PROPN
hjic-847	3	31	szűcs1	szűcs1	PROPN
hjic-847	3	32	1department	1department	NUM
hjic-847	3	33	of	of	ADP
hjic-847	3	34	telecommunications	telecommunication	NOUN
hjic-847	3	35	and	and	CCONJ
hjic-847	3	36	media	medium	NOUN
hjic-847	3	37	informatics	informatic	NOUN
hjic-847	3	38	,	,	PUNCT
hjic-847	3	39	budapest	budapest	PROPN
hjic-847	3	40	university	university	PROPN
hjic-847	3	41	of	of	ADP
hjic-847	3	42	technology	technology	NOUN
hjic-847	3	43	and	and	CCONJ
hjic-847	3	44	economics	economic	NOUN
hjic-847	3	45	,	,	PUNCT
hjic-847	3	46	magyar	magyar	PROPN
hjic-847	3	47	tudósok	tudósok	PROPN
hjic-847	3	48	krt	krt	PROPN
hjic-847	3	49	.	.	PUNCT
hjic-847	4	1	2	2	NUM
hjic-847	4	2	.	.	NUM
hjic-847	4	3	,	,	PUNCT
hjic-847	4	4	h-1117	h-1117	PROPN
hjic-847	4	5	budapest	budapest	NOUN
hjic-847	4	6	,	,	PUNCT
hjic-847	4	7	hungary	hungary	PROPN
hjic-847	4	8	most	most	ADJ
hjic-847	4	9	of	of	ADP
hjic-847	4	10	the	the	DET
hjic-847	4	11	recognition	recognition	NOUN
hjic-847	4	12	systems	system	NOUN
hjic-847	4	13	presume	presume	VERB
hjic-847	4	14	a	a	DET
hjic-847	4	15	controlled	control	VERB
hjic-847	4	16	,	,	PUNCT
hjic-847	4	17	well	well	ADV
hjic-847	4	18	-	-	PUNCT
hjic-847	4	19	defined	define	VERB
hjic-847	4	20	research	research	NOUN
hjic-847	4	21	setting	setting	NOUN
hjic-847	4	22	,	,	PUNCT
hjic-847	4	23	where	where	SCONJ
hjic-847	4	24	all	all	DET
hjic-847	4	25	possible	possible	ADJ
hjic-847	4	26	classes	class	NOUN
hjic-847	4	27	that	that	PRON
hjic-847	4	28	can	can	AUX
hjic-847	4	29	appear	appear	VERB
hjic-847	4	30	during	during	ADP
hjic-847	4	31	a	a	DET
hjic-847	4	32	test	test	NOUN
hjic-847	4	33	are	be	AUX
hjic-847	4	34	known	know	VERB
hjic-847	4	35	a	a	DET
hjic-847	4	36	priori	priori	ADV
hjic-847	4	37	.	.	PUNCT
hjic-847	5	1	this	this	DET
hjic-847	5	2	environment	environment	NOUN
hjic-847	5	3	is	be	AUX
hjic-847	5	4	referred	refer	VERB
hjic-847	5	5	to	to	ADP
hjic-847	5	6	as	as	ADP
hjic-847	5	7	the	the	DET
hjic-847	5	8	“	"	PUNCT
hjic-847	5	9	closed	closed	ADJ
hjic-847	5	10	-	-	PUNCT
hjic-847	5	11	world	world	NOUN
hjic-847	5	12	”	"	PUNCT
hjic-847	5	13	model	model	NOUN
hjic-847	5	14	,	,	PUNCT
hjic-847	5	15	while	while	SCONJ
hjic-847	5	16	the	the	DET
hjic-847	5	17	“	"	PUNCT
hjic-847	5	18	open	open	ADJ
hjic-847	5	19	-	-	PUNCT
hjic-847	5	20	world	world	NOUN
hjic-847	5	21	”	"	PUNCT
hjic-847	5	22	model	model	NOUN
hjic-847	5	23	implies	imply	VERB
hjic-847	5	24	that	that	SCONJ
hjic-847	5	25	unknown	unknown	ADJ
hjic-847	5	26	classes	class	NOUN
hjic-847	5	27	can	can	AUX
hjic-847	5	28	be	be	AUX
hjic-847	5	29	incorporated	incorporate	VERB
hjic-847	5	30	into	into	ADP
hjic-847	5	31	a	a	DET
hjic-847	5	32	recognition	recognition	NOUN
hjic-847	5	33	algorithm	algorithm	NOUN
hjic-847	5	34	whilst	whilst	SCONJ
hjic-847	5	35	being	be	AUX
hjic-847	5	36	predicted	predict	VERB
hjic-847	5	37	.	.	PUNCT
hjic-847	6	1	therefore	therefore	ADV
hjic-847	6	2	,	,	PUNCT
hjic-847	6	3	recognition	recognition	NOUN
hjic-847	6	4	systems	system	NOUN
hjic-847	6	5	that	that	PRON
hjic-847	6	6	operate	operate	VERB
hjic-847	6	7	in	in	ADP
hjic-847	6	8	the	the	DET
hjic-847	6	9	real	real	ADJ
hjic-847	6	10	world	world	NOUN
hjic-847	6	11	have	have	VERB
hjic-847	6	12	to	to	PART
hjic-847	6	13	deal	deal	VERB
hjic-847	6	14	with	with	ADP
hjic-847	6	15	these	these	DET
hjic-847	6	16	unknown	unknown	ADJ
hjic-847	6	17	categories	category	NOUN
hjic-847	6	18	.	.	PUNCT
hjic-847	7	1	our	our	PRON
hjic-847	7	2	objective	objective	NOUN
hjic-847	7	3	was	be	AUX
hjic-847	7	4	not	not	PART
hjic-847	7	5	only	only	ADV
hjic-847	7	6	to	to	PART
hjic-847	7	7	detect	detect	VERB
hjic-847	7	8	data	datum	NOUN
hjic-847	7	9	that	that	PRON
hjic-847	7	10	originate	originate	VERB
hjic-847	7	11	from	from	ADP
hjic-847	7	12	categories	category	NOUN
hjic-847	7	13	unseen	unseen	ADJ
hjic-847	7	14	during	during	ADP
hjic-847	7	15	training	training	NOUN
hjic-847	7	16	,	,	PUNCT
hjic-847	7	17	but	but	CCONJ
hjic-847	7	18	to	to	PART
hjic-847	7	19	identify	identify	VERB
hjic-847	7	20	similarities	similarity	NOUN
hjic-847	7	21	between	between	ADP
hjic-847	7	22	pieces	piece	NOUN
hjic-847	7	23	of	of	ADP
hjic-847	7	24	unknown	unknown	ADJ
hjic-847	7	25	data	datum	NOUN
hjic-847	7	26	and	and	CCONJ
hjic-847	7	27	then	then	ADV
hjic-847	7	28	form	form	VERB
hjic-847	7	29	new	new	ADJ
hjic-847	7	30	classes	class	NOUN
hjic-847	7	31	by	by	ADP
hjic-847	7	32	automatically	automatically	ADV
hjic-847	7	33	labeling	label	VERB
hjic-847	7	34	them	they	PRON
hjic-847	7	35	.	.	PUNCT
hjic-847	8	1	our	our	PRON
hjic-847	8	2	double	double	ADJ
hjic-847	8	3	probability	probability	NOUN
hjic-847	8	4	model	model	NOUN
hjic-847	8	5	was	be	AUX
hjic-847	8	6	extended	extend	VERB
hjic-847	8	7	by	by	ADP
hjic-847	8	8	an	an	DET
hjic-847	8	9	image	image	NOUN
hjic-847	8	10	clustering	cluster	VERB
hjic-847	8	11	algorithm	algorithm	NOUN
hjic-847	8	12	,	,	PUNCT
hjic-847	8	13	in	in	ADP
hjic-847	8	14	which	which	PRON
hjic-847	8	15	kernel	kernel	PROPN
hjic-847	8	16	k	k	X
hjic-847	8	17	-	-	PUNCT
hjic-847	8	18	means	mean	NOUN
hjic-847	8	19	was	be	AUX
hjic-847	8	20	used	use	VERB
hjic-847	8	21	.	.	PUNCT
hjic-847	9	1	a	a	DET
hjic-847	9	2	new	new	ADJ
hjic-847	9	3	procedure	procedure	NOUN
hjic-847	9	4	,	,	PUNCT
hjic-847	9	5	namely	namely	ADV
hjic-847	9	6	the	the	DET
hjic-847	9	7	cluster	cluster	NOUN
hjic-847	9	8	classification	classification	NOUN
hjic-847	9	9	algorithm	algorithm	NOUN
hjic-847	9	10	for	for	ADP
hjic-847	9	11	the	the	DET
hjic-847	9	12	detection	detection	NOUN
hjic-847	9	13	of	of	ADP
hjic-847	9	14	unknowns	unknown	NOUN
hjic-847	9	15	and	and	CCONJ
hjic-847	9	16	automated	automate	VERB
hjic-847	9	17	labeling	labeling	NOUN
hjic-847	9	18	,	,	PUNCT
hjic-847	9	19	is	be	AUX
hjic-847	9	20	proposed	propose	VERB
hjic-847	9	21	.	.	PUNCT
hjic-847	10	1	these	these	DET
hjic-847	10	2	approaches	approach	NOUN
hjic-847	10	3	facilitate	facilitate	VERB
hjic-847	10	4	the	the	DET
hjic-847	10	5	transition	transition	NOUN
hjic-847	10	6	from	from	ADP
hjic-847	10	7	openset	openset	NOUN
hjic-847	10	8	recognition	recognition	NOUN
hjic-847	10	9	to	to	ADP
hjic-847	10	10	an	an	DET
hjic-847	10	11	open	open	ADJ
hjic-847	10	12	-	-	PUNCT
hjic-847	10	13	world	world	NOUN
hjic-847	10	14	problem	problem	NOUN
hjic-847	10	15	.	.	PUNCT
hjic-847	11	1	the	the	DET
hjic-847	11	2	fisher	fisher	PROPN
hjic-847	11	3	vector	vector	PROPN
hjic-847	11	4	(	(	PUNCT
hjic-847	11	5	fv	fv	X
hjic-847	11	6	)	)	PUNCT
hjic-847	11	7	was	be	AUX
hjic-847	11	8	used	use	VERB
hjic-847	11	9	for	for	ADP
hjic-847	11	10	the	the	DET
hjic-847	11	11	mathematical	mathematical	ADJ
hjic-847	11	12	representation	representation	NOUN
hjic-847	11	13	of	of	ADP
hjic-847	11	14	the	the	DET
hjic-847	11	15	images	image	NOUN
hjic-847	11	16	and	and	CCONJ
hjic-847	11	17	then	then	ADV
hjic-847	11	18	a	a	DET
hjic-847	11	19	support	support	NOUN
hjic-847	11	20	vector	vector	NOUN
hjic-847	11	21	machine	machine	NOUN
hjic-847	11	22	introduced	introduce	VERB
hjic-847	11	23	as	as	ADP
hjic-847	11	24	a	a	DET
hjic-847	11	25	classifier	classifier	NOUN
hjic-847	11	26	.	.	PUNCT
hjic-847	12	1	the	the	DET
hjic-847	12	2	measurement	measurement	NOUN
hjic-847	12	3	of	of	ADP
hjic-847	12	4	similarity	similarity	NOUN
hjic-847	12	5	was	be	AUX
hjic-847	12	6	based	base	VERB
hjic-847	12	7	on	on	ADP
hjic-847	12	8	the	the	DET
hjic-847	12	9	fv	fv	PROPN
hjic-847	12	10	representations	representation	NOUN
hjic-847	12	11	.	.	PUNCT
hjic-847	13	1	experiments	experiment	NOUN
hjic-847	13	2	were	be	AUX
hjic-847	13	3	conducted	conduct	VERB
hjic-847	13	4	on	on	ADP
hjic-847	13	5	the	the	DET
hjic-847	13	6	caltech101	caltech101	PROPN
hjic-847	13	7	and	and	CCONJ
hjic-847	13	8	caltech256	caltech256	PROPN
hjic-847	13	9	datasets	dataset	NOUN
hjic-847	13	10	of	of	ADP
hjic-847	13	11	images	image	NOUN
hjic-847	13	12	and	and	CCONJ
hjic-847	13	13	the	the	DET
hjic-847	13	14	rand	rand	NOUN
hjic-847	13	15	index	index	NOUN
hjic-847	13	16	was	be	AUX
hjic-847	13	17	evaluated	evaluate	VERB
hjic-847	13	18	over	over	ADP
hjic-847	13	19	the	the	DET
hjic-847	13	20	unknown	unknown	ADJ
hjic-847	13	21	data	datum	NOUN
hjic-847	13	22	.	.	PUNCT
hjic-847	14	1	the	the	DET
hjic-847	14	2	results	result	NOUN
hjic-847	14	3	showed	show	VERB
hjic-847	14	4	that	that	SCONJ
hjic-847	14	5	our	our	PRON
hjic-847	14	6	proposed	propose	VERB
hjic-847	14	7	cluster	cluster	NOUN
hjic-847	14	8	classification	classification	NOUN
hjic-847	14	9	algorithm	algorithm	NOUN
hjic-847	14	10	was	be	AUX
hjic-847	14	11	able	able	ADJ
hjic-847	14	12	to	to	PART
hjic-847	14	13	yield	yield	VERB
hjic-847	14	14	almost	almost	ADV
hjic-847	14	15	the	the	DET
hjic-847	14	16	same	same	ADJ
hjic-847	14	17	rand	rand	NOUN
hjic-847	14	18	index	index	NOUN
hjic-847	14	19	,	,	PUNCT
hjic-847	14	20	even	even	ADV
hjic-847	14	21	though	though	SCONJ
hjic-847	14	22	the	the	DET
hjic-847	14	23	number	number	NOUN
hjic-847	14	24	of	of	ADP
hjic-847	14	25	unknown	unknown	ADJ
hjic-847	14	26	categories	category	NOUN
hjic-847	14	27	increased	increase	VERB
hjic-847	14	28	.	.	PUNCT
hjic-847	15	1	keywords	keyword	NOUN
hjic-847	15	2	:	:	PUNCT
hjic-847	15	3	open	open	ADJ
hjic-847	15	4	-	-	PUNCT
hjic-847	15	5	world	world	NOUN
hjic-847	15	6	problem	problem	NOUN
hjic-847	15	7	,	,	PUNCT
hjic-847	15	8	cluster	cluster	NOUN
hjic-847	15	9	classification	classification	NOUN
hjic-847	15	10	,	,	PUNCT
hjic-847	15	11	image	image	NOUN
hjic-847	15	12	classification	classification	NOUN
hjic-847	15	13	,	,	PUNCT
hjic-847	15	14	open	open	ADJ
hjic-847	15	15	-	-	PUNCT
hjic-847	15	16	set	set	VERB
hjic-847	15	17	recognition	recognition	NOUN
hjic-847	15	18	,	,	PUNCT
hjic-847	15	19	image	image	NOUN
hjic-847	15	20	clustering	cluster	VERB
hjic-847	15	21	1	1	NUM
hjic-847	15	22	.	.	PUNCT
hjic-847	15	23	introduction	introduction	NOUN
hjic-847	15	24	in	in	ADP
hjic-847	15	25	scenarios	scenario	NOUN
hjic-847	15	26	in	in	ADP
hjic-847	15	27	the	the	DET
hjic-847	15	28	real	real	ADJ
hjic-847	15	29	world	world	NOUN
hjic-847	15	30	,	,	PUNCT
hjic-847	15	31	the	the	DET
hjic-847	15	32	size	size	NOUN
hjic-847	15	33	of	of	ADP
hjic-847	15	34	the	the	DET
hjic-847	15	35	available	available	ADJ
hjic-847	15	36	dataset	dataset	NOUN
hjic-847	15	37	continues	continue	VERB
hjic-847	15	38	to	to	PART
hjic-847	15	39	increase	increase	VERB
hjic-847	15	40	,	,	PUNCT
hjic-847	15	41	therefore	therefore	ADV
hjic-847	15	42	,	,	PUNCT
hjic-847	15	43	any	any	DET
hjic-847	15	44	machine	machine	NOUN
hjic-847	15	45	learning	learn	VERB
hjic-847	15	46	algorithm	algorithm	NOUN
hjic-847	15	47	that	that	PRON
hjic-847	15	48	operates	operate	VERB
hjic-847	15	49	in	in	ADP
hjic-847	15	50	such	such	DET
hjic-847	15	51	an	an	DET
hjic-847	15	52	environment	environment	NOUN
hjic-847	15	53	has	have	VERB
hjic-847	15	54	to	to	PART
hjic-847	15	55	be	be	AUX
hjic-847	15	56	capable	capable	ADJ
hjic-847	15	57	of	of	ADP
hjic-847	15	58	preventing	prevent	VERB
hjic-847	15	59	growth	growth	NOUN
hjic-847	15	60	.	.	PUNCT
hjic-847	16	1	this	this	PRON
hjic-847	16	2	is	be	AUX
hjic-847	16	3	especially	especially	ADV
hjic-847	16	4	true	true	ADJ
hjic-847	16	5	in	in	ADP
hjic-847	16	6	the	the	DET
hjic-847	16	7	case	case	NOUN
hjic-847	16	8	of	of	ADP
hjic-847	16	9	image	image	NOUN
hjic-847	16	10	classification	classification	NOUN
hjic-847	16	11	,	,	PUNCT
hjic-847	16	12	because	because	SCONJ
hjic-847	16	13	the	the	DET
hjic-847	16	14	growing	grow	VERB
hjic-847	16	15	dataset	dataset	NOUN
hjic-847	16	16	of	of	ADP
hjic-847	16	17	tests	test	NOUN
hjic-847	16	18	can	can	AUX
hjic-847	16	19	pose	pose	VERB
hjic-847	16	20	many	many	ADJ
hjic-847	16	21	difficulties	difficulty	NOUN
hjic-847	16	22	,	,	PUNCT
hjic-847	16	23	e.g.	e.g.	ADV
hjic-847	16	24	it	it	PRON
hjic-847	16	25	is	be	AUX
hjic-847	16	26	possible	possible	ADJ
hjic-847	16	27	that	that	SCONJ
hjic-847	16	28	some	some	PRON
hjic-847	16	29	of	of	ADP
hjic-847	16	30	the	the	DET
hjic-847	16	31	test	test	NOUN
hjic-847	16	32	images	image	NOUN
hjic-847	16	33	originate	originate	VERB
hjic-847	16	34	from	from	ADP
hjic-847	16	35	categories	category	NOUN
hjic-847	16	36	that	that	PRON
hjic-847	16	37	are	be	AUX
hjic-847	16	38	unseen	unseen	ADJ
hjic-847	16	39	during	during	ADP
hjic-847	16	40	training	training	NOUN
hjic-847	16	41	.	.	PUNCT
hjic-847	17	1	recognition	recognition	NOUN
hjic-847	17	2	systems	system	NOUN
hjic-847	17	3	should	should	AUX
hjic-847	17	4	detect	detect	VERB
hjic-847	17	5	these	these	DET
hjic-847	17	6	unknown	unknown	ADJ
hjic-847	17	7	images	image	NOUN
hjic-847	17	8	and	and	CCONJ
hjic-847	17	9	handle	handle	VERB
hjic-847	17	10	them	they	PRON
hjic-847	17	11	in	in	ADP
hjic-847	17	12	an	an	DET
hjic-847	17	13	appropriate	appropriate	ADJ
hjic-847	17	14	way	way	NOUN
hjic-847	17	15	.	.	PUNCT
hjic-847	18	1	in	in	ADP
hjic-847	18	2	the	the	DET
hjic-847	18	3	rest	rest	NOUN
hjic-847	18	4	of	of	ADP
hjic-847	18	5	the	the	DET
hjic-847	18	6	paper	paper	NOUN
hjic-847	18	7	the	the	DET
hjic-847	18	8	terms	term	NOUN
hjic-847	18	9	"	"	PUNCT
hjic-847	18	10	unknown	unknown	ADJ
hjic-847	18	11	class	class	NOUN
hjic-847	18	12	or	or	CCONJ
hjic-847	18	13	category	category	NOUN
hjic-847	18	14	"	"	PUNCT
hjic-847	18	15	represent	represent	VERB
hjic-847	18	16	classes	class	NOUN
hjic-847	18	17	or	or	CCONJ
hjic-847	18	18	categories	category	NOUN
hjic-847	18	19	that	that	PRON
hjic-847	18	20	are	be	AUX
hjic-847	18	21	unseen	unseen	ADJ
hjic-847	18	22	during	during	ADP
hjic-847	18	23	training	training	NOUN
hjic-847	18	24	,	,	PUNCT
hjic-847	18	25	and	and	CCONJ
hjic-847	18	26	“	"	PUNCT
hjic-847	18	27	unknown	unknown	ADJ
hjic-847	18	28	image	image	NOUN
hjic-847	18	29	”	"	PUNCT
hjic-847	18	30	denotes	denote	NOUN
hjic-847	18	31	images	image	NOUN
hjic-847	18	32	that	that	PRON
hjic-847	18	33	originate	originate	VERB
hjic-847	18	34	from	from	ADP
hjic-847	18	35	unknown	unknown	ADJ
hjic-847	18	36	classes	class	NOUN
hjic-847	18	37	or	or	CCONJ
hjic-847	18	38	categories	category	NOUN
hjic-847	18	39	.	.	PUNCT
hjic-847	19	1	one	one	NUM
hjic-847	19	2	way	way	NOUN
hjic-847	19	3	of	of	ADP
hjic-847	19	4	handling	handle	VERB
hjic-847	19	5	detected	detect	VERB
hjic-847	19	6	unknown	unknown	ADJ
hjic-847	19	7	images	image	NOUN
hjic-847	19	8	is	be	AUX
hjic-847	19	9	to	to	PART
hjic-847	19	10	measure	measure	VERB
hjic-847	19	11	their	their	PRON
hjic-847	19	12	similarities	similarity	NOUN
hjic-847	19	13	and	and	CCONJ
hjic-847	19	14	identify	identify	VERB
hjic-847	19	15	new	new	ADJ
hjic-847	19	16	categories	category	NOUN
hjic-847	19	17	.	.	PUNCT
hjic-847	20	1	subsequently	subsequently	ADV
hjic-847	20	2	,	,	PUNCT
hjic-847	20	3	these	these	DET
hjic-847	20	4	new	new	ADJ
hjic-847	20	5	categories	category	NOUN
hjic-847	20	6	can	can	AUX
hjic-847	20	7	be	be	AUX
hjic-847	20	8	added	add	VERB
hjic-847	20	9	to	to	ADP
hjic-847	20	10	the	the	DET
hjic-847	20	11	set	set	NOUN
hjic-847	20	12	of	of	ADP
hjic-847	20	13	known	know	VERB
hjic-847	20	14	classes	class	NOUN
hjic-847	20	15	.	.	PUNCT
hjic-847	21	1	based	base	VERB
hjic-847	21	2	on	on	ADP
hjic-847	21	3	this	this	PRON
hjic-847	21	4	,	,	PUNCT
hjic-847	21	5	three	three	NUM
hjic-847	21	6	modules	module	NOUN
hjic-847	21	7	are	be	AUX
hjic-847	21	8	required	require	VERB
hjic-847	21	9	to	to	PART
hjic-847	21	10	solve	solve	VERB
hjic-847	21	11	such	such	ADJ
hjic-847	21	12	problems	problem	NOUN
hjic-847	21	13	in	in	ADP
hjic-847	21	14	the	the	DET
hjic-847	21	15	real	real	ADJ
hjic-847	21	16	world	world	NOUN
hjic-847	21	17	,	,	PUNCT
hjic-847	21	18	namely	namely	ADV
hjic-847	21	19	a	a	DET
hjic-847	21	20	recognition	recognition	NOUN
hjic-847	21	21	system	system	NOUN
hjic-847	21	22	equipped	equip	VERB
hjic-847	21	23	with	with	ADP
hjic-847	21	24	an	an	DET
hjic-847	21	25	unknown	unknown	ADJ
hjic-847	21	26	detector	detector	NOUN
hjic-847	21	27	,	,	PUNCT
hjic-847	21	28	a	a	DET
hjic-847	21	29	labeling	labeling	NOUN
hjic-847	21	30	process	process	NOUN
hjic-847	21	31	and	and	CCONJ
hjic-847	21	32	an	an	DET
hjic-847	21	33	incremental	incremental	ADJ
hjic-847	21	34	learning	learning	NOUN
hjic-847	21	35	process	process	NOUN
hjic-847	21	36	.	.	PUNCT
hjic-847	22	1	*	*	PUNCT
hjic-847	22	2	correspondence	correspondence	NOUN
hjic-847	22	3	:	:	PUNCT
hjic-847	22	4	pappd@tmit.bme.hu	pappd@tmit.bme.hu	NOUN
hjic-847	22	5	let	let	VERB
hjic-847	22	6	us	we	PRON
hjic-847	22	7	assume	assume	VERB
hjic-847	22	8	that	that	SCONJ
hjic-847	22	9	there	there	PRON
hjic-847	22	10	are	be	VERB
hjic-847	22	11	k	k	X
hjic-847	22	12	known	know	VERB
hjic-847	22	13	classes	class	NOUN
hjic-847	22	14	(	(	PUNCT
hjic-847	22	15	c1	c1	PROPN
hjic-847	22	16	,	,	PUNCT
hjic-847	22	17	c2	c2	PROPN
hjic-847	22	18	,	,	PUNCT
hjic-847	22	19	.	.	PUNCT
hjic-847	22	20	.	.	PUNCT
hjic-847	22	21	.	.	PUNCT
hjic-847	23	1	ck	ck	X
hjic-847	23	2	)	)	PUNCT
hjic-847	24	1	and	and	CCONJ
hjic-847	24	2	u	u	NOUN
hjic-847	24	3	unknown	unknown	ADJ
hjic-847	24	4	classes	class	NOUN
hjic-847	24	5	in	in	ADP
hjic-847	24	6	the	the	DET
hjic-847	24	7	test	test	NOUN
hjic-847	24	8	set	set	VERB
hjic-847	24	9	at	at	ADP
hjic-847	24	10	any	any	DET
hjic-847	24	11	given	give	VERB
hjic-847	24	12	moment	moment	NOUN
hjic-847	24	13	,	,	PUNCT
hjic-847	24	14	where	where	SCONJ
hjic-847	24	15	sk	sk	NOUN
hjic-847	24	16	and	and	CCONJ
hjic-847	24	17	su	su	PROPN
hjic-847	24	18	denote	denote	VERB
hjic-847	24	19	the	the	DET
hjic-847	24	20	sets	set	NOUN
hjic-847	24	21	of	of	ADP
hjic-847	24	22	known	known	ADJ
hjic-847	24	23	and	and	CCONJ
hjic-847	24	24	unknown	unknown	ADJ
hjic-847	24	25	classes	class	NOUN
hjic-847	24	26	,	,	PUNCT
hjic-847	24	27	respectively	respectively	ADV
hjic-847	24	28	.	.	PUNCT
hjic-847	25	1	few	few	ADJ
hjic-847	25	2	distinguishable	distinguishable	ADJ
hjic-847	25	3	cases	case	NOUN
hjic-847	25	4	depend	depend	VERB
hjic-847	25	5	on	on	ADP
hjic-847	25	6	the	the	DET
hjic-847	25	7	value	value	NOUN
hjic-847	25	8	of	of	ADP
hjic-847	25	9	u	u	NOUN
hjic-847	25	10	:	:	PUNCT
hjic-847	25	11	1	1	X
hjic-847	25	12	.	.	X
hjic-847	25	13	u	u	NOUN
hjic-847	25	14	=	=	NOUN
hjic-847	25	15	0	0	NUM
hjic-847	25	16	,	,	PUNCT
hjic-847	25	17	2	2	NUM
hjic-847	25	18	.	.	X
hjic-847	25	19	u	u	NOUN
hjic-847	25	20	=	=	NOUN
hjic-847	25	21	1	1	NUM
hjic-847	25	22	,	,	PUNCT
hjic-847	25	23	3	3	NUM
hjic-847	25	24	.	.	X
hjic-847	25	25	u	u	NOUN
hjic-847	25	26	>	>	X
hjic-847	25	27	1	1	X
hjic-847	25	28	.	.	PUNCT
hjic-847	26	1	furthermore	furthermore	ADV
hjic-847	26	2	,	,	PUNCT
hjic-847	26	3	a	a	DET
hjic-847	26	4	few	few	ADJ
hjic-847	26	5	more	more	ADJ
hjic-847	26	6	cases	case	NOUN
hjic-847	26	7	depend	depend	VERB
hjic-847	26	8	on	on	ADP
hjic-847	26	9	the	the	DET
hjic-847	26	10	amount	amount	NOUN
hjic-847	26	11	and	and	CCONJ
hjic-847	26	12	type	type	NOUN
hjic-847	26	13	of	of	ADP
hjic-847	26	14	available	available	ADJ
hjic-847	26	15	information	information	NOUN
hjic-847	26	16	concerning	concern	VERB
hjic-847	26	17	su	su	PROPN
hjic-847	26	18	:	:	PUNCT
hjic-847	26	19	(	(	PUNCT
hjic-847	26	20	a	a	X
hjic-847	26	21	)	)	PUNCT
hjic-847	26	22	training	training	NOUN
hjic-847	26	23	images	image	NOUN
hjic-847	26	24	,	,	PUNCT
hjic-847	26	25	(	(	PUNCT
hjic-847	26	26	b	b	X
hjic-847	26	27	)	)	PUNCT
hjic-847	26	28	set	set	NOUN
hjic-847	26	29	of	of	ADP
hjic-847	26	30	attributes	attribute	NOUN
hjic-847	26	31	,	,	PUNCT
hjic-847	26	32	(	(	PUNCT
hjic-847	26	33	c	c	NOUN
hjic-847	26	34	)	)	PUNCT
hjic-847	26	35	number	number	NOUN
hjic-847	26	36	of	of	ADP
hjic-847	26	37	unknown	unknown	ADJ
hjic-847	26	38	categories	category	NOUN
hjic-847	26	39	(	(	PUNCT
hjic-847	26	40	u	u	NOUN
hjic-847	26	41	)	)	PUNCT
hjic-847	26	42	,	,	PUNCT
hjic-847	26	43	(	(	PUNCT
hjic-847	26	44	d	d	X
hjic-847	26	45	)	)	PUNCT
hjic-847	26	46	nothing	nothing	PRON
hjic-847	26	47	.	.	PUNCT
hjic-847	27	1	the	the	DET
hjic-847	27	2	cases	case	NOUN
hjic-847	27	3	that	that	PRON
hjic-847	27	4	include	include	VERB
hjic-847	27	5	1	1	NUM
hjic-847	27	6	or	or	CCONJ
hjic-847	27	7	a	a	DET
hjic-847	27	8	(	(	PUNCT
hjic-847	27	9	e.g.	e.g.	ADV
hjic-847	27	10	1a	1a	NOUN
hjic-847	27	11	,	,	PUNCT
hjic-847	27	12	2a	2a	NUM
hjic-847	27	13	,	,	PUNCT
hjic-847	27	14	1b	1b	NUM
hjic-847	27	15	,	,	PUNCT
hjic-847	27	16	1c	1c	X
hjic-847	27	17	)	)	PUNCT
hjic-847	27	18	produce	produce	VERB
hjic-847	27	19	the	the	DET
hjic-847	27	20	general	general	ADJ
hjic-847	27	21	multiclass	multiclass	ADJ
hjic-847	27	22	classification	classification	NOUN
hjic-847	27	23	because	because	SCONJ
hjic-847	27	24	all	all	DET
hjic-847	27	25	categories	category	NOUN
hjic-847	27	26	are	be	AUX
hjic-847	27	27	known	know	VERB
hjic-847	27	28	a	a	DET
hjic-847	27	29	priori	priori	ADJ
hjic-847	27	30	and	and	CCONJ
hjic-847	27	31	positive	positive	ADJ
hjic-847	27	32	-	-	PUNCT
hjic-847	27	33	negative	negative	ADJ
hjic-847	27	34	samples	sample	NOUN
hjic-847	27	35	are	be	AUX
hjic-847	27	36	available	available	ADJ
hjic-847	27	37	for	for	ADP
hjic-847	27	38	each	each	DET
hjic-847	27	39	category	category	NOUN
hjic-847	27	40	during	during	ADP
hjic-847	27	41	training	training	NOUN
hjic-847	27	42	.	.	PUNCT
hjic-847	28	1	when	when	SCONJ
hjic-847	28	2	mailto:pappd@tmit.bme.hu	mailto:pappd@tmit.bme.hu	PROPN
hjic-847	28	3	34	34	NUM
hjic-847	28	4	papp	papp	NOUN
hjic-847	28	5	and	and	CCONJ
hjic-847	28	6	szűcs	szűcs	ADJ
hjic-847	28	7	u	u	NOUN
hjic-847	28	8	=	=	PROPN
hjic-847	28	9	1	1	NUM
hjic-847	28	10	,	,	PUNCT
hjic-847	28	11	the	the	DET
hjic-847	28	12	task	task	NOUN
hjic-847	28	13	is	be	AUX
hjic-847	28	14	only	only	ADV
hjic-847	28	15	to	to	PART
hjic-847	28	16	identify	identify	VERB
hjic-847	28	17	the	the	DET
hjic-847	28	18	unknown	unknown	ADJ
hjic-847	28	19	images	image	NOUN
hjic-847	28	20	because	because	SCONJ
hjic-847	28	21	they	they	PRON
hjic-847	28	22	originate	originate	VERB
hjic-847	28	23	from	from	ADP
hjic-847	28	24	the	the	DET
hjic-847	28	25	same	same	ADJ
hjic-847	28	26	category	category	NOUN
hjic-847	28	27	,	,	PUNCT
hjic-847	28	28	therefore	therefore	ADV
hjic-847	28	29	,	,	PUNCT
hjic-847	28	30	the	the	DET
hjic-847	28	31	similarity	similarity	NOUN
hjic-847	28	32	measurement	measurement	NOUN
hjic-847	28	33	is	be	AUX
hjic-847	28	34	unnecessary	unnecessary	ADJ
hjic-847	28	35	.	.	PUNCT
hjic-847	29	1	in	in	ADP
hjic-847	29	2	this	this	DET
hjic-847	29	3	paper	paper	NOUN
hjic-847	29	4	,	,	PUNCT
hjic-847	29	5	the	the	DET
hjic-847	29	6	situation	situation	NOUN
hjic-847	29	7	when	when	SCONJ
hjic-847	29	8	u	u	PROPN
hjic-847	29	9	>	>	X
hjic-847	29	10	1	1	NUM
hjic-847	29	11	is	be	AUX
hjic-847	29	12	considered	consider	VERB
hjic-847	29	13	.	.	PUNCT
hjic-847	30	1	as	as	SCONJ
hjic-847	30	2	has	have	AUX
hjic-847	30	3	been	be	AUX
hjic-847	30	4	mentioned	mention	VERB
hjic-847	30	5	,	,	PUNCT
hjic-847	30	6	3a	3a	PROPN
hjic-847	30	7	represents	represent	VERB
hjic-847	30	8	the	the	DET
hjic-847	30	9	traditional	traditional	ADJ
hjic-847	30	10	multiclass	multiclass	ADJ
hjic-847	30	11	classification	classification	NOUN
hjic-847	30	12	.	.	PUNCT
hjic-847	31	1	3b+3c	3b+3c	NUM
hjic-847	31	2	is	be	AUX
hjic-847	31	3	referred	refer	VERB
hjic-847	31	4	to	to	ADP
hjic-847	31	5	as	as	ADP
hjic-847	31	6	transfer	transfer	NOUN
hjic-847	31	7	learning	learning	NOUN
hjic-847	31	8	or	or	CCONJ
hjic-847	31	9	zero	zero	NUM
hjic-847	31	10	-	-	PUNCT
hjic-847	31	11	shot	shot	NOUN
hjic-847	31	12	learning	learning	NOUN
hjic-847	31	13	[	[	X
hjic-847	31	14	1	1	NUM
hjic-847	31	15	]	]	PUNCT
hjic-847	31	16	,	,	PUNCT
hjic-847	31	17	whereas	whereas	SCONJ
hjic-847	31	18	according	accord	VERB
hjic-847	31	19	to	to	ADP
hjic-847	31	20	the	the	DET
hjic-847	31	21	literature	literature	NOUN
hjic-847	31	22	the	the	DET
hjic-847	31	23	case	case	NOUN
hjic-847	31	24	of	of	ADP
hjic-847	31	25	3c+3d	3c+3d	NUM
hjic-847	31	26	is	be	AUX
hjic-847	31	27	known	know	VERB
hjic-847	31	28	as	as	ADP
hjic-847	31	29	open	open	ADJ
hjic-847	31	30	set	set	VERB
hjic-847	31	31	recognition	recognition	NOUN
hjic-847	32	1	[	[	X
hjic-847	32	2	2,3	2,3	NUM
hjic-847	32	3	]	]	PUNCT
hjic-847	32	4	or	or	CCONJ
hjic-847	32	5	the	the	DET
