I
nte
rna
t
io
na
l J
o
urna
l o
f
Adv
a
nces in Applie
d Science
s
(
I
J
AAS)
Vo
l.
1
5
,
No
.
1
,
Ma
r
ch
20
2
6
,
p
p
.
3
2
2
~
3
3
2
I
SS
N:
2252
-
8
8
1
4
,
DOI
:
1
0
.
1
1
5
9
1
/ijaas
.
v
1
5
.
i
1
.
pp
322
-
3
3
2
322
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
a
s
.
ia
esco
r
e.
co
m
H
y
brid d
eep l
ea
rning
appro
a
ch f
o
r Indo
nesia
n hoa
x
de
tec
tion:
a
com
pa
ra
tive eva
lua
tion with
In
d
o
BE
RT
Siti
M
ujila
h
wa
t
i,
M
o
h.
Ro
s
i
di Z
a
m
ro
ni
,
M
if
t
a
hu
s
S
ho
lih
in
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
c
s
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
Te
c
h
n
o
l
o
g
y
,
U
n
i
v
e
r
si
t
a
s I
sl
a
m
La
m
o
n
g
a
n
,
L
a
m
o
n
g
a
n
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
May
26
,
2
0
2
5
R
ev
is
ed
Sep
23
,
2
0
2
5
Acc
ep
ted
J
an
22
,
2
0
2
6
Th
e
sp
re
a
d
o
f
h
o
a
x
e
s
in
I
n
d
o
n
e
sia
h
a
s
e
sc
a
lat
e
d
sig
n
ifi
c
a
n
tl
y
,
with
o
v
e
r
1
2
,
5
4
7
c
a
se
s
re
c
o
rd
e
d
b
e
twe
e
n
2
0
1
8
a
n
d
2
0
2
3
.
Lo
w
p
u
b
li
c
li
t
e
ra
c
y
a
n
d
u
n
c
o
n
tro
ll
e
d
in
f
o
rm
a
ti
o
n
flo
w
c
o
n
tri
b
u
te
to
t
h
e
ra
p
id
d
isse
m
in
a
ti
o
n
o
f
fa
lse
c
o
n
ten
t
t
h
a
t
fu
e
ls
d
isi
n
fo
rm
a
ti
o
n
a
n
d
so
c
ial
u
n
re
st.
P
re
v
io
u
s
stu
d
ies
h
a
v
e
u
ti
li
z
e
d
a
rti
ficia
l
in
telli
g
e
n
c
e
(AI)
a
p
p
ro
a
c
h
e
s su
c
h
a
s
In
d
o
n
e
sia
b
id
irec
ti
o
n
a
l
e
n
c
o
d
e
r
re
p
re
se
n
tati
o
n
s
fr
o
m
Tr
a
n
sfo
rm
e
rs
(
In
d
o
BERT
)
a
n
d
d
e
e
p
lea
rn
i
n
g
m
o
d
e
ls
li
k
e
l
o
n
g
s
h
o
rt
-
term
m
e
m
o
ry
(L
S
TM
)
,
b
id
irec
ti
o
n
a
l
LS
TM
(BiL
S
TM
),
c
o
n
v
o
lu
t
io
n
a
l
n
e
u
ra
l
n
e
two
r
k
(CNN
),
a
n
d
Tran
sfo
r
m
e
r
-
b
a
se
d
m
e
th
o
d
s.
Ho
we
v
e
r,
m
o
st
re
li
e
d
o
n
a
si
n
g
le
m
o
d
e
li
n
g
p
a
ra
d
ig
m
a
n
d
d
id
n
o
t
a
d
d
re
ss
th
e
trad
e
-
o
ffs
b
e
twe
e
n
c
las
sifica
ti
o
n
p
e
rfo
rm
a
n
c
e
a
n
d
c
o
m
p
u
tati
o
n
a
l
e
fficie
n
c
y
.
Th
is
st
u
d
y
p
ro
p
o
se
s
a
h
y
b
ri
d
a
rc
h
it
e
c
tu
re
th
a
t
in
teg
ra
tes
In
d
o
BERT
with
b
id
i
re
c
ti
o
n
a
l
g
a
t
e
d
re
c
u
rre
n
t
u
n
i
t
(BiG
RU)
a
n
d
BiL
S
TM
t
o
e
n
h
a
n
c
e
I
n
d
o
n
e
sia
n
h
o
a
x
d
e
tec
ti
o
n
.
Us
in
g
4
,
3
1
2
n
e
ws
a
rti
c
les
a
n
d
1
0
-
fo
l
d
c
ro
ss
-
v
a
li
d
a
ti
o
n
,
we
c
o
m
p
a
re
th
e
p
e
rfo
rm
a
n
c
e
o
f
In
d
o
BER
T
–
BiG
RU,
In
d
o
BERT
–
Bi
LS
TM
,
a
n
d
t
h
e
p
ro
p
o
se
d
h
y
b
ri
d
In
d
o
BERT
–
BiG
RU
–
BiL
S
TM
m
o
d
e
l.
Ev
a
l
u
a
ti
o
n
m
e
tr
ics
i
n
c
lu
d
e
a
c
c
u
ra
c
y
,
p
re
c
isio
n
,
r
e
c
a
ll
,
F
1
-
sc
o
re
,
a
n
d
trai
n
in
g
ti
m
e
.
T
h
e
h
y
b
rid
m
o
d
e
l
a
c
h
iev
e
d
th
e
b
e
st
p
e
rfo
rm
a
n
c
e
with
9
8
.
7
3
%
a
c
c
u
ra
c
y
,
9
9
.
0
1
%
re
c
a
ll
,
9
8
.
0
4
%
p
re
c
isio
n
,
a
n
d
9
8
.
9
8
%
F
1
-
sc
o
re
,
wh
il
e
a
lso
re
d
u
c
in
g
trai
n
i
n
g
ti
m
e
c
o
m
p
a
re
d
to
sin
g
le
m
o
d
e
ls.
Th
e
se
f
in
d
i
n
g
s
d
e
m
o
n
stra
te
th
a
t
c
o
m
b
in
i
n
g
BiG
RU
a
n
d
BiL
S
T
M
with
in
t
h
e
In
d
o
BERT
fra
m
e
wo
rk
e
ffe
c
ti
v
e
ly
b
a
lan
c
e
s
p
e
rfo
rm
a
n
c
e
a
n
d
e
fficie
n
c
y
,
m
a
k
in
g
i
t
a
ro
b
u
st
so
lu
ti
o
n
f
o
r
In
d
o
n
e
sia
n
tex
t
c
las
sifica
ti
o
n
.
K
ey
w
o
r
d
s
:
B
id
ir
ec
tio
n
al
g
ated
r
ec
u
r
r
en
t
u
n
it
B
id
ir
ec
tio
n
al
L
STM
Dee
p
lear
n
in
g
Ho
ax
d
etec
tio
n
Hy
b
r
id
m
o
d
el
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Sit
i M
u
jilah
wati
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics E
n
g
in
ee
r
in
g
,
Facu
lty
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
,
Un
iv
e
r
s
itas
I
s
lam
L
am
o
n
g
an
L
am
o
n
g
a
n
,
E
ast J
av
a,
I
n
d
o
n
es
ia
E
m
ail:
m
o
ed
jee@u
n
is
la.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
s
p
r
ea
d
o
f
f
ak
e
n
ews
(
h
o
ax
es)
h
as
b
ec
o
m
e
a
g
lo
b
al
is
s
u
e
th
at
d
em
an
d
s
in
cr
ea
s
in
g
atten
tio
n
,
p
ar
ticu
lar
ly
in
d
ev
elo
p
in
g
co
u
n
tr
ies wh
er
e
d
ig
ital liter
ac
y
le
v
els v
ar
y
s
ig
n
if
ican
tly
am
o
n
g
th
e
p
o
p
u
latio
n
.
