id	sid	tid	token	lemma	pos
10_1101-2021_02_08_430070	1	1	On	on	IN
10_1101-2021_02_08_430070	1	2	the	the	DT
10_1101-2021_02_08_430070	1	3	application	application	NN
10_1101-2021_02_08_430070	1	4	of	of	IN
10_1101-2021_02_08_430070	1	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	1	6	models	model	NNS
10_1101-2021_02_08_430070	1	7	for	for	IN
10_1101-2021_02_08_430070	1	8	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	1	9	methylation	methylation	NN
10_1101-2021_02_08_430070	1	10	detection	detection	NN
10_1101-2021_02_08_430070	1	11	✐	✐	NNP
10_1101-2021_02_08_430070	1	12	✐	✐	NNP
10_1101-2021_02_08_430070	1	13	✐	✐	NNP
10_1101-2021_02_08_430070	1	14	✐	✐	NNP
10_1101-2021_02_08_430070	1	15	✐	✐	NNP
10_1101-2021_02_08_430070	1	16	✐	✐	NNP
10_1101-2021_02_08_430070	1	17	✐	✐	NNP
10_1101-2021_02_08_430070	1	18	✐	✐	NNP
10_1101-2021_02_08_430070	1	19	Genome	Genome	NNP
10_1101-2021_02_08_430070	1	20	Analysis	Analysis	NNP
10_1101-2021_02_08_430070	1	21	On	on	IN
10_1101-2021_02_08_430070	1	22	the	the	DT
10_1101-2021_02_08_430070	1	23	application	application	NN
10_1101-2021_02_08_430070	1	24	of	of	IN
10_1101-2021_02_08_430070	1	25	BERT	BERT	NNP
10_1101-2021_02_08_430070	1	26	models	model	NNS
10_1101-2021_02_08_430070	1	27	for	for	IN
10_1101-2021_02_08_430070	1	28	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	1	29	methylation	methylation	NN
10_1101-2021_02_08_430070	1	30	detection	detection	NN
10_1101-2021_02_08_430070	1	31	Yao	Yao	NNP
10_1101-2021_02_08_430070	1	32	-	-	HYPH
10_1101-2021_02_08_430070	1	33	zhong	zhong	NNP
10_1101-2021_02_08_430070	1	34	Zhang	Zhang	NNP
10_1101-2021_02_08_430070	1	35	1,∗	1,∗	CD
10_1101-2021_02_08_430070	1	36	,	,	,
10_1101-2021_02_08_430070	1	37	Sera	Sera	NNP
10_1101-2021_02_08_430070	1	38	Hatakeyama	Hatakeyama	NNP
10_1101-2021_02_08_430070	1	39	1	1	CD
10_1101-2021_02_08_430070	1	40	,	,	,
10_1101-2021_02_08_430070	1	41	Kiyoshi	Kiyoshi	NNP
10_1101-2021_02_08_430070	1	42	Yamaguchi	Yamaguchi	NNP
10_1101-2021_02_08_430070	1	43	1	1	CD
10_1101-2021_02_08_430070	1	44	,	,	,
10_1101-2021_02_08_430070	1	45	Yoichi	Yoichi	NNP
10_1101-2021_02_08_430070	1	46	Furukawa	Furukawa	NNP
10_1101-2021_02_08_430070	1	47	1	1	CD
10_1101-2021_02_08_430070	1	48	,	,	,
10_1101-2021_02_08_430070	1	49	Satoru	Satoru	NNP
10_1101-2021_02_08_430070	1	50	Miyano	Miyano	NNP
10_1101-2021_02_08_430070	1	51	2	2	CD
10_1101-2021_02_08_430070	1	52	,	,	,
10_1101-2021_02_08_430070	1	53	Rui	Rui	NNP
10_1101-2021_02_08_430070	1	54	Yamaguchi	Yamaguchi	NNP
10_1101-2021_02_08_430070	1	55	3	3	CD
10_1101-2021_02_08_430070	1	56	,	,	,
10_1101-2021_02_08_430070	1	57	and	and	CC
10_1101-2021_02_08_430070	1	58	Seiya	Seiya	NNP
10_1101-2021_02_08_430070	1	59	Imoto	Imoto	NNP
10_1101-2021_02_08_430070	1	60	1,∗	1,∗	CD
10_1101-2021_02_08_430070	1	61	1Institute	1institute	CD
10_1101-2021_02_08_430070	1	62	of	of	IN
10_1101-2021_02_08_430070	1	63	Medical	Medical	NNP
10_1101-2021_02_08_430070	1	64	Science	Science	NNP
10_1101-2021_02_08_430070	1	65	,	,	,
10_1101-2021_02_08_430070	1	66	the	the	DT
10_1101-2021_02_08_430070	1	67	University	University	NNP
10_1101-2021_02_08_430070	1	68	of	of	IN
10_1101-2021_02_08_430070	1	69	Tokyo	Tokyo	NNP
10_1101-2021_02_08_430070	1	70	,	,	,
10_1101-2021_02_08_430070	1	71	Tokyo	Tokyo	NNP
10_1101-2021_02_08_430070	1	72	,	,	,
10_1101-2021_02_08_430070	1	73	108	108	CD
10_1101-2021_02_08_430070	1	74	-	-	HYPH
10_1101-2021_02_08_430070	1	75	0071	0071	CD
10_1101-2021_02_08_430070	1	76	,	,	,
10_1101-2021_02_08_430070	1	77	Japan	Japan	NNP
10_1101-2021_02_08_430070	1	78	2	2	CD
10_1101-2021_02_08_430070	1	79	M&D	M&D	NNP
10_1101-2021_02_08_430070	1	80	Data	Data	NNP
10_1101-2021_02_08_430070	1	81	Science	Science	NNP
10_1101-2021_02_08_430070	1	82	Center	Center	NNP
10_1101-2021_02_08_430070	1	83	,	,	,
10_1101-2021_02_08_430070	1	84	Tokyo	Tokyo	NNP
10_1101-2021_02_08_430070	1	85	Medical	Medical	NNP
10_1101-2021_02_08_430070	1	86	and	and	CC
10_1101-2021_02_08_430070	1	87	Dental	Dental	NNP
10_1101-2021_02_08_430070	1	88	University	University	NNP
10_1101-2021_02_08_430070	1	89	,	,	,
10_1101-2021_02_08_430070	1	90	Tokyo	Tokyo	NNP
10_1101-2021_02_08_430070	1	91	,	,	,
10_1101-2021_02_08_430070	1	92	101	101	CD
10_1101-2021_02_08_430070	1	93	-	-	SYM
10_1101-2021_02_08_430070	1	94	0062	0062	CD
10_1101-2021_02_08_430070	1	95	,	,	,
10_1101-2021_02_08_430070	1	96	Japan	Japan	NNP
10_1101-2021_02_08_430070	1	97	3Aichi	3Aichi	NNP
10_1101-2021_02_08_430070	1	98	Cancer	Cancer	NNP
10_1101-2021_02_08_430070	1	99	Center	Center	NNP
10_1101-2021_02_08_430070	1	100	Research	Research	NNP
10_1101-2021_02_08_430070	1	101	Institute	Institute	NNP
10_1101-2021_02_08_430070	1	102	,	,	,
10_1101-2021_02_08_430070	1	103	Nagoya	Nagoya	NNP
10_1101-2021_02_08_430070	1	104	,	,	,
10_1101-2021_02_08_430070	1	105	464	464	CD
10_1101-2021_02_08_430070	1	106	-	-	SYM
10_1101-2021_02_08_430070	1	107	8681	8681	CD
10_1101-2021_02_08_430070	1	108	,	,	,
10_1101-2021_02_08_430070	1	109	Japan	Japan	NNP
10_1101-2021_02_08_430070	1	110	∗To	∗To	NNP
10_1101-2021_02_08_430070	1	111	whom	whom	WP
10_1101-2021_02_08_430070	1	112	correspondence	correspondence	NN
10_1101-2021_02_08_430070	1	113	should	should	MD
10_1101-2021_02_08_430070	1	114	be	be	VB
10_1101-2021_02_08_430070	1	115	addressed	address	VBN
10_1101-2021_02_08_430070	1	116	.	.	.
10_1101-2021_02_08_430070	2	1	Abstract	Abstract	NNP
10_1101-2021_02_08_430070	2	2	Motivation	Motivation	NNP
10_1101-2021_02_08_430070	2	3	:	:	:
10_1101-2021_02_08_430070	2	4	DNA	dna	NN
10_1101-2021_02_08_430070	2	5	methylation	methylation	NN
10_1101-2021_02_08_430070	2	6	is	be	VBZ
10_1101-2021_02_08_430070	2	7	a	a	DT
10_1101-2021_02_08_430070	2	8	common	common	JJ
10_1101-2021_02_08_430070	2	9	epigenetic	epigenetic	JJ
10_1101-2021_02_08_430070	2	10	modification	modification	NN
10_1101-2021_02_08_430070	2	11	,	,	,
10_1101-2021_02_08_430070	2	12	which	which	WDT
10_1101-2021_02_08_430070	2	13	is	be	VBZ
10_1101-2021_02_08_430070	2	14	widely	widely	RB
10_1101-2021_02_08_430070	2	15	associated	associate	VBN
10_1101-2021_02_08_430070	2	16	with	with	IN
10_1101-2021_02_08_430070	2	17	various	various	JJ
10_1101-2021_02_08_430070	2	18	biological	biological	JJ
10_1101-2021_02_08_430070	2	19	processes	process	NNS
10_1101-2021_02_08_430070	2	20	,	,	,
10_1101-2021_02_08_430070	2	21	such	such	JJ
10_1101-2021_02_08_430070	2	22	as	as	IN
10_1101-2021_02_08_430070	2	23	gene	gene	NN
10_1101-2021_02_08_430070	2	24	expression	expression	NN
10_1101-2021_02_08_430070	2	25	,	,	,
10_1101-2021_02_08_430070	2	26	aging	age	VBG
10_1101-2021_02_08_430070	2	27	,	,	,
10_1101-2021_02_08_430070	2	28	and	and	CC
10_1101-2021_02_08_430070	2	29	disease	disease	NN
10_1101-2021_02_08_430070	2	30	.	.	.
10_1101-2021_02_08_430070	3	1	Nanopore	nanopore	JJ
10_1101-2021_02_08_430070	3	2	sequencing	sequencing	NN
10_1101-2021_02_08_430070	3	3	provides	provide	VBZ
10_1101-2021_02_08_430070	3	4	a	a	DT
10_1101-2021_02_08_430070	3	5	promising	promising	JJ
10_1101-2021_02_08_430070	3	6	methylation	methylation	NN
10_1101-2021_02_08_430070	3	7	detection	detection	NN
10_1101-2021_02_08_430070	3	8	approach	approach	NN
10_1101-2021_02_08_430070	3	9	through	through	IN
10_1101-2021_02_08_430070	3	10	monitoring	monitor	VBG
10_1101-2021_02_08_430070	3	11	abnormal	abnormal	JJ
10_1101-2021_02_08_430070	3	12	signal	signal	NN
10_1101-2021_02_08_430070	3	13	shifts	shift	NNS
10_1101-2021_02_08_430070	3	14	for	for	IN
10_1101-2021_02_08_430070	3	15	detecting	detect	VBG
10_1101-2021_02_08_430070	3	16	modified	modify	VBN
10_1101-2021_02_08_430070	3	17	bases	basis	NNS
10_1101-2021_02_08_430070	3	18	in	in	IN
10_1101-2021_02_08_430070	3	19	target	target	NN
10_1101-2021_02_08_430070	3	20	motif	motif	NN
10_1101-2021_02_08_430070	3	21	regions	region	NNS
10_1101-2021_02_08_430070	3	22	.	.	.
10_1101-2021_02_08_430070	4	1	Recently	recently	RB
10_1101-2021_02_08_430070	4	2	,	,	,
10_1101-2021_02_08_430070	4	3	model	model	NN
10_1101-2021_02_08_430070	4	4	-	-	HYPH
10_1101-2021_02_08_430070	4	5	based	base	VBN
10_1101-2021_02_08_430070	4	6	approaches	approach	NNS
10_1101-2021_02_08_430070	4	7	,	,	,
10_1101-2021_02_08_430070	4	8	especially	especially	RB
10_1101-2021_02_08_430070	4	9	those	those	DT
10_1101-2021_02_08_430070	4	10	with	with	IN
10_1101-2021_02_08_430070	4	11	deep	deep	JJ
10_1101-2021_02_08_430070	4	12	learning	learning	NN
10_1101-2021_02_08_430070	4	13	models	model	NNS
10_1101-2021_02_08_430070	4	14	,	,	,
10_1101-2021_02_08_430070	4	15	have	have	VBP
10_1101-2021_02_08_430070	4	16	achieved	achieve	VBN
10_1101-2021_02_08_430070	4	17	significant	significant	JJ
10_1101-2021_02_08_430070	4	18	performance	performance	NN
10_1101-2021_02_08_430070	4	19	improvements	improvement	NNS
10_1101-2021_02_08_430070	4	20	on	on	IN
10_1101-2021_02_08_430070	4	21	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	4	22	methylation	methylation	NN
10_1101-2021_02_08_430070	4	23	detection	detection	NN
10_1101-2021_02_08_430070	4	24	.	.	.
10_1101-2021_02_08_430070	5	1	In	in	IN
10_1101-2021_02_08_430070	5	2	this	this	DT
10_1101-2021_02_08_430070	5	3	work	work	NN
10_1101-2021_02_08_430070	5	4	,	,	,
10_1101-2021_02_08_430070	5	5	we	-PRON-	PRP
10_1101-2021_02_08_430070	5	6	explore	explore	VBP
10_1101-2021_02_08_430070	5	7	using	use	VBG
10_1101-2021_02_08_430070	5	8	bidirectional	bidirectional	JJ
10_1101-2021_02_08_430070	5	9	encoder	encoder	NN
10_1101-2021_02_08_430070	5	10	representations	representation	NNS
10_1101-2021_02_08_430070	5	11	from	from	IN
10_1101-2021_02_08_430070	5	12	transformers	transformer	NNS
10_1101-2021_02_08_430070	5	13	(	(	-LRB-
10_1101-2021_02_08_430070	5	14	BERT	BERT	NNP
10_1101-2021_02_08_430070	5	15	)	)	-RRB-
10_1101-2021_02_08_430070	5	16	for	for	IN
10_1101-2021_02_08_430070	5	17	doing	do	VBG
10_1101-2021_02_08_430070	5	18	the	the	DT
10_1101-2021_02_08_430070	5	19	task	task	NN
10_1101-2021_02_08_430070	5	20	,	,	,
10_1101-2021_02_08_430070	5	21	which	which	WDT
10_1101-2021_02_08_430070	5	22	can	can	MD
10_1101-2021_02_08_430070	5	23	provide	provide	VB
10_1101-2021_02_08_430070	5	24	non	non	JJ
10_1101-2021_02_08_430070	5	25	-	-	JJ
10_1101-2021_02_08_430070	5	26	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	5	27	neural	neural	JJ
10_1101-2021_02_08_430070	5	28	structures	structure	NNS
10_1101-2021_02_08_430070	5	29	for	for	IN
10_1101-2021_02_08_430070	5	30	fast	fast	JJ
10_1101-2021_02_08_430070	5	31	parallel	parallel	JJ
10_1101-2021_02_08_430070	5	32	computation	computation	NN
10_1101-2021_02_08_430070	5	33	.	.	.
10_1101-2021_02_08_430070	6	1	Results	result	NNS
10_1101-2021_02_08_430070	6	2	:	:	:
10_1101-2021_02_08_430070	6	3	We	-PRON-	PRP
10_1101-2021_02_08_430070	6	4	find	find	VBP
10_1101-2021_02_08_430070	6	5	original	original	JJ
10_1101-2021_02_08_430070	6	6	BERT	BERT	NNP
10_1101-2021_02_08_430070	6	7	architecture	architecture	NN
10_1101-2021_02_08_430070	6	8	does	do	VBZ
10_1101-2021_02_08_430070	6	9	not	not	RB
10_1101-2021_02_08_430070	6	10	work	work	VB
10_1101-2021_02_08_430070	6	11	as	as	RB
10_1101-2021_02_08_430070	6	12	well	well	RB
10_1101-2021_02_08_430070	6	13	as	as	IN
10_1101-2021_02_08_430070	6	14	the	the	DT
10_1101-2021_02_08_430070	6	15	bidirectional	bidirectional	JJ
10_1101-2021_02_08_430070	6	16	recurrent	recurrent	NN
10_1101-2021_02_08_430070	6	17	neural	neural	JJ
10_1101-2021_02_08_430070	6	18	network	network	NN
10_1101-2021_02_08_430070	6	19	(	(	-LRB-
10_1101-2021_02_08_430070	6	20	biRNN	birnn	NN
10_1101-2021_02_08_430070	6	21	)	)	-RRB-
10_1101-2021_02_08_430070	6	22	on	on	IN
10_1101-2021_02_08_430070	6	23	the	the	DT
10_1101-2021_02_08_430070	6	24	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	6	25	methylation	methylation	NN
10_1101-2021_02_08_430070	6	26	prediction	prediction	NN
10_1101-2021_02_08_430070	6	27	task	task	NN
10_1101-2021_02_08_430070	6	28	.	.	.
10_1101-2021_02_08_430070	7	1	Through	through	IN
10_1101-2021_02_08_430070	7	2	further	further	JJ
10_1101-2021_02_08_430070	7	3	analysis	analysis	NN
10_1101-2021_02_08_430070	7	4	,	,	,
10_1101-2021_02_08_430070	7	5	we	-PRON-	PRP
10_1101-2021_02_08_430070	7	6	observe	observe	VBP
10_1101-2021_02_08_430070	7	7	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	7	8	patterns	pattern	NNS
10_1101-2021_02_08_430070	7	9	of	of	IN
10_1101-2021_02_08_430070	7	10	positional	positional	JJ
10_1101-2021_02_08_430070	7	11	-	-	HYPH
10_1101-2021_02_08_430070	7	12	signal	signal	NN
10_1101-2021_02_08_430070	7	13	-	-	HYPH
10_1101-2021_02_08_430070	7	14	shift	shift	NN
10_1101-2021_02_08_430070	7	15	in	in	IN
10_1101-2021_02_08_430070	7	16	the	the	DT
10_1101-2021_02_08_430070	7	17	context	context	NN
10_1101-2021_02_08_430070	7	18	window	window	NN
10_1101-2021_02_08_430070	7	19	surrounding	surround	VBG
10_1101-2021_02_08_430070	7	20	target	target	NN
10_1101-2021_02_08_430070	7	21	5-methylcytosine	5-methylcytosine	CD
10_1101-2021_02_08_430070	7	22	(	(	-LRB-
10_1101-2021_02_08_430070	7	23	5mC	5mc	CD
10_1101-2021_02_08_430070	7	24	)	)	-RRB-
10_1101-2021_02_08_430070	7	25	and	and	CC
10_1101-2021_02_08_430070	7	26	N6-methyladenine	N6-methyladenine	NNP
10_1101-2021_02_08_430070	7	27	(	(	-LRB-
10_1101-2021_02_08_430070	7	28	6mA	6mA	NNP
10_1101-2021_02_08_430070	7	29	)	)	-RRB-
10_1101-2021_02_08_430070	7	30	motifs	motif	NNS
10_1101-2021_02_08_430070	7	31	.	.	.
10_1101-2021_02_08_430070	8	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	8	2	propose	propose	VBP
10_1101-2021_02_08_430070	8	3	a	a	DT
10_1101-2021_02_08_430070	8	4	refined	refined	JJ
10_1101-2021_02_08_430070	8	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	8	6	with	with	IN
10_1101-2021_02_08_430070	8	7	relative	relative	JJ
10_1101-2021_02_08_430070	8	8	position	position	NN
10_1101-2021_02_08_430070	8	9	representation	representation	NN
10_1101-2021_02_08_430070	8	10	and	and	CC
10_1101-2021_02_08_430070	8	11	center	center	NN
10_1101-2021_02_08_430070	8	12	hidden	hide	VBN
10_1101-2021_02_08_430070	8	13	units	units	NNPS
10_1101-2021_02_08_430070	8	14	concatenation	concatenation	NNP
10_1101-2021_02_08_430070	8	15	,	,	,
10_1101-2021_02_08_430070	8	16	which	which	WDT
10_1101-2021_02_08_430070	8	17	takes	take	VBZ
10_1101-2021_02_08_430070	8	18	account	account	NN
10_1101-2021_02_08_430070	8	19	of	of	IN
10_1101-2021_02_08_430070	8	20	task	task	NN
10_1101-2021_02_08_430070	8	21	-	-	HYPH
10_1101-2021_02_08_430070	8	22	specific	specific	JJ
10_1101-2021_02_08_430070	8	23	characters	character	NNS
10_1101-2021_02_08_430070	8	24	into	into	IN
10_1101-2021_02_08_430070	8	25	modeling	modeling	NN
10_1101-2021_02_08_430070	8	26	.	.	.
10_1101-2021_02_08_430070	9	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	9	2	perform	perform	VBP
10_1101-2021_02_08_430070	9	3	systematic	systematic	JJ
10_1101-2021_02_08_430070	9	4	evaluations	evaluation	NNS
10_1101-2021_02_08_430070	9	5	in	in	IN
10_1101-2021_02_08_430070	9	6	-	-	HYPH
10_1101-2021_02_08_430070	9	7	sample	sample	NN
10_1101-2021_02_08_430070	9	8	and	and	CC
10_1101-2021_02_08_430070	9	9	cross	cross	NN
10_1101-2021_02_08_430070	9	10	-	-	NN
10_1101-2021_02_08_430070	9	11	sample	sample	NN
10_1101-2021_02_08_430070	9	12	.	.	.
10_1101-2021_02_08_430070	10	1	The	the	DT
10_1101-2021_02_08_430070	10	2	experiment	experiment	NN
10_1101-2021_02_08_430070	10	3	results	result	NNS
10_1101-2021_02_08_430070	10	4	show	show	VBP
10_1101-2021_02_08_430070	10	5	that	that	IN
10_1101-2021_02_08_430070	10	6	the	the	DT
10_1101-2021_02_08_430070	10	7	refined	refined	JJ
10_1101-2021_02_08_430070	10	8	BERT	BERT	NNP
10_1101-2021_02_08_430070	10	9	model	model	NN
10_1101-2021_02_08_430070	10	10	can	can	MD
10_1101-2021_02_08_430070	10	11	achieve	achieve	VB
10_1101-2021_02_08_430070	10	12	competitive	competitive	JJ
10_1101-2021_02_08_430070	10	13	or	or	CC
10_1101-2021_02_08_430070	10	14	even	even	RB
10_1101-2021_02_08_430070	10	15	better	well	JJR
10_1101-2021_02_08_430070	10	16	results	result	NNS
10_1101-2021_02_08_430070	10	17	than	than	IN
10_1101-2021_02_08_430070	10	18	the	the	DT
10_1101-2021_02_08_430070	10	19	state	state	NN
10_1101-2021_02_08_430070	10	20	-	-	HYPH
10_1101-2021_02_08_430070	10	21	of	of	IN
10_1101-2021_02_08_430070	10	22	-	-	HYPH
10_1101-2021_02_08_430070	10	23	the	the	DT
10_1101-2021_02_08_430070	10	24	-	-	HYPH
10_1101-2021_02_08_430070	10	25	art	art	NN
10_1101-2021_02_08_430070	10	26	biRNN	birnn	NN
10_1101-2021_02_08_430070	10	27	model	model	NN
10_1101-2021_02_08_430070	10	28	,	,	,
10_1101-2021_02_08_430070	10	29	while	while	IN
10_1101-2021_02_08_430070	10	30	the	the	DT
10_1101-2021_02_08_430070	10	31	model	model	NN
10_1101-2021_02_08_430070	10	32	inference	inference	NN
10_1101-2021_02_08_430070	10	33	speed	speed	NN
10_1101-2021_02_08_430070	10	34	is	be	VBZ
10_1101-2021_02_08_430070	10	35	about	about	RB
10_1101-2021_02_08_430070	10	36	6x	6x	CD
10_1101-2021_02_08_430070	10	37	faster	fast	RBR
10_1101-2021_02_08_430070	10	38	.	.	.
10_1101-2021_02_08_430070	11	1	Besides	besides	RB
10_1101-2021_02_08_430070	11	2	,	,	,
10_1101-2021_02_08_430070	11	3	on	on	IN
10_1101-2021_02_08_430070	11	4	the	the	DT
10_1101-2021_02_08_430070	11	5	cross	cross	JJ
10_1101-2021_02_08_430070	11	6	-	-	JJ
10_1101-2021_02_08_430070	11	7	sample	sample	JJ
10_1101-2021_02_08_430070	11	8	evaluation	evaluation	NN
10_1101-2021_02_08_430070	11	9	of	of	IN
10_1101-2021_02_08_430070	11	10	datasets	dataset	NNS
10_1101-2021_02_08_430070	11	11	from	from	IN
10_1101-2021_02_08_430070	11	12	the	the	DT
10_1101-2021_02_08_430070	11	13	different	different	JJ
10_1101-2021_02_08_430070	11	14	research	research	NN
10_1101-2021_02_08_430070	11	15	groups	group	NNS
10_1101-2021_02_08_430070	11	16	,	,	,
10_1101-2021_02_08_430070	11	17	BERT	BERT	NNP
10_1101-2021_02_08_430070	11	18	models	model	NNS
10_1101-2021_02_08_430070	11	19	demonstrate	demonstrate	VBP
10_1101-2021_02_08_430070	11	20	a	a	DT
10_1101-2021_02_08_430070	11	21	good	good	JJ
10_1101-2021_02_08_430070	11	22	generalization	generalization	NN
10_1101-2021_02_08_430070	11	23	performance	performance	NN
10_1101-2021_02_08_430070	11	24	.	.	.
10_1101-2021_02_08_430070	12	1	Availability	availability	NN
10_1101-2021_02_08_430070	12	2	:	:	:
10_1101-2021_02_08_430070	12	3	The	the	DT
10_1101-2021_02_08_430070	12	4	source	source	NN
10_1101-2021_02_08_430070	12	5	code	code	NN
10_1101-2021_02_08_430070	12	6	and	and	CC
10_1101-2021_02_08_430070	12	7	data	datum	NNS
10_1101-2021_02_08_430070	12	8	are	be	VBP
10_1101-2021_02_08_430070	12	9	available	available	JJ
10_1101-2021_02_08_430070	12	10	at	at	IN
10_1101-2021_02_08_430070	12	11	https://github.com/yaozhong/methBERT	https://github.com/yaozhong/methBERT	NNP
10_1101-2021_02_08_430070	12	12	Contact:yaozhong@ims.u-tokyo.ac.jp	Contact:yaozhong@ims.u-tokyo.ac.jp	NNP
10_1101-2021_02_08_430070	12	13	1	1	CD
10_1101-2021_02_08_430070	12	14	Introduction	introduction	NN
10_1101-2021_02_08_430070	12	15	Methylation	Methylation	NNP
10_1101-2021_02_08_430070	12	16	of	of	IN
10_1101-2021_02_08_430070	12	17	DNA	DNA	NNP
10_1101-2021_02_08_430070	12	18	/	/	SYM
10_1101-2021_02_08_430070	12	19	RNA	RNA	NNP
10_1101-2021_02_08_430070	12	20	/	/	SYM
10_1101-2021_02_08_430070	12	21	histone	histone	NN
10_1101-2021_02_08_430070	12	22	is	be	VBZ
10_1101-2021_02_08_430070	12	23	commonly	commonly	RB
10_1101-2021_02_08_430070	12	24	observed	observe	VBN
10_1101-2021_02_08_430070	12	25	in	in	IN
10_1101-2021_02_08_430070	12	26	developmental	developmental	JJ
10_1101-2021_02_08_430070	12	27	disorders	disorder	NNS
10_1101-2021_02_08_430070	12	28	,	,	,
10_1101-2021_02_08_430070	12	29	aging	age	VBG
10_1101-2021_02_08_430070	12	30	,	,	,
10_1101-2021_02_08_430070	12	31	and	and	CC
10_1101-2021_02_08_430070	12	32	genomic	genomic	JJ
10_1101-2021_02_08_430070	12	33	disease	disease	NN
10_1101-2021_02_08_430070	12	34	,	,	,
10_1101-2021_02_08_430070	12	35	such	such	JJ
10_1101-2021_02_08_430070	12	36	as	as	IN
10_1101-2021_02_08_430070	12	37	cancer	cancer	NN
10_1101-2021_02_08_430070	12	38	.	.	.
10_1101-2021_02_08_430070	13	1	Fast	fast	RB
10_1101-2021_02_08_430070	13	2	and	and	CC
10_1101-2021_02_08_430070	13	3	accurately	accurately	RB
10_1101-2021_02_08_430070	13	4	detecting	detect	VBG
10_1101-2021_02_08_430070	13	5	methylation	methylation	NN
10_1101-2021_02_08_430070	13	6	status	status	NN
10_1101-2021_02_08_430070	13	7	has	have	VBZ
10_1101-2021_02_08_430070	13	8	a	a	DT
10_1101-2021_02_08_430070	13	9	fundamental	fundamental	JJ
10_1101-2021_02_08_430070	13	10	requirement	requirement	NN
10_1101-2021_02_08_430070	13	11	to	to	TO
10_1101-2021_02_08_430070	13	12	find	find	VB
10_1101-2021_02_08_430070	13	13	distinctive	distinctive	JJ
10_1101-2021_02_08_430070	13	14	biomarkers	biomarker	NNS
10_1101-2021_02_08_430070	13	15	for	for	IN
10_1101-2021_02_08_430070	13	16	aging	age	VBG
10_1101-2021_02_08_430070	13	17	/	/	SYM
10_1101-2021_02_08_430070	13	18	disease	disease	NN
10_1101-2021_02_08_430070	13	19	profiling	profiling	NN
10_1101-2021_02_08_430070	13	20	.	.	.
10_1101-2021_02_08_430070	14	1	For	for	IN
10_1101-2021_02_08_430070	14	2	a	a	DT
10_1101-2021_02_08_430070	14	3	virome	virome	NN
10_1101-2021_02_08_430070	14	4	/	/	SYM
10_1101-2021_02_08_430070	14	5	metagenome	metagenome	NN
10_1101-2021_02_08_430070	14	6	study	study	NN
10_1101-2021_02_08_430070	14	7	,	,	,
10_1101-2021_02_08_430070	14	8	quick	quick	JJ
10_1101-2021_02_08_430070	14	9	and	and	CC
10_1101-2021_02_08_430070	14	10	accurate	accurate	JJ
10_1101-2021_02_08_430070	14	11	epi	epi	JJ
10_1101-2021_02_08_430070	14	12	-	-	HYPH
10_1101-2021_02_08_430070	14	13	transcriptome	transcriptome	NNP
10_1101-2021_02_08_430070	14	14	detection	detection	NN
10_1101-2021_02_08_430070	14	15	also	also	RB
10_1101-2021_02_08_430070	14	16	plays	play	VBZ
10_1101-2021_02_08_430070	14	17	an	an	DT
10_1101-2021_02_08_430070	14	18	important	important	JJ
10_1101-2021_02_08_430070	14	19	role	role	NN
10_1101-2021_02_08_430070	14	20	in	in	IN
10_1101-2021_02_08_430070	14	21	understanding	understand	VBG
10_1101-2021_02_08_430070	14	22	unseen	unseen	JJ
10_1101-2021_02_08_430070	14	23	strains	strain	NNS
10_1101-2021_02_08_430070	14	24	(	(	-LRB-
10_1101-2021_02_08_430070	14	25	Kim	Kim	NNP
10_1101-2021_02_08_430070	14	26	et	et	FW
10_1101-2021_02_08_430070	14	27	al	al	NNP
10_1101-2021_02_08_430070	14	28	.	.	NNP
10_1101-2021_02_08_430070	14	29	,	,	,
10_1101-2021_02_08_430070	14	30	2020	2020	CD
10_1101-2021_02_08_430070	14	31	)	)	-RRB-
10_1101-2021_02_08_430070	14	32	.	.	.
10_1101-2021_02_08_430070	15	1	One	one	CD
10_1101-2021_02_08_430070	15	2	commonly	commonly	RB
10_1101-2021_02_08_430070	15	3	used	use	VBN
10_1101-2021_02_08_430070	15	4	DNA	dna	NN
10_1101-2021_02_08_430070	15	5	methylation	methylation	NN
10_1101-2021_02_08_430070	15	6	detection	detection	NN
10_1101-2021_02_08_430070	15	7	approach	approach	NN
10_1101-2021_02_08_430070	15	8	is	be	VBZ
10_1101-2021_02_08_430070	15	9	Whole	whole	JJ
10_1101-2021_02_08_430070	15	10	-	-	HYPH
10_1101-2021_02_08_430070	15	11	Genome	genome	JJ
10_1101-2021_02_08_430070	15	12	Bisulfite	Bisulfite	NNP
10_1101-2021_02_08_430070	15	13	Sequencing	Sequencing	NNP
10_1101-2021_02_08_430070	15	14	(	(	-LRB-
10_1101-2021_02_08_430070	15	15	WGBS	WGBS	NNP
10_1101-2021_02_08_430070	15	16	)	)	-RRB-
10_1101-2021_02_08_430070	15	17	.	.	.
10_1101-2021_02_08_430070	16	1	To	to	TO
10_1101-2021_02_08_430070	16	2	detect	detect	VB
10_1101-2021_02_08_430070	16	3	modified	modify	VBN
10_1101-2021_02_08_430070	16	4	bases	basis	NNS
10_1101-2021_02_08_430070	16	5	,	,	,
10_1101-2021_02_08_430070	16	6	WGBS	WGBS	NNP
10_1101-2021_02_08_430070	16	7	first	first	RB
10_1101-2021_02_08_430070	16	8	takes	take	VBZ
10_1101-2021_02_08_430070	16	9	sodium	sodium	NN
10_1101-2021_02_08_430070	16	10	bisulfite	bisulfite	NNP
10_1101-2021_02_08_430070	16	11	conversion	conversion	NN
10_1101-2021_02_08_430070	16	12	before	before	IN
10_1101-2021_02_08_430070	16	13	sequencing	sequencing	NN
10_1101-2021_02_08_430070	16	14	.	.	.
10_1101-2021_02_08_430070	17	1	As	as	IN
10_1101-2021_02_08_430070	17	2	the	the	DT
10_1101-2021_02_08_430070	17	3	pre	pre	JJ
10_1101-2021_02_08_430070	17	4	-	-	JJ
10_1101-2021_02_08_430070	17	5	chemical	chemical	JJ
10_1101-2021_02_08_430070	17	6	bisulfite	bisulfite	NN
10_1101-2021_02_08_430070	17	7	conversion	conversion	NN
10_1101-2021_02_08_430070	17	8	is	be	VBZ
10_1101-2021_02_08_430070	17	9	a	a	DT
10_1101-2021_02_08_430070	17	10	relatively	relatively	RB
10_1101-2021_02_08_430070	17	11	harsh	harsh	JJ
10_1101-2021_02_08_430070	17	12	process	process	NN
10_1101-2021_02_08_430070	17	13	,	,	,
10_1101-2021_02_08_430070	17	14	it	-PRON-	PRP
10_1101-2021_02_08_430070	17	15	makes	make	VBZ
10_1101-2021_02_08_430070	17	16	DNA	dna	NN
10_1101-2021_02_08_430070	17	17	sequences	sequence	NNS
10_1101-2021_02_08_430070	17	18	more	more	RBR
10_1101-2021_02_08_430070	17	19	fragmental	fragmental	JJ
10_1101-2021_02_08_430070	17	20	and	and	CC
10_1101-2021_02_08_430070	17	21	a	a	DT
10_1101-2021_02_08_430070	17	22	large	large	JJ
10_1101-2021_02_08_430070	17	23	amount	amount	NN
10_1101-2021_02_08_430070	17	24	of	of	IN
10_1101-2021_02_08_430070	17	25	DNA	dna	NN
10_1101-2021_02_08_430070	17	26	is	be	VBZ
10_1101-2021_02_08_430070	17	27	usually	usually	RB
10_1101-2021_02_08_430070	17	28	required	require	VBN
10_1101-2021_02_08_430070	17	29	.	.	.
10_1101-2021_02_08_430070	18	1	Also	also	RB
10_1101-2021_02_08_430070	18	2	,	,	,
10_1101-2021_02_08_430070	18	3	limited	limit	VBN
10_1101-2021_02_08_430070	18	4	to	to	IN
10_1101-2021_02_08_430070	18	5	the	the	DT
10_1101-2021_02_08_430070	18	6	read	read	VBN
10_1101-2021_02_08_430070	18	7	length	length	NN
10_1101-2021_02_08_430070	18	8	,	,	,
10_1101-2021_02_08_430070	18	9	it	-PRON-	PRP
10_1101-2021_02_08_430070	18	10	is	be	VBZ
10_1101-2021_02_08_430070	18	11	difficult	difficult	JJ
10_1101-2021_02_08_430070	18	12	to	to	TO
10_1101-2021_02_08_430070	18	13	align	align	VB
10_1101-2021_02_08_430070	18	14	short	short	JJ
10_1101-2021_02_08_430070	18	15	reads	read	NNS
10_1101-2021_02_08_430070	18	16	in	in	IN
10_1101-2021_02_08_430070	18	17	low	low	JJ
10_1101-2021_02_08_430070	18	18	-	-	HYPH
10_1101-2021_02_08_430070	18	19	complex	complex	NN
10_1101-2021_02_08_430070	18	20	regions	region	NNS
10_1101-2021_02_08_430070	18	21	and	and	CC
10_1101-2021_02_08_430070	18	22	analyze	analyze	VB
10_1101-2021_02_08_430070	18	23	methylation	methylation	NN
10_1101-2021_02_08_430070	18	24	patterns	pattern	NNS
10_1101-2021_02_08_430070	18	25	in	in	IN
10_1101-2021_02_08_430070	18	26	a	a	DT
10_1101-2021_02_08_430070	18	27	long-	long-	JJ
10_1101-2021_02_08_430070	18	28	range	range	NN
10_1101-2021_02_08_430070	18	29	.	.	.
10_1101-2021_02_08_430070	19	1	The	the	DT
10_1101-2021_02_08_430070	19	2	data	datum	NNS
10_1101-2021_02_08_430070	19	3	processing	processing	NN
10_1101-2021_02_08_430070	19	4	of	of	IN
10_1101-2021_02_08_430070	19	5	WGBS	WGBS	NNP
10_1101-2021_02_08_430070	19	6	is	be	VBZ
10_1101-2021_02_08_430070	19	7	sophisticated	sophisticated	JJ
10_1101-2021_02_08_430070	19	8	and	and	CC
10_1101-2021_02_08_430070	19	9	time	time	NN
10_1101-2021_02_08_430070	19	10	-	-	HYPH
10_1101-2021_02_08_430070	19	11	consuming	consume	VBG
10_1101-2021_02_08_430070	19	12	.	.	.
10_1101-2021_02_08_430070	20	1	Various	various	JJ
10_1101-2021_02_08_430070	20	2	biases	bias	NNS
10_1101-2021_02_08_430070	20	3	(	(	-LRB-
10_1101-2021_02_08_430070	20	4	e.g.	e.g.	RB
10_1101-2021_02_08_430070	21	1	GC	GC	NNP
10_1101-2021_02_08_430070	21	2	and	and	CC
10_1101-2021_02_08_430070	21	3	fragment	fragment	NN
10_1101-2021_02_08_430070	21	4	length	length	NN
10_1101-2021_02_08_430070	21	5	)	)	-RRB-
10_1101-2021_02_08_430070	21	6	including	include	VBG
10_1101-2021_02_08_430070	21	7	those	those	DT
10_1101-2021_02_08_430070	21	8	introduced	introduce	VBN
10_1101-2021_02_08_430070	21	9	by	by	IN
10_1101-2021_02_08_430070	21	10	bisulfite	bisulfite	NN
10_1101-2021_02_08_430070	21	11	treatment	treatment	NN
10_1101-2021_02_08_430070	21	12	are	be	VBP
10_1101-2021_02_08_430070	21	13	required	require	VBN
10_1101-2021_02_08_430070	21	14	to	to	TO
10_1101-2021_02_08_430070	21	15	be	be	VB
10_1101-2021_02_08_430070	21	16	dealt	deal	VBN
10_1101-2021_02_08_430070	21	17	with	with	IN
10_1101-2021_02_08_430070	21	18	in	in	IN
10_1101-2021_02_08_430070	21	19	the	the	DT
10_1101-2021_02_08_430070	21	20	data	datum	NNS
10_1101-2021_02_08_430070	21	21	analysis	analysis	NN
10_1101-2021_02_08_430070	21	22	.	.	.
10_1101-2021_02_08_430070	22	1	WGBS	WGBS	NNP
10_1101-2021_02_08_430070	22	2	can	can	MD
10_1101-2021_02_08_430070	22	3	only	only	RB
10_1101-2021_02_08_430070	22	4	be	be	VB
10_1101-2021_02_08_430070	22	5	used	use	VBN
10_1101-2021_02_08_430070	22	6	for	for	IN
10_1101-2021_02_08_430070	22	7	DNA	dna	NN
10_1101-2021_02_08_430070	22	8	samples	sample	NNS
10_1101-2021_02_08_430070	22	9	,	,	,
10_1101-2021_02_08_430070	22	10	which	which	WDT
10_1101-2021_02_08_430070	22	11	limits	limit	VBZ
10_1101-2021_02_08_430070	22	12	its	-PRON-	PRP$
10_1101-2021_02_08_430070	22	13	application	application	NN
10_1101-2021_02_08_430070	22	14	of	of	IN
10_1101-2021_02_08_430070	22	15	detecting	detect	VBG
10_1101-2021_02_08_430070	22	16	RNA	RNA	NNP
10_1101-2021_02_08_430070	22	17	methylation	methylation	NN
10_1101-2021_02_08_430070	22	18	.	.	.
10_1101-2021_02_08_430070	23	1	Single	single	JJ
10_1101-2021_02_08_430070	23	2	-	-	HYPH
10_1101-2021_02_08_430070	23	3	molecule	molecule	NN
10_1101-2021_02_08_430070	23	4	sequencing	sequencing	NN
10_1101-2021_02_08_430070	23	5	(	(	-LRB-
10_1101-2021_02_08_430070	23	6	e.g.	e.g.	RB
10_1101-2021_02_08_430070	23	7	,	,	,
10_1101-2021_02_08_430070	23	8	PacBio	PacBio	NNP
10_1101-2021_02_08_430070	23	9	and	and	CC
10_1101-2021_02_08_430070	23	10	Nanopore	Nanopore	NNP
10_1101-2021_02_08_430070	23	11	)	)	-RRB-
10_1101-2021_02_08_430070	23	12	provides	provide	VBZ
10_1101-2021_02_08_430070	23	13	a	a	DT
10_1101-2021_02_08_430070	23	14	promising	promising	JJ
10_1101-2021_02_08_430070	23	15	approach	approach	NN
10_1101-2021_02_08_430070	23	16	through	through	IN
10_1101-2021_02_08_430070	23	17	detecting	detect	VBG
10_1101-2021_02_08_430070	23	18	abnormal	abnormal	JJ
10_1101-2021_02_08_430070	23	19	signals	signal	NNS
10_1101-2021_02_08_430070	23	20	in	in	IN
10_1101-2021_02_08_430070	23	21	target	target	NN
10_1101-2021_02_08_430070	23	22	motif	motif	NN
10_1101-2021_02_08_430070	23	23	regions	region	NNS
10_1101-2021_02_08_430070	23	24	,	,	,
10_1101-2021_02_08_430070	23	25	as	as	IN
10_1101-2021_02_08_430070	23	26	modified	modify	VBN
10_1101-2021_02_08_430070	23	27	bases	basis	NNS
10_1101-2021_02_08_430070	23	28	usually	usually	RB
10_1101-2021_02_08_430070	23	29	have	have	VBP
10_1101-2021_02_08_430070	23	30	different	different	JJ
10_1101-2021_02_08_430070	23	31	current	current	JJ
10_1101-2021_02_08_430070	23	32	signals	signal	NNS
10_1101-2021_02_08_430070	23	33	.	.	.
10_1101-2021_02_08_430070	24	1	Compared	compare	VBN
10_1101-2021_02_08_430070	24	2	with	with	IN
10_1101-2021_02_08_430070	24	3	the	the	DT
10_1101-2021_02_08_430070	24	4	sodium	sodium	NN
10_1101-2021_02_08_430070	24	5	bisulfite	bisulfite	NN
10_1101-2021_02_08_430070	24	6	approach	approach	NN
10_1101-2021_02_08_430070	24	7	,	,	,
10_1101-2021_02_08_430070	24	8	no	no	DT
10_1101-2021_02_08_430070	24	9	extra	extra	JJ
10_1101-2021_02_08_430070	24	10	chemical	chemical	NN
10_1101-2021_02_08_430070	24	11	treatment	treatment	NN
10_1101-2021_02_08_430070	24	12	is	be	VBZ
10_1101-2021_02_08_430070	24	13	required	require	VBN
10_1101-2021_02_08_430070	24	14	,	,	,
10_1101-2021_02_08_430070	24	15	which	which	WDT
10_1101-2021_02_08_430070	24	16	helps	help	VBZ
10_1101-2021_02_08_430070	24	17	to	to	TO
10_1101-2021_02_08_430070	24	18	reduce	reduce	VB
10_1101-2021_02_08_430070	24	19	potential	potential	JJ
10_1101-2021_02_08_430070	24	20	biases	bias	NNS
10_1101-2021_02_08_430070	24	21	.	.	.
10_1101-2021_02_08_430070	25	1	Currently	currently	RB
10_1101-2021_02_08_430070	25	2	exist	exist	VBP
10_1101-2021_02_08_430070	25	3	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	25	4	methylation	methylation	NN
10_1101-2021_02_08_430070	25	5	detection	detection	NN
10_1101-2021_02_08_430070	25	6	methods	method	NNS
10_1101-2021_02_08_430070	25	7	can	can	MD
10_1101-2021_02_08_430070	25	8	be	be	VB
10_1101-2021_02_08_430070	25	9	categorized	categorize	VBN
10_1101-2021_02_08_430070	25	10	into	into	IN
10_1101-2021_02_08_430070	25	11	two	two	CD
10_1101-2021_02_08_430070	25	12	types	type	NNS
10_1101-2021_02_08_430070	25	13	.	.	.
10_1101-2021_02_08_430070	26	1	One	one	CD
10_1101-2021_02_08_430070	26	2	is	be	VBZ
10_1101-2021_02_08_430070	26	3	testing	testing	NN
10_1101-2021_02_08_430070	26	4	-	-	HYPH
10_1101-2021_02_08_430070	26	5	based	base	VBN
10_1101-2021_02_08_430070	26	6	(	(	-LRB-
10_1101-2021_02_08_430070	26	7	e.g	e.g	NNP
10_1101-2021_02_08_430070	26	8	.	.	NNP
10_1101-2021_02_08_430070	26	9	,Tombo	,Tombo	,
10_1101-2021_02_08_430070	26	10	(	(	-LRB-
10_1101-2021_02_08_430070	26	11	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	26	12	et	et	FW
10_1101-2021_02_08_430070	26	13	al	al	NNP
10_1101-2021_02_08_430070	26	14	.	.	NNP
10_1101-2021_02_08_430070	26	15	,	,	,
10_1101-2021_02_08_430070	26	16	2016	2016	CD
10_1101-2021_02_08_430070	26	17	)	)	-RRB-
10_1101-2021_02_08_430070	26	18	)	)	-RRB-
10_1101-2021_02_08_430070	26	19	,	,	,
10_1101-2021_02_08_430070	26	20	the	the	DT
10_1101-2021_02_08_430070	26	21	other	other	JJ
10_1101-2021_02_08_430070	26	22	is	be	VBZ
10_1101-2021_02_08_430070	26	23	model	model	NN
10_1101-2021_02_08_430070	26	24	-	-	HYPH
10_1101-2021_02_08_430070	26	25	based	base	VBN
10_1101-2021_02_08_430070	26	26	(	(	-LRB-
10_1101-2021_02_08_430070	26	27	e.g.	e.g.	RB
10_1101-2021_02_08_430070	26	28	,	,	,
10_1101-2021_02_08_430070	26	29	nanopolish	nanopolish	JJ
10_1101-2021_02_08_430070	26	30	(	(	-LRB-
10_1101-2021_02_08_430070	26	31	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	26	32	et	et	FW
10_1101-2021_02_08_430070	26	33	al	al	NNP
10_1101-2021_02_08_430070	26	34	.	.	NNP
10_1101-2021_02_08_430070	26	35	,	,	,
10_1101-2021_02_08_430070	26	36	2017	2017	CD
10_1101-2021_02_08_430070	26	37	)	)	-RRB-
10_1101-2021_02_08_430070	26	38	,	,	,
10_1101-2021_02_08_430070	26	39	deepMod(Liu	deepMod(Liu	NNP
10_1101-2021_02_08_430070	26	40	et	et	NNP
10_1101-2021_02_08_430070	26	41	al	al	NNP
10_1101-2021_02_08_430070	26	42	.	.	NNP
10_1101-2021_02_08_430070	26	43	,	,	,
10_1101-2021_02_08_430070	26	44	2019	2019	CD
10_1101-2021_02_08_430070	26	45	)	)	-RRB-
10_1101-2021_02_08_430070	26	46	and	and	CC
10_1101-2021_02_08_430070	26	47	deepSignal	deepsignal	JJ
10_1101-2021_02_08_430070	26	48	(	(	-LRB-
10_1101-2021_02_08_430070	26	49	Ni	Ni	NNP
10_1101-2021_02_08_430070	26	50	et	et	FW
10_1101-2021_02_08_430070	26	51	al	al	NNP
10_1101-2021_02_08_430070	26	52	.	.	NNP
10_1101-2021_02_08_430070	26	53	,	,	,
10_1101-2021_02_08_430070	26	54	2019	2019	CD
10_1101-2021_02_08_430070	26	55	)	)	-RRB-
10_1101-2021_02_08_430070	26	56	)	)	-RRB-
10_1101-2021_02_08_430070	26	57	.	.	.
10_1101-2021_02_08_430070	27	1	A	a	DT
10_1101-2021_02_08_430070	27	2	testing-	testing-	NN
10_1101-2021_02_08_430070	27	3	based	base	VBN
10_1101-2021_02_08_430070	27	4	approach	approach	NN
10_1101-2021_02_08_430070	27	5	performs	perform	VBZ
10_1101-2021_02_08_430070	27	6	statistical	statistical	JJ
10_1101-2021_02_08_430070	27	7	test	test	NN
10_1101-2021_02_08_430070	27	8	on	on	IN
10_1101-2021_02_08_430070	27	9	paired	pair	VBN
10_1101-2021_02_08_430070	27	10	signals	signal	NNS
10_1101-2021_02_08_430070	27	11	(	(	-LRB-
10_1101-2021_02_08_430070	27	12	candidate	candidate	NN
10_1101-2021_02_08_430070	27	13	and	and	CC
10_1101-2021_02_08_430070	27	14	reference	reference	NN
10_1101-2021_02_08_430070	27	15	)	)	-RRB-
10_1101-2021_02_08_430070	27	16	and	and	CC
10_1101-2021_02_08_430070	27	17	does	do	VBZ
10_1101-2021_02_08_430070	27	18	not	not	RB
10_1101-2021_02_08_430070	27	19	require	require	VB
10_1101-2021_02_08_430070	27	20	any	any	DT
10_1101-2021_02_08_430070	27	21	training	training	NN
10_1101-2021_02_08_430070	27	22	process	process	NN
10_1101-2021_02_08_430070	27	23	.	.	.
10_1101-2021_02_08_430070	28	1	Also	also	RB
10_1101-2021_02_08_430070	28	2	,	,	,
10_1101-2021_02_08_430070	28	3	it	-PRON-	PRP
10_1101-2021_02_08_430070	28	4	can	can	MD
10_1101-2021_02_08_430070	28	5	be	be	VB
10_1101-2021_02_08_430070	28	6	applied	apply	VBN
10_1101-2021_02_08_430070	28	7	for	for	IN
10_1101-2021_02_08_430070	28	8	any	any	DT
10_1101-2021_02_08_430070	28	9	chemical	chemical	NN
10_1101-2021_02_08_430070	28	10	modifications	modification	NNS
10_1101-2021_02_08_430070	28	11	.	.	.
10_1101-2021_02_08_430070	29	1	A	a	DT
10_1101-2021_02_08_430070	29	2	model	model	NN
10_1101-2021_02_08_430070	29	3	-	-	HYPH
10_1101-2021_02_08_430070	29	4	based	base	VBN
10_1101-2021_02_08_430070	29	5	approach	approach	NN
10_1101-2021_02_08_430070	29	6	trains	train	VBZ
10_1101-2021_02_08_430070	29	7	a	a	DT
10_1101-2021_02_08_430070	29	8	model	model	NN
10_1101-2021_02_08_430070	29	9	1	1	CD
10_1101-2021_02_08_430070	29	10	.license	.license	.