hjic-847	32	6	open	open	ADJ
hjic-847	32	7	-	-	PUNCT
hjic-847	32	8	world	world	NOUN
hjic-847	32	9	problem	problem	NOUN
hjic-847	32	10	[	[	X
hjic-847	32	11	4	4	NUM
hjic-847	32	12	]	]	PUNCT
hjic-847	32	13	.	.	PUNCT
hjic-847	33	1	the	the	DET
hjic-847	33	2	former	former	ADJ
hjic-847	33	3	refers	refer	VERB
hjic-847	33	4	to	to	ADP
hjic-847	33	5	the	the	DET
hjic-847	33	6	detection	detection	NOUN
hjic-847	33	7	of	of	ADP
hjic-847	33	8	images	image	NOUN
hjic-847	33	9	that	that	PRON
hjic-847	33	10	originate	originate	VERB
hjic-847	33	11	from	from	ADP
hjic-847	33	12	unknown	unknown	ADJ
hjic-847	33	13	classes	class	NOUN
hjic-847	33	14	,	,	PUNCT
hjic-847	33	15	and	and	CCONJ
hjic-847	33	16	the	the	DET
hjic-847	33	17	latter	latter	ADJ
hjic-847	33	18	includes	include	VERB
hjic-847	33	19	the	the	DET
hjic-847	33	20	detection	detection	NOUN
hjic-847	33	21	of	of	ADP
hjic-847	33	22	unknown	unknown	ADJ
hjic-847	33	23	images	image	NOUN
hjic-847	33	24	and	and	CCONJ
hjic-847	33	25	a	a	DET
hjic-847	33	26	labeling	labeling	NOUN
hjic-847	33	27	process	process	NOUN
hjic-847	33	28	to	to	PART
hjic-847	33	29	identify	identify	VERB
hjic-847	33	30	new	new	ADJ
hjic-847	33	31	classes	class	NOUN
hjic-847	33	32	,	,	PUNCT
hjic-847	33	33	followed	follow	VERB
hjic-847	33	34	by	by	ADP
hjic-847	33	35	the	the	DET
hjic-847	33	36	incremental	incremental	ADJ
hjic-847	33	37	learning	learning	NOUN
hjic-847	33	38	of	of	ADP
hjic-847	33	39	these	these	DET
hjic-847	33	40	new	new	ADJ
hjic-847	33	41	categories	category	NOUN
hjic-847	33	42	.	.	PUNCT
hjic-847	34	1	our	our	PRON
hjic-847	34	2	goal	goal	NOUN
hjic-847	34	3	was	be	AUX
hjic-847	34	4	to	to	PART
hjic-847	34	5	tackle	tackle	VERB
hjic-847	34	6	the	the	DET
hjic-847	34	7	open	open	ADJ
hjic-847	34	8	-	-	PUNCT
hjic-847	34	9	world	world	NOUN
hjic-847	34	10	problem	problem	NOUN
hjic-847	34	11	as	as	ADV
hjic-847	34	12	well	well	ADV
hjic-847	34	13	as	as	ADP
hjic-847	34	14	develop	develop	VERB
hjic-847	34	15	an	an	DET
hjic-847	34	16	algorithm	algorithm	NOUN
hjic-847	34	17	that	that	PRON
hjic-847	34	18	is	be	AUX
hjic-847	34	19	able	able	ADJ
hjic-847	34	20	to	to	PART
hjic-847	34	21	detect	detect	VERB
hjic-847	34	22	the	the	DET
hjic-847	34	23	unknown	unknown	ADJ
hjic-847	34	24	images	image	NOUN
hjic-847	34	25	and	and	CCONJ
hjic-847	34	26	then	then	ADV
hjic-847	34	27	introduce	introduce	VERB
hjic-847	34	28	new	new	ADJ
hjic-847	34	29	classes	class	NOUN
hjic-847	34	30	by	by	ADP
hjic-847	34	31	automatically	automatically	ADV
hjic-847	34	32	labeling	label	VERB
hjic-847	34	33	the	the	DET
hjic-847	34	34	unknown	unknown	ADJ
hjic-847	34	35	data	datum	NOUN
hjic-847	34	36	using	use	VERB
hjic-847	34	37	unsupervised	unsupervised	ADJ
hjic-847	34	38	learning	learning	NOUN
hjic-847	34	39	.	.	PUNCT
hjic-847	35	1	previously	previously	ADV
hjic-847	35	2	an	an	DET
hjic-847	35	3	algorithm	algorithm	NOUN
hjic-847	35	4	referred	refer	VERB
hjic-847	35	5	to	to	ADP
hjic-847	35	6	as	as	ADP
hjic-847	35	7	the	the	DET
hjic-847	35	8	double	double	ADJ
hjic-847	35	9	probability	probability	NOUN
hjic-847	35	10	model	model	NOUN
hjic-847	35	11	(	(	PUNCT
hjic-847	35	12	dpm	dpm	PROPN
hjic-847	35	13	)	)	PUNCT
hjic-847	36	1	[	[	X
hjic-847	36	2	5	5	NUM
hjic-847	36	3	]	]	PUNCT
hjic-847	36	4	was	be	AUX
hjic-847	36	5	proposed	propose	VERB
hjic-847	36	6	,	,	PUNCT
hjic-847	36	7	which	which	PRON
hjic-847	36	8	is	be	AUX
hjic-847	36	9	suitable	suitable	ADJ
hjic-847	36	10	as	as	ADP
hjic-847	36	11	an	an	DET
hjic-847	36	12	unknown	unknown	ADJ
hjic-847	36	13	detector	detector	NOUN
hjic-847	36	14	in	in	ADP
hjic-847	36	15	an	an	DET
hjic-847	36	16	open	open	ADJ
hjic-847	36	17	-	-	PUNCT
hjic-847	36	18	set	set	VERB
hjic-847	36	19	environment	environment	NOUN
hjic-847	36	20	.	.	PUNCT
hjic-847	37	1	there	there	PRON
hjic-847	37	2	are	be	VERB
hjic-847	37	3	several	several	ADJ
hjic-847	37	4	works	work	NOUN
hjic-847	37	5	that	that	PRON
hjic-847	37	6	use	use	VERB
hjic-847	37	7	a	a	DET
hjic-847	37	8	variant	variant	NOUN
hjic-847	37	9	of	of	ADP
hjic-847	37	10	support	support	NOUN
hjic-847	37	11	vector	vector	NOUN
hjic-847	37	12	machine	machine	NOUN
hjic-847	37	13	(	(	PUNCT
hjic-847	37	14	svm	svm	PROPN
hjic-847	37	15	)	)	PUNCT
hjic-847	37	16	to	to	PART
hjic-847	37	17	solve	solve	VERB
hjic-847	37	18	the	the	DET
hjic-847	37	19	unknown	unknown	ADJ
hjic-847	37	20	detection	detection	NOUN
hjic-847	37	21	problem	problem	NOUN
hjic-847	37	22	,	,	PUNCT
hjic-847	37	23	such	such	ADJ
hjic-847	37	24	as	as	ADP
hjic-847	37	25	the	the	DET
hjic-847	37	26	support	support	NOUN
hjic-847	37	27	vector	vector	NOUN
hjic-847	37	28	data	datum	NOUN
hjic-847	37	29	description	description	NOUN
hjic-847	37	30	[	[	X
hjic-847	37	31	6	6	NUM
hjic-847	37	32	]	]	PUNCT
hjic-847	37	33	,	,	PUNCT
hjic-847	37	34	one	one	NUM
hjic-847	37	35	-	-	PUNCT
hjic-847	37	36	class	class	NOUN
hjic-847	37	37	svm	svm	NOUN
hjic-847	37	38	[	[	X
hjic-847	37	39	7	7	NUM
hjic-847	37	40	,	,	PUNCT
hjic-847	37	41	8	8	NUM
hjic-847	37	42	]	]	PUNCT
hjic-847	37	43	,	,	PUNCT
hjic-847	37	44	reject	reject	VERB
hjic-847	37	45	option	option	NOUN
hjic-847	37	46	svm	svm	NOUN
hjic-847	37	47	(	(	PUNCT
hjic-847	37	48	ro	ro	ADJ
hjic-847	37	49	-	-	ADJ
hjic-847	37	50	svm	svm	ADJ
hjic-847	37	51	)	)	PUNCT
hjic-847	38	1	[	[	X
hjic-847	38	2	9	9	NUM
hjic-847	38	3	]	]	PUNCT
hjic-847	38	4	and	and	CCONJ
hjic-847	38	5	the	the	DET
hjic-847	38	6	novel	novel	ADJ
hjic-847	38	7	weibull	weibull	NOUN
hjic-847	38	8	-	-	PUNCT
hjic-847	38	9	calibrated	calibrate	VERB
hjic-847	38	10	svm	svm	NOUN
hjic-847	38	11	(	(	PUNCT
hjic-847	38	12	w	w	NOUN
hjic-847	38	13	-	-	PUNCT
hjic-847	38	14	svm	svm	ADJ
hjic-847	38	15	)	)	PUNCT
hjic-847	39	1	[	[	X
hjic-847	39	2	3	3	NUM
hjic-847	39	3	]	]	PUNCT
hjic-847	39	4	.	.	PUNCT
hjic-847	40	1	the	the	DET
hjic-847	40	2	latter	latter	ADJ
hjic-847	40	3	one	one	NOUN
hjic-847	40	4	was	be	AUX
hjic-847	40	5	developed	develop	VERB
hjic-847	40	6	to	to	PART
hjic-847	40	7	operate	operate	VERB
hjic-847	40	8	under	under	ADP
hjic-847	40	9	the	the	DET
hjic-847	40	10	compact	compact	ADJ
hjic-847	40	11	abating	abating	NOUN
hjic-847	40	12	probability	probability	NOUN
hjic-847	40	13	model	model	NOUN
hjic-847	40	14	,	,	PUNCT
hjic-847	40	15	where	where	SCONJ
hjic-847	40	16	the	the	DET
hjic-847	40	17	probability	probability	NOUN
hjic-847	40	18	of	of	ADP
hjic-847	40	19	class	class	NOUN
hjic-847	40	20	membership	membership	NOUN
hjic-847	40	21	decreases	decrease	NOUN
hjic-847	40	22	(	(	PUNCT
hjic-847	40	23	abates	abate	NOUN
hjic-847	40	24	)	)	PUNCT
hjic-847	40	25	as	as	ADP
hjic-847	40	26	points	point	NOUN
hjic-847	40	27	move	move	VERB
hjic-847	40	28	from	from	ADP
hjic-847	40	29	known	know	VERB
hjic-847	40	30	data	datum	NOUN
hjic-847	40	31	towards	towards	ADP
hjic-847	40	32	unknown	unknown	ADJ
hjic-847	40	33	space	space	NOUN
hjic-847	40	34	.	.	PUNCT
hjic-847	41	1	scheirer	scheirer	AUX
hjic-847	41	2	et	et	PROPN
hjic-847	41	3	al	al	PROPN
hjic-847	41	4	.	.	PROPN
hjic-847	42	1	claim	claim	VERB
hjic-847	42	2	that	that	SCONJ
hjic-847	42	3	w	w	ADJ
hjic-847	42	4	-	-	PUNCT
hjic-847	42	5	svm	svm	NOUN
hjic-847	42	6	outperforms	outperform	VERB
hjic-847	42	7	their	their	PRON
hjic-847	42	8	previous	previous	ADJ
hjic-847	42	9	solutions	solution	NOUN
hjic-847	42	10	,	,	PUNCT
hjic-847	42	11	namely	namely	ADV
hjic-847	42	12	the	the	DET
hjic-847	42	13	1	1	NUM
hjic-847	42	14	-	-	PUNCT
hjic-847	42	15	vs	vs	ADP
hjic-847	42	16	-	-	PUNCT
hjic-847	42	17	set	set	ADJ
hjic-847	42	18	machine	machine	NOUN
hjic-847	42	19	training	training	NOUN
hjic-847	42	20	algorithm	algorithm	NOUN
hjic-847	43	1	[	[	X
hjic-847	43	2	2	2	NUM
hjic-847	43	3	]	]	PUNCT
hjic-847	43	4	and	and	CCONJ
hjic-847	43	5	the	the	DET
hjic-847	43	6	pi	pi	NOUN
hjic-847	43	7	-	-	PUNCT
hjic-847	43	8	svm	svm	NOUN
hjic-847	43	9	[	[	X
hjic-847	43	10	10	10	NUM
hjic-847	43	11	]	]	PUNCT
hjic-847	43	12	.	.	PUNCT
hjic-847	44	1	on	on	ADP
hjic-847	44	2	the	the	DET
hjic-847	44	3	other	other	ADJ
hjic-847	44	4	hand	hand	NOUN
hjic-847	44	5	,	,	PUNCT
hjic-847	44	6	it	it	PRON
hjic-847	44	7	was	be	AUX
hjic-847	44	8	shown	show	VERB
hjic-847	44	9	that	that	SCONJ
hjic-847	44	10	dpm	dpm	PROPN
hjic-847	44	11	outperforms	outperform	NOUN
hjic-847	44	12	w	w	NOUN
hjic-847	44	13	-	-	PUNCT
hjic-847	44	14	svm	svm	ADJ
hjic-847	44	15	[	[	X
hjic-847	44	16	5	5	NUM
hjic-847	44	17	]	]	PUNCT
hjic-847	44	18	,	,	PUNCT
hjic-847	44	19	therefore	therefore	ADV
hjic-847	44	20	,	,	PUNCT
hjic-847	44	21	in	in	ADP
hjic-847	44	22	this	this	DET
hjic-847	44	23	paper	paper	NOUN
hjic-847	44	24	the	the	DET
hjic-847	44	25	dpm	dpm	PROPN
hjic-847	44	26	was	be	AUX
hjic-847	44	27	used	use	VERB
hjic-847	44	28	for	for	ADP
hjic-847	44	29	unknown	unknown	ADJ
hjic-847	44	30	detection	detection	NOUN
hjic-847	44	31	.	.	PUNCT
hjic-847	45	1	bendale	bendale	NOUN
hjic-847	45	2	and	and	CCONJ
hjic-847	45	3	boult	boult	NOUN
hjic-847	45	4	defined	define	VERB
hjic-847	45	5	open	open	ADJ
hjic-847	45	6	world	world	NOUN
hjic-847	45	7	recognition	recognition	NOUN
hjic-847	45	8	and	and	CCONJ
hjic-847	45	9	presented	present	VERB
hjic-847	45	10	the	the	DET
hjic-847	45	11	nearest	near	ADJ
hjic-847	45	12	non	non	ADJ
hjic-847	45	13	-	-	ADJ
hjic-847	45	14	outlier	outlier	ADJ
hjic-847	45	15	algorithm	algorithm	NOUN
hjic-847	45	16	in	in	ADP
hjic-847	45	17	[	[	X
hjic-847	45	18	4	4	NUM
hjic-847	45	19	]	]	PUNCT
hjic-847	45	20	,	,	PUNCT
hjic-847	45	21	which	which	PRON
hjic-847	45	22	adds	add	VERB
hjic-847	45	23	object	object	VERB
hjic-847	45	24	categories	category	NOUN
hjic-847	45	25	incrementally	incrementally	ADV
hjic-847	45	26	while	while	SCONJ
hjic-847	45	27	detecting	detect	VERB
hjic-847	45	28	outliers	outlier	NOUN
hjic-847	45	29	and	and	CCONJ
hjic-847	45	30	managing	manage	VERB
hjic-847	45	31	open	open	ADJ
hjic-847	45	32	space	space	NOUN
hjic-847	45	33	risk	risk	NOUN
hjic-847	45	34	.	.	PUNCT
hjic-847	46	1	they	they	PRON
hjic-847	46	2	defined	define	VERB
hjic-847	46	3	open	open	ADJ
hjic-847	46	4	world	world	NOUN
hjic-847	46	5	recognition	recognition	NOUN
hjic-847	46	6	in	in	ADP
hjic-847	46	7	the	the	DET
hjic-847	46	8	form	form	NOUN
hjic-847	46	9	of	of	ADP
hjic-847	46	10	three	three	NUM
hjic-847	46	11	sequential	sequential	ADJ
hjic-847	46	12	steps	step	NOUN
hjic-847	46	13	:	:	PUNCT
hjic-847	46	14	a	a	DET
hjic-847	46	15	multiclass	multiclass	ADJ
hjic-847	46	16	open	open	ADJ
hjic-847	46	17	set	set	NOUN
hjic-847	46	18	recognition	recognition	NOUN
hjic-847	46	19	function	function	NOUN
hjic-847	46	20	with	with	ADP
hjic-847	46	21	a	a	DET
hjic-847	46	22	novelty	novelty	NOUN
hjic-847	46	23	detector	detector	NOUN
hjic-847	46	24	,	,	PUNCT
hjic-847	46	25	a	a	DET
hjic-847	46	26	labeling	labeling	NOUN
hjic-847	46	27	process	process	NOUN
hjic-847	46	28	and	and	CCONJ
hjic-847	46	29	an	an	DET
hjic-847	46	30	incremental	incremental	ADJ
hjic-847	46	31	learning	learning	NOUN
hjic-847	46	32	algorithm	algorithm	NOUN
hjic-847	46	33	.	.	PUNCT
hjic-847	47	1	although	although	SCONJ
hjic-847	47	2	all	all	PRON
hjic-847	47	3	of	of	ADP
hjic-847	47	4	these	these	DET
hjic-847	47	5	steps	step	NOUN
hjic-847	47	6	should	should	AUX
hjic-847	47	7	be	be	AUX
hjic-847	47	8	automated	automate	VERB
hjic-847	47	9	,	,	PUNCT
hjic-847	47	10	they	they	PRON
hjic-847	47	11	presumed	presume	VERB
hjic-847	47	12	labels	label	NOUN
hjic-847	47	13	were	be	AUX
hjic-847	47	14	obtained	obtain	VERB
hjic-847	47	15	by	by	ADP
hjic-847	47	16	human	human	ADJ
hjic-847	47	17	labeling	labeling	NOUN
hjic-847	47	18	.	.	PUNCT
hjic-847	48	1	the	the	DET
hjic-847	48	2	main	main	ADJ
hjic-847	48	3	objective	objective	NOUN
hjic-847	48	4	of	of	ADP
hjic-847	48	5	our	our	PRON
hjic-847	48	6	work	work	NOUN
hjic-847	48	7	and	and	CCONJ
hjic-847	48	8	this	this	DET
hjic-847	48	9	paper	paper	NOUN
hjic-847	48	10	is	be	AUX
hjic-847	48	11	to	to	PART
hjic-847	48	12	propose	propose	VERB
hjic-847	48	13	an	an	DET
hjic-847	48	14	automated	automate	VERB
hjic-847	48	15	labeling	labeling	NOUN
hjic-847	48	16	process	process	NOUN
hjic-847	48	17	,	,	PUNCT
hjic-847	48	18	the	the	DET
hjic-847	48	19	so	so	ADV
hjic-847	48	20	-	-	PUNCT
hjic-847	48	21	called	call	VERB
hjic-847	48	22	cluster	cluster	NOUN
hjic-847	48	23	classification	classification	NOUN
hjic-847	48	24	(	(	PUNCT
hjic-847	48	25	cc	cc	NOUN
hjic-847	48	26	)	)	PUNCT
hjic-847	48	27	.	.	PUNCT
hjic-847	49	1	in	in	ADP
hjic-847	49	2	the	the	DET
hjic-847	49	3	next	next	ADJ
hjic-847	49	4	section	section	NOUN
hjic-847	49	5	,	,	PUNCT
hjic-847	49	6	the	the	DET
hjic-847	49	7	dpm	dpm	PROPN
hjic-847	49	8	and	and	CCONJ
hjic-847	49	9	image	image	NOUN
hjic-847	49	10	clustering	clustering	NOUN
hjic-847	49	11	methods	method	NOUN
hjic-847	49	12	are	be	AUX
hjic-847	49	13	reviewed	review	VERB
hjic-847	49	14	,	,	PUNCT
hjic-847	49	15	subsequently	subsequently	ADV
hjic-847	49	16	,	,	PUNCT
hjic-847	49	17	a	a	DET
hjic-847	49	18	baseline	baseline	NOUN
hjic-847	49	19	method	method	NOUN
hjic-847	49	20	is	be	AUX
hjic-847	49	21	suggested	suggest	VERB
hjic-847	49	22	for	for	ADP
hjic-847	49	23	an	an	DET
hjic-847	49	24	open	open	ADJ
hjic-847	49	25	-	-	PUNCT
hjic-847	49	26	world	world	NOUN
hjic-847	49	27	problem	problem	NOUN
hjic-847	49	28	and	and	CCONJ
hjic-847	49	29	finally	finally	ADV
hjic-847	49	30	our	our	PRON
hjic-847	49	31	proposed	propose	VERB
hjic-847	49	32	algorithm	algorithm	NOUN
hjic-847	49	33	,	,	PUNCT
hjic-847	49	34	the	the	DET
hjic-847	49	35	cc	cc	NOUN
hjic-847	49	36	,	,	PUNCT
hjic-847	49	37	is	be	AUX
hjic-847	49	38	presented	present	VERB
hjic-847	49	39	.	.	PUNCT
hjic-847	50	1	the	the	DET
hjic-847	50	2	third	third	ADJ
hjic-847	50	3	section	section	NOUN
hjic-847	50	4	contains	contain	VERB
hjic-847	50	5	experimental	experimental	ADJ
hjic-847	50	6	results	result	NOUN
hjic-847	50	7	and	and	CCONJ
hjic-847	50	8	in	in	ADP
hjic-847	50	9	the	the	DET
hjic-847	50	10	last	last	ADJ
hjic-847	50	11	section	section	NOUN
hjic-847	50	12	our	our	PRON
hjic-847	50	13	conclusion	conclusion	NOUN
hjic-847	50	14	is	be	AUX
hjic-847	50	15	discussed	discuss	VERB
hjic-847	50	16	.	.	PUNCT
hjic-847	51	1	2	2	X
hjic-847	51	2	.	.	NUM
hjic-847	51	3	proposed	propose	VERB
hjic-847	51	4	open	open	ADJ
hjic-847	51	5	-	-	PUNCT
hjic-847	51	6	world	world	NOUN
hjic-847	51	7	recognition	recognition	NOUN
hjic-847	51	8	system	system	NOUN
hjic-847	51	9	2.1	2.1	NUM
hjic-847	51	10	double	double	ADJ
hjic-847	51	11	probability	probability	NOUN
hjic-847	51	12	model	model	NOUN
hjic-847	51	13	the	the	DET
hjic-847	51	14	dpm	dpm	PROPN
hjic-847	52	1	[	[	X
hjic-847	52	2	5	5	NUM
hjic-847	52	3	]	]	PUNCT
hjic-847	52	4	is	be	AUX
hjic-847	52	5	based	base	VERB
hjic-847	52	6	on	on	ADP
hjic-847	52	7	the	the	DET
hjic-847	52	8	likelihood	likelihood	NOUN
hjic-847	52	9	of	of	ADP
hjic-847	52	10	a	a	DET
hjic-847	52	11	classifier	classifier	NOUN
hjic-847	52	12	and	and	CCONJ
hjic-847	52	13	can	can	AUX
hjic-847	52	14	be	be	AUX
hjic-847	52	15	used	use	VERB
hjic-847	52	16	with	with	ADP
hjic-847	52	17	any	any	DET
hjic-847	52	18	kind	kind	NOUN
hjic-847	52	19	of	of	ADP
hjic-847	52	20	classifier	classifier	NOUN
hjic-847	52	21	that	that	PRON
hjic-847	52	22	provides	provide	VERB
hjic-847	52	23	class	class	NOUN
hjic-847	52	24	membership	membership	NOUN
hjic-847	52	25	probabilities	probability	NOUN
hjic-847	52	26	for	for	ADP
hjic-847	52	27	the	the	DET
hjic-847	52	28	images	image	NOUN
hjic-847	52	29	.	.	PUNCT
hjic-847	53	1	as	as	ADP
hjic-847	53	2	a	a	DET
hjic-847	53	3	result	result	NOUN
hjic-847	53	4	,	,	PUNCT
hjic-847	53	5	after	after	ADP
hjic-847	53	6	training	train	VERB
hjic-847	53	7	the	the	DET
hjic-847	53	8	classifier	classifier	NOUN
hjic-847	53	9	,	,	PUNCT
hjic-847	53	10	it	it	PRON
hjic-847	53	11	is	be	AUX
hjic-847	53	12	capable	capable	ADJ
hjic-847	53	13	of	of	ADP
hjic-847	53	14	making	make	VERB
hjic-847	53	15	predictions	prediction	NOUN
hjic-847	53	16	with	with	ADP
hjic-847	53	17	reliability	reliability	NOUN
hjic-847	53	18	values	value	NOUN
hjic-847	53	19	(	(	PUNCT
hjic-847	53	20	scores	score	NOUN
hjic-847	53	21	)	)	PUNCT
hjic-847	53	22	for	for	ADP
hjic-847	53	23	each	each	DET
hjic-847	53	24	class	class	NOUN
hjic-847	53	25	,	,	PUNCT
hjic-847	53	26	i.e.	i.e.	X
hjic-847	53	27	decision	decision	NOUN
hjic-847	53	28	vectors	vector	NOUN
hjic-847	53	29	.	.	PUNCT
hjic-847	54	1	the	the	DET
hjic-847	54	2	range	range	NOUN
hjic-847	54	3	of	of	ADP
hjic-847	54	4	the	the	DET
hjic-847	54	5	scores	score	NOUN
hjic-847	54	6	depends	depend	VERB
hjic-847	54	7	on	on	ADP
hjic-847	54	8	the	the	DET
hjic-847	54	9	type	type	NOUN
hjic-847	54	10	of	of	ADP
hjic-847	54	11	classifier	classifier	NOUN
hjic-847	54	12	(	(	PUNCT
hjic-847	54	13	sometimes	sometimes	ADV
hjic-847	54	14	it	it	PRON
hjic-847	54	15	is	be	AUX
hjic-847	54	16	from	from	ADP
hjic-847	54	17	0	0	NUM
hjic-847	54	18	to	to	ADP
hjic-847	54	19	1	1	NUM
hjic-847	54	20	but	but	CCONJ
hjic-847	54	21	it	it	PRON
hjic-847	54	22	can	can	AUX
hjic-847	54	23	be	be	AUX
hjic-847	54	24	over	over	ADP
hjic-847	54	25	any	any	DET
hjic-847	54	26	range	range	NOUN
hjic-847	54	27	.	.	PUNCT
hjic-847	55	1	only	only	ADV
hjic-847	55	2	one	one	NUM
hjic-847	55	3	condition	condition	NOUN
hjic-847	55	4	is	be	AUX
hjic-847	55	5	required	require	VERB
hjic-847	55	6	,	,	PUNCT
hjic-847	55	7	namely	namely	ADV
hjic-847	55	8	the	the	DET
hjic-847	55	9	larger	large	ADJ
hjic-847	55	10	score	score	NOUN
hjic-847	55	11	for	for	ADP
hjic-847	55	12	class	class	NOUN
hjic-847	55	13	ci	ci	PROPN
hjic-847	55	14	should	should	AUX
hjic-847	55	15	represent	represent	VERB
hjic-847	55	16	the	the	DET
hjic-847	55	17	higher	high	ADJ
hjic-847	55	18	likelihood	likelihood	NOUN
hjic-847	55	19	of	of	ADP
hjic-847	55	20	being	be	AUX
hjic-847	55	21	a	a	DET
hjic-847	55	22	member	member	NOUN
hjic-847	55	23	of	of	ADP
hjic-847	55	24	class	class	PROPN
hjic-847	55	25	ci	ci	PROPN
hjic-847	55	26	.	.	PUNCT
hjic-847	56	1	in	in	ADP
hjic-847	56	2	the	the	DET
hjic-847	56	3	training	training	NOUN
hjic-847	56	4	set	set	NOUN
hjic-847	56	5	or	or	CCONJ
hjic-847	56	6	a	a	DET
hjic-847	56	7	validation	validation	NOUN
hjic-847	56	8	set	set	NOUN
hjic-847	56	9	,	,	PUNCT
hjic-847	56	10	the	the	DET
hjic-847	56	11	instances	instance	NOUN
hjic-847	56	12	with	with	ADP
hjic-847	56	13	corresponding	corresponding	ADJ
hjic-847	56	14	scores	score	NOUN
hjic-847	56	15	are	be	AUX
hjic-847	56	16	investigated	investigate	VERB
hjic-847	56	17	in	in	ADP
hjic-847	56	18	each	each	DET
hjic-847	56	19	class	class	NOUN
hjic-847	56	20	.	.	PUNCT
hjic-847	57	1	the	the	DET
hjic-847	57	2	ground	ground	NOUN
hjic-847	57	3	truth	truth	NOUN
hjic-847	57	4	is	be	AUX
hjic-847	57	5	known	know	VERB
hjic-847	57	6	for	for	ADP
hjic-847	57	7	this	this	DET
hjic-847	57	8	set	set	NOUN
hjic-847	57	9	,	,	PUNCT
hjic-847	57	10	so	so	CCONJ
hjic-847	57	11	the	the	DET
hjic-847	57	12	positive	positive	ADJ
hjic-847	57	13	elements	element	NOUN
hjic-847	57	14	can	can	AUX
hjic-847	57	15	be	be	AUX
hjic-847	57	16	selected	select	VERB
hjic-847	57	17	from	from	ADP
hjic-847	57	18	each	each	DET
hjic-847	57	19	class	class	NOUN
hjic-847	57	20	.	.	PUNCT
hjic-847	58	1	in	in	ADP
hjic-847	58	2	order	order	NOUN
hjic-847	58	3	to	to	PART
hjic-847	58	4	calculate	calculate	VERB
hjic-847	58	5	the	the	DET
hjic-847	58	6	conditional	conditional	ADJ
hjic-847	58	7	probability	probability	NOUN
hjic-847	58	8	that	that	SCONJ
hjic-847	58	9	a	a	DET
hjic-847	58	10	new	new	ADJ
hjic-847	58	11	instance	instance	NOUN
hjic-847	58	12	belongs	belong	VERB
hjic-847	58	13	to	to	ADP
hjic-847	58	14	class	class	NOUN
hjic-847	58	15	ci	ci	NOUN
hjic-847	58	16	according	accord	VERB
hjic-847	58	17	to	to	ADP
hjic-847	58	18	its	its	PRON
hjic-847	58	19	score	score	NOUN
hjic-847	58	20	,	,	PUNCT
hjic-847	58	21	the	the	DET
hjic-847	58	22	cumulative	cumulative	ADJ
hjic-847	58	23	distribution	distribution	NOUN
hjic-847	58	24	function	function	NOUN
hjic-847	58	25	(	(	PUNCT
hjic-847	58	26	cdf	cdf	PROPN
hjic-847	58	27	)	)	PUNCT
hjic-847	58	28	of	of	ADP
hjic-847	58	29	positive	positive	ADJ
hjic-847	58	30	scores	score	NOUN
hjic-847	58	31	should	should	AUX
hjic-847	58	32	be	be	AUX
hjic-847	58	33	determined	determine	VERB
hjic-847	58	34	,	,	PUNCT
hjic-847	58	35	therefore	therefore	ADV
hjic-847	58	36	,	,	PUNCT
hjic-847	58	37	a	a	DET
hjic-847	58	38	reverse	reverse	NOUN
hjic-847	58	39	cdf	cdf	NOUN
hjic-847	58	40	of	of	ADP
hjic-847	58	41	negative	negative	ADJ
hjic-847	58	42	scores	score	NOUN
hjic-847	58	43	was	be	AUX
hjic-847	58	44	created	create	VERB
hjic-847	58	45	:	:	PUNCT
hjic-847	58	46	fpi	fpi	PROPN
hjic-847	58	47	(	(	PUNCT
hjic-847	58	48	x	x	X
hjic-847	58	49	)	)	PUNCT
hjic-847	58	50	=	=	SYM
hjic-847	58	51	p	p	X
hjic-847	58	52	(	(	PUNCT
hjic-847	58	53	ci|score	ci|score	X
hjic-847	58	54	<	<	X
hjic-847	58	55	x	x	X
hjic-847	58	56	)	)	PUNCT
hjic-847	58	57	,	,	PUNCT
hjic-847	58	58	(	(	PUNCT
hjic-847	58	59	1	1	X
hjic-847	58	60	)	)	PUNCT
hjic-847	58	61	fni	fni	NOUN
hjic-847	58	62	(	(	PUNCT
hjic-847	58	63	x	x	NOUN
hjic-847	58	64	)	)	PUNCT
hjic-847	58	65	=	=	SYM
hjic-847	58	66	p	p	NOUN
hjic-847	58	67	(	(	PUNCT
hjic-847	58	68	¬ci|score	¬ci|score	X
hjic-847	58	69	>	>	X
hjic-847	58	70	x	x	X
hjic-847	58	71	)	)	PUNCT
hjic-847	58	72	,	,	PUNCT
hjic-847	58	73	(	(	PUNCT
hjic-847	58	74	2	2	X
hjic-847	58	75	)	)	PUNCT
hjic-847	58	76	where	where	SCONJ
hjic-847	58	77	pi	pi	NOUN
hjic-847	58	78	and	and	CCONJ
hjic-847	58	79	ni	ni	PROPN
hjic-847	58	80	denote	denote	VERB
hjic-847	58	81	the	the	DET
hjic-847	58	82	positive	positive	ADJ
hjic-847	58	83	and	and	CCONJ
hjic-847	58	84	negative	negative	ADJ
hjic-847	58	85	elements	element	NOUN
hjic-847	58	86	,	,	PUNCT
hjic-847	58	87	respectively	respectively	ADV
hjic-847	58	88	.	.	PUNCT
hjic-847	59	1	note	note	VERB
hjic-847	59	2	that	that	SCONJ
hjic-847	59	3	the	the	DET
hjic-847	59	4	sum	sum	NOUN
hjic-847	59	5	of	of	ADP
hjic-847	59	6	these	these	DET
hjic-847	59	7	probabilities	probability	NOUN
hjic-847	59	8	is	be	AUX
hjic-847	59	9	not	not	PART
hjic-847	59	10	always	always	ADV
hjic-847	59	11	equal	equal	ADJ
hjic-847	59	12	to	to	ADP
hjic-847	59	13	1	1	NUM
hjic-847	59	14	(	(	PUNCT
hjic-847	59	15	this	this	PRON
hjic-847	59	16	is	be	AUX
hjic-847	59	17	not	not	PART
hjic-847	59	18	a	a	DET
hjic-847	59	19	requirement	requirement	NOUN
hjic-847	59	20	)	)	PUNCT
hjic-847	59	21	.	.	PUNCT
hjic-847	60	1	a	a	DET
hjic-847	60	2	dpm	dpm	PROPN
hjic-847	60	3	was	be	AUX
hjic-847	60	4	constructed	construct	VERB
hjic-847	60	5	based	base	VERB
hjic-847	60	6	on	on	ADP
hjic-847	60	7	the	the	DET
hjic-847	60	8	cdf	cdf	PROPN
hjic-847	60	9	and	and	CCONJ
hjic-847	60	10	reverse	reverse	VERB
hjic-847	60	11	cdf	cdf	PROPN
hjic-847	60	12	functions	function	NOUN
hjic-847	60	13	.	.	PUNCT
hjic-847	61	1	during	during	ADP
hjic-847	61	2	testing	testing	NOUN
hjic-847	61	3	,	,	PUNCT
hjic-847	61	4	the	the	DET
hjic-847	61	5	focus	focus	NOUN
hjic-847	61	6	is	be	AUX
hjic-847	61	7	on	on	ADP
hjic-847	61	8	the	the	DET
hjic-847	61	9	likelihood	likelihood	NOUN
hjic-847	61	10	of	of	ADP
hjic-847	61	11	the	the	DET
hjic-847	61	12	occurrence	occurrence	NOUN
hjic-847	61	13	of	of	ADP
hjic-847	61	14	an	an	DET
hjic-847	61	15	unknown	unknown	ADJ
hjic-847	61	16	class	class	NOUN
hjic-847	61	17	compared	compare	VERB
hjic-847	61	18	with	with	ADP
hjic-847	61	19	any	any	PRON
hjic-847	61	20	of	of	ADP
hjic-847	61	21	the	the	DET
hjic-847	61	22	known	know	VERB
hjic-847	61	23	classes	class	NOUN
hjic-847	61	24	.	.	PUNCT
hjic-847	62	1	before	before	ADP
hjic-847	62	2	the	the	DET
hjic-847	62	3	comparison	comparison	NOUN
hjic-847	62	4	,	,	PUNCT
hjic-847	62	5	the	the	DET
hjic-847	62	6	probabilities	probability	NOUN
hjic-847	62	7	of	of	ADP
hjic-847	62	8	the	the	DET
hjic-847	62	9	known	know	VERB
hjic-847	62	10	classes	class	NOUN
hjic-847	62	11	should	should	AUX
hjic-847	62	12	be	be	AUX
hjic-847	62	13	calculated	calculate	VERB
hjic-847	62	14	.	.	PUNCT
hjic-847	63	1	scores	score	NOUN
hjic-847	63	2	(	(	PUNCT
hjic-847	63	3	scorei	scorei	ADV
hjic-847	63	4	for	for	ADP
hjic-847	63	5	class	class	NOUN