T
h
e
r
ap
id
d
is
s
em
in
atio
n
o
f
u
n
v
e
r
if
ied
in
f
o
r
m
atio
n
th
r
o
u
g
h
d
ig
ital
p
latf
o
r
m
s
h
as
co
n
tr
ib
u
ted
to
wid
esp
r
ea
d
d
is
in
f
o
r
m
atio
n
,
p
u
b
lic
an
x
iety
,
u
ltima
tely
,
an
d
s
o
cial
d
i
v
is
io
n
[
1
]
.
T
h
is
p
h
en
o
m
en
o
n
h
as
e
s
ca
lated
s
h
ar
p
ly
in
I
n
d
o
n
esia,
with
1
2
,
5
4
7
h
o
ax
c
ases
r
ep
o
r
ted
o
n
lin
e
b
etwe
en
2
0
1
8
an
d
2
0
2
3
.
A
k
ey
co
n
tr
ib
u
tin
g
f
ac
to
r
to
th
is
tr
en
d
is
th
e
lo
w
lev
el
o
f
i
n
f
o
r
m
atio
n
liter
ac
y
,
wh
ic
h
h
a
m
p
er
s
th
e
p
u
b
lic’
s
a
b
ilit
y
to
ef
f
ec
tiv
ely
d
is
tin
g
u
is
h
b
etwe
en
cr
ed
ib
le
s
o
u
r
ce
s
an
d
m
is
lead
in
g
co
n
ten
t
[
2
]
.
T
o
ad
d
r
ess
th
is
ch
allen
g
e,
m
an
y
r
esear
ch
er
s
h
av
e
d
ev
o
ted
t
h
em
s
elv
es
t
o
u
tili
zin
g
ar
tific
ial
in
tellig
en
ce
(
AI
)
,
p
ar
ticu
lar
ly
d
ee
p
lear
n
in
g
a
n
d
n
at
u
r
al
la
n
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P),
to
d
etec
t
f
ak
e
n
ews.
T
r
an
s
f
o
r
m
e
r
-
b
ased
m
o
d
els
li
k
e
b
id
i
r
ec
tio
n
al
e
n
co
d
e
r
r
e
p
r
esen
tatio
n
s
f
r
o
m
T
r
an
s
f
o
r
m
e
r
s
(
B
E
R
T
)
an
d
its
v
ar
ian
ts
in
v
ar
io
u
s
lan
g
u
a
g
e
s
,
s
u
ch
as
I
n
d
o
n
esia
B
E
R
T
(
I
n
d
o
B
E
R
T
)
,
h
av
e
d
e
m
o
n
s
tr
ated
o
u
ts
ta
n
d
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Hyb
r
id
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
fo
r
I
n
d
o
n
esia
n
h
o
a
x
d
etec
tio
n
:
a
c
o
mp
a
r
a
tive
…
(
S
iti Mu
jila
h
w
a
ti
)
323
p
er
f
o
r
m
an
ce
in
T
r
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
s
u
ch
as
B
E
R
T
an
d
its
v
ar
ian
ts
in
v
ar
io
u
s
lan
g
u
ag
es,
s
u
ch
as
I
n
d
o
B
E
R
T
,
h
av
e
s
h
o
wn
e
x
c
ellen
t
p
er
f
o
r
m
an
ce
in
s
u
m
m
ar
izatio
n
ap
p
r
o
ac
h
es
f
o
r
tex
t
-
b
ased
d
ep
r
ess
io
n
d
etec
tio
n
[
3
]
.
R
ah
m
awa
ti
et
a
l
.
[
4
]
t
h
eir
r
esear
ch
co
m
p
ar
ed
s
ev
er
al
o
f
th
e
b
est
m
o
d
els
in
h
o
ax
n
ews
class
if
icatio
n
an
d
co
n
clu
d
ed
t
h
at
I
n
d
o
B
E
R
T
was
th
e
b
est
m
o
d
e
l
o
u
t
o
f
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM)
an
d
n
aiv
e
B
a
y
es.
Similar
ly
,
R
id
h
o
an
d
Yu
lian
ti
[
5
]
in
te
g
r
ated
I
n
d
o
B
E
R
T
with
class
ical
m
ac
h
in
e
lear
n
in
g
m
o
d
els
(
e.
g
.
,
SVM
an
d
n
aïv
e
B
ay
es),
s
h
o
win
g
th
at
th
is
co
m
b
in
atio
n
im
p
r
o
v
ed
class
if
icatio
n
ac
c
u
r
ac
y
b
y
o
v
er
1
0
%
co
m
p
ar
ed
to
m
o
d
els
with
o
u
t
co
n
tex
tu
al
e
m
b
ed
d
in
g
s
.
Nu
g
r
ah
a
a
n
d
Fu
d
h
o
li
[
6
]
ap
p
lied
B
E
R
T
to
d
etec
t
C
OVI
D
-
1
9
m
is
in
f
o
r
m
atio
n
o
n
I
n
d
o
n
esian
T
witter
an
d
f
o
u
n
d
th
at
B
E
R
T
ef
f
ec
tiv
ely
h
an
d
led
in
f
o
r
m
al
an
d
n
o
is
y
tex
t,
ac
h
iev
in
g
s
tr
o
n
g
F1
-
s
co
r
es
with
m
in
im
al
p
r
ep
r
o
ce
s
s
in
g
.
Yu
s
u
f
an
d
Su
y
a
n
to
[
7
]
em
p
lo
y
ed
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
f
o
r
h
o
a
x
class
if
icatio
n
in
I
n
d
o
n
esian
n
ews,
d
em
o
n
s
tr
atin
g
th
e
m
o
d
el’
s
ca
p
a
b
ilit
y
to
lear
n
s
eq
u
en
tial
d
ep
e
n
d
en
c
ies,
alth
o
u
g
h
c
o
n
s
tr
ain
ed
b
y
t
r
ain
in
g
tim
e
an
d
g
r
ad
ien
t
is
s
u
es.
Fak
h
r
u
zz
am
an
an
d
Gu
n
awa
n
[
8
]
u
s
ed
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
with
d
ata
au
g
m
en
tatio
n
tech
n
iq
u
es,
wh
ich
p
r
o
v
e
d
ef
f
ec
tiv
e
i
n
en
h
a
n
cin
g
m
o
d
el
g
e
n
er
aliza
tio
n
o
n
s
m
all
d
atasets
.
Alth
o
u
g
h
th
ese
s
tu
d
ies
d
em
o
n
s
tr
ate
th
e
p
o
ten
tial
o
f
d
ee
p
lea
r
n
in
g
f
o
r
h
o
ax
d
etec
tio
n
,
m
o
s
t
r
ely
o
n
a
s
in
g
le
m
o
d
el
in
g
p
ar
a
d
ig
m
,
ei
th
er
r
ec
u
r
r
e
n
t
n
eu
r
al
n
etwo
r
k
s
(
R
NN)
-
b
ased
o
r
T
r
an
s
f
o
r
m
er
-
b
ased
,
with
o
u
t
a
co
m
p
r
eh
e
n
s
iv
e
an
aly
s
is
o
f
tr
a
d
e
-
o
f
f
s
b
etwe
en
class
if
icatio
n
p
er
f
o
r
m
an
ce
an
d
co
m
p
u
tatio
n
al
ef
f
icien
cy
.
Fo
r
in
s
tan
ce
,
b
id
ir
ec
tio
n
al
L
STM
(
B
iLST
M)
ca
p
tu
r
es
lo
n
g
-
r
an
g
e
d
ep
en
d
en
cies
ef
f
ec
tiv
ely
b
u
t
s
u
f
f
er
s
f
r
o
m
h
ig
h
co
m
p
u
tatio
n
al
c
o
s
t
[
9
]
.
I
n
co
n
tr
ast,
b
id
ir
ec
tio
n
al
g
ated
r
ec
u
r
r
en
t
u
n
it
(
B
iGR
U)
o
f
f
er
s
a
m
o
r
e
lig
h
tweig
h
t
ar
ch
itectu
r
e
with
f
aster
co
n
v
er
g
en
ce
,
th
o
u
g
h
it
m
ay
b
e
l
ess
ex
p
r
ess
iv
e
f
o
r
m
o
d
elin
g
co
m
p
lex
lin
g
u
is
tic
p
atter
n
s
.