10_1101-2021_02_08_430070	29	11	CC	cc	NN
10_1101-2021_02_08_430070	29	12	-	-	HYPH
10_1101-2021_02_08_430070	29	13	BY	BY	NNP
10_1101-2021_02_08_430070	29	14	-	-	HYPH
10_1101-2021_02_08_430070	29	15	NC	NC	NNP
10_1101-2021_02_08_430070	29	16	-	-	HYPH
10_1101-2021_02_08_430070	29	17	ND	ND	NNP
10_1101-2021_02_08_430070	29	18	4.0	4.0	CD
10_1101-2021_02_08_430070	29	19	Internationalpeer	Internationalpeer	NNP
10_1101-2021_02_08_430070	29	20	review	review	NN
10_1101-2021_02_08_430070	29	21	)	)	-RRB-
10_1101-2021_02_08_430070	29	22	is	be	VBZ
10_1101-2021_02_08_430070	29	23	the	the	DT
10_1101-2021_02_08_430070	29	24	author	author	NN
10_1101-2021_02_08_430070	29	25	/	/	SYM
10_1101-2021_02_08_430070	29	26	funder	funder	NN
10_1101-2021_02_08_430070	29	27	,	,	,
10_1101-2021_02_08_430070	29	28	who	who	WP
10_1101-2021_02_08_430070	29	29	has	have	VBZ
10_1101-2021_02_08_430070	29	30	granted	grant	VBN
10_1101-2021_02_08_430070	29	31	bioRxiv	biorxiv	IN
10_1101-2021_02_08_430070	29	32	a	a	DT
10_1101-2021_02_08_430070	29	33	license	license	NN
10_1101-2021_02_08_430070	29	34	to	to	TO
10_1101-2021_02_08_430070	29	35	display	display	VB
10_1101-2021_02_08_430070	29	36	the	the	DT
10_1101-2021_02_08_430070	29	37	preprint	preprint	NN
10_1101-2021_02_08_430070	29	38	in	in	IN
10_1101-2021_02_08_430070	29	39	perpetuity	perpetuity	NN
10_1101-2021_02_08_430070	29	40	.	.	.
10_1101-2021_02_08_430070	30	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	30	2	is	be	VBZ
10_1101-2021_02_08_430070	30	3	made	make	VBN
10_1101-2021_02_08_430070	30	4	available	available	JJ
10_1101-2021_02_08_430070	30	5	under	under	IN
10_1101-2021_02_08_430070	30	6	a	a	DT
10_1101-2021_02_08_430070	30	7	The	the	DT
10_1101-2021_02_08_430070	30	8	copyright	copyright	NN
10_1101-2021_02_08_430070	30	9	holder	holder	NN
10_1101-2021_02_08_430070	30	10	for	for	IN
10_1101-2021_02_08_430070	30	11	this	this	DT
10_1101-2021_02_08_430070	30	12	preprint	preprint	NN
10_1101-2021_02_08_430070	30	13	(	(	-LRB-
10_1101-2021_02_08_430070	30	14	which	which	WDT
10_1101-2021_02_08_430070	30	15	was	be	VBD
10_1101-2021_02_08_430070	30	16	not	not	RB
10_1101-2021_02_08_430070	30	17	certified	certify	VBN
10_1101-2021_02_08_430070	30	18	bythis	bythis	DT
10_1101-2021_02_08_430070	30	19	version	version	NN
10_1101-2021_02_08_430070	30	20	posted	post	VBD
10_1101-2021_02_08_430070	30	21	February	February	NNP
10_1101-2021_02_08_430070	30	22	10	10	CD
10_1101-2021_02_08_430070	30	23	,	,	,
10_1101-2021_02_08_430070	30	24	2021	2021	CD
10_1101-2021_02_08_430070	30	25	.	.	.
10_1101-2021_02_08_430070	30	26	;	;	:
10_1101-2021_02_08_430070	30	27	https://doi.org/10.1101/2021.02.08.430070doi	https://doi.org/10.1101/2021.02.08.430070doi	NFP
10_1101-2021_02_08_430070	30	28	:	:	:
10_1101-2021_02_08_430070	30	29	bioRxiv	biorxiv	VB
10_1101-2021_02_08_430070	30	30	preprint	preprint	NN
10_1101-2021_02_08_430070	30	31	https://doi.org/10.1101/2021.02.08.430070	https://doi.org/10.1101/2021.02.08.430070	UH
10_1101-2021_02_08_430070	30	32	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	30	33	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	30	34	✐	✐	NNP
10_1101-2021_02_08_430070	30	35	✐	✐	NNP
10_1101-2021_02_08_430070	30	36	✐	✐	NNP
10_1101-2021_02_08_430070	30	37	✐	✐	NNP
10_1101-2021_02_08_430070	30	38	✐	✐	NNP
10_1101-2021_02_08_430070	30	39	✐	✐	NNP
10_1101-2021_02_08_430070	30	40	✐	✐	NNP
10_1101-2021_02_08_430070	30	41	✐	✐	NNP
10_1101-2021_02_08_430070	30	42	2	2	CD
10_1101-2021_02_08_430070	30	43	Zhang	Zhang	NNP
10_1101-2021_02_08_430070	30	44	et	et	NNP
10_1101-2021_02_08_430070	30	45	al	al	NNP
10_1101-2021_02_08_430070	30	46	.	.	.
10_1101-2021_02_08_430070	31	1	x	x	SYM
10_1101-2021_02_08_430070	31	2	x	x	SYM
10_1101-2021_02_08_430070	31	3	x	x	NNP
10_1101-2021_02_08_430070	31	4	x	x	NN
10_1101-2021_02_08_430070	31	5	x1	x1	NN
10_1101-2021_02_08_430070	31	6	2	2	CD
10_1101-2021_02_08_430070	31	7	i	i	NN
10_1101-2021_02_08_430070	31	8	n-1	n-1	NNP
10_1101-2021_02_08_430070	31	9	n	n	NN
10_1101-2021_02_08_430070	31	10	......	......	NFP
10_1101-2021_02_08_430070	31	11	......	......	NFP
10_1101-2021_02_08_430070	31	12	Embedding	embed	VBG
10_1101-2021_02_08_430070	31	13	Attention	attention	NN
10_1101-2021_02_08_430070	31	14	Feed	feed	NN
10_1101-2021_02_08_430070	31	15	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	31	16	Attention	Attention	NNP
10_1101-2021_02_08_430070	31	17	Feed	feed	NN
10_1101-2021_02_08_430070	31	18	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	31	19	Attention	Attention	NNP
10_1101-2021_02_08_430070	31	20	Linear	Linear	NNP
10_1101-2021_02_08_430070	31	21	Methylation	Methylation	NNP
10_1101-2021_02_08_430070	31	22	status	status	NN
10_1101-2021_02_08_430070	31	23	Feed	Feed	NNP
10_1101-2021_02_08_430070	31	24	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	31	25	Linear	Linear	NNP
10_1101-2021_02_08_430070	31	26	C	c	NN
10_1101-2021_02_08_430070	31	27	G	g	NN
10_1101-2021_02_08_430070	31	28	5mcC	5mcc	CD
10_1101-2021_02_08_430070	31	29	A	a	NN
10_1101-2021_02_08_430070	31	30	T	t	NN
10_1101-2021_02_08_430070	31	31	A	a	NN
10_1101-2021_02_08_430070	31	32	5	5	CD
10_1101-2021_02_08_430070	31	33	’	’	NN
10_1101-2021_02_08_430070	31	34	3	3	CD
10_1101-2021_02_08_430070	31	35	’	'	''
10_1101-2021_02_08_430070	31	36	DNA	dna	NN
10_1101-2021_02_08_430070	31	37	sequence	sequence	NN
10_1101-2021_02_08_430070	31	38	x	x	SYM
10_1101-2021_02_08_430070	31	39	ix	ix	NNP
10_1101-2021_02_08_430070	31	40	i	i	NNP
10_1101-2021_02_08_430070	31	41	-	-	HYPH
10_1101-2021_02_08_430070	31	42	k	k	NNP
10_1101-2021_02_08_430070	31	43	x	x	NN
10_1101-2021_02_08_430070	31	44	i+k	i+k	NNP
10_1101-2021_02_08_430070	31	45	W	W	NNP
10_1101-2021_02_08_430070	31	46	V	V	NNP
10_1101-2021_02_08_430070	31	47	−	−	NNP
10_1101-2021_02_08_430070	31	48	k	k	NN
10_1101-2021_02_08_430070	31	49	,	,	,
10_1101-2021_02_08_430070	31	50	W	W	NNP
10_1101-2021_02_08_430070	31	51	K	K	NNP
10_1101-2021_02_08_430070	31	52	−	−	NNP
10_1101-2021_02_08_430070	31	53	k	k	NN
10_1101-2021_02_08_430070	31	54	W	w	NN
10_1101-2021_02_08_430070	31	55	V	v	NN
10_1101-2021_02_08_430070	31	56	k	k	NN
10_1101-2021_02_08_430070	31	57	,	,	,
10_1101-2021_02_08_430070	31	58	W	w	NN
10_1101-2021_02_08_430070	31	59	K	K	NNP
10_1101-2021_02_08_430070	31	60	k	k	NN
10_1101-2021_02_08_430070	31	61	......	......	NFP
10_1101-2021_02_08_430070	31	62	............	............	NFP
10_1101-2021_02_08_430070	31	63	......	......	.
10_1101-2021_02_08_430070	32	1	Attention	attention	NN
10_1101-2021_02_08_430070	32	2	Feed	feed	NN
10_1101-2021_02_08_430070	32	3	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	4	Attention	Attention	NNP
10_1101-2021_02_08_430070	32	5	Feed	feed	NN
10_1101-2021_02_08_430070	32	6	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	7	Attention	Attention	NNP
10_1101-2021_02_08_430070	32	8	Concate	Concate	NNP
10_1101-2021_02_08_430070	32	9	Methylation	Methylation	NNP
10_1101-2021_02_08_430070	32	10	status	status	NN
10_1101-2021_02_08_430070	32	11	Feed	Feed	NNP
10_1101-2021_02_08_430070	32	12	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	13	Attention	attention	NN
10_1101-2021_02_08_430070	32	14	Feed	feed	NN
10_1101-2021_02_08_430070	32	15	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	16	Attention	attention	NN
10_1101-2021_02_08_430070	32	17	Feed	feed	NN
10_1101-2021_02_08_430070	32	18	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	19	Attention	Attention	NNP
10_1101-2021_02_08_430070	32	20	Feed	Feed	NNP
10_1101-2021_02_08_430070	32	21	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	22	Linear	Linear	NNP
10_1101-2021_02_08_430070	32	23	(	(	-LRB-
10_1101-2021_02_08_430070	32	24	tanh	tanh	NN
10_1101-2021_02_08_430070	32	25	)	)	-RRB-
10_1101-2021_02_08_430070	32	26	Attention	attention	NN
10_1101-2021_02_08_430070	32	27	Feed	Feed	NNP
10_1101-2021_02_08_430070	32	28	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	29	Attention	attention	NN
10_1101-2021_02_08_430070	32	30	Feed	feed	NN
10_1101-2021_02_08_430070	32	31	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	32	Attention	Attention	NNP
10_1101-2021_02_08_430070	32	33	Feed	Feed	NNP
10_1101-2021_02_08_430070	32	34	Forwad	Forwad	NNP
10_1101-2021_02_08_430070	32	35	relative	relative	JJ
10_1101-2021_02_08_430070	32	36	position	position	NN
10_1101-2021_02_08_430070	32	37	constraint	constraint	NN
10_1101-2021_02_08_430070	32	38	window	window	NN
10_1101-2021_02_08_430070	32	39	x1	x1	NNP
10_1101-2021_02_08_430070	32	40	xn	xn	NNP
10_1101-2021_02_08_430070	32	41	(	(	-LRB-
10_1101-2021_02_08_430070	32	42	a	a	DT
10_1101-2021_02_08_430070	32	43	)	)	-RRB-
10_1101-2021_02_08_430070	32	44	.	.	.
10_1101-2021_02_08_430070	33	1	Basic	basic	JJ
10_1101-2021_02_08_430070	33	2	BERT	BERT	NNP
10_1101-2021_02_08_430070	33	3	for	for	IN
10_1101-2021_02_08_430070	33	4	methyaltion	methyaltion	NN
10_1101-2021_02_08_430070	33	5	detection	detection	NN
10_1101-2021_02_08_430070	33	6	(	(	-LRB-
10_1101-2021_02_08_430070	33	7	b	b	NN
10_1101-2021_02_08_430070	33	8	)	)	-RRB-
10_1101-2021_02_08_430070	33	9	.	.	.
10_1101-2021_02_08_430070	34	1	Refined	refined	JJ
10_1101-2021_02_08_430070	34	2	BERT	BERT	NNP
10_1101-2021_02_08_430070	34	3	with	with	IN
10_1101-2021_02_08_430070	34	4	relative	relative	JJ
10_1101-2021_02_08_430070	34	5	position	position	NN
10_1101-2021_02_08_430070	34	6	representation	representation	NN
10_1101-2021_02_08_430070	34	7	Fig	Fig	NNP
10_1101-2021_02_08_430070	34	8	.	.	.
10_1101-2021_02_08_430070	35	1	1	1	LS
10_1101-2021_02_08_430070	35	2	:	:	:
10_1101-2021_02_08_430070	35	3	Basic	Basic	NNP
10_1101-2021_02_08_430070	35	4	BERT	BERT	NNP
10_1101-2021_02_08_430070	35	5	’s	’s	POS
10_1101-2021_02_08_430070	35	6	and	and	CC
10_1101-2021_02_08_430070	35	7	refined	refine	VBD
10_1101-2021_02_08_430070	35	8	BERT	BERT	NNP
10_1101-2021_02_08_430070	35	9	’s	’s	POS
10_1101-2021_02_08_430070	35	10	model	model	NN
10_1101-2021_02_08_430070	35	11	structure	structure	NN
10_1101-2021_02_08_430070	35	12	used	use	VBN
10_1101-2021_02_08_430070	35	13	for	for	IN
10_1101-2021_02_08_430070	35	14	methylation	methylation	NN
10_1101-2021_02_08_430070	35	15	detection	detection	NN
10_1101-2021_02_08_430070	35	16	.	.	.
10_1101-2021_02_08_430070	36	1	Compared	compare	VBN
10_1101-2021_02_08_430070	36	2	with	with	IN
10_1101-2021_02_08_430070	36	3	the	the	DT
10_1101-2021_02_08_430070	36	4	basic	basic	JJ
10_1101-2021_02_08_430070	36	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	36	6	,	,	,
10_1101-2021_02_08_430070	36	7	enhanced	enhanced	JJ
10_1101-2021_02_08_430070	36	8	constraints	constraint	NNS
10_1101-2021_02_08_430070	36	9	and	and	CC
10_1101-2021_02_08_430070	36	10	additional	additional	JJ
10_1101-2021_02_08_430070	36	11	edges	edge	NNS
10_1101-2021_02_08_430070	36	12	are	be	VBP
10_1101-2021_02_08_430070	36	13	highlighted	highlight	VBN
10_1101-2021_02_08_430070	36	14	in	in	IN
10_1101-2021_02_08_430070	36	15	red	red	JJ
10_1101-2021_02_08_430070	36	16	color	color	NN
10_1101-2021_02_08_430070	36	17	.	.	.
10_1101-2021_02_08_430070	37	1	on	on	IN
10_1101-2021_02_08_430070	37	2	known	know	VBN
10_1101-2021_02_08_430070	37	3	chemical	chemical	NN
10_1101-2021_02_08_430070	37	4	modifications	modification	NNS
10_1101-2021_02_08_430070	37	5	and	and	CC
10_1101-2021_02_08_430070	37	6	makes	make	VBZ
10_1101-2021_02_08_430070	37	7	predictions	prediction	NNS
10_1101-2021_02_08_430070	37	8	whether	whether	IN
10_1101-2021_02_08_430070	37	9	a	a	DT
10_1101-2021_02_08_430070	37	10	signal	signal	JJ
10_1101-2021_02_08_430070	37	11	sequence	sequence	NN
10_1101-2021_02_08_430070	37	12	contains	contain	VBZ
10_1101-2021_02_08_430070	37	13	methylation	methylation	NN
10_1101-2021_02_08_430070	37	14	signals	signal	NNS
10_1101-2021_02_08_430070	37	15	or	or	CC
10_1101-2021_02_08_430070	37	16	not	not	RB
10_1101-2021_02_08_430070	37	17	.	.	.
10_1101-2021_02_08_430070	38	1	Sequential	sequential	JJ
10_1101-2021_02_08_430070	38	2	models	model	NNS
10_1101-2021_02_08_430070	38	3	,	,	,
10_1101-2021_02_08_430070	38	4	such	such	JJ
10_1101-2021_02_08_430070	38	5	as	as	IN
10_1101-2021_02_08_430070	38	6	hidden	hide	VBN
10_1101-2021_02_08_430070	38	7	Markov	Markov	NNP
10_1101-2021_02_08_430070	38	8	model	model	NN
10_1101-2021_02_08_430070	38	9	(	(	-LRB-
10_1101-2021_02_08_430070	38	10	HMM	HMM	NNP
10_1101-2021_02_08_430070	38	11	)	)	-RRB-
10_1101-2021_02_08_430070	38	12	and	and	CC
10_1101-2021_02_08_430070	38	13	bidirectional	bidirectional	JJ
10_1101-2021_02_08_430070	38	14	recurrent	recurrent	NN
10_1101-2021_02_08_430070	38	15	neural	neural	JJ
10_1101-2021_02_08_430070	38	16	network	network	NN
10_1101-2021_02_08_430070	38	17	(	(	-LRB-
10_1101-2021_02_08_430070	38	18	biRNN	biRNN	NNP
10_1101-2021_02_08_430070	38	19	)	)	-RRB-
10_1101-2021_02_08_430070	38	20	,	,	,
10_1101-2021_02_08_430070	38	21	are	be	VBP
10_1101-2021_02_08_430070	38	22	commonly	commonly	RB
10_1101-2021_02_08_430070	38	23	used	use	VBN
10_1101-2021_02_08_430070	38	24	in	in	IN
10_1101-2021_02_08_430070	38	25	the	the	DT
10_1101-2021_02_08_430070	38	26	model	model	NN
10_1101-2021_02_08_430070	38	27	-	-	HYPH
10_1101-2021_02_08_430070	38	28	based	base	VBN
10_1101-2021_02_08_430070	38	29	approach	approach	NN
10_1101-2021_02_08_430070	38	30	.	.	.
10_1101-2021_02_08_430070	39	1	Although	although	IN
10_1101-2021_02_08_430070	39	2	model	model	NN
10_1101-2021_02_08_430070	39	3	-	-	HYPH
10_1101-2021_02_08_430070	39	4	based	base	VBN
10_1101-2021_02_08_430070	39	5	approaches	approach	NNS
10_1101-2021_02_08_430070	39	6	have	have	VBP
10_1101-2021_02_08_430070	39	7	already	already	RB
10_1101-2021_02_08_430070	39	8	achieved	achieve	VBN
10_1101-2021_02_08_430070	39	9	competitive	competitive	JJ
10_1101-2021_02_08_430070	39	10	results	result	NNS
10_1101-2021_02_08_430070	39	11	,	,	,
10_1101-2021_02_08_430070	39	12	the	the	DT
10_1101-2021_02_08_430070	39	13	sequential	sequential	JJ
10_1101-2021_02_08_430070	39	14	computational	computational	JJ
10_1101-2021_02_08_430070	39	15	order	order	NN
10_1101-2021_02_08_430070	39	16	makes	make	VBZ
10_1101-2021_02_08_430070	39	17	them	-PRON-	PRP
10_1101-2021_02_08_430070	39	18	difficult	difficult	JJ
10_1101-2021_02_08_430070	39	19	to	to	TO
10_1101-2021_02_08_430070	39	20	be	be	VB
10_1101-2021_02_08_430070	39	21	optimized	optimize	VBN
10_1101-2021_02_08_430070	39	22	in	in	IN
10_1101-2021_02_08_430070	39	23	parallel	parallel	NN
10_1101-2021_02_08_430070	39	24	for	for	IN
10_1101-2021_02_08_430070	39	25	fast	fast	JJ
10_1101-2021_02_08_430070	39	26	inference	inference	NN
10_1101-2021_02_08_430070	39	27	.	.	.
10_1101-2021_02_08_430070	40	1	Meanwhile	meanwhile	RB
10_1101-2021_02_08_430070	40	2	,	,	,
10_1101-2021_02_08_430070	40	3	finding	find	VBG
10_1101-2021_02_08_430070	40	4	discriminative	discriminative	JJ
10_1101-2021_02_08_430070	40	5	signal	signal	NN
10_1101-2021_02_08_430070	40	6	patterns	pattern	NNS
10_1101-2021_02_08_430070	40	7	for	for	IN
10_1101-2021_02_08_430070	40	8	identifying	identify	VBG
10_1101-2021_02_08_430070	40	9	methylated	methylate	VBN
10_1101-2021_02_08_430070	40	10	signals	signal	NNS
10_1101-2021_02_08_430070	40	11	is	be	VBZ
10_1101-2021_02_08_430070	40	12	also	also	RB
10_1101-2021_02_08_430070	40	13	important	important	JJ
10_1101-2021_02_08_430070	40	14	for	for	IN
10_1101-2021_02_08_430070	40	15	developing	develop	VBG
10_1101-2021_02_08_430070	40	16	novel	novel	JJ
10_1101-2021_02_08_430070	40	17	detection	detection	NN
10_1101-2021_02_08_430070	40	18	algorithms	algorithm	NNS
10_1101-2021_02_08_430070	40	19	.	.	.
10_1101-2021_02_08_430070	41	1	In	in	IN
10_1101-2021_02_08_430070	41	2	this	this	DT
10_1101-2021_02_08_430070	41	3	work	work	NN
10_1101-2021_02_08_430070	41	4	,	,	,
10_1101-2021_02_08_430070	41	5	based	base	VBN
10_1101-2021_02_08_430070	41	6	on	on	IN
10_1101-2021_02_08_430070	41	7	the	the	DT
10_1101-2021_02_08_430070	41	8	bidirectional	bidirectional	JJ
10_1101-2021_02_08_430070	41	9	encoder	encoder	NN
10_1101-2021_02_08_430070	41	10	representations	representation	NNS
10_1101-2021_02_08_430070	41	11	from	from	IN
10_1101-2021_02_08_430070	41	12	transformers	transformer	NNS
10_1101-2021_02_08_430070	41	13	(	(	-LRB-
10_1101-2021_02_08_430070	41	14	BERT	BERT	NNP
10_1101-2021_02_08_430070	41	15	)	)	-RRB-
10_1101-2021_02_08_430070	41	16	,	,	,
10_1101-2021_02_08_430070	41	17	we	-PRON-	PRP
10_1101-2021_02_08_430070	41	18	explore	explore	VBP
10_1101-2021_02_08_430070	41	19	the	the	DT
10_1101-2021_02_08_430070	41	20	non	non	JJ
10_1101-2021_02_08_430070	41	21	-	-	JJ
10_1101-2021_02_08_430070	41	22	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	41	23	modeling	modeling	NN
10_1101-2021_02_08_430070	41	24	approach	approach	NN
10_1101-2021_02_08_430070	41	25	for	for	IN
10_1101-2021_02_08_430070	41	26	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	41	27	methylation	methylation	NN
10_1101-2021_02_08_430070	41	28	detection	detection	NN
10_1101-2021_02_08_430070	41	29	.	.	.
10_1101-2021_02_08_430070	42	1	Though	though	IN
10_1101-2021_02_08_430070	42	2	analyzing	analyze	VBG
10_1101-2021_02_08_430070	42	3	nucleotide	nucleotide	JJ
10_1101-2021_02_08_430070	42	4	sequences	sequence	NNS
10_1101-2021_02_08_430070	42	5	with	with	IN
10_1101-2021_02_08_430070	42	6	both	both	CC
10_1101-2021_02_08_430070	42	7	methylated	methylated	JJ
10_1101-2021_02_08_430070	42	8	and	and	CC
10_1101-2021_02_08_430070	42	9	unmethylated	unmethylated	JJ
10_1101-2021_02_08_430070	42	10	signals	signal	NNS
10_1101-2021_02_08_430070	42	11	,	,	,
10_1101-2021_02_08_430070	42	12	we	-PRON-	PRP
10_1101-2021_02_08_430070	42	13	profile	profile	VBP
10_1101-2021_02_08_430070	42	14	positional	positional	JJ
10_1101-2021_02_08_430070	42	15	signal	signal	NN
10_1101-2021_02_08_430070	42	16	-	-	HYPH
10_1101-2021_02_08_430070	42	17	shift	shift	NN
10_1101-2021_02_08_430070	42	18	for	for	IN
10_1101-2021_02_08_430070	42	19	different	different	JJ
10_1101-2021_02_08_430070	42	20	motifs	motif	NNS
10_1101-2021_02_08_430070	42	21	and	and	CC
10_1101-2021_02_08_430070	42	22	methyltransferases	methyltransferase	NNS
10_1101-2021_02_08_430070	42	23	.	.	.
10_1101-2021_02_08_430070	43	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	43	2	find	find	VBP
10_1101-2021_02_08_430070	43	3	±3bp	±3bp	IN
10_1101-2021_02_08_430070	43	4	region	region	NN
10_1101-2021_02_08_430070	43	5	surrounding	surround	VBG
10_1101-2021_02_08_430070	43	6	the	the	DT
10_1101-2021_02_08_430070	43	7	center	center	NN
10_1101-2021_02_08_430070	43	8	methylation	methylation	NN
10_1101-2021_02_08_430070	43	9	candidate	candidate	NN
10_1101-2021_02_08_430070	43	10	shows	show	VBZ
10_1101-2021_02_08_430070	43	11	significant	significant	JJ
10_1101-2021_02_08_430070	43	12	signal	signal	NN
10_1101-2021_02_08_430070	43	13	-	-	HYPH
10_1101-2021_02_08_430070	43	14	shifts	shift	NNS
10_1101-2021_02_08_430070	43	15	.	.	.
10_1101-2021_02_08_430070	44	1	Different	different	JJ
10_1101-2021_02_08_430070	44	2	methylation	methylation	NN
10_1101-2021_02_08_430070	44	3	types	type	NNS
10_1101-2021_02_08_430070	44	4	,	,	,
10_1101-2021_02_08_430070	44	5	such	such	JJ
10_1101-2021_02_08_430070	44	6	as	as	IN
10_1101-2021_02_08_430070	44	7	5-methylcytosine	5-methylcytosine	CD
10_1101-2021_02_08_430070	44	8	(	(	-LRB-
10_1101-2021_02_08_430070	44	9	5mC	5mc	CD
10_1101-2021_02_08_430070	44	10	)	)	-RRB-
10_1101-2021_02_08_430070	44	11	and	and	CC
10_1101-2021_02_08_430070	44	12	N6-methyladenine	N6-methyladenine	NNP
10_1101-2021_02_08_430070	44	13	(	(	-LRB-
10_1101-2021_02_08_430070	44	14	6mA	6ma	CD
10_1101-2021_02_08_430070	44	15	)	)	-RRB-
10_1101-2021_02_08_430070	44	16	,	,	,
10_1101-2021_02_08_430070	44	17	also	also	RB
10_1101-2021_02_08_430070	44	18	demonstrate	demonstrate	VBP
10_1101-2021_02_08_430070	44	19	different	different	JJ
10_1101-2021_02_08_430070	44	20	signal	signal	JJ
10_1101-2021_02_08_430070	44	21	-	-	HYPH
10_1101-2021_02_08_430070	44	22	shift	shift	NN
10_1101-2021_02_08_430070	44	23	patterns	pattern	NNS
10_1101-2021_02_08_430070	44	24	.	.	.
10_1101-2021_02_08_430070	45	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	45	2	hence	hence	RB
10_1101-2021_02_08_430070	45	3	propose	propose	VBP
10_1101-2021_02_08_430070	45	4	a	a	DT
10_1101-2021_02_08_430070	45	5	refined	refined	JJ
10_1101-2021_02_08_430070	45	6	BERT	BERT	NNP
10_1101-2021_02_08_430070	45	7	model	model	NN
10_1101-2021_02_08_430070	45	8	to	to	TO
10_1101-2021_02_08_430070	45	9	take	take	VB
10_1101-2021_02_08_430070	45	10	account	account	NN
10_1101-2021_02_08_430070	45	11	of	of	IN
10_1101-2021_02_08_430070	45	12	signal	signal	JJ
10_1101-2021_02_08_430070	45	13	-	-	HYPH
10_1101-2021_02_08_430070	45	14	shift	shift	NN
10_1101-2021_02_08_430070	45	15	patterns	pattern	NNS
10_1101-2021_02_08_430070	45	16	in	in	IN
10_1101-2021_02_08_430070	45	17	the	the	DT
10_1101-2021_02_08_430070	45	18	modeling	modeling	NN
10_1101-2021_02_08_430070	45	19	.	.	.
10_1101-2021_02_08_430070	46	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	46	2	evaluate	evaluate	VBP
10_1101-2021_02_08_430070	46	3	the	the	DT
10_1101-2021_02_08_430070	46	4	proposed	propose	VBN
10_1101-2021_02_08_430070	46	5	methods	method	NNS
10_1101-2021_02_08_430070	46	6	on	on	IN
10_1101-2021_02_08_430070	46	7	the	the	DT
10_1101-2021_02_08_430070	46	8	publicly	publicly	RB
10_1101-2021_02_08_430070	46	9	available	available	JJ
10_1101-2021_02_08_430070	46	10	benchmark	benchmark	NN
10_1101-2021_02_08_430070	46	11	dataset	dataset	NN
10_1101-2021_02_08_430070	46	12	.	.	.
10_1101-2021_02_08_430070	47	1	In	in	IN
10_1101-2021_02_08_430070	47	2	both	both	DT
10_1101-2021_02_08_430070	47	3	in	in	IN
10_1101-2021_02_08_430070	47	4	-	-	HYPH
10_1101-2021_02_08_430070	47	5	sample	sample	NN
10_1101-2021_02_08_430070	47	6	and	and	CC
10_1101-2021_02_08_430070	47	7	cross	cross	JJ
10_1101-2021_02_08_430070	47	8	-	-	JJ
10_1101-2021_02_08_430070	47	9	sample	sample	JJ
10_1101-2021_02_08_430070	47	10	evaluation	evaluation	NN
10_1101-2021_02_08_430070	47	11	,	,	,
10_1101-2021_02_08_430070	47	12	the	the	DT
10_1101-2021_02_08_430070	47	13	proposed	propose	VBN
10_1101-2021_02_08_430070	47	14	refined	refine	VBN
10_1101-2021_02_08_430070	47	15	BERT	BERT	NNP
10_1101-2021_02_08_430070	47	16	model	model	NN
10_1101-2021_02_08_430070	47	17	achieves	achieve	VBZ
10_1101-2021_02_08_430070	47	18	a	a	DT
10_1101-2021_02_08_430070	47	19	competitive	competitive	JJ
10_1101-2021_02_08_430070	47	20	or	or	CC
10_1101-2021_02_08_430070	47	21	even	even	RB
10_1101-2021_02_08_430070	47	22	better	well	JJR
10_1101-2021_02_08_430070	47	23	result	result	VB
10_1101-2021_02_08_430070	47	24	when	when	WRB
10_1101-2021_02_08_430070	47	25	compared	compare	VBN
10_1101-2021_02_08_430070	47	26	with	with	IN
10_1101-2021_02_08_430070	47	27	the	the	DT
10_1101-2021_02_08_430070	47	28	state	state	NN
10_1101-2021_02_08_430070	47	29	-	-	HYPH
10_1101-2021_02_08_430070	47	30	of	of	IN
10_1101-2021_02_08_430070	47	31	-	-	HYPH
10_1101-2021_02_08_430070	47	32	the	the	DT
10_1101-2021_02_08_430070	47	33	-	-	HYPH
10_1101-2021_02_08_430070	47	34	art	art	NN
10_1101-2021_02_08_430070	47	35	biRNN	birnn	NN
10_1101-2021_02_08_430070	47	36	model	model	NN
10_1101-2021_02_08_430070	47	37	,	,	,
10_1101-2021_02_08_430070	47	38	while	while	IN
10_1101-2021_02_08_430070	47	39	its	-PRON-	PRP$
10_1101-2021_02_08_430070	47	40	model	model	NN
10_1101-2021_02_08_430070	47	41	inference	inference	NN
10_1101-2021_02_08_430070	47	42	speed	speed	NN
10_1101-2021_02_08_430070	47	43	is	be	VBZ
10_1101-2021_02_08_430070	47	44	about	about	RB
10_1101-2021_02_08_430070	47	45	6x	6x	CD
10_1101-2021_02_08_430070	47	46	faster	fast	RBR
10_1101-2021_02_08_430070	47	47	.	.	.
10_1101-2021_02_08_430070	48	1	In	in	IN
10_1101-2021_02_08_430070	48	2	the	the	DT
10_1101-2021_02_08_430070	48	3	cross	cross	JJ
10_1101-2021_02_08_430070	48	4	-	-	JJ
10_1101-2021_02_08_430070	48	5	sample	sample	JJ
10_1101-2021_02_08_430070	48	6	evaluation	evaluation	NN
10_1101-2021_02_08_430070	48	7	,	,	,
10_1101-2021_02_08_430070	48	8	BERT	BERT	NNP
10_1101-2021_02_08_430070	48	9	models	model	NNS
10_1101-2021_02_08_430070	48	10	also	also	RB
10_1101-2021_02_08_430070	48	11	demonstrate	demonstrate	VBP
10_1101-2021_02_08_430070	48	12	their	-PRON-	PRP$
10_1101-2021_02_08_430070	48	13	transfer	transfer	NN
10_1101-2021_02_08_430070	48	14	learning	learn	VBG
10_1101-2021_02_08_430070	48	15	ability	ability	NN
10_1101-2021_02_08_430070	48	16	across	across	IN
10_1101-2021_02_08_430070	48	17	different	different	JJ
10_1101-2021_02_08_430070	48	18	datasets	dataset	NNS
10_1101-2021_02_08_430070	48	19	.	.	.
10_1101-2021_02_08_430070	49	1	2	2	CD
10_1101-2021_02_08_430070	49	2	Methods	Methods	NNPS
10_1101-2021_02_08_430070	49	3	In	in	IN
10_1101-2021_02_08_430070	49	4	this	this	DT
10_1101-2021_02_08_430070	49	5	section	section	NN
10_1101-2021_02_08_430070	49	6	,	,	,
10_1101-2021_02_08_430070	49	7	we	-PRON-	PRP
10_1101-2021_02_08_430070	49	8	introduce	introduce	VBP
10_1101-2021_02_08_430070	49	9	BERT	BERT	NNP
10_1101-2021_02_08_430070	49	10	(	(	-LRB-
10_1101-2021_02_08_430070	49	11	Devlin	Devlin	NNP
10_1101-2021_02_08_430070	49	12	et	et	NNP
10_1101-2021_02_08_430070	49	13	al	al	NNP
10_1101-2021_02_08_430070	49	14	.	.	NNP
10_1101-2021_02_08_430070	49	15	,	,	,
10_1101-2021_02_08_430070	49	16	2018	2018	CD
10_1101-2021_02_08_430070	49	17	)	)	-RRB-
10_1101-2021_02_08_430070	49	18	and	and	CC
10_1101-2021_02_08_430070	49	19	refined	refine	VBN
10_1101-2021_02_08_430070	49	20	BERT	BERT	NNP
10_1101-2021_02_08_430070	49	21	applied	apply	VBN
10_1101-2021_02_08_430070	49	22	for	for	IN
10_1101-2021_02_08_430070	49	23	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	49	24	methylation	methylation	NN
10_1101-2021_02_08_430070	49	25	detection	detection	NN
10_1101-2021_02_08_430070	49	26	.	.	.
10_1101-2021_02_08_430070	50	1	The	the	DT
10_1101-2021_02_08_430070	50	2	BERT	BERT	NNP
10_1101-2021_02_08_430070	50	3	is	be	VBZ
10_1101-2021_02_08_430070	50	4	built	build	VBN
10_1101-2021_02_08_430070	50	5	on	on	IN
10_1101-2021_02_08_430070	50	6	the	the	DT
10_1101-2021_02_08_430070	50	7	base	base	NN
10_1101-2021_02_08_430070	50	8	of	of	IN
10_1101-2021_02_08_430070	50	9	Transformer	Transformer	NNP
10_1101-2021_02_08_430070	50	10	(	(	-LRB-
10_1101-2021_02_08_430070	50	11	Vaswani	Vaswani	NNP
10_1101-2021_02_08_430070	50	12	et	et	FW
10_1101-2021_02_08_430070	50	13	al	al	NNP
10_1101-2021_02_08_430070	50	14	.	.	NNP
10_1101-2021_02_08_430070	50	15	,	,	,
10_1101-2021_02_08_430070	50	16	2017	2017	CD
10_1101-2021_02_08_430070	50	17	)	)	-RRB-
10_1101-2021_02_08_430070	50	18	,	,	,
10_1101-2021_02_08_430070	50	19	which	which	WDT
10_1101-2021_02_08_430070	50	20	employs	employ	VBZ
10_1101-2021_02_08_430070	50	21	self	self	NN
10_1101-2021_02_08_430070	50	22	-	-	HYPH
10_1101-2021_02_08_430070	50	23	attention	attention	NN
10_1101-2021_02_08_430070	50	24	as	as	IN
10_1101-2021_02_08_430070	50	25	the	the	DT
10_1101-2021_02_08_430070	50	26	core	core	NN
10_1101-2021_02_08_430070	50	27	module	module	NN
10_1101-2021_02_08_430070	50	28	in	in	IN
10_1101-2021_02_08_430070	50	29	its	-PRON-	PRP$
10_1101-2021_02_08_430070	50	30	stacked	stack	VBN
10_1101-2021_02_08_430070	50	31	network	network	NN
10_1101-2021_02_08_430070	50	32	structure	structure	NN
10_1101-2021_02_08_430070	50	33	.	.	.
10_1101-2021_02_08_430070	51	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	51	2	is	be	VBZ
10_1101-2021_02_08_430070	51	3	proposed	propose	VBN
10_1101-2021_02_08_430070	51	4	to	to	TO
10_1101-2021_02_08_430070	51	5	replace	replace	VB
10_1101-2021_02_08_430070	51	6	recurrent	recurrent	NN
10_1101-2021_02_08_430070	51	7	and	and	CC
10_1101-2021_02_08_430070	51	8	convolution	convolution	NN
10_1101-2021_02_08_430070	51	9	operation	operation	NN
10_1101-2021_02_08_430070	51	10	with	with	IN
10_1101-2021_02_08_430070	51	11	purely	purely	RB
10_1101-2021_02_08_430070	51	12	attention	attention	NN
10_1101-2021_02_08_430070	51	13	mechanisms	mechanism	NNS
10_1101-2021_02_08_430070	51	14	.	.	.
10_1101-2021_02_08_430070	52	1	A	a	DT
10_1101-2021_02_08_430070	52	2	typical	typical	JJ
10_1101-2021_02_08_430070	52	3	transformer	transformer	NN
10_1101-2021_02_08_430070	52	4	network	network	NN
10_1101-2021_02_08_430070	52	5	consists	consist	VBZ
10_1101-2021_02_08_430070	52	6	of	of	IN
10_1101-2021_02_08_430070	52	7	encoding	encode	VBG
10_1101-2021_02_08_430070	52	8	and	and	CC
10_1101-2021_02_08_430070	52	9	decoding	decode	VBG
10_1101-2021_02_08_430070	52	10	module	module	NN
10_1101-2021_02_08_430070	52	11	.	.	.
10_1101-2021_02_08_430070	53	1	BERT	BERT	NNP
10_1101-2021_02_08_430070	53	2	only	only	RB
10_1101-2021_02_08_430070	53	3	uses	use	VBZ
10_1101-2021_02_08_430070	53	4	the	the	DT
10_1101-2021_02_08_430070	53	5	encoding	encode	VBG
10_1101-2021_02_08_430070	53	6	module	module	NN
10_1101-2021_02_08_430070	53	7	of	of	IN
10_1101-2021_02_08_430070	53	8	a	a	DT
10_1101-2021_02_08_430070	53	9	typical	typical	JJ
10_1101-2021_02_08_430070	53	10	transformer	transformer	NN
10_1101-2021_02_08_430070	53	11	for	for	IN
10_1101-2021_02_08_430070	53	12	pre-	pre-	NN
10_1101-2021_02_08_430070	53	13	training	training	NN
10_1101-2021_02_08_430070	53	14	on	on	IN
10_1101-2021_02_08_430070	53	15	the	the	DT
10_1101-2021_02_08_430070	53	16	unsupervised	unsupervised	JJ
10_1101-2021_02_08_430070	53	17	data	datum	NNS
10_1101-2021_02_08_430070	53	18	.	.	.
10_1101-2021_02_08_430070	54	1	BERT	BERT	NNP
10_1101-2021_02_08_430070	54	2	has	have	VBZ
10_1101-2021_02_08_430070	54	3	achieved	achieve	VBN
10_1101-2021_02_08_430070	54	4	break	break	NN
10_1101-2021_02_08_430070	54	5	-	-	HYPH
10_1101-2021_02_08_430070	54	6	through	through	RP
10_1101-2021_02_08_430070	54	7	results	result	NNS
10_1101-2021_02_08_430070	54	8	on	on	IN
10_1101-2021_02_08_430070	54	9	many	many	JJ
10_1101-2021_02_08_430070	54	10	natural	natural	JJ
10_1101-2021_02_08_430070	54	11	language	language	NN
10_1101-2021_02_08_430070	54	12	understanding	understanding	NN
10_1101-2021_02_08_430070	54	13	tasks	task	NNS
10_1101-2021_02_08_430070	54	14	.	.	.
10_1101-2021_02_08_430070	55	1	In	in	IN
10_1101-2021_02_08_430070	55	2	this	this	DT
10_1101-2021_02_08_430070	55	3	work	work	NN
10_1101-2021_02_08_430070	55	4	,	,	,
10_1101-2021_02_08_430070	55	5	we	-PRON-	PRP
10_1101-2021_02_08_430070	55	6	explore	explore	VBP
10_1101-2021_02_08_430070	55	7	applying	apply	VBG
10_1101-2021_02_08_430070	55	8	the	the	DT
10_1101-2021_02_08_430070	55	9	BERT	BERT	NNP
10_1101-2021_02_08_430070	55	10	model	model	NN
10_1101-2021_02_08_430070	55	11	for	for	IN
10_1101-2021_02_08_430070	55	12	the	the	DT
10_1101-2021_02_08_430070	55	13	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	55	14	methylation	methylation	NN
10_1101-2021_02_08_430070	55	15	detection	detection	NN
10_1101-2021_02_08_430070	55	16	task	task	NN
10_1101-2021_02_08_430070	55	17	to	to	TO
10_1101-2021_02_08_430070	55	18	leverage	leverage	VB
10_1101-2021_02_08_430070	55	19	the	the	DT
10_1101-2021_02_08_430070	55	20	power	power	NN
10_1101-2021_02_08_430070	55	21	of	of	IN
10_1101-2021_02_08_430070	55	22	advanced	advanced	JJ
10_1101-2021_02_08_430070	55	23	deep	deep	JJ
10_1101-2021_02_08_430070	55	24	learning	learning	NN
10_1101-2021_02_08_430070	55	25	models	model	NNS
10_1101-2021_02_08_430070	55	26	.	.	.
10_1101-2021_02_08_430070	56	1	2.1	2.1	CD
10_1101-2021_02_08_430070	56	2	BERT	BERT	NNS
10_1101-2021_02_08_430070	56	3	and	and	CC
10_1101-2021_02_08_430070	56	4	refined	refine	VBN
10_1101-2021_02_08_430070	56	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	56	6	model	model	NN
10_1101-2021_02_08_430070	56	7	Figure	figure	NN
10_1101-2021_02_08_430070	56	8	1	1	CD
10_1101-2021_02_08_430070	56	9	shows	show	VBZ
10_1101-2021_02_08_430070	56	10	the	the	DT
10_1101-2021_02_08_430070	56	11	model	model	NN
10_1101-2021_02_08_430070	56	12	structures	structure	NNS
10_1101-2021_02_08_430070	56	13	of	of	IN
10_1101-2021_02_08_430070	56	14	BERT	BERT	NNP
10_1101-2021_02_08_430070	56	15	models	model	NNS
10_1101-2021_02_08_430070	56	16	used	use	VBN
10_1101-2021_02_08_430070	56	17	for	for	IN
10_1101-2021_02_08_430070	56	18	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	56	19	methylation	methylation	NN
10_1101-2021_02_08_430070	56	20	detection	detection	NN
10_1101-2021_02_08_430070	56	21	.	.	.
10_1101-2021_02_08_430070	57	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	57	2	explore	explore	VBP
10_1101-2021_02_08_430070	57	3	two	two	CD
10_1101-2021_02_08_430070	57	4	types	type	NNS
10_1101-2021_02_08_430070	57	5	of	of	IN
10_1101-2021_02_08_430070	57	6	BERT	BERT	NNP
10_1101-2021_02_08_430070	57	7	models	model	NNS
10_1101-2021_02_08_430070	57	8	.	.	.
10_1101-2021_02_08_430070	58	1	One	one	CD
10_1101-2021_02_08_430070	58	2	is	be	VBZ
10_1101-2021_02_08_430070	58	3	the	the	DT
10_1101-2021_02_08_430070	58	4	most	most	RBS
10_1101-2021_02_08_430070	58	5	commonly	commonly	RB
10_1101-2021_02_08_430070	58	6	used	use	VBN
10_1101-2021_02_08_430070	58	7	BERT	BERT	NNP
10_1101-2021_02_08_430070	58	8	(	(	-LRB-
10_1101-2021_02_08_430070	58	9	Figure	figure	NN
10_1101-2021_02_08_430070	58	10	1(a	1(a	CD
10_1101-2021_02_08_430070	58	11	)	)	-RRB-
10_1101-2021_02_08_430070	58	12	)	)	-RRB-
10_1101-2021_02_08_430070	58	13	,	,	,
10_1101-2021_02_08_430070	58	14	the	the	DT
10_1101-2021_02_08_430070	58	15	other	other	JJ
10_1101-2021_02_08_430070	58	16	is	be	VBZ
10_1101-2021_02_08_430070	58	17	the	the	DT
10_1101-2021_02_08_430070	58	18	refined	refined	JJ
10_1101-2021_02_08_430070	58	19	BERT	BERT	NNP
10_1101-2021_02_08_430070	58	20	(	(	-LRB-
10_1101-2021_02_08_430070	58	21	Figure	figure	NN
10_1101-2021_02_08_430070	58	22	1(b	1(b	CD
10_1101-2021_02_08_430070	58	23	)	)	-RRB-
10_1101-2021_02_08_430070	58	24	)	)	-RRB-
10_1101-2021_02_08_430070	58	25	,	,	,
10_1101-2021_02_08_430070	58	26	which	which	WDT
10_1101-2021_02_08_430070	58	27	is	be	VBZ
10_1101-2021_02_08_430070	58	28	optimized	optimize	VBN
10_1101-2021_02_08_430070	58	29	for	for	IN
10_1101-2021_02_08_430070	58	30	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	58	31	methylation	methylation	NN
10_1101-2021_02_08_430070	58	32	detection	detection	NN
10_1101-2021_02_08_430070	58	33	.	.	.
10_1101-2021_02_08_430070	59	1	2.1.1	2.1.1	CD
10_1101-2021_02_08_430070	59	2	Embedding	embed	VBG
10_1101-2021_02_08_430070	59	3	module	module	NN
10_1101-2021_02_08_430070	59	4	Given	Given	NNP
10_1101-2021_02_08_430070	59	5	extracted	extract	VBD
10_1101-2021_02_08_430070	59	6	features	feature	NNS
10_1101-2021_02_08_430070	59	7	for	for	IN
10_1101-2021_02_08_430070	59	8	each	each	DT
10_1101-2021_02_08_430070	59	9	position	position	NN
10_1101-2021_02_08_430070	59	10	in	in	IN
10_1101-2021_02_08_430070	59	11	a	a	DT
10_1101-2021_02_08_430070	59	12	sequence	sequence	NN
10_1101-2021_02_08_430070	59	13	,	,	,
10_1101-2021_02_08_430070	59	14	the	the	DT
10_1101-2021_02_08_430070	59	15	embedding	embedding	NN
10_1101-2021_02_08_430070	59	16	layer	layer	NN
10_1101-2021_02_08_430070	59	17	maps	map	VBZ
10_1101-2021_02_08_430070	59	18	input	input	NN
10_1101-2021_02_08_430070	59	19	vectors	vector	NNS
10_1101-2021_02_08_430070	59	20	into	into	IN
10_1101-2021_02_08_430070	59	21	hidden	hide	VBN
10_1101-2021_02_08_430070	59	22	spaces	space	NNS
10_1101-2021_02_08_430070	59	23	.	.	.
10_1101-2021_02_08_430070	60	1	In	in	IN
10_1101-2021_02_08_430070	60	2	the	the	DT
10_1101-2021_02_08_430070	60	3	embedding	embedding	NN
10_1101-2021_02_08_430070	60	4	layer	layer	NN
10_1101-2021_02_08_430070	60	5	,	,	,
10_1101-2021_02_08_430070	60	6	besides	besides	IN
10_1101-2021_02_08_430070	60	7	event	event	NN
10_1101-2021_02_08_430070	60	8	embedding	embedding	NN
10_1101-2021_02_08_430070	60	9	,	,	,
10_1101-2021_02_08_430070	60	10	positional	positional	JJ
10_1101-2021_02_08_430070	60	11	embedding	embedding	NN
10_1101-2021_02_08_430070	60	12	(	(	-LRB-
10_1101-2021_02_08_430070	60	13	PE	PE	NNP
10_1101-2021_02_08_430070	60	14	)	)	-RRB-
10_1101-2021_02_08_430070	60	15	is	be	VBZ
10_1101-2021_02_08_430070	60	16	also	also	RB
10_1101-2021_02_08_430070	60	17	included	include	VBN
10_1101-2021_02_08_430070	60	18	.	.	.
10_1101-2021_02_08_430070	61	1	As	as	IN
10_1101-2021_02_08_430070	61	2	a	a	DT
10_1101-2021_02_08_430070	61	3	BERT	BERT	NNP
10_1101-2021_02_08_430070	61	4	is	be	VBZ
10_1101-2021_02_08_430070	61	5	used	use	VBN
10_1101-2021_02_08_430070	61	6	to	to	TO
10_1101-2021_02_08_430070	61	7	learn	learn	VB
10_1101-2021_02_08_430070	61	8	bidirectional	bidirectional	JJ
10_1101-2021_02_08_430070	61	9	contextual	contextual	JJ
10_1101-2021_02_08_430070	61	10	information	information	NN
10_1101-2021_02_08_430070	61	11	,	,	,
10_1101-2021_02_08_430070	61	12	positional	positional	JJ
10_1101-2021_02_08_430070	61	13	information	information	NN
10_1101-2021_02_08_430070	61	14	is	be	VBZ
10_1101-2021_02_08_430070	61	15	important	important	JJ
10_1101-2021_02_08_430070	61	16	in	in	IN
10_1101-2021_02_08_430070	61	17	the	the	DT
10_1101-2021_02_08_430070	61	18	modeling	modeling	NN
10_1101-2021_02_08_430070	61	19	.	.	.
10_1101-2021_02_08_430070	62	1	The	the	DT
10_1101-2021_02_08_430070	62	2	original	original	JJ
10_1101-2021_02_08_430070	62	3	PE	PE	NNP
10_1101-2021_02_08_430070	62	4	(	(	-LRB-
10_1101-2021_02_08_430070	62	5	Vaswani	Vaswani	NNP
10_1101-2021_02_08_430070	62	6	et	et	FW
10_1101-2021_02_08_430070	62	7	al	al	NNP
10_1101-2021_02_08_430070	62	8	.	.	NNP
10_1101-2021_02_08_430070	62	9	,	,	,
10_1101-2021_02_08_430070	62	10	2017	2017	CD
10_1101-2021_02_08_430070	62	11	)	)	-RRB-
10_1101-2021_02_08_430070	62	12	uses	use	VBZ
10_1101-2021_02_08_430070	62	13	a	a	DT
10_1101-2021_02_08_430070	62	14	sinusoid	sinusoid	JJ
10_1101-2021_02_08_430070	62	15	embedding	embedding	NN
10_1101-2021_02_08_430070	62	16	,	,	,
10_1101-2021_02_08_430070	62	17	which	which	WDT
10_1101-2021_02_08_430070	62	18	is	be	VBZ
10_1101-2021_02_08_430070	62	19	fixed	fix	VBN
10_1101-2021_02_08_430070	62	20	and	and	CC
10_1101-2021_02_08_430070	62	21	not	not	RB
10_1101-2021_02_08_430070	62	22	learnable	learnable	JJ
10_1101-2021_02_08_430070	62	23	.	.	.