hjic-847	63	6	ci	ci	PROPN
hjic-847	63	7	)	)	PUNCT
hjic-847	63	8	for	for	ADP
hjic-847	63	9	a	a	DET
hjic-847	63	10	new	new	ADJ
hjic-847	63	11	instance	instance	NOUN
hjic-847	63	12	are	be	AUX
hjic-847	63	13	obtained	obtain	VERB
hjic-847	63	14	as	as	ADP
hjic-847	63	15	outputs	output	NOUN
hjic-847	63	16	from	from	ADP
hjic-847	63	17	the	the	DET
hjic-847	63	18	original	original	ADJ
hjic-847	63	19	classifier	classifier	NOUN
hjic-847	63	20	,	,	PUNCT
hjic-847	63	21	and	and	CCONJ
hjic-847	63	22	based	base	VERB
hjic-847	63	23	on	on	ADP
hjic-847	63	24	them	they	PRON
hjic-847	63	25	the	the	DET
hjic-847	63	26	probability	probability	NOUN
hjic-847	63	27	of	of	ADP
hjic-847	63	28	class	class	NOUN
hjic-847	63	29	ci	ci	NOUN
hjic-847	63	30	occurring	occur	VERB
hjic-847	63	31	can	can	AUX
hjic-847	63	32	be	be	AUX
hjic-847	63	33	expressed	express	VERB
hjic-847	63	34	as	as	SCONJ
hjic-847	63	35	described	describe	VERB
hjic-847	63	36	in	in	ADP
hjic-847	63	37	pci	pci	PROPN
hjic-847	63	38	=	=	PROPN
hjic-847	63	39	fpi	fpi	PROPN
hjic-847	63	40	(	(	PUNCT
hjic-847	63	41	scorei	scorei	NOUN
hjic-847	63	42	)	)	PUNCT
hjic-847	64	1	k∏	k∏	PROPN
hjic-847	64	2	j=1,j	j=1,j	NOUN
hjic-847	64	3	6	6	NUM
hjic-847	64	4	=	=	NOUN
hjic-847	64	5	i	i	PROPN
hjic-847	64	6	fnj	fnj	NOUN
hjic-847	64	7	(	(	PUNCT
hjic-847	64	8	scorej	scorej	PROPN
hjic-847	64	9	)	)	PUNCT
hjic-847	64	10	.	.	PUNCT
hjic-847	65	1	(	(	PUNCT
hjic-847	65	2	3	3	X
hjic-847	65	3	)	)	PUNCT
hjic-847	65	4	an	an	DET
hjic-847	65	5	expression	expression	NOUN
hjic-847	65	6	for	for	ADP
hjic-847	65	7	the	the	DET
hjic-847	65	8	probability	probability	NOUN
hjic-847	65	9	of	of	ADP
hjic-847	65	10	class	class	NOUN
hjic-847	65	11	ck+1	ck+1	X
hjic-847	65	12	is	be	AUX
hjic-847	65	13	pck+1	pck+1	NOUN
hjic-847	65	14	=	=	PUNCT
hjic-847	65	15	k∏	k∏	PROPN
hjic-847	65	16	j=1	j=1	PROPN
hjic-847	65	17	fnj	fnj	NOUN
hjic-847	65	18	(	(	PUNCT
hjic-847	65	19	scorej	scorej	PROPN
hjic-847	65	20	)	)	PUNCT
hjic-847	65	21	.	.	PUNCT
hjic-847	66	1	(	(	PUNCT
hjic-847	66	2	4	4	X
hjic-847	66	3	)	)	PUNCT
hjic-847	66	4	if	if	SCONJ
hjic-847	66	5	the	the	DET
hjic-847	66	6	probability	probability	NOUN
hjic-847	66	7	of	of	ADP
hjic-847	66	8	being	be	AUX
hjic-847	66	9	a	a	DET
hjic-847	66	10	member	member	NOUN
hjic-847	66	11	of	of	ADP
hjic-847	66	12	class	class	NOUN
hjic-847	66	13	ck+1	ck+1	X
hjic-847	66	14	is	be	AUX
hjic-847	66	15	higher	high	ADJ
hjic-847	66	16	than	than	ADP
hjic-847	66	17	for	for	ADP
hjic-847	66	18	any	any	DET
hjic-847	66	19	other	other	ADJ
hjic-847	66	20	(	(	PUNCT
hjic-847	66	21	known	known	ADJ
hjic-847	66	22	)	)	PUNCT
hjic-847	66	23	class	class	NOUN
hjic-847	66	24	,	,	PUNCT
hjic-847	66	25	then	then	ADV
hjic-847	66	26	the	the	DET
hjic-847	66	27	new	new	ADJ
hjic-847	66	28	instance	instance	NOUN
hjic-847	66	29	will	will	AUX
hjic-847	66	30	be	be	AUX
hjic-847	66	31	a	a	DET
hjic-847	66	32	member	member	NOUN
hjic-847	66	33	of	of	ADP
hjic-847	66	34	the	the	DET
hjic-847	66	35	unknown	unknown	ADJ
hjic-847	66	36	class	class	NOUN
hjic-847	66	37	.	.	PUNCT
hjic-847	67	1	otherwise	otherwise	ADV
hjic-847	67	2	the	the	DET
hjic-847	67	3	prediction	prediction	NOUN
hjic-847	67	4	is	be	AUX
hjic-847	67	5	based	base	VERB
hjic-847	67	6	on	on	ADP
hjic-847	67	7	the	the	DET
hjic-847	67	8	original	original	ADJ
hjic-847	67	9	classifier	classifier	NOUN
hjic-847	67	10	,	,	PUNCT
hjic-847	67	11	i.e.	i.e.	X
hjic-847	67	12	the	the	DET
hjic-847	67	13	class	class	NOUN
hjic-847	67	14	with	with	ADP
hjic-847	67	15	the	the	DET
hjic-847	67	16	largest	large	ADJ
hjic-847	67	17	score	score	NOUN
hjic-847	67	18	will	will	AUX
hjic-847	67	19	be	be	AUX
hjic-847	67	20	selected	select	VERB
hjic-847	67	21	.	.	PUNCT
hjic-847	68	1	the	the	DET
hjic-847	68	2	decision	decision	NOUN
hjic-847	68	3	with	with	ADP
hjic-847	68	4	regard	regard	NOUN
hjic-847	68	5	to	to	ADP
hjic-847	68	6	the	the	DET
hjic-847	68	7	prediction	prediction	NOUN
hjic-847	68	8	of	of	ADP
hjic-847	68	9	test	test	NOUN
hjic-847	68	10	instance	instance	PROPN
hjic-847	68	11	j	j	PROPN
hjic-847	68	12	is	be	AUX
hjic-847	68	13	formalized	formalize	VERB
hjic-847	68	14	as	as	ADP
hjic-847	68	15	dj	dj	NOUN
hjic-847	68	16	=	=	NOUN
hjic-847	68	17	{	{	PUNCT
hjic-847	68	18	ck+1	ck+1	ADV
hjic-847	68	19	|	|	ADV
hjic-847	68	20	pck+1	pck+1	VERB
hjic-847	68	21	>	>	X
hjic-847	68	22	maxi	maxi	ADJ
hjic-847	68	23	{	{	PUNCT
hjic-847	68	24	pci	pci	ADJ
hjic-847	68	25	}	}	PUNCT
hjic-847	68	26	argmaxj	argmaxj	ADV
hjic-847	68	27	{	{	PUNCT
hjic-847	68	28	scorej	scorej	NOUN
hjic-847	68	29	}	}	PUNCT
hjic-847	69	1	|	|	ADV
hjic-847	69	2	otherwise	otherwise	ADV
hjic-847	69	3	(	(	PUNCT
hjic-847	69	4	5	5	X
hjic-847	69	5	)	)	PUNCT
hjic-847	69	6	hungarian	hungarian	ADJ
hjic-847	69	7	journal	journal	NOUN
hjic-847	69	8	of	of	ADP
hjic-847	69	9	industry	industry	NOUN
hjic-847	69	10	and	and	CCONJ
hjic-847	69	11	chemistry	chemistry	NOUN
hjic-847	69	12	automated	automate	VERB
hjic-847	69	13	labeling	labeling	NOUN
hjic-847	69	14	process	process	NOUN
hjic-847	69	15	for	for	ADP
hjic-847	69	16	unknown	unknown	ADJ
hjic-847	69	17	images	image	NOUN
hjic-847	69	18	35	35	NUM
hjic-847	69	19	at	at	ADP
hjic-847	69	20	this	this	DET
hjic-847	69	21	point	point	NOUN
hjic-847	69	22	the	the	DET
hjic-847	69	23	algorithm	algorithm	NOUN
hjic-847	69	24	is	be	AUX
hjic-847	69	25	able	able	ADJ
hjic-847	69	26	to	to	PART
hjic-847	69	27	make	make	VERB
hjic-847	69	28	a	a	DET
hjic-847	69	29	decision	decision	NOUN
hjic-847	69	30	about	about	ADP
hjic-847	69	31	test	test	NOUN
hjic-847	69	32	data	datum	NOUN
hjic-847	69	33	if	if	SCONJ
hjic-847	69	34	it	it	PRON
hjic-847	69	35	originates	originate	VERB
hjic-847	69	36	from	from	ADP
hjic-847	69	37	an	an	DET
hjic-847	69	38	unknown	unknown	ADJ
hjic-847	69	39	category	category	NOUN
hjic-847	69	40	.	.	PUNCT
hjic-847	70	1	also	also	ADV
hjic-847	70	2	,	,	PUNCT
hjic-847	70	3	should	should	AUX
hjic-847	70	4	it	it	PRON
hjic-847	70	5	originate	originate	VERB
hjic-847	70	6	from	from	ADP
hjic-847	70	7	a	a	DET
hjic-847	70	8	known	know	VERB
hjic-847	70	9	category	category	NOUN
hjic-847	70	10	,	,	PUNCT
hjic-847	70	11	then	then	ADV
hjic-847	70	12	based	base	VERB
hjic-847	70	13	on	on	ADP
hjic-847	70	14	the	the	DET
hjic-847	70	15	output	output	NOUN
hjic-847	70	16	of	of	ADP
hjic-847	70	17	the	the	DET
hjic-847	70	18	classifier	classifier	NOUN
hjic-847	70	19	its	its	PRON
hjic-847	70	20	known	know	VERB
hjic-847	70	21	category	category	NOUN
hjic-847	70	22	can	can	AUX
hjic-847	70	23	be	be	AUX
hjic-847	70	24	determined	determine	VERB
hjic-847	70	25	.	.	PUNCT
hjic-847	71	1	2.2	2.2	NUM
hjic-847	71	2	unknown	unknown	ADJ
hjic-847	71	3	image	image	NOUN
hjic-847	71	4	clustering	cluster	VERB
hjic-847	71	5	the	the	DET
hjic-847	71	6	image	image	NOUN
hjic-847	71	7	representations	representation	NOUN
hjic-847	71	8	were	be	AUX
hjic-847	71	9	created	create	VERB
hjic-847	71	10	according	accord	VERB
hjic-847	71	11	to	to	ADP
hjic-847	71	12	the	the	DET
hjic-847	71	13	bag	bag	NOUN
hjic-847	71	14	-	-	PUNCT
hjic-847	71	15	of	of	ADP
hjic-847	71	16	-	-	PUNCT
hjic-847	71	17	words	word	NOUN
hjic-847	71	18	[	[	X
hjic-847	71	19	11	11	NUM
hjic-847	71	20	,	,	PUNCT
hjic-847	71	21	12	12	NUM
hjic-847	71	22	]	]	PUNCT
hjic-847	71	23	model	model	NOUN
hjic-847	71	24	.	.	PUNCT
hjic-847	72	1	based	base	VERB
hjic-847	72	2	on	on	ADP
hjic-847	72	3	their	their	PRON
hjic-847	72	4	visual	visual	ADJ
hjic-847	72	5	content	content	NOUN
hjic-847	72	6	,	,	PUNCT
hjic-847	72	7	each	each	DET
hjic-847	72	8	image	image	NOUN
hjic-847	72	9	was	be	AUX
hjic-847	72	10	represented	represent	VERB
hjic-847	72	11	by	by	ADP
hjic-847	72	12	a	a	DET
hjic-847	72	13	single	single	ADJ
hjic-847	72	14	high	high	ADJ
hjic-847	72	15	dimensional	dimensional	ADJ
hjic-847	72	16	vector	vector	NOUN
hjic-847	72	17	.	.	PUNCT
hjic-847	73	1	in	in	ADP
hjic-847	73	2	order	order	NOUN
hjic-847	73	3	to	to	PART
hjic-847	73	4	create	create	VERB
hjic-847	73	5	these	these	DET
hjic-847	73	6	high	high	ADJ
hjic-847	73	7	-	-	PUNCT
hjic-847	73	8	level	level	NOUN
hjic-847	73	9	descriptors	descriptor	NOUN
hjic-847	73	10	,	,	PUNCT
hjic-847	73	11	the	the	DET
hjic-847	73	12	local	local	ADJ
hjic-847	73	13	attributes	attribute	NOUN
hjic-847	73	14	of	of	ADP
hjic-847	73	15	the	the	DET
hjic-847	73	16	images	image	NOUN
hjic-847	73	17	were	be	AUX
hjic-847	73	18	investigated	investigate	VERB
hjic-847	73	19	by	by	ADP
hjic-847	73	20	calculating	calculate	VERB
hjic-847	73	21	the	the	DET
hjic-847	73	22	low	low	ADJ
hjic-847	73	23	-	-	PUNCT
hjic-847	73	24	level	level	NOUN
hjic-847	73	25	scale	scale	NOUN
hjic-847	73	26	invariant	invariant	ADJ
hjic-847	73	27	feature	feature	NOUN
hjic-847	73	28	transform	transform	NOUN
hjic-847	73	29	(	(	PUNCT
hjic-847	73	30	sift	sift	NOUN
hjic-847	73	31	)	)	PUNCT
hjic-847	74	1	[	[	X
hjic-847	74	2	13	13	NUM
hjic-847	74	3	]	]	PUNCT
hjic-847	74	4	descriptor	descriptor	NOUN
hjic-847	74	5	.	.	PUNCT
hjic-847	75	1	next	next	ADV
hjic-847	75	2	,	,	PUNCT
hjic-847	75	3	the	the	DET
hjic-847	75	4	gaussian	gaussian	ADJ
hjic-847	75	5	mixture	mixture	NOUN
hjic-847	75	6	model	model	NOUN
hjic-847	75	7	(	(	PUNCT
hjic-847	75	8	gmm	gmm	X
hjic-847	75	9	)	)	PUNCT
hjic-847	76	1	[	[	X
hjic-847	76	2	14–16	14–16	NUM
hjic-847	76	3	]	]	PUNCT
hjic-847	76	4	was	be	AUX
hjic-847	76	5	used	use	VERB
hjic-847	76	6	to	to	PART
hjic-847	76	7	define	define	VERB
hjic-847	76	8	the	the	DET
hjic-847	76	9	visual	visual	ADJ
hjic-847	76	10	code	code	NOUN
hjic-847	76	11	words	word	NOUN
hjic-847	76	12	and	and	CCONJ
hjic-847	76	13	the	the	DET
hjic-847	76	14	fisher	fisher	PROPN
hjic-847	76	15	vectors	vector	NOUN
hjic-847	76	16	[	[	X
hjic-847	76	17	17	17	NUM
hjic-847	76	18	,	,	PUNCT
hjic-847	76	19	18	18	NUM
hjic-847	76	20	]	]	PUNCT
hjic-847	76	21	to	to	PART
hjic-847	76	22	encode	encode	VERB
hjic-847	76	23	the	the	DET
hjic-847	76	24	low	low	ADJ
hjic-847	76	25	-	-	PUNCT
hjic-847	76	26	level	level	NOUN
hjic-847	76	27	descriptors	descriptor	NOUN
hjic-847	76	28	into	into	ADP
hjic-847	76	29	high	high	ADJ
hjic-847	76	30	-	-	PUNCT
hjic-847	76	31	level	level	NOUN
hjic-847	76	32	descriptors	descriptor	NOUN
hjic-847	76	33	based	base	VERB
hjic-847	76	34	on	on	ADP
hjic-847	76	35	the	the	DET
hjic-847	76	36	visual	visual	ADJ
hjic-847	76	37	code	code	NOUN
hjic-847	76	38	words	word	NOUN
hjic-847	76	39	.	.	PUNCT
hjic-847	77	1	the	the	DET
hjic-847	77	2	fisher	fisher	PROPN
hjic-847	77	3	vectors	vector	NOUN
hjic-847	77	4	were	be	AUX
hjic-847	77	5	the	the	DET
hjic-847	77	6	final	final	ADJ
hjic-847	77	7	representations	representation	NOUN
hjic-847	77	8	(	(	PUNCT
hjic-847	77	9	image	image	NOUN
hjic-847	77	10	descriptors	descriptor	NOUN
hjic-847	77	11	)	)	PUNCT
hjic-847	77	12	of	of	ADP
hjic-847	77	13	the	the	DET
hjic-847	77	14	images	image	NOUN
hjic-847	77	15	and	and	CCONJ
hjic-847	77	16	were	be	AUX
hjic-847	77	17	used	use	VERB
hjic-847	77	18	as	as	ADP
hjic-847	77	19	the	the	DET
hjic-847	77	20	input	input	NOUN
hjic-847	77	21	data	datum	NOUN
hjic-847	77	22	for	for	ADP
hjic-847	77	23	the	the	DET
hjic-847	77	24	clustering	clustering	ADJ
hjic-847	77	25	algorithm	algorithm	NOUN
hjic-847	77	26	.	.	PUNCT
hjic-847	78	1	after	after	SCONJ
hjic-847	78	2	the	the	DET
hjic-847	78	3	final	final	ADJ
hjic-847	78	4	clusters	cluster	NOUN
hjic-847	78	5	of	of	ADP
hjic-847	78	6	fisher	fisher	PROPN
hjic-847	78	7	vectors	vector	NOUN
hjic-847	78	8	were	be	AUX
hjic-847	78	9	formed	form	VERB
hjic-847	78	10	,	,	PUNCT
hjic-847	78	11	the	the	DET
hjic-847	78	12	image	image	NOUN
hjic-847	78	13	clusters	cluster	NOUN
hjic-847	78	14	could	could	AUX
hjic-847	78	15	be	be	AUX
hjic-847	78	16	produced	produce	VERB
hjic-847	78	17	by	by	ADP
hjic-847	78	18	substituting	substitute	VERB
hjic-847	78	19	the	the	DET
hjic-847	78	20	fisher	fisher	PROPN
hjic-847	78	21	vectors	vector	NOUN
hjic-847	78	22	for	for	ADP
hjic-847	78	23	the	the	DET
hjic-847	78	24	corresponding	corresponding	ADJ
hjic-847	78	25	images	image	NOUN
hjic-847	78	26	.	.	PUNCT
hjic-847	79	1	the	the	DET
hjic-847	79	2	basis	basis	NOUN
hjic-847	79	3	of	of	ADP
hjic-847	79	4	our	our	PRON
hjic-847	79	5	clustering	cluster	VERB
hjic-847	79	6	approach	approach	NOUN
hjic-847	79	7	is	be	AUX
hjic-847	79	8	the	the	DET
hjic-847	79	9	wellknown	wellknown	ADJ
hjic-847	79	10	k	k	NOUN
hjic-847	79	11	-	-	PUNCT
hjic-847	79	12	means	mean	VERB
hjic-847	79	13	clustering	cluster	VERB
hjic-847	79	14	algorithm	algorithm	NOUN
hjic-847	79	15	[	[	X
hjic-847	79	16	19	19	NUM
hjic-847	79	17	]	]	PUNCT
hjic-847	79	18	which	which	PRON
hjic-847	79	19	consists	consist	VERB
hjic-847	79	20	of	of	ADP
hjic-847	79	21	two	two	NUM
hjic-847	79	22	important	important	ADJ
hjic-847	79	23	inputs	input	NOUN
hjic-847	79	24	,	,	PUNCT
hjic-847	79	25	namely	namely	ADV
hjic-847	79	26	the	the	DET
hjic-847	79	27	initial	initial	ADJ
hjic-847	79	28	cluster	cluster	NOUN
hjic-847	79	29	centers	center	NOUN
hjic-847	79	30	and	and	CCONJ
hjic-847	79	31	the	the	DET
hjic-847	79	32	number	number	NOUN
hjic-847	79	33	of	of	ADP
hjic-847	79	34	clusters	cluster	NOUN
hjic-847	79	35	.	.	PUNCT
hjic-847	80	1	the	the	DET
hjic-847	80	2	k	k	NOUN
hjic-847	80	3	-	-	PUNCT
hjic-847	80	4	means	mean	VERB
hjic-847	80	5	clustering	cluster	VERB
hjic-847	80	6	algorithm	algorithm	NOUN
hjic-847	80	7	aims	aim	VERB
hjic-847	80	8	to	to	PART
hjic-847	80	9	minimize	minimize	VERB
hjic-847	80	10	the	the	DET
hjic-847	80	11	sum	sum	NOUN
hjic-847	80	12	of	of	ADP
hjic-847	80	13	squared	squared	ADJ
hjic-847	80	14	distances	distance	NOUN
hjic-847	80	15	from	from	ADP
hjic-847	80	16	all	all	DET
hjic-847	80	17	points	point	NOUN
hjic-847	80	18	to	to	ADP
hjic-847	80	19	their	their	PRON
hjic-847	80	20	cluster	cluster	NOUN
hjic-847	80	21	centers	center	NOUN
hjic-847	80	22	:	:	PUNCT
hjic-847	80	23	e	e	X
hjic-847	80	24	=	=	SYM
hjic-847	80	25	min	min	PROPN
hjic-847	80	26	(	(	PUNCT
hjic-847	80	27	k∑	k∑	VERB
hjic-847	80	28	l=1	l=1	X
hjic-847	80	29	∑	∑	PUNCT
hjic-847	80	30	xi∈cl	xi∈cl	PUNCT
hjic-847	81	1	‖xi	‖xi	NUM
hjic-847	81	2	−	−	PROPN
hjic-847	81	3	zl‖2	zl‖2	PROPN
hjic-847	81	4	)	)	PUNCT
hjic-847	81	5	,	,	PUNCT
hjic-847	81	6	(	(	PUNCT
hjic-847	81	7	6	6	NUM
hjic-847	81	8	)	)	PUNCT
hjic-847	81	9	where	where	SCONJ
hjic-847	81	10	k	k	PROPN
hjic-847	81	11	denotes	denote	VERB
hjic-847	81	12	the	the	DET
hjic-847	81	13	number	number	NOUN
hjic-847	81	14	of	of	ADP
hjic-847	81	15	clusters	cluster	NOUN
hjic-847	81	16	,	,	PUNCT
hjic-847	81	17	xi	xi	PROPN
hjic-847	81	18	represents	represent	VERB
hjic-847	81	19	a	a	DET
hjic-847	81	20	member	member	NOUN
hjic-847	81	21	of	of	ADP
hjic-847	81	22	cluster	cluster	NOUN
hjic-847	81	23	cl	cl	NOUN
hjic-847	81	24	and	and	CCONJ
hjic-847	81	25	zl	zl	X
hjic-847	81	26	stands	stand	VERB
hjic-847	81	27	for	for	ADP
hjic-847	81	28	the	the	DET
hjic-847	81	29	center	center	NOUN
hjic-847	81	30	of	of	ADP
hjic-847	81	31	it	it	PRON
hjic-847	81	32	.	.	PUNCT
hjic-847	82	1	however	however	ADV
hjic-847	82	2	,	,	PUNCT
hjic-847	82	3	the	the	DET
hjic-847	82	4	fisher	fisher	PROPN
hjic-847	82	5	vector	vector	NOUN
hjic-847	82	6	consists	consist	VERB
hjic-847	82	7	of	of	ADP
hjic-847	82	8	65,791	65,791	NUM
hjic-847	82	9	dimensions	dimension	NOUN
hjic-847	82	10	,	,	PUNCT
hjic-847	82	11	and	and	CCONJ
hjic-847	82	12	the	the	DET
hjic-847	82	13	basic	basic	ADJ
hjic-847	82	14	k	k	NOUN
hjic-847	82	15	-	-	PUNCT
hjic-847	82	16	means	mean	VERB
hjic-847	82	17	clustering	clustering	ADJ
hjic-847	82	18	algorithm	algorithm	NOUN
hjic-847	82	19	performs	perform	VERB
hjic-847	82	20	less	less	ADV
hjic-847	82	21	efficiently	efficiently	ADV
hjic-847	82	22	when	when	SCONJ
hjic-847	82	23	the	the	DET
hjic-847	82	24	clusters	cluster	NOUN
hjic-847	82	25	are	be	AUX
hjic-847	82	26	non	non	ADJ
hjic-847	82	27	-	-	ADJ
hjic-847	82	28	linearly	linearly	ADV
hjic-847	82	29	separable	separable	ADJ
hjic-847	82	30	or	or	CCONJ
hjic-847	82	31	the	the	DET
hjic-847	82	32	data	data	NOUN
hjic-847	82	33	contains	contain	VERB
hjic-847	82	34	arbitrarily	arbitrarily	ADV
hjic-847	82	35	shaped	shape	VERB
hjic-847	82	36	clusters	cluster	NOUN
hjic-847	82	37	of	of	ADP
hjic-847	82	38	different	different	ADJ
hjic-847	82	39	densities	density	NOUN
hjic-847	82	40	.	.	PUNCT
hjic-847	83	1	therefore	therefore	ADV
hjic-847	83	2	,	,	PUNCT
hjic-847	83	3	an	an	DET
hjic-847	83	4	upgraded	upgrade	VERB
hjic-847	83	5	version	version	NOUN
hjic-847	83	6	of	of	ADP
hjic-847	83	7	the	the	DET
hjic-847	83	8	k	k	NOUN
hjic-847	83	9	-	-	PUNCT
hjic-847	83	10	means	mean	VERB
hjic-847	83	11	clustering	clustering	ADJ
hjic-847	83	12	algorithm	algorithm	NOUN
hjic-847	83	13	was	be	AUX
hjic-847	83	14	applied	apply	VERB
hjic-847	83	15	in	in	ADP
hjic-847	83	16	the	the	DET
hjic-847	83	17	recognition	recognition	NOUN
hjic-847	83	18	system	system	NOUN
hjic-847	83	19	referred	refer	VERB
hjic-847	83	20	to	to	ADP
hjic-847	83	21	as	as	SCONJ
hjic-847	83	22	kernel	kernel	PROPN
hjic-847	83	23	k	k	X
hjic-847	83	24	-	-	PUNCT
hjic-847	83	25	means	mean	VERB
hjic-847	83	26	[	[	X
hjic-847	83	27	20–22	20–22	NUM
hjic-847	83	28	]	]	X
hjic-847	83	29	.	.	PUNCT
hjic-847	84	1	the	the	DET
hjic-847	84	2	objective	objective	ADJ
hjic-847	84	3	function	function	NOUN
hjic-847	84	4	of	of	ADP
hjic-847	84	5	kernel	kernel	PROPN
hjic-847	84	6	k	k	X
hjic-847	84	7	-	-	PUNCT
hjic-847	84	8	means	means	NOUN
hjic-847	84	9	is	be	AUX
hjic-847	84	10	still	still	ADV
hjic-847	84	11	to	to	PART
hjic-847	84	12	minimize	minimize	VERB
hjic-847	84	13	the	the	DET
hjic-847	84	14	sum	sum	NOUN
hjic-847	84	15	of	of	ADP
hjic-847	84	16	squared	squared	ADJ
hjic-847	84	17	distances	distance	NOUN
hjic-847	84	18	,	,	PUNCT
hjic-847	84	19	but	but	CCONJ
hjic-847	84	20	it	it	PRON
hjic-847	84	21	uses	use	VERB
hjic-847	84	22	the	the	DET
hjic-847	84	23	kernel	kernel	NOUN
hjic-847	84	24	trick	trick	NOUN
hjic-847	84	25	to	to	PART
hjic-847	84	26	transform	transform	VERB
hjic-847	84	27	the	the	DET
hjic-847	84	28	data	data	NOUN
hjic-847	84	29	points	point	NOUN
hjic-847	84	30	into	into	ADP
hjic-847	84	31	infinite	infinite	ADJ
hjic-847	84	32	feature	feature	NOUN
hjic-847	84	33	space	space	NOUN
hjic-847	84	34	xi	xi	INTJ
hjic-847	84	35	→	→	SYM
hjic-847	84	36	ϑ	ϑ	X
hjic-847	84	37	(	(	PUNCT
hjic-847	84	38	xi	xi	PROPN
hjic-847	84	39	)	)	PUNCT
hjic-847	84	40	,	,	PUNCT
hjic-847	84	41	as	as	SCONJ
hjic-847	84	42	can	can	AUX
hjic-847	84	43	be	be	AUX
hjic-847	84	44	seen	see	VERB
hjic-847	84	45	in	in	ADP
hjic-847	84	46	e	e	PROPN
hjic-847	84	47	=	=	SYM
hjic-847	84	48	min	min	PROPN
hjic-847	84	49			X
hjic-847	84	50	k∑	k∑	VERB
hjic-847	85	1	l=1	l=1	X
hjic-847	85	2	∑	∑	PUNCT
hjic-847	85	3	xi∈cl	xi∈cl	PUNCT
hjic-847	86	1	∥∥∥∥∥∥∥ϑ	∥∥∥∥∥∥∥ϑ	PUNCT
hjic-847	86	2	(	(	PUNCT
hjic-847	86	3	xi)−	xi)−	PUNCT
hjic-847	86	4	∑	∑	PUNCT
hjic-847	86	5	xj∈cl	xj∈cl	X
hjic-847	86	6	ϑ	ϑ	X
hjic-847	86	7	(	(	PUNCT
hjic-847	86	8	xj	xj	PROPN
hjic-847	86	9	)	)	PUNCT
hjic-847	86	10	nl	nl	PROPN
hjic-847	86	11	∥∥∥∥∥∥∥	∥∥∥∥∥∥∥	PROPN
hjic-847	86	12	2	2	PROPN
hjic-847	86	13	,	,	PUNCT
hjic-847	86	14	(	(	PUNCT
hjic-847	86	15	7	7	X
hjic-847	86	16	)	)	PUNCT
hjic-847	86	17	where	where	SCONJ
hjic-847	86	18	nl	nl	PROPN
hjic-847	86	19	denotes	denote	VERB
hjic-847	86	20	the	the	DET
hjic-847	86	21	number	number	NOUN
hjic-847	86	22	of	of	ADP
hjic-847	86	23	images	image	NOUN
hjic-847	86	24	in	in	ADP
hjic-847	86	25	cluster	cluster	NOUN
hjic-847	86	26	cl	cl	NOUN
hjic-847	86	27	.	.	PUNCT
hjic-847	87	1	the	the	DET
hjic-847	87	2	trick	trick	NOUN
hjic-847	87	3	here	here	ADV
hjic-847	87	4	is	be	AUX
hjic-847	87	5	that	that	DET
hjic-847	87	6	explicit	explicit	ADJ
hjic-847	87	7	calculations	calculation	NOUN
hjic-847	87	8	in	in	ADP
hjic-847	87	9	the	the	DET
hjic-847	87	10	feature	feature	NOUN
hjic-847	87	11	space	space	NOUN
hjic-847	87	12	are	be	AUX
hjic-847	87	13	never	never	ADV
hjic-847	87	14	required	require	VERB
hjic-847	87	15	,	,	PUNCT
hjic-847	87	16	since	since	SCONJ
hjic-847	87	17	transformed	transform	VERB
hjic-847	87	18	data	data	NOUN
hjic-847	87	19	points	point	NOUN
hjic-847	87	20	are	be	AUX
hjic-847	87	21	only	only	ADV
hjic-847	87	22	present	present	ADJ
hjic-847	87	23	as	as	ADP
hjic-847	87	24	part	part	NOUN
hjic-847	87	25	of	of	ADP
hjic-847	87	26	an	an	DET
hjic-847	87	27	inner	inner	ADJ
hjic-847	87	28	product	product	NOUN
hjic-847	87	29	.	.	PUNCT
hjic-847	88	1	therefore	therefore	ADV
hjic-847	88	2	,	,	PUNCT
hjic-847	88	3	they	they	PRON
hjic-847	88	4	can	can	AUX
hjic-847	88	5	be	be	AUX
hjic-847	88	6	substituted	substitute	VERB
hjic-847	88	7	for	for	ADP
hjic-847	88	8	their	their	PRON
hjic-847	88	9	kernel	kernel	NOUN
hjic-847	88	10	representatives	representative	NOUN
hjic-847	88	11	(	(	PUNCT
hjic-847	88	12	the	the	DET
hjic-847	88	13	gaussian	gaussian	ADJ
hjic-847	88	14	kernel	kernel	PROPN
hjic-847	88	15	was	be	AUX
hjic-847	88	16	implemented	implement	VERB
hjic-847	88	17	here	here	ADV
hjic-847	88	18	)	)	PUNCT
hjic-847	88	19	.	.	PUNCT
hjic-847	89	1	in	in	ADP
hjic-847	89	2	order	order	NOUN
hjic-847	89	3	to	to	PART
hjic-847	89	4	reduce	reduce	VERB
hjic-847	89	5	the	the	DET
hjic-847	89	6	randomness	randomness	NOUN
hjic-847	89	7	of	of	ADP
hjic-847	89	8	final	final	ADJ
hjic-847	89	9	clusters	cluster	NOUN
hjic-847	89	10	,	,	PUNCT
hjic-847	89	11	the	the	DET
hjic-847	89	12	plusplus	plusplus	ADJ
hjic-847	89	13	cluster	cluster	NOUN
hjic-847	89	14	center	center	PROPN
hjic-847	89	15	initialization	initialization	NOUN
hjic-847	89	16	algorithm	algorithm	NOUN
hjic-847	89	17	was	be	AUX
hjic-847	89	18	used	use	VERB
hjic-847	89	19	before	before	ADP
hjic-847	89	20	the	the	DET
hjic-847	89	21	iterative	iterative	NOUN
hjic-847	89	22	steps	step	NOUN
hjic-847	89	23	,	,	PUNCT
hjic-847	89	24	which	which	PRON
hjic-847	89	25	was	be	AUX
hjic-847	89	26	proposed	propose	VERB
hjic-847	89	27	by	by	ADP
hjic-847	89	28	d.	d.	PROPN
hjic-847	89	29	arthur	arthur	PROPN
hjic-847	89	30	and	and	CCONJ
hjic-847	89	31	s.	s.	PROPN
hjic-847	89	32	vassilvitskii	vassilvitskii	PROPN
hjic-847	90	1	[	[	X
hjic-847	90	2	23	23	NUM
hjic-847	90	3	]	]	PUNCT
hjic-847	90	4	.	.	PUNCT
hjic-847	91	1	this	this	DET
hjic-847	91	2	approach	approach	NOUN
hjic-847	91	3	aims	aim	VERB
hjic-847	91	4	to	to	PART
hjic-847	91	5	spread	spread	VERB
hjic-847	91	6	out	out	ADP
hjic-847	91	7	the	the	DET
hjic-847	91	8	initial	initial	ADJ
hjic-847	91	9	cluster	cluster	NOUN
hjic-847	91	10	centers	center	NOUN
hjic-847	91	11	and	and	CCONJ
hjic-847	91	12	accelerate	accelerate	VERB
hjic-847	91	13	their	their	PRON
hjic-847	91	14	convergence	convergence	NOUN
hjic-847	91	15	.	.	PUNCT
hjic-847	92	1	the	the	DET
hjic-847	92	2	first	first	ADJ
hjic-847	92	3	cluster	cluster	NOUN
hjic-847	92	4	center	center	NOUN
hjic-847	92	5	is	be	AUX
hjic-847	92	6	randomly	randomly	ADV
hjic-847	92	7	selected	select	VERB
hjic-847	92	8	from	from	ADP
hjic-847	92	9	the	the	DET
hjic-847	92	10	data	data	NOUN
hjic-847	92	11	points	point	NOUN
hjic-847	92	12	,	,	PUNCT
hjic-847	92	13	after	after	ADP
hjic-847	92	14	that	that	PRON
hjic-847	92	15	each	each	DET
hjic-847	92	16	subsequent	subsequent	ADJ
hjic-847	92	17	cluster	cluster	NOUN
hjic-847	92	18	center	center	NOUN
hjic-847	92	19	is	be	AUX
hjic-847	92	20	chosen	choose	VERB
hjic-847	92	21	from	from	ADP
hjic-847	92	22	the	the	DET
hjic-847	92	23	data	data	NOUN
hjic-847	92	24	points	point	NOUN
hjic-847	92	25	with	with	ADP
hjic-847	92	26	a	a	DET
hjic-847	92	27	probability	probability	NOUN
hjic-847	92	28	proportional	proportional	ADJ
hjic-847	92	29	to	to	ADP
hjic-847	92	30	its	its	PRON
hjic-847	92	31	squared	square	VERB
hjic-847	92	32	distance	distance	NOUN
hjic-847	92	33	from	from	ADP
hjic-847	92	34	the	the	DET
hjic-847	92	35	closest	close	ADJ
hjic-847	92	36	existing	exist	VERB
hjic-847	92	37	cluster	cluster	NOUN
hjic-847	92	38	center	center	NOUN
hjic-847	92	39	.	.	PUNCT
hjic-847	93	1	in	in	ADP
hjic-847	93	2	the	the	DET
hjic-847	93	3	following	follow	VERB
hjic-847	93	4	sub	sub	NOUN
hjic-847	93	5	-	-	NOUN
hjic-847	93	6	sections	section	NOUN
hjic-847	93	7	,	,	PUNCT
hjic-847	93	8	the	the	DET