T
r
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
lik
e
I
n
d
o
B
E
R
T
p
r
o
v
id
e
r
ich
co
n
tex
tu
al
r
ep
r
esen
tatio
n
s
,
y
et
th
e
y
in
h
er
en
tly
lack
s
eq
u
e
n
tial
m
o
d
elin
g
ca
p
ab
ilit
ies.
T
h
ese
lim
itatio
n
s
ar
e
f
u
r
th
er
am
p
lifie
d
in
th
e
I
n
d
o
n
esian
lan
g
u
ag
e,
wh
ich
f
ea
tu
r
es
co
m
p
lex
m
o
r
p
h
o
lo
g
y
,
in
f
lectio
n
,
an
d
h
ig
h
ly
f
lex
ib
le
wo
r
d
o
r
d
er
.
T
h
er
e
f
o
r
e,
a
m
o
d
el
th
at
co
m
b
i
n
es
th
e
s
tr
en
g
th
s
o
f
b
o
th
r
ec
u
r
r
en
t
an
d
T
r
an
s
f
o
r
m
e
r
-
b
ased
ap
p
r
o
ac
h
e
s
,
wh
ile
ad
d
r
ess
in
g
th
eir
r
esp
ec
tiv
e
lim
itatio
n
s
,
is
ess
en
tial.
W
h
ile
p
r
ev
io
u
s
s
tu
d
ies
h
av
e
ex
p
lo
r
ed
th
e
u
s
e
o
f
eith
er
T
r
an
s
f
o
r
m
e
r
-
b
ased
m
o
d
els
o
r
R
NNs
f
o
r
I
n
d
o
n
esian
h
o
ax
d
etec
tio
n
,
th
ey
h
av
e
n
o
t
e
x
p
licitly
a
d
d
r
ess
ed
th
e
tr
ad
e
-
o
f
f
s
b
et
wee
n
class
if
icatio
n
e
f
f
ec
tiv
en
ess
an
d
co
m
p
u
tatio
n
al
ef
f
icien
c
y
,
n
o
r
p
r
o
p
o
s
ed
a
u
n
if
ied
h
y
b
r
id
f
r
am
ewo
r
k
th
at
lev
e
r
ag
es
th
e
s
tr
en
g
th
s
o
f
b
o
th
.
T
o
ad
d
r
ess
th
is
g
ap
,
th
is
p
ap
er
p
r
esen
ts
a
n
o
v
el
h
y
b
r
id
d
ee
p
lear
n
in
g
m
o
d
el
th
at
in
teg
r
ates
B
iG
R
U
an
d
B
iL
STM
lay
er
s
with
I
n
d
o
B
E
R
T
em
b
ed
d
in
g
s
f
o
r
I
n
d
o
n
esian
f
ak
e
n
ews
d
et
ec
tio
n
.
T
h
e
h
y
b
r
id
ar
ch
itectu
r
e
is
d
esig
n
e
d
to
lev
er
ag
e
th
e
co
m
p
le
m
en
tar
y
s
tr
en
g
th
s
o
f
ea
ch
co
m
p
o
n
en
t:
B
iGR
U
o
f
f
er
s
ef
f
icien
t
tr
ain
in
g
an
d
f
aster
c
o
n
v
e
r
g
en
c
e,
wh
ile
B
iLST
M
ca
p
tu
r
es
co
m
p
lex
an
d
lo
n
g
-
r
a
n
g
e
d
ep
en
d
en
cies
in
s
e
q
u
en
tial
d
ata.
No
tab
ly
,
B
iGR
U
h
as
d
em
o
n
s
tr
ated
s
u
p
er
io
r
v
ali
d
atio
n
ac
cu
r
ac
y
co
m
p
a
r
ed
to
B
iLST
M
(
0
.
9
6
9
v
s
.
0
.
9
4
7
)
an
d
ex
h
ib
its
f
aster
co
n
v
er
g
en
ce
d
u
r
in
g
tr
ain
in
g
[
1
0
]
.
W
h
en
co
m
b
in
e
d
with
co
n
tex
tu
alize
d
em
b
ed
d
in
g
s
f
r
o
m
I
n
d
o
B
E
R
T
,
th
e
h
y
b
r
id
co
n
f
ig
u
r
atio
n
is
ex
p
ec
ted
to
o
u
tp
er
f
o
r
m
m
o
d
els
th
at
r
ely
s
o
lely
o
n
T
r
a
n
s
f
o
r
m
er
-
b
ased
o
r
n
o
n
-
c
o
n
tex
tu
al
ar
c
h
itectu
r
e
s
.
Fo
r
e
x
am
p
le,
T
r
an
s
f
o
r
m
er
-
b
ased
h
y
b
r
id
m
o
d
els
in
co
r
p
o
r
atin
g
B
iGR
U
h
av
e
ac
h
iev
ed
u
p
to
9
9
.
7
%
ac
cu
r
ac
y
in
f
a
k
e
n
ews
d
etec
tio
n
b
ef
o
r
e
th
e
1
0
t
h
ep
o
c
h
[
1
1
]
.
T
h
is
co
m
b
in
atio
n
o
f
f
er
s
a
b
alan
ce
d
t
r
ad
e
-
o
f
f
b
etwe
en
ac
cu
r
ac
y
,
co
n
v
er
g
e
n
ce
,
a
n
d
co
m
p
u
tatio
n
al
ef
f
icien
c
y
,
m
a
k
in
g
it
a
p
r
ac
tical
s
o
lu
tio
n
f
o
r
NL
P task
s
in
I
n
d
o
n
esian
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
is
ev
alu
a
ted
u
s
in
g
a
d
ataset
o
f
4
,
3
1
2
n
ews
ar
ticles,
with
p
er
f
o
r
m
an
c
e
v
alid
ated
th
r
o
u
g
h
1
0
-
f
o
ld
c
r
o
s
s
-
v
alid
a
tio
n
.
E
v
alu
atio
n
m
etr
ics
in
clu
d
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
an
d
tr
ain
in
g
tim
e.
Desp
ite
th
e
p
r
o
m
is
in
g
p
r
o
g
r
ess
o
f
p
r
e
v
io
u
s
s
tu
d
ies,
two
cr
itical
g
ap
s
r
em
ai
n
u
n
r
eso
lv
ed
.
First,
m
o
s
t
ap
p
r
o
ac
h
es
h
av
e
f
o
c
u
s
ed
o
n
a
s
in
g
le
m
o
d
elin
g
p
ar
ad
ig
m
,
eith
er
T
r
an
s
f
o
r
m
e
r
-
b
ase
d
o
r
R
NN
,
with
o
u
t
ex
p
li
citly
ad
d
r
ess
in
g
th
e
tr
a
d
e
-
o
f
f
s
b
etwe
en
class
if
icatio
n
p
e
r
f
o
r
m
a
n
ce
an
d
c
o
m
p
u
tatio
n
al
ef
f
icien
cy
.
Seco
n
d
,
th
e
co
m
p
lem
e
n
tar
y
s
tr
en
g
th
s
o
f
B
iGR
U
an
d
B
iLST
M
h
a
v
e
n
o
t
b
ee
n
s
y
s
tem
atica
lly
in
teg
r
ated
with
in
t
h
e
I
n
d
o
B
E
R
T
f
r
am
ew
o
r
k
f
o
r
I
n
d
o
n
esian
h
o
ax
d
etec
tio
n
.