10_1101-2021_02_08_430070	63	1	PE(pos	pe(pos	UH
10_1101-2021_02_08_430070	63	2	,	,	,
10_1101-2021_02_08_430070	63	3	2i	2i	NNP
10_1101-2021_02_08_430070	63	4	)	)	-RRB-
10_1101-2021_02_08_430070	63	5	=	=	NFP
10_1101-2021_02_08_430070	63	6	sin	sin	NN
10_1101-2021_02_08_430070	63	7	pos	pos	NNP
10_1101-2021_02_08_430070	63	8	100002i	100002i	NNP
10_1101-2021_02_08_430070	63	9	/	/	SYM
10_1101-2021_02_08_430070	63	10	dmodel	dmodel	NNP
10_1101-2021_02_08_430070	63	11	PE(pos	pe(pos	CD
10_1101-2021_02_08_430070	63	12	,	,	,
10_1101-2021_02_08_430070	63	13	2i	2i	NNP
10_1101-2021_02_08_430070	63	14	+	+	SYM
10_1101-2021_02_08_430070	63	15	1	1	CD
10_1101-2021_02_08_430070	63	16	)	)	-RRB-
10_1101-2021_02_08_430070	63	17	=	=	NFP
10_1101-2021_02_08_430070	63	18	cos	cos	NNP
10_1101-2021_02_08_430070	63	19	pos	pos	NNP
10_1101-2021_02_08_430070	63	20	100002i	100002i	NNP
10_1101-2021_02_08_430070	63	21	/	/	SYM
10_1101-2021_02_08_430070	63	22	dmodel	dmodel	NN
10_1101-2021_02_08_430070	63	23	,	,	,
10_1101-2021_02_08_430070	63	24	where	where	WRB
10_1101-2021_02_08_430070	63	25	pos	pos	NNP
10_1101-2021_02_08_430070	63	26	is	be	VBZ
10_1101-2021_02_08_430070	63	27	the	the	DT
10_1101-2021_02_08_430070	63	28	position	position	NN
10_1101-2021_02_08_430070	63	29	and	and	CC
10_1101-2021_02_08_430070	63	30	i	i	PRP
10_1101-2021_02_08_430070	63	31	is	be	VBZ
10_1101-2021_02_08_430070	63	32	the	the	DT
10_1101-2021_02_08_430070	63	33	embedding	embedding	NN
10_1101-2021_02_08_430070	63	34	dimension	dimension	NN
10_1101-2021_02_08_430070	63	35	.	.	.
10_1101-2021_02_08_430070	64	1	For	for	IN
10_1101-2021_02_08_430070	64	2	any	any	DT
10_1101-2021_02_08_430070	64	3	fixed	fix	VBN
10_1101-2021_02_08_430070	64	4	offset	offset	NN
10_1101-2021_02_08_430070	64	5	k	k	NN
10_1101-2021_02_08_430070	64	6	,	,	,
10_1101-2021_02_08_430070	64	7	PEpos+k	PEpos+k	NNP
10_1101-2021_02_08_430070	64	8	can	can	MD
10_1101-2021_02_08_430070	64	9	be	be	VB
10_1101-2021_02_08_430070	64	10	represented	represent	VBN
10_1101-2021_02_08_430070	64	11	as	as	IN
10_1101-2021_02_08_430070	64	12	a	a	DT
10_1101-2021_02_08_430070	64	13	linear	linear	JJ
10_1101-2021_02_08_430070	64	14	function	function	NN
10_1101-2021_02_08_430070	64	15	of	of	IN
10_1101-2021_02_08_430070	64	16	PEpos	pepos	PRP
10_1101-2021_02_08_430070	64	17	.	.	.
10_1101-2021_02_08_430070	65	1	According	accord	VBG
10_1101-2021_02_08_430070	65	2	to	to	IN
10_1101-2021_02_08_430070	65	3	the	the	DT
10_1101-2021_02_08_430070	65	4	recent	recent	JJ
10_1101-2021_02_08_430070	65	5	progress	progress	NN
10_1101-2021_02_08_430070	65	6	(	(	-LRB-
10_1101-2021_02_08_430070	65	7	Huang	Huang	NNP
10_1101-2021_02_08_430070	65	8	et	et	NNP
10_1101-2021_02_08_430070	65	9	al	al	NNP
10_1101-2021_02_08_430070	65	10	.	.	NNP
10_1101-2021_02_08_430070	65	11	,	,	,
10_1101-2021_02_08_430070	65	12	2020	2020	CD
10_1101-2021_02_08_430070	65	13	)	)	-RRB-
10_1101-2021_02_08_430070	65	14	,	,	,
10_1101-2021_02_08_430070	65	15	learnable	learnable	JJ
10_1101-2021_02_08_430070	65	16	PE	PE	NNP
10_1101-2021_02_08_430070	65	17	and	and	CC
10_1101-2021_02_08_430070	65	18	relative	relative	JJ
10_1101-2021_02_08_430070	65	19	position	position	NN
10_1101-2021_02_08_430070	65	20	embedding	embedding	NN
10_1101-2021_02_08_430070	65	21	can	can	MD
10_1101-2021_02_08_430070	65	22	help	help	VB
10_1101-2021_02_08_430070	65	23	to	to	TO
10_1101-2021_02_08_430070	65	24	further	further	RB
10_1101-2021_02_08_430070	65	25	improve	improve	VB
10_1101-2021_02_08_430070	65	26	BERT	BERT	NNP
10_1101-2021_02_08_430070	65	27	’s	’s	POS
10_1101-2021_02_08_430070	65	28	performances	performance	NNS
10_1101-2021_02_08_430070	65	29	.	.	.
10_1101-2021_02_08_430070	66	1	Therefore	therefore	RB
10_1101-2021_02_08_430070	66	2	,	,	,
10_1101-2021_02_08_430070	66	3	in	in	IN
10_1101-2021_02_08_430070	66	4	the	the	DT
10_1101-2021_02_08_430070	66	5	refined	refined	JJ
10_1101-2021_02_08_430070	66	6	BERT	BERT	NNP
10_1101-2021_02_08_430070	66	7	model	model	NN
10_1101-2021_02_08_430070	66	8	,	,	,
10_1101-2021_02_08_430070	66	9	we	-PRON-	PRP
10_1101-2021_02_08_430070	66	10	use	use	VBP
10_1101-2021_02_08_430070	66	11	learnable	learnable	JJ
10_1101-2021_02_08_430070	66	12	PE	PE	NNP
10_1101-2021_02_08_430070	66	13	and	and	CC
10_1101-2021_02_08_430070	66	14	relative	relative	JJ
10_1101-2021_02_08_430070	66	15	position	position	NN
10_1101-2021_02_08_430070	66	16	representation	representation	NN
10_1101-2021_02_08_430070	66	17	.	.	.
10_1101-2021_02_08_430070	67	1	The	the	DT
10_1101-2021_02_08_430070	67	2	learnable	learnable	JJ
10_1101-2021_02_08_430070	67	3	PE	PE	NNP
10_1101-2021_02_08_430070	67	4	takes	take	VBZ
10_1101-2021_02_08_430070	67	5	positional	positional	JJ
10_1101-2021_02_08_430070	67	6	embedding	embed	VBG
10_1101-2021_02_08_430070	67	7	vectors	vector	NNS
10_1101-2021_02_08_430070	67	8	as	as	IN
10_1101-2021_02_08_430070	67	9	parameters	parameter	NNS
10_1101-2021_02_08_430070	67	10	,	,	,
10_1101-2021_02_08_430070	67	11	which	which	WDT
10_1101-2021_02_08_430070	67	12	are	be	VBP
10_1101-2021_02_08_430070	67	13	updated	update	VBN
10_1101-2021_02_08_430070	67	14	during	during	IN
10_1101-2021_02_08_430070	67	15	the	the	DT
10_1101-2021_02_08_430070	67	16	learning	learning	NN
10_1101-2021_02_08_430070	67	17	process	process	NN
10_1101-2021_02_08_430070	67	18	.	.	.
10_1101-2021_02_08_430070	68	1	.license	.license	NNP
10_1101-2021_02_08_430070	68	2	CC	CC	NNP
10_1101-2021_02_08_430070	68	3	-	-	HYPH
10_1101-2021_02_08_430070	68	4	BY	BY	NNP
10_1101-2021_02_08_430070	68	5	-	-	HYPH
10_1101-2021_02_08_430070	68	6	NC	NC	NNP
10_1101-2021_02_08_430070	68	7	-	-	HYPH
10_1101-2021_02_08_430070	68	8	ND	ND	NNP
10_1101-2021_02_08_430070	68	9	4.0	4.0	CD
10_1101-2021_02_08_430070	68	10	Internationalpeer	Internationalpeer	NNP
10_1101-2021_02_08_430070	68	11	review	review	NN
10_1101-2021_02_08_430070	68	12	)	)	-RRB-
10_1101-2021_02_08_430070	68	13	is	be	VBZ
10_1101-2021_02_08_430070	68	14	the	the	DT
10_1101-2021_02_08_430070	68	15	author	author	NN
10_1101-2021_02_08_430070	68	16	/	/	SYM
10_1101-2021_02_08_430070	68	17	funder	funder	NN
10_1101-2021_02_08_430070	68	18	,	,	,
10_1101-2021_02_08_430070	68	19	who	who	WP
10_1101-2021_02_08_430070	68	20	has	have	VBZ
10_1101-2021_02_08_430070	68	21	granted	grant	VBN
10_1101-2021_02_08_430070	68	22	bioRxiv	biorxiv	IN
10_1101-2021_02_08_430070	68	23	a	a	DT
10_1101-2021_02_08_430070	68	24	license	license	NN
10_1101-2021_02_08_430070	68	25	to	to	TO
10_1101-2021_02_08_430070	68	26	display	display	VB
10_1101-2021_02_08_430070	68	27	the	the	DT
10_1101-2021_02_08_430070	68	28	preprint	preprint	NN
10_1101-2021_02_08_430070	68	29	in	in	IN
10_1101-2021_02_08_430070	68	30	perpetuity	perpetuity	NN
10_1101-2021_02_08_430070	68	31	.	.	.
10_1101-2021_02_08_430070	69	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	69	2	is	be	VBZ
10_1101-2021_02_08_430070	69	3	made	make	VBN
10_1101-2021_02_08_430070	69	4	available	available	JJ
10_1101-2021_02_08_430070	69	5	under	under	IN
10_1101-2021_02_08_430070	69	6	a	a	DT
10_1101-2021_02_08_430070	69	7	The	the	DT
10_1101-2021_02_08_430070	69	8	copyright	copyright	NN
10_1101-2021_02_08_430070	69	9	holder	holder	NN
10_1101-2021_02_08_430070	69	10	for	for	IN
10_1101-2021_02_08_430070	69	11	this	this	DT
10_1101-2021_02_08_430070	69	12	preprint	preprint	NN
10_1101-2021_02_08_430070	69	13	(	(	-LRB-
10_1101-2021_02_08_430070	69	14	which	which	WDT
10_1101-2021_02_08_430070	69	15	was	be	VBD
10_1101-2021_02_08_430070	69	16	not	not	RB
10_1101-2021_02_08_430070	69	17	certified	certify	VBN
10_1101-2021_02_08_430070	69	18	bythis	bythis	DT
10_1101-2021_02_08_430070	69	19	version	version	NN
10_1101-2021_02_08_430070	69	20	posted	post	VBD
10_1101-2021_02_08_430070	69	21	February	February	NNP
10_1101-2021_02_08_430070	69	22	10	10	CD
10_1101-2021_02_08_430070	69	23	,	,	,
10_1101-2021_02_08_430070	69	24	2021	2021	CD
10_1101-2021_02_08_430070	69	25	.	.	.
10_1101-2021_02_08_430070	69	26	;	;	:
10_1101-2021_02_08_430070	69	27	https://doi.org/10.1101/2021.02.08.430070doi	https://doi.org/10.1101/2021.02.08.430070doi	NFP
10_1101-2021_02_08_430070	69	28	:	:	:
10_1101-2021_02_08_430070	69	29	bioRxiv	biorxiv	VB
10_1101-2021_02_08_430070	69	30	preprint	preprint	NN
10_1101-2021_02_08_430070	69	31	https://doi.org/10.1101/2021.02.08.430070	https://doi.org/10.1101/2021.02.08.430070	UH
10_1101-2021_02_08_430070	69	32	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	69	33	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	69	34	✐	✐	NNP
10_1101-2021_02_08_430070	69	35	✐	✐	NNP
10_1101-2021_02_08_430070	69	36	✐	✐	NNP
10_1101-2021_02_08_430070	69	37	✐	✐	NNP
10_1101-2021_02_08_430070	69	38	✐	✐	NNP
10_1101-2021_02_08_430070	69	39	✐	✐	NNP
10_1101-2021_02_08_430070	69	40	✐	✐	NNP
10_1101-2021_02_08_430070	69	41	✐	✐	NNP
10_1101-2021_02_08_430070	69	42	BERT	BERT	NNP
10_1101-2021_02_08_430070	69	43	for	for	IN
10_1101-2021_02_08_430070	69	44	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	69	45	methylation	methylation	NN
10_1101-2021_02_08_430070	69	46	detection	detection	NN
10_1101-2021_02_08_430070	69	47	3	3	CD
10_1101-2021_02_08_430070	69	48	2.1.2	2.1.2	CD
10_1101-2021_02_08_430070	69	49	Self	self	NN
10_1101-2021_02_08_430070	69	50	-	-	HYPH
10_1101-2021_02_08_430070	69	51	attention	attention	NN
10_1101-2021_02_08_430070	69	52	module	module	NN
10_1101-2021_02_08_430070	69	53	Following	follow	VBG
10_1101-2021_02_08_430070	69	54	the	the	DT
10_1101-2021_02_08_430070	69	55	embedding	embedding	NN
10_1101-2021_02_08_430070	69	56	layer	layer	NN
10_1101-2021_02_08_430070	69	57	,	,	,
10_1101-2021_02_08_430070	69	58	there	there	EX
10_1101-2021_02_08_430070	69	59	are	be	VBP
10_1101-2021_02_08_430070	69	60	three	three	CD
10_1101-2021_02_08_430070	69	61	stacked	stack	VBN
10_1101-2021_02_08_430070	69	62	transformer	transformer	NN
10_1101-2021_02_08_430070	69	63	blocks	block	NNS
10_1101-2021_02_08_430070	69	64	.	.	.
10_1101-2021_02_08_430070	70	1	Each	each	DT
10_1101-2021_02_08_430070	70	2	transformer	transformer	NN
10_1101-2021_02_08_430070	70	3	block	block	NN
10_1101-2021_02_08_430070	70	4	consists	consist	VBZ
10_1101-2021_02_08_430070	70	5	of	of	IN
10_1101-2021_02_08_430070	70	6	a	a	DT
10_1101-2021_02_08_430070	70	7	multi	multi	JJ
10_1101-2021_02_08_430070	70	8	-	-	JJ
10_1101-2021_02_08_430070	70	9	head	head	JJ
10_1101-2021_02_08_430070	70	10	self	self	NN
10_1101-2021_02_08_430070	70	11	-	-	HYPH
10_1101-2021_02_08_430070	70	12	attention	attention	NN
10_1101-2021_02_08_430070	70	13	layer	layer	NN
10_1101-2021_02_08_430070	70	14	and	and	CC
10_1101-2021_02_08_430070	70	15	position	position	NN
10_1101-2021_02_08_430070	70	16	-	-	HYPH
10_1101-2021_02_08_430070	70	17	wise	wise	JJ
10_1101-2021_02_08_430070	70	18	fully	fully	RB
10_1101-2021_02_08_430070	70	19	connected	connect	VBN
10_1101-2021_02_08_430070	70	20	feed	feed	NN
10_1101-2021_02_08_430070	70	21	-	-	HYPH
10_1101-2021_02_08_430070	70	22	forward	forward	NN
10_1101-2021_02_08_430070	70	23	network	network	NN
10_1101-2021_02_08_430070	70	24	.	.	.
10_1101-2021_02_08_430070	71	1	The	the	DT
10_1101-2021_02_08_430070	71	2	self	self	NN
10_1101-2021_02_08_430070	71	3	-	-	HYPH
10_1101-2021_02_08_430070	71	4	attention	attention	NN
10_1101-2021_02_08_430070	71	5	mechanism	mechanism	NN
10_1101-2021_02_08_430070	71	6	is	be	VBZ
10_1101-2021_02_08_430070	71	7	a	a	DT
10_1101-2021_02_08_430070	71	8	modeling	modeling	JJ
10_1101-2021_02_08_430070	71	9	approach	approach	NN
10_1101-2021_02_08_430070	71	10	of	of	IN
10_1101-2021_02_08_430070	71	11	describing	describe	VBG
10_1101-2021_02_08_430070	71	12	context	context	NN
10_1101-2021_02_08_430070	71	13	information	information	NN
10_1101-2021_02_08_430070	71	14	for	for	IN
10_1101-2021_02_08_430070	71	15	different	different	JJ
10_1101-2021_02_08_430070	71	16	positions	position	NNS
10_1101-2021_02_08_430070	71	17	of	of	IN
10_1101-2021_02_08_430070	71	18	inputs	input	NNS
10_1101-2021_02_08_430070	71	19	under	under	IN
10_1101-2021_02_08_430070	71	20	a	a	DT
10_1101-2021_02_08_430070	71	21	deep	deep	JJ
10_1101-2021_02_08_430070	71	22	learning	learning	NN
10_1101-2021_02_08_430070	71	23	framework	framework	NN
10_1101-2021_02_08_430070	71	24	.	.	.
10_1101-2021_02_08_430070	72	1	The	the	DT
10_1101-2021_02_08_430070	72	2	self-	self-	NN
10_1101-2021_02_08_430070	72	3	attention	attention	NN
10_1101-2021_02_08_430070	72	4	mechanism	mechanism	NN
10_1101-2021_02_08_430070	72	5	imitates	imitate	VBZ
10_1101-2021_02_08_430070	72	6	the	the	DT
10_1101-2021_02_08_430070	72	7	human	human	JJ
10_1101-2021_02_08_430070	72	8	sight	sight	NN
10_1101-2021_02_08_430070	72	9	mechanism	mechanism	NN
10_1101-2021_02_08_430070	72	10	and	and	CC
10_1101-2021_02_08_430070	72	11	provides	provide	VBZ
10_1101-2021_02_08_430070	72	12	a	a	DT
10_1101-2021_02_08_430070	72	13	model	model	NN
10_1101-2021_02_08_430070	72	14	with	with	IN
10_1101-2021_02_08_430070	72	15	the	the	DT
10_1101-2021_02_08_430070	72	16	ability	ability	NN
10_1101-2021_02_08_430070	72	17	to	to	TO
10_1101-2021_02_08_430070	72	18	zoom	zoom	VB
10_1101-2021_02_08_430070	72	19	in	in	RP
10_1101-2021_02_08_430070	72	20	or	or	CC
10_1101-2021_02_08_430070	72	21	out	out	RB
10_1101-2021_02_08_430070	72	22	in	in	IN
10_1101-2021_02_08_430070	72	23	a	a	DT
10_1101-2021_02_08_430070	72	24	particular	particular	JJ
10_1101-2021_02_08_430070	72	25	position	position	NN
10_1101-2021_02_08_430070	72	26	of	of	IN
10_1101-2021_02_08_430070	72	27	an	an	DT
10_1101-2021_02_08_430070	72	28	input	input	NN
10_1101-2021_02_08_430070	72	29	sequence	sequence	NN
10_1101-2021_02_08_430070	72	30	.	.	.
10_1101-2021_02_08_430070	73	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	73	2	demonstrates	demonstrate	VBZ
10_1101-2021_02_08_430070	73	3	the	the	DT
10_1101-2021_02_08_430070	73	4	effectiveness	effectiveness	NN
10_1101-2021_02_08_430070	73	5	in	in	IN
10_1101-2021_02_08_430070	73	6	many	many	JJ
10_1101-2021_02_08_430070	73	7	different	different	JJ
10_1101-2021_02_08_430070	73	8	tasks	task	NNS
10_1101-2021_02_08_430070	73	9	including	include	VBG
10_1101-2021_02_08_430070	73	10	natural	natural	JJ
10_1101-2021_02_08_430070	73	11	language	language	NN
10_1101-2021_02_08_430070	73	12	understanding	understanding	NN
10_1101-2021_02_08_430070	73	13	,	,	,
10_1101-2021_02_08_430070	73	14	image	image	NN
10_1101-2021_02_08_430070	73	15	recognition	recognition	NN
10_1101-2021_02_08_430070	73	16	,	,	,
10_1101-2021_02_08_430070	73	17	and	and	CC
10_1101-2021_02_08_430070	73	18	several	several	JJ
10_1101-2021_02_08_430070	73	19	bioinformatics	bioinformatic	NNS
10_1101-2021_02_08_430070	73	20	applications	application	NNS
10_1101-2021_02_08_430070	73	21	.	.	.
10_1101-2021_02_08_430070	74	1	Attention	attention	NN
10_1101-2021_02_08_430070	74	2	function	function	NN
10_1101-2021_02_08_430070	74	3	is	be	VBZ
10_1101-2021_02_08_430070	74	4	described	describe	VBN
10_1101-2021_02_08_430070	74	5	as	as	IN
10_1101-2021_02_08_430070	74	6	mapping	mapping	NN
10_1101-2021_02_08_430070	74	7	Q	Q	NNP
10_1101-2021_02_08_430070	74	8	and	and	CC
10_1101-2021_02_08_430070	74	9	a	a	DT
10_1101-2021_02_08_430070	74	10	set	set	NN
10_1101-2021_02_08_430070	74	11	of	of	IN
10_1101-2021_02_08_430070	74	12	key	key	JJ
10_1101-2021_02_08_430070	74	13	-	-	HYPH
10_1101-2021_02_08_430070	74	14	value	value	NN
10_1101-2021_02_08_430070	74	15	(	(	-LRB-
10_1101-2021_02_08_430070	74	16	K	K	NNP
10_1101-2021_02_08_430070	74	17	,	,	,
10_1101-2021_02_08_430070	74	18	V	V	NNP
10_1101-2021_02_08_430070	74	19	)	)	-RRB-
10_1101-2021_02_08_430070	74	20	pairs	pair	NNS
10_1101-2021_02_08_430070	74	21	to	to	IN
10_1101-2021_02_08_430070	74	22	an	an	DT
10_1101-2021_02_08_430070	74	23	output	output	NN
10_1101-2021_02_08_430070	74	24	.	.	.
10_1101-2021_02_08_430070	75	1	Formally	formally	RB
10_1101-2021_02_08_430070	75	2	,	,	,
10_1101-2021_02_08_430070	75	3	for	for	IN
10_1101-2021_02_08_430070	75	4	an	an	DT
10_1101-2021_02_08_430070	75	5	input	input	NN
10_1101-2021_02_08_430070	75	6	x	x	SYM
10_1101-2021_02_08_430070	75	7	=	=	NFP
10_1101-2021_02_08_430070	75	8	(	(	-LRB-
10_1101-2021_02_08_430070	75	9	x1	x1	NNP
10_1101-2021_02_08_430070	75	10	,	,	,
10_1101-2021_02_08_430070	75	11	...	...	:
10_1101-2021_02_08_430070	75	12	,	,	,
10_1101-2021_02_08_430070	75	13	xn	xn	NNP
10_1101-2021_02_08_430070	75	14	)	)	-RRB-
10_1101-2021_02_08_430070	75	15	of	of	IN
10_1101-2021_02_08_430070	75	16	n	n	DT
10_1101-2021_02_08_430070	75	17	elements	element	NNS
10_1101-2021_02_08_430070	75	18	where	where	WRB
10_1101-2021_02_08_430070	75	19	xi	xi	NNP
10_1101-2021_02_08_430070	75	20	∈	∈	NNP
10_1101-2021_02_08_430070	75	21	Rdx	Rdx	NNP
10_1101-2021_02_08_430070	75	22	,	,	,
10_1101-2021_02_08_430070	75	23	we	-PRON-	PRP
10_1101-2021_02_08_430070	75	24	calculate	calculate	VBP
10_1101-2021_02_08_430070	75	25	query	query	NN
10_1101-2021_02_08_430070	75	26	Q	q	NN
10_1101-2021_02_08_430070	75	27	,	,	,
10_1101-2021_02_08_430070	75	28	key	key	JJ
10_1101-2021_02_08_430070	75	29	K	k	NN
10_1101-2021_02_08_430070	75	30	and	and	CC
10_1101-2021_02_08_430070	75	31	value	value	NN
10_1101-2021_02_08_430070	75	32	V	V	NNP
10_1101-2021_02_08_430070	75	33	vectors	vector	NNS
10_1101-2021_02_08_430070	75	34	of	of	IN
10_1101-2021_02_08_430070	75	35	dimension	dimension	NN
10_1101-2021_02_08_430070	75	36	dk	dk	NN
10_1101-2021_02_08_430070	75	37	based	base	VBN
10_1101-2021_02_08_430070	75	38	on	on	IN
10_1101-2021_02_08_430070	75	39	the	the	DT
10_1101-2021_02_08_430070	75	40	embedding	embed	VBG
10_1101-2021_02_08_430070	75	41	vector	vector	NN
10_1101-2021_02_08_430070	75	42	of	of	IN
10_1101-2021_02_08_430070	75	43	embed(x	embed(x	NNP
10_1101-2021_02_08_430070	75	44	)	)	-RRB-
10_1101-2021_02_08_430070	75	45	.	.	.
10_1101-2021_02_08_430070	76	1	The	the	DT
10_1101-2021_02_08_430070	76	2	attention	attention	NN
10_1101-2021_02_08_430070	76	3	module	module	NN
10_1101-2021_02_08_430070	76	4	generates	generate	VBZ
10_1101-2021_02_08_430070	76	5	a	a	DT
10_1101-2021_02_08_430070	76	6	new	new	JJ
10_1101-2021_02_08_430070	76	7	sequence	sequence	NN
10_1101-2021_02_08_430070	76	8	z	z	NN
10_1101-2021_02_08_430070	76	9	=	=	NFP
10_1101-2021_02_08_430070	76	10	(	(	-LRB-
10_1101-2021_02_08_430070	76	11	z1	z1	NNP
10_1101-2021_02_08_430070	76	12	,	,	,
10_1101-2021_02_08_430070	76	13	...	...	:
10_1101-2021_02_08_430070	76	14	,	,	,
10_1101-2021_02_08_430070	76	15	zn	zn	NNP
10_1101-2021_02_08_430070	76	16	)	)	-RRB-
10_1101-2021_02_08_430070	76	17	of	of	IN
10_1101-2021_02_08_430070	76	18	the	the	DT
10_1101-2021_02_08_430070	76	19	same	same	JJ
10_1101-2021_02_08_430070	76	20	length	length	NN
10_1101-2021_02_08_430070	76	21	as	as	IN
10_1101-2021_02_08_430070	76	22	x.	x.	NNP
10_1101-2021_02_08_430070	76	23	zi	zi	NNP
10_1101-2021_02_08_430070	76	24	is	be	VBZ
10_1101-2021_02_08_430070	76	25	calculated	calculate	VBN
10_1101-2021_02_08_430070	76	26	as	as	IN
10_1101-2021_02_08_430070	76	27	a	a	DT
10_1101-2021_02_08_430070	76	28	weighted	weighted	JJ
10_1101-2021_02_08_430070	76	29	sum	sum	NN
10_1101-2021_02_08_430070	76	30	of	of	IN
10_1101-2021_02_08_430070	76	31	linearly	linearly	JJ
10_1101-2021_02_08_430070	76	32	transformed	transform	VBN
10_1101-2021_02_08_430070	76	33	input	input	NN
10_1101-2021_02_08_430070	76	34	elements	element	NNS
10_1101-2021_02_08_430070	76	35	as	as	IN
10_1101-2021_02_08_430070	76	36	follows	follow	VBZ
10_1101-2021_02_08_430070	76	37	:	:	:
10_1101-2021_02_08_430070	76	38	zi	zi	NN
10_1101-2021_02_08_430070	76	39	=	=	SYM
10_1101-2021_02_08_430070	76	40	n∑	n∑	VB
10_1101-2021_02_08_430070	76	41	j=1	j=1	NNS
10_1101-2021_02_08_430070	76	42	aij(xjW	aij(xjW	NNP
10_1101-2021_02_08_430070	76	43	V	v	NN
10_1101-2021_02_08_430070	76	44	)	)	-RRB-
10_1101-2021_02_08_430070	76	45	aij	aij	NN
10_1101-2021_02_08_430070	76	46	=	=	SYM
10_1101-2021_02_08_430070	76	47	exp	exp	NN
10_1101-2021_02_08_430070	76	48	eij∑n	eij∑n	NNP
10_1101-2021_02_08_430070	76	49	k=1	k=1	JJ
10_1101-2021_02_08_430070	76	50	exp	exp	NN
10_1101-2021_02_08_430070	76	51	eik	eik	NNP
10_1101-2021_02_08_430070	76	52	eij	eij	NNP
10_1101-2021_02_08_430070	76	53	=	=	NFP
10_1101-2021_02_08_430070	76	54	(	(	-LRB-
10_1101-2021_02_08_430070	76	55	xiW	xiW	NNP
10_1101-2021_02_08_430070	76	56	Q)(xjW	q)(xjw	NN
10_1101-2021_02_08_430070	76	57	T	t	NN
10_1101-2021_02_08_430070	76	58	)	)	-RRB-
10_1101-2021_02_08_430070	76	59	T	T	NNP
10_1101-2021_02_08_430070	76	60	√	√	NNP
10_1101-2021_02_08_430070	76	61	dz	dz	NN
10_1101-2021_02_08_430070	76	62	,	,	,
10_1101-2021_02_08_430070	76	63	where	where	WRB
10_1101-2021_02_08_430070	76	64	W	W	NNP
10_1101-2021_02_08_430070	76	65	Q	Q	NNP
10_1101-2021_02_08_430070	76	66	,	,	,
10_1101-2021_02_08_430070	76	67	W	W	NNP
10_1101-2021_02_08_430070	76	68	K	k	NN
10_1101-2021_02_08_430070	76	69	,	,	,
10_1101-2021_02_08_430070	76	70	W	w	NN
10_1101-2021_02_08_430070	76	71	T	t	NN
10_1101-2021_02_08_430070	76	72	∈	∈	JJ
10_1101-2021_02_08_430070	76	73	Rdx×dz	rdx×dz	NN
10_1101-2021_02_08_430070	76	74	are	be	VBP
10_1101-2021_02_08_430070	76	75	parameter	parameter	NN
10_1101-2021_02_08_430070	76	76	matrices	matrix	NNS
10_1101-2021_02_08_430070	76	77	.	.	.
10_1101-2021_02_08_430070	77	1	The	the	DT
10_1101-2021_02_08_430070	77	2	self	self	NN
10_1101-2021_02_08_430070	77	3	-	-	HYPH
10_1101-2021_02_08_430070	77	4	attention	attention	NN
10_1101-2021_02_08_430070	77	5	computes	compute	NNS
10_1101-2021_02_08_430070	77	6	a	a	DT
10_1101-2021_02_08_430070	77	7	pairwise	pairwise	NN
10_1101-2021_02_08_430070	77	8	correlation	correlation	NN
10_1101-2021_02_08_430070	77	9	of	of	IN
10_1101-2021_02_08_430070	77	10	embed(xi	embed(xi	ADD
10_1101-2021_02_08_430070	77	11	)	)	-RRB-
10_1101-2021_02_08_430070	77	12	and	and	CC
10_1101-2021_02_08_430070	77	13	embed(xj	embed(xj	ADD
10_1101-2021_02_08_430070	77	14	)	)	-RRB-
10_1101-2021_02_08_430070	77	15	,	,	,
10_1101-2021_02_08_430070	77	16	which	which	WDT
10_1101-2021_02_08_430070	77	17	can	can	MD
10_1101-2021_02_08_430070	77	18	be	be	VB
10_1101-2021_02_08_430070	77	19	calculated	calculate	VBN
10_1101-2021_02_08_430070	77	20	in	in	IN
10_1101-2021_02_08_430070	77	21	a	a	DT
10_1101-2021_02_08_430070	77	22	parallel	parallel	JJ
10_1101-2021_02_08_430070	77	23	way	way	NN
10_1101-2021_02_08_430070	77	24	.	.	.
10_1101-2021_02_08_430070	78	1	While	while	IN
10_1101-2021_02_08_430070	78	2	in	in	IN
10_1101-2021_02_08_430070	78	3	a	a	DT
10_1101-2021_02_08_430070	78	4	biRNN	birnn	JJ
10_1101-2021_02_08_430070	78	5	,	,	,
10_1101-2021_02_08_430070	78	6	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	78	7	hidden	hide	VBN
10_1101-2021_02_08_430070	78	8	units	unit	NNS
10_1101-2021_02_08_430070	78	9	are	be	VBP
10_1101-2021_02_08_430070	78	10	required	require	VBN
10_1101-2021_02_08_430070	78	11	to	to	TO
10_1101-2021_02_08_430070	78	12	be	be	VB
10_1101-2021_02_08_430070	78	13	calculated	calculate	VBN
10_1101-2021_02_08_430070	78	14	successively	successively	RB
10_1101-2021_02_08_430070	78	15	.	.	.
10_1101-2021_02_08_430070	79	1	This	this	DT
10_1101-2021_02_08_430070	79	2	architecture	architecture	NN
10_1101-2021_02_08_430070	79	3	difference	difference	NN
10_1101-2021_02_08_430070	79	4	makes	make	VBZ
10_1101-2021_02_08_430070	79	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	79	6	can	can	MD
10_1101-2021_02_08_430070	79	7	be	be	VB
10_1101-2021_02_08_430070	79	8	optimized	optimize	VBN
10_1101-2021_02_08_430070	79	9	for	for	IN
10_1101-2021_02_08_430070	79	10	fast	fast	JJ
10_1101-2021_02_08_430070	79	11	inference	inference	NN
10_1101-2021_02_08_430070	79	12	.	.	.
10_1101-2021_02_08_430070	80	1	2.1.3	2.1.3	CD
10_1101-2021_02_08_430070	80	2	Relative	relative	JJ
10_1101-2021_02_08_430070	80	3	position	position	NN
10_1101-2021_02_08_430070	80	4	representation	representation	NN
10_1101-2021_02_08_430070	80	5	in	in	IN
10_1101-2021_02_08_430070	80	6	self	self	NN
10_1101-2021_02_08_430070	80	7	-	-	HYPH
10_1101-2021_02_08_430070	80	8	attention	attention	NN
10_1101-2021_02_08_430070	80	9	heads	head	NNS
10_1101-2021_02_08_430070	80	10	For	for	IN
10_1101-2021_02_08_430070	80	11	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	80	12	sequencing	sequencing	NN
10_1101-2021_02_08_430070	80	13	,	,	,
10_1101-2021_02_08_430070	80	14	signals	signal	NNS
10_1101-2021_02_08_430070	80	15	are	be	VBP
10_1101-2021_02_08_430070	80	16	supposed	suppose	VBN
10_1101-2021_02_08_430070	80	17	to	to	TO
10_1101-2021_02_08_430070	80	18	be	be	VB
10_1101-2021_02_08_430070	80	19	more	more	RBR
10_1101-2021_02_08_430070	80	20	affected	affected	JJ
10_1101-2021_02_08_430070	80	21	by	by	IN
10_1101-2021_02_08_430070	80	22	the	the	DT
10_1101-2021_02_08_430070	80	23	nucleotide	nucleotide	JJ
10_1101-2021_02_08_430070	80	24	passing	pass	VBG
10_1101-2021_02_08_430070	80	25	through	through	IN
10_1101-2021_02_08_430070	80	26	the	the	DT
10_1101-2021_02_08_430070	80	27	pore	pore	NN
10_1101-2021_02_08_430070	80	28	.	.	.
10_1101-2021_02_08_430070	81	1	Its	-PRON-	PRP$
10_1101-2021_02_08_430070	81	2	surrounding	surround	VBG
10_1101-2021_02_08_430070	81	3	nucleotides	nucleotide	NNS
10_1101-2021_02_08_430070	81	4	may	may	MD
10_1101-2021_02_08_430070	81	5	also	also	RB
10_1101-2021_02_08_430070	81	6	have	have	VB
10_1101-2021_02_08_430070	81	7	effects	effect	NNS
10_1101-2021_02_08_430070	81	8	on	on	IN
10_1101-2021_02_08_430070	81	9	the	the	DT
10_1101-2021_02_08_430070	81	10	current	current	JJ
10_1101-2021_02_08_430070	81	11	signals	signal	NNS
10_1101-2021_02_08_430070	81	12	.	.	.
10_1101-2021_02_08_430070	82	1	For	for	IN
10_1101-2021_02_08_430070	82	2	those	those	DT
10_1101-2021_02_08_430070	82	3	nucleotides	nucleotide	NNS
10_1101-2021_02_08_430070	82	4	that	that	WDT
10_1101-2021_02_08_430070	82	5	are	be	VBP
10_1101-2021_02_08_430070	82	6	too	too	RB
10_1101-2021_02_08_430070	82	7	far	far	RB
10_1101-2021_02_08_430070	82	8	away	away	RB
10_1101-2021_02_08_430070	82	9	in	in	IN
10_1101-2021_02_08_430070	82	10	a	a	DT
10_1101-2021_02_08_430070	82	11	context	context	NN
10_1101-2021_02_08_430070	82	12	window	window	NN
10_1101-2021_02_08_430070	82	13	,	,	,
10_1101-2021_02_08_430070	82	14	it	-PRON-	PRP
10_1101-2021_02_08_430070	82	15	is	be	VBZ
10_1101-2021_02_08_430070	82	16	intuitive	intuitive	JJ
10_1101-2021_02_08_430070	82	17	to	to	TO
10_1101-2021_02_08_430070	82	18	assume	assume	VB
10_1101-2021_02_08_430070	82	19	they	-PRON-	PRP
10_1101-2021_02_08_430070	82	20	have	have	VBP
10_1101-2021_02_08_430070	82	21	less	less	JJR
10_1101-2021_02_08_430070	82	22	effect	effect	NN
10_1101-2021_02_08_430070	82	23	on	on	IN
10_1101-2021_02_08_430070	82	24	the	the	DT
10_1101-2021_02_08_430070	82	25	detected	detected	JJ
10_1101-2021_02_08_430070	82	26	current	current	JJ
10_1101-2021_02_08_430070	82	27	signals	signal	NNS
10_1101-2021_02_08_430070	82	28	.	.	.
10_1101-2021_02_08_430070	83	1	In	in	IN
10_1101-2021_02_08_430070	83	2	the	the	DT
10_1101-2021_02_08_430070	83	3	refined	refined	JJ
10_1101-2021_02_08_430070	83	4	BERT	BERT	NNP
10_1101-2021_02_08_430070	83	5	model	model	NN
10_1101-2021_02_08_430070	83	6	,	,	,
10_1101-2021_02_08_430070	83	7	we	-PRON-	PRP
10_1101-2021_02_08_430070	83	8	add	add	VBP
10_1101-2021_02_08_430070	83	9	relative	relative	JJ
10_1101-2021_02_08_430070	83	10	position	position	NN
10_1101-2021_02_08_430070	83	11	representation	representation	NN
10_1101-2021_02_08_430070	83	12	in	in	IN
10_1101-2021_02_08_430070	83	13	the	the	DT
10_1101-2021_02_08_430070	83	14	attention	attention	NN
10_1101-2021_02_08_430070	83	15	module	module	NN
10_1101-2021_02_08_430070	83	16	following	follow	VBG
10_1101-2021_02_08_430070	83	17	the	the	DT
10_1101-2021_02_08_430070	83	18	method	method	NN
10_1101-2021_02_08_430070	83	19	proposed	propose	VBN
10_1101-2021_02_08_430070	83	20	by	by	IN
10_1101-2021_02_08_430070	83	21	Shaw	Shaw	NNP
10_1101-2021_02_08_430070	83	22	et	et	NNP
10_1101-2021_02_08_430070	83	23	al	al	NNP
10_1101-2021_02_08_430070	83	24	.	.	.
10_1101-2021_02_08_430070	84	1	(	(	-LRB-
10_1101-2021_02_08_430070	84	2	2018	2018	CD
10_1101-2021_02_08_430070	84	3	)	)	-RRB-
10_1101-2021_02_08_430070	84	4	.	.	.
10_1101-2021_02_08_430070	85	1	For	for	IN
10_1101-2021_02_08_430070	85	2	any	any	DT
10_1101-2021_02_08_430070	85	3	two	two	CD
10_1101-2021_02_08_430070	85	4	input	input	NN
10_1101-2021_02_08_430070	85	5	elements	element	NNS
10_1101-2021_02_08_430070	85	6	xi	xi	NNP
10_1101-2021_02_08_430070	85	7	and	and	CC
10_1101-2021_02_08_430070	85	8	xj	xj	NNP
10_1101-2021_02_08_430070	85	9	,	,	,
10_1101-2021_02_08_430070	85	10	the	the	DT
10_1101-2021_02_08_430070	85	11	relative	relative	JJ
10_1101-2021_02_08_430070	85	12	position	position	NN
10_1101-2021_02_08_430070	85	13	information	information	NN
10_1101-2021_02_08_430070	85	14	is	be	VBZ
10_1101-2021_02_08_430070	85	15	modeled	model	VBN
10_1101-2021_02_08_430070	85	16	with	with	IN
10_1101-2021_02_08_430070	85	17	two	two	CD
10_1101-2021_02_08_430070	85	18	distinct	distinct	JJ
10_1101-2021_02_08_430070	85	19	edge	edge	NN
10_1101-2021_02_08_430070	85	20	representations	representation	NNS
10_1101-2021_02_08_430070	85	21	aVij	aVij	NNP
10_1101-2021_02_08_430070	85	22	,	,	,
10_1101-2021_02_08_430070	85	23	a	a	DT
10_1101-2021_02_08_430070	85	24	K	K	NNP
10_1101-2021_02_08_430070	85	25	ij	ij	NN
10_1101-2021_02_08_430070	85	26	.	.	.
10_1101-2021_02_08_430070	86	1	For	for	IN
10_1101-2021_02_08_430070	86	2	linear	linear	JJ
10_1101-2021_02_08_430070	86	3	sequences	sequence	NNS
10_1101-2021_02_08_430070	86	4	,	,	,
10_1101-2021_02_08_430070	86	5	those	those	DT
10_1101-2021_02_08_430070	86	6	edges	edge	NNS
10_1101-2021_02_08_430070	86	7	are	be	VBP
10_1101-2021_02_08_430070	86	8	used	use	VBN
10_1101-2021_02_08_430070	86	9	to	to	TO
10_1101-2021_02_08_430070	86	10	capture	capture	VB
10_1101-2021_02_08_430070	86	11	the	the	DT
10_1101-2021_02_08_430070	86	12	relative	relative	JJ
10_1101-2021_02_08_430070	86	13	position	position	NN
10_1101-2021_02_08_430070	86	14	differences	difference	NNS
10_1101-2021_02_08_430070	86	15	between	between	IN
10_1101-2021_02_08_430070	86	16	input	input	NN
10_1101-2021_02_08_430070	86	17	elements	element	NNS
10_1101-2021_02_08_430070	86	18	.	.	.
10_1101-2021_02_08_430070	87	1	As	as	IN
10_1101-2021_02_08_430070	87	2	the	the	DT
10_1101-2021_02_08_430070	87	3	precise	precise	JJ
10_1101-2021_02_08_430070	87	4	relative	relative	JJ
10_1101-2021_02_08_430070	87	5	position	position	NN
10_1101-2021_02_08_430070	87	6	is	be	VBZ
10_1101-2021_02_08_430070	87	7	not	not	RB
10_1101-2021_02_08_430070	87	8	useful	useful	JJ
10_1101-2021_02_08_430070	87	9	beyond	beyond	IN
10_1101-2021_02_08_430070	87	10	a	a	DT
10_1101-2021_02_08_430070	87	11	certain	certain	JJ
10_1101-2021_02_08_430070	87	12	distance	distance	NN
10_1101-2021_02_08_430070	87	13	,	,	,
10_1101-2021_02_08_430070	87	14	we	-PRON-	PRP
10_1101-2021_02_08_430070	87	15	clip	clip	VBP
10_1101-2021_02_08_430070	87	16	the	the	DT
10_1101-2021_02_08_430070	87	17	maximum	maximum	JJ
10_1101-2021_02_08_430070	87	18	distance	distance	NN
10_1101-2021_02_08_430070	87	19	(	(	-LRB-
10_1101-2021_02_08_430070	87	20	e.g.	e.g.	RB
10_1101-2021_02_08_430070	88	1	±3bp	±3bp	NNP
10_1101-2021_02_08_430070	88	2	)	)	-RRB-
10_1101-2021_02_08_430070	88	3	in	in	IN
10_1101-2021_02_08_430070	88	4	calculating	calculate	VBG
10_1101-2021_02_08_430070	88	5	attention	attention	NN
10_1101-2021_02_08_430070	88	6	aij	aij	NN
10_1101-2021_02_08_430070	88	7	∈	∈	NNP
10_1101-2021_02_08_430070	88	8	A.	a.	NN
10_1101-2021_02_08_430070	88	9	a	a	DT
10_1101-2021_02_08_430070	88	10	K	k	NN
10_1101-2021_02_08_430070	88	11	ij	ij	NN
10_1101-2021_02_08_430070	88	12	=	=	NFP
10_1101-2021_02_08_430070	88	13	W	W	NNP
10_1101-2021_02_08_430070	88	14	K	K	NNP
10_1101-2021_02_08_430070	88	15	clip(j−i	clip(j−i	NNP
10_1101-2021_02_08_430070	88	16	,	,	,
10_1101-2021_02_08_430070	88	17	k	k	LS
10_1101-2021_02_08_430070	88	18	)	)	-RRB-
10_1101-2021_02_08_430070	88	19	a	a	DT
10_1101-2021_02_08_430070	88	20	V	v	NN
10_1101-2021_02_08_430070	88	21	ij	ij	NN
10_1101-2021_02_08_430070	88	22	=	=	NN
10_1101-2021_02_08_430070	88	23	W	w	NN
10_1101-2021_02_08_430070	88	24	V	V	NNP
10_1101-2021_02_08_430070	88	25	clip(j−i	clip(j−i	NNP
10_1101-2021_02_08_430070	88	26	,	,	,
10_1101-2021_02_08_430070	88	27	k	k	NN
10_1101-2021_02_08_430070	88	28	)	)	-RRB-
10_1101-2021_02_08_430070	88	29	clip(x	clip(x	NN
10_1101-2021_02_08_430070	88	30	,	,	,
10_1101-2021_02_08_430070	88	31	k	k	LS
10_1101-2021_02_08_430070	88	32	)	)	-RRB-
10_1101-2021_02_08_430070	88	33	=	=	SYM
10_1101-2021_02_08_430070	88	34	max(−k	max(−k	NNP
10_1101-2021_02_08_430070	88	35	,	,	,
10_1101-2021_02_08_430070	88	36	min(k	min(k	NN
10_1101-2021_02_08_430070	88	37	,	,	,
10_1101-2021_02_08_430070	88	38	x	x	NNP
10_1101-2021_02_08_430070	88	39	)	)	-RRB-
10_1101-2021_02_08_430070	88	40	)	)	-RRB-
10_1101-2021_02_08_430070	88	41	2.1.4	2.1.4	CD
10_1101-2021_02_08_430070	88	42	Final	final	JJ
10_1101-2021_02_08_430070	88	43	full	full	JJ
10_1101-2021_02_08_430070	88	44	connection	connection	NN
10_1101-2021_02_08_430070	88	45	layer	layer	NN
10_1101-2021_02_08_430070	88	46	After	after	IN
10_1101-2021_02_08_430070	88	47	the	the	DT
10_1101-2021_02_08_430070	88	48	stacked	stacked	JJ
10_1101-2021_02_08_430070	88	49	transformer	transformer	NN
10_1101-2021_02_08_430070	88	50	blocks	block	NNS
10_1101-2021_02_08_430070	88	51	,	,	,
10_1101-2021_02_08_430070	88	52	hidden	hide	VBN
10_1101-2021_02_08_430070	88	53	units	unit	NNS
10_1101-2021_02_08_430070	88	54	of	of	IN
10_1101-2021_02_08_430070	88	55	the	the	DT
10_1101-2021_02_08_430070	88	56	center	center	NN
10_1101-2021_02_08_430070	88	57	position	position	NN
10_1101-2021_02_08_430070	88	58	feed	feed	VBP
10_1101-2021_02_08_430070	88	59	to	to	IN
10_1101-2021_02_08_430070	88	60	a	a	DT
10_1101-2021_02_08_430070	88	61	full	full	JJ
10_1101-2021_02_08_430070	88	62	connection	connection	NN
10_1101-2021_02_08_430070	88	63	linear	linear	NN
10_1101-2021_02_08_430070	88	64	layer	layer	NN
10_1101-2021_02_08_430070	88	65	that	that	WDT
10_1101-2021_02_08_430070	88	66	makes	make	VBZ
10_1101-2021_02_08_430070	88	67	the	the	DT
10_1101-2021_02_08_430070	88	68	final	final	JJ
10_1101-2021_02_08_430070	88	69	prediction	prediction	NN
10_1101-2021_02_08_430070	88	70	of	of	IN
10_1101-2021_02_08_430070	88	71	whether	whether	IN
10_1101-2021_02_08_430070	88	72	a	a	DT
10_1101-2021_02_08_430070	88	73	given	give	VBN
10_1101-2021_02_08_430070	88	74	input	input	NN
10_1101-2021_02_08_430070	88	75	contains	contain	VBZ
10_1101-2021_02_08_430070	88	76	a	a	DT
10_1101-2021_02_08_430070	88	77	methylated	methylated	JJ
10_1101-2021_02_08_430070	88	78	motif	motif	NN
10_1101-2021_02_08_430070	88	79	or	or	CC
10_1101-2021_02_08_430070	88	80	not	not	RB
10_1101-2021_02_08_430070	88	81	.	.	.
10_1101-2021_02_08_430070	89	1	In	in	IN
10_1101-2021_02_08_430070	89	2	the	the	DT
10_1101-2021_02_08_430070	89	3	refined	refined	JJ
10_1101-2021_02_08_430070	89	4	BERT	BERT	NNP
10_1101-2021_02_08_430070	89	5	,	,	,
10_1101-2021_02_08_430070	89	6	besides	besides	IN
10_1101-2021_02_08_430070	89	7	the	the	DT
10_1101-2021_02_08_430070	89	8	hidden	hide	VBN
10_1101-2021_02_08_430070	89	9	units	unit	NNS
10_1101-2021_02_08_430070	89	10	of	of	IN
10_1101-2021_02_08_430070	89	11	the	the	DT
10_1101-2021_02_08_430070	89	12	center	center	NN
10_1101-2021_02_08_430070	89	13	position	position	NN
10_1101-2021_02_08_430070	89	14	,	,	,
10_1101-2021_02_08_430070	89	15	hidden	hide	VBN
10_1101-2021_02_08_430070	89	16	units	unit	NNS
10_1101-2021_02_08_430070	89	17	in	in	IN
10_1101-2021_02_08_430070	89	18	its	-PRON-	PRP$
10_1101-2021_02_08_430070	89	19	surrounding	surround	VBG
10_1101-2021_02_08_430070	89	20	window	window	NN
10_1101-2021_02_08_430070	89	21	(	(	-LRB-
10_1101-2021_02_08_430070	89	22	e.g.	e.g.	RB
10_1101-2021_02_08_430070	89	23	,	,	,
10_1101-2021_02_08_430070	89	24	±3bp	±3bp	NN
10_1101-2021_02_08_430070	89	25	)	)	-RRB-
10_1101-2021_02_08_430070	89	26	are	be	VBP
10_1101-2021_02_08_430070	89	27	concatenated	concatenate	VBN
10_1101-2021_02_08_430070	89	28	as	as	IN
10_1101-2021_02_08_430070	89	29	the	the	DT
10_1101-2021_02_08_430070	89	30	input	input	NN
10_1101-2021_02_08_430070	89	31	of	of	IN
10_1101-2021_02_08_430070	89	32	the	the	DT
10_1101-2021_02_08_430070	89	33	final	final	JJ
10_1101-2021_02_08_430070	89	34	full	full	JJ
10_1101-2021_02_08_430070	89	35	connection	connection	NN
10_1101-2021_02_08_430070	89	36	layer	layer	NN
10_1101-2021_02_08_430070	89	37	.	.	.