hjic-847	93	9	usage	usage	NOUN
hjic-847	93	10	of	of	ADP
hjic-847	93	11	the	the	DET
hjic-847	93	12	presented	present	VERB
hjic-847	93	13	methods	method	NOUN
hjic-847	93	14	is	be	AUX
hjic-847	93	15	discussed	discuss	VERB
hjic-847	93	16	.	.	PUNCT
hjic-847	94	1	2.3	2.3	NUM
hjic-847	94	2	baseline	baseline	NOUN
hjic-847	94	3	method	method	NOUN
hjic-847	94	4	in	in	ADP
hjic-847	94	5	this	this	DET
hjic-847	94	6	section	section	NOUN
hjic-847	94	7	,	,	PUNCT
hjic-847	94	8	a	a	DET
hjic-847	94	9	baseline	baseline	NOUN
hjic-847	94	10	method	method	NOUN
hjic-847	94	11	of	of	ADP
hjic-847	94	12	open	open	ADJ
hjic-847	94	13	world	world	NOUN
hjic-847	94	14	recognition	recognition	NOUN
hjic-847	94	15	is	be	AUX
hjic-847	94	16	presented	present	VERB
hjic-847	94	17	.	.	PUNCT
hjic-847	95	1	first	first	ADV
hjic-847	95	2	,	,	PUNCT
hjic-847	95	3	at	at	ADP
hjic-847	95	4	training	training	NOUN
hjic-847	95	5	time	time	NOUN
hjic-847	95	6	the	the	DET
hjic-847	95	7	classifier	classifier	NOUN
hjic-847	95	8	of	of	ADP
hjic-847	95	9	the	the	DET
hjic-847	95	10	training	training	NOUN
hjic-847	95	11	data	datum	NOUN
hjic-847	95	12	is	be	AUX
hjic-847	95	13	trained	train	VERB
hjic-847	95	14	with	with	ADP
hjic-847	95	15	k	k	PROPN
hjic-847	95	16	known	know	VERB
hjic-847	95	17	classes	class	NOUN
hjic-847	95	18	,	,	PUNCT
hjic-847	95	19	then	then	ADV
hjic-847	95	20	,	,	PUNCT
hjic-847	95	21	at	at	ADP
hjic-847	95	22	testing	testing	NOUN
hjic-847	95	23	time	time	NOUN
hjic-847	95	24	classification	classification	NOUN
hjic-847	95	25	of	of	ADP
hjic-847	95	26	the	the	DET
hjic-847	95	27	test	test	NOUN
hjic-847	95	28	data	datum	NOUN
hjic-847	95	29	(	(	PUNCT
hjic-847	95	30	k+u	k+u	PROPN
hjic-847	95	31	classes	class	NOUN
hjic-847	95	32	)	)	PUNCT
hjic-847	95	33	is	be	AUX
hjic-847	95	34	performed	perform	VERB
hjic-847	95	35	.	.	PUNCT
hjic-847	96	1	the	the	DET
hjic-847	96	2	dpm	dpm	PROPN
hjic-847	96	3	is	be	AUX
hjic-847	96	4	applied	apply	VERB
hjic-847	96	5	to	to	ADP
hjic-847	96	6	the	the	DET
hjic-847	96	7	output	output	NOUN
hjic-847	96	8	of	of	ADP
hjic-847	96	9	the	the	DET
hjic-847	96	10	classifier	classifier	NOUN
hjic-847	96	11	to	to	PART
hjic-847	96	12	detect	detect	VERB
hjic-847	96	13	unknown	unknown	ADJ
hjic-847	96	14	images	image	NOUN
hjic-847	96	15	udpm	udpm	NOUN
hjic-847	96	16	:	:	PUNCT
hjic-847	96	17	udpm	udpm	PROPN
hjic-847	96	18	=	=	PUNCT
hjic-847	96	19	nu⋃	nu⋃	PROPN
hjic-847	96	20	j=1	j=1	PROPN
hjic-847	96	21	{	{	PUNCT
hjic-847	96	22	ij	ij	INTJ
hjic-847	96	23	|dj	|dj	NOUN
hjic-847	96	24	=	=	PUNCT
hjic-847	96	25	ck+1	ck+1	X
hjic-847	96	26	}	}	PUNCT
hjic-847	96	27	(	(	PUNCT
hjic-847	96	28	8)	8)	NUM
hjic-847	96	29	where	where	SCONJ
hjic-847	96	30	ij	ij	NOUN
hjic-847	96	31	represents	represent	VERB
hjic-847	96	32	test	test	NOUN
hjic-847	96	33	instance	instance	PROPN
hjic-847	96	34	j	j	PROPN
hjic-847	96	35	,	,	PUNCT
hjic-847	96	36	nu	nu	PROPN
hjic-847	96	37	denotes	denote	VERB
hjic-847	96	38	the	the	DET
hjic-847	96	39	number	number	NOUN
hjic-847	96	40	of	of	ADP
hjic-847	96	41	test	test	NOUN
hjic-847	96	42	instances	instance	NOUN
hjic-847	96	43	in	in	ADP
hjic-847	96	44	the	the	DET
hjic-847	96	45	test	test	NOUN
hjic-847	96	46	data	datum	NOUN
hjic-847	96	47	,	,	PUNCT
hjic-847	96	48	dj	dj	NOUN
hjic-847	96	49	stands	stand	VERB
hjic-847	96	50	for	for	ADP
hjic-847	96	51	the	the	DET
hjic-847	96	52	decision	decision	NOUN
hjic-847	96	53	of	of	ADP
hjic-847	96	54	the	the	DET
hjic-847	96	55	dpm	dpm	PROPN
hjic-847	96	56	,	,	PUNCT
hjic-847	96	57	and	and	CCONJ
hjic-847	96	58	⋃	⋃	PROPN
hjic-847	96	59	{	{	PUNCT
hjic-847	96	60	.	.	PUNCT
hjic-847	96	61	.	.	PUNCT
hjic-847	96	62	.	.	PUNCT
hjic-847	97	1	}	}	PUNCT
hjic-847	97	2	is	be	AUX
hjic-847	97	3	the	the	DET
hjic-847	97	4	operation	operation	NOUN
hjic-847	97	5	of	of	ADP
hjic-847	97	6	union	union	NOUN
hjic-847	97	7	.	.	PUNCT
hjic-847	98	1	now	now	ADV
hjic-847	98	2	,	,	PUNCT
hjic-847	98	3	let	let	VERB
hjic-847	98	4	us	we	PRON
hjic-847	98	5	assume	assume	VERB
hjic-847	98	6	that	that	SCONJ
hjic-847	98	7	information	information	NOUN
hjic-847	98	8	concerning	concern	VERB
hjic-847	98	9	u	u	NOUN
hjic-847	98	10	was	be	AUX
hjic-847	98	11	provided	provide	VERB
hjic-847	98	12	(	(	PUNCT
hjic-847	98	13	as	as	ADP
hjic-847	98	14	in	in	ADP
hjic-847	98	15	the	the	DET
hjic-847	98	16	case	case	NOUN
hjic-847	98	17	3c	3c	NUM
hjic-847	98	18	)	)	PUNCT
hjic-847	98	19	,	,	PUNCT
hjic-847	98	20	and	and	CCONJ
hjic-847	98	21	u	u	NOUN
hjic-847	98	22	was	be	AUX
hjic-847	98	23	used	use	VERB
hjic-847	98	24	as	as	ADP
hjic-847	98	25	the	the	DET
hjic-847	98	26	number	number	NOUN
hjic-847	98	27	of	of	ADP
hjic-847	98	28	clusters	cluster	NOUN
hjic-847	98	29	.	.	PUNCT
hjic-847	99	1	the	the	DET
hjic-847	99	2	kernel	kernel	PROPN
hjic-847	99	3	k	k	PROPN
hjic-847	99	4	-	-	PUNCT
hjic-847	99	5	means	means	PROPN
hjic-847	99	6	plusplus	plusplus	NOUN
hjic-847	99	7	cluster	cluster	NOUN
hjic-847	99	8	center	center	NOUN
hjic-847	99	9	initialization	initialization	NOUN
hjic-847	99	10	algorithm	algorithm	NOUN
hjic-847	99	11	(	(	PUNCT
hjic-847	99	12	kk++	kk++	NOUN
hjic-847	99	13	)	)	PUNCT
hjic-847	99	14	was	be	AUX
hjic-847	99	15	performed	perform	VERB
hjic-847	99	16	on	on	ADP
hjic-847	99	17	udpm	udpm	PROPN
hjic-847	99	18	with	with	ADP
hjic-847	99	19	k	k	PROPN
hjic-847	99	20	=	=	PUNCT
hjic-847	99	21	u	u	PROPN
hjic-847	99	22	clusters	cluster	NOUN
hjic-847	99	23	(	(	PUNCT
hjic-847	99	24	which	which	PRON
hjic-847	99	25	is	be	AUX
hjic-847	99	26	the	the	DET
hjic-847	99	27	input	input	NOUN
hjic-847	99	28	parameter	parameter	NOUN
hjic-847	99	29	for	for	ADP
hjic-847	99	30	the	the	DET
hjic-847	99	31	kk++	kk++	NOUN
hjic-847	99	32	)	)	PUNCT
hjic-847	99	33	,	,	PUNCT
hjic-847	99	34	and	and	CCONJ
hjic-847	99	35	then	then	ADV
hjic-847	99	36	the	the	DET
hjic-847	99	37	appropriate	appropriate	ADJ
hjic-847	99	38	labels	label	NOUN
hjic-847	99	39	were	be	AUX
hjic-847	99	40	assigned	assign	VERB
hjic-847	99	41	to	to	ADP
hjic-847	99	42	the	the	DET
hjic-847	99	43	unknown	unknown	ADJ
hjic-847	99	44	images	image	NOUN
hjic-847	99	45	:	:	PUNCT
hjic-847	100	1	lj	lj	ADP
hjic-847	100	2	=	=	PUNCT
hjic-847	100	3	ck+i|i	ck+i|i	NOUN
hjic-847	100	4	=	=	PUNCT
hjic-847	100	5	out	out	ADP
hjic-847	100	6	(	(	PUNCT
hjic-847	100	7	kk++	kk++	NOUN
hjic-847	100	8	)	)	PUNCT
hjic-847	100	9	j	j	PROPN
hjic-847	101	1	=	=	NOUN
hjic-847	101	2	1	1	NUM
hjic-847	101	3	.	.	PUNCT
hjic-847	101	4	.	.	PUNCT
hjic-847	102	1	.m	.m	PROPN
hjic-847	102	2	,	,	PUNCT
hjic-847	102	3	i	i	PRON
hjic-847	102	4	=	=	NOUN
hjic-847	102	5	1	1	X
hjic-847	102	6	.	.	PUNCT
hjic-847	102	7	.	.	PUNCT
hjic-847	102	8	.	.	PUNCT
hjic-847	103	1	u	u	NOUN
hjic-847	103	2	(	(	PUNCT
hjic-847	103	3	9	9	NUM
hjic-847	103	4	)	)	PUNCT
hjic-847	103	5	where	where	SCONJ
hjic-847	103	6	m	m	NOUN
hjic-847	103	7	represents	represent	VERB
hjic-847	103	8	the	the	DET
hjic-847	103	9	number	number	NOUN
hjic-847	103	10	of	of	ADP
hjic-847	103	11	unknown	unknown	ADJ
hjic-847	103	12	images	image	NOUN
hjic-847	103	13	;	;	PUNCT
hjic-847	103	14	lj	lj	PROPN
hjic-847	103	15	and	and	CCONJ
hjic-847	103	16	ci	ci	PROPN
hjic-847	103	17	denote	denote	VERB
hjic-847	103	18	the	the	DET
hjic-847	103	19	label	label	NOUN
hjic-847	103	20	of	of	ADP
hjic-847	103	21	unknown	unknown	ADJ
hjic-847	103	22	image	image	NOUN
hjic-847	103	23	udpm	udpm	PROPN
hjic-847	103	24	j	j	PROPN
hjic-847	103	25	and	and	CCONJ
hjic-847	103	26	cluster	cluster	NOUN
hjic-847	103	27	identity	identity	NOUN
hjic-847	103	28	,	,	PUNCT
hjic-847	103	29	respectively	respectively	ADV
hjic-847	103	30	.	.	PUNCT
hjic-847	104	1	this	this	PRON
hjic-847	104	2	concludes	conclude	VERB
hjic-847	104	3	the	the	DET
hjic-847	104	4	baseline	baseline	NOUN
hjic-847	104	5	method	method	NOUN
hjic-847	104	6	for	for	ADP
hjic-847	104	7	automated	automate	VERB
hjic-847	104	8	labeling	labeling	NOUN
hjic-847	104	9	.	.	PUNCT
hjic-847	105	1	at	at	ADP
hjic-847	105	2	this	this	DET
hjic-847	105	3	point	point	NOUN
hjic-847	105	4	the	the	DET
hjic-847	105	5	classifier	classifier	NOUN
hjic-847	105	6	can	can	AUX
hjic-847	105	7	be	be	AUX
hjic-847	105	8	retrained	retrain	VERB
hjic-847	105	9	based	base	VERB
hjic-847	105	10	on	on	ADP
hjic-847	105	11	the	the	DET
hjic-847	105	12	previously	previously	ADV
hjic-847	105	13	known	know	VERB
hjic-847	105	14	and	and	CCONJ
hjic-847	105	15	new	new	ADJ
hjic-847	105	16	labels	label	NOUN
hjic-847	105	17	,	,	PUNCT
hjic-847	105	18	and	and	CCONJ
hjic-847	105	19	then	then	ADV
hjic-847	105	20	the	the	DET
hjic-847	105	21	new	new	ADJ
hjic-847	105	22	test	test	NOUN
hjic-847	105	23	data	datum	NOUN
hjic-847	105	24	classified	classify	VERB
hjic-847	105	25	.	.	PUNCT
hjic-847	106	1	2.4	2.4	NUM
hjic-847	106	2	cluster	cluster	NOUN
hjic-847	106	3	classification	classification	NOUN
hjic-847	106	4	in	in	ADP
hjic-847	106	5	this	this	DET
hjic-847	106	6	section	section	NOUN
hjic-847	106	7	our	our	PRON
hjic-847	106	8	proposed	propose	VERB
hjic-847	106	9	cc	cc	NOUN
hjic-847	106	10	approach	approach	NOUN
hjic-847	106	11	is	be	AUX
hjic-847	106	12	presented	present	VERB
hjic-847	106	13	,	,	PUNCT
hjic-847	106	14	which	which	PRON
hjic-847	106	15	is	be	AUX
hjic-847	106	16	suitable	suitable	ADJ
hjic-847	106	17	for	for	ADP
hjic-847	106	18	unknown	unknown	ADJ
hjic-847	106	19	detection	detection	NOUN
hjic-847	106	20	and	and	CCONJ
hjic-847	106	21	automated	automate	VERB
hjic-847	106	22	labeling	labeling	NOUN
hjic-847	106	23	.	.	PUNCT
hjic-847	107	1	this	this	DET
hjic-847	107	2	algorithm	algorithm	NOUN
hjic-847	107	3	contains	contain	VERB
hjic-847	107	4	extended	extended	ADJ
hjic-847	107	5	training	training	NOUN
hjic-847	107	6	and	and	CCONJ
hjic-847	107	7	testing	testing	NOUN
hjic-847	107	8	phases	phase	NOUN
hjic-847	107	9	.	.	PUNCT
hjic-847	108	1	in	in	ADP
hjic-847	108	2	training	training	NOUN
hjic-847	108	3	time	time	NOUN
hjic-847	108	4	,	,	PUNCT
hjic-847	108	5	a	a	DET
hjic-847	108	6	classifier	classifier	NOUN
hjic-847	108	7	of	of	ADP
hjic-847	108	8	the	the	DET
hjic-847	108	9	training	training	NOUN
hjic-847	108	10	data	datum	NOUN
hjic-847	108	11	is	be	AUX
hjic-847	108	12	trained	train	VERB
hjic-847	108	13	with	with	ADP
hjic-847	108	14	k	k	PROPN
hjic-847	108	15	known	know	VERB
hjic-847	108	16	classes	class	NOUN
hjic-847	108	17	,	,	PUNCT
hjic-847	108	18	then	then	ADV
hjic-847	108	19	a	a	DET
hjic-847	108	20	pseudocluster	pseudocluster	NOUN
hjic-847	108	21	is	be	AUX
hjic-847	108	22	also	also	ADV
hjic-847	108	23	created	create	VERB
hjic-847	108	24	based	base	VERB
hjic-847	108	25	on	on	ADP
hjic-847	108	26	the	the	DET
hjic-847	108	27	k	k	PROPN
hjic-847	108	28	known	know	VERB
hjic-847	108	29	categories	category	NOUN
hjic-847	108	30	.	.	PUNCT
hjic-847	109	1	this	this	PRON
hjic-847	109	2	means	mean	VERB
hjic-847	109	3	that	that	SCONJ
hjic-847	109	4	the	the	DET
hjic-847	109	5	ground	ground	NOUN
hjic-847	109	6	truth	truth	NOUN
hjic-847	109	7	class	class	NOUN
hjic-847	109	8	labels	label	NOUN
hjic-847	109	9	are	be	AUX
hjic-847	109	10	implemented	implement	VERB
hjic-847	109	11	rather	rather	ADV
hjic-847	109	12	than	than	ADP
hjic-847	109	13	a	a	DET
hjic-847	109	14	clustering	cluster	VERB
hjic-847	109	15	algorithm	algorithm	NOUN
hjic-847	109	16	(	(	PUNCT
hjic-847	109	17	to	to	PART
hjic-847	109	18	determine	determine	VERB
hjic-847	109	19	the	the	DET
hjic-847	109	20	final	final	ADJ
hjic-847	109	21	clusters	cluster	NOUN
hjic-847	109	22	)	)	PUNCT
hjic-847	109	23	,	,	PUNCT
hjic-847	109	24	i.e.	i.e.	X
hjic-847	109	25	each	each	DET
hjic-847	109	26	category	category	NOUN
hjic-847	109	27	is	be	AUX
hjic-847	109	28	a	a	DET
hjic-847	109	29	cluster	cluster	NOUN
hjic-847	109	30	.	.	PUNCT
hjic-847	110	1	subsequently	subsequently	ADV
hjic-847	110	2	,	,	PUNCT
hjic-847	110	3	the	the	DET
hjic-847	110	4	images	image	NOUN
hjic-847	110	5	are	be	AUX
hjic-847	110	6	substituted	substitute	VERB
hjic-847	110	7	for	for	ADP
hjic-847	110	8	their	their	PRON
hjic-847	110	9	fisher	fisher	PROPN
hjic-847	110	10	vector	vector	NOUN
hjic-847	110	11	47(1	47(1	PROPN
hjic-847	110	12	)	)	PUNCT
hjic-847	110	13	pp	pp	ADV
hjic-847	110	14	.	.	PUNCT
hjic-847	111	1	33–39	33–39	NUM
hjic-847	111	2	(	(	PUNCT
hjic-847	111	3	2019	2019	NUM
hjic-847	111	4	)	)	PUNCT
hjic-847	111	5	36	36	NUM
hjic-847	111	6	papp	papp	NOUN
hjic-847	111	7	and	and	CCONJ
hjic-847	111	8	szűcs	szűcs	ADJ
hjic-847	111	9	representations	representation	NOUN
hjic-847	111	10	and	and	CCONJ
hjic-847	111	11	the	the	DET
hjic-847	111	12	cluster	cluster	NOUN
hjic-847	111	13	centers	center	NOUN
hjic-847	111	14	calculated	calculate	VERB
hjic-847	111	15	which	which	PRON
hjic-847	111	16	will	will	AUX
hjic-847	111	17	be	be	AUX
hjic-847	111	18	used	use	VERB
hjic-847	111	19	in	in	ADP
hjic-847	111	20	the	the	DET
hjic-847	111	21	testing	testing	NOUN
hjic-847	111	22	phase	phase	NOUN
hjic-847	111	23	.	.	PUNCT
hjic-847	112	1	let	let	VERB
hjic-847	112	2	us	we	PRON
hjic-847	112	3	assume	assume	VERB
hjic-847	112	4	t	t	PROPN
hjic-847	112	5	categories	category	NOUN
hjic-847	112	6	are	be	AUX
hjic-847	112	7	found	find	VERB
hjic-847	112	8	in	in	ADP
hjic-847	112	9	the	the	DET
hjic-847	112	10	testing	testing	NOUN
hjic-847	112	11	phase	phase	NOUN
hjic-847	112	12	,	,	PUNCT
hjic-847	112	13	and	and	CCONJ
hjic-847	112	14	that	that	SCONJ
hjic-847	112	15	t	t	PROPN
hjic-847	112	16	>	>	X
hjic-847	112	17	k.	k.	PROPN
hjic-847	113	1	the	the	DET
hjic-847	113	2	test	test	NOUN
hjic-847	113	3	data	datum	NOUN
hjic-847	113	4	is	be	AUX
hjic-847	113	5	classified	classify	VERB
hjic-847	113	6	into	into	ADP
hjic-847	113	7	the	the	DET
hjic-847	113	8	k	k	PROPN
hjic-847	113	9	known	know	VERB
hjic-847	113	10	categories	category	NOUN
hjic-847	113	11	and	and	CCONJ
hjic-847	113	12	a	a	DET
hjic-847	113	13	dpm	dpm	NOUN
hjic-847	113	14	applied	apply	VERB
hjic-847	113	15	based	base	VERB
hjic-847	113	16	on	on	ADP
hjic-847	113	17	the	the	DET
hjic-847	113	18	decision	decision	NOUN
hjic-847	113	19	vectors	vector	NOUN
hjic-847	113	20	to	to	PART
hjic-847	113	21	detect	detect	VERB
hjic-847	113	22	the	the	DET
hjic-847	113	23	unknown	unknown	ADJ
hjic-847	113	24	images	image	NOUN
hjic-847	113	25	udpm	udpm	NOUN
hjic-847	113	26	.	.	PUNCT
hjic-847	114	1	the	the	DET
hjic-847	114	2	next	next	ADJ
hjic-847	114	3	step	step	NOUN
hjic-847	114	4	is	be	AUX
hjic-847	114	5	to	to	PART
hjic-847	114	6	form	form	VERB
hjic-847	114	7	clusters	cluster	NOUN
hjic-847	114	8	using	use	VERB
hjic-847	114	9	the	the	DET
hjic-847	114	10	kernel	kernel	NOUN
hjic-847	114	11	k	k	X
hjic-847	114	12	-	-	PUNCT
hjic-847	114	13	means	mean	VERB
hjic-847	114	14	clustering	cluster	VERB
hjic-847	114	15	algorithm	algorithm	NOUN
hjic-847	114	16	starting	start	VERB
hjic-847	114	17	from	from	ADP
hjic-847	114	18	the	the	DET
hjic-847	114	19	k	k	PROPN
hjic-847	114	20	cluster	cluster	NOUN
hjic-847	114	21	centers	center	NOUN
hjic-847	114	22	that	that	PRON
hjic-847	114	23	were	be	AUX
hjic-847	114	24	calculated	calculate	VERB
hjic-847	114	25	at	at	ADP
hjic-847	114	26	training	training	NOUN
hjic-847	114	27	time	time	NOUN
hjic-847	114	28	from	from	ADP
hjic-847	114	29	the	the	DET
hjic-847	114	30	pseudocluster	pseudocluster	NOUN
hjic-847	114	31	.	.	PUNCT
hjic-847	115	1	afterwards	afterwards	ADV
hjic-847	115	2	,	,	PUNCT
hjic-847	115	3	the	the	DET
hjic-847	115	4	remaining	remain	VERB
hjic-847	115	5	t	t	NOUN
hjic-847	115	6	-k	-k	PUNCT
hjic-847	115	7	cluster	cluster	NOUN
hjic-847	115	8	centers	center	NOUN
hjic-847	115	9	are	be	AUX
hjic-847	115	10	determined	determine	VERB
hjic-847	115	11	following	follow	VERB
hjic-847	115	12	the	the	DET
hjic-847	115	13	plusplus	plusplus	ADJ
hjic-847	115	14	initiation	initiation	NOUN
hjic-847	115	15	protocol	protocol	NOUN
hjic-847	115	16	.	.	PUNCT
hjic-847	116	1	furthermore	furthermore	ADV
hjic-847	116	2	,	,	PUNCT
hjic-847	116	3	the	the	DET
hjic-847	116	4	training	training	NOUN
hjic-847	116	5	and	and	CCONJ
hjic-847	116	6	test	test	NOUN
hjic-847	116	7	datasets	dataset	NOUN
hjic-847	116	8	were	be	AUX
hjic-847	116	9	used	use	VERB
hjic-847	116	10	together	together	ADV
hjic-847	116	11	as	as	ADP
hjic-847	116	12	the	the	DET
hjic-847	116	13	input	input	NOUN
hjic-847	116	14	data	datum	NOUN
hjic-847	116	15	.	.	PUNCT
hjic-847	117	1	basically	basically	ADV
hjic-847	117	2	,	,	PUNCT
hjic-847	117	3	with	with	ADP
hjic-847	117	4	these	these	DET
hjic-847	117	5	modifications	modification	NOUN
hjic-847	117	6	it	it	PRON
hjic-847	117	7	was	be	AUX
hjic-847	117	8	possible	possible	ADJ
hjic-847	117	9	to	to	PART
hjic-847	117	10	guide	guide	VERB
hjic-847	117	11	the	the	DET
hjic-847	117	12	clustering	clustering	ADJ
hjic-847	117	13	algorithm	algorithm	NOUN
hjic-847	117	14	,	,	PUNCT
hjic-847	117	15	therefore	therefore	ADV
hjic-847	117	16	,	,	PUNCT
hjic-847	117	17	create	create	VERB
hjic-847	117	18	more	more	ADV
hjic-847	117	19	accurate	accurate	ADJ
hjic-847	117	20	clusters	cluster	NOUN
hjic-847	117	21	.	.	PUNCT
hjic-847	118	1	the	the	DET
hjic-847	118	2	following	follow	VERB
hjic-847	118	3	step	step	NOUN
hjic-847	118	4	of	of	ADP
hjic-847	118	5	the	the	DET
hjic-847	118	6	testing	testing	NOUN
hjic-847	118	7	phase	phase	NOUN
hjic-847	118	8	is	be	AUX
hjic-847	118	9	to	to	PART
hjic-847	118	10	classify	classify	VERB
hjic-847	118	11	the	the	DET
hjic-847	118	12	clusters	cluster	NOUN
hjic-847	118	13	{	{	PUNCT
hjic-847	118	14	ci	ci	NOUN
hjic-847	118	15	}	}	PUNCT
hjic-847	118	16	by	by	ADP
hjic-847	118	17	weighted	weight	VERB
hjic-847	118	18	majority	majority	NOUN
hjic-847	118	19	voting	voting	NOUN
hjic-847	118	20	of	of	ADP
hjic-847	118	21	the	the	DET
hjic-847	118	22	members	member	NOUN
hjic-847	118	23	of	of	ADP
hjic-847	118	24	the	the	DET
hjic-847	118	25	cluster	cluster	NOUN
hjic-847	118	26	.	.	PUNCT
hjic-847	119	1	the	the	DET
hjic-847	119	2	vote	vote	NOUN
hjic-847	119	3	is	be	AUX
hjic-847	119	4	based	base	VERB
hjic-847	119	5	on	on	ADP
hjic-847	119	6	the	the	DET
hjic-847	119	7	class	class	NOUN
hjic-847	119	8	membership	membership	NOUN
hjic-847	119	9	probabilities	probability	NOUN
hjic-847	119	10	(	(	PUNCT
hjic-847	119	11	pci	pci	ADJ
hjic-847	119	12	;	;	PUNCT
hjic-847	119	13	i	i	PRON
hjic-847	119	14	=	=	NOUN
hjic-847	119	15	1	1	X
hjic-847	119	16	.	.	PUNCT
hjic-847	119	17	.	.	PUNCT
hjic-847	120	1	.k	.k	PROPN
hjic-847	121	1	+	+	CCONJ
hjic-847	121	2	1	1	X
hjic-847	121	3	)	)	PUNCT
hjic-847	121	4	calculated	calculate	VERB
hjic-847	121	5	in	in	ADP
hjic-847	121	6	eqs	eqs	PROPN
hjic-847	121	7	.	.	PROPN
hjic-847	122	1	3–4	3–4	NUM
hjic-847	122	2	.	.	PUNCT
hjic-847	122	3	as	as	SCONJ
hjic-847	122	4	was	be	AUX
hjic-847	122	5	seen	see	VERB
hjic-847	122	6	in	in	ADP
hjic-847	122	7	section	section	NOUN
hjic-847	122	8	1	1	NUM
hjic-847	122	9	,	,	PUNCT
hjic-847	122	10	the	the	DET
hjic-847	122	11	definition	definition	NOUN
hjic-847	122	12	of	of	ADP
hjic-847	122	13	problem	problem	NOUN
hjic-847	122	14	3c	3c	NUM
hjic-847	122	15	assumes	assume	VERB
hjic-847	122	16	that	that	SCONJ
hjic-847	122	17	the	the	DET
hjic-847	122	18	number	number	NOUN
hjic-847	122	19	of	of	ADP
hjic-847	122	20	unknown	unknown	ADJ
hjic-847	122	21	categories	category	NOUN
hjic-847	122	22	exceeds	exceed	VERB
hjic-847	122	23	1	1	NUM
hjic-847	122	24	.	.	PUNCT
hjic-847	123	1	nonetheless	nonetheless	ADV
hjic-847	123	2	,	,	PUNCT
hjic-847	123	3	the	the	DET
hjic-847	123	4	output	output	NOUN
hjic-847	123	5	of	of	ADP
hjic-847	123	6	the	the	DET
hjic-847	123	7	dpm	dpm	PROPN
hjic-847	123	8	only	only	ADV
hjic-847	123	9	yields	yield	NOUN
hjic-847	123	10	k	k	PROPN
hjic-847	124	1	+	+	CCONJ
hjic-847	124	2	1	1	NUM
hjic-847	124	3	alternatives	alternative	NOUN
hjic-847	124	4	instead	instead	ADV
hjic-847	124	5	of	of	ADP
hjic-847	124	6	t	t	PROPN
hjic-847	124	7	.	.	PUNCT
hjic-847	125	1	in	in	ADP
hjic-847	125	2	spite	spite	NOUN
hjic-847	125	3	of	of	ADP
hjic-847	125	4	this	this	PRON
hjic-847	125	5	,	,	PUNCT
hjic-847	125	6	the	the	DET
hjic-847	125	7	classification	classification	NOUN
hjic-847	125	8	of	of	ADP
hjic-847	125	9	clusters	cluster	NOUN
hjic-847	125	10	that	that	PRON
hjic-847	125	11	depend	depend	VERB
hjic-847	125	12	on	on	ADP
hjic-847	125	13	{	{	PUNCT
hjic-847	125	14	pci	pci	ADJ
hjic-847	125	15	}	}	PUNCT
hjic-847	125	16	can	can	AUX
hjic-847	125	17	increase	increase	VERB
hjic-847	125	18	the	the	DET
hjic-847	125	19	number	number	NOUN
hjic-847	125	20	of	of	ADP
hjic-847	125	21	alternatives	alternative	NOUN
hjic-847	125	22	to	to	ADP
hjic-847	125	23	t	t	PROPN
hjic-847	125	24	as	as	SCONJ
hjic-847	125	25	will	will	AUX
hjic-847	125	26	be	be	AUX
hjic-847	125	27	seen	see	VERB
hjic-847	125	28	later	later	ADV
hjic-847	125	29	.	.	PUNCT
hjic-847	126	1	in	in	ADP
hjic-847	126	2	section	section	NOUN
hjic-847	126	3	1	1	NUM
hjic-847	126	4	,	,	PUNCT
hjic-847	126	5	a	a	DET
hjic-847	126	6	differentiation	differentiation	NOUN
hjic-847	126	7	was	be	AUX
hjic-847	126	8	made	make	VERB
hjic-847	126	9	between	between	ADP
hjic-847	126	10	known	known	ADJ
hjic-847	126	11	and	and	CCONJ
hjic-847	126	12	unknown	unknown	ADJ
hjic-847	126	13	images	image	NOUN
hjic-847	126	14	,	,	PUNCT
hjic-847	126	15	and	and	CCONJ
hjic-847	126	16	now	now	ADV
hjic-847	126	17	this	this	DET
hjic-847	126	18	differentiation	differentiation	NOUN
hjic-847	126	19	is	be	AUX
hjic-847	126	20	broken	break	VERB
hjic-847	126	21	down	down	ADP
hjic-847	126	22	even	even	ADV
hjic-847	126	23	more	more	ADV
hjic-847	126	24	.	.	PUNCT
hjic-847	127	1	the	the	DET
hjic-847	127	2	training	training	NOUN
hjic-847	127	3	data	data	NOUN
hjic-847	127	4	contains	contain	VERB
hjic-847	127	5	only	only	ADV
hjic-847	127	6	known	know	VERB
hjic-847	127	7	images	image	NOUN
hjic-847	127	8	,	,	PUNCT
hjic-847	127	9	because	because	SCONJ
hjic-847	127	10	each	each	PRON
hjic-847	127	11	of	of	ADP
hjic-847	127	12	them	they	PRON
hjic-847	127	13	belongs	belong	VERB
hjic-847	127	14	to	to	ADP
hjic-847	127	15	one	one	NUM
hjic-847	127	16	of	of	ADP
hjic-847	127	17	the	the	DET
hjic-847	127	18	set	set	NOUN
hjic-847	127	19	of	of	ADP
hjic-847	127	20	known	know	VERB
hjic-847	127	21	categories	category	NOUN
hjic-847	127	22	(	(	PUNCT
hjic-847	127	23	sk	sk	NOUN
hjic-847	127	24	)	)	PUNCT
hjic-847	127	25	.	.	PUNCT
hjic-847	128	1	from	from	ADP
hjic-847	128	2	now	now	ADV
hjic-847	128	3	on	on	ADV
hjic-847	128	4	,	,	PUNCT
hjic-847	128	5	the	the	DET
hjic-847	128	6	union	union	NOUN
hjic-847	128	7	of	of	ADP
hjic-847	128	8	known	know	VERB
hjic-847	128	9	images	image	NOUN
hjic-847	128	10	of	of	ADP
hjic-847	128	11	the	the	DET
hjic-847	128	12	training	training	NOUN
hjic-847	128	13	data	datum	NOUN
hjic-847	128	14	will	will	AUX
hjic-847	128	15	be	be	AUX
hjic-847	128	16	denoted	denote	VERB
hjic-847	128	17	by	by	ADP
hjic-847	128	18	kgt	kgt	PROPN
hjic-847	128	19	as	as	SCONJ
hjic-847	128	20	can	can	AUX
hjic-847	128	21	be	be	AUX
hjic-847	128	22	seen	see	VERB
hjic-847	128	23	in	in	ADP
hjic-847	128	24	:	:	PUNCT
hjic-847	128	25	kgt	kgt	PROPN
hjic-847	128	26	=	=	SYM
hjic-847	128	27	nk⋃	nk⋃	PROPN
hjic-847	128	28	j=1	j=1	PROPN
hjic-847	128	29	{	{	PUNCT
hjic-847	128	30	ij	ij	NOUN
hjic-847	128	31	}	}	PUNCT
hjic-847	128	32	(	(	PUNCT
hjic-847	128	33	10	10	NUM
hjic-847	128	34	)	)	PUNCT
hjic-847	128	35	where	where	SCONJ
hjic-847	128	36	nk	nk	PROPN
hjic-847	128	37	stands	stand	VERB
hjic-847	128	38	for	for	ADP
hjic-847	128	39	the	the	DET
hjic-847	128	40	number	number	NOUN
hjic-847	128	41	of	of	ADP
hjic-847	128	42	images	image	NOUN
hjic-847	128	43	in	in	ADP
hjic-847	128	44	the	the	DET
hjic-847	128	45	training	training	NOUN
hjic-847	128	46	data	datum	NOUN
hjic-847	128	47	.	.	PUNCT
hjic-847	129	1	on	on	ADP
hjic-847	129	2	the	the	DET
hjic-847	129	3	other	other	ADJ
hjic-847	129	4	hand	hand	NOUN
hjic-847	129	5	,	,	PUNCT
hjic-847	129	6	the	the	DET
hjic-847	129	7	test	test	NOUN
hjic-847	129	8	data	data	NOUN
hjic-847	129	9	contains	contain	VERB
hjic-847	129	10	both	both	DET
hjic-847	129	11	known	known	ADJ
hjic-847	129	12	and	and	CCONJ
hjic-847	129	13	unknown	unknown	ADJ
hjic-847	129	14	images	image	NOUN
hjic-847	129	15	.	.	PUNCT