T
o
b
r
id
g
e
th
ese
g
a
p
s
,
th
is
s
tu
d
y
in
tr
o
d
u
ce
s
a
h
y
b
r
id
I
n
d
o
B
E
R
T
-
B
iGR
U
-
B
iLST
M
ar
ch
itectu
r
e
an
d
m
ak
es th
e
f
o
l
lo
win
g
co
n
tr
i
b
u
tio
n
s
:
i)
No
v
el
h
y
b
r
id
f
r
am
ewo
r
k
:
we
d
esig
n
an
d
im
p
lem
e
n
t
a
h
y
b
r
i
d
ar
ch
itectu
r
e
th
at
in
teg
r
ates
co
n
tex
tu
alize
d
em
b
ed
d
in
g
s
f
r
o
m
I
n
d
o
B
E
R
T
with
B
iG
R
U
an
d
B
iLST
M
la
y
er
s
,
en
ab
lin
g
ef
f
icien
t
co
n
v
e
r
g
en
ce
wh
ile
p
r
eser
v
in
g
t
h
e
ab
ilit
y
to
ca
p
tu
r
e
lo
n
g
-
ter
m
d
ep
e
n
d
en
cies.
ii)
C
o
m
p
r
eh
en
s
iv
e
e
v
alu
atio
n
:
we
s
y
s
tem
atica
lly
ev
alu
ate
th
r
ee
m
o
d
el
v
a
r
ian
ts
(
I
n
d
o
B
E
R
T
-
B
i
GR
U
,
I
n
d
o
B
E
R
T
-
B
iLST
M,
an
d
t
h
e
p
r
o
p
o
s
ed
h
y
b
r
id
I
n
d
o
B
E
R
T
-
B
iG
R
U
-
B
iL
STM
)
o
n
a
d
ataset
o
f
4
,
3
1
2
I
n
d
o
n
esian
n
ews a
r
ticle
s
u
s
in
g
1
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
.
iii)
Per
f
o
r
m
an
ce
an
d
ef
f
icien
cy
: w
e
d
em
o
n
s
tr
ate
th
at
t
h
e
h
y
b
r
i
d
m
o
d
el
ac
h
iev
es
s
u
p
er
i
o
r
ac
c
u
r
ac
y
(
9
8
.
7
3
%
)
an
d
F1
-
s
co
r
e
(
9
8
.
9
8
%)
wh
ile
r
ed
u
cin
g
tr
ain
i
n
g
tim
e
co
m
p
ar
ed
to
s
tan
d
alo
n
e
m
o
d
els,
th
er
eb
y
b
alan
cin
g
ef
f
ec
tiv
en
ess
an
d
ef
f
icien
cy
.
iv
)
Gen
er
alis
ab
ilit
y
an
d
p
r
ac
tical
im
p
licatio
n
s
:
b
ey
o
n
d
h
o
ax
d
et
ec
tio
n
,
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
f
r
am
ewo
r
k
ca
n
b
e
ex
te
n
d
ed
to
o
th
er
NL
P
tas
k
s
in
l
o
w
-
r
eso
u
r
ce
an
d
m
o
r
p
h
o
lo
g
ically
r
ich
lan
g
u
ag
es
,
s
u
p
p
o
r
tin
g
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
s
u
ch
as m
is
in
f
o
r
m
atio
n
f
ilter
in
g
,
m
ed
ia
v
er
if
icatio
n
,
an
d
d
ig
ital liter
ac
y
in
itiativ
es.
Alth
o
u
g
h
th
e
e
x
p
er
im
e
n
tal
ev
alu
atio
n
in
th
is
s
tu
d
y
f
o
cu
s
es
o
n
I
n
d
o
n
esian
tex
ts
,
th
e
p
r
o
p
o
s
ed
h
y
b
r
i
d
I
n
d
o
B
E
R
T
-
B
iGR
U
-
B
iLST
M
f
r
am
ewo
r
k
is
co
n
ce
p
t
u
ally
g
e
n
er
aliza
b
le.
I
ts
d
esig
n
is
p
ar
ti
cu
lar
ly
r
elev
a
n
t
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
2
0
2
6
:
322
-
3
3
2
324
m
u
ltil
in
g
u
al
h
o
a
x
d
etec
tio
n
in
lo
w
-
r
eso
u
r
ce
an
d
m
o
r
p
h
o
l
o
g
i
ca
lly
r
ich
lan
g
u
a
g
es,
wh
er
e
s
im
ilar
ch
allen
g
es
o
f
d
ata
s
ca
r
city
,
n
o
is
y
tex
t,
a
n
d
c
o
m
p
lex
lin
g
u
is
tic
s
tr
u
ctu
r
es o
cc
u
r
.
T
h
e
r
em
ain
d
er
o
f
th
is
p
ap
e
r
is
o
r
g
an
ized
as
f
o
llo
ws
:
s
ec
tio
n
2
p
r
esen
ts
th
e
m
eth
o
d
o
lo
g
y
a
n
d
ar
ch
itectu
r
e
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el.
Sectio
n
3
d
escr
ib
es
th
e
d
atasets
an
d
e
x
p
er
im
e
n
t
al
s
etu
p
.
Sectio
n
4
d
is
cu
s
s
es
th
e
r
esu
lts
an
d
p
e
r
f
o
r
m
a
n
ce
ev
al
u
atio
n
.
Fin
ally
,
s
ec
ti
o
n
5
c
o
n
clu
d
es
th
e
p
ap
er
a
n
d
s
u
g
g
ests
d
ir
ec
tio
n
s
f
o
r
f
u
tu
r
e
r
esear
ch
.
2.
M
E
T
H
O
D
T
h
is
s
tu
d
y
co
m
p
ar
es
th
e
p
e
r
f
o
r
m
an
ce
o
f
th
r
ee
I
n
d
o
B
E
R
T
-
b
ased
m
o
d
els
,
B
iGR
U,
B
iLS
T
M,
an
d
a
h
y
b
r
id
B
iGR
U
-
B
iLST
M
f
o
r
I
n
d
o
n
esian
h
o
ax
d
etec
tio
n
.
T
h
e
m
eth
o
d
o
lo
g
ical
wo
r
k
f
lo
w
co
v
er
s
f
iv
e
s
tag
es:
d
ata
co
llectio
n
,
p
r
ep
r
o
ce
s
s
in
g
,
f
ea
tu
r
e
ex
tr
ac
tio
n
with
I
n
d
o
B
E
R
T
,
m
o
d
el
tr
ain
in
g
,
an
d
p
er
f
o
r
m
an
ce
ev
alu
atio
n
.
T
h
e
h
y
b
r
id
d
esig
n
is
m
o
tiv
ated
b
y
p
r
io
r
wo
r
k
s
h
o
win
g
th
e
ef
f
ec
tiv
e
n
ess
o
f
co
m
b
in
in
g
T
r
an
s
f
o
r
m
e
r
em
b
ed
d
in
g
s
w
ith
r
ec
u
r
r
en
t
n
etwo
r
k
s
(
e.
g
.
,
B
E
R
T
-
B
iGR
U
f
o
r
s
en
tim
en
t
an
aly
s
is
an
d
B
E
R
T
-
C
NN
-
L
STM
f
o
r
h
o
ax
d
etec
tio
n
)
.
Acc
o
r
d
i
n
g
ly
,
th
i
s
s
tu
d
y
in
teg
r
ates
I
n
d
o
B
E
R
T
with
B
iG
R
U
an
d
B
iLST
M
lay
er
s
to
ex
p
lo
it th
eir
co
m
p
lem
e
n
tar
y
s
tr
en
g
th
s
in
ca
p
tu
r
in
g
c
o
n
tex
tu
al
an
d
s
eq
u
en
tial in
f
o
r
m
atio
n
.
Fig
u
r
e
1
illu
s
t
r
ates
th
e
ar
c
h
itectu
r
e
o
f
th
e
r
esear
ch
p
i
p
elin
e.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
with
d
ata
p
r
ep
ar
atio
n
,
in
cl
u
d
in
g
p
r
e
p
r
o
c
ess
in
g
s
tep
s
s
u
ch
as
clea
n
in
g
an
d
to
k
en
izatio
n
.