10_1101-2021_02_08_430070	90	1	2.2	2.2	CD
10_1101-2021_02_08_430070	90	2	Applying	apply	VBG
10_1101-2021_02_08_430070	90	3	BERT	BERT	NNP
10_1101-2021_02_08_430070	90	4	models	model	NNS
10_1101-2021_02_08_430070	90	5	for	for	IN
10_1101-2021_02_08_430070	90	6	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	90	7	methylation	methylation	NN
10_1101-2021_02_08_430070	90	8	detection	detection	NN
10_1101-2021_02_08_430070	90	9	The	the	DT
10_1101-2021_02_08_430070	90	10	BERT	BERT	NNP
10_1101-2021_02_08_430070	90	11	models	model	NNS
10_1101-2021_02_08_430070	90	12	are	be	VBP
10_1101-2021_02_08_430070	90	13	then	then	RB
10_1101-2021_02_08_430070	90	14	applied	apply	VBN
10_1101-2021_02_08_430070	90	15	to	to	TO
10_1101-2021_02_08_430070	90	16	replace	replace	VB
10_1101-2021_02_08_430070	90	17	different	different	JJ
10_1101-2021_02_08_430070	90	18	classification	classification	NN
10_1101-2021_02_08_430070	90	19	models	model	NNS
10_1101-2021_02_08_430070	90	20	(	(	-LRB-
10_1101-2021_02_08_430070	90	21	e.g.	e.g.	RB
10_1101-2021_02_08_430070	91	1	biRNN	birnn	LS
10_1101-2021_02_08_430070	91	2	)	)	-RRB-
10_1101-2021_02_08_430070	91	3	in	in	IN
10_1101-2021_02_08_430070	91	4	a	a	DT
10_1101-2021_02_08_430070	91	5	typical	typical	JJ
10_1101-2021_02_08_430070	91	6	model	model	NN
10_1101-2021_02_08_430070	91	7	-	-	HYPH
10_1101-2021_02_08_430070	91	8	based	base	VBN
10_1101-2021_02_08_430070	91	9	methylation	methylation	NN
10_1101-2021_02_08_430070	91	10	detection	detection	NN
10_1101-2021_02_08_430070	91	11	framework	framework	NN
10_1101-2021_02_08_430070	91	12	.	.	.
10_1101-2021_02_08_430070	92	1	In	in	IN
10_1101-2021_02_08_430070	92	2	this	this	DT
10_1101-2021_02_08_430070	92	3	framework	framework	NN
10_1101-2021_02_08_430070	92	4	,	,	,
10_1101-2021_02_08_430070	92	5	raw	raw	JJ
10_1101-2021_02_08_430070	92	6	signals	signal	NNS
10_1101-2021_02_08_430070	92	7	of	of	IN
10_1101-2021_02_08_430070	92	8	each	each	DT
10_1101-2021_02_08_430070	92	9	read	read	NN
10_1101-2021_02_08_430070	92	10	are	be	VBP
10_1101-2021_02_08_430070	92	11	first	first	RB
10_1101-2021_02_08_430070	92	12	translated	translate	VBN
10_1101-2021_02_08_430070	92	13	into	into	IN
10_1101-2021_02_08_430070	92	14	nucleotide	nucleotide	JJ
10_1101-2021_02_08_430070	92	15	sequences	sequence	NNS
10_1101-2021_02_08_430070	92	16	(	(	-LRB-
10_1101-2021_02_08_430070	92	17	basecalling	basecalle	VBG
10_1101-2021_02_08_430070	92	18	)	)	-RRB-
10_1101-2021_02_08_430070	92	19	.	.	.
10_1101-2021_02_08_430070	93	1	Signals	signal	NNS
10_1101-2021_02_08_430070	93	2	are	be	VBP
10_1101-2021_02_08_430070	93	3	then	then	RB
10_1101-2021_02_08_430070	93	4	aligned	align	VBN
10_1101-2021_02_08_430070	93	5	to	to	IN
10_1101-2021_02_08_430070	93	6	corresponding	correspond	VBG
10_1101-2021_02_08_430070	93	7	reference	reference	NN
10_1101-2021_02_08_430070	93	8	nucleotides	nucleotide	NNS
10_1101-2021_02_08_430070	93	9	through	through	IN
10_1101-2021_02_08_430070	93	10	the	the	DT
10_1101-2021_02_08_430070	93	11	re	re	JJ
10_1101-2021_02_08_430070	93	12	-	-	NN
10_1101-2021_02_08_430070	93	13	squiggle	squiggle	JJ
10_1101-2021_02_08_430070	93	14	process	process	NN
10_1101-2021_02_08_430070	93	15	.	.	.
10_1101-2021_02_08_430070	94	1	After	after	IN
10_1101-2021_02_08_430070	94	2	that	that	DT
10_1101-2021_02_08_430070	94	3	,	,	,
10_1101-2021_02_08_430070	94	4	the	the	DT
10_1101-2021_02_08_430070	94	5	target	target	NN
10_1101-2021_02_08_430070	94	6	motif	motif	NN
10_1101-2021_02_08_430070	94	7	(	(	-LRB-
10_1101-2021_02_08_430070	94	8	e.g.	e.g.	RB
10_1101-2021_02_08_430070	95	1	CpG	CpG	NNP
10_1101-2021_02_08_430070	95	2	)	)	-RRB-
10_1101-2021_02_08_430070	95	3	and	and	CC
10_1101-2021_02_08_430070	95	4	its	-PRON-	PRP$
10_1101-2021_02_08_430070	95	5	context	context	NN
10_1101-2021_02_08_430070	95	6	regions	region	NNS
10_1101-2021_02_08_430070	95	7	are	be	VBP
10_1101-2021_02_08_430070	95	8	localized	localize	VBN
10_1101-2021_02_08_430070	95	9	through	through	IN
10_1101-2021_02_08_430070	95	10	nucleotide	nucleotide	JJ
10_1101-2021_02_08_430070	95	11	matching	matching	NN
10_1101-2021_02_08_430070	95	12	and	and	CC
10_1101-2021_02_08_430070	95	13	signals	signal	NNS
10_1101-2021_02_08_430070	95	14	in	in	IN
10_1101-2021_02_08_430070	95	15	a	a	DT
10_1101-2021_02_08_430070	95	16	context	context	NN
10_1101-2021_02_08_430070	95	17	window	window	NN
10_1101-2021_02_08_430070	95	18	of	of	IN
10_1101-2021_02_08_430070	95	19	a	a	DT
10_1101-2021_02_08_430070	95	20	fixed	fix	VBN
10_1101-2021_02_08_430070	95	21	length	length	NN
10_1101-2021_02_08_430070	95	22	(	(	-LRB-
10_1101-2021_02_08_430070	95	23	e.g.	e.g.	RB
10_1101-2021_02_08_430070	96	1	21bp	21bp	LS
10_1101-2021_02_08_430070	96	2	)	)	-RRB-
10_1101-2021_02_08_430070	96	3	are	be	VBP
10_1101-2021_02_08_430070	96	4	transformed	transform	VBN
10_1101-2021_02_08_430070	96	5	into	into	IN
10_1101-2021_02_08_430070	96	6	event	event	NN
10_1101-2021_02_08_430070	96	7	-	-	HYPH
10_1101-2021_02_08_430070	96	8	based	base	VBN
10_1101-2021_02_08_430070	96	9	features	feature	NNS
10_1101-2021_02_08_430070	96	10	as	as	IN
10_1101-2021_02_08_430070	96	11	the	the	DT
10_1101-2021_02_08_430070	96	12	input	input	NN
10_1101-2021_02_08_430070	96	13	of	of	IN
10_1101-2021_02_08_430070	96	14	methylation	methylation	NN
10_1101-2021_02_08_430070	96	15	callers	caller	NNS
10_1101-2021_02_08_430070	96	16	.	.	.
10_1101-2021_02_08_430070	97	1	Typical	typical	JJ
10_1101-2021_02_08_430070	97	2	event	event	NN
10_1101-2021_02_08_430070	97	3	-	-	HYPH
10_1101-2021_02_08_430070	97	4	based	base	VBN
10_1101-2021_02_08_430070	97	5	features	feature	NNS
10_1101-2021_02_08_430070	97	6	include	include	VBP
10_1101-2021_02_08_430070	97	7	signal	signal	NNP
10_1101-2021_02_08_430070	97	8	mean	mean	NNP
10_1101-2021_02_08_430070	97	9	,	,	,
10_1101-2021_02_08_430070	97	10	signal	signal	NNP
10_1101-2021_02_08_430070	97	11	standard	standard	JJ
10_1101-2021_02_08_430070	97	12	deviation	deviation	NN
10_1101-2021_02_08_430070	97	13	,	,	,
10_1101-2021_02_08_430070	97	14	event	event	NN
10_1101-2021_02_08_430070	97	15	length	length	NN
10_1101-2021_02_08_430070	97	16	,	,	,
10_1101-2021_02_08_430070	97	17	and	and	CC
10_1101-2021_02_08_430070	97	18	nucleotide	nucleotide	JJ
10_1101-2021_02_08_430070	97	19	information	information	NN
10_1101-2021_02_08_430070	97	20	(	(	-LRB-
10_1101-2021_02_08_430070	97	21	Liu	Liu	NNP
10_1101-2021_02_08_430070	97	22	et	et	NNP
10_1101-2021_02_08_430070	97	23	al	al	NNP
10_1101-2021_02_08_430070	97	24	.	.	NNP
10_1101-2021_02_08_430070	97	25	,	,	,
10_1101-2021_02_08_430070	97	26	2019	2019	CD
10_1101-2021_02_08_430070	97	27	)	)	-RRB-
10_1101-2021_02_08_430070	97	28	.	.	.
10_1101-2021_02_08_430070	98	1	Here	here	RB
10_1101-2021_02_08_430070	98	2	,	,	,
10_1101-2021_02_08_430070	98	3	we	-PRON-	PRP
10_1101-2021_02_08_430070	98	4	utilize	utilize	VBP
10_1101-2021_02_08_430070	98	5	the	the	DT
10_1101-2021_02_08_430070	98	6	framework	framework	NN
10_1101-2021_02_08_430070	98	7	of	of	IN
10_1101-2021_02_08_430070	98	8	deepMOD	deepMOD	NNP
10_1101-2021_02_08_430070	98	9	and	and	CC
10_1101-2021_02_08_430070	98	10	perform	perform	VB
10_1101-2021_02_08_430070	98	11	the	the	DT
10_1101-2021_02_08_430070	98	12	same	same	JJ
10_1101-2021_02_08_430070	98	13	pre	pre	NN
10_1101-2021_02_08_430070	98	14	-	-	NN
10_1101-2021_02_08_430070	98	15	process	process	NN
10_1101-2021_02_08_430070	98	16	for	for	IN
10_1101-2021_02_08_430070	98	17	the	the	DT
10_1101-2021_02_08_430070	98	18	data	datum	NNS
10_1101-2021_02_08_430070	98	19	.	.	.
10_1101-2021_02_08_430070	99	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	99	2	use	use	VBP
10_1101-2021_02_08_430070	99	3	Tombo	Tombo	NNP
10_1101-2021_02_08_430070	99	4	(	(	-LRB-
10_1101-2021_02_08_430070	99	5	Ver	Ver	NNP
10_1101-2021_02_08_430070	99	6	1.5.1	1.5.1	NNP
10_1101-2021_02_08_430070	99	7	)	)	-RRB-
10_1101-2021_02_08_430070	99	8	to	to	TO
10_1101-2021_02_08_430070	99	9	perform	perform	VB
10_1101-2021_02_08_430070	99	10	re-	re-	JJ
10_1101-2021_02_08_430070	99	11	squiggling	squiggle	VBG
10_1101-2021_02_08_430070	99	12	and	and	CC
10_1101-2021_02_08_430070	99	13	utilize	utilize	VB
10_1101-2021_02_08_430070	99	14	Minimap2	Minimap2	NNP
10_1101-2021_02_08_430070	99	15	(	(	-LRB-
10_1101-2021_02_08_430070	99	16	Ver	Ver	NNP
10_1101-2021_02_08_430070	99	17	2.17-r941	2.17-r941	CD
10_1101-2021_02_08_430070	99	18	)	)	-RRB-
10_1101-2021_02_08_430070	99	19	to	to	TO
10_1101-2021_02_08_430070	99	20	align	align	VB
10_1101-2021_02_08_430070	99	21	events	event	NNS
10_1101-2021_02_08_430070	99	22	to	to	IN
10_1101-2021_02_08_430070	99	23	the	the	DT
10_1101-2021_02_08_430070	99	24	reference	reference	NN
10_1101-2021_02_08_430070	99	25	genome	genome	JJ
10_1101-2021_02_08_430070	99	26	.	.	.
10_1101-2021_02_08_430070	100	1	Here	here	RB
10_1101-2021_02_08_430070	100	2	,	,	,
10_1101-2021_02_08_430070	100	3	we	-PRON-	PRP
10_1101-2021_02_08_430070	100	4	use	use	VBP
10_1101-2021_02_08_430070	100	5	E.coli	e.coli	JJ
10_1101-2021_02_08_430070	100	6	K-12	k-12	CD
10_1101-2021_02_08_430070	100	7	MG1655	mg1655	CD
10_1101-2021_02_08_430070	100	8	and	and	CC
10_1101-2021_02_08_430070	100	9	H.Sapiens	H.Sapiens	NNP
10_1101-2021_02_08_430070	100	10	GRCh38	GRCh38	NNP
10_1101-2021_02_08_430070	100	11	as	as	IN
10_1101-2021_02_08_430070	100	12	the	the	DT
10_1101-2021_02_08_430070	100	13	reference	reference	NN
10_1101-2021_02_08_430070	100	14	genomes	genome	VBZ
10_1101-2021_02_08_430070	100	15	.	.	.
10_1101-2021_02_08_430070	101	1	3	3	LS
10_1101-2021_02_08_430070	101	2	Experiments	experiment	NNS
10_1101-2021_02_08_430070	101	3	We	-PRON-	PRP
10_1101-2021_02_08_430070	101	4	compare	compare	VBP
10_1101-2021_02_08_430070	101	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	101	6	models	model	NNS
10_1101-2021_02_08_430070	101	7	with	with	IN
10_1101-2021_02_08_430070	101	8	the	the	DT
10_1101-2021_02_08_430070	101	9	state	state	NN
10_1101-2021_02_08_430070	101	10	-	-	HYPH
10_1101-2021_02_08_430070	101	11	of	of	IN
10_1101-2021_02_08_430070	101	12	-	-	HYPH
10_1101-2021_02_08_430070	101	13	the	the	DT
10_1101-2021_02_08_430070	101	14	-	-	HYPH
10_1101-2021_02_08_430070	101	15	art	art	NN
10_1101-2021_02_08_430070	101	16	biRNN	birnn	NN
10_1101-2021_02_08_430070	101	17	model	model	NN
10_1101-2021_02_08_430070	101	18	,	,	,
10_1101-2021_02_08_430070	101	19	which	which	WDT
10_1101-2021_02_08_430070	101	20	is	be	VBZ
10_1101-2021_02_08_430070	101	21	used	use	VBN
10_1101-2021_02_08_430070	101	22	as	as	IN
10_1101-2021_02_08_430070	101	23	the	the	DT
10_1101-2021_02_08_430070	101	24	basic	basic	JJ
10_1101-2021_02_08_430070	101	25	network	network	NN
10_1101-2021_02_08_430070	101	26	structure	structure	NN
10_1101-2021_02_08_430070	101	27	in	in	IN
10_1101-2021_02_08_430070	101	28	DeepMOD	DeepMOD	NNP
10_1101-2021_02_08_430070	101	29	(	(	-LRB-
10_1101-2021_02_08_430070	101	30	Liu	Liu	NNP
10_1101-2021_02_08_430070	101	31	et	et	NNP
10_1101-2021_02_08_430070	101	32	al	al	NNP
10_1101-2021_02_08_430070	101	33	.	.	NNP
10_1101-2021_02_08_430070	101	34	,	,	,
10_1101-2021_02_08_430070	101	35	2019	2019	CD
10_1101-2021_02_08_430070	101	36	)	)	-RRB-
10_1101-2021_02_08_430070	101	37	and	and	CC
10_1101-2021_02_08_430070	101	38	DeepSignal	DeepSignal	NNP
10_1101-2021_02_08_430070	101	39	(	(	-LRB-
10_1101-2021_02_08_430070	101	40	Ni	Ni	NNP
10_1101-2021_02_08_430070	101	41	et	et	FW
10_1101-2021_02_08_430070	101	42	al	al	NNP
10_1101-2021_02_08_430070	101	43	.	.	NNP
10_1101-2021_02_08_430070	101	44	,	,	,
10_1101-2021_02_08_430070	101	45	2019	2019	CD
10_1101-2021_02_08_430070	101	46	)	)	-RRB-
10_1101-2021_02_08_430070	101	47	.	.	.
10_1101-2021_02_08_430070	102	1	To	to	TO
10_1101-2021_02_08_430070	102	2	compare	compare	VB
10_1101-2021_02_08_430070	102	3	with	with	IN
10_1101-2021_02_08_430070	102	4	other	other	JJ
10_1101-2021_02_08_430070	102	5	non	non	JJ
10_1101-2021_02_08_430070	102	6	-	-	JJ
10_1101-2021_02_08_430070	102	7	deep	deep	JJ
10_1101-2021_02_08_430070	102	8	-	-	HYPH
10_1101-2021_02_08_430070	102	9	learning-	learning-	NN
10_1101-2021_02_08_430070	102	10	based	base	VBN
10_1101-2021_02_08_430070	102	11	methods	method	NNS
10_1101-2021_02_08_430070	102	12	,	,	,
10_1101-2021_02_08_430070	102	13	we	-PRON-	PRP
10_1101-2021_02_08_430070	102	14	utilized	utilize	VBD
10_1101-2021_02_08_430070	102	15	the	the	DT
10_1101-2021_02_08_430070	102	16	CpG	CpG	NNP
10_1101-2021_02_08_430070	102	17	benchmark	benchmark	NNP
10_1101-2021_02_08_430070	102	18	pipeline	pipeline	NN
10_1101-2021_02_08_430070	102	19	(	(	-LRB-
10_1101-2021_02_08_430070	102	20	Yuen	Yuen	NNP
10_1101-2021_02_08_430070	102	21	et	et	FW
10_1101-2021_02_08_430070	102	22	al	al	NNP
10_1101-2021_02_08_430070	102	23	.	.	NNP
10_1101-2021_02_08_430070	102	24	,	,	,
10_1101-2021_02_08_430070	102	25	2020	2020	CD
10_1101-2021_02_08_430070	102	26	)	)	-RRB-
10_1101-2021_02_08_430070	102	27	as	as	IN
10_1101-2021_02_08_430070	102	28	a	a	DT
10_1101-2021_02_08_430070	102	29	pivot	pivot	NN
10_1101-2021_02_08_430070	102	30	.	.	.
10_1101-2021_02_08_430070	103	1	3.1	3.1	CD
10_1101-2021_02_08_430070	103	2	Data	datum	NNS
10_1101-2021_02_08_430070	103	3	and	and	CC
10_1101-2021_02_08_430070	103	4	model	model	NN
10_1101-2021_02_08_430070	103	5	parameters	parameter	NNS
10_1101-2021_02_08_430070	103	6	We	-PRON-	PRP
10_1101-2021_02_08_430070	103	7	train	train	VBP
10_1101-2021_02_08_430070	103	8	and	and	CC
10_1101-2021_02_08_430070	103	9	test	test	VBP
10_1101-2021_02_08_430070	103	10	the	the	DT
10_1101-2021_02_08_430070	103	11	models	model	NNS
10_1101-2021_02_08_430070	103	12	on	on	IN
10_1101-2021_02_08_430070	103	13	the	the	DT
10_1101-2021_02_08_430070	103	14	public	public	JJ
10_1101-2021_02_08_430070	103	15	accessible	accessible	JJ
10_1101-2021_02_08_430070	103	16	5mC	5mc	CD
10_1101-2021_02_08_430070	103	17	(	(	-LRB-
10_1101-2021_02_08_430070	103	18	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	103	19	et	et	FW
10_1101-2021_02_08_430070	103	20	al	al	NNP
10_1101-2021_02_08_430070	103	21	.	.	NNP
10_1101-2021_02_08_430070	103	22	,	,	,
10_1101-2021_02_08_430070	103	23	2016	2016	CD
10_1101-2021_02_08_430070	103	24	;	;	:
10_1101-2021_02_08_430070	103	25	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	103	26	et	et	FW
10_1101-2021_02_08_430070	103	27	al	al	NNP
10_1101-2021_02_08_430070	103	28	.	.	NNP
10_1101-2021_02_08_430070	103	29	,	,	,
10_1101-2021_02_08_430070	103	30	2017	2017	CD
10_1101-2021_02_08_430070	103	31	)	)	-RRB-
10_1101-2021_02_08_430070	103	32	and	and	CC
10_1101-2021_02_08_430070	103	33	6mA	6ma	CD
10_1101-2021_02_08_430070	103	34	(	(	-LRB-
10_1101-2021_02_08_430070	103	35	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	103	36	et	et	FW
10_1101-2021_02_08_430070	103	37	al	al	NNP
10_1101-2021_02_08_430070	103	38	.	.	NNP
10_1101-2021_02_08_430070	103	39	,	,	,
10_1101-2021_02_08_430070	103	40	2016	2016	CD
10_1101-2021_02_08_430070	103	41	)	)	-RRB-
10_1101-2021_02_08_430070	103	42	datasets	dataset	VBZ
10_1101-2021_02_08_430070	103	43	.	.	.
10_1101-2021_02_08_430070	104	1	The	the	DT
10_1101-2021_02_08_430070	104	2	datasets	dataset	NNS
10_1101-2021_02_08_430070	104	3	include	include	VBP
10_1101-2021_02_08_430070	104	4	samples	sample	NNS
10_1101-2021_02_08_430070	104	5	of	of	IN
10_1101-2021_02_08_430070	104	6	E.coli	e.coli	JJ
10_1101-2021_02_08_430070	104	7	K-12	k-12	CD
10_1101-2021_02_08_430070	104	8	MG1655	mg1655	NN
10_1101-2021_02_08_430070	104	9	,	,	,
10_1101-2021_02_08_430070	104	10	K-12	k-12	CD
10_1101-2021_02_08_430070	104	11	ER2925	ER2925	NNP
10_1101-2021_02_08_430070	104	12	,	,	,
10_1101-2021_02_08_430070	104	13	and	and	CC
10_1101-2021_02_08_430070	104	14	H.sapiens	H.sapiens	NNP
10_1101-2021_02_08_430070	104	15	NA12878	NA12878	NNP
10_1101-2021_02_08_430070	104	16	.	.	.
10_1101-2021_02_08_430070	105	1	Negative	negative	JJ
10_1101-2021_02_08_430070	105	2	control	control	NN
10_1101-2021_02_08_430070	105	3	samples	sample	NNS
10_1101-2021_02_08_430070	105	4	are	be	VBP
10_1101-2021_02_08_430070	105	5	amplified	amplify	VBN
10_1101-2021_02_08_430070	105	6	with	with	IN
10_1101-2021_02_08_430070	105	7	PCR	PCR	NNP
10_1101-2021_02_08_430070	105	8	and	and	CC
10_1101-2021_02_08_430070	105	9	no	no	DT
10_1101-2021_02_08_430070	105	10	modified	modify	VBN
10_1101-2021_02_08_430070	105	11	bases	basis	NNS
10_1101-2021_02_08_430070	105	12	are	be	VBP
10_1101-2021_02_08_430070	105	13	included	include	VBN
10_1101-2021_02_08_430070	105	14	.	.	.
10_1101-2021_02_08_430070	106	1	Positive	positive	JJ
10_1101-2021_02_08_430070	106	2	control	control	NN
10_1101-2021_02_08_430070	106	3	samples	sample	NNS
10_1101-2021_02_08_430070	106	4	are	be	VBP
10_1101-2021_02_08_430070	106	5	synthetically	synthetically	RB
10_1101-2021_02_08_430070	106	6	introduced	introduce	VBN
10_1101-2021_02_08_430070	106	7	by	by	IN
10_1101-2021_02_08_430070	106	8	specific	specific	JJ
10_1101-2021_02_08_430070	106	9	enzymes	enzyme	NNS
10_1101-2021_02_08_430070	106	10	after	after	IN
10_1101-2021_02_08_430070	106	11	PCR	PCR	NNP
10_1101-2021_02_08_430070	106	12	amplification	amplification	NN
10_1101-2021_02_08_430070	106	13	,	,	,
10_1101-2021_02_08_430070	106	14	which	which	WDT
10_1101-2021_02_08_430070	106	15	includes	include	VBZ
10_1101-2021_02_08_430070	106	16	SssI	SssI	NNP
10_1101-2021_02_08_430070	106	17	,	,	,
10_1101-2021_02_08_430070	106	18	Hhal	Hhal	NNP
10_1101-2021_02_08_430070	106	19	,	,	,
10_1101-2021_02_08_430070	106	20	MpeI	MpeI	NNP
10_1101-2021_02_08_430070	106	21	methylases	methylase	VBZ
10_1101-2021_02_08_430070	106	22	for	for	IN
10_1101-2021_02_08_430070	106	23	5mC	5mc	CD
10_1101-2021_02_08_430070	106	24	,	,	,
10_1101-2021_02_08_430070	106	25	and	and	CC
10_1101-2021_02_08_430070	106	26	TaqI	TaqI	NNP
10_1101-2021_02_08_430070	106	27	,	,	,
10_1101-2021_02_08_430070	106	28	EcoRI	EcoRI	NNP
10_1101-2021_02_08_430070	106	29	,	,	,
10_1101-2021_02_08_430070	106	30	and	and	CC
10_1101-2021_02_08_430070	106	31	Dam	Dam	NNP
10_1101-2021_02_08_430070	106	32	for	for	IN
10_1101-2021_02_08_430070	106	33	6mA	6ma	CD
10_1101-2021_02_08_430070	106	34	modification	modification	NN
10_1101-2021_02_08_430070	106	35	.	.	.
10_1101-2021_02_08_430070	107	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	107	2	use	use	VBP
10_1101-2021_02_08_430070	107	3	the	the	DT
10_1101-2021_02_08_430070	107	4	samples	sample	NNS
10_1101-2021_02_08_430070	107	5	that	that	WDT
10_1101-2021_02_08_430070	107	6	are	be	VBP
10_1101-2021_02_08_430070	107	7	sequenced	sequence	VBN
10_1101-2021_02_08_430070	107	8	with	with	IN
10_1101-2021_02_08_430070	107	9	Oxford	Oxford	NNP
10_1101-2021_02_08_430070	107	10	Nanopore	Nanopore	NNP
10_1101-2021_02_08_430070	107	11	R9	r9	JJ
10_1101-2021_02_08_430070	107	12	flow	flow	NN
10_1101-2021_02_08_430070	107	13	cells	cell	NNS
10_1101-2021_02_08_430070	107	14	.	.	.
10_1101-2021_02_08_430070	108	1	For	for	IN
10_1101-2021_02_08_430070	108	2	each	each	DT
10_1101-2021_02_08_430070	108	3	dataset	dataset	NN
10_1101-2021_02_08_430070	108	4	,	,	,
10_1101-2021_02_08_430070	108	5	we	-PRON-	PRP
10_1101-2021_02_08_430070	108	6	randomly	randomly	RB
10_1101-2021_02_08_430070	108	7	shuffle	shuffle	VBP
10_1101-2021_02_08_430070	108	8	reads	read	VBZ
10_1101-2021_02_08_430070	108	9	in	in	IN
10_1101-2021_02_08_430070	108	10	positive	positive	JJ
10_1101-2021_02_08_430070	108	11	and	and	CC
10_1101-2021_02_08_430070	108	12	negative	negative	JJ
10_1101-2021_02_08_430070	108	13	controls	control	NNS
10_1101-2021_02_08_430070	108	14	and	and	CC
10_1101-2021_02_08_430070	108	15	construct	construct	VB
10_1101-2021_02_08_430070	108	16	the	the	DT
10_1101-2021_02_08_430070	108	17	training	training	NN
10_1101-2021_02_08_430070	108	18	,	,	,
10_1101-2021_02_08_430070	108	19	validate	validate	NN
10_1101-2021_02_08_430070	108	20	and	and	CC
10_1101-2021_02_08_430070	108	21	test	test	NN
10_1101-2021_02_08_430070	108	22	set	set	NN
10_1101-2021_02_08_430070	108	23	according	accord	VBG
10_1101-2021_02_08_430070	108	24	to	to	IN
10_1101-2021_02_08_430070	108	25	a	a	DT
10_1101-2021_02_08_430070	108	26	split	split	JJ
10_1101-2021_02_08_430070	108	27	proportion	proportion	NN
10_1101-2021_02_08_430070	108	28	of	of	IN
10_1101-2021_02_08_430070	108	29	80/10/10	80/10/10	CD
10_1101-2021_02_08_430070	108	30	for	for	IN
10_1101-2021_02_08_430070	108	31	in	in	IN
10_1101-2021_02_08_430070	108	32	-	-	HYPH
10_1101-2021_02_08_430070	108	33	sample	sample	NN
10_1101-2021_02_08_430070	108	34	evaluation	evaluation	NN
10_1101-2021_02_08_430070	108	35	.	.	.
10_1101-2021_02_08_430070	109	1	For	for	IN
10_1101-2021_02_08_430070	109	2	the	the	DT
10_1101-2021_02_08_430070	109	3	cross	cross	JJ
10_1101-2021_02_08_430070	109	4	-	-	JJ
10_1101-2021_02_08_430070	109	5	sample	sample	JJ
10_1101-2021_02_08_430070	109	6	evaluation	evaluation	NN
10_1101-2021_02_08_430070	109	7	,	,	,
10_1101-2021_02_08_430070	109	8	we	-PRON-	PRP
10_1101-2021_02_08_430070	109	9	train	train	VBP
10_1101-2021_02_08_430070	109	10	models	model	NNS
10_1101-2021_02_08_430070	109	11	on	on	IN
10_1101-2021_02_08_430070	109	12	one	one	CD
10_1101-2021_02_08_430070	109	13	dataset	dataset	NN
10_1101-2021_02_08_430070	109	14	and	and	CC
10_1101-2021_02_08_430070	109	15	test	test	VB
10_1101-2021_02_08_430070	109	16	on	on	IN
10_1101-2021_02_08_430070	109	17	the	the	DT
10_1101-2021_02_08_430070	109	18	other	other	JJ
10_1101-2021_02_08_430070	109	19	dataset	dataset	NN
10_1101-2021_02_08_430070	109	20	.	.	.
10_1101-2021_02_08_430070	110	1	BiRNN	birnn	DT
10_1101-2021_02_08_430070	110	2	uses	use	VBZ
10_1101-2021_02_08_430070	110	3	the	the	DT
10_1101-2021_02_08_430070	110	4	default	default	NN
10_1101-2021_02_08_430070	110	5	model	model	NN
10_1101-2021_02_08_430070	110	6	architecture	architecture	NN
10_1101-2021_02_08_430070	110	7	and	and	CC
10_1101-2021_02_08_430070	110	8	parameter	parameter	NN
10_1101-2021_02_08_430070	110	9	setting	setting	NN
10_1101-2021_02_08_430070	110	10	of	of	IN
10_1101-2021_02_08_430070	110	11	DeepMOD	DeepMOD	NNP
10_1101-2021_02_08_430070	110	12	,	,	,
10_1101-2021_02_08_430070	110	13	which	which	WDT
10_1101-2021_02_08_430070	110	14	consists	consist	VBZ
10_1101-2021_02_08_430070	110	15	of	of	IN
10_1101-2021_02_08_430070	110	16	three	three	CD
10_1101-2021_02_08_430070	110	17	stacked	stack	VBN
10_1101-2021_02_08_430070	110	18	bi	bi	JJ
10_1101-2021_02_08_430070	110	19	-	-	JJ
10_1101-2021_02_08_430070	110	20	directional	directional	JJ
10_1101-2021_02_08_430070	110	21	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	110	22	layers	layer	NNS
10_1101-2021_02_08_430070	110	23	(	(	-LRB-
10_1101-2021_02_08_430070	110	24	hidden_size=100	hidden_size=100	NNP
10_1101-2021_02_08_430070	110	25	)	)	-RRB-
10_1101-2021_02_08_430070	110	26	and	and	CC
10_1101-2021_02_08_430070	110	27	one	one	CD
10_1101-2021_02_08_430070	110	28	full	full	JJ
10_1101-2021_02_08_430070	110	29	connection	connection	NN
10_1101-2021_02_08_430070	110	30	layer	layer	NN
10_1101-2021_02_08_430070	110	31	for	for	IN
10_1101-2021_02_08_430070	110	32	the	the	DT
10_1101-2021_02_08_430070	110	33	center	center	NN
10_1101-2021_02_08_430070	110	34	position	position	NN
10_1101-2021_02_08_430070	110	35	.	.	.
10_1101-2021_02_08_430070	111	1	The	the	DT
10_1101-2021_02_08_430070	111	2	total	total	JJ
10_1101-2021_02_08_430070	111	3	number	number	NN
10_1101-2021_02_08_430070	111	4	of	of	IN
10_1101-2021_02_08_430070	111	5	biRNN	birnn	JJ
10_1101-2021_02_08_430070	111	6	parameters	parameter	NNS
10_1101-2021_02_08_430070	111	7	is	be	VBZ
10_1101-2021_02_08_430070	111	8	570,802	570,802	CD
10_1101-2021_02_08_430070	111	9	for	for	IN
10_1101-2021_02_08_430070	111	10	an	an	DT
10_1101-2021_02_08_430070	111	11	input	input	JJ
10_1101-2021_02_08_430070	111	12	length	length	NN
10_1101-2021_02_08_430070	111	13	of	of	IN
10_1101-2021_02_08_430070	111	14	21bp	21bp	NN
10_1101-2021_02_08_430070	111	15	.	.	.
10_1101-2021_02_08_430070	112	1	BERTs	bert	NNS
10_1101-2021_02_08_430070	112	2	use	use	VBP
10_1101-2021_02_08_430070	112	3	three	three	CD
10_1101-2021_02_08_430070	112	4	attention	attention	NN
10_1101-2021_02_08_430070	112	5	layers	layer	NNS
10_1101-2021_02_08_430070	112	6	(	(	-LRB-
10_1101-2021_02_08_430070	112	7	hidden_size=100	hidden_size=100	UH
10_1101-2021_02_08_430070	112	8	,	,	,
10_1101-2021_02_08_430070	112	9	attention_head=4	attention_head=4	NNP
10_1101-2021_02_08_430070	112	10	)	)	-RRB-
10_1101-2021_02_08_430070	112	11	and	and	CC
10_1101-2021_02_08_430070	112	12	one	one	CD
10_1101-2021_02_08_430070	112	13	full	full	JJ
10_1101-2021_02_08_430070	112	14	connection	connection	NN
10_1101-2021_02_08_430070	112	15	layer	layer	NN
10_1101-2021_02_08_430070	112	16	.	.	.
10_1101-2021_02_08_430070	113	1	For	for	IN
10_1101-2021_02_08_430070	113	2	the	the	DT
10_1101-2021_02_08_430070	113	3	refined	refined	JJ
10_1101-2021_02_08_430070	113	4	BERT	BERT	NNP
10_1101-2021_02_08_430070	113	5	,	,	,
10_1101-2021_02_08_430070	113	6	learnable	learnable	JJ
10_1101-2021_02_08_430070	113	7	positional	positional	JJ
10_1101-2021_02_08_430070	113	8	encoding	encoding	NN
10_1101-2021_02_08_430070	113	9	,	,	,
10_1101-2021_02_08_430070	113	10	attention	attention	NN
10_1101-2021_02_08_430070	113	11	with	with	IN
10_1101-2021_02_08_430070	113	12	relative	relative	JJ
10_1101-2021_02_08_430070	113	13	position	position	NN
10_1101-2021_02_08_430070	113	14	representation	representation	NN
10_1101-2021_02_08_430070	113	15	and	and	CC
10_1101-2021_02_08_430070	113	16	center	center	NN
10_1101-2021_02_08_430070	113	17	-	-	HYPH
10_1101-2021_02_08_430070	113	18	hidden	hide	VBN
10_1101-2021_02_08_430070	113	19	-	-	HYPH
10_1101-2021_02_08_430070	113	20	concatenation	concatenation	NN
10_1101-2021_02_08_430070	113	21	are	be	VBP
10_1101-2021_02_08_430070	113	22	used	use	VBN
10_1101-2021_02_08_430070	113	23	.	.	.
10_1101-2021_02_08_430070	114	1	For	for	IN
10_1101-2021_02_08_430070	114	2	BERT	BERT	NNP
10_1101-2021_02_08_430070	114	3	and	and	CC
10_1101-2021_02_08_430070	114	4	refined	refine	VBN
10_1101-2021_02_08_430070	114	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	114	6	,	,	,
10_1101-2021_02_08_430070	114	7	there	there	EX
10_1101-2021_02_08_430070	114	8	are	be	VBP
10_1101-2021_02_08_430070	114	9	total	total	NN
10_1101-2021_02_08_430070	114	10	of	of	IN
10_1101-2021_02_08_430070	114	11	364,902	364,902	CD
10_1101-2021_02_08_430070	114	12	and	and	CC
10_1101-2021_02_08_430070	114	13	368,202	368,202	CD
10_1101-2021_02_08_430070	114	14	parameters	parameter	NNS
10_1101-2021_02_08_430070	114	15	,	,	,
10_1101-2021_02_08_430070	114	16	which	which	WDT
10_1101-2021_02_08_430070	114	17	are	be	VBP
10_1101-2021_02_08_430070	114	18	around	around	RB
10_1101-2021_02_08_430070	114	19	35	35	CD
10_1101-2021_02_08_430070	114	20	%	%	NN
10_1101-2021_02_08_430070	114	21	less	less	JJR
10_1101-2021_02_08_430070	114	22	than	than	IN
10_1101-2021_02_08_430070	114	23	that	that	DT
10_1101-2021_02_08_430070	114	24	of	of	IN
10_1101-2021_02_08_430070	114	25	biRNN	birnn	NN
10_1101-2021_02_08_430070	114	26	.	.	.
10_1101-2021_02_08_430070	115	1	More	more	RBR
10_1101-2021_02_08_430070	115	2	detailed	detailed	JJ
10_1101-2021_02_08_430070	115	3	information	information	NN
10_1101-2021_02_08_430070	115	4	on	on	IN
10_1101-2021_02_08_430070	115	5	the	the	DT
10_1101-2021_02_08_430070	115	6	model	model	NN
10_1101-2021_02_08_430070	115	7	structures	structure	NNS
10_1101-2021_02_08_430070	115	8	is	be	VBZ
10_1101-2021_02_08_430070	115	9	described	describe	VBN
10_1101-2021_02_08_430070	115	10	in	in	IN
10_1101-2021_02_08_430070	115	11	the	the	DT
10_1101-2021_02_08_430070	115	12	supplement	supplement	NN
10_1101-2021_02_08_430070	115	13	material	material	NN
10_1101-2021_02_08_430070	115	14	.	.	.
10_1101-2021_02_08_430070	116	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	116	2	implement	implement	VBP
10_1101-2021_02_08_430070	116	3	the	the	DT
10_1101-2021_02_08_430070	116	4	three	three	CD
10_1101-2021_02_08_430070	116	5	models	model	NNS
10_1101-2021_02_08_430070	116	6	using	use	VBG
10_1101-2021_02_08_430070	116	7	Pytorch	Pytorch	NNP
10_1101-2021_02_08_430070	116	8	.	.	.
10_1101-2021_02_08_430070	117	1	All	all	PDT
10_1101-2021_02_08_430070	117	2	the	the	DT
10_1101-2021_02_08_430070	117	3	models	model	NNS
10_1101-2021_02_08_430070	117	4	are	be	VBP
10_1101-2021_02_08_430070	117	5	optimized	optimize	VBN
10_1101-2021_02_08_430070	117	6	using	use	VBG
10_1101-2021_02_08_430070	117	7	Adam	Adam	NNP
10_1101-2021_02_08_430070	117	8	optimizer	optimizer	NN
10_1101-2021_02_08_430070	117	9	(	(	-LRB-
10_1101-2021_02_08_430070	117	10	Kingma	Kingma	NNP
10_1101-2021_02_08_430070	117	11	and	and	CC
10_1101-2021_02_08_430070	117	12	Ba	Ba	NNP
10_1101-2021_02_08_430070	117	13	,	,	,
10_1101-2021_02_08_430070	117	14	2014	2014	CD
10_1101-2021_02_08_430070	117	15	)	)	-RRB-
10_1101-2021_02_08_430070	117	16	with	with	IN
10_1101-2021_02_08_430070	117	17	the	the	DT
10_1101-2021_02_08_430070	117	18	learning	learning	NN
10_1101-2021_02_08_430070	117	19	rate	rate	NN
10_1101-2021_02_08_430070	117	20	of	of	IN
10_1101-2021_02_08_430070	117	21	1e	1e	CD
10_1101-2021_02_08_430070	117	22	−	−	NNP
10_1101-2021_02_08_430070	117	23	4	4	CD
10_1101-2021_02_08_430070	117	24	and	and	CC
10_1101-2021_02_08_430070	117	25	maximum	maximum	JJ
10_1101-2021_02_08_430070	117	26	iteration	iteration	NN
10_1101-2021_02_08_430070	117	27	epoch	epoch	NN
10_1101-2021_02_08_430070	117	28	of	of	IN
10_1101-2021_02_08_430070	117	29	50	50	CD
10_1101-2021_02_08_430070	117	30	.	.	.
10_1101-2021_02_08_430070	118	1	Model	model	NN
10_1101-2021_02_08_430070	118	2	parameters	parameter	NNS
10_1101-2021_02_08_430070	118	3	are	be	VBP
10_1101-2021_02_08_430070	118	4	selected	select	VBN
10_1101-2021_02_08_430070	118	5	based	base	VBN
10_1101-2021_02_08_430070	118	6	on	on	IN
10_1101-2021_02_08_430070	118	7	the	the	DT
10_1101-2021_02_08_430070	118	8	minimum	minimum	JJ
10_1101-2021_02_08_430070	118	9	validation	validation	NN
10_1101-2021_02_08_430070	118	10	loss	loss	NN
10_1101-2021_02_08_430070	118	11	.	.	.
10_1101-2021_02_08_430070	119	1	3.2	3.2	CD
10_1101-2021_02_08_430070	119	2	Exploring	Exploring	NNP
10_1101-2021_02_08_430070	119	3	differentiated	differentiated	JJ
10_1101-2021_02_08_430070	119	4	signal	signal	NN
10_1101-2021_02_08_430070	119	5	positions	position	NNS
10_1101-2021_02_08_430070	119	6	in	in	IN
10_1101-2021_02_08_430070	119	7	the	the	DT
10_1101-2021_02_08_430070	119	8	context	context	NN
10_1101-2021_02_08_430070	119	9	window	window	NN
10_1101-2021_02_08_430070	119	10	surrounding	surround	VBG
10_1101-2021_02_08_430070	119	11	target	target	NN
10_1101-2021_02_08_430070	119	12	motifs	motif	NNS
10_1101-2021_02_08_430070	119	13	Ideally	ideally	RB
10_1101-2021_02_08_430070	119	14	,	,	,
10_1101-2021_02_08_430070	119	15	we	-PRON-	PRP
10_1101-2021_02_08_430070	119	16	assume	assume	VBP
10_1101-2021_02_08_430070	119	17	a	a	DT
10_1101-2021_02_08_430070	119	18	modified	modify	VBN
10_1101-2021_02_08_430070	119	19	nucleotide	nucleotide	NN
10_1101-2021_02_08_430070	119	20	(	(	-LRB-
10_1101-2021_02_08_430070	119	21	e.g.	e.g.	RB
10_1101-2021_02_08_430070	119	22	,	,	,
10_1101-2021_02_08_430070	119	23	the	the	DT
10_1101-2021_02_08_430070	119	24	center	center	JJ
10_1101-2021_02_08_430070	119	25	position	position	NN
10_1101-2021_02_08_430070	119	26	of	of	IN
10_1101-2021_02_08_430070	119	27	XXXXXXXXXXC5mCGXXXXXXXXX	XXXXXXXXXXC5mCGXXXXXXXXX	NNP
10_1101-2021_02_08_430070	119	28	)	)	-RRB-
10_1101-2021_02_08_430070	119	29	has	have	VBZ
10_1101-2021_02_08_430070	119	30	different	different	JJ
10_1101-2021_02_08_430070	119	31	current	current	JJ
10_1101-2021_02_08_430070	119	32	signals	signal	NNS
10_1101-2021_02_08_430070	119	33	,	,	,
10_1101-2021_02_08_430070	119	34	.license	.license	.
10_1101-2021_02_08_430070	119	35	CC	cc	NN
10_1101-2021_02_08_430070	119	36	-	-	HYPH
10_1101-2021_02_08_430070	119	37	BY	BY	NNP
10_1101-2021_02_08_430070	119	38	-	-	HYPH
10_1101-2021_02_08_430070	119	39	NC	NC	NNP
10_1101-2021_02_08_430070	119	40	-	-	HYPH
10_1101-2021_02_08_430070	119	41	ND	ND	NNP
10_1101-2021_02_08_430070	119	42	4.0	4.0	CD
10_1101-2021_02_08_430070	119	43	Internationalpeer	Internationalpeer	NNP
10_1101-2021_02_08_430070	119	44	review	review	NN
10_1101-2021_02_08_430070	119	45	)	)	-RRB-
10_1101-2021_02_08_430070	119	46	is	be	VBZ
10_1101-2021_02_08_430070	119	47	the	the	DT
10_1101-2021_02_08_430070	119	48	author	author	NN
10_1101-2021_02_08_430070	119	49	/	/	SYM
10_1101-2021_02_08_430070	119	50	funder	funder	NN
10_1101-2021_02_08_430070	119	51	,	,	,
10_1101-2021_02_08_430070	119	52	who	who	WP
10_1101-2021_02_08_430070	119	53	has	have	VBZ
10_1101-2021_02_08_430070	119	54	granted	grant	VBN
10_1101-2021_02_08_430070	119	55	bioRxiv	biorxiv	IN
10_1101-2021_02_08_430070	119	56	a	a	DT
10_1101-2021_02_08_430070	119	57	license	license	NN
10_1101-2021_02_08_430070	119	58	to	to	TO
10_1101-2021_02_08_430070	119	59	display	display	VB
10_1101-2021_02_08_430070	119	60	the	the	DT
10_1101-2021_02_08_430070	119	61	preprint	preprint	NN
10_1101-2021_02_08_430070	119	62	in	in	IN
10_1101-2021_02_08_430070	119	63	perpetuity	perpetuity	NN
10_1101-2021_02_08_430070	119	64	.	.	.
10_1101-2021_02_08_430070	120	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	120	2	is	be	VBZ
10_1101-2021_02_08_430070	120	3	made	make	VBN
10_1101-2021_02_08_430070	120	4	available	available	JJ
10_1101-2021_02_08_430070	120	5	under	under	IN
10_1101-2021_02_08_430070	120	6	a	a	DT
10_1101-2021_02_08_430070	120	7	The	the	DT
10_1101-2021_02_08_430070	120	8	copyright	copyright	NN
10_1101-2021_02_08_430070	120	9	holder	holder	NN
10_1101-2021_02_08_430070	120	10	for	for	IN
10_1101-2021_02_08_430070	120	11	this	this	DT
10_1101-2021_02_08_430070	120	12	preprint	preprint	NN
10_1101-2021_02_08_430070	120	13	(	(	-LRB-
10_1101-2021_02_08_430070	120	14	which	which	WDT
10_1101-2021_02_08_430070	120	15	was	be	VBD
10_1101-2021_02_08_430070	120	16	not	not	RB
10_1101-2021_02_08_430070	120	17	certified	certify	VBN
10_1101-2021_02_08_430070	120	18	bythis	bythis	DT
10_1101-2021_02_08_430070	120	19	version	version	NN
10_1101-2021_02_08_430070	120	20	posted	post	VBD
10_1101-2021_02_08_430070	120	21	February	February	NNP
10_1101-2021_02_08_430070	120	22	10	10	CD
10_1101-2021_02_08_430070	120	23	,	,	,
10_1101-2021_02_08_430070	120	24	2021	2021	CD
10_1101-2021_02_08_430070	120	25	.	.	.
10_1101-2021_02_08_430070	120	26	;	;	:
10_1101-2021_02_08_430070	120	27	https://doi.org/10.1101/2021.02.08.430070doi	https://doi.org/10.1101/2021.02.08.430070doi	NFP
10_1101-2021_02_08_430070	120	28	:	:	:
10_1101-2021_02_08_430070	120	29	bioRxiv	biorxiv	VB
10_1101-2021_02_08_430070	120	30	preprint	preprint	NN
10_1101-2021_02_08_430070	120	31	https://doi.org/10.1101/2021.02.08.430070	https://doi.org/10.1101/2021.02.08.430070	UH
10_1101-2021_02_08_430070	120	32	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	120	33	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	120	34	✐	✐	NNP
10_1101-2021_02_08_430070	120	35	✐	✐	NNP
10_1101-2021_02_08_430070	120	36	✐	✐	NNP
10_1101-2021_02_08_430070	120	37	✐	✐	NNP
10_1101-2021_02_08_430070	120	38	✐	✐	NNP
10_1101-2021_02_08_430070	120	39	✐	✐	NNP
10_1101-2021_02_08_430070	120	40	✐	✐	NNP
10_1101-2021_02_08_430070	120	41	✐	✐	NNP
10_1101-2021_02_08_430070	120	42	4	4	CD
10_1101-2021_02_08_430070	120	43	Zhang	Zhang	NNP
10_1101-2021_02_08_430070	120	44	et	et	NNP
10_1101-2021_02_08_430070	120	45	al	al	NNP
10_1101-2021_02_08_430070	120	46	.	.	.
10_1101-2021_02_08_430070	121	1	(	(	-LRB-
10_1101-2021_02_08_430070	121	2	a1	a1	NNP
10_1101-2021_02_08_430070	121	3	)	)	-RRB-
10_1101-2021_02_08_430070	121	4	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	121	5	-	-	HYPH
10_1101-2021_02_08_430070	121	6	E.coli_Cg_SssI	E.coli_Cg_SssI	NNP
10_1101-2021_02_08_430070	121	7	(	(	-LRB-
10_1101-2021_02_08_430070	121	8	a2	a2	NNP
10_1101-2021_02_08_430070	121	9	)	)	-RRB-
10_1101-2021_02_08_430070	121	10	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	121	11	-	-	HYPH
10_1101-2021_02_08_430070	121	12	E.coli_Cg_MpeI	E.coli_Cg_MpeI	NNP
10_1101-2021_02_08_430070	121	13	(	(	-LRB-
10_1101-2021_02_08_430070	121	14	a3	a3	NNP
10_1101-2021_02_08_430070	121	15	)	)	-RRB-
10_1101-2021_02_08_430070	121	16	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	121	17	-	-	HYPH
10_1101-2021_02_08_430070	121	18	E.coli_gCgc_Hhal	e.coli_gcgc_hhal	NN
10_1101-2021_02_08_430070	121	19	(	(	-LRB-
10_1101-2021_02_08_430070	121	20	b1	b1	NN
10_1101-2021_02_08_430070	121	21	)	)	-RRB-
10_1101-2021_02_08_430070	121	22	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	121	23	-	-	HYPH
10_1101-2021_02_08_430070	121	24	E.coli_Cg_SssI	E.coli_Cg_SssI	NNP
10_1101-2021_02_08_430070	121	25	(	(	-LRB-
10_1101-2021_02_08_430070	121	26	b2	b2	NN
10_1101-2021_02_08_430070	121	27	)	)	-RRB-
10_1101-2021_02_08_430070	121	28	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	121	29	-	-	HYPH
10_1101-2021_02_08_430070	121	30	H.Sapiens_Cg_SssI	H.Sapiens_Cg_SssI	NNP
10_1101-2021_02_08_430070	121	31	(	(	-LRB-
10_1101-2021_02_08_430070	121	32	c1	c1	NN
10_1101-2021_02_08_430070	121	33	)	)	-RRB-
10_1101-2021_02_08_430070	121	34	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	121	35	-	-	HYPH
10_1101-2021_02_08_430070	121	36	E.coli_gaAttc_EcoRI	E.coli_gaAttc_EcoRI	NNP
10_1101-2021_02_08_430070	121	37	(	(	-LRB-
10_1101-2021_02_08_430070	121	38	c2	c2	NN
10_1101-2021_02_08_430070	121	39	)	)	-RRB-
10_1101-2021_02_08_430070	121	40	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	121	41	-	-	HYPH
10_1101-2021_02_08_430070	121	42	E.coli_tcgA_TaqI	E.coli_tcgA_TaqI	NNP
10_1101-2021_02_08_430070	121	43	(	(	-LRB-
10_1101-2021_02_08_430070	121	44	c3	c3	NN
10_1101-2021_02_08_430070	121	45	)	)	-RRB-
10_1101-2021_02_08_430070	121	46	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	121	47	-	-	HYPH
10_1101-2021_02_08_430070	121	48	E.coli_gAtc_Dam	E.coli_gAtc_Dam	NNP
10_1101-2021_02_08_430070	121	49	Fig	Fig	NNP
10_1101-2021_02_08_430070	121	50	.	.	.