hjic-847	130	1	furthermore	furthermore	ADV
hjic-847	130	2	,	,	PUNCT
hjic-847	130	3	based	base	VERB
hjic-847	130	4	on	on	ADP
hjic-847	130	5	the	the	DET
hjic-847	130	6	output	output	NOUN
hjic-847	130	7	of	of	ADP
hjic-847	130	8	dpm	dpm	PROPN
hjic-847	130	9	,	,	PUNCT
hjic-847	130	10	the	the	DET
hjic-847	130	11	test	test	NOUN
hjic-847	130	12	data	datum	NOUN
hjic-847	130	13	can	can	AUX
hjic-847	130	14	be	be	AUX
hjic-847	130	15	divided	divide	VERB
hjic-847	130	16	into	into	ADP
hjic-847	130	17	two	two	NUM
hjic-847	130	18	different	different	ADJ
hjic-847	130	19	subsets	subset	NOUN
hjic-847	130	20	,	,	PUNCT
hjic-847	130	21	namely	namely	ADV
hjic-847	130	22	predicted	predict	VERB
hjic-847	130	23	known	know	VERB
hjic-847	130	24	images	image	NOUN
hjic-847	130	25	(	(	PUNCT
hjic-847	130	26	kdpm	kdpm	PROPN
hjic-847	130	27	)	)	PUNCT
hjic-847	130	28	and	and	CCONJ
hjic-847	130	29	predicted	predict	VERB
hjic-847	130	30	unknown	unknown	ADJ
hjic-847	130	31	images	image	NOUN
hjic-847	130	32	(	(	PUNCT
hjic-847	130	33	udpm	udpm	PROPN
hjic-847	130	34	)	)	PUNCT
hjic-847	130	35	,	,	PUNCT
hjic-847	130	36	as	as	SCONJ
hjic-847	130	37	can	can	AUX
hjic-847	130	38	be	be	AUX
hjic-847	130	39	seen	see	VERB
hjic-847	130	40	in	in	ADP
hjic-847	130	41	eqs	eqs	PROPN
hjic-847	130	42	.	.	PROPN
hjic-847	130	43	11	11	NUM
hjic-847	130	44	and	and	CCONJ
hjic-847	130	45	8	8	NUM
hjic-847	130	46	,	,	PUNCT
hjic-847	130	47	respectively	respectively	ADV
hjic-847	130	48	.	.	PUNCT
hjic-847	131	1	kdpm	kdpm	PROPN
hjic-847	131	2	=	=	PUNCT
hjic-847	131	3	nu⋃	nu⋃	PROPN
hjic-847	131	4	j=1	j=1	PROPN
hjic-847	131	5	{	{	PUNCT
hjic-847	131	6	ij	ij	INTJ
hjic-847	131	7	|dj	|dj	NOUN
hjic-847	131	8	6=	6=	NOUN
hjic-847	131	9	ck+1	ck+1	NUM
hjic-847	131	10	}	}	PUNCT
hjic-847	131	11	(	(	PUNCT
hjic-847	131	12	11	11	NUM
hjic-847	131	13	)	)	PUNCT
hjic-847	131	14	the	the	DET
hjic-847	131	15	weight	weight	NOUN
hjic-847	131	16	of	of	ADP
hjic-847	131	17	the	the	DET
hjic-847	131	18	images	image	NOUN
hjic-847	131	19	can	can	AUX
hjic-847	131	20	be	be	AUX
hjic-847	131	21	calculated	calculate	VERB
hjic-847	131	22	based	base	VERB
hjic-847	131	23	on	on	ADP
hjic-847	131	24	the	the	DET
hjic-847	131	25	cluster	cluster	NOUN
hjic-847	131	26	coherence	coherence	NOUN
hjic-847	131	27	.	.	PUNCT
hjic-847	132	1	the	the	DET
hjic-847	132	2	coherence	coherence	NOUN
hjic-847	132	3	of	of	ADP
hjic-847	132	4	a	a	DET
hjic-847	132	5	cluster	cluster	NOUN
hjic-847	132	6	can	can	AUX
hjic-847	132	7	be	be	AUX
hjic-847	132	8	determined	determine	VERB
hjic-847	132	9	by	by	ADP
hjic-847	132	10	comparing	compare	VERB
hjic-847	132	11	the	the	DET
hjic-847	132	12	number	number	NOUN
hjic-847	132	13	of	of	ADP
hjic-847	132	14	known	know	VERB
hjic-847	132	15	images	image	NOUN
hjic-847	132	16	to	to	ADP
hjic-847	132	17	the	the	DET
hjic-847	132	18	number	number	NOUN
hjic-847	132	19	of	of	ADP
hjic-847	132	20	predicted	predict	VERB
hjic-847	132	21	unknown	unknown	ADJ
hjic-847	132	22	images	image	NOUN
hjic-847	132	23	inside	inside	ADP
hjic-847	132	24	that	that	PRON
hjic-847	132	25	given	give	VERB
hjic-847	132	26	cluster	cluster	NOUN
hjic-847	132	27	.	.	PUNCT
hjic-847	133	1	it	it	PRON
hjic-847	133	2	should	should	AUX
hjic-847	133	3	be	be	AUX
hjic-847	133	4	noted	note	VERB
hjic-847	133	5	that	that	SCONJ
hjic-847	133	6	known	know	VERB
hjic-847	133	7	images	image	NOUN
hjic-847	133	8	inside	inside	ADP
hjic-847	133	9	the	the	DET
hjic-847	133	10	clusters	cluster	NOUN
hjic-847	133	11	either	either	CCONJ
hjic-847	133	12	originate	originate	VERB
hjic-847	133	13	from	from	ADP
hjic-847	133	14	kgt	kgt	PROPN
hjic-847	133	15	or	or	CCONJ
hjic-847	133	16	kdpm	kdpm	PROPN
hjic-847	133	17	,	,	PUNCT
hjic-847	133	18	while	while	SCONJ
hjic-847	133	19	the	the	DET
hjic-847	133	20	predicted	predict	VERB
hjic-847	133	21	unknown	unknown	ADJ
hjic-847	133	22	images	image	NOUN
hjic-847	133	23	are	be	AUX
hjic-847	133	24	all	all	PRON
hjic-847	133	25	part	part	NOUN
hjic-847	133	26	of	of	ADP
hjic-847	133	27	udpm	udpm	NOUN
hjic-847	133	28	.	.	PUNCT
hjic-847	134	1	if	if	SCONJ
hjic-847	134	2	the	the	DET
hjic-847	134	3	number	number	NOUN
hjic-847	134	4	of	of	ADP
hjic-847	134	5	known	know	VERB
hjic-847	134	6	images	image	NOUN
hjic-847	134	7	exceeds	exceed	VERB
hjic-847	134	8	the	the	DET
hjic-847	134	9	number	number	NOUN
hjic-847	134	10	of	of	ADP
hjic-847	134	11	unknown	unknown	ADJ
hjic-847	134	12	images	image	NOUN
hjic-847	134	13	it	it	PRON
hjic-847	134	14	is	be	AUX
hjic-847	134	15	implied	imply	VERB
hjic-847	134	16	that	that	SCONJ
hjic-847	134	17	a	a	DET
hjic-847	134	18	cluster	cluster	NOUN
hjic-847	134	19	exhibits	exhibit	VERB
hjic-847	134	20	“	"	PUNCT
hjic-847	134	21	known	know	VERB
hjic-847	134	22	coherence	coherence	NOUN
hjic-847	134	23	”	"	PUNCT
hjic-847	134	24	(	(	PUNCT
hjic-847	134	25	kc	kc	PROPN
hjic-847	134	26	)	)	PUNCT
hjic-847	134	27	,	,	PUNCT
hjic-847	134	28	and	and	CCONJ
hjic-847	134	29	“	"	PUNCT
hjic-847	134	30	unknown	unknown	ADJ
hjic-847	134	31	coherence	coherence	NOUN
hjic-847	134	32	”	"	PUNCT
hjic-847	134	33	(	(	PUNCT
hjic-847	134	34	uc	uc	NOUN
hjic-847	134	35	)	)	PUNCT
hjic-847	134	36	vice	vice	NOUN
hjic-847	134	37	versa	versa	ADV
hjic-847	134	38	,	,	PUNCT
hjic-847	134	39	as	as	SCONJ
hjic-847	134	40	described	describe	VERB
hjic-847	134	41	in	in	ADP
hjic-847	134	42	:	:	PUNCT
hjic-847	134	43	ccoh	ccoh	NOUN
hjic-847	134	44	=	=	SYM
hjic-847	134	45	{	{	PUNCT
hjic-847	134	46	kc	kc	PROPN
hjic-847	134	47	|	|	ADV
hjic-847	134	48	∥∥{kgt	∥∥{kgt	VERB
hjic-847	134	49	∪kdpm	∪kdpm	X
hjic-847	134	50	}	}	PUNCT
hjic-847	134	51	∥∥	∥∥	X
hjic-847	134	52	≥	≥	PRON
hjic-847	134	53	∥∥udpm	∥∥udpm	PROPN
hjic-847	134	54	∥∥	∥∥	AUX
hjic-847	134	55	uc	uc	INTJ
hjic-847	134	56	|	|	ADV
hjic-847	134	57	∥∥{kgt	∥∥{kgt	X
hjic-847	134	58	∪kdpm	∪kdpm	X
hjic-847	134	59	}	}	PUNCT
hjic-847	134	60	∥∥	∥∥	X
hjic-847	134	61	<	<	X
hjic-847	134	62	∥∥udpm	∥∥udpm	PROPN
hjic-847	134	63	∥∥	∥∥	X
hjic-847	134	64	(	(	PUNCT
hjic-847	134	65	12	12	NUM
hjic-847	134	66	)	)	PUNCT
hjic-847	134	67	where	where	SCONJ
hjic-847	134	68	‖x‖	‖x‖	PROPN
hjic-847	134	69	represents	represent	VERB
hjic-847	134	70	the	the	DET
hjic-847	134	71	number	number	NOUN
hjic-847	134	72	of	of	ADP
hjic-847	134	73	elements	element	NOUN
hjic-847	134	74	in	in	ADP
hjic-847	134	75	x	x	PUNCT
hjic-847	134	76	,	,	PUNCT
hjic-847	134	77	and	and	CCONJ
hjic-847	134	78	the	the	DET
hjic-847	134	79	superscript	superscript	ADJ
hjic-847	134	80	coh	coh	NOUN
hjic-847	134	81	indicates	indicate	VERB
hjic-847	134	82	the	the	DET
hjic-847	134	83	coherence	coherence	NOUN
hjic-847	134	84	of	of	ADP
hjic-847	134	85	cluster	cluster	NOUN
hjic-847	134	86	c.	c.	NOUN
hjic-847	134	87	the	the	DET
hjic-847	134	88	weights	weight	NOUN
hjic-847	134	89	can	can	AUX
hjic-847	134	90	be	be	AUX
hjic-847	134	91	calculated	calculate	VERB
hjic-847	134	92	as	as	SCONJ
hjic-847	134	93	described	describe	VERB
hjic-847	134	94	in	in	ADP
hjic-847	134	95	eqs	eqs	PROPN
hjic-847	134	96	.	.	PROPN
hjic-847	134	97	13	13	NUM
hjic-847	134	98	and	and	CCONJ
hjic-847	134	99	14	14	NUM
hjic-847	134	100	.	.	PUNCT
hjic-847	135	1	intuitively	intuitively	ADV
hjic-847	135	2	,	,	PUNCT
hjic-847	135	3	if	if	SCONJ
hjic-847	135	4	an	an	DET
hjic-847	135	5	image	image	NOUN
hjic-847	135	6	is	be	AUX
hjic-847	135	7	known	know	VERB
hjic-847	135	8	and	and	CCONJ
hjic-847	135	9	located	locate	VERB
hjic-847	135	10	inside	inside	ADP
hjic-847	135	11	cluster	cluster	NOUN
hjic-847	135	12	uc	uc	PROPN
hjic-847	135	13	,	,	PUNCT
hjic-847	135	14	then	then	ADV
hjic-847	135	15	it	it	PRON
hjic-847	135	16	is	be	AUX
hjic-847	135	17	“	"	PUNCT
hjic-847	135	18	punished	punish	VERB
hjic-847	135	19	”	"	PUNCT
hjic-847	135	20	by	by	ADP
hjic-847	135	21	assigning	assign	VERB
hjic-847	135	22	a	a	DET
hjic-847	135	23	lower	low	ADJ
hjic-847	135	24	weight	weight	NOUN
hjic-847	135	25	to	to	ADP
hjic-847	135	26	it	it	PRON
hjic-847	135	27	;	;	PUNCT
hjic-847	135	28	and	and	CCONJ
hjic-847	135	29	vice	vice	ADV
hjic-847	135	30	versa	versa	ADV
hjic-847	135	31	,	,	PUNCT
hjic-847	135	32	an	an	DET
hjic-847	135	33	unknown	unknown	ADJ
hjic-847	135	34	image	image	NOUN
hjic-847	135	35	is	be	AUX
hjic-847	135	36	given	give	VERB
hjic-847	135	37	a	a	DET
hjic-847	135	38	lower	low	ADJ
hjic-847	135	39	weight	weight	NOUN
hjic-847	135	40	inside	inside	ADP
hjic-847	135	41	cluster	cluster	NOUN
hjic-847	135	42	kc	kc	PROPN
hjic-847	135	43	.	.	PUNCT
hjic-847	136	1	moreover	moreover	ADV
hjic-847	136	2	,	,	PUNCT
hjic-847	136	3	the	the	PRON
hjic-847	136	4	larger	large	ADJ
hjic-847	136	5	the	the	DET
hjic-847	136	6	difference	difference	NOUN
hjic-847	136	7	between	between	ADP
hjic-847	136	8	the	the	DET
hjic-847	136	9	numbers	number	NOUN
hjic-847	136	10	of	of	ADP
hjic-847	136	11	known	known	ADJ
hjic-847	136	12	and	and	CCONJ
hjic-847	136	13	unknown	unknown	ADJ
hjic-847	136	14	images	image	NOUN
hjic-847	136	15	implies	imply	VERB
hjic-847	136	16	a	a	DET
hjic-847	136	17	more	more	ADV
hjic-847	136	18	severe	severe	ADJ
hjic-847	136	19	punishment	punishment	NOUN
hjic-847	136	20	with	with	ADP
hjic-847	136	21	regard	regard	NOUN
hjic-847	136	22	to	to	ADP
hjic-847	136	23	the	the	DET
hjic-847	136	24	value	value	NOUN
hjic-847	136	25	of	of	ADP
hjic-847	136	26	weights	weight	NOUN
hjic-847	136	27	.	.	PUNCT
hjic-847	137	1	wkc	wkc	PROPN
hjic-847	137	2	j	j	PROPN
hjic-847	137	3	=	=	PUNCT
hjic-847	137	4			PROPN
hjic-847	137	5	1	1	NUM
hjic-847	137	6	+	+	CCONJ
hjic-847	137	7	(	(	PUNCT
hjic-847	137	8	#	#	NOUN
hjic-847	137	9	known−#unknown	known−#unknown	NOUN
hjic-847	137	10	)	)	PUNCT
hjic-847	137	11	(	(	PUNCT
hjic-847	137	12	#	#	SYM
hjic-847	137	13	known+#unknown	known+#unknown	NOUN
hjic-847	137	14	)	)	PUNCT
hjic-847	138	1	|	|	ADV
hjic-847	138	2	ij	ij	INTJ
hjic-847	138	3	/∈	/∈	PUNCT
hjic-847	139	1	udpm	udpm	PROPN
hjic-847	139	2	1−	1−	NUM
hjic-847	139	3	(	(	PUNCT
hjic-847	139	4	#	#	SYM
hjic-847	139	5	known−#unknown	known−#unknown	NOUN
hjic-847	139	6	)	)	PUNCT
hjic-847	139	7	(	(	PUNCT
hjic-847	139	8	#	#	SYM
hjic-847	139	9	known+#unknown	known+#unknown	NOUN
hjic-847	139	10	)	)	PUNCT
hjic-847	139	11	|	|	ADV
hjic-847	139	12	ij	ij	INTJ
hjic-847	139	13	∈	∈	PROPN
hjic-847	139	14	udpm	udpm	X
hjic-847	139	15	(	(	PUNCT
hjic-847	139	16	13	13	NUM
hjic-847	139	17	)	)	PUNCT
hjic-847	139	18	wuc	wuc	PROPN
hjic-847	139	19	j	j	PROPN
hjic-847	140	1	=	=	PUNCT
hjic-847	140	2			PROPN
hjic-847	140	3	1	1	NUM
hjic-847	140	4	+	+	CCONJ
hjic-847	140	5	(	(	PUNCT
hjic-847	140	6	#	#	NOUN
hjic-847	140	7	known−#unknown	known−#unknown	NOUN
hjic-847	140	8	)	)	PUNCT
hjic-847	140	9	(	(	PUNCT
hjic-847	140	10	#	#	SYM
hjic-847	140	11	known+#unknown	known+#unknown	NOUN
hjic-847	140	12	)	)	PUNCT
hjic-847	141	1	|	|	ADV
hjic-847	141	2	ij	ij	INTJ
hjic-847	141	3	∈	∈	PROPN
hjic-847	141	4	udpm	udpm	PROPN
hjic-847	141	5	1−	1−	NUM
hjic-847	141	6	(	(	PUNCT
hjic-847	141	7	#	#	SYM
hjic-847	141	8	known−#unknown	known−#unknown	NOUN
hjic-847	141	9	)	)	PUNCT
hjic-847	141	10	(	(	PUNCT
hjic-847	141	11	#	#	SYM
hjic-847	141	12	known+#unknown	known+#unknown	NOUN
hjic-847	141	13	)	)	PUNCT
hjic-847	142	1	|	|	ADV
hjic-847	142	2	ij	ij	INTJ
hjic-847	142	3	/∈	/∈	PUNCT
hjic-847	143	1	udpm	udpm	PROPN
hjic-847	143	2	(	(	PUNCT
hjic-847	143	3	14	14	NUM
hjic-847	143	4	)	)	PUNCT
hjic-847	143	5	thereafter	thereafter	ADV
hjic-847	143	6	the	the	DET
hjic-847	143	7	final	final	ADJ
hjic-847	143	8	decision	decision	NOUN
hjic-847	143	9	vector	vector	NOUN
hjic-847	143	10	of	of	ADP
hjic-847	143	11	cluster	cluster	NOUN
hjic-847	143	12	ci	ci	NOUN
hjic-847	143	13	can	can	AUX
hjic-847	143	14	be	be	AUX
hjic-847	143	15	calculated	calculate	VERB
hjic-847	143	16	as	as	ADP
hjic-847	143	17	:	:	PUNCT
hjic-847	143	18	vi	vi	PROPN
hjic-847	143	19	=	=	SYM
hjic-847	143	20	1	1	NUM
hjic-847	143	21	ni	ni	NOUN
hjic-847	143	22	ni∑	ni∑	PROPN
hjic-847	143	23	j=1	j=1	NOUN
hjic-847	143	24	wj	wj	X
hjic-847	143	25	×	×	NOUN
hjic-847	143	26	dj	dj	X
hjic-847	143	27	(	(	PUNCT
hjic-847	143	28	15	15	NUM
hjic-847	143	29	)	)	PUNCT
hjic-847	143	30	where	where	SCONJ
hjic-847	143	31	ni	ni	PROPN
hjic-847	143	32	denotes	denote	VERB
hjic-847	143	33	the	the	DET
hjic-847	143	34	number	number	NOUN
hjic-847	143	35	of	of	ADP
hjic-847	143	36	images	image	NOUN
hjic-847	143	37	in	in	ADP
hjic-847	143	38	cluster	cluster	NOUN
hjic-847	143	39	ci	ci	PROPN
hjic-847	143	40	,	,	PUNCT
hjic-847	143	41	wj	wj	PROPN
hjic-847	143	42	represents	represent	VERB
hjic-847	143	43	the	the	DET
hjic-847	143	44	weight	weight	NOUN
hjic-847	143	45	and	and	CCONJ
hjic-847	143	46	dj	dj	NOUN
hjic-847	143	47	stands	stand	VERB
hjic-847	143	48	for	for	ADP
hjic-847	143	49	the	the	DET
hjic-847	143	50	decision	decision	NOUN
hjic-847	143	51	vector	vector	NOUN
hjic-847	143	52	(	(	PUNCT
hjic-847	143	53	{	{	PUNCT
hjic-847	143	54	pci	pci	PROPN
hjic-847	143	55	}	}	PUNCT
hjic-847	143	56	)	)	PUNCT
hjic-847	143	57	of	of	ADP
hjic-847	143	58	image	image	NOUN
hjic-847	144	1	j.	j.	PROPN
hjic-847	144	2	note	note	VERB
hjic-847	144	3	that	that	SCONJ
hjic-847	144	4	dj	dj	NOUN
hjic-847	144	5	possesses	possess	VERB
hjic-847	144	6	k	k	X
hjic-847	144	7	+	+	CCONJ
hjic-847	144	8	1	1	NUM
hjic-847	144	9	elements	element	NOUN
hjic-847	144	10	(	(	PUNCT
hjic-847	144	11	+1	+1	PROPN
hjic-847	144	12	from	from	ADP
hjic-847	144	13	dpm	dpm	PROPN
hjic-847	144	14	)	)	PUNCT
hjic-847	144	15	,	,	PUNCT
hjic-847	144	16	therefore	therefore	ADV
hjic-847	144	17	,	,	PUNCT
hjic-847	144	18	vector	vector	PROPN
hjic-847	144	19	vi	vi	PROPN
hjic-847	144	20	also	also	ADV
hjic-847	144	21	possesses	possess	VERB
hjic-847	144	22	k	k	X
hjic-847	144	23	+	+	CCONJ
hjic-847	144	24	1	1	NUM
hjic-847	144	25	elements	element	NOUN
hjic-847	144	26	.	.	PUNCT
hjic-847	145	1	consequently	consequently	ADV
hjic-847	145	2	,	,	PUNCT
hjic-847	145	3	the	the	DET
hjic-847	145	4	element	element	NOUN
hjic-847	145	5	with	with	ADP
hjic-847	145	6	the	the	DET
hjic-847	145	7	maximum	maximum	ADJ
hjic-847	145	8	value	value	NOUN
hjic-847	145	9	of	of	ADP
hjic-847	145	10	vi	vi	PROPN
hjic-847	145	11	determines	determine	VERB
hjic-847	145	12	the	the	DET
hjic-847	145	13	category	category	NOUN
hjic-847	145	14	of	of	ADP
hjic-847	145	15	cluster	cluster	NOUN
hjic-847	145	16	ci	ci	PROPN
hjic-847	145	17	.	.	PUNCT
hjic-847	146	1	the	the	DET
hjic-847	146	2	classification	classification	NOUN
hjic-847	146	3	of	of	ADP
hjic-847	146	4	cluster	cluster	NOUN
hjic-847	146	5	ci	ci	NOUN
hjic-847	146	6	is	be	AUX
hjic-847	146	7	formalized	formalize	VERB
hjic-847	146	8	in	in	ADP
hjic-847	146	9	:	:	PUNCT
hjic-847	146	10	di	di	X
hjic-847	146	11	=	=	PUNCT
hjic-847	146	12	{	{	PUNCT
hjic-847	146	13	new	new	ADJ
hjic-847	146	14	class	class	NOUN
hjic-847	146	15	|	|	ADV
hjic-847	146	16	vk+1	vk+1	ADV
hjic-847	146	17	=	=	SYM
hjic-847	146	18	maxj{vj	maxj{vj	NOUN
hjic-847	146	19	}	}	PUNCT
hjic-847	146	20	argmaxi{vi	argmaxi{vi	NOUN
hjic-847	146	21	}	}	PUNCT
hjic-847	147	1	|	|	ADV
hjic-847	147	2	otherwise	otherwise	ADV
hjic-847	147	3	(	(	PUNCT
hjic-847	147	4	16	16	NUM
hjic-847	147	5	)	)	PUNCT
hjic-847	147	6	the	the	DET
hjic-847	147	7	results	result	NOUN
hjic-847	147	8	of	of	ADP
hjic-847	147	9	the	the	DET
hjic-847	147	10	classification	classification	NOUN
hjic-847	147	11	of	of	ADP
hjic-847	147	12	the	the	DET
hjic-847	147	13	clusters	cluster	NOUN
hjic-847	147	14	can	can	AUX
hjic-847	147	15	be	be	AUX
hjic-847	147	16	considered	consider	VERB
hjic-847	147	17	as	as	ADP
hjic-847	147	18	a	a	DET
hjic-847	147	19	labeling	labeling	NOUN
hjic-847	147	20	proposal	proposal	NOUN
hjic-847	147	21	,	,	PUNCT
hjic-847	147	22	i.e.	i.e.	X
hjic-847	147	23	label	label	VERB
hjic-847	147	24	each	each	DET
hjic-847	147	25	image	image	NOUN
hjic-847	147	26	inside	inside	ADP
hjic-847	147	27	cluster	cluster	NOUN
hjic-847	147	28	ci	ci	NOUN
hjic-847	147	29	according	accord	VERB
hjic-847	147	30	to	to	ADP
hjic-847	147	31	di	di	NOUN
hjic-847	147	32	.	.	PUNCT
hjic-847	148	1	when	when	SCONJ
hjic-847	148	2	decision	decision	NOUN
hjic-847	148	3	di	di	NOUN
hjic-847	148	4	for	for	ADP
hjic-847	148	5	cluster	cluster	NOUN
hjic-847	148	6	ci	ci	NOUN
hjic-847	148	7	is	be	AUX
hjic-847	148	8	that	that	SCONJ
hjic-847	148	9	it	it	PRON
hjic-847	148	10	is	be	AUX
hjic-847	148	11	part	part	NOUN
hjic-847	148	12	of	of	ADP
hjic-847	148	13	a	a	DET
hjic-847	148	14	known	know	VERB
hjic-847	148	15	category	category	NOUN
hjic-847	148	16	,	,	PUNCT
hjic-847	148	17	then	then	ADV
hjic-847	148	18	each	each	DET
hjic-847	148	19	image	image	NOUN
hjic-847	148	20	inside	inside	ADP
hjic-847	148	21	ci	ci	PROPN
hjic-847	148	22	gets	get	AUX
hjic-847	148	23	labeled	label	VERB
hjic-847	148	24	with	with	ADP
hjic-847	148	25	the	the	DET
hjic-847	148	26	same	same	ADJ
hjic-847	148	27	category	category	NOUN
hjic-847	148	28	.	.	PUNCT
hjic-847	149	1	on	on	ADP
hjic-847	149	2	the	the	DET
hjic-847	149	3	other	other	ADJ
hjic-847	149	4	hand	hand	NOUN
hjic-847	149	5	,	,	PUNCT
hjic-847	149	6	when	when	SCONJ
hjic-847	149	7	di	di	X
hjic-847	149	8	=	=	NOUN
hjic-847	149	9	a	a	DET
hjic-847	149	10	new	new	ADJ
hjic-847	149	11	class	class	NOUN
hjic-847	149	12	,	,	PUNCT
hjic-847	149	13	a	a	DET
hjic-847	149	14	new	new	ADJ
hjic-847	149	15	category	category	NOUN
hjic-847	149	16	is	be	AUX
hjic-847	149	17	created	create	VERB
hjic-847	149	18	and	and	CCONJ
hjic-847	149	19	each	each	DET
hjic-847	149	20	image	image	NOUN
hjic-847	149	21	in	in	ADP
hjic-847	149	22	ci	ci	NOUN
hjic-847	149	23	gets	get	AUX
hjic-847	149	24	labeled	label	VERB
hjic-847	149	25	with	with	ADP
hjic-847	149	26	the	the	DET
hjic-847	149	27	new	new	ADJ
hjic-847	149	28	category	category	NOUN
hjic-847	149	29	.	.	PUNCT
hjic-847	150	1	basically	basically	ADV
hjic-847	150	2	,	,	PUNCT
hjic-847	150	3	the	the	DET
hjic-847	150	4	cc	cc	PROPN
hjic-847	150	5	algorithm	algorithm	PROPN
hjic-847	150	6	follows	follow	VERB
hjic-847	150	7	this	this	DET
hjic-847	150	8	labeling	labeling	NOUN
hjic-847	150	9	proposal	proposal	NOUN
hjic-847	150	10	.	.	PUNCT
hjic-847	151	1	3	3	X
hjic-847	151	2	.	.	X
hjic-847	151	3	experimental	experimental	ADJ
hjic-847	151	4	results	result	NOUN
hjic-847	151	5	in	in	ADP
hjic-847	151	6	order	order	NOUN
hjic-847	151	7	to	to	PART
hjic-847	151	8	measure	measure	VERB
hjic-847	151	9	the	the	DET
hjic-847	151	10	efficiency	efficiency	NOUN
hjic-847	151	11	of	of	ADP
hjic-847	151	12	the	the	DET
hjic-847	151	13	labeling	labeling	NOUN
hjic-847	151	14	process	process	NOUN
hjic-847	151	15	,	,	PUNCT
hjic-847	151	16	experiments	experiment	NOUN
hjic-847	151	17	were	be	AUX
hjic-847	151	18	conducted	conduct	VERB
hjic-847	151	19	on	on	ADP
hjic-847	151	20	the	the	DET
hjic-847	151	21	caltech101	caltech101	PROPN
hjic-847	151	22	[	[	X
hjic-847	151	23	24	24	NUM
hjic-847	151	24	]	]	PUNCT
hjic-847	151	25	and	and	CCONJ
hjic-847	151	26	hungarian	hungarian	ADJ
hjic-847	151	27	journal	journal	NOUN
hjic-847	151	28	of	of	ADP
hjic-847	151	29	industry	industry	NOUN
hjic-847	151	30	and	and	CCONJ
hjic-847	151	31	chemistry	chemistry	NOUN
hjic-847	151	32	automated	automate	VERB
hjic-847	151	33	labeling	labeling	NOUN
hjic-847	151	34	process	process	NOUN
hjic-847	151	35	for	for	ADP
hjic-847	151	36	unknown	unknown	ADJ
hjic-847	151	37	images	image	NOUN
hjic-847	151	38	37	37	NUM
hjic-847	151	39	figure	figure	NOUN
hjic-847	151	40	1	1	NUM
hjic-847	151	41	:	:	PUNCT
hjic-847	151	42	example	example	NOUN
hjic-847	151	43	images	image	NOUN
hjic-847	151	44	from	from	ADP
hjic-847	151	45	the	the	DET
hjic-847	151	46	caltech101	caltech101	PROPN
hjic-847	151	47	and	and	CCONJ
hjic-847	151	48	caltech256	caltech256	PROPN
hjic-847	151	49	datasets	dataset	NOUN
hjic-847	151	50	.	.	PUNCT
hjic-847	152	1	the	the	DET
hjic-847	152	2	airplane	airplane	NOUN
hjic-847	152	3	,	,	PUNCT
hjic-847	152	4	butterfly	butterfly	NOUN
hjic-847	152	5	and	and	CCONJ
hjic-847	152	6	windmill	windmill	NOUN
hjic-847	152	7	categories	category	NOUN
hjic-847	152	8	are	be	AUX
hjic-847	152	9	represented	represent	VERB
hjic-847	152	10	by	by	ADP
hjic-847	152	11	the	the	DET
hjic-847	152	12	left	left	ADJ
hjic-847	152	13	,	,	PUNCT
hjic-847	152	14	middle	middle	ADJ
hjic-847	152	15	and	and	CCONJ
hjic-847	152	16	right	right	ADJ
hjic-847	152	17	columns	column	NOUN
hjic-847	152	18	,	,	PUNCT
hjic-847	152	19	respectively	respectively	ADV
hjic-847	152	20	.	.	PUNCT
hjic-847	153	1	caltech256	caltech256	PROPN
hjic-847	154	1	[	[	X
hjic-847	154	2	25	25	NUM
hjic-847	154	3	]	]	PUNCT
hjic-847	154	4	datasets	dataset	NOUN
hjic-847	154	5	.	.	PUNCT
hjic-847	155	1	example	example	NOUN
hjic-847	155	2	images	image	NOUN
hjic-847	155	3	from	from	ADP
hjic-847	155	4	these	these	DET
hjic-847	155	5	datasets	dataset	NOUN
hjic-847	155	6	are	be	AUX
hjic-847	155	7	shown	show	VERB
hjic-847	155	8	in	in	ADP
hjic-847	155	9	fig	fig	NOUN
hjic-847	155	10	.	.	PUNCT
hjic-847	156	1	1	1	X
hjic-847	156	2	.	.	PUNCT
hjic-847	156	3	the	the	DET
hjic-847	156	4	former	former	ADJ
hjic-847	156	5	consists	consist	VERB
hjic-847	156	6	of	of	ADP
hjic-847	156	7	101	101	NUM
hjic-847	156	8	categories	category	NOUN
hjic-847	156	9	and	and	CCONJ
hjic-847	156	10	8	8	NUM
hjic-847	156	11	,	,	PUNCT
hjic-847	156	12	677	677	NUM
hjic-847	156	13	images	image	NOUN
hjic-847	156	14	,	,	PUNCT
hjic-847	156	15	while	while	SCONJ
hjic-847	156	16	the	the	DET
hjic-847	156	17	latter	latter	ADJ
hjic-847	156	18	is	be	AUX
hjic-847	156	19	composed	compose	VERB
hjic-847	156	20	of	of	ADP
hjic-847	156	21	30	30	NUM
hjic-847	156	22	,	,	PUNCT
hjic-847	156	23	607	607	NUM
hjic-847	156	24	images	image	NOUN
hjic-847	156	25	from	from	ADP
hjic-847	156	26	256	256	NUM
hjic-847	156	27	different	different	ADJ
hjic-847	156	28	classes	class	NOUN
hjic-847	156	29	.	.	PUNCT
hjic-847	157	1	to	to	PART
hjic-847	157	2	create	create	VERB
hjic-847	157	3	an	an	DET
hjic-847	157	4	open	open	ADJ
hjic-847	157	5	-	-	PUNCT
hjic-847	157	6	world	world	NOUN
hjic-847	157	7	environment	environment	NOUN
hjic-847	157	8	,	,	PUNCT
hjic-847	157	9	50	50	NUM
hjic-847	157	10	known	know	VERB
hjic-847	157	11	and	and	CCONJ
hjic-847	157	12	50	50	NUM
hjic-847	157	13	unknown	unknown	ADJ
hjic-847	157	14	categories	category	NOUN
hjic-847	157	15	were	be	AUX
hjic-847	157	16	randomly	randomly	ADV
hjic-847	157	17	selected	select	VERB
hjic-847	157	18	from	from	ADP
hjic-847	157	19	the	the	DET
hjic-847	157	20	caltech101	caltech101	PROPN
hjic-847	157	21	dataset	dataset	NOUN
hjic-847	157	22	,	,	PUNCT
hjic-847	157	23	and	and	CCONJ
hjic-847	157	24	100	100	NUM
hjic-847	157	25	of	of	ADP
hjic-847	157	26	both	both	DET
hjic-847	157	27	categories	category	NOUN
hjic-847	157	28	from	from	ADP
hjic-847	157	29	the	the	DET
hjic-847	157	30	caltech256	caltech256	PROPN
hjic-847	157	31	dataset	dataset	NOUN
hjic-847	157	32	.	.	PUNCT
hjic-847	158	1	these	these	DET
hjic-847	158	2	ranfigure	ranfigure	NOUN
hjic-847	158	3	2	2	NUM
hjic-847	158	4	:	:	PUNCT
hjic-847	158	5	averaged	average	VERB
hjic-847	158	6	results	result	NOUN
hjic-847	158	7	of	of	ADP
hjic-847	158	8	the	the	DET
hjic-847	158	9	5	5	NUM
hjic-847	158	10	-	-	SYM
hjic-847	158	11	5	5	NUM
hjic-847	158	12	different	different	ADJ
hjic-847	158	13	test	test	NOUN
hjic-847	158	14	datasets	dataset	NOUN
hjic-847	158	15	that	that	PRON
hjic-847	158	16	were	be	AUX
hjic-847	158	17	randomly	randomly	ADV
hjic-847	158	18	selected	select	VERB
hjic-847	158	19	from	from	ADP
hjic-847	158	20	the	the	DET
hjic-847	158	21	caltech101	caltech101	PROPN
hjic-847	158	22	and	and	CCONJ
hjic-847	158	23	caltech256	caltech256	PROPN
hjic-847	158	24	datasets	dataset	NOUN
hjic-847	158	25	.	.	PUNCT
hjic-847	159	1	the	the	DET
hjic-847	159	2	ri	ri	PROPN
hjic-847	159	3	is	be	AUX
hjic-847	159	4	plotted	plot	VERB
hjic-847	159	5	against	against	ADP
hjic-847	159	6	the	the	DET
hjic-847	159	7	number	number	NOUN
hjic-847	159	8	of	of	ADP
hjic-847	159	9	unknown	unknown	ADJ
hjic-847	159	10	categories	category	NOUN
hjic-847	159	11	.	.	PUNCT
hjic-847	160	1	the	the	DET
hjic-847	160	2	diagrams	diagram	NOUN
hjic-847	160	3	compare	compare	VERB
hjic-847	160	4	the	the	DET