T
h
e
clea
n
ed
d
ata
is
th
en
p
ass
ed
th
r
o
u
g
h
th
e
I
n
d
o
B
E
R
T
m
o
d
el
to
o
b
tain
co
n
tex
tu
al
em
b
e
d
d
i
n
g
s
,
wh
ich
ar
e
s
u
b
s
eq
u
e
n
tly
f
ed
in
to
th
r
ee
ty
p
es
o
f
d
ee
p
lear
n
in
g
class
if
ier
s
:
B
iLST
M,
B
iG
R
U,
an
d
a
h
y
b
r
id
B
iGR
U
-
B
iLST
M
m
o
d
el.
T
h
e
o
u
t
p
u
ts
ar
e
ev
alu
ated
u
s
in
g
class
if
icatio
n
m
etr
ics in
clu
d
in
g
ac
c
u
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
a
n
d
t
r
ain
in
g
tim
e.
Fig
u
r
e
1
.
Ar
c
h
itectu
r
e
o
f
th
e
r
esear
ch
2
.
1
.
Da
t
a
s
et
T
h
e
d
atas
et
w
as
c
o
m
p
i
le
d
f
r
o
m
tw
o
r
el
ia
b
l
e
s
o
u
r
c
es:
T
u
r
n
B
a
c
k
H
o
a
x
.
id
(
l
ab
ele
d
as
h
o
a
x
)
an
d
Deti
k
.
c
o
m
(
l
a
b
el
ed
as
n
o
n
-
h
o
a
x
)
.
T
h
e
D
eti
k
.
c
o
m
s
u
b
s
e
t
c
o
v
e
r
s
t
h
r
ee
ca
te
g
o
r
i
es:
p
o
li
tics
,
s
p
o
r
ts
,
an
d
tec
h
n
o
lo
g
y
.
I
n
to
tal
,
4
,
3
1
2
n
e
ws
a
r
ti
cles
we
r
e
c
o
ll
ec
t
ed
.
F
o
r
m
o
d
el
d
e
v
el
o
p
m
e
n
t
,
th
e
d
at
as
et
w
as
d
i
v
i
d
e
d
i
n
t
o
tr
a
in
in
g
(
7
0
%
)
,
v
ali
d
ati
o
n
(
2
0
%),
a
n
d
test
in
g
(
1
0
%)
s
e
ts
to
e
n
s
u
r
e
r
o
b
u
s
t
a
n
d
r
e
lia
b
l
e
e
v
al
u
ati
o
n
(
s
e
e
T
a
b
le
1
)
.
T
ab
le
1
.
Data
s
et
s
p
litt
in
g
o
v
er
v
iew
D
a
t
a
C
o
u
n
t
Tr
a
i
n
7
0
%
3
,
01
9
Te
st
1
0
%
4
3
2
V
a
l
i
d
a
t
i
o
n
2
0
%
8
6
1
2
.
2
P
re
pro
ce
s
s
ing
I
n
th
is
s
tag
e,
th
e
d
ataset
u
n
d
er
g
o
es
clea
n
in
g
an
d
lab
elin
g
to
p
r
ep
ar
e
f
o
r
m
o
d
el
tr
ain
i
n
g
.
C
lean
in
g
in
clu
d
es
lo
wer
ca
s
in
g
,
r
e
m
o
v
al
o
f
n
o
n
-
alp
h
ab
etic
p
u
n
ctu
a
tio
n
,
n
o
r
m
aliza
tio
n
,
s
to
p
wo
r
d
elim
in
atio
n
,
a
n
d
to
k
en
izatio
n
[
1
2
]
.
L
a
b
elin
g
is
s
o
u
r
ce
-
b
ased
:
n
ews
f
r
o
m
T
u
r
n
B
ac
k
Ho
ax
.
id
is
ass
ig
n
ed
lab
el
0
(
h
o
a
x
)
,
wh
ile
n
ews
f
r
o
m
Detik
.
co
m
i
s
ass
ig
n
ed
lab
el
1
(
n
o
n
-
h
o
ax
)
.
T
h
e
tex
ts
ar
e
th
en
to
k
en
ized
w
ith
th
e
I
n
d
o
B
E
R
T
W
o
r
d
Piece
to
k
en
izer
,
p
ad
d
ed
o
r
tr
u
n
ca
ted
t
o
a
m
ax
im
u
m
s
eq
u
en
ce
len
g
t
h
o
f
1
2
8
to
k
en
s
,
an
d
co
n
v
er
ted
in
to
in
p
u
t_
id
s
an
d
atten
tio
n
_
m
ask
ten
s
o
r
s
.
Fin
ally
,
th
e
4
,
3
1
2
n
e
ws
ar
ticles
ar
e
s
p
lit
in
to
tr
ain
i
n
g
(
7
0
%),
v
alid
atio
n
(
2
0
%),
an
d
test
(
1
0
%)
s
ets to
en
s
u
r
e
r
eliab
le
ev
alu
atio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Hyb
r
id
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
fo
r
I
n
d
o
n
esia
n
h
o
a
x
d
etec
tio
n
:
a
c
o
mp
a
r
a
tive
…
(
S
iti Mu
jila
h
w
a
ti
)
325
2
.
3
.
E
m
bedd
ing
ex
t
ra
ct
i
o
n wit
h
I
nd
o
B
E
R
T
T
h
e
I
n
d
o
B
E
R
T
m
o
d
el
em
p
lo
y
ed
in
t
h
is
s
tu
d
y
f
o
llo
ws
th
e
s
tan
d
ar
d
B
E
R
T
T
r
an
s
f
o
r
m
er
en
co
d
e
r
ar
ch
itectu
r
e,
as
a
d
ap
ted
f
o
r
th
e
I
n
d
o
n
esian
lan
g
u
ag
e
[
1
3
]
.
E
ac
h
in
p
u
t
tex
t
is
to
k
en
ized
u
s
in
g
th
e
W
o
r
d
Piece
to
k
en
izer
an
d
en
clo
s
ed
b
y
s
p
ec
ial
to
k
en
s
[
C
L
S]
an
d
[
SEP]
.
T
h
e
to
k
en
s
ar
e
tr
an
s
f
o
r
m
ed
in
t
o
v
ec
to
r
r
ep
r
esen
tat
io
n
s
th
r
o
u
g
h
a
co
m
b
in
atio
n
o
f
to
k
en
,
s
eg
m
e
n
t,
an
d
p
o
s
itio
n
al
em
b
e
d
d
in
g
s
,
wh
ich
ar
e
th
e
n
p
r
o
ce
s
s
ed
b
y
m
u
ltip
le
b
id
ir
ec
t
io
n
al
T
r
an
s
f
o
r
m
er
e
n
co
d
er
lay
er
s
to
ca
p
tu
r
e
c
o
n
tex
tu
al
r
elatio
n
s
h
ip
s
with
in
th
e
s
en
ten
ce
.
T
h
e
r
esu
ltin
g
co
n
te
x
tu
al
em
b
ed
d
i
n
g
s
(
in
clu
d
in
g
t
h
e
[
C
L
S]
to
k
en
r
ep
r
esen
tatio
n
)
s
er
v
e
as
th
e
in
p
u
t
f
ea
tu
r
es
f
o
r
th
e
s
u
b
s
eq
u
en
t
d
e
ep
lear
n
in
g
m
o
d
els.
Un
lik
e
ty
p
ical
I
n
d
o
B
E
R
T
class
if
ica
tio
n
s
etu
p
s
th
at
d
ir
ec
tly
ap
p
ly
a
So
f
tm
ax
lay
er
,
in
th
i
s
s
tu
d
y
,
th
e
em
b
ed
d
in
g
s
ar
e
p
ass
ed
in
to
B
iG
R
U,
B
iLST
M
,
an
d
th
e
p
r
o
p
o
s
ed
B
iG
R
U
-
B
iLST
M
h
y
b
r
id
ar
ch
i
tectu
r
es to
en
h
an
ce
s
eq
u
en
tial
m
o
d
elin
g
(
s
ee
Fig
u
r
e
2
)
.
Fig
u
r
e
2
.
I
n
d
o
B
E
R
T
ar
ch
itect
u
r
e
2
.