10_1101-2021_02_08_430070	122	1	2	2	LS
10_1101-2021_02_08_430070	122	2	:	:	:
10_1101-2021_02_08_430070	122	3	Boxplot	boxplot	NN
10_1101-2021_02_08_430070	122	4	of	of	IN
10_1101-2021_02_08_430070	122	5	positional	positional	JJ
10_1101-2021_02_08_430070	122	6	signal	signal	NN
10_1101-2021_02_08_430070	122	7	-	-	HYPH
10_1101-2021_02_08_430070	122	8	shift	shift	NN
10_1101-2021_02_08_430070	122	9	for	for	IN
10_1101-2021_02_08_430070	122	10	5mC	5mc	CD
10_1101-2021_02_08_430070	122	11	and	and	CC
10_1101-2021_02_08_430070	122	12	6mA	6ma	CD
10_1101-2021_02_08_430070	122	13	datasets	dataset	NNS
10_1101-2021_02_08_430070	122	14	of	of	IN
10_1101-2021_02_08_430070	122	15	the	the	DT
10_1101-2021_02_08_430070	122	16	specific	specific	JJ
10_1101-2021_02_08_430070	122	17	motif	motif	NN
10_1101-2021_02_08_430070	122	18	and	and	CC
10_1101-2021_02_08_430070	122	19	methyltransferase	methyltransferase	NN
10_1101-2021_02_08_430070	122	20	.	.	.
10_1101-2021_02_08_430070	123	1	(	(	-LRB-
10_1101-2021_02_08_430070	123	2	a1),(a2	a1),(a2	NNP
10_1101-2021_02_08_430070	123	3	)	)	-RRB-
10_1101-2021_02_08_430070	123	4	and	and	CC
10_1101-2021_02_08_430070	123	5	(	(	-LRB-
10_1101-2021_02_08_430070	123	6	a3	a3	NNP
10_1101-2021_02_08_430070	123	7	)	)	-RRB-
10_1101-2021_02_08_430070	123	8	are	be	VBP
10_1101-2021_02_08_430070	123	9	on	on	IN
10_1101-2021_02_08_430070	123	10	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	123	11	’s	’s	POS
10_1101-2021_02_08_430070	123	12	E.coli	e.coli	JJ
10_1101-2021_02_08_430070	123	13	5mC	5mc	CD
10_1101-2021_02_08_430070	123	14	dataset	dataset	NN
10_1101-2021_02_08_430070	123	15	.	.	.
10_1101-2021_02_08_430070	124	1	(	(	-LRB-
10_1101-2021_02_08_430070	124	2	b1	b1	NN
10_1101-2021_02_08_430070	124	3	)	)	-RRB-
10_1101-2021_02_08_430070	124	4	and	and	CC
10_1101-2021_02_08_430070	124	5	(	(	-LRB-
10_1101-2021_02_08_430070	124	6	b2	b2	NN
10_1101-2021_02_08_430070	124	7	)	)	-RRB-
10_1101-2021_02_08_430070	124	8	are	be	VBP
10_1101-2021_02_08_430070	124	9	on	on	IN
10_1101-2021_02_08_430070	124	10	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	124	11	’s	’s	NNP
10_1101-2021_02_08_430070	124	12	5mC	5mc	CD
10_1101-2021_02_08_430070	124	13	dataset	dataset	NN
10_1101-2021_02_08_430070	124	14	.	.	.
10_1101-2021_02_08_430070	125	1	(	(	-LRB-
10_1101-2021_02_08_430070	125	2	c1	c1	NN
10_1101-2021_02_08_430070	125	3	)	)	-RRB-
10_1101-2021_02_08_430070	125	4	,	,	,
10_1101-2021_02_08_430070	125	5	(	(	-LRB-
10_1101-2021_02_08_430070	125	6	c2	c2	NNP
10_1101-2021_02_08_430070	125	7	)	)	-RRB-
10_1101-2021_02_08_430070	125	8	and	and	CC
10_1101-2021_02_08_430070	125	9	(	(	-LRB-
10_1101-2021_02_08_430070	125	10	c3	c3	NN
10_1101-2021_02_08_430070	125	11	)	)	-RRB-
10_1101-2021_02_08_430070	125	12	are	be	VBP
10_1101-2021_02_08_430070	125	13	on	on	IN
10_1101-2021_02_08_430070	125	14	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	125	15	’s	’s	POS
10_1101-2021_02_08_430070	125	16	E.coli	e.coli	JJ
10_1101-2021_02_08_430070	125	17	6mA	6ma	CD
10_1101-2021_02_08_430070	125	18	dataset	dataset	NN
10_1101-2021_02_08_430070	125	19	.	.	.
10_1101-2021_02_08_430070	126	1	Each	each	DT
10_1101-2021_02_08_430070	126	2	dataset	dataset	NN
10_1101-2021_02_08_430070	126	3	is	be	VBZ
10_1101-2021_02_08_430070	126	4	represented	represent	VBN
10_1101-2021_02_08_430070	126	5	in	in	IN
10_1101-2021_02_08_430070	126	6	a	a	DT
10_1101-2021_02_08_430070	126	7	format	format	NN
10_1101-2021_02_08_430070	126	8	of	of	IN
10_1101-2021_02_08_430070	126	9	dataSource_motif_methltansferase	datasource_motif_methltansferase	ADD
10_1101-2021_02_08_430070	126	10	.	.	.
10_1101-2021_02_08_430070	127	1	when	when	WRB
10_1101-2021_02_08_430070	127	2	compared	compare	VBN
10_1101-2021_02_08_430070	127	3	with	with	IN
10_1101-2021_02_08_430070	127	4	the	the	DT
10_1101-2021_02_08_430070	127	5	unmodified	unmodified	JJ
10_1101-2021_02_08_430070	127	6	one	one	NN
10_1101-2021_02_08_430070	127	7	.	.	.
10_1101-2021_02_08_430070	128	1	As	as	IN
10_1101-2021_02_08_430070	128	2	the	the	DT
10_1101-2021_02_08_430070	128	3	boundary	boundary	NN
10_1101-2021_02_08_430070	128	4	of	of	IN
10_1101-2021_02_08_430070	128	5	nucleotide	nucleotide	NNP
10_1101-2021_02_08_430070	128	6	/	/	SYM
10_1101-2021_02_08_430070	128	7	k-	k-	NNP
10_1101-2021_02_08_430070	128	8	mer	mer	NNP
10_1101-2021_02_08_430070	128	9	signals	signal	NNS
10_1101-2021_02_08_430070	128	10	are	be	VBP
10_1101-2021_02_08_430070	128	11	not	not	RB
10_1101-2021_02_08_430070	128	12	rigorous	rigorous	JJ
10_1101-2021_02_08_430070	128	13	and	and	CC
10_1101-2021_02_08_430070	128	14	surrounding	surround	VBG
10_1101-2021_02_08_430070	128	15	nucleotides	nucleotide	NNS
10_1101-2021_02_08_430070	128	16	may	may	MD
10_1101-2021_02_08_430070	128	17	also	also	RB
10_1101-2021_02_08_430070	128	18	be	be	VB
10_1101-2021_02_08_430070	128	19	affected	affect	VBN
10_1101-2021_02_08_430070	128	20	,	,	,
10_1101-2021_02_08_430070	128	21	it	-PRON-	PRP
10_1101-2021_02_08_430070	128	22	is	be	VBZ
10_1101-2021_02_08_430070	128	23	worthwhile	worthwhile	JJ
10_1101-2021_02_08_430070	128	24	investigating	investigate	VBG
10_1101-2021_02_08_430070	128	25	signal	signal	JJ
10_1101-2021_02_08_430070	128	26	-	-	HYPH
10_1101-2021_02_08_430070	128	27	shift	shift	NN
10_1101-2021_02_08_430070	128	28	patterns	pattern	NNS
10_1101-2021_02_08_430070	128	29	related	relate	VBN
10_1101-2021_02_08_430070	128	30	to	to	IN
10_1101-2021_02_08_430070	128	31	methylation	methylation	NN
10_1101-2021_02_08_430070	128	32	in	in	IN
10_1101-2021_02_08_430070	128	33	a	a	DT
10_1101-2021_02_08_430070	128	34	large	large	JJ
10_1101-2021_02_08_430070	128	35	context	context	NN
10_1101-2021_02_08_430070	128	36	.	.	.
10_1101-2021_02_08_430070	129	1	To	to	TO
10_1101-2021_02_08_430070	129	2	identify	identify	VB
10_1101-2021_02_08_430070	129	3	signal	signal	NN
10_1101-2021_02_08_430070	129	4	-	-	HYPH
10_1101-2021_02_08_430070	129	5	shift	shift	NN
10_1101-2021_02_08_430070	129	6	affected	affect	VBN
10_1101-2021_02_08_430070	129	7	by	by	IN
10_1101-2021_02_08_430070	129	8	methylation	methylation	NN
10_1101-2021_02_08_430070	129	9	for	for	IN
10_1101-2021_02_08_430070	129	10	a	a	DT
10_1101-2021_02_08_430070	129	11	specific	specific	JJ
10_1101-2021_02_08_430070	129	12	dataset	dataset	NN
10_1101-2021_02_08_430070	129	13	,	,	,
10_1101-2021_02_08_430070	129	14	we	-PRON-	PRP
10_1101-2021_02_08_430070	129	15	use	use	VBP
10_1101-2021_02_08_430070	129	16	a	a	DT
10_1101-2021_02_08_430070	129	17	simple	simple	JJ
10_1101-2021_02_08_430070	129	18	quantification	quantification	NN
10_1101-2021_02_08_430070	129	19	approach	approach	NN
10_1101-2021_02_08_430070	129	20	to	to	TO
10_1101-2021_02_08_430070	129	21	calculate	calculate	VB
10_1101-2021_02_08_430070	129	22	significant	significant	JJ
10_1101-2021_02_08_430070	129	23	signal	signal	NN
10_1101-2021_02_08_430070	129	24	changes	change	NNS
10_1101-2021_02_08_430070	129	25	of	of	IN
10_1101-2021_02_08_430070	129	26	each	each	DT
10_1101-2021_02_08_430070	129	27	position	position	NN
10_1101-2021_02_08_430070	129	28	in	in	IN
10_1101-2021_02_08_430070	129	29	the	the	DT
10_1101-2021_02_08_430070	129	30	context	context	NN
10_1101-2021_02_08_430070	129	31	window	window	NN
10_1101-2021_02_08_430070	129	32	.	.	.
10_1101-2021_02_08_430070	130	1	Given	give	VBN
10_1101-2021_02_08_430070	130	2	a	a	DT
10_1101-2021_02_08_430070	130	3	dataset	dataset	NN
10_1101-2021_02_08_430070	130	4	of	of	IN
10_1101-2021_02_08_430070	130	5	a	a	DT
10_1101-2021_02_08_430070	130	6	specific	specific	JJ
10_1101-2021_02_08_430070	130	7	motif	motif	NN
10_1101-2021_02_08_430070	130	8	and	and	CC
10_1101-2021_02_08_430070	130	9	methyltransferase	methyltransferase	NN
10_1101-2021_02_08_430070	130	10	,	,	,
10_1101-2021_02_08_430070	130	11	we	-PRON-	PRP
10_1101-2021_02_08_430070	130	12	first	first	RB
10_1101-2021_02_08_430070	130	13	cluster	cluster	VBD
10_1101-2021_02_08_430070	130	14	instances	instance	NNS
10_1101-2021_02_08_430070	130	15	with	with	IN
10_1101-2021_02_08_430070	130	16	the	the	DT
10_1101-2021_02_08_430070	130	17	same	same	JJ
10_1101-2021_02_08_430070	130	18	nucleotide	nucleotide	JJ
10_1101-2021_02_08_430070	130	19	sequence	sequence	NN
10_1101-2021_02_08_430070	130	20	to	to	TO
10_1101-2021_02_08_430070	130	21	avoid	avoid	VB
10_1101-2021_02_08_430070	130	22	the	the	DT
10_1101-2021_02_08_430070	130	23	effect	effect	NN
10_1101-2021_02_08_430070	130	24	of	of	IN
10_1101-2021_02_08_430070	130	25	nucleotide	nucleotide	JJ
10_1101-2021_02_08_430070	130	26	sequences	sequence	NNS
10_1101-2021_02_08_430070	130	27	.	.	.
10_1101-2021_02_08_430070	131	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	131	2	reserve	reserve	VBP
10_1101-2021_02_08_430070	131	3	sequence	sequence	NN
10_1101-2021_02_08_430070	131	4	clusters	cluster	NNS
10_1101-2021_02_08_430070	131	5	that	that	WDT
10_1101-2021_02_08_430070	131	6	contain	contain	VBP
10_1101-2021_02_08_430070	131	7	both	both	DT
10_1101-2021_02_08_430070	131	8	methylation	methylation	NN
10_1101-2021_02_08_430070	131	9	and	and	CC
10_1101-2021_02_08_430070	131	10	unmethylation	unmethylation	NN
10_1101-2021_02_08_430070	131	11	instances	instance	NNS
10_1101-2021_02_08_430070	131	12	(	(	-LRB-
10_1101-2021_02_08_430070	131	13	≥	≥	UH
10_1101-2021_02_08_430070	131	14	1	1	CD
10_1101-2021_02_08_430070	131	15	)	)	-RRB-
10_1101-2021_02_08_430070	131	16	.	.	.
10_1101-2021_02_08_430070	132	1	For	for	IN
10_1101-2021_02_08_430070	132	2	each	each	DT
10_1101-2021_02_08_430070	132	3	sequence	sequence	NN
10_1101-2021_02_08_430070	132	4	cluster	cluster	NN
10_1101-2021_02_08_430070	132	5	,	,	,
10_1101-2021_02_08_430070	132	6	we	-PRON-	PRP
10_1101-2021_02_08_430070	132	7	normalize	normalize	VBP
10_1101-2021_02_08_430070	132	8	event	event	NN
10_1101-2021_02_08_430070	132	9	signal	signal	NN
10_1101-2021_02_08_430070	132	10	values	value	NNS
10_1101-2021_02_08_430070	132	11	of	of	IN
10_1101-2021_02_08_430070	132	12	methylation	methylation	NN
10_1101-2021_02_08_430070	132	13	samples	sample	NNS
10_1101-2021_02_08_430070	132	14	with	with	IN
10_1101-2021_02_08_430070	132	15	their	-PRON-	PRP$
10_1101-2021_02_08_430070	132	16	according	accord	VBG
10_1101-2021_02_08_430070	132	17	unmodified	unmodified	JJ
10_1101-2021_02_08_430070	132	18	averaged	averaged	JJ
10_1101-2021_02_08_430070	132	19	event	event	NN
10_1101-2021_02_08_430070	132	20	signal	signal	NN
10_1101-2021_02_08_430070	132	21	values	value	NNS
10_1101-2021_02_08_430070	132	22	for	for	IN
10_1101-2021_02_08_430070	132	23	each	each	DT
10_1101-2021_02_08_430070	132	24	position	position	NN
10_1101-2021_02_08_430070	132	25	.	.	.
10_1101-2021_02_08_430070	133	1	The	the	DT
10_1101-2021_02_08_430070	133	2	i	i	NNP
10_1101-2021_02_08_430070	133	3	-	-	HYPH
10_1101-2021_02_08_430070	133	4	th	th	NNP
10_1101-2021_02_08_430070	133	5	positional	positional	JJ
10_1101-2021_02_08_430070	133	6	signal	signal	JJ
10_1101-2021_02_08_430070	133	7	-	-	HYPH
10_1101-2021_02_08_430070	133	8	shift	shift	NN
10_1101-2021_02_08_430070	133	9	is	be	VBZ
10_1101-2021_02_08_430070	133	10	then	then	RB
10_1101-2021_02_08_430070	133	11	calculated	calculate	VBN
10_1101-2021_02_08_430070	133	12	as	as	IN
10_1101-2021_02_08_430070	133	13	smethi	smethi	NN
10_1101-2021_02_08_430070	133	14	−	−	NNP
10_1101-2021_02_08_430070	133	15	avg(s	avg(s	UH
10_1101-2021_02_08_430070	133	16	unmeth	unmeth	JJ
10_1101-2021_02_08_430070	133	17	i	i	PRP
10_1101-2021_02_08_430070	133	18	)	)	-RRB-
10_1101-2021_02_08_430070	133	19	.	.	.
10_1101-2021_02_08_430070	134	1	For	for	IN
10_1101-2021_02_08_430070	134	2	those	those	DT
10_1101-2021_02_08_430070	134	3	normalized	normalized	JJ
10_1101-2021_02_08_430070	134	4	methylation	methylation	NN
10_1101-2021_02_08_430070	134	5	samples	sample	NNS
10_1101-2021_02_08_430070	134	6	,	,	,
10_1101-2021_02_08_430070	134	7	we	-PRON-	PRP
10_1101-2021_02_08_430070	134	8	calculate	calculate	VBP
10_1101-2021_02_08_430070	134	9	basic	basic	JJ
10_1101-2021_02_08_430070	134	10	statistics	statistic	NNS
10_1101-2021_02_08_430070	134	11	of	of	IN
10_1101-2021_02_08_430070	134	12	signal	signal	NN
10_1101-2021_02_08_430070	134	13	-	-	HYPH
10_1101-2021_02_08_430070	134	14	shift	shift	NN
10_1101-2021_02_08_430070	134	15	for	for	IN
10_1101-2021_02_08_430070	134	16	each	each	DT
10_1101-2021_02_08_430070	134	17	position	position	NN
10_1101-2021_02_08_430070	134	18	and	and	CC
10_1101-2021_02_08_430070	134	19	draw	draw	VB
10_1101-2021_02_08_430070	134	20	boxplots	boxplot	NNS
10_1101-2021_02_08_430070	134	21	for	for	IN
10_1101-2021_02_08_430070	134	22	5mC	5mc	CD
10_1101-2021_02_08_430070	134	23	and	and	CC
10_1101-2021_02_08_430070	134	24	6mA	6ma	CD
10_1101-2021_02_08_430070	134	25	training	training	NN
10_1101-2021_02_08_430070	134	26	sets	set	NNS
10_1101-2021_02_08_430070	134	27	.	.	.
10_1101-2021_02_08_430070	135	1	Shown	show	VBN
10_1101-2021_02_08_430070	135	2	in	in	IN
10_1101-2021_02_08_430070	135	3	Figure	Figure	NNP
10_1101-2021_02_08_430070	135	4	2	2	CD
10_1101-2021_02_08_430070	135	5	,	,	,
10_1101-2021_02_08_430070	135	6	for	for	IN
10_1101-2021_02_08_430070	135	7	all	all	DT
10_1101-2021_02_08_430070	135	8	datasets	dataset	NNS
10_1101-2021_02_08_430070	135	9	,	,	,
10_1101-2021_02_08_430070	135	10	we	-PRON-	PRP
10_1101-2021_02_08_430070	135	11	can	can	MD
10_1101-2021_02_08_430070	135	12	observed	observe	VBN
10_1101-2021_02_08_430070	135	13	positions	position	NNS
10_1101-2021_02_08_430070	135	14	of	of	IN
10_1101-2021_02_08_430070	135	15	significantly	significantly	RB
10_1101-2021_02_08_430070	135	16	signal	signal	NN
10_1101-2021_02_08_430070	135	17	-	-	HYPH
10_1101-2021_02_08_430070	135	18	shift	shift	NN
10_1101-2021_02_08_430070	135	19	are	be	VBP
10_1101-2021_02_08_430070	135	20	located	locate	VBN
10_1101-2021_02_08_430070	135	21	in	in	IN
10_1101-2021_02_08_430070	135	22	a	a	DT
10_1101-2021_02_08_430070	135	23	range	range	NN
10_1101-2021_02_08_430070	135	24	of	of	IN
10_1101-2021_02_08_430070	135	25	±3bp	±3bp	NN
10_1101-2021_02_08_430070	135	26	to	to	IN
10_1101-2021_02_08_430070	135	27	the	the	DT
10_1101-2021_02_08_430070	135	28	center	center	JJ
10_1101-2021_02_08_430070	135	29	position	position	NN
10_1101-2021_02_08_430070	135	30	(	(	-LRB-
10_1101-2021_02_08_430070	135	31	the	the	DT
10_1101-2021_02_08_430070	135	32	11th	11th	NN
10_1101-2021_02_08_430070	135	33	)	)	-RRB-
10_1101-2021_02_08_430070	135	34	in	in	IN
10_1101-2021_02_08_430070	135	35	which	which	WDT
10_1101-2021_02_08_430070	135	36	the	the	DT
10_1101-2021_02_08_430070	135	37	target	target	NN
10_1101-2021_02_08_430070	135	38	nucleotide	nucleotide	RB
10_1101-2021_02_08_430070	135	39	is	be	VBZ
10_1101-2021_02_08_430070	135	40	located	locate	VBN
10_1101-2021_02_08_430070	135	41	.	.	.
10_1101-2021_02_08_430070	136	1	For	for	IN
10_1101-2021_02_08_430070	136	2	the	the	DT
10_1101-2021_02_08_430070	136	3	rest	rest	NN
10_1101-2021_02_08_430070	136	4	off	off	IN
10_1101-2021_02_08_430070	136	5	-	-	HYPH
10_1101-2021_02_08_430070	136	6	center	center	NN
10_1101-2021_02_08_430070	136	7	positions	position	NNS
10_1101-2021_02_08_430070	136	8	,	,	,
10_1101-2021_02_08_430070	136	9	the	the	DT
10_1101-2021_02_08_430070	136	10	averaged	average	VBN
10_1101-2021_02_08_430070	136	11	signal	signal	JJ
10_1101-2021_02_08_430070	136	12	-	-	HYPH
10_1101-2021_02_08_430070	136	13	shift	shift	NN
10_1101-2021_02_08_430070	136	14	values	value	NNS
10_1101-2021_02_08_430070	136	15	are	be	VBP
10_1101-2021_02_08_430070	136	16	close	close	JJ
10_1101-2021_02_08_430070	136	17	to	to	IN
10_1101-2021_02_08_430070	136	18	0	0	CD
10_1101-2021_02_08_430070	136	19	.	.	.
10_1101-2021_02_08_430070	137	1	This	this	DT
10_1101-2021_02_08_430070	137	2	indicates	indicate	VBZ
10_1101-2021_02_08_430070	137	3	a	a	DT
10_1101-2021_02_08_430070	137	4	modified	modify	VBN
10_1101-2021_02_08_430070	137	5	nucleotide	nucleotide	RB
10_1101-2021_02_08_430070	137	6	not	not	RB
10_1101-2021_02_08_430070	137	7	only	only	RB
10_1101-2021_02_08_430070	137	8	affect	affect	VB
10_1101-2021_02_08_430070	137	9	its	-PRON-	PRP$
10_1101-2021_02_08_430070	137	10	corresponding	correspond	VBG
10_1101-2021_02_08_430070	137	11	current	current	JJ
10_1101-2021_02_08_430070	137	12	signals	signal	NNS
10_1101-2021_02_08_430070	137	13	but	but	CC
10_1101-2021_02_08_430070	137	14	also	also	RB
10_1101-2021_02_08_430070	137	15	the	the	DT
10_1101-2021_02_08_430070	137	16	signals	signal	NNS
10_1101-2021_02_08_430070	137	17	of	of	IN
10_1101-2021_02_08_430070	137	18	its	-PRON-	PRP$
10_1101-2021_02_08_430070	137	19	surrounding	surround	VBG
10_1101-2021_02_08_430070	137	20	nucleotides	nucleotide	NNS
10_1101-2021_02_08_430070	137	21	.	.	.
10_1101-2021_02_08_430070	138	1	Besides	besides	RB
10_1101-2021_02_08_430070	138	2	,	,	,
10_1101-2021_02_08_430070	138	3	5mC	5mc	CD
10_1101-2021_02_08_430070	138	4	and	and	CC
10_1101-2021_02_08_430070	138	5	6mA	6mA	NNP
10_1101-2021_02_08_430070	138	6	datasets	dataset	NNS
10_1101-2021_02_08_430070	138	7	show	show	VBP
10_1101-2021_02_08_430070	138	8	different	different	JJ
10_1101-2021_02_08_430070	138	9	positional	positional	JJ
10_1101-2021_02_08_430070	138	10	-	-	HYPH
10_1101-2021_02_08_430070	138	11	signal	signal	NN
10_1101-2021_02_08_430070	138	12	-	-	HYPH
10_1101-2021_02_08_430070	138	13	shift	shift	NN
10_1101-2021_02_08_430070	138	14	patterns	pattern	NNS
10_1101-2021_02_08_430070	138	15	.	.	.
10_1101-2021_02_08_430070	139	1	Specific	specific	JJ
10_1101-2021_02_08_430070	139	2	positions	position	NNS
10_1101-2021_02_08_430070	139	3	,	,	,
10_1101-2021_02_08_430070	139	4	such	such	JJ
10_1101-2021_02_08_430070	139	5	as	as	IN
10_1101-2021_02_08_430070	139	6	-2bp	-2bp	NN
10_1101-2021_02_08_430070	139	7	position	position	NN
10_1101-2021_02_08_430070	139	8	(	(	-LRB-
10_1101-2021_02_08_430070	139	9	9th	9th	NN
10_1101-2021_02_08_430070	139	10	)	)	-RRB-
10_1101-2021_02_08_430070	139	11	in	in	IN
10_1101-2021_02_08_430070	139	12	the	the	DT
10_1101-2021_02_08_430070	139	13	5mC	5mc	CD
10_1101-2021_02_08_430070	139	14	dataset	dataset	NN
10_1101-2021_02_08_430070	139	15	and	and	CC
10_1101-2021_02_08_430070	139	16	+1bp	+1bp	NN
10_1101-2021_02_08_430070	139	17	position	position	NN
10_1101-2021_02_08_430070	139	18	(	(	-LRB-
10_1101-2021_02_08_430070	139	19	12th	12th	NN
10_1101-2021_02_08_430070	139	20	)	)	-RRB-
10_1101-2021_02_08_430070	139	21	in	in	IN
10_1101-2021_02_08_430070	139	22	the	the	DT
10_1101-2021_02_08_430070	139	23	6mA	6ma	CD
10_1101-2021_02_08_430070	139	24	dataset	dataset	NN
10_1101-2021_02_08_430070	139	25	,	,	,
10_1101-2021_02_08_430070	139	26	have	have	VBP
10_1101-2021_02_08_430070	139	27	larger	large	JJR
10_1101-2021_02_08_430070	139	28	averaged	average	VBN
10_1101-2021_02_08_430070	139	29	signal-	signal-	NN
10_1101-2021_02_08_430070	139	30	shift	shift	NN
10_1101-2021_02_08_430070	139	31	values	value	NNS
10_1101-2021_02_08_430070	139	32	.	.	.
10_1101-2021_02_08_430070	140	1	Such	such	JJ
10_1101-2021_02_08_430070	140	2	pattern	pattern	NN
10_1101-2021_02_08_430070	140	3	can	can	MD
10_1101-2021_02_08_430070	140	4	be	be	VB
10_1101-2021_02_08_430070	140	5	generalized	generalize	VBN
10_1101-2021_02_08_430070	140	6	across	across	IN
10_1101-2021_02_08_430070	140	7	the	the	DT
10_1101-2021_02_08_430070	140	8	different	different	JJ
10_1101-2021_02_08_430070	140	9	dataset	dataset	NN
10_1101-2021_02_08_430070	140	10	with	with	IN
10_1101-2021_02_08_430070	140	11	the	the	DT
10_1101-2021_02_08_430070	140	12	same	same	JJ
10_1101-2021_02_08_430070	140	13	motif	motif	NN
10_1101-2021_02_08_430070	140	14	and	and	CC
10_1101-2021_02_08_430070	140	15	methyltransferase	methyltransferase	NN
10_1101-2021_02_08_430070	140	16	.	.	.
10_1101-2021_02_08_430070	141	1	For	for	IN
10_1101-2021_02_08_430070	141	2	example	example	NN
10_1101-2021_02_08_430070	141	3	,	,	,
10_1101-2021_02_08_430070	141	4	Figure	figure	NN
10_1101-2021_02_08_430070	141	5	2	2	CD
10_1101-2021_02_08_430070	141	6	(	(	-LRB-
10_1101-2021_02_08_430070	141	7	a1	a1	NNP
10_1101-2021_02_08_430070	141	8	)	)	-RRB-
10_1101-2021_02_08_430070	141	9	,	,	,
10_1101-2021_02_08_430070	141	10	(	(	-LRB-
10_1101-2021_02_08_430070	141	11	b1	b1	NN
10_1101-2021_02_08_430070	141	12	)	)	-RRB-
10_1101-2021_02_08_430070	141	13	and	and	CC
10_1101-2021_02_08_430070	141	14	(	(	-LRB-
10_1101-2021_02_08_430070	141	15	b2	b2	NN
10_1101-2021_02_08_430070	141	16	)	)	-RRB-
10_1101-2021_02_08_430070	141	17	show	show	VBP
10_1101-2021_02_08_430070	141	18	a	a	DT
10_1101-2021_02_08_430070	141	19	similar	similar	JJ
10_1101-2021_02_08_430070	141	20	positional	positional	JJ
10_1101-2021_02_08_430070	141	21	signal	signal	JJ
10_1101-2021_02_08_430070	141	22	-	-	HYPH
10_1101-2021_02_08_430070	141	23	shift	shift	NN
10_1101-2021_02_08_430070	141	24	pattern	pattern	NN
10_1101-2021_02_08_430070	141	25	.	.	.
10_1101-2021_02_08_430070	142	1	For	for	IN
10_1101-2021_02_08_430070	142	2	different	different	JJ
10_1101-2021_02_08_430070	142	3	methyltransferases	methyltransferase	NNS
10_1101-2021_02_08_430070	142	4	,	,	,
10_1101-2021_02_08_430070	142	5	such	such	JJ
10_1101-2021_02_08_430070	142	6	as	as	IN
10_1101-2021_02_08_430070	142	7	Hhal	Hhal	NNP
10_1101-2021_02_08_430070	142	8	(	(	-LRB-
10_1101-2021_02_08_430070	142	9	Figure	Figure	NNP
10_1101-2021_02_08_430070	142	10	2(a3	2(a3	NNP
10_1101-2021_02_08_430070	142	11	)	)	-RRB-
10_1101-2021_02_08_430070	142	12	)	)	-RRB-
10_1101-2021_02_08_430070	142	13	also	also	RB
10_1101-2021_02_08_430070	142	14	shows	show	VBZ
10_1101-2021_02_08_430070	142	15	a	a	DT
10_1101-2021_02_08_430070	142	16	similar	similar	JJ
10_1101-2021_02_08_430070	142	17	pattern	pattern	NN
10_1101-2021_02_08_430070	142	18	as	as	IN
10_1101-2021_02_08_430070	142	19	in	in	IN
10_1101-2021_02_08_430070	142	20	SssI	SssI	NNP
10_1101-2021_02_08_430070	142	21	,	,	,
10_1101-2021_02_08_430070	142	22	while	while	IN
10_1101-2021_02_08_430070	142	23	MpeI	MpeI	NNP
10_1101-2021_02_08_430070	142	24	does	do	VBZ
10_1101-2021_02_08_430070	142	25	not	not	RB
10_1101-2021_02_08_430070	142	26	have	have	VB
10_1101-2021_02_08_430070	142	27	a	a	DT
10_1101-2021_02_08_430070	142	28	similar	similar	JJ
10_1101-2021_02_08_430070	142	29	pattern	pattern	NN
10_1101-2021_02_08_430070	142	30	obviously	obviously	RB
10_1101-2021_02_08_430070	142	31	(	(	-LRB-
10_1101-2021_02_08_430070	142	32	Figure	figure	NN
10_1101-2021_02_08_430070	142	33	2(a2	2(a2	NNP
10_1101-2021_02_08_430070	142	34	)	)	-RRB-
10_1101-2021_02_08_430070	142	35	)	)	-RRB-
10_1101-2021_02_08_430070	142	36	.	.	.
10_1101-2021_02_08_430070	143	1	Those	those	DT
10_1101-2021_02_08_430070	143	2	positional	positional	JJ
10_1101-2021_02_08_430070	143	3	signal	signal	JJ
10_1101-2021_02_08_430070	143	4	patterns	pattern	NNS
10_1101-2021_02_08_430070	143	5	can	can	MD
10_1101-2021_02_08_430070	143	6	be	be	VB
10_1101-2021_02_08_430070	143	7	directly	directly	RB
10_1101-2021_02_08_430070	143	8	modeled	model	VBN
10_1101-2021_02_08_430070	143	9	by	by	IN
10_1101-2021_02_08_430070	143	10	a	a	DT
10_1101-2021_02_08_430070	143	11	biRNN	birnn	NN
10_1101-2021_02_08_430070	143	12	,	,	,
10_1101-2021_02_08_430070	143	13	while	while	IN
10_1101-2021_02_08_430070	143	14	for	for	IN
10_1101-2021_02_08_430070	143	15	the	the	DT
10_1101-2021_02_08_430070	143	16	basic	basic	JJ
10_1101-2021_02_08_430070	143	17	BERT	BERT	NNP
10_1101-2021_02_08_430070	143	18	,	,	,
10_1101-2021_02_08_430070	143	19	they	-PRON-	PRP
10_1101-2021_02_08_430070	143	20	are	be	VBP
10_1101-2021_02_08_430070	143	21	not	not	RB
10_1101-2021_02_08_430070	143	22	specifically	specifically	RB
10_1101-2021_02_08_430070	143	23	considered	consider	VBN
10_1101-2021_02_08_430070	143	24	in	in	IN
10_1101-2021_02_08_430070	143	25	its	-PRON-	PRP$
10_1101-2021_02_08_430070	143	26	model	model	NN
10_1101-2021_02_08_430070	143	27	structure	structure	NN
10_1101-2021_02_08_430070	143	28	.	.	.
10_1101-2021_02_08_430070	144	1	In	in	IN
10_1101-2021_02_08_430070	144	2	a	a	DT
10_1101-2021_02_08_430070	144	3	biRNN	birnn	NN
10_1101-2021_02_08_430070	144	4	,	,	,
10_1101-2021_02_08_430070	144	5	such	such	JJ
10_1101-2021_02_08_430070	144	6	as	as	IN
10_1101-2021_02_08_430070	144	7	the	the	DT
10_1101-2021_02_08_430070	144	8	implementation	implementation	NN
10_1101-2021_02_08_430070	144	9	of	of	IN
10_1101-2021_02_08_430070	144	10	deepMOD	deepMOD	NNP
10_1101-2021_02_08_430070	144	11	,	,	,
10_1101-2021_02_08_430070	144	12	the	the	DT
10_1101-2021_02_08_430070	144	13	last	last	JJ
10_1101-2021_02_08_430070	144	14	full	full	JJ
10_1101-2021_02_08_430070	144	15	connection	connection	NN
10_1101-2021_02_08_430070	144	16	layer	layer	NN
10_1101-2021_02_08_430070	144	17	uses	use	VBZ
10_1101-2021_02_08_430070	144	18	hidden	hide	VBN
10_1101-2021_02_08_430070	144	19	units	unit	NNS
10_1101-2021_02_08_430070	144	20	of	of	IN
10_1101-2021_02_08_430070	144	21	the	the	DT
10_1101-2021_02_08_430070	144	22	center	center	NN
10_1101-2021_02_08_430070	144	23	time	time	NN
10_1101-2021_02_08_430070	144	24	step	step	NN
10_1101-2021_02_08_430070	144	25	as	as	IN
10_1101-2021_02_08_430070	144	26	the	the	DT
10_1101-2021_02_08_430070	144	27	input	input	NN
10_1101-2021_02_08_430070	144	28	.	.	.
10_1101-2021_02_08_430070	145	1	Meanwhile	meanwhile	RB
10_1101-2021_02_08_430070	145	2	,	,	,
10_1101-2021_02_08_430070	145	3	the	the	DT
10_1101-2021_02_08_430070	145	4	bi	bi	JJ
10_1101-2021_02_08_430070	145	5	-	-	JJ
10_1101-2021_02_08_430070	145	6	directional	directional	JJ
10_1101-2021_02_08_430070	145	7	structure	structure	NN
10_1101-2021_02_08_430070	145	8	and	and	CC
10_1101-2021_02_08_430070	145	9	the	the	DT
10_1101-2021_02_08_430070	145	10	information	information	NN
10_1101-2021_02_08_430070	145	11	decay	decay	NN
10_1101-2021_02_08_430070	145	12	from	from	IN
10_1101-2021_02_08_430070	145	13	both	both	DT
10_1101-2021_02_08_430070	145	14	ends	end	NNS
10_1101-2021_02_08_430070	145	15	to	to	IN
10_1101-2021_02_08_430070	145	16	the	the	DT
10_1101-2021_02_08_430070	145	17	center	center	JJ
10_1101-2021_02_08_430070	145	18	position	position	NN
10_1101-2021_02_08_430070	145	19	render	render	VB
10_1101-2021_02_08_430070	145	20	the	the	DT
10_1101-2021_02_08_430070	145	21	model	model	NN
10_1101-2021_02_08_430070	145	22	focusing	focus	VBG
10_1101-2021_02_08_430070	145	23	more	more	RBR
10_1101-2021_02_08_430070	145	24	on	on	IN
10_1101-2021_02_08_430070	145	25	center	center	NN
10_1101-2021_02_08_430070	145	26	positions	position	NNS
10_1101-2021_02_08_430070	145	27	.	.	.
10_1101-2021_02_08_430070	146	1	For	for	IN
10_1101-2021_02_08_430070	146	2	the	the	DT
10_1101-2021_02_08_430070	146	3	basic	basic	JJ
10_1101-2021_02_08_430070	146	4	BERT	BERT	NNP
10_1101-2021_02_08_430070	146	5	,	,	,
10_1101-2021_02_08_430070	146	6	as	as	IN
10_1101-2021_02_08_430070	146	7	any	any	DT
10_1101-2021_02_08_430070	146	8	arbitrary	arbitrary	JJ
10_1101-2021_02_08_430070	146	9	time-	time-	NN
10_1101-2021_02_08_430070	146	10	step	step	NN
10_1101-2021_02_08_430070	146	11	pair	pair	NN
10_1101-2021_02_08_430070	146	12	is	be	VBZ
10_1101-2021_02_08_430070	146	13	processed	process	VBN
10_1101-2021_02_08_430070	146	14	with	with	IN
10_1101-2021_02_08_430070	146	15	the	the	DT
10_1101-2021_02_08_430070	146	16	same	same	JJ
10_1101-2021_02_08_430070	146	17	attention	attention	NN
10_1101-2021_02_08_430070	146	18	module	module	NN
10_1101-2021_02_08_430070	146	19	,	,	,
10_1101-2021_02_08_430070	146	20	the	the	DT
10_1101-2021_02_08_430070	146	21	importance	importance	NN
10_1101-2021_02_08_430070	146	22	of	of	IN
10_1101-2021_02_08_430070	146	23	center	center	NN
10_1101-2021_02_08_430070	146	24	positions	position	NNS
10_1101-2021_02_08_430070	146	25	are	be	VBP
10_1101-2021_02_08_430070	146	26	not	not	RB
10_1101-2021_02_08_430070	146	27	specifically	specifically	RB
10_1101-2021_02_08_430070	146	28	considered	consider	VBN
10_1101-2021_02_08_430070	146	29	in	in	IN
10_1101-2021_02_08_430070	146	30	the	the	DT
10_1101-2021_02_08_430070	146	31	model	model	NN
10_1101-2021_02_08_430070	146	32	.	.	.
10_1101-2021_02_08_430070	147	1	Therefore	therefore	RB
10_1101-2021_02_08_430070	147	2	,	,	,
10_1101-2021_02_08_430070	147	3	we	-PRON-	PRP
10_1101-2021_02_08_430070	147	4	propose	propose	VBP
10_1101-2021_02_08_430070	147	5	a	a	DT
10_1101-2021_02_08_430070	147	6	refined	refined	JJ
10_1101-2021_02_08_430070	147	7	BERT	BERT	NNP
10_1101-2021_02_08_430070	147	8	model	model	NN
10_1101-2021_02_08_430070	147	9	to	to	TO
10_1101-2021_02_08_430070	147	10	solve	solve	VB
10_1101-2021_02_08_430070	147	11	this	this	DT
10_1101-2021_02_08_430070	147	12	problem	problem	NN
10_1101-2021_02_08_430070	147	13	.	.	.
10_1101-2021_02_08_430070	148	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	148	2	incorporate	incorporate	VBP
10_1101-2021_02_08_430070	148	3	relative	relative	JJ
10_1101-2021_02_08_430070	148	4	-	-	HYPH
10_1101-2021_02_08_430070	148	5	position	position	NN
10_1101-2021_02_08_430070	148	6	attention	attention	NN
10_1101-2021_02_08_430070	148	7	and	and	CC
10_1101-2021_02_08_430070	148	8	center	center	NN
10_1101-2021_02_08_430070	148	9	-	-	HYPH
10_1101-2021_02_08_430070	148	10	hidden	hide	VBN
10_1101-2021_02_08_430070	148	11	-	-	HYPH
10_1101-2021_02_08_430070	148	12	units	unit	NNS
10_1101-2021_02_08_430070	148	13	concatenation	concatenation	NN
10_1101-2021_02_08_430070	148	14	to	to	TO
10_1101-2021_02_08_430070	148	15	enable	enable	VB
10_1101-2021_02_08_430070	148	16	a	a	DT
10_1101-2021_02_08_430070	148	17	BERT	BERT	NNP
10_1101-2021_02_08_430070	148	18	model	model	NN
10_1101-2021_02_08_430070	148	19	to	to	TO
10_1101-2021_02_08_430070	148	20	pay	pay	VB
10_1101-2021_02_08_430070	148	21	more	more	JJR
10_1101-2021_02_08_430070	148	22	attention	attention	NN
10_1101-2021_02_08_430070	148	23	to	to	IN
10_1101-2021_02_08_430070	148	24	center	center	NN
10_1101-2021_02_08_430070	148	25	positions	position	NNS
10_1101-2021_02_08_430070	148	26	.	.	.
10_1101-2021_02_08_430070	149	1	3.3	3.3	CD
10_1101-2021_02_08_430070	149	2	In	in	IN
10_1101-2021_02_08_430070	149	3	-	-	HYPH
10_1101-2021_02_08_430070	149	4	sample	sample	NN
10_1101-2021_02_08_430070	149	5	evaluation	evaluation	NN
10_1101-2021_02_08_430070	149	6	To	to	TO
10_1101-2021_02_08_430070	149	7	evaluate	evaluate	VB
10_1101-2021_02_08_430070	149	8	model	model	NN
10_1101-2021_02_08_430070	149	9	performance	performance	NN
10_1101-2021_02_08_430070	149	10	,	,	,
10_1101-2021_02_08_430070	149	11	we	-PRON-	PRP
10_1101-2021_02_08_430070	149	12	first	first	RB
10_1101-2021_02_08_430070	149	13	perform	perform	VBP
10_1101-2021_02_08_430070	149	14	the	the	DT
10_1101-2021_02_08_430070	149	15	in	in	IN
10_1101-2021_02_08_430070	149	16	-	-	HYPH
10_1101-2021_02_08_430070	149	17	sample	sample	NN
10_1101-2021_02_08_430070	149	18	evaluation	evaluation	NN
10_1101-2021_02_08_430070	149	19	on	on	IN
10_1101-2021_02_08_430070	149	20	5mC	5mC	NNP
10_1101-2021_02_08_430070	149	21	and	and	CC
10_1101-2021_02_08_430070	149	22	6mA	6mA	NNP
10_1101-2021_02_08_430070	149	23	datasets	dataset	NNS
10_1101-2021_02_08_430070	149	24	.	.	.
10_1101-2021_02_08_430070	150	1	The	the	DT
10_1101-2021_02_08_430070	150	2	predictions	prediction	NNS
10_1101-2021_02_08_430070	150	3	of	of	IN
10_1101-2021_02_08_430070	150	4	different	different	JJ
10_1101-2021_02_08_430070	150	5	models	model	NNS
10_1101-2021_02_08_430070	150	6	are	be	VBP
10_1101-2021_02_08_430070	150	7	evaluated	evaluate	VBN
10_1101-2021_02_08_430070	150	8	on	on	IN
10_1101-2021_02_08_430070	150	9	the	the	DT
10_1101-2021_02_08_430070	150	10	read	read	JJ
10_1101-2021_02_08_430070	150	11	and	and	CC
10_1101-2021_02_08_430070	150	12	genomic	genomic	JJ
10_1101-2021_02_08_430070	150	13	level	level	NN
10_1101-2021_02_08_430070	150	14	.	.	.
10_1101-2021_02_08_430070	151	1	For	for	IN
10_1101-2021_02_08_430070	151	2	the	the	DT
10_1101-2021_02_08_430070	151	3	genomic	genomic	JJ
10_1101-2021_02_08_430070	151	4	level	level	NN
10_1101-2021_02_08_430070	151	5	evaluation	evaluation	NN
10_1101-2021_02_08_430070	151	6	,	,	,
10_1101-2021_02_08_430070	151	7	we	-PRON-	PRP
10_1101-2021_02_08_430070	151	8	group	group	VBP
10_1101-2021_02_08_430070	151	9	all	all	DT
10_1101-2021_02_08_430070	151	10	reads	read	VBZ
10_1101-2021_02_08_430070	151	11	aligned	align	VBN
10_1101-2021_02_08_430070	151	12	to	to	IN
10_1101-2021_02_08_430070	151	13	the	the	DT
10_1101-2021_02_08_430070	151	14	same	same	JJ
10_1101-2021_02_08_430070	151	15	genomic	genomic	JJ
10_1101-2021_02_08_430070	151	16	coordinate	coordinate	NN
10_1101-2021_02_08_430070	151	17	,	,	,
10_1101-2021_02_08_430070	151	18	and	and	CC
10_1101-2021_02_08_430070	151	19	uses	use	VBZ
10_1101-2021_02_08_430070	151	20	a	a	DT
10_1101-2021_02_08_430070	151	21	threshold	threshold	NN
10_1101-2021_02_08_430070	151	22	of	of	IN
10_1101-2021_02_08_430070	151	23	prediction	prediction	NN
10_1101-2021_02_08_430070	151	24	methylation	methylation	NN
10_1101-2021_02_08_430070	151	25	percentage	percentage	NN
10_1101-2021_02_08_430070	151	26	≥	≥	CD
10_1101-2021_02_08_430070	151	27	0.1	0.1	CD
10_1101-2021_02_08_430070	151	28	(	(	-LRB-
10_1101-2021_02_08_430070	151	29	same	same	JJ
10_1101-2021_02_08_430070	151	30	as	as	IN
10_1101-2021_02_08_430070	151	31	deepMOD	deepMOD	NNP
10_1101-2021_02_08_430070	151	32	)	)	-RRB-
10_1101-2021_02_08_430070	151	33	as	as	IN
10_1101-2021_02_08_430070	151	34	a	a	DT
10_1101-2021_02_08_430070	151	35	genomic	genomic	JJ
10_1101-2021_02_08_430070	151	36	position	position	NN
10_1101-2021_02_08_430070	151	37	prediction	prediction	NN
10_1101-2021_02_08_430070	151	38	.	.	.
10_1101-2021_02_08_430070	152	1	In	in	IN
10_1101-2021_02_08_430070	152	2	general	general	JJ
10_1101-2021_02_08_430070	152	3	,	,	,
10_1101-2021_02_08_430070	152	4	on	on	IN
10_1101-2021_02_08_430070	152	5	the	the	DT
10_1101-2021_02_08_430070	152	6	five	five	CD
10_1101-2021_02_08_430070	152	7	5mC	5mc	CD
10_1101-2021_02_08_430070	152	8	datasets	dataset	NNS
10_1101-2021_02_08_430070	152	9	,	,	,
10_1101-2021_02_08_430070	152	10	the	the	DT
10_1101-2021_02_08_430070	152	11	AUC	AUC	NNP
10_1101-2021_02_08_430070	152	12	performance	performance	NN
10_1101-2021_02_08_430070	152	13	of	of	IN
10_1101-2021_02_08_430070	152	14	the	the	DT
10_1101-2021_02_08_430070	152	15	three	three	CD
10_1101-2021_02_08_430070	152	16	models	model	NNS
10_1101-2021_02_08_430070	152	17	are	be	VBP
10_1101-2021_02_08_430070	152	18	relatively	relatively	RB
10_1101-2021_02_08_430070	152	19	close	close	JJ
10_1101-2021_02_08_430070	152	20	on	on	IN
10_1101-2021_02_08_430070	152	21	both	both	DT
10_1101-2021_02_08_430070	152	22	read	read	VBN
10_1101-2021_02_08_430070	152	23	level	level	NN
10_1101-2021_02_08_430070	152	24	and	and	CC
10_1101-2021_02_08_430070	152	25	genomic	genomic	JJ
10_1101-2021_02_08_430070	152	26	level	level	NN
10_1101-2021_02_08_430070	152	27	.	.	.
10_1101-2021_02_08_430070	153	1	The	the	DT
10_1101-2021_02_08_430070	153	2	basic	basic	JJ
10_1101-2021_02_08_430070	153	3	BERT	BERT	NNP
10_1101-2021_02_08_430070	153	4	model	model	NN
10_1101-2021_02_08_430070	153	5	does	do	VBZ
10_1101-2021_02_08_430070	153	6	not	not	RB
10_1101-2021_02_08_430070	153	7	work	work	VB
10_1101-2021_02_08_430070	153	8	as	as	RB
10_1101-2021_02_08_430070	153	9	well	well	RB
10_1101-2021_02_08_430070	153	10	as	as	IN
10_1101-2021_02_08_430070	153	11	the	the	DT
10_1101-2021_02_08_430070	153	12	biRNN	birnn	JJ
10_1101-2021_02_08_430070	153	13	model	model	NN
10_1101-2021_02_08_430070	153	14	that	that	IN
10_1101-2021_02_08_430070	153	15	AUC	AUC	NNP
10_1101-2021_02_08_430070	153	16	scores	score	NNS
10_1101-2021_02_08_430070	153	17	are	be	VBP
10_1101-2021_02_08_430070	153	18	lower	low	JJR
10_1101-2021_02_08_430070	153	19	.	.	.
10_1101-2021_02_08_430070	154	1	The	the	DT
10_1101-2021_02_08_430070	154	2	refined	refined	JJ
10_1101-2021_02_08_430070	154	3	BERT	BERT	NNP
10_1101-2021_02_08_430070	154	4	model	model	NN
10_1101-2021_02_08_430070	154	5	achieves	achieve	VBZ
10_1101-2021_02_08_430070	154	6	equivalent	equivalent	JJ
10_1101-2021_02_08_430070	154	7	or	or	CC
10_1101-2021_02_08_430070	154	8	better	well	JJR
10_1101-2021_02_08_430070	154	9	AUC	AUC	NNP
10_1101-2021_02_08_430070	154	10	scores	score	NNS
10_1101-2021_02_08_430070	154	11	on	on	IN
10_1101-2021_02_08_430070	154	12	the	the	DT
10_1101-2021_02_08_430070	154	13	genomic-	genomic-	JJ
10_1101-2021_02_08_430070	154	14	level	level	NN
10_1101-2021_02_08_430070	154	15	.	.	.