hjic-847	160	5	labeling	labeling	NOUN
hjic-847	160	6	performance	performance	NOUN
hjic-847	160	7	of	of	ADP
hjic-847	160	8	the	the	DET
hjic-847	160	9	dpm	dpm	PROPN
hjic-847	160	10	with	with	ADP
hjic-847	160	11	kernel	kernel	PROPN
hjic-847	160	12	k	k	X
hjic-847	160	13	-	-	PUNCT
hjic-847	160	14	means	means	PROPN
hjic-847	160	15	(	(	PUNCT
hjic-847	160	16	dpm+kk	dpm+kk	PROPN
hjic-847	160	17	)	)	PUNCT
hjic-847	160	18	against	against	ADP
hjic-847	160	19	the	the	DET
hjic-847	160	20	cc	cc	PROPN
hjic-847	160	21	.	.	PROPN
hjic-847	160	22	table	table	NOUN
hjic-847	161	1	1	1	NUM
hjic-847	161	2	:	:	PUNCT
hjic-847	161	3	summary	summary	NOUN
hjic-847	161	4	of	of	ADP
hjic-847	161	5	the	the	DET
hjic-847	161	6	results	result	NOUN
hjic-847	161	7	obtained	obtain	VERB
hjic-847	161	8	from	from	ADP
hjic-847	161	9	the	the	DET
hjic-847	161	10	test	test	NOUN
hjic-847	161	11	data	datum	NOUN
hjic-847	161	12	with	with	ADP
hjic-847	161	13	the	the	DET
hjic-847	161	14	baseline	baseline	NOUN
hjic-847	161	15	(	(	PUNCT
hjic-847	161	16	dpm+kk	dpm+kk	PROPN
hjic-847	161	17	)	)	PUNCT
hjic-847	161	18	and	and	CCONJ
hjic-847	161	19	cc	cc	ADP
hjic-847	161	20	methods	method	NOUN
hjic-847	161	21	using	use	VERB
hjic-847	161	22	the	the	DET
hjic-847	161	23	caltech101	caltech101	PROPN
hjic-847	161	24	and	and	CCONJ
hjic-847	161	25	caltech256	caltech256	PROPN
hjic-847	161	26	datasets	dataset	NOUN
hjic-847	161	27	.	.	PUNCT
hjic-847	162	1	the	the	DET
hjic-847	162	2	baseline	baseline	PROPN
hjic-847	162	3	column	column	NOUN
hjic-847	162	4	contains	contain	VERB
hjic-847	162	5	the	the	DET
hjic-847	162	6	ri	ri	PROPN
hjic-847	162	7	values	value	NOUN
hjic-847	162	8	evaluated	evaluate	VERB
hjic-847	162	9	which	which	PRON
hjic-847	162	10	depend	depend	VERB
hjic-847	162	11	on	on	ADP
hjic-847	162	12	the	the	DET
hjic-847	162	13	number	number	NOUN
hjic-847	162	14	of	of	ADP
hjic-847	162	15	unknown	unknown	ADJ
hjic-847	162	16	categories	category	NOUN
hjic-847	162	17	(	(	PUNCT
hjic-847	162	18	un	un	PROPN
hjic-847	162	19	.	.	PROPN
hjic-847	162	20	cat	cat	NOUN
hjic-847	162	21	.	.	PUNCT
hjic-847	162	22	)	)	PUNCT
hjic-847	162	23	,	,	PUNCT
hjic-847	162	24	and	and	CCONJ
hjic-847	162	25	the	the	DET
hjic-847	162	26	cc	cc	PROPN
hjic-847	162	27	column	column	NOUN
hjic-847	162	28	presents	present	VERB
hjic-847	162	29	the	the	DET
hjic-847	162	30	improvements	improvement	NOUN
hjic-847	162	31	that	that	PRON
hjic-847	162	32	result	result	VERB
hjic-847	162	33	from	from	ADP
hjic-847	162	34	cc	cc	PROPN
hjic-847	162	35	as	as	ADP
hjic-847	162	36	a	a	DET
hjic-847	162	37	percentage	percentage	NOUN
hjic-847	162	38	.	.	PUNCT
hjic-847	163	1	caltech101	caltech101	PROPN
hjic-847	163	2	caltech256	caltech256	PROPN
hjic-847	163	3	un	un	PROPN
hjic-847	163	4	.	.	PROPN
hjic-847	163	5	cat	cat	PROPN
hjic-847	163	6	.	.	PUNCT
hjic-847	164	1	base	base	NOUN
hjic-847	164	2	-	-	PUNCT
hjic-847	164	3	line	line	NOUN
hjic-847	164	4	cc	cc	NOUN
hjic-847	164	5	(	(	PUNCT
hjic-847	164	6	%	%	INTJ
hjic-847	164	7	)	)	PUNCT
hjic-847	164	8	un	un	PROPN
hjic-847	164	9	.	.	PROPN
hjic-847	164	10	cat	cat	PROPN
hjic-847	164	11	.	.	PUNCT
hjic-847	165	1	base	base	NOUN
hjic-847	165	2	-	-	PUNCT
hjic-847	165	3	line	line	NOUN
hjic-847	165	4	cc	cc	NOUN
hjic-847	165	5	(	(	PUNCT
hjic-847	165	6	%	%	INTJ
hjic-847	165	7	)	)	PUNCT
hjic-847	165	8	5	5	NUM
hjic-847	165	9	0.629	0.629	NUM
hjic-847	165	10	6	6	NUM
hjic-847	165	11	10	10	NUM
hjic-847	165	12	0.514	0.514	NUM
hjic-847	165	13	1	1	NUM
hjic-847	165	14	10	10	NUM
hjic-847	165	15	0.594	0.594	NUM
hjic-847	165	16	13	13	NUM
hjic-847	165	17	20	20	NUM
hjic-847	165	18	0.489	0.489	NUM
hjic-847	165	19	9	9	NUM
hjic-847	165	20	15	15	NUM
hjic-847	165	21	0.567	0.567	NUM
hjic-847	165	22	15	15	NUM
hjic-847	165	23	30	30	NUM
hjic-847	165	24	0.484	0.484	NUM
hjic-847	165	25	6	6	NUM
hjic-847	165	26	20	20	NUM
hjic-847	165	27	0.561	0.561	NUM
hjic-847	165	28	16	16	NUM
hjic-847	165	29	40	40	NUM
hjic-847	165	30	0.478	0.478	NUM
hjic-847	165	31	9	9	NUM
hjic-847	165	32	25	25	NUM
hjic-847	165	33	0.550	0.550	NUM
hjic-847	165	34	17	17	NUM
hjic-847	165	35	50	50	NUM
hjic-847	165	36	0.452	0.452	NUM
hjic-847	165	37	13	13	NUM
hjic-847	165	38	30	30	NUM
hjic-847	165	39	0.514	0.514	NUM
hjic-847	165	40	28	28	NUM
hjic-847	165	41	60	60	NUM
hjic-847	165	42	0.448	0.448	NUM
hjic-847	165	43	13	13	NUM
hjic-847	165	44	35	35	NUM
hjic-847	165	45	0.536	0.536	NUM
hjic-847	165	46	18	18	NUM
hjic-847	165	47	70	70	NUM
hjic-847	165	48	0.433	0.433	NUM
hjic-847	165	49	19	19	NUM
hjic-847	165	50	40	40	NUM
hjic-847	165	51	0.522	0.522	NUM
hjic-847	165	52	23	23	NUM
hjic-847	165	53	80	80	NUM
hjic-847	165	54	0.426	0.426	NUM
hjic-847	165	55	17	17	NUM
hjic-847	165	56	45	45	NUM
hjic-847	165	57	0.505	0.505	NUM
hjic-847	165	58	24	24	NUM
hjic-847	165	59	90	90	NUM
hjic-847	165	60	0.412	0.412	NUM
hjic-847	165	61	22	22	NUM
hjic-847	165	62	50	50	NUM
hjic-847	165	63	0.497	0.497	NUM
hjic-847	165	64	25	25	NUM
hjic-847	165	65	100	100	NUM
hjic-847	165	66	0.397	0.397	NUM
hjic-847	165	67	21	21	NUM
hjic-847	165	68	dom	dom	NOUN
hjic-847	165	69	selections	selection	NOUN
hjic-847	165	70	were	be	AUX
hjic-847	165	71	repeated	repeat	VERB
hjic-847	165	72	5	5	NUM
hjic-847	165	73	times	time	NOUN
hjic-847	165	74	in	in	ADP
hjic-847	165	75	order	order	NOUN
hjic-847	165	76	to	to	PART
hjic-847	165	77	calculate	calculate	VERB
hjic-847	165	78	the	the	DET
hjic-847	165	79	average	average	NOUN
hjic-847	165	80	of	of	ADP
hjic-847	165	81	the	the	DET
hjic-847	165	82	results	result	NOUN
hjic-847	165	83	of	of	ADP
hjic-847	165	84	each	each	DET
hjic-847	165	85	experiment	experiment	NOUN
hjic-847	165	86	to	to	PART
hjic-847	165	87	obtain	obtain	VERB
hjic-847	165	88	a	a	DET
hjic-847	165	89	more	more	ADV
hjic-847	165	90	comprehensive	comprehensive	ADJ
hjic-847	165	91	overview	overview	NOUN
hjic-847	165	92	of	of	ADP
hjic-847	165	93	the	the	DET
hjic-847	165	94	efficiency	efficiency	NOUN
hjic-847	165	95	of	of	ADP
hjic-847	165	96	the	the	DET
hjic-847	165	97	cc	cc	PROPN
hjic-847	165	98	algorithm	algorithm	NOUN
hjic-847	165	99	with	with	ADP
hjic-847	165	100	regard	regard	NOUN
hjic-847	165	101	to	to	ADP
hjic-847	165	102	these	these	DET
hjic-847	165	103	datasets	dataset	NOUN
hjic-847	165	104	.	.	PUNCT
hjic-847	166	1	all	all	PRON
hjic-847	166	2	of	of	ADP
hjic-847	166	3	the	the	DET
hjic-847	166	4	known	know	VERB
hjic-847	166	5	categories	category	NOUN
hjic-847	166	6	were	be	AUX
hjic-847	166	7	available	available	ADJ
hjic-847	166	8	from	from	ADP
hjic-847	166	9	the	the	DET
hjic-847	166	10	beginning	beginning	NOUN
hjic-847	166	11	of	of	ADP
hjic-847	166	12	the	the	DET
hjic-847	166	13	tests	test	NOUN
hjic-847	166	14	,	,	PUNCT
hjic-847	166	15	but	but	CCONJ
hjic-847	166	16	the	the	DET
hjic-847	166	17	unknown	unknown	ADJ
hjic-847	166	18	categories	category	NOUN
hjic-847	166	19	were	be	AUX
hjic-847	166	20	added	add	VERB
hjic-847	166	21	incrementally	incrementally	ADV
hjic-847	166	22	over	over	ADP
hjic-847	166	23	10	10	NUM
hjic-847	166	24	steps	step	NOUN
hjic-847	166	25	,	,	PUNCT
hjic-847	166	26	and	and	CCONJ
hjic-847	166	27	in	in	ADP
hjic-847	166	28	each	each	DET
hjic-847	166	29	step	step	NOUN
hjic-847	166	30	the	the	DET
hjic-847	166	31	rand	rand	NOUN
hjic-847	166	32	index	index	NOUN
hjic-847	166	33	(	(	PUNCT
hjic-847	166	34	ri	ri	PROPN
hjic-847	166	35	)	)	PUNCT
hjic-847	166	36	,	,	PUNCT
hjic-847	166	37	ri	ri	PROPN
hjic-847	167	1	=	=	PUNCT
hjic-847	167	2	tp	tp	PART
hjic-847	167	3	+	+	CCONJ
hjic-847	167	4	tn	tn	NOUN
hjic-847	167	5	tp+	tp+	NOUN
hjic-847	167	6	fp	fp	X
hjic-847	167	7	+	+	NUM
hjic-847	167	8	tn+	tn+	NOUN
hjic-847	167	9	fn	fn	NOUN
hjic-847	167	10	,	,	PUNCT
hjic-847	167	11	(	(	PUNCT
hjic-847	167	12	17	17	NUM
hjic-847	167	13	)	)	PUNCT
hjic-847	167	14	was	be	AUX
hjic-847	167	15	evaluated	evaluate	VERB
hjic-847	167	16	over	over	ADP
hjic-847	167	17	the	the	DET
hjic-847	167	18	unknown	unknown	ADJ
hjic-847	167	19	images	image	NOUN
hjic-847	167	20	,	,	PUNCT
hjic-847	167	21	where	where	SCONJ
hjic-847	167	22	tp	tp	NOUN
hjic-847	167	23	,	,	PUNCT
hjic-847	167	24	tn	tn	PROPN
hjic-847	167	25	,	,	PUNCT
hjic-847	167	26	fp	fp	NOUN
hjic-847	167	27	,	,	PUNCT
hjic-847	167	28	and	and	CCONJ
hjic-847	167	29	fn	fn	NOUN
hjic-847	167	30	denote	denote	VERB
hjic-847	167	31	the	the	DET
hjic-847	167	32	number	number	NOUN
hjic-847	167	33	of	of	ADP
hjic-847	167	34	true	true	ADJ
hjic-847	167	35	positive	positive	ADJ
hjic-847	167	36	,	,	PUNCT
hjic-847	167	37	true	true	ADJ
hjic-847	167	38	negative	negative	ADJ
hjic-847	167	39	,	,	PUNCT
hjic-847	167	40	false	false	ADJ
hjic-847	167	41	positive	positive	ADJ
hjic-847	167	42	and	and	CCONJ
hjic-847	167	43	false	false	ADJ
hjic-847	167	44	negative	negative	ADJ
hjic-847	167	45	decisions	decision	NOUN
hjic-847	167	46	,	,	PUNCT
hjic-847	167	47	respectively	respectively	ADV
hjic-847	167	48	.	.	PUNCT
hjic-847	168	1	the	the	DET
hjic-847	168	2	ri	ri	PROPN
hjic-847	168	3	measures	measure	VERB
hjic-847	168	4	the	the	DET
hjic-847	168	5	similarity	similarity	NOUN
hjic-847	168	6	between	between	ADP
hjic-847	168	7	the	the	DET
hjic-847	168	8	ground	ground	NOUN
hjic-847	168	9	truth	truth	NOUN
hjic-847	168	10	and	and	CCONJ
hjic-847	168	11	predicted	predict	VERB
hjic-847	168	12	labels	label	NOUN
hjic-847	168	13	of	of	ADP
hjic-847	168	14	the	the	DET
hjic-847	168	15	unknown	unknown	ADJ
hjic-847	168	16	images	image	NOUN
hjic-847	168	17	,	,	PUNCT
hjic-847	168	18	in	in	ADP
hjic-847	168	19	other	other	ADJ
hjic-847	168	20	words	word	NOUN
hjic-847	168	21	,	,	PUNCT
hjic-847	168	22	the	the	DET
hjic-847	168	23	percentage	percentage	NOUN
hjic-847	168	24	of	of	ADP
hjic-847	168	25	correct	correct	ADJ
hjic-847	168	26	decisions	decision	NOUN
hjic-847	168	27	.	.	PUNCT
hjic-847	169	1	two	two	NUM
hjic-847	169	2	methods	method	NOUN
hjic-847	169	3	were	be	AUX
hjic-847	169	4	assessed	assess	VERB
hjic-847	169	5	and	and	CCONJ
hjic-847	169	6	compared	compare	VERB
hjic-847	169	7	,	,	PUNCT
hjic-847	169	8	namely	namely	ADV
hjic-847	169	9	the	the	DET
hjic-847	169	10	baseline	baseline	NOUN
hjic-847	169	11	method	method	NOUN
hjic-847	169	12	(	(	PUNCT
hjic-847	169	13	dpm+kk	dpm+kk	PROPN
hjic-847	169	14	)	)	PUNCT
hjic-847	169	15	and	and	CCONJ
hjic-847	169	16	the	the	DET
hjic-847	169	17	cc	cc	PROPN
hjic-847	169	18	,	,	PUNCT
hjic-847	169	19	which	which	PRON
hjic-847	169	20	were	be	AUX
hjic-847	169	21	discussed	discuss	VERB
hjic-847	169	22	in	in	ADP
hjic-847	169	23	section	section	NOUN
hjic-847	169	24	2.3	2.3	NUM
hjic-847	169	25	and	and	CCONJ
hjic-847	169	26	2.4	2.4	NUM
hjic-847	169	27	,	,	PUNCT
hjic-847	169	28	respectively	respectively	ADV
hjic-847	169	29	.	.	PUNCT
hjic-847	170	1	both	both	DET
hjic-847	170	2	procedures	procedure	NOUN
hjic-847	170	3	used	use	VERB
hjic-847	170	4	fisher	fisher	PROPN
hjic-847	170	5	vectors	vector	NOUN
hjic-847	170	6	to	to	PART
hjic-847	170	7	mathematically	mathematically	ADV
hjic-847	170	8	represent	represent	VERB
hjic-847	170	9	the	the	DET
hjic-847	170	10	images	image	NOUN
hjic-847	170	11	encoded	encode	VERB
hjic-847	170	12	from	from	ADP
hjic-847	170	13	128	128	NUM
hjic-847	170	14	dimensional	dimensional	ADJ
hjic-847	170	15	sift	sift	ADJ
hjic-847	170	16	descriptors	descriptor	NOUN
hjic-847	170	17	using	use	VERB
hjic-847	170	18	a	a	DET
hjic-847	170	19	gmm	gmm	NOUN
hjic-847	170	20	consisting	consist	VERB
hjic-847	170	21	of	of	ADP
hjic-847	170	22	256	256	NUM
hjic-847	170	23	code	code	NOUN
hjic-847	170	24	words	word	NOUN
hjic-847	170	25	;	;	PUNCT
hjic-847	170	26	a	a	DET
hjic-847	170	27	svm	svm	NOUN
hjic-847	170	28	equipped	equip	VERB
hjic-847	170	29	with	with	ADP
hjic-847	170	30	a	a	DET
hjic-847	170	31	radial	radial	ADJ
hjic-847	170	32	basis	basis	NOUN
hjic-847	170	33	function	function	NOUN
hjic-847	170	34	(	(	PUNCT
hjic-847	170	35	rbf	rbf	PROPN
hjic-847	170	36	)	)	PUNCT
hjic-847	170	37	kernel	kernel	PROPN
hjic-847	170	38	was	be	AUX
hjic-847	170	39	applied	apply	VERB
hjic-847	170	40	as	as	ADP
hjic-847	170	41	a	a	DET
hjic-847	170	42	classifier	classifier	NOUN
hjic-847	170	43	.	.	PUNCT
hjic-847	171	1	the	the	DET
hjic-847	171	2	results	result	NOUN
hjic-847	171	3	can	can	AUX
hjic-847	171	4	be	be	AUX
hjic-847	171	5	seen	see	VERB
hjic-847	171	6	in	in	ADP
hjic-847	171	7	fig	fig	NOUN
hjic-847	171	8	.	.	PUNCT
hjic-847	172	1	2	2	NUM
hjic-847	172	2	and	and	CCONJ
hjic-847	172	3	table	table	NOUN
hjic-847	172	4	1	1	NUM
hjic-847	172	5	.	.	PUNCT
hjic-847	173	1	the	the	DET
hjic-847	173	2	first	first	ADJ
hjic-847	173	3	diagram	diagram	NOUN
hjic-847	173	4	shows	show	VERB
hjic-847	173	5	the	the	DET
hjic-847	173	6	results	result	NOUN
hjic-847	173	7	obtained	obtain	VERB
hjic-847	173	8	from	from	ADP
hjic-847	173	9	the	the	DET
hjic-847	173	10	caltech101	caltech101	PROPN
hjic-847	173	11	dataset	dataset	NOUN
hjic-847	173	12	and	and	CCONJ
hjic-847	173	13	the	the	DET
hjic-847	173	14	second	second	ADJ
hjic-847	173	15	from	from	ADP
hjic-847	173	16	the	the	DET
hjic-847	173	17	caltech256	caltech256	PROPN
hjic-847	173	18	dataset	dataset	NOUN
hjic-847	173	19	.	.	PUNCT
hjic-847	174	1	the	the	DET
hjic-847	174	2	dpm	dpm	PROPN
hjic-847	174	3	with	with	ADP
hjic-847	174	4	kernel	kernel	PROPN
hjic-847	174	5	k	k	X
hjic-847	174	6	-	-	PUNCT
hjic-847	174	7	means	mean	NOUN
hjic-847	174	8	and	and	CCONJ
hjic-847	174	9	the	the	DET
hjic-847	174	10	cc	cc	NOUN
hjic-847	174	11	are	be	AUX
hjic-847	174	12	represented	represent	VERB
hjic-847	174	13	by	by	ADP
hjic-847	174	14	dashed	dash	VERB
hjic-847	174	15	and	and	CCONJ
hjic-847	174	16	solid	solid	ADJ
hjic-847	174	17	lines	line	NOUN
hjic-847	174	18	,	,	PUNCT
hjic-847	174	19	respectively	respectively	ADV
hjic-847	174	20	.	.	PUNCT
hjic-847	175	1	in	in	ADP
hjic-847	175	2	both	both	DET
hjic-847	175	3	experiments	experiment	NOUN
hjic-847	175	4	,	,	PUNCT
hjic-847	175	5	the	the	DET
hjic-847	175	6	cc	cc	PROPN
hjic-847	175	7	algorithm	algorithm	PROPN
hjic-847	175	8	yielded	yield	VERB
hjic-847	175	9	a	a	DET
hjic-847	175	10	higher	high	ADJ
hjic-847	175	11	ri	ri	NOUN
hjic-847	175	12	,	,	PUNCT
hjic-847	175	13	although	although	SCONJ
hjic-847	175	14	during	during	ADP
hjic-847	175	15	the	the	DET
hjic-847	175	16	first	first	ADJ
hjic-847	175	17	step	step	NOUN
hjic-847	175	18	the	the	DET
hjic-847	175	19	difference	difference	NOUN
hjic-847	175	20	between	between	ADP
hjic-847	175	21	the	the	DET
hjic-847	175	22	two	two	NUM
hjic-847	175	23	methods	method	NOUN
hjic-847	175	24	was	be	AUX
hjic-847	175	25	minimal	minimal	ADJ
hjic-847	175	26	.	.	PUNCT
hjic-847	176	1	it	it	PRON
hjic-847	176	2	can	can	AUX
hjic-847	176	3	be	be	AUX
hjic-847	176	4	seen	see	VERB
hjic-847	176	5	that	that	SCONJ
hjic-847	176	6	the	the	DET
hjic-847	176	7	ri	ri	PROPN
hjic-847	176	8	of	of	ADP
hjic-847	176	9	dpm+kk	dpm+kk	PROPN
hjic-847	176	10	starts	start	VERB
hjic-847	176	11	to	to	PART
hjic-847	176	12	decrease	decrease	VERB
hjic-847	176	13	as	as	ADP
hjic-847	176	14	the	the	DET
hjic-847	176	15	number	number	NOUN
hjic-847	176	16	of	of	ADP
hjic-847	176	17	unknown	unknown	ADJ
hjic-847	176	18	categories	category	NOUN
hjic-847	176	19	increases	increase	NOUN
hjic-847	176	20	,	,	PUNCT
hjic-847	176	21	while	while	SCONJ
hjic-847	176	22	the	the	DET
hjic-847	176	23	cc	cc	NOUN
hjic-847	176	24	remains	remain	VERB
hjic-847	176	25	by	by	ADP
hjic-847	176	26	and	and	CCONJ
hjic-847	176	27	large	large	ADJ
hjic-847	176	28	unchanged	unchanged	ADJ
hjic-847	176	29	.	.	PUNCT
hjic-847	177	1	47(1	47(1	NUM
hjic-847	177	2	)	)	PUNCT
hjic-847	177	3	pp	pp	ADV
hjic-847	177	4	.	.	PUNCT
hjic-847	178	1	33–39	33–39	NUM
hjic-847	178	2	(	(	PUNCT
hjic-847	178	3	2019	2019	NUM
hjic-847	178	4	)	)	PUNCT
hjic-847	178	5	38	38	NUM
hjic-847	178	6	papp	papp	NOUN
hjic-847	178	7	and	and	CCONJ
hjic-847	178	8	szűcs	szűcs	ADJ
hjic-847	178	9	4	4	NUM
hjic-847	178	10	.	.	PUNCT
hjic-847	178	11	conclusion	conclusion	NOUN
hjic-847	178	12	in	in	ADP
hjic-847	178	13	this	this	DET
hjic-847	178	14	paper	paper	NOUN
hjic-847	178	15	,	,	PUNCT
hjic-847	178	16	the	the	DET
hjic-847	178	17	problem	problem	NOUN
hjic-847	178	18	of	of	ADP
hjic-847	178	19	open	open	ADJ
hjic-847	178	20	world	world	NOUN
hjic-847	178	21	recognition	recognition	NOUN
hjic-847	178	22	was	be	AUX
hjic-847	178	23	reviewed	review	VERB
hjic-847	178	24	and	and	CCONJ
hjic-847	178	25	the	the	DET
hjic-847	178	26	possible	possible	ADJ
hjic-847	178	27	cases	case	NOUN
hjic-847	178	28	were	be	AUX
hjic-847	178	29	differentiated	differentiate	VERB
hjic-847	178	30	based	base	VERB
hjic-847	178	31	on	on	ADP
hjic-847	178	32	our	our	PRON
hjic-847	178	33	prior	prior	ADJ
hjic-847	178	34	knowledge	knowledge	NOUN
hjic-847	178	35	and	and	CCONJ
hjic-847	178	36	actual	actual	ADJ
hjic-847	178	37	information	information	NOUN
hjic-847	178	38	about	about	ADP
hjic-847	178	39	the	the	DET
hjic-847	178	40	test	test	NOUN
hjic-847	178	41	data	datum	NOUN
hjic-847	178	42	and	and	CCONJ
hjic-847	178	43	,	,	PUNCT
hjic-847	178	44	thus	thus	ADV
hjic-847	178	45	,	,	PUNCT
hjic-847	178	46	the	the	DET
hjic-847	178	47	unknown	unknown	ADJ
hjic-847	178	48	space	space	NOUN
hjic-847	178	49	.	.	PUNCT
hjic-847	179	1	the	the	DET
hjic-847	179	2	dpm	dpm	PROPN
hjic-847	179	3	and	and	CCONJ
hjic-847	179	4	kernel	kernel	PROPN
hjic-847	179	5	k	k	X
hjic-847	179	6	-	-	PUNCT
hjic-847	179	7	means	means	NOUN
hjic-847	179	8	algorithm	algorithm	NOUN
hjic-847	179	9	were	be	AUX
hjic-847	179	10	also	also	ADV
hjic-847	179	11	reviewed	review	VERB
hjic-847	179	12	in	in	ADP
hjic-847	179	13	brief	brief	NOUN
hjic-847	179	14	,	,	PUNCT
hjic-847	179	15	followed	follow	VERB
hjic-847	179	16	by	by	ADP
hjic-847	179	17	the	the	DET
hjic-847	179	18	presentation	presentation	NOUN
hjic-847	179	19	of	of	ADP
hjic-847	179	20	two	two	NUM
hjic-847	179	21	approaches	approach	NOUN
hjic-847	179	22	,	,	PUNCT
hjic-847	179	23	which	which	PRON
hjic-847	179	24	perform	perform	VERB
hjic-847	179	25	multi	multi	ADJ
hjic-847	179	26	-	-	ADJ
hjic-847	179	27	class	class	ADJ
hjic-847	179	28	classification	classification	NOUN
hjic-847	179	29	,	,	PUNCT
hjic-847	179	30	automatically	automatically	ADV
hjic-847	179	31	detect	detect	VERB
hjic-847	179	32	unknown	unknown	ADJ
hjic-847	179	33	images	image	NOUN
hjic-847	179	34	and	and	CCONJ
hjic-847	179	35	propose	propose	VERB
hjic-847	179	36	a	a	DET
hjic-847	179	37	labeling	labeling	NOUN
hjic-847	179	38	for	for	ADP
hjic-847	179	39	them	they	PRON
hjic-847	179	40	.	.	PUNCT
hjic-847	180	1	the	the	DET
hjic-847	180	2	first	first	ADJ
hjic-847	180	3	method	method	NOUN
hjic-847	180	4	is	be	AUX
hjic-847	180	5	a	a	DET
hjic-847	180	6	baseline	baseline	NOUN
hjic-847	180	7	technique	technique	NOUN
hjic-847	180	8	where	where	SCONJ
hjic-847	180	9	dpm	dpm	PROPN
hjic-847	180	10	was	be	AUX
hjic-847	180	11	sequentially	sequentially	ADV
hjic-847	180	12	applied	apply	VERB
hjic-847	180	13	followed	follow	VERB
hjic-847	180	14	by	by	ADP
hjic-847	180	15	kernel	kernel	PROPN
hjic-847	180	16	k	k	X
hjic-847	180	17	-	-	PUNCT
hjic-847	180	18	means	means	NOUN
hjic-847	180	19	with	with	ADP
hjic-847	180	20	a	a	DET
hjic-847	180	21	plusplus	plusplus	ADJ
hjic-847	180	22	cluster	cluster	NOUN
hjic-847	180	23	center	center	NOUN
hjic-847	180	24	initialization	initialization	NOUN
hjic-847	180	25	algorithm	algorithm	NOUN
hjic-847	180	26	.	.	PUNCT
hjic-847	181	1	however	however	ADV
hjic-847	181	2	,	,	PUNCT
hjic-847	181	3	our	our	PRON
hjic-847	181	4	proposed	propose	VERB
hjic-847	181	5	cc	cc	NOUN
hjic-847	181	6	is	be	AUX
hjic-847	181	7	a	a	DET
hjic-847	181	8	complex	complex	ADJ
hjic-847	181	9	method	method	NOUN
hjic-847	181	10	of	of	ADP
hjic-847	181	11	combining	combine	VERB
hjic-847	181	12	the	the	DET
hjic-847	181	13	unknown	unknown	ADJ
hjic-847	181	14	detector	detector	NOUN
hjic-847	181	15	and	and	CCONJ
hjic-847	181	16	clustering	cluster	VERB
hjic-847	181	17	algorithm	algorithm	NOUN
hjic-847	181	18	that	that	PRON
hjic-847	181	19	seeks	seek	VERB
hjic-847	181	20	to	to	PART
hjic-847	181	21	determine	determine	VERB
hjic-847	181	22	the	the	DET
hjic-847	181	23	identity	identity	NOUN
hjic-847	181	24	of	of	ADP
hjic-847	181	25	formed	form	VERB
hjic-847	181	26	clusters	cluster	NOUN
hjic-847	181	27	,	,	PUNCT
hjic-847	181	28	while	while	SCONJ
hjic-847	181	29	refining	refine	VERB
hjic-847	181	30	the	the	DET
hjic-847	181	31	decisions	decision	NOUN
hjic-847	181	32	made	make	VERB
hjic-847	181	33	by	by	ADP
hjic-847	181	34	the	the	DET
hjic-847	181	35	classifier	classifier	NOUN
hjic-847	181	36	and	and	CCONJ
hjic-847	181	37	unknown	unknown	ADJ
hjic-847	181	38	detector	detector	NOUN
hjic-847	181	39	.	.	PUNCT
hjic-847	182	1	the	the	DET
hjic-847	182	2	cc	cc	PROPN
hjic-847	182	3	algorithm	algorithm	PROPN
hjic-847	182	4	constructs	construct	VERB
hjic-847	182	5	a	a	DET
hjic-847	182	6	specific	specific	ADJ
hjic-847	182	7	weight	weight	NOUN
hjic-847	182	8	system	system	NOUN
hjic-847	182	9	to	to	PART
hjic-847	182	10	reward	reward	VERB
hjic-847	182	11	or	or	CCONJ
hjic-847	182	12	punish	punish	VERB
hjic-847	182	13	images	image	NOUN
hjic-847	182	14	which	which	PRON
hjic-847	182	15	were	be	AUX
hjic-847	182	16	placed	place	VERB
hjic-847	182	17	into	into	ADP
hjic-847	182	18	a	a	DET
hjic-847	182	19	category	category	NOUN
hjic-847	182	20	that	that	PRON
hjic-847	182	21	is	be	AUX
hjic-847	182	22	presumably	presumably	ADV
hjic-847	182	23	unsuitable	unsuitable	ADJ
hjic-847	182	24	for	for	ADP
hjic-847	182	25	their	their	PRON
hjic-847	182	26	estimated	estimate	VERB
hjic-847	182	27	identity	identity	NOUN
hjic-847	182	28	.	.	PUNCT
hjic-847	183	1	multiple	multiple	ADJ
hjic-847	183	2	experiments	experiment	NOUN
hjic-847	183	3	were	be	AUX
hjic-847	183	4	conducted	conduct	VERB
hjic-847	183	5	on	on	ADP
hjic-847	183	6	two	two	NUM
hjic-847	183	7	large	large	ADJ
hjic-847	183	8	datasets	dataset	NOUN
hjic-847	183	9	(	(	PUNCT
hjic-847	183	10	caltech101	caltech101	PROPN
hjic-847	183	11	and	and	CCONJ
hjic-847	183	12	caltech256	caltech256	PROPN
hjic-847	183	13	)	)	PUNCT
hjic-847	183	14	,	,	PUNCT
hjic-847	183	15	and	and	CCONJ
hjic-847	183	16	the	the	DET
hjic-847	183	17	ri	ri	PROPN
hjic-847	183	18	evaluated	evaluate	VERB
hjic-847	183	19	with	with	ADP
hjic-847	183	20	regard	regard	NOUN
hjic-847	183	21	to	to	ADP
hjic-847	183	22	the	the	DET
hjic-847	183	23	unknown	unknown	ADJ
hjic-847	183	24	images	image	NOUN
hjic-847	183	25	.	.	PUNCT
hjic-847	184	1	the	the	DET
hjic-847	184	2	results	result	NOUN
hjic-847	184	3	showed	show	VERB
hjic-847	184	4	that	that	SCONJ
hjic-847	184	5	the	the	DET
hjic-847	184	6	cc	cc	PROPN
hjic-847	184	7	outperformed	outperform	VERB
hjic-847	184	8	the	the	DET
hjic-847	184	9	baseline	baseline	NOUN
hjic-847	184	10	method	method	NOUN
hjic-847	184	11	,	,	PUNCT
hjic-847	184	12	and	and	CCONJ
hjic-847	184	13	was	be	AUX
hjic-847	184	14	able	able	ADJ
hjic-847	184	15	to	to	PART
hjic-847	184	16	maintain	maintain	VERB
hjic-847	184	17	almost	almost	ADV
hjic-847	184	18	the	the	DET
hjic-847	184	19	same	same	ADJ
hjic-847	184	20	ri	ri	NOUN
hjic-847	184	21	,	,	PUNCT
hjic-847	184	22	while	while	SCONJ
hjic-847	184	23	the	the	DET
hjic-847	184	24	number	number	NOUN
hjic-847	184	25	of	of	ADP
hjic-847	184	26	unknown	unknown	ADJ
hjic-847	184	27	categories	category	NOUN
hjic-847	184	28	increased	increase	VERB
hjic-847	184	29	.	.	PUNCT
hjic-847	185	1	acknowledgement	acknowledgement	NOUN
hjic-847	185	2	the	the	DET
hjic-847	185	3	research	research	NOUN
hjic-847	185	4	was	be	AUX
hjic-847	185	5	supported	support	VERB
hjic-847	185	6	by	by	ADP
hjic-847	185	7	the	the	DET
hjic-847	185	8	únkp-18	únkp-18	PROPN
hjic-847	185	9	-	-	PUNCT
hjic-847	185	10	3	3	NUM
hjic-847	185	11	new	new	ADJ
hjic-847	185	12	national	national	PROPN
hjic-847	185	13	excellence	excellence	PROPN
hjic-847	185	14	program	program	NOUN
hjic-847	185	15	of	of	ADP
hjic-847	185	16	the	the	DET
hjic-847	185	17	ministry	ministry	PROPN
hjic-847	185	18	of	of	ADP
hjic-847	185	19	human	human	ADJ
hjic-847	185	20	capacities	capacity	NOUN
hjic-847	185	21	.	.	PUNCT
hjic-847	186	1	references	reference	NOUN