4
.
Desig
n mo
del
2
.
4
.
1
.
B
idi
re
ct
io
na
l lo
ng
s
ho
rt
-
t
er
m m
e
m
o
ry
a
rc
hite
ct
ure
T
h
e
B
iLST
M
m
o
d
el
ex
ten
d
s
th
e
c
o
n
v
e
n
tio
n
al
L
STM
b
y
p
r
o
ce
s
s
in
g
th
e
in
p
u
t
s
eq
u
en
ce
in
b
o
th
f
o
r
war
d
an
d
b
ac
k
war
d
d
ir
ec
tio
n
s
.
T
h
is
b
id
ir
ec
tio
n
al
s
tr
u
ctu
r
e
allo
ws
th
e
m
o
d
el
to
ca
p
tu
r
e
p
ast
an
d
f
u
tu
r
e
co
n
tex
t
s
im
u
ltan
eo
u
s
ly
,
p
r
o
d
u
cin
g
r
ich
er
s
eq
u
en
ce
r
ep
r
esen
t
atio
n
s
[
1
4
]
.
T
h
e
o
u
tp
u
ts
f
r
o
m
th
e
two
d
ir
ec
tio
n
s
ar
e
co
n
ca
ten
ated
at
ea
ch
tim
e
s
tep
an
d
th
en
p
ass
ed
to
a
d
en
s
e
o
r
class
if
icatio
n
lay
er
.
C
o
m
p
ar
ed
to
s
ta
n
d
ar
d
L
STM
,
B
iLST
M
is
m
o
r
e
ef
f
ec
tiv
e
in
m
o
d
elin
g
lo
n
g
-
ter
m
d
ep
en
d
en
cies
in
s
eq
u
en
tial
d
ata
,
m
ak
in
g
it
s
u
itab
le
f
o
r
tex
t
class
if
icatio
n
task
s
.
T
h
e
o
v
er
all
d
esig
n
is
illu
s
tr
ated
in
Fig
u
r
e
3
.
T
h
is
ar
ch
itectu
r
e
h
as
b
ee
n
wid
ely
ap
p
lied
i
n
NL
P
task
s
s
u
c
h
as
s
en
t
im
en
t
an
al
y
s
is
,
m
ac
h
in
e
tr
a
n
s
latio
n
,
an
d
f
a
k
e
n
ews
d
etec
t
io
n
,
d
em
o
n
s
tr
atin
g
its
ef
f
ec
tiv
en
ess
in
ca
p
tu
r
in
g
s
eq
u
en
tial d
ep
e
n
d
en
cies.
Fig
u
r
e
3
.
B
iLST
M
ar
ch
itectu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
2
0
2
6
:
322
-
3
3
2
326
2
.
4
.
2
.
B
idi
re
ct
io
na
l g
a
t
ed
re
c
urre
nt
un
it
a
rc
hite
ct
ure
B
iG
R
U
f
u
n
ctio
n
s
s
im
ilar
ly
to
B
iLST
M
in
th
at
b
o
th
p
r
o
ce
s
s
es
d
ata
in
two
d
ir
ec
tio
n
s
,
f
o
r
war
d
an
d
b
ac
k
war
d
,
s
im
u
ltan
eo
u
s
ly
.
T
h
is
allo
ws
f
o
r
th
e
r
ec
o
g
n
itio
n
o
f
d
ep
e
n
d
en
c
y
r
elatio
n
s
h
ip
s
in
b
o
th
d
ir
ec
tio
n
s
a
n
d
th
e
h
an
d
lin
g
o
f
co
m
p
le
x
tem
p
o
r
al
d
y
n
am
ics
[
1
5
]
,
[
1
6
]
.
I
n
th
e
B
iGR
U
ar
ch
itectu
r
e,
th
e
in
p
u
t
s
eq
u
en
ce
is
d
iv
id
ed
in
t
o
two
p
ath
s
:
o
n
e
f
o
llo
win
g
th
e
o
r
ig
in
al
s
eq
u
e
n
c
e
(
f
o
r
war
d
p
ath
)
an
d
o
n
e
f
o
ll
o
win
g
th
e
r
e
v
er
s
e
s
eq
u
en
ce
(
b
ac
k
war
d
p
ath
)
[
1
7
]
.
E
ac
h
p
ath
way
c
o
m
p
r
is
es
GR
U
u
n
its
th
at
u
tili
ze
g
atin
g
m
ec
h
a
n
is
m
s
to
r
eg
u
late
th
e
f
lo
w
o
f
in
f
o
r
m
a
tio
n
(
s
ee
Fig
u
r
e
4
)
.
T
h
e
f
o
r
war
d
p
ath
GR
U
ca
p
tu
r
es
in
f
o
r
m
atio
n
f
r
o
m
th
e
p
r
ev
io
u
s
tim
e
t
o
th
e
cu
r
r
en
t
ti
m
e,
wh
ile
th
e
b
ac
k
war
d
p
ath
GR
U
ca
p
tu
r
es
in
f
o
r
m
atio
n
f
r
o
m
th
e
later
tim
e
to
th
e
cu
r
r
en
t
tim
e.
T
h
e
r
esu
lts
o
f
th
ese
two
p
ath
s
ar
e
th
en
co
m
b
in
ed
to
f
o
r
m
th
e
f
in
al
r
e
p
r
esen
tatio
n
at
ea
ch
p
o
in
t in
tim
e.
C
o
m
p
ar
e
d
to
B
iLST
M,
B
iG
R
U
u
s
es f
ewe
r
p
ar
a
m
eter
s
,
wh
ich
r
ed
u
ce
s
co
m
p
u
tatio
n
al
co
s
t w
h
ile
m
ain
tain
in
g
co
m
p
etitiv
e
p
er
f
o
r
m
an
ce
in
s
eq
u
en
tial m
o
d
elin
g
task
s
.
Fig
u
r
e
4
.
B
iGR
U
ar
c
h
itectu
r
e
[
1
8
]
2
.
4
.
3
.
P
ro
po
s
ed
hy
brid
I
n
ad
d
itio
n
t
o
ev
alu
ati
n
g
B
iGR
U
an
d
B
iLST
M
s
ep
ar
ately
,
t
h
is
s
tu
d
y
p
r
o
p
o
s
es
a
h
y
b
r
id
ar
ch
itectu
r
e
th
at
in
teg
r
ates
b
o
th
o
n
t
o
p
o
f
I
n
d
o
B
E
R
T
em
b
ed
d
in
g
s
.
T
h
e
aim
is
to
co
m
b
in
e
B
iGR
U’
s
tr
ain
in
g
ef
f
icien
cy
with
B
iLST
M’
s
ab
ilit
y
to
c
ap
tu
r
e
lo
n
g
-
ter
m
s
eq
u
en
tial
d
ep
en
d
e
n
cies,
th
er
eb
y
ad
d
r
e
s
s
in
g
th
e
tr
ad
e
-
o
f
f
b
etwe
en
co
m
p
u
tatio
n
al
s
p
ee
d
an
d
co
n
tex
tu
al
ac
cu
r
ac
y
in
I
n
d
o
n
esian
h
o
ax
d
etec
tio
n
.
T
h
e
o
v
er
all
d
esig
n
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
is
illu
s
tr
ated
in
Fig
u
r
e
5
.
Fig
u
r
e
5
.
Hy
b
r
id
m
o
d
el
ar
c
h
itectu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Hyb
r
id
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
fo
r
I
n
d
o
n
esia
n
h
o
a
x
d
etec
tio
n
:
a
c
o
mp
a
r
a
tive
…
(
S
iti Mu
jila
h
w
a
ti
)
327
W
e
s
tack
a
b
id
ir
ec
t
io
n
al
GR
U
f
o
llo
wed
b
y
a
B
iLST
M
o
n
to
p
o
f
I
n
d
o
B
E
R
T
em
b
ed
d
in
g
s
to
co
m
b
in
e
f
ast
co
n
v
er
g
en
ce
(
B
iGR
U)
with
lo
n
g
-
r
an
g
e
d
e
p
en
d
e
n
cy
m
o
d
elin
g
(
B
iLST
M)
.