10_1101-2021_02_08_430070	155	1	Note	note	VB
10_1101-2021_02_08_430070	155	2	that	that	IN
10_1101-2021_02_08_430070	155	3	on	on	IN
10_1101-2021_02_08_430070	155	4	the	the	DT
10_1101-2021_02_08_430070	155	5	dataset	dataset	JJ
10_1101-2021_02_08_430070	155	6	Stoiber_E.coli_CG_MpeI	Stoiber_E.coli_CG_MpeI	NNP
10_1101-2021_02_08_430070	155	7	and	and	CC
10_1101-2021_02_08_430070	155	8	.license	.license	NNP
10_1101-2021_02_08_430070	155	9	CC	CC	NNP
10_1101-2021_02_08_430070	155	10	-	-	HYPH
10_1101-2021_02_08_430070	155	11	BY	BY	NNP
10_1101-2021_02_08_430070	155	12	-	-	HYPH
10_1101-2021_02_08_430070	155	13	NC	NC	NNP
10_1101-2021_02_08_430070	155	14	-	-	HYPH
10_1101-2021_02_08_430070	155	15	ND	ND	NNP
10_1101-2021_02_08_430070	155	16	4.0	4.0	CD
10_1101-2021_02_08_430070	155	17	Internationalpeer	Internationalpeer	NNP
10_1101-2021_02_08_430070	155	18	review	review	NN
10_1101-2021_02_08_430070	155	19	)	)	-RRB-
10_1101-2021_02_08_430070	155	20	is	be	VBZ
10_1101-2021_02_08_430070	155	21	the	the	DT
10_1101-2021_02_08_430070	155	22	author	author	NN
10_1101-2021_02_08_430070	155	23	/	/	SYM
10_1101-2021_02_08_430070	155	24	funder	funder	NN
10_1101-2021_02_08_430070	155	25	,	,	,
10_1101-2021_02_08_430070	155	26	who	who	WP
10_1101-2021_02_08_430070	155	27	has	have	VBZ
10_1101-2021_02_08_430070	155	28	granted	grant	VBN
10_1101-2021_02_08_430070	155	29	bioRxiv	biorxiv	IN
10_1101-2021_02_08_430070	155	30	a	a	DT
10_1101-2021_02_08_430070	155	31	license	license	NN
10_1101-2021_02_08_430070	155	32	to	to	TO
10_1101-2021_02_08_430070	155	33	display	display	VB
10_1101-2021_02_08_430070	155	34	the	the	DT
10_1101-2021_02_08_430070	155	35	preprint	preprint	NN
10_1101-2021_02_08_430070	155	36	in	in	IN
10_1101-2021_02_08_430070	155	37	perpetuity	perpetuity	NN
10_1101-2021_02_08_430070	155	38	.	.	.
10_1101-2021_02_08_430070	156	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	156	2	is	be	VBZ
10_1101-2021_02_08_430070	156	3	made	make	VBN
10_1101-2021_02_08_430070	156	4	available	available	JJ
10_1101-2021_02_08_430070	156	5	under	under	IN
10_1101-2021_02_08_430070	156	6	a	a	DT
10_1101-2021_02_08_430070	156	7	The	the	DT
10_1101-2021_02_08_430070	156	8	copyright	copyright	NN
10_1101-2021_02_08_430070	156	9	holder	holder	NN
10_1101-2021_02_08_430070	156	10	for	for	IN
10_1101-2021_02_08_430070	156	11	this	this	DT
10_1101-2021_02_08_430070	156	12	preprint	preprint	NN
10_1101-2021_02_08_430070	156	13	(	(	-LRB-
10_1101-2021_02_08_430070	156	14	which	which	WDT
10_1101-2021_02_08_430070	156	15	was	be	VBD
10_1101-2021_02_08_430070	156	16	not	not	RB
10_1101-2021_02_08_430070	156	17	certified	certify	VBN
10_1101-2021_02_08_430070	156	18	bythis	bythis	DT
10_1101-2021_02_08_430070	156	19	version	version	NN
10_1101-2021_02_08_430070	156	20	posted	post	VBD
10_1101-2021_02_08_430070	156	21	February	February	NNP
10_1101-2021_02_08_430070	156	22	10	10	CD
10_1101-2021_02_08_430070	156	23	,	,	,
10_1101-2021_02_08_430070	156	24	2021	2021	CD
10_1101-2021_02_08_430070	156	25	.	.	.
10_1101-2021_02_08_430070	156	26	;	;	:
10_1101-2021_02_08_430070	156	27	https://doi.org/10.1101/2021.02.08.430070doi	https://doi.org/10.1101/2021.02.08.430070doi	NFP
10_1101-2021_02_08_430070	156	28	:	:	:
10_1101-2021_02_08_430070	156	29	bioRxiv	biorxiv	VB
10_1101-2021_02_08_430070	156	30	preprint	preprint	NN
10_1101-2021_02_08_430070	156	31	https://doi.org/10.1101/2021.02.08.430070	https://doi.org/10.1101/2021.02.08.430070	UH
10_1101-2021_02_08_430070	156	32	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	156	33	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	156	34	✐	✐	NNP
10_1101-2021_02_08_430070	156	35	✐	✐	NNP
10_1101-2021_02_08_430070	156	36	✐	✐	NNP
10_1101-2021_02_08_430070	156	37	✐	✐	NNP
10_1101-2021_02_08_430070	156	38	✐	✐	NNP
10_1101-2021_02_08_430070	156	39	✐	✐	NNP
10_1101-2021_02_08_430070	156	40	✐	✐	NNP
10_1101-2021_02_08_430070	156	41	✐	✐	NNP
10_1101-2021_02_08_430070	156	42	BERT	BERT	NNP
10_1101-2021_02_08_430070	156	43	for	for	IN
10_1101-2021_02_08_430070	156	44	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	156	45	methylation	methylation	NN
10_1101-2021_02_08_430070	156	46	detection	detection	NN
10_1101-2021_02_08_430070	156	47	5	5	CD
10_1101-2021_02_08_430070	156	48	Dataset	Dataset	NNP
10_1101-2021_02_08_430070	156	49	Species	Species	NNP
10_1101-2021_02_08_430070	156	50	Motif_Methyltransferase	Motif_Methyltransferase	NNP
10_1101-2021_02_08_430070	156	51	Model	Model	NNP
10_1101-2021_02_08_430070	156	52	Single	Single	NNP
10_1101-2021_02_08_430070	156	53	(	(	-LRB-
10_1101-2021_02_08_430070	156	54	read	read	VBN
10_1101-2021_02_08_430070	156	55	-	-	HYPH
10_1101-2021_02_08_430070	156	56	level	level	NN
10_1101-2021_02_08_430070	156	57	)	)	-RRB-
10_1101-2021_02_08_430070	156	58	Group	group	NN
10_1101-2021_02_08_430070	156	59	(	(	-LRB-
10_1101-2021_02_08_430070	156	60	>	>	XX
10_1101-2021_02_08_430070	156	61	=	=	SYM
10_1101-2021_02_08_430070	156	62	1	1	CD
10_1101-2021_02_08_430070	156	63	,	,	,
10_1101-2021_02_08_430070	156	64	genomic	genomic	JJ
10_1101-2021_02_08_430070	156	65	-	-	HYPH
10_1101-2021_02_08_430070	156	66	level	level	NN
10_1101-2021_02_08_430070	156	67	)	)	-RRB-
10_1101-2021_02_08_430070	156	68	AUC	AUC	NNP
10_1101-2021_02_08_430070	156	69	Precision	Precision	NNP
10_1101-2021_02_08_430070	156	70	Recall	Recall	NNP
10_1101-2021_02_08_430070	156	71	AUC	AUC	NNP
10_1101-2021_02_08_430070	156	72	Precision	Precision	NNP
10_1101-2021_02_08_430070	156	73	Recall	Recall	NNP
10_1101-2021_02_08_430070	156	74	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	156	75	E.coli	e.coli	JJ
10_1101-2021_02_08_430070	156	76	GCGC_HhaI	GCGC_HhaI	VBZ
10_1101-2021_02_08_430070	156	77	biRNN	birnn	NN
10_1101-2021_02_08_430070	156	78	0.9205	0.9205	CD
10_1101-2021_02_08_430070	156	79	0.9545	0.9545	CD
10_1101-2021_02_08_430070	156	80	0.8593	0.8593	CD
10_1101-2021_02_08_430070	156	81	0.9322	0.9322	CD
10_1101-2021_02_08_430070	156	82	0.9320	0.9320	CD
10_1101-2021_02_08_430070	156	83	0.9134	0.9134	CD
10_1101-2021_02_08_430070	156	84	BERT_basic	BERT_basic	NNP
10_1101-2021_02_08_430070	156	85	0.9183	0.9183	CD
10_1101-2021_02_08_430070	156	86	0.9528	0.9528	CD
10_1101-2021_02_08_430070	156	87	0.8556	0.8556	CD
10_1101-2021_02_08_430070	156	88	0.9305	0.9305	CD
10_1101-2021_02_08_430070	156	89	0.9299	0.9299	CD
10_1101-2021_02_08_430070	156	90	0.9113	0.9113	CD
10_1101-2021_02_08_430070	156	91	BERT_refined	BERT_refined	NNP
10_1101-2021_02_08_430070	156	92	0.9239	0.9239	CD
10_1101-2021_02_08_430070	156	93	0.9563	0.9563	CD
10_1101-2021_02_08_430070	156	94	0.8655	0.8655	CD
10_1101-2021_02_08_430070	156	95	0.9351	0.9351	CD
10_1101-2021_02_08_430070	156	96	0.9341	0.9341	CD
10_1101-2021_02_08_430070	156	97	0.9177	0.9177	CD
10_1101-2021_02_08_430070	156	98	CG_MpeI	CG_MpeI	NNP
10_1101-2021_02_08_430070	156	99	BiRNN	BiRNN	NNP
10_1101-2021_02_08_430070	156	100	0.7184	0.7184	CD
10_1101-2021_02_08_430070	156	101	0.8943	0.8943	CD
10_1101-2021_02_08_430070	156	102	0.4555	0.4555	CD
10_1101-2021_02_08_430070	156	103	0.7482	0.7482	CD
10_1101-2021_02_08_430070	156	104	0.8764	0.8764	CD
10_1101-2021_02_08_430070	156	105	0.5452	0.5452	CD
10_1101-2021_02_08_430070	156	106	BERT_basic	BERT_basic	NNP
10_1101-2021_02_08_430070	156	107	0.7045	0.7045	CD
10_1101-2021_02_08_430070	156	108	0.8682	0.8682	CD
10_1101-2021_02_08_430070	156	109	0.4316	0.4316	CD
10_1101-2021_02_08_430070	156	110	0.7312	0.7312	CD
10_1101-2021_02_08_430070	156	111	0.8494	0.8494	CD
10_1101-2021_02_08_430070	156	112	0.5211	0.5211	CD
10_1101-2021_02_08_430070	156	113	BERT_refined	bert_refine	VBN
10_1101-2021_02_08_430070	156	114	0.717	0.717	CD
10_1101-2021_02_08_430070	156	115	0.9017	0.9017	CD
10_1101-2021_02_08_430070	156	116	0.4511	0.4511	CD
10_1101-2021_02_08_430070	156	117	0.7482	0.7482	CD
10_1101-2021_02_08_430070	156	118	0.8848	0.8848	CD
10_1101-2021_02_08_430070	156	119	0.5412	0.5412	CD
10_1101-2021_02_08_430070	156	120	CG_SssI	CG_SssI	NNP
10_1101-2021_02_08_430070	156	121	BiRNN	birnn	DT
10_1101-2021_02_08_430070	156	122	0.9017	0.9017	CD
10_1101-2021_02_08_430070	156	123	0.9576	0.9576	CD
10_1101-2021_02_08_430070	156	124	0.8097	0.8097	CD
10_1101-2021_02_08_430070	156	125	0.9127	0.9127	CD
10_1101-2021_02_08_430070	156	126	0.9508	0.9508	CD
10_1101-2021_02_08_430070	156	127	0.8420	0.8420	CD
10_1101-2021_02_08_430070	156	128	BERT_basic	BERT_basic	NNP
10_1101-2021_02_08_430070	156	129	0.9001	0.9001	CD
10_1101-2021_02_08_430070	156	130	0.9534	0.9534	CD
10_1101-2021_02_08_430070	156	131	0.8071	0.8071	CD
10_1101-2021_02_08_430070	156	132	0.9107	0.9107	CD
10_1101-2021_02_08_430070	156	133	0.9463	0.9463	CD
10_1101-2021_02_08_430070	156	134	0.8395	0.8395	CD
10_1101-2021_02_08_430070	156	135	BERT_refined	bert_refine	VBN
10_1101-2021_02_08_430070	156	136	0.9068	0.9068	CD
10_1101-2021_02_08_430070	156	137	0.9509	0.9509	CD
10_1101-2021_02_08_430070	156	138	0.821	0.821	CD
10_1101-2021_02_08_430070	156	139	0.9162	0.9162	CD
10_1101-2021_02_08_430070	156	140	0.9433	0.9433	CD
10_1101-2021_02_08_430070	156	141	0.852	0.852	CD
10_1101-2021_02_08_430070	156	142	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	156	143	E.	E.	NNP
10_1101-2021_02_08_430070	156	144	coli	coli	NNS
10_1101-2021_02_08_430070	156	145	CG_SssI	CG_SssI	NNP
10_1101-2021_02_08_430070	156	146	BiRNN	birnn	DT
10_1101-2021_02_08_430070	156	147	0.9514	0.9514	CD
10_1101-2021_02_08_430070	156	148	0.9512	0.9512	CD
10_1101-2021_02_08_430070	156	149	0.9316	0.9316	CD
10_1101-2021_02_08_430070	156	150	0.9284	0.9284	CD
10_1101-2021_02_08_430070	156	151	0.8805	0.8805	CD
10_1101-2021_02_08_430070	156	152	0.9854	0.9854	CD
10_1101-2021_02_08_430070	156	153	BERT_basic	BERT_basic	NNP
10_1101-2021_02_08_430070	156	154	0.9477	0.9477	CD
10_1101-2021_02_08_430070	156	155	0.9469	0.9469	CD
10_1101-2021_02_08_430070	156	156	0.9268	0.9268	CD
10_1101-2021_02_08_430070	156	157	0.9227	0.9227	CD
10_1101-2021_02_08_430070	156	158	0.8718	0.8718	CD
10_1101-2021_02_08_430070	156	159	0.9845	0.9845	CD
10_1101-2021_02_08_430070	156	160	BERT_refined	bert_refine	VBN
10_1101-2021_02_08_430070	156	161	0.9464	0.9464	CD
10_1101-2021_02_08_430070	156	162	0.9656	0.9656	CD
10_1101-2021_02_08_430070	156	163	0.9124	0.9124	CD
10_1101-2021_02_08_430070	156	164	0.9456	0.9456	CD
10_1101-2021_02_08_430070	156	165	0.9135	0.9135	CD
10_1101-2021_02_08_430070	156	166	0.9803	0.9803	CD
10_1101-2021_02_08_430070	156	167	H.Sapiens	H.Sapiens	NNP
10_1101-2021_02_08_430070	156	168	CG_SssI	CG_SssI	NNS
10_1101-2021_02_08_430070	156	169	BiRNN	birnn	DT
10_1101-2021_02_08_430070	156	170	0.9004	0.9004	CD
10_1101-2021_02_08_430070	156	171	0.8891	0.8891	CD
10_1101-2021_02_08_430070	156	172	0.9230	0.9230	CD
10_1101-2021_02_08_430070	156	173	0.9010	0.9010	CD
10_1101-2021_02_08_430070	156	174	0.8900	0.8900	CD
10_1101-2021_02_08_430070	156	175	0.9240	0.9240	CD
10_1101-2021_02_08_430070	156	176	BERT_basic	bert_basic	NN
10_1101-2021_02_08_430070	156	177	0.8962	0.8962	CD
10_1101-2021_02_08_430070	156	178	0.8813	0.8813	CD
10_1101-2021_02_08_430070	156	179	0.9248	0.9248	CD
10_1101-2021_02_08_430070	156	180	0.8969	0.8969	CD
10_1101-2021_02_08_430070	156	181	0.8823	0.8823	CD
10_1101-2021_02_08_430070	156	182	0.9256	0.9256	CD
10_1101-2021_02_08_430070	156	183	BERT_refined	BERT_refined	NNP
10_1101-2021_02_08_430070	156	184	0.9045	0.9045	CD
10_1101-2021_02_08_430070	156	185	0.9143	0.9143	CD
10_1101-2021_02_08_430070	156	186	0.8984	0.8984	CD
10_1101-2021_02_08_430070	156	187	0.9053	0.9053	CD
10_1101-2021_02_08_430070	156	188	0.9147	0.9147	CD
10_1101-2021_02_08_430070	156	189	0.9003	0.9003	CD
10_1101-2021_02_08_430070	156	190	Table	table	NN
10_1101-2021_02_08_430070	156	191	1	1	CD
10_1101-2021_02_08_430070	156	192	.	.	.
10_1101-2021_02_08_430070	157	1	In	in	IN
10_1101-2021_02_08_430070	157	2	-	-	HYPH
10_1101-2021_02_08_430070	157	3	sample	sample	NN
10_1101-2021_02_08_430070	157	4	evaluation	evaluation	NN
10_1101-2021_02_08_430070	157	5	of	of	IN
10_1101-2021_02_08_430070	157	6	different	different	JJ
10_1101-2021_02_08_430070	157	7	deep	deep	JJ
10_1101-2021_02_08_430070	157	8	learning	learning	NN
10_1101-2021_02_08_430070	157	9	models	model	NNS
10_1101-2021_02_08_430070	157	10	on	on	IN
10_1101-2021_02_08_430070	157	11	5mC	5mc	CD
10_1101-2021_02_08_430070	157	12	datasets	dataset	NNS
10_1101-2021_02_08_430070	157	13	.	.	.
10_1101-2021_02_08_430070	158	1	The	the	DT
10_1101-2021_02_08_430070	158	2	best	good	JJS
10_1101-2021_02_08_430070	158	3	score	score	NN
10_1101-2021_02_08_430070	158	4	of	of	IN
10_1101-2021_02_08_430070	158	5	each	each	DT
10_1101-2021_02_08_430070	158	6	dataset	dataset	NN
10_1101-2021_02_08_430070	158	7	is	be	VBZ
10_1101-2021_02_08_430070	158	8	highlighted	highlight	VBN
10_1101-2021_02_08_430070	158	9	in	in	IN
10_1101-2021_02_08_430070	158	10	bold	bold	JJ
10_1101-2021_02_08_430070	158	11	.	.	.
10_1101-2021_02_08_430070	159	1	Dataset	Dataset	NNP
10_1101-2021_02_08_430070	159	2	Species	Species	NNP
10_1101-2021_02_08_430070	159	3	Motif_Methyltransferase	Motif_Methyltransferase	NNP
10_1101-2021_02_08_430070	159	4	Model	Model	NNP
10_1101-2021_02_08_430070	159	5	Single	Single	NNP
10_1101-2021_02_08_430070	159	6	(	(	-LRB-
10_1101-2021_02_08_430070	159	7	read	read	VBN
10_1101-2021_02_08_430070	159	8	-	-	HYPH
10_1101-2021_02_08_430070	159	9	level	level	NN
10_1101-2021_02_08_430070	159	10	)	)	-RRB-
10_1101-2021_02_08_430070	159	11	Group	group	NN
10_1101-2021_02_08_430070	159	12	(	(	-LRB-
10_1101-2021_02_08_430070	159	13	>	>	XX
10_1101-2021_02_08_430070	159	14	=	=	SYM
10_1101-2021_02_08_430070	159	15	1	1	CD
10_1101-2021_02_08_430070	159	16	,	,	,
10_1101-2021_02_08_430070	159	17	genomic	genomic	JJ
10_1101-2021_02_08_430070	159	18	level	level	NN
10_1101-2021_02_08_430070	159	19	)	)	-RRB-
10_1101-2021_02_08_430070	159	20	AUC	AUC	NNP
10_1101-2021_02_08_430070	159	21	Precision	Precision	NNP
10_1101-2021_02_08_430070	159	22	Recall	Recall	NNP
10_1101-2021_02_08_430070	159	23	AUC	AUC	NNP
10_1101-2021_02_08_430070	159	24	Precision	Precision	NNP
10_1101-2021_02_08_430070	159	25	Recall	Recall	NNP
10_1101-2021_02_08_430070	159	26	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	159	27	E.coli	e.coli	JJ
10_1101-2021_02_08_430070	159	28	gaAttc_EcoRI	gaattc_ecori	NN
10_1101-2021_02_08_430070	159	29	BiRNN	birnn	JJ
10_1101-2021_02_08_430070	159	30	0.8524	0.8524	CD
10_1101-2021_02_08_430070	159	31	0.8088	0.8088	CD
10_1101-2021_02_08_430070	159	32	0.7497	0.7497	CD
10_1101-2021_02_08_430070	159	33	0.8429	0.8429	CD
10_1101-2021_02_08_430070	159	34	0.7797	0.7797	CD
10_1101-2021_02_08_430070	159	35	0.8035	0.8035	CD
10_1101-2021_02_08_430070	159	36	BERT_basic	bert_basic	JJ
10_1101-2021_02_08_430070	159	37	0.8607	0.8607	CD
10_1101-2021_02_08_430070	159	38	0.8151	0.8151	CD
10_1101-2021_02_08_430070	159	39	0.7653	0.7653	CD
10_1101-2021_02_08_430070	159	40	0.8591	0.8591	CD
10_1101-2021_02_08_430070	159	41	0.7969	0.7969	CD
10_1101-2021_02_08_430070	159	42	0.8277	0.8277	CD
10_1101-2021_02_08_430070	159	43	BERT_refined	BERT_refined	NNP
10_1101-2021_02_08_430070	159	44	0.8611	0.8611	CD
10_1101-2021_02_08_430070	159	45	0.8826	0.8826	CD
10_1101-2021_02_08_430070	159	46	0.7473	0.7473	CD
10_1101-2021_02_08_430070	159	47	0.8655	0.8655	CD
10_1101-2021_02_08_430070	159	48	0.8596	0.8596	CD
10_1101-2021_02_08_430070	159	49	0.7987	0.7987	CD
10_1101-2021_02_08_430070	159	50	tcgA_TaqI	tcga_taqi	NN
10_1101-2021_02_08_430070	159	51	BiRNN	BiRNN	NNP
10_1101-2021_02_08_430070	159	52	0.7722	0.7722	CD
10_1101-2021_02_08_430070	159	53	0.7922	0.7922	CD
10_1101-2021_02_08_430070	159	54	0.5750	0.5750	CD
10_1101-2021_02_08_430070	159	55	0.7750	0.7750	CD
10_1101-2021_02_08_430070	159	56	0.7789	0.7789	CD
10_1101-2021_02_08_430070	159	57	0.6290	0.6290	CD
10_1101-2021_02_08_430070	159	58	BERT_basic	BERT_basic	NNP
10_1101-2021_02_08_430070	159	59	0.7573	0.7573	CD
10_1101-2021_02_08_430070	159	60	0.8168	0.8168	CD
10_1101-2021_02_08_430070	159	61	0.5392	0.5392	CD
10_1101-2021_02_08_430070	159	62	0.7653	0.7653	CD
10_1101-2021_02_08_430070	159	63	0.8063	0.8063	CD
10_1101-2021_02_08_430070	159	64	0.5937	0.5937	CD
10_1101-2021_02_08_430070	159	65	BERT_refined	BERT_refined	NNP
10_1101-2021_02_08_430070	159	66	0.7857	0.7857	CD
10_1101-2021_02_08_430070	159	67	0.7788	0.7788	CD
10_1101-2021_02_08_430070	159	68	0.6064	0.6064	CD
10_1101-2021_02_08_430070	159	69	0.7843	0.7843	CD
10_1101-2021_02_08_430070	159	70	0.7643	0.7643	CD
10_1101-2021_02_08_430070	159	71	0.6586	0.6586	CD
10_1101-2021_02_08_430070	159	72	gAtc_Dam	gatc_dam	CD
10_1101-2021_02_08_430070	159	73	BiRNN	BiRNN	NNP
10_1101-2021_02_08_430070	159	74	0.6123	0.6123	CD
10_1101-2021_02_08_430070	159	75	0.7656	0.7656	CD
10_1101-2021_02_08_430070	159	76	0.247	0.247	CD
10_1101-2021_02_08_430070	159	77	0.6337	0.6337	CD
10_1101-2021_02_08_430070	159	78	0.7631	0.7631	CD
10_1101-2021_02_08_430070	159	79	0.3241	0.3241	CD
10_1101-2021_02_08_430070	159	80	BERT_basic	BERT_basic	NNP
10_1101-2021_02_08_430070	159	81	0.6128	0.6128	CD
10_1101-2021_02_08_430070	159	82	0.7329	0.7329	CD
10_1101-2021_02_08_430070	159	83	0.2529	0.2529	CD
10_1101-2021_02_08_430070	159	84	0.631	0.631	CD
10_1101-2021_02_08_430070	159	85	0.7311	0.7311	CD
10_1101-2021_02_08_430070	159	86	0.3305	0.3305	CD
10_1101-2021_02_08_430070	159	87	BERT_refined	bert_refine	VBN
10_1101-2021_02_08_430070	159	88	0.6188	0.6188	CD
10_1101-2021_02_08_430070	159	89	0.7513	0.7513	CD
10_1101-2021_02_08_430070	159	90	0.2634	0.2634	CD
10_1101-2021_02_08_430070	159	91	0.6385	0.6385	CD
10_1101-2021_02_08_430070	159	92	0.7471	0.7471	CD
10_1101-2021_02_08_430070	159	93	0.3421	0.3421	CD
10_1101-2021_02_08_430070	159	94	Table	table	NN
10_1101-2021_02_08_430070	159	95	2	2	CD
10_1101-2021_02_08_430070	159	96	.	.	.
10_1101-2021_02_08_430070	160	1	In	in	IN
10_1101-2021_02_08_430070	160	2	-	-	HYPH
10_1101-2021_02_08_430070	160	3	sample	sample	NN
10_1101-2021_02_08_430070	160	4	evaluation	evaluation	NN
10_1101-2021_02_08_430070	160	5	of	of	IN
10_1101-2021_02_08_430070	160	6	different	different	JJ
10_1101-2021_02_08_430070	160	7	deep	deep	JJ
10_1101-2021_02_08_430070	160	8	learning	learning	NN
10_1101-2021_02_08_430070	160	9	models	model	NNS
10_1101-2021_02_08_430070	160	10	on	on	IN
10_1101-2021_02_08_430070	160	11	6mA	6ma	CD
10_1101-2021_02_08_430070	160	12	datasets	dataset	NNS
10_1101-2021_02_08_430070	160	13	.	.	.
10_1101-2021_02_08_430070	160	14	The	the	DT
10_1101-2021_02_08_430070	160	15	best	good	JJS
10_1101-2021_02_08_430070	160	16	score	score	NN
10_1101-2021_02_08_430070	160	17	of	of	IN
10_1101-2021_02_08_430070	160	18	each	each	DT
10_1101-2021_02_08_430070	160	19	dataset	dataset	NN
10_1101-2021_02_08_430070	160	20	is	be	VBZ
10_1101-2021_02_08_430070	160	21	highlighted	highlight	VBN
10_1101-2021_02_08_430070	160	22	in	in	IN
10_1101-2021_02_08_430070	160	23	bold	bold	JJ
10_1101-2021_02_08_430070	160	24	.	.	.
10_1101-2021_02_08_430070	161	1	Simpson_E.coli_CG_SssI	Simpson_E.coli_CG_SssI	NNP
10_1101-2021_02_08_430070	161	2	,	,	,
10_1101-2021_02_08_430070	161	3	although	although	IN
10_1101-2021_02_08_430070	161	4	the	the	DT
10_1101-2021_02_08_430070	161	5	read	read	JJ
10_1101-2021_02_08_430070	161	6	-	-	HYPH
10_1101-2021_02_08_430070	161	7	level	level	NN
10_1101-2021_02_08_430070	161	8	AUC	auc	NN
10_1101-2021_02_08_430070	161	9	of	of	IN
10_1101-2021_02_08_430070	161	10	the	the	DT
10_1101-2021_02_08_430070	161	11	refined	refined	JJ
10_1101-2021_02_08_430070	161	12	BERT	BERT	NNP
10_1101-2021_02_08_430070	161	13	are	be	VBP
10_1101-2021_02_08_430070	161	14	0.0014	0.0014	CD
10_1101-2021_02_08_430070	161	15	and	and	CC
10_1101-2021_02_08_430070	161	16	0.005	0.005	CD
10_1101-2021_02_08_430070	161	17	lower	low	JJR
10_1101-2021_02_08_430070	161	18	than	than	IN
10_1101-2021_02_08_430070	161	19	that	that	DT
10_1101-2021_02_08_430070	161	20	of	of	IN
10_1101-2021_02_08_430070	161	21	biRNN	birnn	NN
10_1101-2021_02_08_430070	161	22	,	,	,
10_1101-2021_02_08_430070	161	23	the	the	DT
10_1101-2021_02_08_430070	161	24	genomic	genomic	JJ
10_1101-2021_02_08_430070	161	25	-	-	HYPH
10_1101-2021_02_08_430070	161	26	level	level	NN
10_1101-2021_02_08_430070	161	27	performance	performance	NN
10_1101-2021_02_08_430070	161	28	of	of	IN
10_1101-2021_02_08_430070	161	29	the	the	DT
10_1101-2021_02_08_430070	161	30	refined	refined	JJ
10_1101-2021_02_08_430070	161	31	BERT	BERT	NNP
10_1101-2021_02_08_430070	161	32	is	be	VBZ
10_1101-2021_02_08_430070	161	33	equal	equal	JJ
10_1101-2021_02_08_430070	161	34	or	or	CC
10_1101-2021_02_08_430070	161	35	significantly	significantly	RB
10_1101-2021_02_08_430070	161	36	better	well	JJR
10_1101-2021_02_08_430070	161	37	than	than	IN
10_1101-2021_02_08_430070	161	38	biRNN	birnn	NN
10_1101-2021_02_08_430070	161	39	.	.	.
10_1101-2021_02_08_430070	162	1	This	this	DT
10_1101-2021_02_08_430070	162	2	can	can	MD
10_1101-2021_02_08_430070	162	3	be	be	VB
10_1101-2021_02_08_430070	162	4	explained	explain	VBN
10_1101-2021_02_08_430070	162	5	by	by	IN
10_1101-2021_02_08_430070	162	6	the	the	DT
10_1101-2021_02_08_430070	162	7	more	more	RBR
10_1101-2021_02_08_430070	162	8	accurate	accurate	JJ
10_1101-2021_02_08_430070	162	9	prediction	prediction	NN
10_1101-2021_02_08_430070	162	10	in	in	IN
10_1101-2021_02_08_430070	162	11	several	several	JJ
10_1101-2021_02_08_430070	162	12	low	low	JJ
10_1101-2021_02_08_430070	162	13	read	read	NN
10_1101-2021_02_08_430070	162	14	-	-	HYPH
10_1101-2021_02_08_430070	162	15	coverage	coverage	NN
10_1101-2021_02_08_430070	162	16	regions	region	NNS
10_1101-2021_02_08_430070	162	17	.	.	.
10_1101-2021_02_08_430070	163	1	On	on	IN
10_1101-2021_02_08_430070	163	2	the	the	DT
10_1101-2021_02_08_430070	163	3	6mA	6ma	CD
10_1101-2021_02_08_430070	163	4	dataset	dataset	NN
10_1101-2021_02_08_430070	163	5	,	,	,
10_1101-2021_02_08_430070	163	6	the	the	DT
10_1101-2021_02_08_430070	163	7	refined	refined	JJ
10_1101-2021_02_08_430070	163	8	BERT	BERT	NNP
10_1101-2021_02_08_430070	163	9	model	model	NN
10_1101-2021_02_08_430070	163	10	achieves	achieve	VBZ
10_1101-2021_02_08_430070	163	11	the	the	DT
10_1101-2021_02_08_430070	163	12	best	good	JJS
10_1101-2021_02_08_430070	163	13	AUC	AUC	NNP
10_1101-2021_02_08_430070	163	14	performance	performance	NN
10_1101-2021_02_08_430070	163	15	on	on	IN
10_1101-2021_02_08_430070	163	16	both	both	DT
10_1101-2021_02_08_430070	163	17	read	read	NN
10_1101-2021_02_08_430070	163	18	-	-	HYPH
10_1101-2021_02_08_430070	163	19	level	level	NN
10_1101-2021_02_08_430070	163	20	and	and	CC
10_1101-2021_02_08_430070	163	21	genomic	genomic	JJ
10_1101-2021_02_08_430070	163	22	-	-	HYPH
10_1101-2021_02_08_430070	163	23	level	level	NN
10_1101-2021_02_08_430070	163	24	.	.	.
10_1101-2021_02_08_430070	164	1	The	the	DT
10_1101-2021_02_08_430070	164	2	performance	performance	NN
10_1101-2021_02_08_430070	164	3	of	of	IN
10_1101-2021_02_08_430070	164	4	the	the	DT
10_1101-2021_02_08_430070	164	5	basic	basic	JJ
10_1101-2021_02_08_430070	164	6	BERT	BERT	NNP
10_1101-2021_02_08_430070	164	7	model	model	NN
10_1101-2021_02_08_430070	164	8	is	be	VBZ
10_1101-2021_02_08_430070	164	9	variant	variant	JJ
10_1101-2021_02_08_430070	164	10	and	and	CC
10_1101-2021_02_08_430070	164	11	unstable	unstable	JJ
10_1101-2021_02_08_430070	164	12	.	.	.
10_1101-2021_02_08_430070	165	1	On	on	IN
10_1101-2021_02_08_430070	165	2	Stobier_E.coli_gaAttc_EcoRI	Stobier_E.coli_gaAttc_EcoRI	NNP
10_1101-2021_02_08_430070	165	3	and	and	CC
10_1101-2021_02_08_430070	165	4	Stoiber_E.coli_gAtc_Dam	Stoiber_E.coli_gAtc_Dam	NNP
10_1101-2021_02_08_430070	165	5	,	,	,
10_1101-2021_02_08_430070	165	6	the	the	DT
10_1101-2021_02_08_430070	165	7	basic	basic	JJ
10_1101-2021_02_08_430070	165	8	BERT	BERT	NNP
10_1101-2021_02_08_430070	165	9	performs	perform	VBZ
10_1101-2021_02_08_430070	165	10	slightly	slightly	RB
10_1101-2021_02_08_430070	165	11	better	well	JJR
10_1101-2021_02_08_430070	165	12	than	than	IN
10_1101-2021_02_08_430070	165	13	biRNN	birnn	NN
10_1101-2021_02_08_430070	165	14	on	on	IN
10_1101-2021_02_08_430070	165	15	the	the	DT
10_1101-2021_02_08_430070	165	16	read	read	JJ
10_1101-2021_02_08_430070	165	17	-	-	HYPH
10_1101-2021_02_08_430070	165	18	level	level	NN
10_1101-2021_02_08_430070	165	19	AUC	AUC	NNP
10_1101-2021_02_08_430070	165	20	,	,	,
10_1101-2021_02_08_430070	165	21	but	but	CC
10_1101-2021_02_08_430070	165	22	has	have	VBZ
10_1101-2021_02_08_430070	165	23	a	a	DT
10_1101-2021_02_08_430070	165	24	large	large	JJ
10_1101-2021_02_08_430070	165	25	performance	performance	NN
10_1101-2021_02_08_430070	165	26	gap	gap	NN
10_1101-2021_02_08_430070	165	27	on	on	IN
10_1101-2021_02_08_430070	165	28	Stoiber_E.coli_gaAttc_EcoRI	stoiber_e.coli_gaattc_ecori	NN
10_1101-2021_02_08_430070	165	29	.	.	.
10_1101-2021_02_08_430070	166	1	In	in	IN
10_1101-2021_02_08_430070	166	2	summary	summary	NN
10_1101-2021_02_08_430070	166	3	,	,	,
10_1101-2021_02_08_430070	166	4	in	in	IN
10_1101-2021_02_08_430070	166	5	the	the	DT
10_1101-2021_02_08_430070	166	6	in	in	IN
10_1101-2021_02_08_430070	166	7	-	-	HYPH
10_1101-2021_02_08_430070	166	8	sample	sample	NN
10_1101-2021_02_08_430070	166	9	evaluation	evaluation	NN
10_1101-2021_02_08_430070	166	10	,	,	,
10_1101-2021_02_08_430070	166	11	the	the	DT
10_1101-2021_02_08_430070	166	12	refined	refined	JJ
10_1101-2021_02_08_430070	166	13	BERT	BERT	NNP
10_1101-2021_02_08_430070	166	14	model	model	NN
10_1101-2021_02_08_430070	166	15	can	can	MD
10_1101-2021_02_08_430070	166	16	achieve	achieve	VB
10_1101-2021_02_08_430070	166	17	competitive	competitive	JJ
10_1101-2021_02_08_430070	166	18	or	or	CC
10_1101-2021_02_08_430070	166	19	better	well	JJR
10_1101-2021_02_08_430070	166	20	results	result	NNS
10_1101-2021_02_08_430070	166	21	when	when	WRB
10_1101-2021_02_08_430070	166	22	compared	compare	VBN
10_1101-2021_02_08_430070	166	23	with	with	IN
10_1101-2021_02_08_430070	166	24	the	the	DT
10_1101-2021_02_08_430070	166	25	biRNN	biRNN	NNP
10_1101-2021_02_08_430070	166	26	model	model	NN
10_1101-2021_02_08_430070	166	27	on	on	IN
10_1101-2021_02_08_430070	166	28	benchmark	benchmark	JJ
10_1101-2021_02_08_430070	166	29	5mC	5mc	CD
10_1101-2021_02_08_430070	166	30	and	and	CC
10_1101-2021_02_08_430070	166	31	6mA	6mA	NNP
10_1101-2021_02_08_430070	166	32	datasets	dataset	NNS
10_1101-2021_02_08_430070	166	33	.	.	.
10_1101-2021_02_08_430070	167	1	3.4	3.4	CD
10_1101-2021_02_08_430070	167	2	Cross	cross	JJ
10_1101-2021_02_08_430070	167	3	-	-	JJ
10_1101-2021_02_08_430070	167	4	sample	sample	NN
10_1101-2021_02_08_430070	167	5	evaluation	evaluation	NN
10_1101-2021_02_08_430070	167	6	We	-PRON-	PRP
10_1101-2021_02_08_430070	167	7	then	then	RB
10_1101-2021_02_08_430070	167	8	conduct	conduct	VBP
10_1101-2021_02_08_430070	167	9	the	the	DT
10_1101-2021_02_08_430070	167	10	cross	cross	JJ
10_1101-2021_02_08_430070	167	11	-	-	JJ
10_1101-2021_02_08_430070	167	12	sample	sample	JJ
10_1101-2021_02_08_430070	167	13	evaluation	evaluation	NN
10_1101-2021_02_08_430070	167	14	.	.	.
10_1101-2021_02_08_430070	168	1	To	to	TO
10_1101-2021_02_08_430070	168	2	compare	compare	VB
10_1101-2021_02_08_430070	168	3	with	with	IN
10_1101-2021_02_08_430070	168	4	other	other	JJ
10_1101-2021_02_08_430070	168	5	non-	non-	NN
10_1101-2021_02_08_430070	168	6	deep	deep	RB
10_1101-2021_02_08_430070	168	7	-	-	HYPH
10_1101-2021_02_08_430070	168	8	learning	learn	VBG
10_1101-2021_02_08_430070	168	9	based	base	VBN
10_1101-2021_02_08_430070	168	10	methods	method	NNS
10_1101-2021_02_08_430070	168	11	,	,	,
10_1101-2021_02_08_430070	168	12	we	-PRON-	PRP
10_1101-2021_02_08_430070	168	13	utilize	utilize	VBP
10_1101-2021_02_08_430070	168	14	the	the	DT
10_1101-2021_02_08_430070	168	15	benchmark	benchmark	NN
10_1101-2021_02_08_430070	168	16	pipeline	pipeline	NN
10_1101-2021_02_08_430070	168	17	(	(	-LRB-
10_1101-2021_02_08_430070	168	18	Yuen	Yuen	NNP
10_1101-2021_02_08_430070	168	19	et	et	FW
10_1101-2021_02_08_430070	168	20	al	al	NNP
10_1101-2021_02_08_430070	168	21	.	.	NNP
10_1101-2021_02_08_430070	168	22	,	,	,
10_1101-2021_02_08_430070	168	23	2020	2020	CD
10_1101-2021_02_08_430070	168	24	)	)	-RRB-
10_1101-2021_02_08_430070	168	25	as	as	IN
10_1101-2021_02_08_430070	168	26	a	a	DT
10_1101-2021_02_08_430070	168	27	pivot	pivot	NN
10_1101-2021_02_08_430070	168	28	.	.	.
10_1101-2021_02_08_430070	169	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	169	2	test	test	VBP
10_1101-2021_02_08_430070	169	3	models	model	NNS
10_1101-2021_02_08_430070	169	4	on	on	IN
10_1101-2021_02_08_430070	169	5	the	the	DT
10_1101-2021_02_08_430070	169	6	same	same	JJ
10_1101-2021_02_08_430070	169	7	benchmark	benchmark	JJ
10_1101-2021_02_08_430070	169	8	dataset1	dataset1	NN
10_1101-2021_02_08_430070	169	9	,	,	,
10_1101-2021_02_08_430070	169	10	which	which	WDT
10_1101-2021_02_08_430070	169	11	is	be	VBZ
10_1101-2021_02_08_430070	169	12	generated	generate	VBN
10_1101-2021_02_08_430070	169	13	based	base	VBN
10_1101-2021_02_08_430070	169	14	on	on	IN
10_1101-2021_02_08_430070	169	15	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	169	16	’s	’s	POS
10_1101-2021_02_08_430070	169	17	E.coli	e.coli	JJ
10_1101-2021_02_08_430070	169	18	dataset	dataset	NN
10_1101-2021_02_08_430070	169	19	with	with	IN
10_1101-2021_02_08_430070	169	20	different	different	JJ
10_1101-2021_02_08_430070	169	21	methylation	methylation	NN
10_1101-2021_02_08_430070	169	22	levels	level	NNS
10_1101-2021_02_08_430070	169	23	.	.	.
10_1101-2021_02_08_430070	170	1	In	in	IN
10_1101-2021_02_08_430070	170	2	the	the	DT
10_1101-2021_02_08_430070	170	3	dataset	dataset	NN
10_1101-2021_02_08_430070	170	4	,	,	,
10_1101-2021_02_08_430070	170	5	100	100	CD
10_1101-2021_02_08_430070	170	6	arbitrary	arbitrary	JJ
10_1101-2021_02_08_430070	170	7	sites	site	NNS
10_1101-2021_02_08_430070	170	8	are	be	VBP
10_1101-2021_02_08_430070	170	9	selected	select	VBN
10_1101-2021_02_08_430070	170	10	,	,	,
10_1101-2021_02_08_430070	170	11	which	which	WDT
10_1101-2021_02_08_430070	170	12	contain	contain	VBP
10_1101-2021_02_08_430070	170	13	singleton	singleton	NNP
10_1101-2021_02_08_430070	170	14	CpG	CpG	NNP
10_1101-2021_02_08_430070	170	15	in	in	IN
10_1101-2021_02_08_430070	170	16	a	a	DT
10_1101-2021_02_08_430070	170	17	window	window	NN
10_1101-2021_02_08_430070	170	18	of	of	IN
10_1101-2021_02_08_430070	170	19	10nt	10nt	NN
10_1101-2021_02_08_430070	170	20	from	from	IN
10_1101-2021_02_08_430070	170	21	both	both	CC
10_1101-2021_02_08_430070	170	22	methylated	methylated	JJ
10_1101-2021_02_08_430070	170	23	and	and	CC
10_1101-2021_02_08_430070	170	24	unmethylated	unmethylated	JJ
10_1101-2021_02_08_430070	170	25	instances	instance	NNS
10_1101-2021_02_08_430070	170	26	in	in	IN
10_1101-2021_02_08_430070	170	27	the	the	DT
10_1101-2021_02_08_430070	170	28	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	170	29	’s	’s	POS
10_1101-2021_02_08_430070	170	30	E.coli	e.coli	JJ
10_1101-2021_02_08_430070	170	31	dataset	dataset	NN
10_1101-2021_02_08_430070	170	32	.	.	.
10_1101-2021_02_08_430070	171	1	Yuen	Yuen	NNP
10_1101-2021_02_08_430070	171	2	et	et	FW
10_1101-2021_02_08_430070	171	3	al	al	NNP
10_1101-2021_02_08_430070	171	4	.	.	.
10_1101-2021_02_08_430070	172	1	created	create	VBN
10_1101-2021_02_08_430070	172	2	11	11	CD
10_1101-2021_02_08_430070	172	3	specific	specific	JJ
10_1101-2021_02_08_430070	172	4	mixtures	mixture	NNS
10_1101-2021_02_08_430070	172	5	of	of	IN
10_1101-2021_02_08_430070	172	6	methylated	methylated	JJ
10_1101-2021_02_08_430070	172	7	and	and	CC
10_1101-2021_02_08_430070	172	8	unmethylated	unmethylated	JJ
10_1101-2021_02_08_430070	172	9	reads	read	NNS
10_1101-2021_02_08_430070	172	10	,	,	,
10_1101-2021_02_08_430070	172	11	containing	contain	VBG
10_1101-2021_02_08_430070	172	12	0	0	CD
10_1101-2021_02_08_430070	172	13	%	%	NN
10_1101-2021_02_08_430070	172	14	,	,	,
10_1101-2021_02_08_430070	172	15	10	10	CD
10_1101-2021_02_08_430070	172	16	%	%	NN
10_1101-2021_02_08_430070	172	17	,	,	,
10_1101-2021_02_08_430070	172	18	...	...	:
10_1101-2021_02_08_430070	172	19	,	,	,
10_1101-2021_02_08_430070	172	20	100	100	CD
10_1101-2021_02_08_430070	172	21	%	%	NN
10_1101-2021_02_08_430070	172	22	of	of	IN
10_1101-2021_02_08_430070	172	23	methylated	methylated	JJ
10_1101-2021_02_08_430070	172	24	reads	read	NNS
10_1101-2021_02_08_430070	172	25	.	.	.
10_1101-2021_02_08_430070	173	1	Each	each	DT
10_1101-2021_02_08_430070	173	2	mixture	mixture	NN
10_1101-2021_02_08_430070	173	3	contains	contain	VBZ
10_1101-2021_02_08_430070	173	4	approximately	approximately	RB
10_1101-2021_02_08_430070	173	5	2400	2400	CD
10_1101-2021_02_08_430070	173	6	reads	read	NNS
10_1101-2021_02_08_430070	173	7	.	.	.
10_1101-2021_02_08_430070	174	1	More	more	RBR
10_1101-2021_02_08_430070	174	2	detailed	detailed	JJ
10_1101-2021_02_08_430070	174	3	information	information	NN
10_1101-2021_02_08_430070	174	4	can	can	MD
10_1101-2021_02_08_430070	174	5	be	be	VB
10_1101-2021_02_08_430070	174	6	found	find	VBN
10_1101-2021_02_08_430070	174	7	in	in	IN
10_1101-2021_02_08_430070	174	8	(	(	-LRB-
10_1101-2021_02_08_430070	174	9	Yuen	Yuen	NNP
10_1101-2021_02_08_430070	174	10	et	et	FW
10_1101-2021_02_08_430070	174	11	al	al	NNP
10_1101-2021_02_08_430070	174	12	.	.	NNP
10_1101-2021_02_08_430070	174	13	,	,	,
10_1101-2021_02_08_430070	174	14	2020	2020	CD
10_1101-2021_02_08_430070	174	15	)	)	-RRB-
10_1101-2021_02_08_430070	174	16	.	.	.
10_1101-2021_02_08_430070	175	1	Different	different	JJ
10_1101-2021_02_08_430070	175	2	from	from	IN
10_1101-2021_02_08_430070	175	3	the	the	DT
10_1101-2021_02_08_430070	175	4	deepMOD	deepMOD	NNP
10_1101-2021_02_08_430070	175	5	model	model	NN
10_1101-2021_02_08_430070	175	6	used	use	VBN
10_1101-2021_02_08_430070	175	7	in	in	IN
10_1101-2021_02_08_430070	175	8	the	the	DT
10_1101-2021_02_08_430070	175	9	original	original	JJ
10_1101-2021_02_08_430070	175	10	benchmark	benchmark	NN
10_1101-2021_02_08_430070	175	11	pipeline	pipeline	NN
10_1101-2021_02_08_430070	175	12	,	,	,
10_1101-2021_02_08_430070	175	13	which	which	WDT
10_1101-2021_02_08_430070	175	14	is	be	VBZ
10_1101-2021_02_08_430070	175	15	pre	pre	VBN
10_1101-2021_02_08_430070	175	16	-	-	VBN
10_1101-2021_02_08_430070	175	17	trained	train	VBN
10_1101-2021_02_08_430070	175	18	on	on	IN
10_1101-2021_02_08_430070	175	19	a	a	DT
10_1101-2021_02_08_430070	175	20	mixture	mixture	NN
10_1101-2021_02_08_430070	175	21	dataset	dataset	NN
10_1101-2021_02_08_430070	175	22	of	of	IN
10_1101-2021_02_08_430070	175	23	all	all	DT
10_1101-2021_02_08_430070	175	24	5mC	5mc	CD
10_1101-2021_02_08_430070	175	25	positive	positive	JJ
10_1101-2021_02_08_430070	175	26	(	(	-LRB-
10_1101-2021_02_08_430070	175	27	Cg_SssI	Cg_SssI	NNP
10_1101-2021_02_08_430070	175	28	,	,	,
10_1101-2021_02_08_430070	175	29	Cg_MpeI	Cg_MpeI	NNP
10_1101-2021_02_08_430070	175	30	,	,	,
10_1101-2021_02_08_430070	175	31	and	and	CC
10_1101-2021_02_08_430070	175	32	gCgc_Hhal	gcgc_hhal	CD
10_1101-2021_02_08_430070	175	33	)	)	-RRB-
10_1101-2021_02_08_430070	175	34	and	and	CC
10_1101-2021_02_08_430070	175	35	negative	negative	JJ
10_1101-2021_02_08_430070	175	36	controls	control	NNS
10_1101-2021_02_08_430070	175	37	(	(	-LRB-
10_1101-2021_02_08_430070	175	38	UMR	UMR	NNP
10_1101-2021_02_08_430070	175	39	,	,	,
10_1101-2021_02_08_430070	175	40	con1	con1	NNP
10_1101-2021_02_08_430070	175	41	,	,	,
10_1101-2021_02_08_430070	175	42	and	and	CC
10_1101-2021_02_08_430070	175	43	con2	con2	NN
10_1101-2021_02_08_430070	175	44	)	)	-RRB-
10_1101-2021_02_08_430070	175	45	.	.	.
10_1101-2021_02_08_430070	176	1	Here	here	RB
10_1101-2021_02_08_430070	176	2	,	,	,
10_1101-2021_02_08_430070	176	3	we	-PRON-	PRP
10_1101-2021_02_08_430070	176	4	test	test	VBP
10_1101-2021_02_08_430070	176	5	two	two	CD
10_1101-2021_02_08_430070	176	6	different	different	JJ
10_1101-2021_02_08_430070	176	7	models	model	NNS
10_1101-2021_02_08_430070	176	8	trained	train	VBN
10_1101-2021_02_08_430070	176	9	on	on	IN
10_1101-2021_02_08_430070	176	10	a	a	DT
10_1101-2021_02_08_430070	176	11	single	single	JJ
10_1101-2021_02_08_430070	176	12	dataset	dataset	NN
10_1101-2021_02_08_430070	176	13	with	with	IN
10_1101-2021_02_08_430070	176	14	the	the	DT
10_1101-2021_02_08_430070	176	15	same	same	JJ
10_1101-2021_02_08_430070	176	16	methyltransferase	methyltransferase	NN
10_1101-2021_02_08_430070	176	17	to	to	TO
10_1101-2021_02_08_430070	176	18	reduce	reduce	VB
10_1101-2021_02_08_430070	176	19	potential	potential	JJ
10_1101-2021_02_08_430070	176	20	overlapping	overlapping	NN
10_1101-2021_02_08_430070	176	21	between	between	IN
10_1101-2021_02_08_430070	176	22	the	the	DT
10_1101-2021_02_08_430070	176	23	training	training	NN
10_1101-2021_02_08_430070	176	24	and	and	CC
10_1101-2021_02_08_430070	176	25	testing	testing	NN
10_1101-2021_02_08_430070	176	26	set	set	NN
10_1101-2021_02_08_430070	176	27	.	.	.
10_1101-2021_02_08_430070	177	1	All	all	DT
10_1101-2021_02_08_430070	177	2	three	three	CD
10_1101-2021_02_08_430070	177	3	models	model	NNS
10_1101-2021_02_08_430070	177	4	are	be	VBP
10_1101-2021_02_08_430070	177	5	trained	train	VBN
10_1101-2021_02_08_430070	177	6	on	on	IN
10_1101-2021_02_08_430070	177	7	Stoiber_Ecoli_CG_SssI	Stoiber_Ecoli_CG_SssI	NNP
10_1101-2021_02_08_430070	177	8	and	and	CC
10_1101-2021_02_08_430070	177	9	Simpson_Hsapiens_CG_SssI	Simpson_Hsapiens_CG_SssI	NNP
10_1101-2021_02_08_430070	177	10	,	,	,
10_1101-2021_02_08_430070	177	11	separately	separately	RB
10_1101-2021_02_08_430070	177	12	.	.	.