hjic-847	186	2	[	[	X
hjic-847	186	3	1	1	NUM
hjic-847	186	4	]	]	X
hjic-847	186	5	lampert	lampert	PROPN
hjic-847	186	6	,	,	PUNCT
hjic-847	186	7	c.	c.	PROPN
hjic-847	186	8	h.	h.	PROPN
hjic-847	186	9	;	;	PUNCT
hjic-847	186	10	nickisch	nickisch	PROPN
hjic-847	186	11	,	,	PUNCT
hjic-847	186	12	h.	h.	PROPN
hjic-847	186	13	;	;	PUNCT
hjic-847	186	14	harmeling	harmele	VERB
hjic-847	186	15	,	,	PUNCT
hjic-847	186	16	s.	s.	PROPN
hjic-847	186	17	:	:	PUNCT
hjic-847	186	18	learning	learn	VERB
hjic-847	186	19	to	to	PART
hjic-847	186	20	detect	detect	VERB
hjic-847	186	21	unseen	unseen	ADJ
hjic-847	186	22	object	object	NOUN
hjic-847	186	23	classes	class	NOUN
hjic-847	186	24	by	by	ADP
hjic-847	186	25	betweenclass	betweenclass	NOUN
hjic-847	186	26	attribute	attribute	NOUN
hjic-847	186	27	transfer	transfer	NOUN
hjic-847	186	28	,	,	PUNCT
hjic-847	186	29	2009	2009	NUM
hjic-847	186	30	ieee	ieee	NOUN
hjic-847	186	31	conference	conference	NOUN
hjic-847	186	32	on	on	ADP
hjic-847	186	33	computer	computer	NOUN
hjic-847	186	34	vision	vision	NOUN
hjic-847	186	35	and	and	CCONJ
hjic-847	186	36	pattern	pattern	NOUN
hjic-847	186	37	recognition	recognition	NOUN
hjic-847	186	38	,	,	PUNCT
hjic-847	186	39	2009	2009	NUM
hjic-847	186	40	,	,	PUNCT
hjic-847	186	41	pp	pp	ADV
hjic-847	186	42	.	.	PUNCT
hjic-847	187	1	951–958	951–958	NUM
hjic-847	187	2	isbn	isbn	NOUN
hjic-847	187	3	:	:	PUNCT
hjic-847	187	4	978	978	NUM
hjic-847	187	5	-	-	SYM
hjic-847	187	6	1	1	NUM
hjic-847	187	7	-	-	PUNCT
hjic-847	187	8	4244	4244	NUM
hjic-847	187	9	-	-	PUNCT
hjic-847	187	10	3992	3992	NUM
hjic-847	187	11	-	-	SYM
hjic-847	187	12	8	8	NUM
hjic-847	187	13	doi	doi	NOUN
hjic-847	187	14	:	:	PUNCT
hjic-847	187	15	10.1109	10.1109	NUM
hjic-847	187	16	/	/	SYM
hjic-847	187	17	cvpr.2009.5206594	cvpr.2009.5206594	NOUN
hjic-847	187	18	[	[	X
hjic-847	187	19	2	2	NUM
hjic-847	187	20	]	]	PUNCT
hjic-847	187	21	scheirer	scheirer	ADV
hjic-847	187	22	,	,	PUNCT
hjic-847	187	23	w.	w.	PROPN
hjic-847	187	24	j.	j.	PROPN
hjic-847	187	25	;	;	PUNCT
hjic-847	187	26	de	de	PROPN
hjic-847	187	27	rezende	rezende	PROPN
hjic-847	187	28	rocha	rocha	PROPN
hjic-847	187	29	,	,	PUNCT
hjic-847	187	30	a.	a.	NOUN
hjic-847	187	31	;	;	PUNCT
hjic-847	187	32	sapkota	sapkota	NOUN
hjic-847	187	33	,	,	PUNCT
hjic-847	187	34	a.	a.	NOUN
hjic-847	187	35	;	;	PUNCT
hjic-847	187	36	boult	boult	NOUN
hjic-847	187	37	,	,	PUNCT
hjic-847	187	38	t.	t.	PROPN
hjic-847	187	39	e.	e.	PROPN
hjic-847	187	40	:	:	PUNCT
hjic-847	187	41	toward	toward	ADP
hjic-847	187	42	open	open	ADJ
hjic-847	187	43	set	set	NOUN
hjic-847	187	44	recognition	recognition	NOUN
hjic-847	187	45	,	,	PUNCT
hjic-847	187	46	ieee	ieee	NOUN
hjic-847	187	47	t.	t.	NOUN
hjic-847	187	48	pattern	pattern	NOUN
hjic-847	187	49	anal	anal	PROPN
hjic-847	187	50	.	.	PUNCT
hjic-847	187	51	,	,	PUNCT
hjic-847	187	52	2013	2013	NUM
hjic-847	187	53	35(7	35(7	NUM
hjic-847	187	54	)	)	PUNCT
hjic-847	187	55	,	,	PUNCT
hjic-847	187	56	1757–1772	1757–1772	NUM
hjic-847	187	57	doi	doi	NOUN
hjic-847	187	58	:	:	PUNCT
hjic-847	187	59	10.1109	10.1109	NUM
hjic-847	187	60	/	/	SYM
hjic-847	187	61	tpami.2012.256	tpami.2012.256	PROPN
hjic-847	188	1	[	[	X
hjic-847	188	2	3	3	NUM
hjic-847	188	3	]	]	PUNCT
hjic-847	188	4	scheirer	scheirer	ADV
hjic-847	188	5	,	,	PUNCT
hjic-847	188	6	w.	w.	PROPN
hjic-847	188	7	j.	j.	PROPN
hjic-847	188	8	;	;	PUNCT
hjic-847	188	9	jain	jain	PROPN
hjic-847	188	10	,	,	PUNCT
hjic-847	188	11	l.	l.	PROPN
hjic-847	188	12	p.	p.	PROPN
hjic-847	188	13	;	;	PUNCT
hjic-847	188	14	boult	boult	NOUN
hjic-847	188	15	,	,	PUNCT
hjic-847	188	16	t.	t.	PROPN
hjic-847	188	17	e.	e.	PROPN
hjic-847	188	18	:	:	PUNCT
hjic-847	188	19	probability	probability	NOUN
hjic-847	188	20	models	model	NOUN
hjic-847	188	21	for	for	ADP
hjic-847	188	22	open	open	ADJ
hjic-847	188	23	set	set	NOUN
hjic-847	188	24	recognition	recognition	NOUN
hjic-847	188	25	,	,	PUNCT
hjic-847	188	26	ieee	ieee	NOUN
hjic-847	188	27	t.	t.	NOUN
hjic-847	188	28	pattern	pattern	NOUN
hjic-847	188	29	anal	anal	PROPN
hjic-847	188	30	.	.	PUNCT
hjic-847	188	31	,	,	PUNCT
hjic-847	188	32	2014	2014	NUM
hjic-847	188	33	36(11	36(11	NUM
hjic-847	188	34	)	)	PUNCT
hjic-847	188	35	,	,	PUNCT
hjic-847	188	36	2317–2324	2317–2324	PROPN
hjic-847	188	37	doi	doi	NOUN
hjic-847	188	38	:	:	PUNCT
hjic-847	188	39	10.1109	10.1109	NUM
hjic-847	188	40	/	/	SYM
hjic-847	188	41	tpami.2014.2321392	tpami.2014.2321392	PRON
hjic-847	188	42	[	[	X
hjic-847	188	43	4	4	NUM
hjic-847	188	44	]	]	X
hjic-847	188	45	bendale	bendale	NOUN
hjic-847	188	46	,	,	PUNCT
hjic-847	188	47	a.	a.	NOUN
hjic-847	188	48	;	;	PUNCT
hjic-847	188	49	boult	boult	NOUN
hjic-847	188	50	,	,	PUNCT
hjic-847	188	51	t.	t.	PROPN
hjic-847	188	52	:	:	PUNCT
hjic-847	188	53	towards	towards	ADP
hjic-847	188	54	open	open	ADJ
hjic-847	188	55	world	world	NOUN
hjic-847	188	56	recognition	recognition	NOUN
hjic-847	188	57	,	,	PUNCT
hjic-847	188	58	2015	2015	NUM
hjic-847	188	59	ieee	ieee	NOUN
hjic-847	188	60	conference	conference	NOUN
hjic-847	188	61	on	on	ADP
hjic-847	188	62	computer	computer	NOUN
hjic-847	188	63	vision	vision	NOUN
hjic-847	188	64	and	and	CCONJ
hjic-847	188	65	pattern	pattern	NOUN
hjic-847	188	66	recognition	recognition	NOUN
hjic-847	188	67	,	,	PUNCT
hjic-847	188	68	2015	2015	NUM
hjic-847	188	69	,	,	PUNCT
hjic-847	188	70	pp	pp	ADJ
hjic-847	188	71	.	.	PUNCT
hjic-847	189	1	1893–1902	1893–1902	NUM
hjic-847	189	2	isbn	isbn	NOUN
hjic-847	189	3	:	:	PUNCT
hjic-847	189	4	978	978	NUM
hjic-847	189	5	-	-	SYM
hjic-847	189	6	1	1	NUM
hjic-847	189	7	-	-	PUNCT
hjic-847	189	8	4673	4673	NUM
hjic-847	189	9	-	-	PUNCT
hjic-847	189	10	6964	6964	NUM
hjic-847	189	11	-	-	SYM
hjic-847	189	12	0	0	NUM
hjic-847	189	13	doi	doi	NOUN
hjic-847	189	14	:	:	PUNCT
hjic-847	189	15	10.1109	10.1109	NUM
hjic-847	189	16	/	/	SYM
hjic-847	189	17	cvpr.2015.7298799	cvpr.2015.7298799	PRON
hjic-847	190	1	[	[	X
hjic-847	190	2	5	5	NUM
hjic-847	190	3	]	]	X
hjic-847	190	4	papp	papp	NOUN
hjic-847	190	5	,	,	PUNCT
hjic-847	190	6	d.	d.	PROPN
hjic-847	190	7	;	;	PUNCT
hjic-847	190	8	szűcs	szűcs	PROPN
hjic-847	190	9	,	,	PUNCT
hjic-847	190	10	g.	g.	PROPN
hjic-847	190	11	:	:	PUNCT
hjic-847	190	12	double	double	ADJ
hjic-847	190	13	probability	probability	NOUN
hjic-847	190	14	model	model	NOUN
hjic-847	190	15	for	for	ADP
hjic-847	190	16	open	open	ADJ
hjic-847	190	17	set	set	VERB
hjic-847	190	18	problem	problem	NOUN
hjic-847	190	19	at	at	ADP
hjic-847	190	20	image	image	NOUN
hjic-847	190	21	classification	classification	NOUN
hjic-847	190	22	,	,	PUNCT
hjic-847	190	23	informatica	informatica	PROPN
hjic-847	190	24	,	,	PUNCT
hjic-847	190	25	2018	2018	NUM
hjic-847	190	26	29(2	29(2	NOUN
hjic-847	190	27	)	)	PUNCT
hjic-847	190	28	,	,	PUNCT
hjic-847	190	29	353–369	353–369	NUM
hjic-847	190	30	doi	doi	NOUN
hjic-847	190	31	:	:	PUNCT
hjic-847	190	32	10.15388	10.15388	NUM
hjic-847	190	33	/	/	SYM
hjic-847	190	34	informatica.2018.171	informatica.2018.171	PROPN
hjic-847	191	1	[	[	X
hjic-847	191	2	6	6	NUM
hjic-847	191	3	]	]	SYM
hjic-847	191	4	tax	tax	NOUN
hjic-847	191	5	,	,	PUNCT
hjic-847	191	6	d.	d.	PROPN
hjic-847	191	7	m.	m.	PROPN
hjic-847	191	8	;	;	PUNCT
hjic-847	191	9	duin	duin	PROPN
hjic-847	191	10	,	,	PUNCT
hjic-847	191	11	r.	r.	PROPN
hjic-847	191	12	p.	p.	PROPN
hjic-847	191	13	:	:	PUNCT
hjic-847	192	1	support	support	NOUN
hjic-847	192	2	vector	vector	NOUN
hjic-847	192	3	data	datum	NOUN
hjic-847	192	4	description	description	NOUN
hjic-847	192	5	,	,	PUNCT
hjic-847	192	6	machine	machine	NOUN
hjic-847	192	7	learning	learning	NOUN
hjic-847	192	8	,	,	PUNCT
hjic-847	192	9	2004	2004	NUM
hjic-847	192	10	54(1	54(1	NUM
hjic-847	192	11	)	)	PUNCT
hjic-847	192	12	,	,	PUNCT
hjic-847	192	13	45–66	45–66	NUM
hjic-847	192	14	doi	doi	NOUN
hjic-847	192	15	:	:	PUNCT
hjic-847	192	16	10.1023	10.1023	NUM
hjic-847	192	17	/	/	SYM
hjic-847	192	18	b	b	NOUN
hjic-847	192	19	:	:	PUNCT
hjic-847	192	20	mach.0000008084.60811.49	mach.0000008084.60811.49	PROPN
hjic-847	193	1	[	[	X
hjic-847	193	2	7	7	NUM
hjic-847	193	3	]	]	X
hjic-847	193	4	cevikalp	cevikalp	NOUN
hjic-847	193	5	,	,	PUNCT
hjic-847	193	6	h.	h.	PROPN
hjic-847	193	7	;	;	PUNCT
hjic-847	193	8	triggs	triggs	PROPN
hjic-847	193	9	,	,	PUNCT
hjic-847	193	10	b.	b.	PROPN
hjic-847	193	11	:	:	PUNCT
hjic-847	193	12	efficient	efficient	ADJ
hjic-847	193	13	object	object	NOUN
hjic-847	193	14	detection	detection	NOUN
hjic-847	193	15	using	use	VERB
hjic-847	193	16	cascades	cascade	NOUN
hjic-847	193	17	of	of	ADP
hjic-847	193	18	nearest	near	ADJ
hjic-847	193	19	convex	convex	ADJ
hjic-847	193	20	model	model	NOUN
hjic-847	193	21	classifiers	classifier	NOUN
hjic-847	193	22	,	,	PUNCT
hjic-847	193	23	2012	2012	NUM
hjic-847	193	24	ieee	ieee	NOUN
hjic-847	193	25	conference	conference	NOUN
hjic-847	193	26	on	on	ADP
hjic-847	193	27	computer	computer	NOUN
hjic-847	193	28	vision	vision	NOUN
hjic-847	193	29	and	and	CCONJ
hjic-847	193	30	pattern	pattern	NOUN
hjic-847	193	31	recognition	recognition	NOUN
hjic-847	193	32	,	,	PUNCT
hjic-847	193	33	2012	2012	NUM
hjic-847	193	34	,	,	PUNCT
hjic-847	193	35	pp	pp	ADJ
hjic-847	193	36	.	.	PUNCT
hjic-847	194	1	3138–3145	3138–3145	NUM
hjic-847	194	2	isbn	isbn	ADJ
hjic-847	194	3	:	:	PUNCT
hjic-847	194	4	978	978	NUM
hjic-847	194	5	-	-	SYM
hjic-847	194	6	1	1	NUM
hjic-847	194	7	-	-	PUNCT
hjic-847	194	8	4673	4673	NUM
hjic-847	194	9	-	-	PUNCT
hjic-847	194	10	1226	1226	NUM
hjic-847	194	11	-	-	SYM
hjic-847	194	12	4	4	NUM
hjic-847	194	13	doi	doi	NOUN
hjic-847	194	14	:	:	PUNCT
hjic-847	194	15	10.1109	10.1109	NUM
hjic-847	194	16	/	/	SYM
hjic-847	194	17	cvpr.2012.6248047	cvpr.2012.6248047	ADJ
hjic-847	194	18	[	[	X
hjic-847	194	19	8	8	NUM
hjic-847	194	20	]	]	PUNCT
hjic-847	194	21	schölkopf	schölkopf	NOUN
hjic-847	194	22	,	,	PUNCT
hjic-847	194	23	b.	b.	PROPN
hjic-847	194	24	;	;	PUNCT
hjic-847	194	25	platt	platt	PROPN
hjic-847	194	26	,	,	PUNCT
hjic-847	194	27	j.	j.	PROPN
hjic-847	194	28	c.	c.	PROPN
hjic-847	194	29	;	;	PUNCT
hjic-847	194	30	shawe	shawe	NOUN
hjic-847	194	31	-	-	PUNCT
hjic-847	194	32	taylor	taylor	PROPN
hjic-847	194	33	,	,	PUNCT
hjic-847	194	34	j.	j.	PROPN
hjic-847	194	35	;	;	PUNCT
hjic-847	194	36	smola	smola	PROPN
hjic-847	194	37	,	,	PUNCT
hjic-847	194	38	a.	a.	PROPN
hjic-847	194	39	j.	j.	PROPN
hjic-847	194	40	;	;	PUNCT
hjic-847	194	41	williamson	williamson	PROPN
hjic-847	194	42	,	,	PUNCT
hjic-847	194	43	r.	r.	PROPN
hjic-847	194	44	c.	c.	PROPN
hjic-847	194	45	:	:	PUNCT
hjic-847	194	46	estimating	estimate	VERB
hjic-847	194	47	the	the	DET
hjic-847	194	48	support	support	NOUN
hjic-847	194	49	of	of	ADP
hjic-847	194	50	a	a	DET
hjic-847	194	51	high	high	ADJ
hjic-847	194	52	-	-	PUNCT
hjic-847	194	53	dimensional	dimensional	ADJ
hjic-847	194	54	distribution	distribution	NOUN
hjic-847	194	55	,	,	PUNCT
hjic-847	194	56	neural	neural	ADJ
hjic-847	194	57	comput	comput	NOUN
hjic-847	194	58	.	.	PUNCT
hjic-847	194	59	,	,	PUNCT
hjic-847	194	60	2001	2001	NUM
hjic-847	194	61	13(7	13(7	NUM
hjic-847	194	62	)	)	PUNCT
hjic-847	194	63	,	,	PUNCT
hjic-847	194	64	1443–1471	1443–1471	PROPN
hjic-847	194	65	doi	doi	NOUN
hjic-847	194	66	:	:	PUNCT
hjic-847	194	67	10.1162/089976601750264965	10.1162/089976601750264965	NUM
hjic-847	194	68	[	[	X
hjic-847	194	69	9	9	NUM
hjic-847	194	70	]	]	SYM
hjic-847	194	71	zhang	zhang	PROPN
hjic-847	194	72	,	,	PUNCT
hjic-847	194	73	r.	r.	PROPN
hjic-847	194	74	;	;	PUNCT
hjic-847	194	75	metaxas	metaxas	PROPN
hjic-847	194	76	,	,	PUNCT
hjic-847	194	77	d.	d.	PROPN
hjic-847	194	78	n.	n.	PROPN
hjic-847	194	79	:	:	PUNCT
hjic-847	194	80	ro	ro	ADJ
hjic-847	194	81	-	-	ADJ
hjic-847	194	82	svm	svm	ADJ
hjic-847	194	83	:	:	PUNCT
hjic-847	194	84	support	support	NOUN
hjic-847	194	85	vector	vector	NOUN
hjic-847	194	86	machine	machine	NOUN
hjic-847	194	87	with	with	ADP
hjic-847	194	88	reject	reject	NOUN
hjic-847	194	89	option	option	NOUN
hjic-847	194	90	for	for	ADP
hjic-847	194	91	image	image	NOUN
hjic-847	194	92	categorization	categorization	NOUN
hjic-847	194	93	,	,	PUNCT
hjic-847	194	94	in	in	ADP
hjic-847	194	95	:	:	PUNCT
hjic-847	194	96	chantler	chantler	NOUN
hjic-847	194	97	,	,	PUNCT
hjic-847	194	98	m.	m.	NOUN
hjic-847	194	99	;	;	PUNCT
hjic-847	194	100	fisher	fisher	PROPN
hjic-847	194	101	,	,	PUNCT
hjic-847	194	102	b.	b.	PROPN
hjic-847	194	103	;	;	PUNCT
hjic-847	194	104	trucco	trucco	PROPN
hjic-847	194	105	,	,	PUNCT
hjic-847	194	106	m.	m.	NOUN
hjic-847	194	107	;	;	PUNCT
hjic-847	194	108	(	(	PUNCT
hjic-847	194	109	eds	ed	NOUN
hjic-847	194	110	.	.	PUNCT
hjic-847	194	111	):	):	PUNCT
hjic-847	194	112	proceedings	proceeding	NOUN
hjic-847	194	113	of	of	ADP
hjic-847	194	114	the	the	DET
hjic-847	194	115	british	british	ADJ
hjic-847	194	116	machine	machine	NOUN
hjic-847	194	117	conference	conference	PROPN
hjic-847	194	118	(	(	PUNCT
hjic-847	194	119	bmva	bmva	NOUN
hjic-847	194	120	press	press	NOUN
hjic-847	194	121	,	,	PUNCT
hjic-847	194	122	uk	uk	PROPN
hjic-847	194	123	)	)	PUNCT
hjic-847	194	124	2006	2006	NUM
hjic-847	194	125	,	,	PUNCT
hjic-847	194	126	pp	pp	ADV
hjic-847	194	127	.	.	PUNCT
hjic-847	195	1	123.1	123.1	NUM
hjic-847	195	2	-	-	SYM
hjic-847	195	3	123.10	123.10	NUM
hjic-847	195	4	.	.	PUNCT
hjic-847	196	1	isbn	isbn	ADJ
hjic-847	196	2	:	:	PUNCT
hjic-847	196	3	1	1	NUM
hjic-847	196	4	-	-	NUM
hjic-847	196	5	901725	901725	NUM
hjic-847	196	6	-	-	SYM
hjic-847	196	7	32	32	NUM
hjic-847	196	8	-	-	SYM
hjic-847	196	9	4	4	NUM
hjic-847	196	10	doi	doi	NOUN
hjic-847	196	11	:	:	PUNCT
hjic-847	197	1	10.5244	10.5244	NUM
hjic-847	197	2	/	/	SYM
hjic-847	197	3	c.20.123	c.20.123	NOUN
hjic-847	197	4	[	[	X
hjic-847	197	5	10	10	NUM
hjic-847	197	6	]	]	X
hjic-847	197	7	jain	jain	PROPN
hjic-847	197	8	,	,	PUNCT
hjic-847	197	9	l.	l.	PROPN
hjic-847	197	10	p.	p.	PROPN
hjic-847	197	11	;	;	PUNCT
hjic-847	197	12	scheirer	scheirer	ADV
hjic-847	197	13	,	,	PUNCT
hjic-847	197	14	w.	w.	PROPN
hjic-847	197	15	j.	j.	PROPN
hjic-847	197	16	;	;	PUNCT
hjic-847	197	17	boult	boult	NOUN
hjic-847	197	18	,	,	PUNCT
hjic-847	197	19	t.	t.	PROPN
hjic-847	197	20	e.	e.	PROPN
hjic-847	197	21	:	:	PUNCT
hjic-847	197	22	multiclass	multiclass	ADJ
hjic-847	197	23	open	open	ADJ
hjic-847	197	24	set	set	NOUN
hjic-847	197	25	recognition	recognition	NOUN
hjic-847	197	26	using	use	VERB
hjic-847	197	27	probability	probability	NOUN
hjic-847	197	28	of	of	ADP
hjic-847	197	29	inclusion	inclusion	NOUN
hjic-847	197	30	,	,	PUNCT
hjic-847	197	31	in	in	ADP
hjic-847	197	32	:	:	PUNCT
hjic-847	197	33	fleet	fleet	PROPN
hjic-847	197	34	,	,	PUNCT
hjic-847	197	35	d.	d.	PROPN
hjic-847	197	36	;	;	PUNCT
hjic-847	197	37	pajdla	pajdla	NOUN
hjic-847	197	38	,	,	PUNCT
hjic-847	197	39	t.	t.	PROPN
hjic-847	197	40	;	;	PUNCT
hjic-847	197	41	schiele	schiele	PROPN
hjic-847	197	42	,	,	PUNCT
hjic-847	197	43	b.	b.	PROPN
hjic-847	197	44	;	;	PUNCT
hjic-847	197	45	tuytelaars	tuytelaar	NOUN
hjic-847	197	46	,	,	PUNCT
hjic-847	197	47	t.	t.	PROPN
hjic-847	197	48	;	;	PUNCT
hjic-847	197	49	(	(	PUNCT
hjic-847	197	50	eds	ed	NOUN
hjic-847	197	51	.	.	PROPN
hjic-847	197	52	):	):	PUNCT
hjic-847	197	53	computer	computer	NOUN
hjic-847	197	54	vision	vision	NOUN
hjic-847	197	55	–	–	PUNCT
hjic-847	197	56	eccv	eccv	ADV
hjic-847	197	57	2014	2014	NUM
hjic-847	197	58	.	.	PUNCT
hjic-847	197	59	,	,	PUNCT
hjic-847	197	60	eccv	eccv	ADV
hjic-847	197	61	2014	2014	NUM
hjic-847	197	62	.	.	PUNCT
hjic-847	198	1	lecture	lecture	NOUN
hjic-847	198	2	notes	note	NOUN
hjic-847	198	3	in	in	ADP
hjic-847	198	4	computer	computer	NOUN
hjic-847	198	5	science	science	NOUN
hjic-847	198	6	,	,	PUNCT
hjic-847	198	7	8691	8691	NUM
hjic-847	198	8	(	(	PUNCT
hjic-847	198	9	springer	springer	NOUN
hjic-847	198	10	,	,	PUNCT
hjic-847	198	11	cham	cham	PROPN
hjic-847	198	12	,	,	PUNCT
hjic-847	198	13	switzerland	switzerland	PROPN
hjic-847	198	14	)	)	PUNCT
hjic-847	198	15	2014	2014	NUM
hjic-847	198	16	,	,	PUNCT
hjic-847	198	17	pp	pp	ADP
hjic-847	198	18	.	.	PUNCT
hjic-847	199	1	393–409	393–409	NUM
hjic-847	199	2	.	.	PUNCT
hjic-847	200	1	isbn	isbn	ADJ
hjic-847	200	2	:	:	PUNCT
hjic-847	200	3	978	978	NUM
hjic-847	200	4	-	-	SYM
hjic-847	200	5	3	3	NUM
hjic-847	200	6	-	-	NUM
hjic-847	200	7	319	319	NUM
hjic-847	200	8	-	-	PUNCT
hjic-847	200	9	10577	10577	NUM
hjic-847	200	10	-	-	SYM
hjic-847	200	11	2	2	NUM
hjic-847	200	12	doi	doi	NOUN
hjic-847	200	13	:	:	PUNCT
hjic-847	200	14	10.1007/9783	10.1007/9783	NUM
hjic-847	200	15	-	-	NUM
hjic-847	200	16	319	319	NUM
hjic-847	200	17	-	-	PUNCT
hjic-847	200	18	10578	10578	NUM
hjic-847	200	19	-	-	SYM
hjic-847	200	20	9_26	9_26	NUM
hjic-847	200	21	[	[	PUNCT
hjic-847	200	22	11	11	NUM
hjic-847	200	23	]	]	X
hjic-847	200	24	fei	fei	PROPN
hjic-847	200	25	-	-	PUNCT
hjic-847	200	26	fei	fei	PROPN
hjic-847	200	27	,	,	PUNCT
hjic-847	200	28	l.	l.	PROPN
hjic-847	200	29	;	;	PUNCT
hjic-847	200	30	fergus	fergus	PROPN
hjic-847	200	31	,	,	PUNCT
hjic-847	200	32	r.	r.	PROPN
hjic-847	200	33	;	;	PUNCT
hjic-847	200	34	torralba	torralba	PROPN
hjic-847	200	35	,	,	PUNCT
hjic-847	200	36	a.	a.	NOUN
hjic-847	200	37	:	:	PUNCT
hjic-847	200	38	recognizing	recognize	VERB
hjic-847	200	39	and	and	CCONJ
hjic-847	200	40	learning	learn	VERB
hjic-847	200	41	object	object	NOUN
hjic-847	200	42	categories	category	NOUN
hjic-847	200	43	,	,	PUNCT
hjic-847	200	44	2007	2007	NUM
hjic-847	200	45	ieee	ieee	PROPN
hjic-847	200	46	computer	computer	NOUN
hjic-847	200	47	society	society	PROPN
hjic-847	200	48	conference	conference	NOUN
hjic-847	200	49	on	on	ADP
hjic-847	200	50	computer	computer	NOUN
hjic-847	200	51	vision	vision	NOUN
hjic-847	200	52	and	and	CCONJ
hjic-847	200	53	pattern	pattern	NOUN
hjic-847	200	54	recognition	recognition	NOUN
hjic-847	200	55	,	,	PUNCT
hjic-847	200	56	short	short	ADJ
hjic-847	200	57	course	course	NOUN
hjic-847	200	58	,	,	PUNCT
hjic-847	200	59	2007	2007	NUM
hjic-847	200	60	http://	http://	PROPN
hjic-847	200	61	people.csail.mit.edu/torralba/shortcourserloc/	people.csail.mit.edu/torralba/shortcourserloc/	X
hjic-847	200	62	[	[	X
hjic-847	200	63	12	12	NUM
hjic-847	200	64	]	]	PUNCT
hjic-847	200	65	lazebnik	lazebnik	X
hjic-847	200	66	,	,	PUNCT
hjic-847	200	67	s.	s.	PROPN
hjic-847	200	68	;	;	PUNCT
hjic-847	200	69	schmid	schmid	PROPN
hjic-847	200	70	,	,	PUNCT
hjic-847	200	71	c.	c.	PROPN
hjic-847	200	72	;	;	PUNCT
hjic-847	200	73	ponce	ponce	PROPN
hjic-847	200	74	,	,	PUNCT
hjic-847	200	75	j.	j.	PROPN
hjic-847	200	76	:	:	PUNCT
hjic-847	200	77	beyond	beyond	ADP
hjic-847	200	78	bags	bag	NOUN
hjic-847	200	79	of	of	ADP
hjic-847	200	80	features	feature	NOUN
hjic-847	200	81	:	:	PUNCT
hjic-847	200	82	spatial	spatial	ADJ
hjic-847	200	83	pyramid	pyramid	NOUN
hjic-847	200	84	matching	match	VERB
hjic-847	200	85	for	for	ADP
hjic-847	200	86	recognizing	recognize	VERB
hjic-847	200	87	natural	natural	ADJ
hjic-847	200	88	scene	scene	NOUN
hjic-847	200	89	categories	category	NOUN
hjic-847	200	90	,	,	PUNCT
hjic-847	200	91	2006	2006	NUM
hjic-847	200	92	ieee	ieee	NOUN
hjic-847	200	93	conference	conference	NOUN
hjic-847	200	94	on	on	ADP
hjic-847	200	95	computer	computer	NOUN
hjic-847	200	96	vision	vision	NOUN
hjic-847	200	97	and	and	CCONJ
hjic-847	200	98	pattern	pattern	NOUN
hjic-847	200	99	recognition	recognition	NOUN
hjic-847	200	100	,	,	PUNCT
hjic-847	200	101	2006	2006	NUM
hjic-847	200	102	2	2	NUM
hjic-847	200	103	,	,	PUNCT
hjic-847	200	104	pp	pp	ADJ
hjic-847	200	105	.	.	PUNCT
hjic-847	201	1	2169–2178	2169–2178	NUM
hjic-847	201	2	isbn	isbn	ADJ
hjic-847	201	3	:	:	PUNCT
hjic-847	201	4	0	0	NUM
hjic-847	201	5	-	-	SYM
hjic-847	201	6	7695	7695	NUM
hjic-847	201	7	-	-	PUNCT
hjic-847	201	8	2597	2597	NUM
hjic-847	201	9	-	-	SYM
hjic-847	201	10	0	0	NUM
hjic-847	201	11	doi	doi	NOUN
hjic-847	201	12	:	:	PUNCT
hjic-847	201	13	10.1109	10.1109	NUM
hjic-847	201	14	/	/	SYM
hjic-847	201	15	cvpr.2006.68	cvpr.2006.68	NOUN
hjic-847	201	16	[	[	X
hjic-847	201	17	13	13	NUM
hjic-847	201	18	]	]	X
hjic-847	201	19	lowe	lowe	PROPN
hjic-847	201	20	,	,	PUNCT
hjic-847	201	21	d.	d.	PROPN
hjic-847	201	22	g.	g.	PROPN
hjic-847	201	23	:	:	PUNCT
hjic-847	201	24	distinctive	distinctive	ADJ
hjic-847	201	25	image	image	NOUN
hjic-847	201	26	features	feature	NOUN
hjic-847	201	27	from	from	ADP
hjic-847	201	28	scaleinvariant	scaleinvariant	ADJ
hjic-847	201	29	keypoints	keypoint	NOUN
hjic-847	201	30	,	,	PUNCT
hjic-847	201	31	int	int	NOUN
hjic-847	201	32	.	.	PUNCT
hjic-847	202	1	j.	j.	PROPN
hjic-847	202	2	comput	comput	PROPN
hjic-847	202	3	.	.	PUNCT
hjic-847	203	1	vision	vision	PROPN
hjic-847	203	2	,	,	PUNCT
hjic-847	203	3	2004	2004	NUM
hjic-847	203	4	60(2	60(2	NUM
hjic-847	203	5	)	)	PUNCT
hjic-847	203	6	,	,	PUNCT
hjic-847	203	7	91–110	91–110	PROPN
hjic-847	203	8	doi	doi	NOUN
hjic-847	203	9	:	:	PUNCT
hjic-847	203	10	10.1023	10.1023	NUM
hjic-847	203	11	/	/	SYM
hjic-847	203	12	b	b	NOUN
hjic-847	203	13	:	:	PUNCT
hjic-847	203	14	visi.0000029664.99615.94	visi.0000029664.99615.94	PROPN
hjic-847	203	15	[	[	X
hjic-847	203	16	14	14	NUM
hjic-847	203	17	]	]	X
hjic-847	203	18	reynolds	reynolds	PROPN
hjic-847	203	19	,	,	PUNCT
hjic-847	203	20	d.	d.	PROPN
hjic-847	203	21	a.	a.	PROPN
hjic-847	203	22	:	:	PUNCT
hjic-847	203	23	gaussian	gaussian	ADJ
hjic-847	203	24	mixture	mixture	NOUN
hjic-847	203	25	models	model	NOUN
hjic-847	203	26	,	,	PUNCT
hjic-847	203	27	in	in	ADP
hjic-847	203	28	:	:	PUNCT
hjic-847	203	29	li	li	PROPN
hjic-847	203	30	,	,	PUNCT
hjic-847	203	31	s.	s.	PROPN
hjic-847	203	32	z.	z.	PROPN
hjic-847	203	33	;	;	PUNCT
hjic-847	203	34	(	(	PUNCT
hjic-847	203	35	ed	ed	NOUN
hjic-847	203	36	.	.	PUNCT
hjic-847	203	37	):	):	PUNCT
hjic-847	203	38	encyclopedia	encyclopedia	NOUN
hjic-847	203	39	of	of	ADP
hjic-847	203	40	biometric	biometric	ADJ
hjic-847	203	41	recognition	recognition	NOUN
hjic-847	203	42	,	,	PUNCT
hjic-847	203	43	1st	1st	PROPN
hjic-847	203	44	ed	ed	NOUN
hjic-847	203	45	.	.	PROPN
hjic-847	203	46	,	,	PUNCT
hjic-847	203	47	(	(	PUNCT
hjic-847	203	48	springer	springer	NOUN
hjic-847	203	49	,	,	PUNCT
hjic-847	203	50	boston	boston	PROPN
hjic-847	203	51	,	,	PUNCT
hjic-847	203	52	usa	usa	PROPN
hjic-847	203	53	)	)	PUNCT
hjic-847	203	54	2009	2009	NUM
hjic-847	203	55	,	,	PUNCT
hjic-847	203	56	pp	pp	ADV
hjic-847	203	57	.	.	PUNCT
hjic-847	204	1	659–663	659–663	NUM
hjic-847	204	2	isbn	isbn	NOUN
hjic-847	204	3	:	:	PUNCT
hjic-847	204	4	978	978	NUM
hjic-847	204	5	-	-	SYM
hjic-847	204	6	0	0	NUM
hjic-847	204	7	-	-	PUNCT
hjic-847	204	8	387	387	NUM
hjic-847	204	9	-	-	PUNCT
hjic-847	204	10	73003	73003	NUM
hjic-847	204	11	-	-	SYM
hjic-847	204	12	5	5	NUM
hjic-847	204	13	doi	doi	NOUN
hjic-847	204	14	:	:	PUNCT
hjic-847	204	15	10.1007/978	10.1007/978	NUM
hjic-847	204	16	-	-	SYM
hjic-847	204	17	14899	14899	NUM
hjic-847	204	18	-	-	PUNCT
hjic-847	204	19	7488	7488	NUM
hjic-847	204	20	-	-	SYM
hjic-847	204	21	4_196	4_196	NUM
hjic-847	204	22	[	[	X
hjic-847	204	23	15	15	NUM
hjic-847	204	24	]	]	X
hjic-847	204	25	tomasi	tomasi	NOUN
hjic-847	204	26	,	,	PUNCT
hjic-847	204	27	c.	c.	NOUN
hjic-847	204	28	:	:	PUNCT
hjic-847	204	29	estimating	estimate	VERB
hjic-847	204	30	gaussian	gaussian	ADJ
hjic-847	204	31	mixture	mixture	NOUN
hjic-847	204	32	densities	density	NOUN
hjic-847	204	33	with	with	ADP
hjic-847	204	34	em	em	PRON
hjic-847	204	35	:	:	PUNCT
hjic-847	204	36	a	a	DET
hjic-847	204	37	tutorial	tutorial	NOUN
hjic-847	204	38	(	(	PUNCT
hjic-847	204	39	tech	tech	NOUN
hjic-847	204	40	.	.	PUNCT
hjic-847	205	1	rep	rep	AUX
hjic-847	205	2	.	.	PROPN
hjic-847	205	3	,	,	PUNCT
hjic-847	205	4	duke	duke	PROPN
hjic-847	205	5	university	university	PROPN
hjic-847	205	6	)	)	PUNCT
hjic-847	205	7	2004	2004	NUM
hjic-847	205	8	https://www2.cs.duke.edu/courses/	https://www2.cs.duke.edu/courses/	PROPN
hjic-847	205	9	spring04	spring04	NOUN
hjic-847	205	10	/	/	SYM
hjic-847	205	11	cps196.1	cps196.1	PROPN
hjic-847	205	12	/	/	SYM
hjic-847	205	13	handouts	handout	NOUN
hjic-847	205	14	/	/	SYM
hjic-847	205	15	em	em	NOUN
hjic-847	205	16	/	/	SYM
hjic-847	205	17	tomasiem.pdf	tomasiem.pdf	X
hjic-847	205	18	[	[	X
hjic-847	205	19	16	16	NUM