T
h
e
s
en
ten
ce
em
b
e
d
d
in
g
is
th
en
p
o
o
led
an
d
f
ed
to
a
lig
h
t
weig
h
t c
lass
if
ier
.
Pip
elin
e.
1)
I
n
p
u
t te
x
t →
I
n
d
o
B
E
R
T
to
k
en
i
ze
r
→
in
p
u
t_
id
s
,
atten
tio
n
_
m
ask
(
m
ax
s
eq
len
e
.
g
.
,
1
2
8
)
.
2)
I
n
d
o
B
E
R
T
en
co
d
e
r
→
co
n
te
x
tu
al
to
k
en
e
m
b
ed
d
in
g
s
(
h
i
d
d
en
s
ize
p
er
to
k
en
p
er
s
tep
)
.
3)
B
iG
R
U
(
b
id
ir
ec
tio
n
al)
→
ca
p
t
u
r
es e
f
f
icien
t b
id
ir
ec
ti
o
n
al
p
at
ter
n
s
.
4)
B
iLST
M
(
b
id
ir
ec
tio
n
al)
→
en
r
ich
es lo
n
g
-
ter
m
s
eq
u
en
t
ial
d
e
p
en
d
en
cies.
5)
Glo
b
al
av
er
ag
e
p
o
o
lin
g
→
B
atch
No
r
m
→
d
r
o
p
o
u
t
.
6)
Den
s
e
(
R
eL
U)
→
Den
s
e
(
So
f
tm
ax
,
2
class
es: h
o
ax
/
n
o
n
-
h
o
ax
)
.
Desig
n
r
atio
n
ale.
–
B
iG
R
U
→
B
iLST
M
o
r
d
er
:
GR
U’
s
lig
h
ter
g
atin
g
ac
ce
ler
ate
s
ea
r
ly
co
n
v
er
g
en
ce
an
d
s
tab
ilizes
g
r
ad
ien
ts
;
L
STM
th
en
ca
p
tu
r
es lo
n
g
er
d
e
p
en
d
en
cies th
at
GR
U
m
ay
m
is
s
.
–
P
o
o
lin
g
an
d
r
e
g
u
lar
izatio
n
:
g
l
o
b
al
p
o
o
lin
g
,
b
atch
n
o
r
m
aliza
tio
n
,
an
d
d
r
o
p
o
u
t
r
ed
u
ce
d
im
e
n
s
io
n
ality
an
d
m
itig
ate
o
v
er
f
itti
n
g
w
h
ile
k
ee
p
in
g
th
e
class
if
ier
co
m
p
ac
t.
–
C
o
m
p
atib
ilit
y
with
I
n
d
o
B
E
R
T
:
lev
er
ag
es
co
n
tex
tu
al
em
b
ed
d
in
g
s
with
o
u
t
h
ea
v
y
task
-
s
p
ec
if
ic
h
ea
d
s
,
p
r
eser
v
in
g
T
r
an
s
f
o
r
m
er
s
em
an
tics
.
As
s
u
m
m
ar
ized
in
T
ab
le
2
,
to
k
en
ized
in
p
u
ts
ar
e
en
co
d
ed
b
y
I
n
d
o
B
E
R
T
in
to
(
1
2
8
,
7
6
8
)
co
n
te
x
tu
al
em
b
ed
d
in
g
s
,
p
ass
ed
th
r
o
u
g
h
a
b
id
ir
ec
tio
n
al
GR
U
-
L
STM
s
eq
u
en
ce
la
y
er
,
p
o
o
le
d
to
a
f
i
x
ed
-
len
g
t
h
v
ec
t
o
r
,
r
eg
u
lar
ized
,
an
d
f
in
ally
class
if
ied
v
ia
a
two
-
lay
e
r
d
en
s
e
h
ea
d
with
So
f
tm
ax
.
T
ab
le
2
.
Hy
b
r
id
m
o
d
el
lay
er
c
o
n
f
ig
u
r
atio
n
La
y
e
r
t
y
p
e
N
a
me/
d
e
s
c
r
i
p
t
i
o
n
O
u
t
p
u
t
s
h
a
p
e
P
a
r
a
me
t
e
r
s
I
n
p
u
t
La
y
e
r
i
n
p
u
t
_
i
d
s,
a
t
t
e
n
t
i
o
n
_
mas
k
(
N
o
n
e
,
1
2
8
)
0
La
mb
d
a
-
(
N
o
n
e
,
1
2
8
,
7
6
8
)
0
B
i
d
i
r
e
c
t
i
o
n
a
l
(
G
R
U
+
LST
M
)
-
(
N
o
n
e
,
1
2
8
,
2
5
6
)
3
9
4
,
2
4
0
G
l
o
b
a
l
A
v
e
r
a
g
e
P
o
o
l
i
n
g
1
D
-
(
N
o
n
e
,
2
5
6
)
0
D
r
o
p
o
u
t
-
(
N
o
n
e
,
1
2
8
)
0
D
r
o
p
o
u
t
-
(
N
o
n
e
,
2
5
6
)
0
D
e
n
se
-
(
N
o
n
e
,
1
2
8
)
3
2
,
8
9
6
D
e
n
se
S
o
f
t
ma
x
o
u
t
p
u
t
(
2
c
l
a
s
ses)
(
N
o
n
e
,
2
)
2
5
8
2
.
5
.
M
o
del
ev
a
lua
t
i
o
n
T
h
e
ev
alu
atio
n
o
f
m
o
d
el
p
er
f
o
r
m
a
n
ce
is
co
n
d
u
cted
u
s
in
g
co
n
v
en
tio
n
al
class
if
icatio
n
m
etr
ics,
in
clu
d
in
g
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
a
n
d
th
e
F1
-
s
co
r
e.
I
n
th
is
s
ce
n
ar
io
,
(
tr
u
e
tr
u
e
)
d
en
o
te
s
th
e
q
u
an
tity
o
f
d
ata
th
at
is
ac
cu
r
ately
id
en
tifie
d
as
tr
u
e
(
tr
u
e
p
o
s
itiv
e)
,
wh
er
ea
s
(
tr
u
e
f
alse
)
s
ig
n
if
ies
th
e
am
o
u
n
t
o
f
d
ata
th
at
is
er
r
o
n
eo
u
s
ly
class
if
ied
as
tr
u
e
(
f
alse
n
eg
ativ
e
)
.
Fu
r
t
h
er
m
o
r
e,
(
f
alse
tr
u
e
)
d
e
n
o
tes
t
h
e
am
o
u
n
t
o
f
d
ata
th
at
is
in
co
r
r
ec
tly
class
if
ied
as
f
alse
(
f
alse
p
o
s
itiv
e)
,
an
d
(
f
alse
f
alse
)
d
en
o
tes
th
e
am
o
u
n
t
o
f
d
ata
t
h
at
is
co
r
r
ec
tly
class
if
ied
as f
alse (
tr
u
e
n
eg
ativ
e)
.
=
+
+
+
+
(
1
)
=
+
(
2
)
=
+
(
3
)
1
−
=
2
∗
∗
+
(
4
)
2
.
6
.
M
o
del
v
a
lid
a
t
io
n
T
o
p
r
ev
e
n
t
o
v
er
f
itti
n
g
an
d
en
s
u
r
e
th
e
g
en
er
aliza
b
ilit
y
o
f
th
e
m
o
d
el,
a
k
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
tech
n
iq
u
e
was
em
p
lo
y
ed
with
1
0
f
o
ld
s
.
E
ac
h
m
o
d
el
was
tr
ain
ed
f
o
r
2
0
ep
o
c
h
s
p
er
f
o
ld
,
r
e
s
u
ltin
g
in
a
to
tal
o
f
2
0
0
tr
ain
i
n
g
iter
atio
n
s
ac
r
o
s
s
th
e
f
o
ld
s
.