10_1101-2021_02_08_430070	178	1	Simpson_Hsapiens_CG_SssI	Simpson_Hsapiens_CG_SssI	NNP
10_1101-2021_02_08_430070	178	2	is	be	VBZ
10_1101-2021_02_08_430070	178	3	sequenced	sequence	VBN
10_1101-2021_02_08_430070	178	4	by	by	IN
10_1101-2021_02_08_430070	178	5	the	the	DT
10_1101-2021_02_08_430070	178	6	same	same	JJ
10_1101-2021_02_08_430070	178	7	group	group	NN
10_1101-2021_02_08_430070	178	8	on	on	IN
10_1101-2021_02_08_430070	178	9	different	different	JJ
10_1101-2021_02_08_430070	178	10	species	specie	NNS
10_1101-2021_02_08_430070	178	11	,	,	,
10_1101-2021_02_08_430070	178	12	while	while	IN
10_1101-2021_02_08_430070	178	13	Stoiber_Ecoli_CG_SssI	Stoiber_Ecoli_CG_SssI	NNP
10_1101-2021_02_08_430070	178	14	is	be	VBZ
10_1101-2021_02_08_430070	178	15	sequenced	sequence	VBN
10_1101-2021_02_08_430070	178	16	by	by	IN
10_1101-2021_02_08_430070	178	17	a	a	DT
10_1101-2021_02_08_430070	178	18	different	different	JJ
10_1101-2021_02_08_430070	178	19	group	group	NN
10_1101-2021_02_08_430070	178	20	on	on	IN
10_1101-2021_02_08_430070	178	21	the	the	DT
10_1101-2021_02_08_430070	178	22	same	same	JJ
10_1101-2021_02_08_430070	178	23	species	species	NN
10_1101-2021_02_08_430070	178	24	.	.	.
10_1101-2021_02_08_430070	179	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	179	2	use	use	VBP
10_1101-2021_02_08_430070	179	3	METEORE	METEORE	NNP
10_1101-2021_02_08_430070	179	4	pipeline	pipeline	NN
10_1101-2021_02_08_430070	179	5	(	(	-LRB-
10_1101-2021_02_08_430070	179	6	Yuen	Yuen	NNP
10_1101-2021_02_08_430070	179	7	et	et	FW
10_1101-2021_02_08_430070	179	8	al	al	NNP
10_1101-2021_02_08_430070	179	9	.	.	NNP
10_1101-2021_02_08_430070	179	10	,	,	,
10_1101-2021_02_08_430070	179	11	2020	2020	CD
10_1101-2021_02_08_430070	179	12	)	)	-RRB-
10_1101-2021_02_08_430070	179	13	to	to	TO
10_1101-2021_02_08_430070	179	14	generate	generate	VB
10_1101-2021_02_08_430070	179	15	violin	violin	NN
10_1101-2021_02_08_430070	179	16	plots	plot	NNS
10_1101-2021_02_08_430070	179	17	for	for	IN
10_1101-2021_02_08_430070	179	18	model	model	NN
10_1101-2021_02_08_430070	179	19	predictions	prediction	NNS
10_1101-2021_02_08_430070	179	20	on	on	IN
10_1101-2021_02_08_430070	179	21	each	each	DT
10_1101-2021_02_08_430070	179	22	mixture	mixture	NN
10_1101-2021_02_08_430070	179	23	.	.	.
10_1101-2021_02_08_430070	180	1	The	the	DT
10_1101-2021_02_08_430070	180	2	Pearson	Pearson	NNP
10_1101-2021_02_08_430070	180	3	’s	’s	POS
10_1101-2021_02_08_430070	180	4	correlation	correlation	NN
10_1101-2021_02_08_430070	180	5	r	r	NN
10_1101-2021_02_08_430070	180	6	,	,	,
10_1101-2021_02_08_430070	180	7	coefficient	coefficient	NN
10_1101-2021_02_08_430070	180	8	of	of	IN
10_1101-2021_02_08_430070	180	9	determination	determination	NN
10_1101-2021_02_08_430070	180	10	r2	r2	NN
10_1101-2021_02_08_430070	180	11	and	and	CC
10_1101-2021_02_08_430070	180	12	root	root	NN
10_1101-2021_02_08_430070	180	13	mean	mean	VBP
10_1101-2021_02_08_430070	180	14	square	square	JJ
10_1101-2021_02_08_430070	180	15	error	error	NN
10_1101-2021_02_08_430070	180	16	(	(	-LRB-
10_1101-2021_02_08_430070	180	17	RMSE	RMSE	NNP
10_1101-2021_02_08_430070	180	18	)	)	-RRB-
10_1101-2021_02_08_430070	180	19	are	be	VBP
10_1101-2021_02_08_430070	180	20	used	use	VBN
10_1101-2021_02_08_430070	180	21	as	as	IN
10_1101-2021_02_08_430070	180	22	the	the	DT
10_1101-2021_02_08_430070	180	23	evaluation	evaluation	NN
10_1101-2021_02_08_430070	180	24	metrics	metric	NNS
10_1101-2021_02_08_430070	180	25	for	for	IN
10_1101-2021_02_08_430070	180	26	each	each	DT
10_1101-2021_02_08_430070	180	27	model	model	NN
10_1101-2021_02_08_430070	180	28	.	.	.
10_1101-2021_02_08_430070	181	1	With	with	IN
10_1101-2021_02_08_430070	181	2	the	the	DT
10_1101-2021_02_08_430070	181	3	training	training	NN
10_1101-2021_02_08_430070	181	4	data	datum	NNS
10_1101-2021_02_08_430070	181	5	of	of	IN
10_1101-2021_02_08_430070	181	6	Simpson_Hsapiens_CG_SssI	Simpson_Hsapiens_CG_SssI	NNP
10_1101-2021_02_08_430070	181	7	,	,	,
10_1101-2021_02_08_430070	181	8	all	all	DT
10_1101-2021_02_08_430070	181	9	three	three	CD
10_1101-2021_02_08_430070	181	10	models	model	NNS
10_1101-2021_02_08_430070	181	11	achieve	achieve	VBP
10_1101-2021_02_08_430070	181	12	performances	performance	NNS
10_1101-2021_02_08_430070	181	13	ranked	rank	VBD
10_1101-2021_02_08_430070	181	14	next	next	RB
10_1101-2021_02_08_430070	181	15	to	to	IN
10_1101-2021_02_08_430070	181	16	the	the	DT
10_1101-2021_02_08_430070	181	17	best	good	JJS
10_1101-2021_02_08_430070	181	18	reported	report	VBN
10_1101-2021_02_08_430070	181	19	results	result	NNS
10_1101-2021_02_08_430070	181	20	of	of	IN
10_1101-2021_02_08_430070	181	21	Megalodon	Megalodon	NNP
10_1101-2021_02_08_430070	181	22	(	(	-LRB-
10_1101-2021_02_08_430070	181	23	r=0.9860	r=0.9860	NNP
10_1101-2021_02_08_430070	181	24	,	,	,
10_1101-2021_02_08_430070	181	25	r2	r2	NNP
10_1101-2021_02_08_430070	181	26	=	=	SYM
10_1101-2021_02_08_430070	181	27	0.9723	0.9723	CD
10_1101-2021_02_08_430070	181	28	,	,	,
10_1101-2021_02_08_430070	181	29	RMSE=0.0758	RMSE=0.0758	NNP
10_1101-2021_02_08_430070	181	30	)	)	-RRB-
10_1101-2021_02_08_430070	181	31	on	on	IN
10_1101-2021_02_08_430070	181	32	the	the	DT
10_1101-2021_02_08_430070	181	33	dataset	dataset	NN
10_1101-2021_02_08_430070	181	34	(	(	-LRB-
10_1101-2021_02_08_430070	181	35	Yuen	Yuen	NNP
10_1101-2021_02_08_430070	181	36	et	et	FW
10_1101-2021_02_08_430070	181	37	al	al	NNP
10_1101-2021_02_08_430070	181	38	.	.	NNP
10_1101-2021_02_08_430070	181	39	,	,	,
10_1101-2021_02_08_430070	181	40	2020	2020	CD
10_1101-2021_02_08_430070	181	41	)	)	-RRB-
10_1101-2021_02_08_430070	181	42	.	.	.
10_1101-2021_02_08_430070	182	1	BiRNN	birnn	DT
10_1101-2021_02_08_430070	182	2	achieves	achieve	VBZ
10_1101-2021_02_08_430070	182	3	the	the	DT
10_1101-2021_02_08_430070	182	4	best	good	JJS
10_1101-2021_02_08_430070	182	5	Pearson	Pearson	NNP
10_1101-2021_02_08_430070	182	6	correlation	correlation	NN
10_1101-2021_02_08_430070	182	7	r=0.9828	r=0.9828	NNP
10_1101-2021_02_08_430070	182	8	and	and	CC
10_1101-2021_02_08_430070	182	9	r2=0.9658	r2=0.9658	NNP
10_1101-2021_02_08_430070	182	10	,	,	,
10_1101-2021_02_08_430070	182	11	while	while	IN
10_1101-2021_02_08_430070	182	12	refine	refine	JJ
10_1101-2021_02_08_430070	182	13	BERT	BERT	NNP
10_1101-2021_02_08_430070	182	14	achieves	achieve	VBZ
10_1101-2021_02_08_430070	182	15	minimal	minimal	JJ
10_1101-2021_02_08_430070	182	16	RMSE	rmse	NN
10_1101-2021_02_08_430070	182	17	of	of	IN
10_1101-2021_02_08_430070	182	18	0.0732	0.0732	CD
10_1101-2021_02_08_430070	182	19	among	among	IN
10_1101-2021_02_08_430070	182	20	the	the	DT
10_1101-2021_02_08_430070	182	21	evaluated	evaluated	JJ
10_1101-2021_02_08_430070	182	22	three	three	CD
10_1101-2021_02_08_430070	182	23	models	model	NNS
10_1101-2021_02_08_430070	182	24	.	.	.
10_1101-2021_02_08_430070	183	1	When	when	WRB
10_1101-2021_02_08_430070	183	2	using	use	VBG
10_1101-2021_02_08_430070	183	3	Stoiber_Ecoli_CG_SssI	Stoiber_Ecoli_CG_SssI	NNP
10_1101-2021_02_08_430070	183	4	for	for	IN
10_1101-2021_02_08_430070	183	5	training	training	NN
10_1101-2021_02_08_430070	183	6	models	model	NNS
10_1101-2021_02_08_430070	183	7	,	,	,
10_1101-2021_02_08_430070	183	8	the	the	DT
10_1101-2021_02_08_430070	183	9	performances	performance	NNS
10_1101-2021_02_08_430070	183	10	of	of	IN
10_1101-2021_02_08_430070	183	11	all	all	DT
10_1101-2021_02_08_430070	183	12	three	three	CD
10_1101-2021_02_08_430070	183	13	models	model	NNS
10_1101-2021_02_08_430070	183	14	decrease	decrease	VBP
10_1101-2021_02_08_430070	183	15	.	.	.
10_1101-2021_02_08_430070	184	1	This	this	DT
10_1101-2021_02_08_430070	184	2	indicates	indicate	VBZ
10_1101-2021_02_08_430070	184	3	the	the	DT
10_1101-2021_02_08_430070	184	4	challenge	challenge	NN
10_1101-2021_02_08_430070	184	5	of	of	IN
10_1101-2021_02_08_430070	184	6	using	use	VBG
10_1101-2021_02_08_430070	184	7	datasets	dataset	NNS
10_1101-2021_02_08_430070	184	8	sequenced	sequence	VBN
10_1101-2021_02_08_430070	184	9	by	by	IN
10_1101-2021_02_08_430070	184	10	different	different	JJ
10_1101-2021_02_08_430070	184	11	research	research	NN
10_1101-2021_02_08_430070	184	12	groups	group	NNS
10_1101-2021_02_08_430070	184	13	.	.	.
10_1101-2021_02_08_430070	185	1	Here	here	RB
10_1101-2021_02_08_430070	185	2	,	,	,
10_1101-2021_02_08_430070	185	3	both	both	DT
10_1101-2021_02_08_430070	185	4	BERT	BERT	NNP
10_1101-2021_02_08_430070	185	5	models	model	NNS
10_1101-2021_02_08_430070	185	6	show	show	VBP
10_1101-2021_02_08_430070	185	7	better	well	JJR
10_1101-2021_02_08_430070	185	8	performances	performance	NNS
10_1101-2021_02_08_430070	185	9	than	than	IN
10_1101-2021_02_08_430070	185	10	biRNN	birnn	NN
10_1101-2021_02_08_430070	185	11	,	,	,
10_1101-2021_02_08_430070	185	12	as	as	IN
10_1101-2021_02_08_430070	185	13	in	in	IN
10_1101-2021_02_08_430070	185	14	Figure	Figure	NNP
10_1101-2021_02_08_430070	185	15	3b	3b	NNP
10_1101-2021_02_08_430070	185	16	.	.	.
10_1101-2021_02_08_430070	186	1	The	the	DT
10_1101-2021_02_08_430070	186	2	refined	refined	JJ
10_1101-2021_02_08_430070	186	3	.license	.license	.
10_1101-2021_02_08_430070	186	4	CC	cc	NN
10_1101-2021_02_08_430070	186	5	-	-	HYPH
10_1101-2021_02_08_430070	186	6	BY	BY	NNP
10_1101-2021_02_08_430070	186	7	-	-	HYPH
10_1101-2021_02_08_430070	186	8	NC	NC	NNP
10_1101-2021_02_08_430070	186	9	-	-	HYPH
10_1101-2021_02_08_430070	186	10	ND	ND	NNP
10_1101-2021_02_08_430070	186	11	4.0	4.0	CD
10_1101-2021_02_08_430070	186	12	Internationalpeer	Internationalpeer	NNP
10_1101-2021_02_08_430070	186	13	review	review	NN
10_1101-2021_02_08_430070	186	14	)	)	-RRB-
10_1101-2021_02_08_430070	186	15	is	be	VBZ
10_1101-2021_02_08_430070	186	16	the	the	DT
10_1101-2021_02_08_430070	186	17	author	author	NN
10_1101-2021_02_08_430070	186	18	/	/	SYM
10_1101-2021_02_08_430070	186	19	funder	funder	NN
10_1101-2021_02_08_430070	186	20	,	,	,
10_1101-2021_02_08_430070	186	21	who	who	WP
10_1101-2021_02_08_430070	186	22	has	have	VBZ
10_1101-2021_02_08_430070	186	23	granted	grant	VBN
10_1101-2021_02_08_430070	186	24	bioRxiv	biorxiv	IN
10_1101-2021_02_08_430070	186	25	a	a	DT
10_1101-2021_02_08_430070	186	26	license	license	NN
10_1101-2021_02_08_430070	186	27	to	to	TO
10_1101-2021_02_08_430070	186	28	display	display	VB
10_1101-2021_02_08_430070	186	29	the	the	DT
10_1101-2021_02_08_430070	186	30	preprint	preprint	NN
10_1101-2021_02_08_430070	186	31	in	in	IN
10_1101-2021_02_08_430070	186	32	perpetuity	perpetuity	NN
10_1101-2021_02_08_430070	186	33	.	.	.
10_1101-2021_02_08_430070	187	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	187	2	is	be	VBZ
10_1101-2021_02_08_430070	187	3	made	make	VBN
10_1101-2021_02_08_430070	187	4	available	available	JJ
10_1101-2021_02_08_430070	187	5	under	under	IN
10_1101-2021_02_08_430070	187	6	a	a	DT
10_1101-2021_02_08_430070	187	7	The	the	DT
10_1101-2021_02_08_430070	187	8	copyright	copyright	NN
10_1101-2021_02_08_430070	187	9	holder	holder	NN
10_1101-2021_02_08_430070	187	10	for	for	IN
10_1101-2021_02_08_430070	187	11	this	this	DT
10_1101-2021_02_08_430070	187	12	preprint	preprint	NN
10_1101-2021_02_08_430070	187	13	(	(	-LRB-
10_1101-2021_02_08_430070	187	14	which	which	WDT
10_1101-2021_02_08_430070	187	15	was	be	VBD
10_1101-2021_02_08_430070	187	16	not	not	RB
10_1101-2021_02_08_430070	187	17	certified	certify	VBN
10_1101-2021_02_08_430070	187	18	bythis	bythis	DT
10_1101-2021_02_08_430070	187	19	version	version	NN
10_1101-2021_02_08_430070	187	20	posted	post	VBD
10_1101-2021_02_08_430070	187	21	February	February	NNP
10_1101-2021_02_08_430070	187	22	10	10	CD
10_1101-2021_02_08_430070	187	23	,	,	,
10_1101-2021_02_08_430070	187	24	2021	2021	CD
10_1101-2021_02_08_430070	187	25	.	.	.
10_1101-2021_02_08_430070	187	26	;	;	:
10_1101-2021_02_08_430070	187	27	https://doi.org/10.1101/2021.02.08.430070doi	https://doi.org/10.1101/2021.02.08.430070doi	NFP
10_1101-2021_02_08_430070	187	28	:	:	:
10_1101-2021_02_08_430070	187	29	bioRxiv	biorxiv	VB
10_1101-2021_02_08_430070	187	30	preprint	preprint	NN
10_1101-2021_02_08_430070	187	31	https://doi.org/10.1101/2021.02.08.430070	https://doi.org/10.1101/2021.02.08.430070	UH
10_1101-2021_02_08_430070	187	32	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	187	33	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	187	34	✐	✐	NNP
10_1101-2021_02_08_430070	187	35	✐	✐	NNP
10_1101-2021_02_08_430070	187	36	✐	✐	NNP
10_1101-2021_02_08_430070	187	37	✐	✐	NNP
10_1101-2021_02_08_430070	187	38	✐	✐	NNP
10_1101-2021_02_08_430070	187	39	✐	✐	NNP
10_1101-2021_02_08_430070	187	40	✐	✐	NNP
10_1101-2021_02_08_430070	187	41	✐	✐	NNP
10_1101-2021_02_08_430070	187	42	6	6	CD
10_1101-2021_02_08_430070	187	43	Zhang	Zhang	NNP
10_1101-2021_02_08_430070	187	44	et	et	NNP
10_1101-2021_02_08_430070	187	45	al	al	NNP
10_1101-2021_02_08_430070	187	46	.	.	.
10_1101-2021_02_08_430070	188	1	(	(	-LRB-
10_1101-2021_02_08_430070	188	2	a	a	LS
10_1101-2021_02_08_430070	188	3	)	)	-RRB-
10_1101-2021_02_08_430070	188	4	Models	model	NNS
10_1101-2021_02_08_430070	188	5	trained	train	VBN
10_1101-2021_02_08_430070	188	6	with	with	IN
10_1101-2021_02_08_430070	188	7	Simpson_Hsapiens_CG_SssI	simpson_hsapiens_cg_sssi	JJ
10_1101-2021_02_08_430070	188	8	dataset	dataset	NN
10_1101-2021_02_08_430070	188	9	.	.	.
10_1101-2021_02_08_430070	189	1	(	(	-LRB-
10_1101-2021_02_08_430070	189	2	b	b	LS
10_1101-2021_02_08_430070	189	3	)	)	-RRB-
10_1101-2021_02_08_430070	189	4	Models	model	NNS
10_1101-2021_02_08_430070	189	5	trained	train	VBN
10_1101-2021_02_08_430070	189	6	with	with	IN
10_1101-2021_02_08_430070	189	7	Stoiber_Ecoli_CG_SssI	Stoiber_Ecoli_CG_SssI	NNP
10_1101-2021_02_08_430070	189	8	dataset	dataset	NN
10_1101-2021_02_08_430070	189	9	.	.	.
10_1101-2021_02_08_430070	190	1	Fig	fig	NN
10_1101-2021_02_08_430070	190	2	.	.	.
10_1101-2021_02_08_430070	191	1	3	3	LS
10_1101-2021_02_08_430070	191	2	:	:	:
10_1101-2021_02_08_430070	191	3	Violin	Violin	NNP
10_1101-2021_02_08_430070	191	4	plots	plot	NNS
10_1101-2021_02_08_430070	191	5	of	of	IN
10_1101-2021_02_08_430070	191	6	prediction	prediction	NN
10_1101-2021_02_08_430070	191	7	results	result	NNS
10_1101-2021_02_08_430070	191	8	of	of	IN
10_1101-2021_02_08_430070	191	9	models	model	NNS
10_1101-2021_02_08_430070	191	10	trained	train	VBN
10_1101-2021_02_08_430070	191	11	on	on	IN
10_1101-2021_02_08_430070	191	12	different	different	JJ
10_1101-2021_02_08_430070	191	13	datasets	dataset	NNS
10_1101-2021_02_08_430070	191	14	.	.	.
10_1101-2021_02_08_430070	192	1	BERT	BERT	NNP
10_1101-2021_02_08_430070	192	2	achieves	achieve	VBZ
10_1101-2021_02_08_430070	192	3	the	the	DT
10_1101-2021_02_08_430070	192	4	best	good	JJS
10_1101-2021_02_08_430070	192	5	r=0.9446	r=0.9446	NNPS
10_1101-2021_02_08_430070	192	6	,	,	,
10_1101-2021_02_08_430070	192	7	r2=0.8924	r2=0.8924	NNP
10_1101-2021_02_08_430070	192	8	and	and	CC
10_1101-2021_02_08_430070	192	9	RMSE	RMSE	NNP
10_1101-2021_02_08_430070	192	10	of	of	IN
10_1101-2021_02_08_430070	192	11	0.1449	0.1449	CD
10_1101-2021_02_08_430070	192	12	among	among	IN
10_1101-2021_02_08_430070	192	13	the	the	DT
10_1101-2021_02_08_430070	192	14	three	three	CD
10_1101-2021_02_08_430070	192	15	models	model	NNS
10_1101-2021_02_08_430070	192	16	,	,	,
10_1101-2021_02_08_430070	192	17	which	which	WDT
10_1101-2021_02_08_430070	192	18	demonstrate	demonstrate	VBP
10_1101-2021_02_08_430070	192	19	the	the	DT
10_1101-2021_02_08_430070	192	20	generalization	generalization	NN
10_1101-2021_02_08_430070	192	21	ability	ability	NN
10_1101-2021_02_08_430070	192	22	on	on	IN
10_1101-2021_02_08_430070	192	23	datasets	dataset	NNS
10_1101-2021_02_08_430070	192	24	sequenced	sequence	VBN
10_1101-2021_02_08_430070	192	25	by	by	IN
10_1101-2021_02_08_430070	192	26	different	different	JJ
10_1101-2021_02_08_430070	192	27	research	research	NN
10_1101-2021_02_08_430070	192	28	groups	group	NNS
10_1101-2021_02_08_430070	192	29	.	.	.
10_1101-2021_02_08_430070	193	1	Based	base	VBN
10_1101-2021_02_08_430070	193	2	on	on	IN
10_1101-2021_02_08_430070	193	3	the	the	DT
10_1101-2021_02_08_430070	193	4	reported	report	VBN
10_1101-2021_02_08_430070	193	5	benchmark	benchmark	JJ
10_1101-2021_02_08_430070	193	6	results	result	NNS
10_1101-2021_02_08_430070	193	7	,	,	,
10_1101-2021_02_08_430070	193	8	the	the	DT
10_1101-2021_02_08_430070	193	9	Pearson	Pearson	NNP
10_1101-2021_02_08_430070	193	10	correlation	correlation	NN
10_1101-2021_02_08_430070	193	11	ranks	rank	NNS
10_1101-2021_02_08_430070	193	12	between	between	IN
10_1101-2021_02_08_430070	193	13	reported	report	VBN
10_1101-2021_02_08_430070	193	14	deepMOD	deepMOD	NNP
10_1101-2021_02_08_430070	193	15	and	and	CC
10_1101-2021_02_08_430070	193	16	deepSignal	deepsignal	JJ
10_1101-2021_02_08_430070	193	17	(	(	-LRB-
10_1101-2021_02_08_430070	193	18	Megalodon	Megalodon	NNP
10_1101-2021_02_08_430070	193	19	>	>	XX
10_1101-2021_02_08_430070	193	20	DeepMODmixModel	DeepMODmixModel	NNP
10_1101-2021_02_08_430070	193	21	(	(	-LRB-
10_1101-2021_02_08_430070	193	22	0.9467	0.9467	CD
10_1101-2021_02_08_430070	193	23	)	)	-RRB-
10_1101-2021_02_08_430070	193	24	>	>	XX
10_1101-2021_02_08_430070	193	25	refined	refined	JJ
10_1101-2021_02_08_430070	193	26	BERT	BERT	NNP
10_1101-2021_02_08_430070	193	27	>	>	XX
10_1101-2021_02_08_430070	193	28	DeepSignalhuman_hx1	DeepSignalhuman_hx1	NNP
10_1101-2021_02_08_430070	193	29	(	(	-LRB-
10_1101-2021_02_08_430070	193	30	0.9420	0.9420	CD
10_1101-2021_02_08_430070	193	31	)	)	-RRB-
10_1101-2021_02_08_430070	193	32	>	>	XX
10_1101-2021_02_08_430070	193	33	Guppy	Guppy	NNP
10_1101-2021_02_08_430070	193	34	>	>	XX
10_1101-2021_02_08_430070	193	35	Nanopolish	nanopolish	JJ
10_1101-2021_02_08_430070	193	36	>	>	NN
10_1101-2021_02_08_430070	193	37	Tombo	Tombo	NNP
10_1101-2021_02_08_430070	193	38	)	)	-RRB-
10_1101-2021_02_08_430070	193	39	.	.	.
10_1101-2021_02_08_430070	194	1	3.5	3.5	CD
10_1101-2021_02_08_430070	194	2	Model	model	NN
10_1101-2021_02_08_430070	194	3	inference	inference	NN
10_1101-2021_02_08_430070	194	4	speed	speed	NN
10_1101-2021_02_08_430070	194	5	The	the	DT
10_1101-2021_02_08_430070	194	6	main	main	JJ
10_1101-2021_02_08_430070	194	7	motivation	motivation	NN
10_1101-2021_02_08_430070	194	8	of	of	IN
10_1101-2021_02_08_430070	194	9	applying	apply	VBG
10_1101-2021_02_08_430070	194	10	BERT	BERT	NNP
10_1101-2021_02_08_430070	194	11	models	model	NNS
10_1101-2021_02_08_430070	194	12	is	be	VBZ
10_1101-2021_02_08_430070	194	13	to	to	TO
10_1101-2021_02_08_430070	194	14	use	use	VB
10_1101-2021_02_08_430070	194	15	a	a	DT
10_1101-2021_02_08_430070	194	16	non	non	JJ
10_1101-2021_02_08_430070	194	17	-	-	JJ
10_1101-2021_02_08_430070	194	18	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	194	19	modeling	modeling	NN
10_1101-2021_02_08_430070	194	20	approach	approach	NN
10_1101-2021_02_08_430070	194	21	for	for	IN
10_1101-2021_02_08_430070	194	22	the	the	DT
10_1101-2021_02_08_430070	194	23	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	194	24	methylation	methylation	NN
10_1101-2021_02_08_430070	194	25	detection	detection	NN
10_1101-2021_02_08_430070	194	26	task	task	NN
10_1101-2021_02_08_430070	194	27	to	to	TO
10_1101-2021_02_08_430070	194	28	improve	improve	VB
10_1101-2021_02_08_430070	194	29	the	the	DT
10_1101-2021_02_08_430070	194	30	model	model	NN
10_1101-2021_02_08_430070	194	31	inference	inference	NN
10_1101-2021_02_08_430070	194	32	speed	speed	NN
10_1101-2021_02_08_430070	194	33	.	.	.
10_1101-2021_02_08_430070	195	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	195	2	performed	perform	VBD
10_1101-2021_02_08_430070	195	3	a	a	DT
10_1101-2021_02_08_430070	195	4	speed	speed	NN
10_1101-2021_02_08_430070	195	5	test	test	NN
10_1101-2021_02_08_430070	195	6	on	on	IN
10_1101-2021_02_08_430070	195	7	a	a	DT
10_1101-2021_02_08_430070	195	8	server	server	NN
10_1101-2021_02_08_430070	195	9	with	with	IN
10_1101-2021_02_08_430070	195	10	24	24	CD
10_1101-2021_02_08_430070	195	11	CPU	cpu	NN
10_1101-2021_02_08_430070	195	12	cores	core	NNS
10_1101-2021_02_08_430070	195	13	(	(	-LRB-
10_1101-2021_02_08_430070	195	14	Intel(R	intel(r	NN
10_1101-2021_02_08_430070	195	15	)	)	-RRB-
10_1101-2021_02_08_430070	195	16	Xeon(R	Xeon(R	NNP
10_1101-2021_02_08_430070	195	17	)	)	-RRB-
10_1101-2021_02_08_430070	195	18	Gold	gold	NN
10_1101-2021_02_08_430070	195	19	6126	6126	CD
10_1101-2021_02_08_430070	195	20	CPU	CPU	NNP
10_1101-2021_02_08_430070	195	21	@	@	SYM
10_1101-2021_02_08_430070	195	22	2.60GHz	2.60ghz	CD
10_1101-2021_02_08_430070	195	23	)	)	-RRB-
10_1101-2021_02_08_430070	195	24	and	and	CC
10_1101-2021_02_08_430070	195	25	.license	.license	NNP
10_1101-2021_02_08_430070	195	26	CC	cc	NN
10_1101-2021_02_08_430070	195	27	-	-	HYPH
10_1101-2021_02_08_430070	195	28	BY	BY	NNP
10_1101-2021_02_08_430070	195	29	-	-	HYPH
10_1101-2021_02_08_430070	195	30	NC	NC	NNP
10_1101-2021_02_08_430070	195	31	-	-	HYPH
10_1101-2021_02_08_430070	195	32	ND	ND	NNP
10_1101-2021_02_08_430070	195	33	4.0	4.0	CD
10_1101-2021_02_08_430070	195	34	Internationalpeer	Internationalpeer	NNP
10_1101-2021_02_08_430070	195	35	review	review	NN
10_1101-2021_02_08_430070	195	36	)	)	-RRB-
10_1101-2021_02_08_430070	195	37	is	be	VBZ
10_1101-2021_02_08_430070	195	38	the	the	DT
10_1101-2021_02_08_430070	195	39	author	author	NN
10_1101-2021_02_08_430070	195	40	/	/	SYM
10_1101-2021_02_08_430070	195	41	funder	funder	NN
10_1101-2021_02_08_430070	195	42	,	,	,
10_1101-2021_02_08_430070	195	43	who	who	WP
10_1101-2021_02_08_430070	195	44	has	have	VBZ
10_1101-2021_02_08_430070	195	45	granted	grant	VBN
10_1101-2021_02_08_430070	195	46	bioRxiv	biorxiv	IN
10_1101-2021_02_08_430070	195	47	a	a	DT
10_1101-2021_02_08_430070	195	48	license	license	NN
10_1101-2021_02_08_430070	195	49	to	to	TO
10_1101-2021_02_08_430070	195	50	display	display	VB
10_1101-2021_02_08_430070	195	51	the	the	DT
10_1101-2021_02_08_430070	195	52	preprint	preprint	NN
10_1101-2021_02_08_430070	195	53	in	in	IN
10_1101-2021_02_08_430070	195	54	perpetuity	perpetuity	NN
10_1101-2021_02_08_430070	195	55	.	.	.
10_1101-2021_02_08_430070	196	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	196	2	is	be	VBZ
10_1101-2021_02_08_430070	196	3	made	make	VBN
10_1101-2021_02_08_430070	196	4	available	available	JJ
10_1101-2021_02_08_430070	196	5	under	under	IN
10_1101-2021_02_08_430070	196	6	a	a	DT
10_1101-2021_02_08_430070	196	7	The	the	DT
10_1101-2021_02_08_430070	196	8	copyright	copyright	NN
10_1101-2021_02_08_430070	196	9	holder	holder	NN
10_1101-2021_02_08_430070	196	10	for	for	IN
10_1101-2021_02_08_430070	196	11	this	this	DT
10_1101-2021_02_08_430070	196	12	preprint	preprint	NN
10_1101-2021_02_08_430070	196	13	(	(	-LRB-
10_1101-2021_02_08_430070	196	14	which	which	WDT
10_1101-2021_02_08_430070	196	15	was	be	VBD
10_1101-2021_02_08_430070	196	16	not	not	RB
10_1101-2021_02_08_430070	196	17	certified	certify	VBN
10_1101-2021_02_08_430070	196	18	bythis	bythis	DT
10_1101-2021_02_08_430070	196	19	version	version	NN
10_1101-2021_02_08_430070	196	20	posted	post	VBD
10_1101-2021_02_08_430070	196	21	February	February	NNP
10_1101-2021_02_08_430070	196	22	10	10	CD
10_1101-2021_02_08_430070	196	23	,	,	,
10_1101-2021_02_08_430070	196	24	2021	2021	CD
10_1101-2021_02_08_430070	196	25	.	.	.
10_1101-2021_02_08_430070	196	26	;	;	:
10_1101-2021_02_08_430070	196	27	https://doi.org/10.1101/2021.02.08.430070doi	https://doi.org/10.1101/2021.02.08.430070doi	NFP
10_1101-2021_02_08_430070	196	28	:	:	:
10_1101-2021_02_08_430070	196	29	bioRxiv	biorxiv	VB
10_1101-2021_02_08_430070	196	30	preprint	preprint	NN
10_1101-2021_02_08_430070	196	31	https://doi.org/10.1101/2021.02.08.430070	https://doi.org/10.1101/2021.02.08.430070	UH
10_1101-2021_02_08_430070	196	32	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	196	33	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	196	34	✐	✐	NNP
10_1101-2021_02_08_430070	196	35	✐	✐	NNP
10_1101-2021_02_08_430070	196	36	✐	✐	NNP
10_1101-2021_02_08_430070	196	37	✐	✐	NNP
10_1101-2021_02_08_430070	196	38	✐	✐	NNP
10_1101-2021_02_08_430070	196	39	✐	✐	NNP
10_1101-2021_02_08_430070	196	40	✐	✐	NNP
10_1101-2021_02_08_430070	196	41	✐	✐	NNP
10_1101-2021_02_08_430070	196	42	BERT	BERT	NNP
10_1101-2021_02_08_430070	196	43	for	for	IN
10_1101-2021_02_08_430070	196	44	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	196	45	methylation	methylation	NN
10_1101-2021_02_08_430070	196	46	detection	detection	NN
10_1101-2021_02_08_430070	196	47	7	7	CD
10_1101-2021_02_08_430070	196	48	Model	Model	NNP
10_1101-2021_02_08_430070	196	49	Model	Model	NNP
10_1101-2021_02_08_430070	196	50	inference	inference	NN
10_1101-2021_02_08_430070	196	51	time	time	NN
10_1101-2021_02_08_430070	196	52	Total	total	JJ
10_1101-2021_02_08_430070	196	53	running	running	NN
10_1101-2021_02_08_430070	196	54	time	time	NN
10_1101-2021_02_08_430070	196	55	biRNN	birnn	NN
10_1101-2021_02_08_430070	196	56	162.91	162.91	NNP
10_1101-2021_02_08_430070	196	57	s	s	NNP
10_1101-2021_02_08_430070	196	58	711.56	711.56	NNP
10_1101-2021_02_08_430070	196	59	s	s	NNP
10_1101-2021_02_08_430070	196	60	BERT_basic	BERT_basic	NNP
10_1101-2021_02_08_430070	196	61	22.71	22.71	NNP
10_1101-2021_02_08_430070	196	62	s	s	NNP
10_1101-2021_02_08_430070	196	63	615.36	615.36	CD
10_1101-2021_02_08_430070	196	64	s	s	NNP
10_1101-2021_02_08_430070	196	65	BERT_refined	BERT_refined	NNP
10_1101-2021_02_08_430070	196	66	27.29	27.29	CD
10_1101-2021_02_08_430070	196	67	s	s	NNPS
10_1101-2021_02_08_430070	196	68	622.73	622.73	NNP
10_1101-2021_02_08_430070	196	69	s	s	NNP
10_1101-2021_02_08_430070	196	70	Table	table	NN
10_1101-2021_02_08_430070	196	71	3	3	CD
10_1101-2021_02_08_430070	196	72	.	.	.
10_1101-2021_02_08_430070	197	1	Model	model	NN
10_1101-2021_02_08_430070	197	2	inference	inference	NN
10_1101-2021_02_08_430070	197	3	and	and	CC
10_1101-2021_02_08_430070	197	4	total	total	JJ
10_1101-2021_02_08_430070	197	5	running	running	NN
10_1101-2021_02_08_430070	197	6	time	time	NN
10_1101-2021_02_08_430070	197	7	on	on	IN
10_1101-2021_02_08_430070	197	8	the	the	DT
10_1101-2021_02_08_430070	197	9	benchmark	benchmark	JJ
10_1101-2021_02_08_430070	197	10	dataset1	dataset1	NN
10_1101-2021_02_08_430070	197	11	for	for	IN
10_1101-2021_02_08_430070	197	12	all	all	DT
10_1101-2021_02_08_430070	197	13	26402	26402	CD
10_1101-2021_02_08_430070	197	14	reads	read	NNS
10_1101-2021_02_08_430070	197	15	.	.	.
10_1101-2021_02_08_430070	198	1	one	one	CD
10_1101-2021_02_08_430070	198	2	V100	V100	NNP
10_1101-2021_02_08_430070	198	3	NIVIDA	NIVIDA	NNP
10_1101-2021_02_08_430070	198	4	GPU	GPU	NNP
10_1101-2021_02_08_430070	198	5	card	card	NN
10_1101-2021_02_08_430070	198	6	.	.	.
10_1101-2021_02_08_430070	199	1	In	in	IN
10_1101-2021_02_08_430070	199	2	the	the	DT
10_1101-2021_02_08_430070	199	3	running	running	NN
10_1101-2021_02_08_430070	199	4	,	,	,
10_1101-2021_02_08_430070	199	5	CPUs	cpu	NNS
10_1101-2021_02_08_430070	199	6	are	be	VBP
10_1101-2021_02_08_430070	199	7	responsible	responsible	JJ
10_1101-2021_02_08_430070	199	8	for	for	IN
10_1101-2021_02_08_430070	199	9	data	datum	NNS
10_1101-2021_02_08_430070	199	10	loading	loading	NN
10_1101-2021_02_08_430070	199	11	and	and	CC
10_1101-2021_02_08_430070	199	12	feature	feature	NN
10_1101-2021_02_08_430070	199	13	extraction	extraction	NN
10_1101-2021_02_08_430070	199	14	,	,	,
10_1101-2021_02_08_430070	199	15	while	while	IN
10_1101-2021_02_08_430070	199	16	GPU	GPU	NNP
10_1101-2021_02_08_430070	199	17	works	work	VBZ
10_1101-2021_02_08_430070	199	18	for	for	IN
10_1101-2021_02_08_430070	199	19	model	model	NN
10_1101-2021_02_08_430070	199	20	inference	inference	NN
10_1101-2021_02_08_430070	199	21	.	.	.
10_1101-2021_02_08_430070	200	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	200	2	tested	test	VBD
10_1101-2021_02_08_430070	200	3	the	the	DT
10_1101-2021_02_08_430070	200	4	model	model	NN
10_1101-2021_02_08_430070	200	5	inference	inference	NN
10_1101-2021_02_08_430070	200	6	time	time	NN
10_1101-2021_02_08_430070	200	7	and	and	CC
10_1101-2021_02_08_430070	200	8	total	total	JJ
10_1101-2021_02_08_430070	200	9	running	running	NN
10_1101-2021_02_08_430070	200	10	time	time	NN
10_1101-2021_02_08_430070	200	11	of	of	IN
10_1101-2021_02_08_430070	200	12	the	the	DT
10_1101-2021_02_08_430070	200	13	three	three	CD
10_1101-2021_02_08_430070	200	14	models	model	NNS
10_1101-2021_02_08_430070	200	15	on	on	IN
10_1101-2021_02_08_430070	200	16	the	the	DT
10_1101-2021_02_08_430070	200	17	benchmark	benchmark	JJ
10_1101-2021_02_08_430070	200	18	dataset1	dataset1	NN
10_1101-2021_02_08_430070	200	19	.	.	.
10_1101-2021_02_08_430070	201	1	For	for	IN
10_1101-2021_02_08_430070	201	2	each	each	DT
10_1101-2021_02_08_430070	201	3	mixture	mixture	NN
10_1101-2021_02_08_430070	201	4	split	split	NN
10_1101-2021_02_08_430070	201	5	,	,	,
10_1101-2021_02_08_430070	201	6	we	-PRON-	PRP
10_1101-2021_02_08_430070	201	7	repeated	repeat	VBD
10_1101-2021_02_08_430070	201	8	5	5	CD
10_1101-2021_02_08_430070	201	9	times	time	NNS
10_1101-2021_02_08_430070	201	10	running	run	VBG
10_1101-2021_02_08_430070	201	11	and	and	CC
10_1101-2021_02_08_430070	201	12	took	take	VBD
10_1101-2021_02_08_430070	201	13	the	the	DT
10_1101-2021_02_08_430070	201	14	averaged	average	VBN
10_1101-2021_02_08_430070	201	15	value	value	NN
10_1101-2021_02_08_430070	201	16	.	.	.
10_1101-2021_02_08_430070	202	1	As	as	IN
10_1101-2021_02_08_430070	202	2	shown	show	VBN
10_1101-2021_02_08_430070	202	3	in	in	IN
10_1101-2021_02_08_430070	202	4	Table	table	NN
10_1101-2021_02_08_430070	202	5	3	3	CD
10_1101-2021_02_08_430070	202	6	,	,	,
10_1101-2021_02_08_430070	202	7	the	the	DT
10_1101-2021_02_08_430070	202	8	model	model	NN
10_1101-2021_02_08_430070	202	9	inference	inference	NN
10_1101-2021_02_08_430070	202	10	speed	speed	NN
10_1101-2021_02_08_430070	202	11	of	of	IN
10_1101-2021_02_08_430070	202	12	BERT	BERT	NNP
10_1101-2021_02_08_430070	202	13	models	model	NNS
10_1101-2021_02_08_430070	202	14	is	be	VBZ
10_1101-2021_02_08_430070	202	15	around	around	RB
10_1101-2021_02_08_430070	202	16	6x∼7x	6x∼7x	CD
10_1101-2021_02_08_430070	202	17	faster	fast	JJR
10_1101-2021_02_08_430070	202	18	than	than	IN
10_1101-2021_02_08_430070	202	19	biRNN	biRNN	NNP
10_1101-2021_02_08_430070	202	20	model	model	NNP
10_1101-2021_02_08_430070	202	21	(	(	-LRB-
10_1101-2021_02_08_430070	202	22	BERT_refined:5.96x	BERT_refined:5.96x	NNP
10_1101-2021_02_08_430070	202	23	,	,	,
10_1101-2021_02_08_430070	202	24	BERT_basic:7.16x	BERT_basic:7.16x	NNP
10_1101-2021_02_08_430070	202	25	)	)	-RRB-
10_1101-2021_02_08_430070	202	26	.	.	.
10_1101-2021_02_08_430070	203	1	The	the	DT
10_1101-2021_02_08_430070	203	2	inference	inference	NN
10_1101-2021_02_08_430070	203	3	time	time	NN
10_1101-2021_02_08_430070	203	4	of	of	IN
10_1101-2021_02_08_430070	203	5	refined	refined	JJ
10_1101-2021_02_08_430070	203	6	BERT	BERT	NNP
10_1101-2021_02_08_430070	203	7	is	be	VBZ
10_1101-2021_02_08_430070	203	8	only	only	RB
10_1101-2021_02_08_430070	203	9	slightly	slightly	RB
10_1101-2021_02_08_430070	203	10	slower	slow	JJR
10_1101-2021_02_08_430070	203	11	than	than	IN
10_1101-2021_02_08_430070	203	12	the	the	DT
10_1101-2021_02_08_430070	203	13	basic	basic	JJ
10_1101-2021_02_08_430070	203	14	BERT	BERT	NNP
10_1101-2021_02_08_430070	203	15	model	model	NN
10_1101-2021_02_08_430070	203	16	.	.	.
10_1101-2021_02_08_430070	204	1	The	the	DT
10_1101-2021_02_08_430070	204	2	gap	gap	NN
10_1101-2021_02_08_430070	204	3	of	of	IN
10_1101-2021_02_08_430070	204	4	the	the	DT
10_1101-2021_02_08_430070	204	5	total	total	JJ
10_1101-2021_02_08_430070	204	6	time	time	NN
10_1101-2021_02_08_430070	204	7	is	be	VBZ
10_1101-2021_02_08_430070	204	8	not	not	RB
10_1101-2021_02_08_430070	204	9	that	that	RB
10_1101-2021_02_08_430070	204	10	large	large	JJ
10_1101-2021_02_08_430070	204	11	(	(	-LRB-
10_1101-2021_02_08_430070	204	12	BERT_refined:1.14x	BERT_refined:1.14x	NNP
10_1101-2021_02_08_430070	204	13	,	,	,
10_1101-2021_02_08_430070	204	14	BERT_basic:1.16x	bert_basic:1.16x	ADD
10_1101-2021_02_08_430070	204	15	)	)	-RRB-
10_1101-2021_02_08_430070	204	16	,	,	,
10_1101-2021_02_08_430070	204	17	as	as	IN
10_1101-2021_02_08_430070	204	18	the	the	DT
10_1101-2021_02_08_430070	204	19	data	datum	NNS
10_1101-2021_02_08_430070	204	20	I	-PRON-	PRP
10_1101-2021_02_08_430070	204	21	/	/	SYM
10_1101-2021_02_08_430070	204	22	O	o	NN
10_1101-2021_02_08_430070	204	23	and	and	CC
10_1101-2021_02_08_430070	204	24	feature	feature	NN
10_1101-2021_02_08_430070	204	25	extraction	extraction	NN
10_1101-2021_02_08_430070	204	26	take	take	VBP
10_1101-2021_02_08_430070	204	27	major	major	JJ
10_1101-2021_02_08_430070	204	28	time	time	NN
10_1101-2021_02_08_430070	204	29	.	.	.
10_1101-2021_02_08_430070	205	1	In	in	IN
10_1101-2021_02_08_430070	205	2	the	the	DT
10_1101-2021_02_08_430070	205	3	current	current	JJ
10_1101-2021_02_08_430070	205	4	implementation	implementation	NN
10_1101-2021_02_08_430070	205	5	of	of	IN
10_1101-2021_02_08_430070	205	6	BERT	BERT	NNP
10_1101-2021_02_08_430070	205	7	,	,	,
10_1101-2021_02_08_430070	205	8	we	-PRON-	PRP
10_1101-2021_02_08_430070	205	9	use	use	VBP
10_1101-2021_02_08_430070	205	10	reads	read	VBZ
10_1101-2021_02_08_430070	205	11	as	as	IN
10_1101-2021_02_08_430070	205	12	the	the	DT
10_1101-2021_02_08_430070	205	13	basic	basic	JJ
10_1101-2021_02_08_430070	205	14	data	datum	NNS
10_1101-2021_02_08_430070	205	15	unit	unit	NN
10_1101-2021_02_08_430070	205	16	and	and	CC
10_1101-2021_02_08_430070	205	17	integrate	integrate	VB
10_1101-2021_02_08_430070	205	18	the	the	DT
10_1101-2021_02_08_430070	205	19	data	data	NN
10_1101-2021_02_08_430070	205	20	pre	pre	JJ
10_1101-2021_02_08_430070	205	21	-	-	JJ
10_1101-2021_02_08_430070	205	22	processing	processing	JJ
10_1101-2021_02_08_430070	205	23	part	part	NN
10_1101-2021_02_08_430070	205	24	during	during	IN
10_1101-2021_02_08_430070	205	25	a	a	DT
10_1101-2021_02_08_430070	205	26	read	read	JJ
10_1101-2021_02_08_430070	205	27	-	-	HYPH
10_1101-2021_02_08_430070	205	28	batch	batch	NN
10_1101-2021_02_08_430070	205	29	loading	loading	NN
10_1101-2021_02_08_430070	205	30	process	process	NN
10_1101-2021_02_08_430070	205	31	.	.	.
10_1101-2021_02_08_430070	206	1	The	the	DT
10_1101-2021_02_08_430070	206	2	data	datum	NNS
10_1101-2021_02_08_430070	206	3	I	i	NN
10_1101-2021_02_08_430070	206	4	/	/	SYM
10_1101-2021_02_08_430070	206	5	O	o	NN
10_1101-2021_02_08_430070	206	6	and	and	CC
10_1101-2021_02_08_430070	206	7	feature	feature	NN
10_1101-2021_02_08_430070	206	8	extraction	extraction	NN
10_1101-2021_02_08_430070	206	9	part	part	NN
10_1101-2021_02_08_430070	206	10	can	can	MD
10_1101-2021_02_08_430070	206	11	be	be	VB
10_1101-2021_02_08_430070	206	12	further	further	RB
10_1101-2021_02_08_430070	206	13	accelerated	accelerate	VBN
10_1101-2021_02_08_430070	206	14	.	.	.
10_1101-2021_02_08_430070	207	1	4	4	LS
10_1101-2021_02_08_430070	207	2	Discussion	Discussion	NNP
10_1101-2021_02_08_430070	207	3	A	A	NNP
10_1101-2021_02_08_430070	207	4	BERT	BERT	NNP
10_1101-2021_02_08_430070	207	5	commonly	commonly	RB
10_1101-2021_02_08_430070	207	6	works	work	VBZ
10_1101-2021_02_08_430070	207	7	in	in	IN
10_1101-2021_02_08_430070	207	8	a	a	DT
10_1101-2021_02_08_430070	207	9	pre	pre	JJ
10_1101-2021_02_08_430070	207	10	-	-	JJ
10_1101-2021_02_08_430070	207	11	training	training	JJ
10_1101-2021_02_08_430070	207	12	and	and	CC
10_1101-2021_02_08_430070	207	13	fine	fine	JJ
10_1101-2021_02_08_430070	207	14	-	-	HYPH
10_1101-2021_02_08_430070	207	15	tuning	tuning	NN
10_1101-2021_02_08_430070	207	16	approach	approach	NN
10_1101-2021_02_08_430070	207	17	.	.	.
10_1101-2021_02_08_430070	208	1	In	in	IN
10_1101-2021_02_08_430070	208	2	the	the	DT
10_1101-2021_02_08_430070	208	3	pre	pre	JJ
10_1101-2021_02_08_430070	208	4	-	-	JJ
10_1101-2021_02_08_430070	208	5	training	training	JJ
10_1101-2021_02_08_430070	208	6	phase	phase	NN
10_1101-2021_02_08_430070	208	7	,	,	,
10_1101-2021_02_08_430070	208	8	a	a	DT
10_1101-2021_02_08_430070	208	9	BERT	BERT	NNP
10_1101-2021_02_08_430070	208	10	learns	learn	VBZ
10_1101-2021_02_08_430070	208	11	bi	bi	JJ
10_1101-2021_02_08_430070	208	12	-	-	JJ
10_1101-2021_02_08_430070	208	13	directional	directional	JJ
10_1101-2021_02_08_430070	208	14	representations	representation	NNS
10_1101-2021_02_08_430070	208	15	from	from	IN
10_1101-2021_02_08_430070	208	16	unlabeled	unlabeled	JJ
10_1101-2021_02_08_430070	208	17	data	datum	NNS
10_1101-2021_02_08_430070	208	18	.	.	.
10_1101-2021_02_08_430070	209	1	After	after	IN
10_1101-2021_02_08_430070	209	2	that	that	DT
10_1101-2021_02_08_430070	209	3	,	,	,
10_1101-2021_02_08_430070	209	4	learned	learn	VBN
10_1101-2021_02_08_430070	209	5	feature	feature	NN
10_1101-2021_02_08_430070	209	6	representations	representation	NNS
10_1101-2021_02_08_430070	209	7	are	be	VBP
10_1101-2021_02_08_430070	209	8	used	use	VBN
10_1101-2021_02_08_430070	209	9	on	on	IN
10_1101-2021_02_08_430070	209	10	task-	task-	NN
10_1101-2021_02_08_430070	209	11	specific	specific	JJ
10_1101-2021_02_08_430070	209	12	data	datum	NNS
10_1101-2021_02_08_430070	209	13	for	for	IN
10_1101-2021_02_08_430070	209	14	further	further	JJ
10_1101-2021_02_08_430070	209	15	fine	fine	NN
10_1101-2021_02_08_430070	209	16	-	-	HYPH
10_1101-2021_02_08_430070	209	17	tuning	tuning	NN
10_1101-2021_02_08_430070	209	18	.	.	.