hjic-847	205	20	]	]	X
hjic-847	205	21	browne	browne	PROPN
hjic-847	205	22	,	,	PUNCT
hjic-847	205	23	r.	r.	PROPN
hjic-847	205	24	p.	p.	PROPN
hjic-847	205	25	;	;	PUNCT
hjic-847	205	26	mcnicholas	mcnichola	NOUN
hjic-847	205	27	,	,	PUNCT
hjic-847	205	28	p.	p.	PROPN
hjic-847	205	29	d.	d.	PROPN
hjic-847	205	30	;	;	PUNCT
hjic-847	205	31	sparling	sparling	PROPN
hjic-847	205	32	,	,	PUNCT
hjic-847	205	33	m.	m.	NOUN
hjic-847	205	34	d.	d.	PROPN
hjic-847	205	35	:	:	PUNCT
hjic-847	205	36	model	model	NOUN
hjic-847	205	37	-	-	PUNCT
hjic-847	205	38	based	base	VERB
hjic-847	205	39	learning	learning	NOUN
hjic-847	205	40	using	use	VERB
hjic-847	205	41	a	a	DET
hjic-847	205	42	mixture	mixture	NOUN
hjic-847	205	43	of	of	ADP
hjic-847	205	44	mixtures	mixture	NOUN
hjic-847	205	45	of	of	ADP
hjic-847	205	46	gaussian	gaussian	ADJ
hjic-847	205	47	and	and	CCONJ
hjic-847	205	48	uniform	uniform	ADJ
hjic-847	205	49	distributions	distribution	NOUN
hjic-847	205	50	,	,	PUNCT
hjic-847	205	51	ieee	ieee	NOUN
hjic-847	205	52	t.	t.	NOUN
hjic-847	205	53	pattern	pattern	NOUN
hjic-847	205	54	anal	anal	PROPN
hjic-847	205	55	.	.	PUNCT
hjic-847	205	56	,	,	PUNCT
hjic-847	205	57	2012	2012	NUM
hjic-847	205	58	34(4	34(4	NOUN
hjic-847	205	59	)	)	PUNCT
hjic-847	205	60	,	,	PUNCT
hjic-847	205	61	814–817	814–817	NUM
hjic-847	205	62	doi	doi	NOUN
hjic-847	205	63	:	:	PUNCT
hjic-847	205	64	10.1109	10.1109	NUM
hjic-847	205	65	/	/	SYM
hjic-847	205	66	tpami.2011.199	tpami.2011.199	NOUN
hjic-847	205	67	[	[	X
hjic-847	205	68	17	17	NUM
hjic-847	205	69	]	]	SYM
hjic-847	205	70	perronnin	perronnin	VERB
hjic-847	205	71	,	,	PUNCT
hjic-847	205	72	f.	f.	PROPN
hjic-847	205	73	;	;	PUNCT
hjic-847	205	74	dance	dance	PROPN
hjic-847	205	75	,	,	PUNCT
hjic-847	205	76	c.	c.	PROPN
hjic-847	205	77	:	:	PUNCT
hjic-847	205	78	fisher	fisher	PROPN
hjic-847	205	79	kernel	kernel	PROPN
hjic-847	205	80	on	on	ADP
hjic-847	205	81	visual	visual	ADJ
hjic-847	205	82	vocabularies	vocabulary	NOUN
hjic-847	205	83	for	for	ADP
hjic-847	205	84	image	image	NOUN
hjic-847	205	85	categorization	categorization	NOUN
hjic-847	205	86	,	,	PUNCT
hjic-847	205	87	2007	2007	NUM
hjic-847	205	88	ieee	ieee	NOUN
hjic-847	205	89	conference	conference	NOUN
hjic-847	205	90	on	on	ADP
hjic-847	205	91	computer	computer	NOUN
hjic-847	205	92	vision	vision	NOUN
hjic-847	205	93	and	and	CCONJ
hjic-847	205	94	pattern	pattern	NOUN
hjic-847	205	95	hungarian	hungarian	ADJ
hjic-847	205	96	journal	journal	NOUN
hjic-847	205	97	of	of	ADP
hjic-847	205	98	industry	industry	NOUN
hjic-847	205	99	and	and	CCONJ
hjic-847	205	100	chemistry	chemistry	NOUN
hjic-847	205	101	https://doi.org/10.1109/cvpr.2009.5206594	https://doi.org/10.1109/cvpr.2009.5206594	PROPN
hjic-847	205	102	https://doi.org/10.1109/cvpr.2009.5206594	https://doi.org/10.1109/cvpr.2009.5206594	PROPN
hjic-847	205	103	https://doi.org/10.1109/tpami.2012.256	https://doi.org/10.1109/tpami.2012.256	ADJ
hjic-847	205	104	https://doi.org/10.1109/tpami.2012.256	https://doi.org/10.1109/tpami.2012.256	VERB
hjic-847	205	105	https://doi.org/10.1109/tpami.2014.2321392	https://doi.org/10.1109/tpami.2014.2321392	X
hjic-847	205	106	https://doi.org/10.1109/tpami.2014.2321392	https://doi.org/10.1109/tpami.2014.2321392	PROPN
hjic-847	205	107	https://doi.org/10.1109/cvpr.2015.7298799	https://doi.org/10.1109/cvpr.2015.7298799	PROPN
hjic-847	205	108	https://doi.org/10.15388/informatica.2018.171	https://doi.org/10.15388/informatica.2018.171	PROPN
hjic-847	205	109	https://doi.org/10.15388/informatica.2018.171	https://doi.org/10.15388/informatica.2018.171	PROPN
hjic-847	205	110	https://doi.org/10.1023/b:mach.0000008084.60811.49	https://doi.org/10.1023/b:mach.0000008084.60811.49	PRON
hjic-847	205	111	https://doi.org/10.1023/b:mach.0000008084.60811.49	https://doi.org/10.1023/b:mach.0000008084.60811.49	PROPN
hjic-847	205	112	https://doi.org/10.1109/cvpr.2012.6248047	https://doi.org/10.1109/cvpr.2012.6248047	PROPN
hjic-847	205	113	https://doi.org/10.1162/089976601750264965	https://doi.org/10.1162/089976601750264965	PART
hjic-847	205	114	https://doi.org/10.1162/089976601750264965	https://doi.org/10.1162/089976601750264965	NOUN
hjic-847	205	115	https://doi.org/10.5244/c.20.123	https://doi.org/10.5244/c.20.123	PROPN
hjic-847	205	116	https://doi.org/10.1007/978-3-319-10578-9_26	https://doi.org/10.1007/978-3-319-10578-9_26	PROPN
hjic-847	205	117	https://doi.org/10.1007/978-3-319-10578-9_26	https://doi.org/10.1007/978-3-319-10578-9_26	PROPN
hjic-847	205	118	http://people.csail.mit.edu/torralba/shortcourserloc/	http://people.csail.mit.edu/torralba/shortcourserloc/	NUM
hjic-847	205	119	http://people.csail.mit.edu/torralba/shortcourserloc/	http://people.csail.mit.edu/torralba/shortcourserloc/	PROPN
hjic-847	205	120	https://doi.org/10.1109/cvpr.2006.68	https://doi.org/10.1109/cvpr.2006.68	PROPN
hjic-847	205	121	https://doi.org/10.1109/cvpr.2006.68	https://doi.org/10.1109/cvpr.2006.68	PROPN
hjic-847	205	122	https://doi.org/10.1023/b:visi.0000029664.99615.94	https://doi.org/10.1023/b:visi.0000029664.99615.94	VERB
hjic-847	205	123	https://doi.org/10.1007/978-1-4899-7488-4_196	https://doi.org/10.1007/978-1-4899-7488-4_196	ADJ
hjic-847	205	124	https://doi.org/10.1007/978-1-4899-7488-4_196	https://doi.org/10.1007/978-1-4899-7488-4_196	NOUN
hjic-847	205	125	https://www2.cs.duke.edu/courses/spring04/cps196.1/handouts/em/tomasiem.pdf	https://www2.cs.duke.edu/courses/spring04/cps196.1/handouts/em/tomasiem.pdf	NOUN
hjic-847	205	126	https://www2.cs.duke.edu/courses/spring04/cps196.1/handouts/em/tomasiem.pdf	https://www2.cs.duke.edu/courses/spring04/cps196.1/handouts/em/tomasiem.pdf	NOUN
hjic-847	205	127	https://doi.org/10.1109/tpami.2011.199	https://doi.org/10.1109/tpami.2011.199	PROPN
hjic-847	205	128	https://doi.org/10.1109/tpami.2011.199	https://doi.org/10.1109/tpami.2011.199	VERB
hjic-847	205	129	automated	automate	VERB
hjic-847	205	130	labeling	labeling	NOUN
hjic-847	205	131	process	process	NOUN
hjic-847	205	132	for	for	ADP
hjic-847	205	133	unknown	unknown	ADJ
hjic-847	205	134	images	image	NOUN
hjic-847	205	135	39	39	NUM
hjic-847	205	136	recognition	recognition	NOUN
hjic-847	205	137	,	,	PUNCT
hjic-847	205	138	2007	2007	NUM
hjic-847	205	139	,	,	PUNCT
hjic-847	205	140	pp	pp	ADV
hjic-847	205	141	.	.	PUNCT
hjic-847	206	1	1–8	1–8	X
hjic-847	206	2	isbn	isbn	NOUN
hjic-847	206	3	:	:	PUNCT
hjic-847	206	4	1	1	NUM
hjic-847	206	5	-	-	PUNCT
hjic-847	206	6	4244	4244	NUM
hjic-847	206	7	-	-	PUNCT
hjic-847	206	8	1179	1179	NUM
hjic-847	206	9	-	-	SYM
hjic-847	206	10	3	3	NUM
hjic-847	206	11	doi	doi	NOUN
hjic-847	206	12	:	:	PUNCT
hjic-847	206	13	10.1109	10.1109	NUM
hjic-847	206	14	/	/	SYM
hjic-847	206	15	cvpr.2007.383266	cvpr.2007.383266	X
hjic-847	206	16	[	[	X
hjic-847	206	17	18	18	NUM
hjic-847	206	18	]	]	SYM
hjic-847	206	19	perronnin	perronnin	VERB
hjic-847	206	20	,	,	PUNCT
hjic-847	206	21	f.	f.	PROPN
hjic-847	206	22	;	;	PUNCT
hjic-847	206	23	sánchez	sánchez	PROPN
hjic-847	206	24	,	,	PUNCT
hjic-847	206	25	j.	j.	PROPN
hjic-847	206	26	;	;	PUNCT
hjic-847	206	27	mensink	mensink	PROPN
hjic-847	206	28	,	,	PUNCT
hjic-847	206	29	t.	t.	PROPN
hjic-847	206	30	:	:	PUNCT
hjic-847	206	31	improving	improve	VERB
hjic-847	206	32	the	the	DET
hjic-847	206	33	fisher	fisher	PROPN
hjic-847	206	34	kernel	kernel	NOUN
hjic-847	206	35	for	for	ADP
hjic-847	206	36	large	large	ADJ
hjic-847	206	37	-	-	PUNCT
hjic-847	206	38	scale	scale	NOUN
hjic-847	206	39	image	image	NOUN
hjic-847	206	40	classification	classification	NOUN
hjic-847	206	41	,	,	PUNCT
hjic-847	206	42	in	in	ADP
hjic-847	206	43	:	:	PUNCT
hjic-847	206	44	daniilidis	daniilidi	NOUN
hjic-847	206	45	,	,	PUNCT
hjic-847	206	46	k	k	X
hjic-847	206	47	;	;	PUNCT
hjic-847	206	48	maragos	marago	NOUN
hjic-847	206	49	,	,	PUNCT
hjic-847	206	50	p.	p.	NOUN
hjic-847	206	51	;	;	PUNCT
hjic-847	206	52	paragios	paragio	NOUN
hjic-847	206	53	,	,	PUNCT
hjic-847	206	54	n.	n.	NOUN
hjic-847	206	55	;	;	PUNCT
hjic-847	206	56	(	(	PUNCT
hjic-847	206	57	eds	ed	NOUN
hjic-847	206	58	.	.	PROPN
hjic-847	206	59	):	):	PUNCT
hjic-847	206	60	computer	computer	NOUN
hjic-847	206	61	vision	vision	NOUN
hjic-847	206	62	–	–	PUNCT
hjic-847	206	63	eccv	eccv	ADV
hjic-847	206	64	2010	2010	NUM
hjic-847	206	65	.	.	PUNCT
hjic-847	206	66	,	,	PUNCT
hjic-847	206	67	eccv	eccv	ADV
hjic-847	206	68	2010	2010	NUM
hjic-847	206	69	.	.	PUNCT
hjic-847	207	1	lecture	lecture	NOUN
hjic-847	207	2	notes	note	NOUN
hjic-847	207	3	in	in	ADP
hjic-847	207	4	computer	computer	NOUN
hjic-847	207	5	science	science	NOUN
hjic-847	207	6	,	,	PUNCT
hjic-847	207	7	6314	6314	NUM
hjic-847	207	8	(	(	PUNCT
hjic-847	207	9	springer	springer	NOUN
hjic-847	207	10	,	,	PUNCT
hjic-847	207	11	berlin	berlin	PROPN
hjic-847	207	12	,	,	PUNCT
hjic-847	207	13	germany	germany	PROPN
hjic-847	207	14	)	)	PUNCT
hjic-847	207	15	2010	2010	NUM
hjic-847	207	16	,	,	PUNCT
hjic-847	207	17	pp	pp	ADP
hjic-847	207	18	.	.	PUNCT
hjic-847	208	1	143–156	143–156	NUM
hjic-847	208	2	isbn	isbn	NOUN
hjic-847	208	3	:	:	PUNCT
hjic-847	208	4	978	978	NUM
hjic-847	208	5	-	-	SYM
hjic-847	208	6	3	3	NUM
hjic-847	208	7	-	-	PUNCT
hjic-847	208	8	642	642	NUM
hjic-847	208	9	-	-	PUNCT
hjic-847	208	10	15560	15560	NUM
hjic-847	208	11	-	-	SYM
hjic-847	208	12	4	4	NUM
hjic-847	208	13	doi	doi	NOUN
hjic-847	208	14	:	:	PUNCT
hjic-847	208	15	10.1007/978	10.1007/978	NUM
hjic-847	208	16	-	-	SYM
hjic-847	208	17	3	3	NUM
hjic-847	208	18	-	-	PUNCT
hjic-847	208	19	64215561	64215561	NUM
hjic-847	208	20	-	-	SYM
hjic-847	208	21	1_11	1_11	NUM
hjic-847	208	22	[	[	X
hjic-847	208	23	19	19	NUM
hjic-847	208	24	]	]	SYM
hjic-847	208	25	macqueen	macqueen	PROPN
hjic-847	208	26	,	,	PUNCT
hjic-847	208	27	j.	j.	PROPN
hjic-847	208	28	:	:	PUNCT
hjic-847	208	29	some	some	DET
hjic-847	208	30	methods	method	NOUN
hjic-847	208	31	for	for	ADP
hjic-847	208	32	classification	classification	NOUN
hjic-847	208	33	and	and	CCONJ
hjic-847	208	34	analysis	analysis	NOUN
hjic-847	208	35	of	of	ADP
hjic-847	208	36	multivariate	multivariate	NOUN
hjic-847	208	37	observations	observation	NOUN
hjic-847	208	38	,	,	PUNCT
hjic-847	208	39	in	in	ADP
hjic-847	208	40	:	:	PUNCT
hjic-847	208	41	le	le	X
hjic-847	208	42	cam	cam	PROPN
hjic-847	208	43	,	,	PUNCT
hjic-847	208	44	l.	l.	PROPN
hjic-847	208	45	m.	m.	PROPN
hjic-847	208	46	;	;	PUNCT
hjic-847	208	47	neyman	neyman	PROPN
hjic-847	208	48	,	,	PUNCT
hjic-847	208	49	j.	j.	PROPN
hjic-847	208	50	;	;	PUNCT
hjic-847	208	51	(	(	PUNCT
hjic-847	208	52	eds	ed	NOUN
hjic-847	208	53	.	.	PUNCT
hjic-847	208	54	):	):	PUNCT
hjic-847	208	55	proceedings	proceeding	NOUN
hjic-847	208	56	of	of	ADP
hjic-847	208	57	the	the	DET
hjic-847	208	58	fifth	fifth	ADJ
hjic-847	208	59	berkeley	berkeley	PROPN
hjic-847	208	60	symposium	symposium	NOUN
hjic-847	208	61	on	on	ADP
hjic-847	208	62	mathematical	mathematical	ADJ
hjic-847	208	63	statistics	statistic	NOUN
hjic-847	208	64	and	and	CCONJ
hjic-847	208	65	probability	probability	NOUN
hjic-847	208	66	,	,	PUNCT
hjic-847	208	67	1	1	NUM
hjic-847	208	68	(	(	PUNCT
hjic-847	208	69	university	university	NOUN
hjic-847	208	70	of	of	ADP
hjic-847	208	71	california	california	PROPN
hjic-847	208	72	press	press	PROPN
hjic-847	208	73	,	,	PUNCT
hjic-847	208	74	berkeley	berkeley	PROPN
hjic-847	208	75	,	,	PUNCT
hjic-847	208	76	usa	usa	PROPN
hjic-847	208	77	)	)	PUNCT
hjic-847	208	78	1967	1967	NUM
hjic-847	208	79	pp	pp	NOUN
hjic-847	208	80	.	.	PUNCT
hjic-847	209	1	281–297	281–297	NUM
hjic-847	209	2	[	[	SYM
hjic-847	209	3	20	20	NUM
hjic-847	209	4	]	]	X
hjic-847	209	5	chitta	chitta	PROPN
hjic-847	209	6	,	,	PUNCT
hjic-847	209	7	r.	r.	PROPN
hjic-847	209	8	;	;	PUNCT
hjic-847	209	9	jin	jin	PROPN
hjic-847	209	10	,	,	PUNCT
hjic-847	209	11	r.	r.	PROPN
hjic-847	209	12	;	;	PUNCT
hjic-847	209	13	havens	haven	NOUN
hjic-847	209	14	,	,	PUNCT
hjic-847	209	15	t.	t.	PROPN
hjic-847	209	16	c.	c.	PROPN
hjic-847	209	17	;	;	PUNCT
hjic-847	209	18	jain	jain	PROPN
hjic-847	209	19	,	,	PUNCT
hjic-847	209	20	a.	a.	PROPN
hjic-847	209	21	k.	k.	PROPN
hjic-847	209	22	:	:	PUNCT
hjic-847	209	23	approximate	approximate	ADJ
hjic-847	209	24	kernel	kernel	PROPN
hjic-847	209	25	k	k	X
hjic-847	209	26	-	-	PUNCT
hjic-847	209	27	means	mean	NOUN
hjic-847	209	28	:	:	PUNCT
hjic-847	209	29	solution	solution	NOUN
hjic-847	209	30	to	to	ADP
hjic-847	209	31	large	large	ADJ
hjic-847	209	32	scale	scale	NOUN
hjic-847	209	33	kernel	kernel	NOUN
hjic-847	209	34	clustering	clustering	NOUN
hjic-847	209	35	,	,	PUNCT
hjic-847	209	36	in	in	ADP
hjic-847	209	37	:	:	PUNCT
hjic-847	209	38	proceedings	proceeding	NOUN
hjic-847	209	39	of	of	ADP
hjic-847	209	40	the	the	DET
hjic-847	209	41	17th	17th	ADJ
hjic-847	209	42	acm	acm	PROPN
hjic-847	209	43	sigkdd	sigkdd	NOUN
hjic-847	209	44	international	international	ADJ
hjic-847	209	45	conference	conference	NOUN
hjic-847	209	46	on	on	ADP
hjic-847	209	47	knowledge	knowledge	NOUN
hjic-847	209	48	discovery	discovery	PROPN
hjic-847	209	49	and	and	CCONJ
hjic-847	209	50	data	datum	NOUN
hjic-847	209	51	mining	mining	NOUN
hjic-847	209	52	,	,	PUNCT
hjic-847	209	53	(	(	PUNCT
hjic-847	209	54	acm	acm	PROPN
hjic-847	209	55	,	,	PUNCT
hjic-847	209	56	new	new	PROPN
hjic-847	209	57	york	york	PROPN
hjic-847	209	58	,	,	PUNCT
hjic-847	209	59	usa	usa	PROPN
hjic-847	209	60	)	)	PUNCT
hjic-847	209	61	2011	2011	NUM
hjic-847	209	62	,	,	PUNCT
hjic-847	209	63	pp	pp	ADV
hjic-847	209	64	.	.	PUNCT
hjic-847	210	1	895–903	895–903	NUM
hjic-847	210	2	isbn	isbn	NOUN
hjic-847	210	3	:	:	PUNCT
hjic-847	210	4	978	978	NUM
hjic-847	210	5	-	-	SYM
hjic-847	210	6	1	1	NUM
hjic-847	210	7	-	-	PUNCT
hjic-847	210	8	4503	4503	NUM
hjic-847	210	9	-	-	PUNCT
hjic-847	210	10	0813	0813	NUM
hjic-847	210	11	-	-	PUNCT
hjic-847	210	12	7	7	NUM
hjic-847	210	13	doi	doi	NOUN
hjic-847	210	14	:	:	PUNCT
hjic-847	210	15	10.1145/2020408.2020558	10.1145/2020408.2020558	NUM
hjic-847	210	16	[	[	X
hjic-847	210	17	21	21	NUM
hjic-847	210	18	]	]	SYM
hjic-847	210	19	dhillon	dhillon	PROPN
hjic-847	210	20	,	,	PUNCT
hjic-847	210	21	i.	i.	PROPN
hjic-847	210	22	s.	s.	PROPN
hjic-847	210	23	;	;	PUNCT
hjic-847	210	24	guan	guan	PROPN
hjic-847	210	25	,	,	PUNCT
hjic-847	210	26	y.	y.	PROPN
hjic-847	210	27	;	;	PUNCT
hjic-847	210	28	kulis	kulis	PROPN
hjic-847	210	29	,	,	PUNCT
hjic-847	210	30	b.	b.	PROPN
hjic-847	210	31	;	;	PUNCT
hjic-847	210	32	kernel	kernel	PROPN
hjic-847	210	33	k	k	PROPN
hjic-847	210	34	-	-	PUNCT
hjic-847	210	35	means	means	ADV
hjic-847	210	36	:	:	PUNCT
hjic-847	210	37	spectral	spectral	ADJ
hjic-847	210	38	clustering	clustering	NOUN
hjic-847	210	39	,	,	PUNCT
hjic-847	210	40	and	and	CCONJ
hjic-847	210	41	normalized	normalize	VERB
hjic-847	210	42	cuts	cut	NOUN
hjic-847	210	43	,	,	PUNCT
hjic-847	210	44	in	in	ADP
hjic-847	210	45	:	:	PUNCT
hjic-847	210	46	proceedings	proceeding	NOUN
hjic-847	210	47	of	of	ADP
hjic-847	210	48	the	the	DET
hjic-847	210	49	10th	10th	ADJ
hjic-847	210	50	acm	acm	PROPN
hjic-847	210	51	sigkdd	sigkdd	NOUN
hjic-847	210	52	international	international	ADJ
hjic-847	210	53	conference	conference	NOUN
hjic-847	210	54	on	on	ADP
hjic-847	210	55	knowledge	knowledge	NOUN
hjic-847	210	56	discovery	discovery	PROPN
hjic-847	210	57	and	and	CCONJ
hjic-847	210	58	data	datum	NOUN
hjic-847	210	59	mining	mining	NOUN
hjic-847	210	60	,	,	PUNCT
hjic-847	210	61	(	(	PUNCT
hjic-847	210	62	acm	acm	PROPN
hjic-847	210	63	,	,	PUNCT
hjic-847	210	64	new	new	PROPN
hjic-847	210	65	york	york	PROPN
hjic-847	210	66	,	,	PUNCT
hjic-847	210	67	usa	usa	PROPN
hjic-847	210	68	)	)	PUNCT
hjic-847	210	69	2004	2004	NUM
hjic-847	210	70	,	,	PUNCT
hjic-847	210	71	pp	pp	ADJ
hjic-847	210	72	.	.	PUNCT
hjic-847	211	1	551	551	NUM
hjic-847	211	2	–	–	PUNCT
hjic-847	211	3	556	556	NUM
hjic-847	211	4	isbn	isbn	NOUN
hjic-847	211	5	:	:	PUNCT
hjic-847	211	6	1	1	NUM
hjic-847	211	7	-	-	SYM
hjic-847	211	8	58113	58113	NUM
hjic-847	211	9	-	-	PUNCT
hjic-847	211	10	888	888	NUM
hjic-847	211	11	-	-	SYM
hjic-847	211	12	1	1	NUM
hjic-847	211	13	doi	doi	NOUN
hjic-847	211	14	:	:	PUNCT
hjic-847	211	15	10.1145/1014052.1014118	10.1145/1014052.1014118	NUM
hjic-847	212	1	[	[	X
hjic-847	212	2	22	22	NUM
hjic-847	212	3	]	]	X
hjic-847	212	4	papp	papp	NOUN
hjic-847	212	5	,	,	PUNCT
hjic-847	212	6	d.	d.	PROPN
hjic-847	212	7	;	;	PUNCT
hjic-847	212	8	szűcs	szűcs	PROPN
hjic-847	212	9	,	,	PUNCT
hjic-847	212	10	g.	g.	PROPN
hjic-847	212	11	:	:	PUNCT
hjic-847	212	12	mmkk++	mmkk++	VERB
hjic-847	212	13	algorithm	algorithm	NOUN
hjic-847	212	14	for	for	ADP
hjic-847	212	15	clustering	cluster	VERB
hjic-847	212	16	heterogeneous	heterogeneous	ADJ
hjic-847	212	17	images	image	NOUN
hjic-847	212	18	into	into	ADP
hjic-847	212	19	an	an	DET
hjic-847	212	20	unknown	unknown	ADJ
hjic-847	212	21	number	number	NOUN
hjic-847	212	22	of	of	ADP
hjic-847	212	23	clusters	cluster	NOUN
hjic-847	212	24	,	,	PUNCT
hjic-847	212	25	elcvia	elcvia	PROPN
hjic-847	212	26	:	:	PUNCT
hjic-847	212	27	electronic	electronic	ADJ
hjic-847	212	28	letters	letter	NOUN
hjic-847	212	29	on	on	ADP
hjic-847	212	30	computer	computer	NOUN
hjic-847	212	31	vision	vision	NOUN
hjic-847	212	32	and	and	CCONJ
hjic-847	212	33	image	image	NOUN
hjic-847	212	34	analysis	analysis	NOUN
hjic-847	212	35	,	,	PUNCT
hjic-847	212	36	2017	2017	NUM
hjic-847	212	37	16(3	16(3	NOUN
hjic-847	212	38	)	)	PUNCT
hjic-847	212	39	,	,	PUNCT
hjic-847	212	40	30–45	30–45	NUM
hjic-847	212	41	doi	doi	NOUN
hjic-847	212	42	:	:	PUNCT
hjic-847	212	43	10.5565	10.5565	NUM
hjic-847	212	44	/	/	SYM
hjic-847	212	45	rev	rev	PROPN
hjic-847	212	46	/	/	SYM
hjic-847	212	47	elcvia.1054	elcvia.1054	PROPN
hjic-847	213	1	[	[	X
hjic-847	213	2	23	23	NUM
hjic-847	213	3	]	]	X
hjic-847	213	4	arthur	arthur	PROPN
hjic-847	213	5	,	,	PUNCT
hjic-847	213	6	d.	d.	PROPN
hjic-847	213	7	;	;	PUNCT
hjic-847	213	8	vassilvitskii	vassilvitskii	PROPN
hjic-847	213	9	,	,	PUNCT
hjic-847	213	10	s.	s.	PROPN
hjic-847	213	11	:	:	PUNCT
hjic-847	213	12	k	k	X
hjic-847	213	13	-	-	PUNCT
hjic-847	213	14	means++	means++	NOUN
hjic-847	213	15	:	:	PUNCT
hjic-847	213	16	the	the	DET
hjic-847	213	17	advantages	advantage	NOUN
hjic-847	213	18	of	of	ADP
hjic-847	213	19	careful	careful	ADJ
hjic-847	213	20	seeding	seeding	NOUN
hjic-847	213	21	,	,	PUNCT
hjic-847	213	22	in	in	ADP
hjic-847	213	23	:	:	PUNCT
hjic-847	213	24	proceedings	proceeding	NOUN
hjic-847	213	25	of	of	ADP
hjic-847	213	26	the	the	DET
hjic-847	213	27	eighteenth	eighteenth	ADJ
hjic-847	213	28	annual	annual	ADJ
hjic-847	213	29	acm	acm	NOUN
hjic-847	213	30	-	-	PUNCT
hjic-847	213	31	siam	siam	NOUN
hjic-847	213	32	symposium	symposium	NOUN
hjic-847	213	33	on	on	ADP
hjic-847	213	34	discrete	discrete	ADJ
hjic-847	213	35	algorithms	algorithm	NOUN
hjic-847	213	36	,	,	PUNCT
hjic-847	213	37	(	(	PUNCT
hjic-847	213	38	society	society	NOUN
hjic-847	213	39	for	for	ADP
hjic-847	213	40	industrial	industrial	ADJ
hjic-847	213	41	and	and	CCONJ
hjic-847	213	42	applied	applied	ADJ
hjic-847	213	43	mathematics	mathematic	NOUN
hjic-847	213	44	,	,	PUNCT
hjic-847	213	45	philadelphia	philadelphia	PROPN
hjic-847	213	46	,	,	PUNCT
hjic-847	213	47	usa	usa	PROPN
hjic-847	213	48	)	)	PUNCT
hjic-847	213	49	2007	2007	NUM
hjic-847	213	50	,	,	PUNCT
hjic-847	213	51	pp	pp	ADJ
hjic-847	213	52	.	.	PUNCT
hjic-847	214	1	1027–1035	1027–1035	NUM
hjic-847	214	2	isbn	isbn	ADJ
hjic-847	214	3	:	:	PUNCT
hjic-847	214	4	978	978	NUM
hjic-847	214	5	-	-	SYM
hjic-847	214	6	0	0	NUM
hjic-847	214	7	-	-	PUNCT
hjic-847	214	8	898716	898716	NUM
hjic-847	214	9	-	-	SYM
hjic-847	214	10	24	24	NUM
hjic-847	214	11	-	-	SYM
hjic-847	214	12	5	5	NUM
hjic-847	214	13	[	[	SYM
hjic-847	214	14	24	24	NUM
hjic-847	214	15	]	]	X
hjic-847	214	16	fei	fei	PROPN
hjic-847	214	17	-	-	PUNCT
hjic-847	214	18	fei	fei	PROPN
hjic-847	214	19	,	,	PUNCT
hjic-847	214	20	l.	l.	PROPN
hjic-847	214	21	;	;	PUNCT
hjic-847	214	22	fergus	fergus	PROPN
hjic-847	214	23	,	,	PUNCT
hjic-847	214	24	r.	r.	PROPN
hjic-847	214	25	;	;	PUNCT
hjic-847	214	26	perona	perona	PROPN
hjic-847	214	27	,	,	PUNCT
hjic-847	214	28	p.	p.	NOUN
hjic-847	214	29	:	:	PUNCT
hjic-847	215	1	learning	learn	VERB
hjic-847	215	2	generative	generative	ADJ
hjic-847	215	3	visual	visual	ADJ
hjic-847	215	4	models	model	NOUN
hjic-847	215	5	from	from	ADP
hjic-847	215	6	few	few	ADJ
hjic-847	215	7	training	training	NOUN
hjic-847	215	8	examples	example	NOUN
hjic-847	215	9	:	:	PUNCT
hjic-847	215	10	an	an	DET
hjic-847	215	11	incremental	incremental	ADJ
hjic-847	215	12	bayesian	bayesian	NOUN
hjic-847	215	13	approach	approach	NOUN
hjic-847	215	14	tested	test	VERB
hjic-847	215	15	on	on	ADP
hjic-847	215	16	101	101	NUM
hjic-847	215	17	object	object	NOUN
hjic-847	215	18	categories	category	NOUN
hjic-847	215	19	,	,	PUNCT
hjic-847	215	20	comput	comput	NOUN
hjic-847	215	21	.	.	PUNCT
hjic-847	216	1	vis	vis	X
hjic-847	216	2	.	.	PUNCT
hjic-847	216	3	image	image	NOUN
hjic-847	216	4	und	und	PROPN
hjic-847	216	5	.	.	PUNCT
hjic-847	216	6	,	,	PUNCT
hjic-847	216	7	2007	2007	NUM
hjic-847	216	8	106(1	106(1	NUM
hjic-847	216	9	)	)	PUNCT
hjic-847	216	10	,	,	PUNCT
hjic-847	216	11	59–70	59–70	NUM
hjic-847	216	12	.	.	PUNCT
hjic-847	217	1	doi	doi	NOUN
hjic-847	217	2	:	:	PUNCT
hjic-847	217	3	10.1016	10.1016	NUM
hjic-847	217	4	/	/	SYM
hjic-847	217	5	j.cviu.2005.09.012	j.cviu.2005.09.012	NOUN
hjic-847	217	6	[	[	X
hjic-847	217	7	25	25	NUM
hjic-847	217	8	]	]	PUNCT
hjic-847	217	9	griffin	griffin	NOUN
hjic-847	217	10	,	,	PUNCT
hjic-847	217	11	g.	g.	PROPN
hjic-847	217	12	;	;	PUNCT
hjic-847	217	13	holub	holub	PROPN
hjic-847	217	14	,	,	PUNCT
hjic-847	217	15	a.	a.	NOUN
hjic-847	217	16	;	;	PUNCT
hjic-847	217	17	perona	perona	PROPN
hjic-847	217	18	,	,	PUNCT
hjic-847	217	19	p.	p.	NOUN
hjic-847	217	20	:	:	PUNCT
hjic-847	218	1	the	the	DET
hjic-847	218	2	caltech	caltech	PROPN
hjic-847	218	3	256	256	NUM
hjic-847	218	4	,	,	PUNCT
hjic-847	218	5	caltech	caltech	NOUN
hjic-847	218	6	,	,	PUNCT
hjic-847	218	7	tech	tech	NOUN
hjic-847	218	8	.	.	PUNCT
hjic-847	218	9	rep	rep	PROPN
hjic-847	218	10	.	.	PROPN
hjic-847	218	11	,	,	PUNCT
hjic-847	218	12	2012	2012	NUM
hjic-847	218	13	47(1	47(1	NUM
hjic-847	218	14	)	)	PUNCT
hjic-847	218	15	pp	pp	ADV
hjic-847	218	16	.	.	PUNCT
hjic-847	219	1	33–39	33–39	NUM
hjic-847	219	2	(	(	PUNCT
hjic-847	219	3	2019	2019	NUM
hjic-847	219	4	)	)	PUNCT
hjic-847	220	1	https://doi.org/10.1109/cvpr.2007.383266	https://doi.org/10.1109/cvpr.2007.383266	PROPN
hjic-847	220	2	https://doi.org/10.1109/cvpr.2007.383266	https://doi.org/10.1109/cvpr.2007.383266	PROPN
hjic-847	220	3	https://doi.org/10.1007/978-3-642-15561-1_11	https://doi.org/10.1007/978-3-642-15561-1_11	VERB
hjic-847	220	4	https://doi.org/10.1007/978-3-642-15561-1_11	https://doi.org/10.1007/978-3-642-15561-1_11	PROPN
hjic-847	220	5	https://doi.org/10.1145/2020408.2020558	https://doi.org/10.1145/2020408.2020558	NOUN
hjic-847	220	6	https://doi.org/10.1145/2020408.2020558	https://doi.org/10.1145/2020408.2020558	NOUN
hjic-847	220	7	https://doi.org/10.1145/1014052.1014118	https://doi.org/10.1145/1014052.1014118	PROPN
hjic-847	221	1	https://doi.org/10.5565/rev/elcvia.1054	https://doi.org/10.5565/rev/elcvia.1054	X
hjic-847	221	2	https://doi.org/10.1016/j.cviu.2005.09.012	https://doi.org/10.1016/j.cviu.2005.09.012	NUM
hjic-847	221	3	introduction	introduction	NOUN
hjic-847	221	4	proposed	propose	VERB
hjic-847	221	5	open	open	ADJ
hjic-847	221	6	-	-	PUNCT
hjic-847	221	7	world	world	NOUN
hjic-847	221	8	recognition	recognition	NOUN
hjic-847	221	9	system	system	NOUN
hjic-847	221	10	double	double	ADJ
hjic-847	221	11	probability	probability	NOUN
hjic-847	221	12	model	model	NOUN
hjic-847	221	13	unknown	unknown	ADJ
hjic-847	221	14	image	image	NOUN
hjic-847	221	15	clustering	cluster	VERB
hjic-847	221	16	baseline	baseline	NOUN
hjic-847	221	17	method	method	NOUN
hjic-847	221	18	cluster	cluster	NOUN
hjic-847	221	19	classification	classification	NOUN
hjic-847	221	20	experimental	experimental	ADJ
hjic-847	221	21	results	result	NOUN
hjic-847	221	22	conclusion	conclusion	NOUN