10_1101-2021_02_08_430070	210	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	210	2	has	have	VBZ
10_1101-2021_02_08_430070	210	3	lead	lead	NN
10_1101-2021_02_08_430070	210	4	to	to	IN
10_1101-2021_02_08_430070	210	5	several	several	JJ
10_1101-2021_02_08_430070	210	6	state	state	NN
10_1101-2021_02_08_430070	210	7	-	-	HYPH
10_1101-2021_02_08_430070	210	8	of	of	IN
10_1101-2021_02_08_430070	210	9	-	-	HYPH
10_1101-2021_02_08_430070	210	10	the	the	DT
10_1101-2021_02_08_430070	210	11	-	-	HYPH
10_1101-2021_02_08_430070	210	12	art	art	NN
10_1101-2021_02_08_430070	210	13	results	result	NNS
10_1101-2021_02_08_430070	210	14	on	on	IN
10_1101-2021_02_08_430070	210	15	many	many	JJ
10_1101-2021_02_08_430070	210	16	downstream	downstream	JJ
10_1101-2021_02_08_430070	210	17	tasks	task	NNS
10_1101-2021_02_08_430070	210	18	in	in	IN
10_1101-2021_02_08_430070	210	19	language	language	NN
10_1101-2021_02_08_430070	210	20	understanding	understanding	NN
10_1101-2021_02_08_430070	210	21	.	.	.
10_1101-2021_02_08_430070	211	1	According	accord	VBG
10_1101-2021_02_08_430070	211	2	to	to	IN
10_1101-2021_02_08_430070	211	3	the	the	DT
10_1101-2021_02_08_430070	211	4	data	data	NN
10_1101-2021_02_08_430070	211	5	scale	scale	NN
10_1101-2021_02_08_430070	211	6	,	,	,
10_1101-2021_02_08_430070	211	7	the	the	DT
10_1101-2021_02_08_430070	211	8	number	number	NN
10_1101-2021_02_08_430070	211	9	of	of	IN
10_1101-2021_02_08_430070	211	10	BERT	BERT	NNP
10_1101-2021_02_08_430070	211	11	parameters	parameter	NNS
10_1101-2021_02_08_430070	211	12	is	be	VBZ
10_1101-2021_02_08_430070	211	13	usually	usually	RB
10_1101-2021_02_08_430070	211	14	large	large	JJ
10_1101-2021_02_08_430070	211	15	,	,	,
10_1101-2021_02_08_430070	211	16	and	and	CC
10_1101-2021_02_08_430070	211	17	training	train	VBG
10_1101-2021_02_08_430070	211	18	such	such	PDT
10_1101-2021_02_08_430070	211	19	a	a	DT
10_1101-2021_02_08_430070	211	20	model	model	NN
10_1101-2021_02_08_430070	211	21	requires	require	VBZ
10_1101-2021_02_08_430070	211	22	a	a	DT
10_1101-2021_02_08_430070	211	23	huge	huge	JJ
10_1101-2021_02_08_430070	211	24	amount	amount	NN
10_1101-2021_02_08_430070	211	25	of	of	IN
10_1101-2021_02_08_430070	211	26	computational	computational	JJ
10_1101-2021_02_08_430070	211	27	resources	resource	NNS
10_1101-2021_02_08_430070	211	28	.	.	.
10_1101-2021_02_08_430070	212	1	For	for	IN
10_1101-2021_02_08_430070	212	2	example	example	NN
10_1101-2021_02_08_430070	212	3	,	,	,
10_1101-2021_02_08_430070	212	4	the	the	DT
10_1101-2021_02_08_430070	212	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	212	6	used	use	VBN
10_1101-2021_02_08_430070	212	7	for	for	IN
10_1101-2021_02_08_430070	212	8	natural	natural	JJ
10_1101-2021_02_08_430070	212	9	language	language	NN
10_1101-2021_02_08_430070	212	10	modeling	modeling	NN
10_1101-2021_02_08_430070	212	11	has	have	VBZ
10_1101-2021_02_08_430070	212	12	a	a	DT
10_1101-2021_02_08_430070	212	13	parameter	parameter	NN
10_1101-2021_02_08_430070	212	14	scale	scale	NN
10_1101-2021_02_08_430070	212	15	ranging	range	VBG
10_1101-2021_02_08_430070	212	16	from	from	IN
10_1101-2021_02_08_430070	212	17	110	110	CD
10_1101-2021_02_08_430070	212	18	M	m	NN
10_1101-2021_02_08_430070	212	19	to	to	IN
10_1101-2021_02_08_430070	212	20	340	340	CD
10_1101-2021_02_08_430070	212	21	M	m	NN
10_1101-2021_02_08_430070	212	22	(	(	-LRB-
10_1101-2021_02_08_430070	212	23	Devlin	Devlin	NNP
10_1101-2021_02_08_430070	212	24	et	et	NNP
10_1101-2021_02_08_430070	212	25	al	al	NNP
10_1101-2021_02_08_430070	212	26	.	.	NNP
10_1101-2021_02_08_430070	212	27	,	,	,
10_1101-2021_02_08_430070	212	28	2018	2018	CD
10_1101-2021_02_08_430070	212	29	)	)	-RRB-
10_1101-2021_02_08_430070	212	30	.	.	.
10_1101-2021_02_08_430070	213	1	In	in	IN
10_1101-2021_02_08_430070	213	2	this	this	DT
10_1101-2021_02_08_430070	213	3	work	work	NN
10_1101-2021_02_08_430070	213	4	,	,	,
10_1101-2021_02_08_430070	213	5	we	-PRON-	PRP
10_1101-2021_02_08_430070	213	6	did	do	VBD
10_1101-2021_02_08_430070	213	7	not	not	RB
10_1101-2021_02_08_430070	213	8	follow	follow	VB
10_1101-2021_02_08_430070	213	9	this	this	DT
10_1101-2021_02_08_430070	213	10	schema	schema	NN
10_1101-2021_02_08_430070	213	11	.	.	.
10_1101-2021_02_08_430070	214	1	Instead	instead	RB
10_1101-2021_02_08_430070	214	2	,	,	,
10_1101-2021_02_08_430070	214	3	we	-PRON-	PRP
10_1101-2021_02_08_430070	214	4	utilized	utilize	VBD
10_1101-2021_02_08_430070	214	5	the	the	DT
10_1101-2021_02_08_430070	214	6	model	model	NN
10_1101-2021_02_08_430070	214	7	architecture	architecture	NN
10_1101-2021_02_08_430070	214	8	of	of	IN
10_1101-2021_02_08_430070	214	9	BERT	BERT	NNP
10_1101-2021_02_08_430070	214	10	to	to	TO
10_1101-2021_02_08_430070	214	11	provide	provide	VB
10_1101-2021_02_08_430070	214	12	a	a	DT
10_1101-2021_02_08_430070	214	13	lightweight	lightweight	JJ
10_1101-2021_02_08_430070	214	14	and	and	CC
10_1101-2021_02_08_430070	214	15	non	non	JJ
10_1101-2021_02_08_430070	214	16	-	-	JJ
10_1101-2021_02_08_430070	214	17	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	214	18	solution	solution	NN
10_1101-2021_02_08_430070	214	19	to	to	TO
10_1101-2021_02_08_430070	214	20	replace	replace	VB
10_1101-2021_02_08_430070	214	21	the	the	DT
10_1101-2021_02_08_430070	214	22	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	214	23	biRNN	biRNN	NNP
10_1101-2021_02_08_430070	214	24	model	model	NN
10_1101-2021_02_08_430070	214	25	.	.	.
10_1101-2021_02_08_430070	215	1	In	in	IN
10_1101-2021_02_08_430070	215	2	our	-PRON-	PRP$
10_1101-2021_02_08_430070	215	3	experiment	experiment	NN
10_1101-2021_02_08_430070	215	4	,	,	,
10_1101-2021_02_08_430070	215	5	the	the	DT
10_1101-2021_02_08_430070	215	6	BERT	BERT	NNP
10_1101-2021_02_08_430070	215	7	uses	use	VBZ
10_1101-2021_02_08_430070	215	8	three	three	CD
10_1101-2021_02_08_430070	215	9	attention	attention	NN
10_1101-2021_02_08_430070	215	10	layers	layer	NNS
10_1101-2021_02_08_430070	215	11	with	with	IN
10_1101-2021_02_08_430070	215	12	4	4	CD
10_1101-2021_02_08_430070	215	13	attention	attention	NN
10_1101-2021_02_08_430070	215	14	heads	head	NNS
10_1101-2021_02_08_430070	215	15	and	and	CC
10_1101-2021_02_08_430070	215	16	100	100	CD
10_1101-2021_02_08_430070	215	17	hidden	hide	VBN
10_1101-2021_02_08_430070	215	18	units	unit	NNS
10_1101-2021_02_08_430070	215	19	for	for	IN
10_1101-2021_02_08_430070	215	20	each	each	DT
10_1101-2021_02_08_430070	215	21	layer	layer	NN
10_1101-2021_02_08_430070	215	22	.	.	.
10_1101-2021_02_08_430070	216	1	The	the	DT
10_1101-2021_02_08_430070	216	2	total	total	JJ
10_1101-2021_02_08_430070	216	3	number	number	NN
10_1101-2021_02_08_430070	216	4	of	of	IN
10_1101-2021_02_08_430070	216	5	model	model	NN
10_1101-2021_02_08_430070	216	6	parameters	parameter	NNS
10_1101-2021_02_08_430070	216	7	is	be	VBZ
10_1101-2021_02_08_430070	216	8	around	around	RB
10_1101-2021_02_08_430070	216	9	0.37	0.37	CD
10_1101-2021_02_08_430070	216	10	M	M	NNP
10_1101-2021_02_08_430070	216	11	,	,	,
10_1101-2021_02_08_430070	216	12	which	which	WDT
10_1101-2021_02_08_430070	216	13	is	be	VBZ
10_1101-2021_02_08_430070	216	14	even	even	RB
10_1101-2021_02_08_430070	216	15	less	less	JJR
10_1101-2021_02_08_430070	216	16	than	than	IN
10_1101-2021_02_08_430070	216	17	that	that	DT
10_1101-2021_02_08_430070	216	18	of	of	IN
10_1101-2021_02_08_430070	216	19	biRNN	birnn	NN
10_1101-2021_02_08_430070	216	20	(	(	-LRB-
10_1101-2021_02_08_430070	216	21	0.57	0.57	CD
10_1101-2021_02_08_430070	216	22	M	m	NN
10_1101-2021_02_08_430070	216	23	)	)	-RRB-
10_1101-2021_02_08_430070	216	24	.	.	.
10_1101-2021_02_08_430070	217	1	In	in	IN
10_1101-2021_02_08_430070	217	2	the	the	DT
10_1101-2021_02_08_430070	217	3	future	future	NN
10_1101-2021_02_08_430070	217	4	,	,	,
10_1101-2021_02_08_430070	217	5	when	when	WRB
10_1101-2021_02_08_430070	217	6	more	more	JJR
10_1101-2021_02_08_430070	217	7	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	217	8	methylation	methylation	NN
10_1101-2021_02_08_430070	217	9	data	datum	NNS
10_1101-2021_02_08_430070	217	10	becomes	become	VBZ
10_1101-2021_02_08_430070	217	11	available	available	JJ
10_1101-2021_02_08_430070	217	12	,	,	,
10_1101-2021_02_08_430070	217	13	a	a	DT
10_1101-2021_02_08_430070	217	14	larger	large	JJR
10_1101-2021_02_08_430070	217	15	BERT	BERT	NNP
10_1101-2021_02_08_430070	217	16	model	model	NN
10_1101-2021_02_08_430070	217	17	and	and	CC
10_1101-2021_02_08_430070	217	18	pre	pre	JJ
10_1101-2021_02_08_430070	217	19	-	-	JJ
10_1101-2021_02_08_430070	217	20	training	training	JJ
10_1101-2021_02_08_430070	217	21	and	and	CC
10_1101-2021_02_08_430070	217	22	fine	fine	JJ
10_1101-2021_02_08_430070	217	23	-	-	HYPH
10_1101-2021_02_08_430070	217	24	tuning	tuning	NN
10_1101-2021_02_08_430070	217	25	scheme	scheme	NN
10_1101-2021_02_08_430070	217	26	can	can	MD
10_1101-2021_02_08_430070	217	27	be	be	VB
10_1101-2021_02_08_430070	217	28	further	further	RB
10_1101-2021_02_08_430070	217	29	explored	explore	VBN
10_1101-2021_02_08_430070	217	30	.	.	.
10_1101-2021_02_08_430070	218	1	5	5	CD
10_1101-2021_02_08_430070	218	2	Conclusion	Conclusion	NNP
10_1101-2021_02_08_430070	218	3	In	in	IN
10_1101-2021_02_08_430070	218	4	this	this	DT
10_1101-2021_02_08_430070	218	5	work	work	NN
10_1101-2021_02_08_430070	218	6	,	,	,
10_1101-2021_02_08_430070	218	7	we	-PRON-	PRP
10_1101-2021_02_08_430070	218	8	explored	explore	VBD
10_1101-2021_02_08_430070	218	9	applying	apply	VBG
10_1101-2021_02_08_430070	218	10	BERT	BERT	NNP
10_1101-2021_02_08_430070	218	11	models	model	NNS
10_1101-2021_02_08_430070	218	12	for	for	IN
10_1101-2021_02_08_430070	218	13	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	218	14	methylation	methylation	NN
10_1101-2021_02_08_430070	218	15	detection	detection	NN
10_1101-2021_02_08_430070	218	16	,	,	,
10_1101-2021_02_08_430070	218	17	which	which	WDT
10_1101-2021_02_08_430070	218	18	aims	aim	VBZ
10_1101-2021_02_08_430070	218	19	to	to	TO
10_1101-2021_02_08_430070	218	20	use	use	VB
10_1101-2021_02_08_430070	218	21	a	a	DT
10_1101-2021_02_08_430070	218	22	non	non	JJ
10_1101-2021_02_08_430070	218	23	-	-	JJ
10_1101-2021_02_08_430070	218	24	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	218	25	modeling	modeling	NN
10_1101-2021_02_08_430070	218	26	approach	approach	NN
10_1101-2021_02_08_430070	218	27	for	for	IN
10_1101-2021_02_08_430070	218	28	fast	fast	JJ
10_1101-2021_02_08_430070	218	29	inference	inference	NN
10_1101-2021_02_08_430070	218	30	.	.	.
10_1101-2021_02_08_430070	219	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	219	2	quantified	quantify	VBD
10_1101-2021_02_08_430070	219	3	positional	positional	JJ
10_1101-2021_02_08_430070	219	4	signal	signal	NN
10_1101-2021_02_08_430070	219	5	-	-	HYPH
10_1101-2021_02_08_430070	219	6	shift	shift	NN
10_1101-2021_02_08_430070	219	7	related	relate	VBN
10_1101-2021_02_08_430070	219	8	to	to	IN
10_1101-2021_02_08_430070	219	9	methylation	methylation	NN
10_1101-2021_02_08_430070	219	10	for	for	IN
10_1101-2021_02_08_430070	219	11	different	different	JJ
10_1101-2021_02_08_430070	219	12	datasets	dataset	NNS
10_1101-2021_02_08_430070	219	13	of	of	IN
10_1101-2021_02_08_430070	219	14	specific	specific	JJ
10_1101-2021_02_08_430070	219	15	motif	motif	NN
10_1101-2021_02_08_430070	219	16	/	/	SYM
10_1101-2021_02_08_430070	219	17	methylase	methylase	NN
10_1101-2021_02_08_430070	219	18	and	and	CC
10_1101-2021_02_08_430070	219	19	found	find	VBD
10_1101-2021_02_08_430070	219	20	patterns	pattern	NNS
10_1101-2021_02_08_430070	219	21	across	across	IN
10_1101-2021_02_08_430070	219	22	datasets	dataset	NNS
10_1101-2021_02_08_430070	219	23	.	.	.
10_1101-2021_02_08_430070	220	1	In	in	IN
10_1101-2021_02_08_430070	220	2	the	the	DT
10_1101-2021_02_08_430070	220	3	process	process	NN
10_1101-2021_02_08_430070	220	4	of	of	IN
10_1101-2021_02_08_430070	220	5	evaluation	evaluation	NN
10_1101-2021_02_08_430070	220	6	,	,	,
10_1101-2021_02_08_430070	220	7	we	-PRON-	PRP
10_1101-2021_02_08_430070	220	8	found	find	VBD
10_1101-2021_02_08_430070	220	9	the	the	DT
10_1101-2021_02_08_430070	220	10	original	original	JJ
10_1101-2021_02_08_430070	220	11	BERT	BERT	NNP
10_1101-2021_02_08_430070	220	12	architecture	architecture	NN
10_1101-2021_02_08_430070	220	13	does	do	VBZ
10_1101-2021_02_08_430070	220	14	not	not	RB
10_1101-2021_02_08_430070	220	15	work	work	VB
10_1101-2021_02_08_430070	220	16	as	as	RB
10_1101-2021_02_08_430070	220	17	well	well	RB
10_1101-2021_02_08_430070	220	18	as	as	IN
10_1101-2021_02_08_430070	220	19	biRNN	birnn	NN
10_1101-2021_02_08_430070	220	20	.	.	.
10_1101-2021_02_08_430070	221	1	We	-PRON-	PRP
10_1101-2021_02_08_430070	221	2	proposed	propose	VBD
10_1101-2021_02_08_430070	221	3	a	a	DT
10_1101-2021_02_08_430070	221	4	refined	refined	JJ
10_1101-2021_02_08_430070	221	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	221	6	considering	consider	VBG
10_1101-2021_02_08_430070	221	7	task	task	NN
10_1101-2021_02_08_430070	221	8	-	-	HYPH
10_1101-2021_02_08_430070	221	9	specific	specific	JJ
10_1101-2021_02_08_430070	221	10	characters	character	NNS
10_1101-2021_02_08_430070	221	11	into	into	IN
10_1101-2021_02_08_430070	221	12	the	the	DT
10_1101-2021_02_08_430070	221	13	modeling	modeling	NN
10_1101-2021_02_08_430070	221	14	.	.	.
10_1101-2021_02_08_430070	222	1	Compared	compare	VBN
10_1101-2021_02_08_430070	222	2	with	with	IN
10_1101-2021_02_08_430070	222	3	the	the	DT
10_1101-2021_02_08_430070	222	4	original	original	JJ
10_1101-2021_02_08_430070	222	5	BERT	BERT	NNP
10_1101-2021_02_08_430070	222	6	,	,	,
10_1101-2021_02_08_430070	222	7	the	the	DT
10_1101-2021_02_08_430070	222	8	refined	refined	JJ
10_1101-2021_02_08_430070	222	9	BERT	BERT	NNP
10_1101-2021_02_08_430070	222	10	uses	use	VBZ
10_1101-2021_02_08_430070	222	11	learnable	learnable	JJ
10_1101-2021_02_08_430070	222	12	positional	positional	JJ
10_1101-2021_02_08_430070	222	13	encoding	encoding	NN
10_1101-2021_02_08_430070	222	14	and	and	CC
10_1101-2021_02_08_430070	222	15	self	self	NN
10_1101-2021_02_08_430070	222	16	-	-	HYPH
10_1101-2021_02_08_430070	222	17	attention	attention	NN
10_1101-2021_02_08_430070	222	18	with	with	IN
10_1101-2021_02_08_430070	222	19	relative	relative	JJ
10_1101-2021_02_08_430070	222	20	position	position	NN
10_1101-2021_02_08_430070	222	21	representation	representation	NN
10_1101-2021_02_08_430070	222	22	,	,	,
10_1101-2021_02_08_430070	222	23	and	and	CC
10_1101-2021_02_08_430070	222	24	focuses	focus	VBZ
10_1101-2021_02_08_430070	222	25	more	more	RBR
10_1101-2021_02_08_430070	222	26	on	on	IN
10_1101-2021_02_08_430070	222	27	the	the	DT
10_1101-2021_02_08_430070	222	28	center	center	NN
10_1101-2021_02_08_430070	222	29	positions	position	NNS
10_1101-2021_02_08_430070	222	30	in	in	IN
10_1101-2021_02_08_430070	222	31	a	a	DT
10_1101-2021_02_08_430070	222	32	±3bp	±3bp	NN
10_1101-2021_02_08_430070	222	33	range	range	NN
10_1101-2021_02_08_430070	222	34	.	.	.
10_1101-2021_02_08_430070	223	1	The	the	DT
10_1101-2021_02_08_430070	223	2	experiment	experiment	NN
10_1101-2021_02_08_430070	223	3	results	result	NNS
10_1101-2021_02_08_430070	223	4	show	show	VBP
10_1101-2021_02_08_430070	223	5	that	that	IN
10_1101-2021_02_08_430070	223	6	the	the	DT
10_1101-2021_02_08_430070	223	7	refined	refined	JJ
10_1101-2021_02_08_430070	223	8	BERT	BERT	NNP
10_1101-2021_02_08_430070	223	9	can	can	MD
10_1101-2021_02_08_430070	223	10	achieve	achieve	VB
10_1101-2021_02_08_430070	223	11	competitive	competitive	JJ
10_1101-2021_02_08_430070	223	12	and	and	CC
10_1101-2021_02_08_430070	223	13	even	even	RB
10_1101-2021_02_08_430070	223	14	better	well	JJR
10_1101-2021_02_08_430070	223	15	results	result	NNS
10_1101-2021_02_08_430070	223	16	than	than	IN
10_1101-2021_02_08_430070	223	17	the	the	DT
10_1101-2021_02_08_430070	223	18	state-	state-	NN
10_1101-2021_02_08_430070	223	19	of	of	IN
10_1101-2021_02_08_430070	223	20	-	-	HYPH
10_1101-2021_02_08_430070	223	21	the	the	DT
10_1101-2021_02_08_430070	223	22	-	-	HYPH
10_1101-2021_02_08_430070	223	23	art	art	NN
10_1101-2021_02_08_430070	223	24	biRNN	birnn	NN
10_1101-2021_02_08_430070	223	25	model	model	NN
10_1101-2021_02_08_430070	223	26	on	on	IN
10_1101-2021_02_08_430070	223	27	a	a	DT
10_1101-2021_02_08_430070	223	28	set	set	NN
10_1101-2021_02_08_430070	223	29	of	of	IN
10_1101-2021_02_08_430070	223	30	5mC	5mc	CD
10_1101-2021_02_08_430070	223	31	and	and	CC
10_1101-2021_02_08_430070	223	32	6mA	6ma	CD
10_1101-2021_02_08_430070	223	33	benchmark	benchmark	NN
10_1101-2021_02_08_430070	223	34	datasets	dataset	NNS
10_1101-2021_02_08_430070	223	35	,	,	,
10_1101-2021_02_08_430070	223	36	while	while	IN
10_1101-2021_02_08_430070	223	37	the	the	DT
10_1101-2021_02_08_430070	223	38	model	model	NN
10_1101-2021_02_08_430070	223	39	inference	inference	NN
10_1101-2021_02_08_430070	223	40	speed	speed	NN
10_1101-2021_02_08_430070	223	41	is	be	VBZ
10_1101-2021_02_08_430070	223	42	about	about	RB
10_1101-2021_02_08_430070	223	43	6x	6x	CD
10_1101-2021_02_08_430070	223	44	faster	fast	RBR
10_1101-2021_02_08_430070	223	45	.	.	.
10_1101-2021_02_08_430070	224	1	On	on	IN
10_1101-2021_02_08_430070	224	2	the	the	DT
10_1101-2021_02_08_430070	224	3	cross	cross	JJ
10_1101-2021_02_08_430070	224	4	-	-	JJ
10_1101-2021_02_08_430070	224	5	sample	sample	JJ
10_1101-2021_02_08_430070	224	6	evaluation	evaluation	NN
10_1101-2021_02_08_430070	224	7	,	,	,
10_1101-2021_02_08_430070	224	8	for	for	IN
10_1101-2021_02_08_430070	224	9	the	the	DT
10_1101-2021_02_08_430070	224	10	case	case	NN
10_1101-2021_02_08_430070	224	11	that	that	WDT
10_1101-2021_02_08_430070	224	12	train	train	NN
10_1101-2021_02_08_430070	224	13	and	and	CC
10_1101-2021_02_08_430070	224	14	test	test	NN
10_1101-2021_02_08_430070	224	15	data	datum	NNS
10_1101-2021_02_08_430070	224	16	from	from	IN
10_1101-2021_02_08_430070	224	17	different	different	JJ
10_1101-2021_02_08_430070	224	18	research	research	NN
10_1101-2021_02_08_430070	224	19	groups	group	NNS
10_1101-2021_02_08_430070	224	20	,	,	,
10_1101-2021_02_08_430070	224	21	BERTs	BERTs	NNP
10_1101-2021_02_08_430070	224	22	(	(	-LRB-
10_1101-2021_02_08_430070	224	23	include	include	VBP
10_1101-2021_02_08_430070	224	24	the	the	DT
10_1101-2021_02_08_430070	224	25	original	original	JJ
10_1101-2021_02_08_430070	224	26	BERT	BERT	NNP
10_1101-2021_02_08_430070	224	27	)	)	-RRB-
10_1101-2021_02_08_430070	224	28	show	show	VBP
10_1101-2021_02_08_430070	224	29	a	a	DT
10_1101-2021_02_08_430070	224	30	better	well	JJR
10_1101-2021_02_08_430070	224	31	performance	performance	NN
10_1101-2021_02_08_430070	224	32	than	than	IN
10_1101-2021_02_08_430070	224	33	biRNN	birnn	NN
10_1101-2021_02_08_430070	224	34	.	.	.
10_1101-2021_02_08_430070	225	1	Acknowledgements	acknowledgement	NNS
10_1101-2021_02_08_430070	225	2	We	-PRON-	PRP
10_1101-2021_02_08_430070	225	3	would	would	MD
10_1101-2021_02_08_430070	225	4	like	like	VB
10_1101-2021_02_08_430070	225	5	to	to	TO
10_1101-2021_02_08_430070	225	6	thank	thank	VB
10_1101-2021_02_08_430070	225	7	Marcus	Marcus	NNP
10_1101-2021_02_08_430070	225	8	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	225	9	and	and	CC
10_1101-2021_02_08_430070	225	10	Jared	Jared	NNP
10_1101-2021_02_08_430070	225	11	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	225	12	for	for	IN
10_1101-2021_02_08_430070	225	13	making	make	VBG
10_1101-2021_02_08_430070	225	14	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	225	15	methylation	methylation	NN
10_1101-2021_02_08_430070	225	16	data	datum	NNS
10_1101-2021_02_08_430070	225	17	publicly	publicly	RB
10_1101-2021_02_08_430070	225	18	available	available	JJ
10_1101-2021_02_08_430070	225	19	,	,	,
10_1101-2021_02_08_430070	225	20	Zaka	Zaka	NNP
10_1101-2021_02_08_430070	225	21	Wing	Wing	NNP
10_1101-2021_02_08_430070	225	22	-	-	HYPH
10_1101-2021_02_08_430070	225	23	Sze	Sze	NNP
10_1101-2021_02_08_430070	225	24	Yuen	Yuen	NNP
10_1101-2021_02_08_430070	225	25	for	for	IN
10_1101-2021_02_08_430070	225	26	providing	provide	VBG
10_1101-2021_02_08_430070	225	27	the	the	DT
10_1101-2021_02_08_430070	225	28	benchmark	benchmark	JJ
10_1101-2021_02_08_430070	225	29	dataset	dataset	NN
10_1101-2021_02_08_430070	225	30	and	and	CC
10_1101-2021_02_08_430070	225	31	pipeline	pipeline	NN
10_1101-2021_02_08_430070	225	32	,	,	,
10_1101-2021_02_08_430070	225	33	authors	author	NNS
10_1101-2021_02_08_430070	225	34	of	of	IN
10_1101-2021_02_08_430070	225	35	deepMOD	deepMOD	NNP
10_1101-2021_02_08_430070	225	36	and	and	CC
10_1101-2021_02_08_430070	225	37	deepSignal	deepsignal	JJ
10_1101-2021_02_08_430070	225	38	for	for	IN
10_1101-2021_02_08_430070	225	39	providing	provide	VBG
10_1101-2021_02_08_430070	225	40	their	-PRON-	PRP$
10_1101-2021_02_08_430070	225	41	source	source	NN
10_1101-2021_02_08_430070	225	42	codes	code	NNS
10_1101-2021_02_08_430070	225	43	.	.	.
10_1101-2021_02_08_430070	226	1	References	References	NNPS
10_1101-2021_02_08_430070	226	2	Devlin	Devlin	NNP
10_1101-2021_02_08_430070	226	3	,	,	,
10_1101-2021_02_08_430070	226	4	J.	J.	NNP
10_1101-2021_02_08_430070	226	5	et	et	NNP
10_1101-2021_02_08_430070	226	6	al	al	NNP
10_1101-2021_02_08_430070	226	7	.	.	.
10_1101-2021_02_08_430070	227	1	(	(	-LRB-
10_1101-2021_02_08_430070	227	2	2018	2018	CD
10_1101-2021_02_08_430070	227	3	)	)	-RRB-
10_1101-2021_02_08_430070	227	4	.	.	.
10_1101-2021_02_08_430070	228	1	Bert	Bert	NNP
10_1101-2021_02_08_430070	228	2	:	:	:
10_1101-2021_02_08_430070	228	3	Pre	pre	JJ
10_1101-2021_02_08_430070	228	4	-	-	JJ
10_1101-2021_02_08_430070	228	5	training	training	NN
10_1101-2021_02_08_430070	228	6	of	of	IN
10_1101-2021_02_08_430070	228	7	deep	deep	JJ
10_1101-2021_02_08_430070	228	8	bidirectional	bidirectional	JJ
10_1101-2021_02_08_430070	228	9	transformers	transformer	NNS
10_1101-2021_02_08_430070	228	10	for	for	IN
10_1101-2021_02_08_430070	228	11	language	language	NN
10_1101-2021_02_08_430070	228	12	understanding	understanding	NN
10_1101-2021_02_08_430070	228	13	.	.	.
10_1101-2021_02_08_430070	229	1	arXiv	arXiv	NNP
10_1101-2021_02_08_430070	229	2	preprint	preprint	NN
10_1101-2021_02_08_430070	229	3	arXiv:1810.04805	arXiv:1810.04805	NNP
10_1101-2021_02_08_430070	229	4	.	.	.
10_1101-2021_02_08_430070	230	1	Huang	Huang	NNP
10_1101-2021_02_08_430070	230	2	,	,	,
10_1101-2021_02_08_430070	230	3	Z.	Z.	NNP
10_1101-2021_02_08_430070	230	4	et	et	FW
10_1101-2021_02_08_430070	230	5	al	al	NNP
10_1101-2021_02_08_430070	230	6	.	.	.
10_1101-2021_02_08_430070	231	1	(	(	-LRB-
10_1101-2021_02_08_430070	231	2	2020	2020	CD
10_1101-2021_02_08_430070	231	3	)	)	-RRB-
10_1101-2021_02_08_430070	231	4	.	.	.
10_1101-2021_02_08_430070	232	1	Improve	improve	VB
10_1101-2021_02_08_430070	232	2	transformer	transformer	NN
10_1101-2021_02_08_430070	232	3	models	model	NNS
10_1101-2021_02_08_430070	232	4	with	with	IN
10_1101-2021_02_08_430070	232	5	better	well	JJR
10_1101-2021_02_08_430070	232	6	relative	relative	JJ
10_1101-2021_02_08_430070	232	7	position	position	NN
10_1101-2021_02_08_430070	232	8	embeddings	embedding	NNS
10_1101-2021_02_08_430070	232	9	.	.	.
10_1101-2021_02_08_430070	233	1	arXiv	arXiv	NNP
10_1101-2021_02_08_430070	233	2	preprint	preprint	NN
10_1101-2021_02_08_430070	233	3	arXiv:2009.13658	arxiv:2009.13658	VB
10_1101-2021_02_08_430070	233	4	.	.	.
10_1101-2021_02_08_430070	234	1	Kim	Kim	NNP
10_1101-2021_02_08_430070	234	2	,	,	,
10_1101-2021_02_08_430070	234	3	D.	D.	NNP
10_1101-2021_02_08_430070	234	4	et	et	NNP
10_1101-2021_02_08_430070	234	5	al	al	NNP
10_1101-2021_02_08_430070	234	6	.	.	.
10_1101-2021_02_08_430070	235	1	(	(	-LRB-
10_1101-2021_02_08_430070	235	2	2020	2020	CD
10_1101-2021_02_08_430070	235	3	)	)	-RRB-
10_1101-2021_02_08_430070	235	4	.	.	.
10_1101-2021_02_08_430070	236	1	The	the	DT
10_1101-2021_02_08_430070	236	2	architecture	architecture	NN
10_1101-2021_02_08_430070	236	3	of	of	IN
10_1101-2021_02_08_430070	236	4	sars	sars	NNP
10_1101-2021_02_08_430070	236	5	-	-	HYPH
10_1101-2021_02_08_430070	236	6	cov-2	cov-2	NNP
10_1101-2021_02_08_430070	236	7	transcriptome	transcriptome	DT
10_1101-2021_02_08_430070	236	8	.	.	.
10_1101-2021_02_08_430070	237	1	Cell	cell	NN
10_1101-2021_02_08_430070	237	2	,	,	,
10_1101-2021_02_08_430070	237	3	181(4	181(4	CD
10_1101-2021_02_08_430070	237	4	)	)	-RRB-
10_1101-2021_02_08_430070	237	5	,	,	,
10_1101-2021_02_08_430070	237	6	914–921	914–921	CD
10_1101-2021_02_08_430070	237	7	.	.	.
10_1101-2021_02_08_430070	238	1	Kingma	Kingma	NNP
10_1101-2021_02_08_430070	238	2	,	,	,
10_1101-2021_02_08_430070	238	3	D.	D.	NNP
10_1101-2021_02_08_430070	238	4	P.	P.	NNP
10_1101-2021_02_08_430070	238	5	and	and	CC
10_1101-2021_02_08_430070	238	6	Ba	Ba	NNP
10_1101-2021_02_08_430070	238	7	,	,	,
10_1101-2021_02_08_430070	238	8	J.	J.	NNP
10_1101-2021_02_08_430070	239	1	(	(	-LRB-
10_1101-2021_02_08_430070	239	2	2014	2014	CD
10_1101-2021_02_08_430070	239	3	)	)	-RRB-
10_1101-2021_02_08_430070	239	4	.	.	.
10_1101-2021_02_08_430070	240	1	Adam	Adam	NNP
10_1101-2021_02_08_430070	240	2	:	:	:
10_1101-2021_02_08_430070	240	3	A	a	DT
10_1101-2021_02_08_430070	240	4	method	method	NN
10_1101-2021_02_08_430070	240	5	for	for	IN
10_1101-2021_02_08_430070	240	6	stochastic	stochastic	JJ
10_1101-2021_02_08_430070	240	7	optimization	optimization	NN
10_1101-2021_02_08_430070	240	8	.	.	.
10_1101-2021_02_08_430070	241	1	arXiv	arXiv	NNP
10_1101-2021_02_08_430070	241	2	preprint	preprint	NN
10_1101-2021_02_08_430070	241	3	arXiv:1412.6980	arXiv:1412.6980	NNP
10_1101-2021_02_08_430070	241	4	.	.	.
10_1101-2021_02_08_430070	242	1	Liu	Liu	NNP
10_1101-2021_02_08_430070	242	2	,	,	,
10_1101-2021_02_08_430070	242	3	Q.	Q.	NNP
10_1101-2021_02_08_430070	242	4	et	et	NNP
10_1101-2021_02_08_430070	242	5	al	al	NNP
10_1101-2021_02_08_430070	242	6	.	.	.
10_1101-2021_02_08_430070	243	1	(	(	-LRB-
10_1101-2021_02_08_430070	243	2	2019	2019	CD
10_1101-2021_02_08_430070	243	3	)	)	-RRB-
10_1101-2021_02_08_430070	243	4	.	.	.
10_1101-2021_02_08_430070	244	1	Detection	detection	NN
10_1101-2021_02_08_430070	244	2	of	of	IN
10_1101-2021_02_08_430070	244	3	dna	dna	NN
10_1101-2021_02_08_430070	244	4	base	base	NN
10_1101-2021_02_08_430070	244	5	modifications	modification	NNS
10_1101-2021_02_08_430070	244	6	by	by	IN
10_1101-2021_02_08_430070	244	7	deep	deep	JJ
10_1101-2021_02_08_430070	244	8	recurrent	recurrent	JJ
10_1101-2021_02_08_430070	244	9	neural	neural	JJ
10_1101-2021_02_08_430070	244	10	network	network	NN
10_1101-2021_02_08_430070	244	11	on	on	IN
10_1101-2021_02_08_430070	244	12	oxford	oxford	NNP
10_1101-2021_02_08_430070	244	13	nanopore	nanopore	NNP
10_1101-2021_02_08_430070	244	14	sequencing	sequence	VBG
10_1101-2021_02_08_430070	244	15	data	datum	NNS
10_1101-2021_02_08_430070	244	16	.	.	.
10_1101-2021_02_08_430070	245	1	Nature	nature	NN
10_1101-2021_02_08_430070	245	2	communications	communication	NNS
10_1101-2021_02_08_430070	245	3	,	,	,
10_1101-2021_02_08_430070	245	4	10(1	10(1	CD
10_1101-2021_02_08_430070	245	5	)	)	-RRB-
10_1101-2021_02_08_430070	245	6	,	,	,
10_1101-2021_02_08_430070	245	7	1–11	1–11	NNP
10_1101-2021_02_08_430070	245	8	.	.	.
10_1101-2021_02_08_430070	246	1	Ni	Ni	NNP
10_1101-2021_02_08_430070	246	2	,	,	,
10_1101-2021_02_08_430070	246	3	P.	P.	NNP
10_1101-2021_02_08_430070	246	4	et	et	NNP
10_1101-2021_02_08_430070	246	5	al	al	NNP
10_1101-2021_02_08_430070	246	6	.	.	.
10_1101-2021_02_08_430070	247	1	(	(	-LRB-
10_1101-2021_02_08_430070	247	2	2019	2019	CD
10_1101-2021_02_08_430070	247	3	)	)	-RRB-
10_1101-2021_02_08_430070	247	4	.	.	.
10_1101-2021_02_08_430070	248	1	Deepsignal	deepsignal	JJ
10_1101-2021_02_08_430070	248	2	:	:	:
10_1101-2021_02_08_430070	248	3	detecting	detect	VBG
10_1101-2021_02_08_430070	248	4	dna	dna	NN
10_1101-2021_02_08_430070	248	5	methylation	methylation	NN
10_1101-2021_02_08_430070	248	6	state	state	NN
10_1101-2021_02_08_430070	248	7	from	from	IN
10_1101-2021_02_08_430070	248	8	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	248	9	sequencing	sequencing	NN
10_1101-2021_02_08_430070	248	10	reads	read	NNS
10_1101-2021_02_08_430070	248	11	using	use	VBG
10_1101-2021_02_08_430070	248	12	deep	deep	JJ
10_1101-2021_02_08_430070	248	13	-	-	HYPH
10_1101-2021_02_08_430070	248	14	learning	learning	NN
10_1101-2021_02_08_430070	248	15	.	.	.
10_1101-2021_02_08_430070	249	1	Bioinformatics	bioinformatic	NNS
10_1101-2021_02_08_430070	249	2	,	,	,
10_1101-2021_02_08_430070	249	3	35(22	35(22	CD
10_1101-2021_02_08_430070	249	4	)	)	-RRB-
10_1101-2021_02_08_430070	249	5	,	,	,
10_1101-2021_02_08_430070	249	6	4586–4595	4586–4595	CD
10_1101-2021_02_08_430070	249	7	.	.	.
10_1101-2021_02_08_430070	250	1	Shaw	Shaw	NNP
10_1101-2021_02_08_430070	250	2	,	,	,
10_1101-2021_02_08_430070	250	3	P.	P.	NNP
10_1101-2021_02_08_430070	250	4	et	et	NNP
10_1101-2021_02_08_430070	250	5	al	al	NNP
10_1101-2021_02_08_430070	250	6	.	.	.
10_1101-2021_02_08_430070	251	1	(	(	-LRB-
10_1101-2021_02_08_430070	251	2	2018	2018	CD
10_1101-2021_02_08_430070	251	3	)	)	-RRB-
10_1101-2021_02_08_430070	251	4	.	.	.
10_1101-2021_02_08_430070	252	1	Self	self	NN
10_1101-2021_02_08_430070	252	2	-	-	HYPH
10_1101-2021_02_08_430070	252	3	attention	attention	NN
10_1101-2021_02_08_430070	252	4	with	with	IN
10_1101-2021_02_08_430070	252	5	relative	relative	JJ
10_1101-2021_02_08_430070	252	6	position	position	NN
10_1101-2021_02_08_430070	252	7	representations	representation	NNS
10_1101-2021_02_08_430070	252	8	.	.	.
10_1101-2021_02_08_430070	253	1	arXiv	arXiv	NNP
10_1101-2021_02_08_430070	253	2	preprint	preprint	NN
10_1101-2021_02_08_430070	253	3	arXiv:1803.02155	arXiv:1803.02155	NNP
10_1101-2021_02_08_430070	253	4	.	.	.
10_1101-2021_02_08_430070	254	1	Simpson	Simpson	NNP
10_1101-2021_02_08_430070	254	2	,	,	,
10_1101-2021_02_08_430070	254	3	J.	J.	NNP
10_1101-2021_02_08_430070	254	4	T.	T.	NNP
10_1101-2021_02_08_430070	254	5	et	et	NNP
10_1101-2021_02_08_430070	254	6	al	al	NNP
10_1101-2021_02_08_430070	254	7	.	.	.
10_1101-2021_02_08_430070	255	1	(	(	-LRB-
10_1101-2021_02_08_430070	255	2	2017	2017	CD
10_1101-2021_02_08_430070	255	3	)	)	-RRB-
10_1101-2021_02_08_430070	255	4	.	.	.
10_1101-2021_02_08_430070	256	1	Detecting	detect	VBG
10_1101-2021_02_08_430070	256	2	dna	dna	NN
10_1101-2021_02_08_430070	256	3	cytosine	cytosine	NN
10_1101-2021_02_08_430070	256	4	methylation	methylation	NN
10_1101-2021_02_08_430070	256	5	using	use	VBG
10_1101-2021_02_08_430070	256	6	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	256	7	sequencing	sequencing	NN
10_1101-2021_02_08_430070	256	8	.	.	.
10_1101-2021_02_08_430070	257	1	Nature	nature	NN
10_1101-2021_02_08_430070	257	2	methods	method	NNS
10_1101-2021_02_08_430070	257	3	,	,	,
10_1101-2021_02_08_430070	257	4	14(4	14(4	CD
10_1101-2021_02_08_430070	257	5	)	)	-RRB-
10_1101-2021_02_08_430070	257	6	,	,	,
10_1101-2021_02_08_430070	257	7	407	407	CD
10_1101-2021_02_08_430070	257	8	.	.	.
10_1101-2021_02_08_430070	258	1	Stoiber	Stoiber	NNP
10_1101-2021_02_08_430070	258	2	,	,	,
10_1101-2021_02_08_430070	258	3	M.	M.	NNP
10_1101-2021_02_08_430070	258	4	H.	H.	NNP
10_1101-2021_02_08_430070	258	5	et	et	NNP
10_1101-2021_02_08_430070	258	6	al	al	NNP
10_1101-2021_02_08_430070	258	7	.	.	.
10_1101-2021_02_08_430070	259	1	(	(	-LRB-
10_1101-2021_02_08_430070	259	2	2016	2016	CD
10_1101-2021_02_08_430070	259	3	)	)	-RRB-
10_1101-2021_02_08_430070	259	4	.	.	.
10_1101-2021_02_08_430070	260	1	De	De	NNP
10_1101-2021_02_08_430070	260	2	novo	novo	NNP
10_1101-2021_02_08_430070	260	3	identification	identification	NN
10_1101-2021_02_08_430070	260	4	of	of	IN
10_1101-2021_02_08_430070	260	5	dna	dna	NNP
10_1101-2021_02_08_430070	260	6	modifications	modification	NNS
10_1101-2021_02_08_430070	260	7	enabled	enable	VBN
10_1101-2021_02_08_430070	260	8	by	by	IN
10_1101-2021_02_08_430070	260	9	genome	genome	NN
10_1101-2021_02_08_430070	260	10	-	-	HYPH
10_1101-2021_02_08_430070	260	11	guided	guide	VBN
10_1101-2021_02_08_430070	260	12	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	260	13	signal	signal	NN
10_1101-2021_02_08_430070	260	14	processing	processing	NN
10_1101-2021_02_08_430070	260	15	.	.	.
10_1101-2021_02_08_430070	261	1	BioRxiv	biorxiv	NN
10_1101-2021_02_08_430070	261	2	,	,	,
10_1101-2021_02_08_430070	261	3	page	page	NN
10_1101-2021_02_08_430070	261	4	094672	094672	CD
10_1101-2021_02_08_430070	261	5	.	.	.
10_1101-2021_02_08_430070	262	1	Vaswani	Vaswani	NNP
10_1101-2021_02_08_430070	262	2	,	,	,
10_1101-2021_02_08_430070	262	3	A.	A.	NNP
10_1101-2021_02_08_430070	262	4	et	et	FW
10_1101-2021_02_08_430070	262	5	al	al	NNP
10_1101-2021_02_08_430070	262	6	.	.	.
10_1101-2021_02_08_430070	263	1	(	(	-LRB-
10_1101-2021_02_08_430070	263	2	2017	2017	CD
10_1101-2021_02_08_430070	263	3	)	)	-RRB-
10_1101-2021_02_08_430070	263	4	.	.	.
10_1101-2021_02_08_430070	264	1	Attention	attention	NN
10_1101-2021_02_08_430070	264	2	is	be	VBZ
10_1101-2021_02_08_430070	264	3	all	all	DT
10_1101-2021_02_08_430070	264	4	you	-PRON-	PRP
10_1101-2021_02_08_430070	264	5	need	need	VBP
10_1101-2021_02_08_430070	264	6	.	.	.
10_1101-2021_02_08_430070	265	1	pages	page	NNS
10_1101-2021_02_08_430070	265	2	5998–6008	5998–6008	CD
10_1101-2021_02_08_430070	265	3	.	.	.
10_1101-2021_02_08_430070	266	1	Yuen	Yuen	NNP
10_1101-2021_02_08_430070	266	2	,	,	,
10_1101-2021_02_08_430070	266	3	Z.	Z.	NNP
10_1101-2021_02_08_430070	266	4	W.-S.	W.-S.	NNP
10_1101-2021_02_08_430070	266	5	et	et	NNP
10_1101-2021_02_08_430070	266	6	al	al	NNP
10_1101-2021_02_08_430070	266	7	.	.	.
10_1101-2021_02_08_430070	267	1	(	(	-LRB-
10_1101-2021_02_08_430070	267	2	2020	2020	CD
10_1101-2021_02_08_430070	267	3	)	)	-RRB-
10_1101-2021_02_08_430070	267	4	.	.	.
10_1101-2021_02_08_430070	268	1	Systematic	systematic	JJ
10_1101-2021_02_08_430070	268	2	benchmarking	benchmarking	NN
10_1101-2021_02_08_430070	268	3	of	of	IN
10_1101-2021_02_08_430070	268	4	tools	tool	NNS
10_1101-2021_02_08_430070	268	5	for	for	IN
10_1101-2021_02_08_430070	268	6	cpg	cpg	NNP
10_1101-2021_02_08_430070	268	7	methylation	methylation	NN
10_1101-2021_02_08_430070	268	8	detection	detection	NN
10_1101-2021_02_08_430070	268	9	from	from	IN
10_1101-2021_02_08_430070	268	10	nanopore	nanopore	JJ
10_1101-2021_02_08_430070	268	11	sequencing	sequencing	NN
10_1101-2021_02_08_430070	268	12	.	.	.
10_1101-2021_02_08_430070	269	1	bioRxiv	biorxiv	NN
10_1101-2021_02_08_430070	269	2	.	.	.
10_1101-2021_02_08_430070	270	1	.license	.license	NNP
10_1101-2021_02_08_430070	270	2	CC	CC	NNP
10_1101-2021_02_08_430070	270	3	-	-	HYPH
10_1101-2021_02_08_430070	270	4	BY	BY	NNP
10_1101-2021_02_08_430070	270	5	-	-	HYPH
10_1101-2021_02_08_430070	270	6	NC	NC	NNP
10_1101-2021_02_08_430070	270	7	-	-	HYPH
10_1101-2021_02_08_430070	270	8	ND	ND	NNP
10_1101-2021_02_08_430070	270	9	4.0	4.0	CD
10_1101-2021_02_08_430070	270	10	Internationalpeer	Internationalpeer	NNP
10_1101-2021_02_08_430070	270	11	review	review	NN
10_1101-2021_02_08_430070	270	12	)	)	-RRB-
10_1101-2021_02_08_430070	270	13	is	be	VBZ
10_1101-2021_02_08_430070	270	14	the	the	DT
10_1101-2021_02_08_430070	270	15	author	author	NN
10_1101-2021_02_08_430070	270	16	/	/	SYM
10_1101-2021_02_08_430070	270	17	funder	funder	NN
10_1101-2021_02_08_430070	270	18	,	,	,
10_1101-2021_02_08_430070	270	19	who	who	WP
10_1101-2021_02_08_430070	270	20	has	have	VBZ
10_1101-2021_02_08_430070	270	21	granted	grant	VBN
10_1101-2021_02_08_430070	270	22	bioRxiv	biorxiv	IN
10_1101-2021_02_08_430070	270	23	a	a	DT
10_1101-2021_02_08_430070	270	24	license	license	NN
10_1101-2021_02_08_430070	270	25	to	to	TO
10_1101-2021_02_08_430070	270	26	display	display	VB
10_1101-2021_02_08_430070	270	27	the	the	DT
10_1101-2021_02_08_430070	270	28	preprint	preprint	NN
10_1101-2021_02_08_430070	270	29	in	in	IN
10_1101-2021_02_08_430070	270	30	perpetuity	perpetuity	NN
10_1101-2021_02_08_430070	270	31	.	.	.
10_1101-2021_02_08_430070	271	1	It	-PRON-	PRP
10_1101-2021_02_08_430070	271	2	is	be	VBZ
10_1101-2021_02_08_430070	271	3	made	make	VBN
10_1101-2021_02_08_430070	271	4	available	available	JJ
10_1101-2021_02_08_430070	271	5	under	under	IN
10_1101-2021_02_08_430070	271	6	a	a	DT
10_1101-2021_02_08_430070	271	7	The	the	DT
10_1101-2021_02_08_430070	271	8	copyright	copyright	NN
10_1101-2021_02_08_430070	271	9	holder	holder	NN
10_1101-2021_02_08_430070	271	10	for	for	IN
10_1101-2021_02_08_430070	271	11	this	this	DT
10_1101-2021_02_08_430070	271	12	preprint	preprint	NN
10_1101-2021_02_08_430070	271	13	(	(	-LRB-
10_1101-2021_02_08_430070	271	14	which	which	WDT
10_1101-2021_02_08_430070	271	15	was	be	VBD
10_1101-2021_02_08_430070	271	16	not	not	RB
10_1101-2021_02_08_430070	271	17	certified	certify	VBN
10_1101-2021_02_08_430070	271	18	bythis	bythis	DT
10_1101-2021_02_08_430070	271	19	version	version	NN
10_1101-2021_02_08_430070	271	20	posted	post	VBD
10_1101-2021_02_08_430070	271	21	February	February	NNP
10_1101-2021_02_08_430070	271	22	10	10	CD
10_1101-2021_02_08_430070	271	23	,	,	,
10_1101-2021_02_08_430070	271	24	2021	2021	CD
10_1101-2021_02_08_430070	271	25	.	.	.
10_1101-2021_02_08_430070	271	26	;	;	:
10_1101-2021_02_08_430070	271	27	https://doi.org/10.1101/2021.02.08.430070doi	https://doi.org/10.1101/2021.02.08.430070doi	NFP
10_1101-2021_02_08_430070	271	28	:	:	:
10_1101-2021_02_08_430070	271	29	bioRxiv	biorxiv	VB
10_1101-2021_02_08_430070	271	30	preprint	preprint	NN
10_1101-2021_02_08_430070	271	31	https://doi.org/10.1101/2021.02.08.430070	https://doi.org/10.1101/2021.02.08.430070	UH
10_1101-2021_02_08_430070	271	32	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	NN
10_1101-2021_02_08_430070	271	33	http://creativecommons.org/licenses/by-nc-nd/4.0/	http://creativecommons.org/licenses/by-nc-nd/4.0/	CD