E
ar
ly
s
to
p
p
in
g
was
ap
p
lied
d
u
r
in
g
tr
ai
n
in
g
to
m
o
n
i
to
r
v
alid
atio
n
lo
s
s
an
d
h
alt
th
e
tr
ain
in
g
p
r
o
ce
s
s
if
n
o
im
p
r
o
v
em
en
t
was
o
b
s
er
v
ed
o
v
er
a
p
r
e
d
ef
in
e
d
p
atien
ce
th
r
esh
o
ld
.
T
h
is
ap
p
r
o
ac
h
n
o
t
o
n
ly
m
in
im
izes
th
e
r
is
k
o
f
o
v
e
r
f
itti
n
g
b
u
t
a
ls
o
en
s
u
r
es
m
o
r
e
s
tab
le
an
d
r
eliab
le
ev
alu
atio
n
r
esu
lts
ac
r
o
s
s
d
if
f
er
en
t d
ata
p
a
r
titi
o
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
2
0
2
6
:
322
-
3
3
2
328
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
I
n
d
o
B
E
R
T
-
b
ased
B
iG
R
U
-
B
iL
STM
h
y
b
r
id
m
o
d
el
was
tr
ain
ed
an
d
test
ed
u
s
in
g
a
cr
o
s
s
-
v
alid
a
tio
n
tech
n
i
q
u
e
f
o
r
1
0
f
o
ld
s
(
k
-
f
o
ld
=1
0
)
to
e
n
s
u
r
e
th
e
s
tab
ilit
y
an
d
g
en
e
r
aliza
tio
n
o
f
th
e
m
o
d
el
to
th
e
d
ata
u
s
ed
.
T
h
e
tr
ain
in
g
was
co
n
d
u
cted
f
o
r
2
0
ep
o
ch
s
with
a
b
atch
s
ize
o
f
3
2
in
ea
ch
f
o
ld
.
T
h
e
ev
al
u
atio
n
r
esu
lts
d
em
o
n
s
tr
ate
th
e
m
o
d
el's
ca
p
ac
ity
to
g
en
e
r
ate
co
n
s
is
ten
t
p
er
f
o
r
m
an
ce
in
th
e
t
r
ain
in
g
,
test
in
g
,
a
n
d
v
alid
atio
n
p
r
o
ce
s
s
es,
with
ac
cu
r
ac
y
v
al
u
es
an
d
o
t
h
er
ev
al
u
atio
n
m
etr
ics
ex
h
ib
itin
g
r
ela
tiv
e
s
tab
ilit
y
ac
r
o
s
s
ea
ch
f
o
l
d
(
s
ee
Fig
u
r
e
6
)
.
T
h
is
f
in
d
in
g
s
u
g
g
ests
th
at
th
e
h
y
b
r
id
m
o
d
el
p
o
s
s
ess
e
s
ef
f
ec
tiv
e
g
en
er
aliza
tio
n
ca
p
ab
ilit
ies
an
d
ca
n
ef
f
icien
tly
h
an
d
le
d
ata
v
ar
iatio
n
s
.
T
h
e
em
p
lo
y
m
en
t
o
f
k
-
f
o
ld
v
alid
a
tio
n
o
f
f
er
s
a
m
o
r
e
co
m
p
r
eh
e
n
s
iv
e
d
ep
ictio
n
o
f
t
h
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
a
n
d
ass
is
ts
in
m
itig
atin
g
th
e
p
o
ten
tial
f
o
r
b
ias
ar
is
in
g
f
r
o
m
r
a
n
d
o
m
d
ata
d
is
tr
ib
u
tio
n
.
Fig
u
r
e
6
.
Gr
a
p
h
ic
tr
ain
in
g
an
d
v
alid
atio
n
ac
cu
r
a
cy
f
o
r
1
0
-
f
o
ld
I
n
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
m
o
d
e
l,
B
iGR
U
an
d
B
iLST
M
ar
e
ar
r
an
g
e
d
s
eq
u
en
tially
to
c
o
m
b
in
e
th
eir
co
m
p
lem
en
tar
y
s
tr
en
g
th
s
.
B
iGR
U,
p
lace
d
f
ir
s
t,
ac
ce
ler
ates
co
n
v
er
g
e
n
ce
with
f
ewe
r
p
ar
am
eter
s
wh
ile
ef
f
ec
tiv
ely
m
o
d
elin
g
b
id
ir
ec
ti
o
n
al
d
ep
en
d
en
cies.
B
iLST
M
th
en
en
h
an
ce
s
th
e
ab
ilit
y
to
ca
p
tu
r
e
lo
n
g
-
ter
m
d
ep
en
d
e
n
cie
s
an
d
co
m
p
lex
c
o
n
tex
tu
al
p
atter
n
s
.
T
h
is
co
n
f
i
g
u
r
atio
n
en
a
b
les
r
ich
er
s
eq
u
en
ce
r
ep
r
esen
tatio
n
s
an
d
lead
s
to
im
p
r
o
v
e
d
class
if
i
ca
tio
n
p
er
f
o
r
m
an
ce
,
as
s
u
m
m
ar
ized
in
T
ab
le
3
.
Fig
u
r
e
7
s
h
o
ws
a
g
r
ap
h
o
f
th
e
r
esu
lts
o
f
th
e
I
n
d
o
B
E
R
T
-
b
ase
d
B
iGR
U
-
B
iLST
M
h
y
b
r
id
m
o
d
el
tr
ain
in
g
p
r
o
ce
s
s
,
wh
ich
i
n
clu
d
es
ch
an
g
es
in
ac
cu
r
ac
y
an
d
lo
s
s
v
alu
es o
v
er
2
0
ep
o
c
h
s
.
T
h
is
g
r
ap
h
p
r
o
v
id
e
s
an
o
v
er
v
iew
o
f
th
e
s
tab
ilit
y
a
n
d
co
n
v
er
g
e
n
ce
o
f
th
e
m
o
d
el
d
u
r
in
g
th
e
tr
ain
i
n
g
an
d
v
alid
atio
n
p
r
o
ce
s
s
.
T
r
ain
in
g
ac
c
u
r
ac
y
in
c
r
ea
s
es
g
r
ad
u
ally
f
r
o
m
th
e
b
e
g
in
n
i
n
g
to
th
e
en
d
o
f
th
e
ep
o
ch
,
i
n
d
icatin
g
th
at
th
e
m
o
d
el
ca
n
lear
n
p
atter
n
s
f
r
o
m
th
e
tr
ain
in
g
d
ata
ef
f
ec
tiv
el
y
.
T
h
e
v
ali
d
atio
n
ac
cu
r
ac
y
d
em
o
n
s
tr
ates
s
tab
ilit
y
an
d
h
ig
h
v
al
u
e
f
r
o
m
th
e
o
u
ts
et,
ex
h
ib
itin
g
s
lig
h
t
v
ar
iatio
n
s
b
u
t
m
ain
tain
i
n
g
a
co
n
s
is
ten
t
u
p
war
d
tr
e
n
d
.
T
h
e
ab
s
en
ce
o
f
an
y
d
is
ce
r
n
ib
le
i
n
d
icatio
n
o
f
o
v
er
f
itti
n
g
is
ev
id
en
ce
d
b
y
th
e
lack
o
f
a
s
u
b
s
tan
tial
d
ec
lin
e
in
v
alid
atio
n
ac
cu
r
ac
y
,
d
esp
ite
t
h
e
s
u
s
tain
ed
in
cr
ea
s
e
i
n
tr
ain
i
n
g
ac
cu
r
ac
y
.
Me
an
wh
ile,
th
e
s
ec
o
n
d
Fig
u
r
e
8
is
a
co
n
f
u
s
io
n
m
atr
ix
th
at
illu
s
tr
ates
th
e
d
is
tr
ib
u
tio
n
o
f
m
o
d
el
p
r
ed
ictio
n
s
f
o
r
ea
c
h
o
f
th
e
class
es
an
d
id
en
tifie
s
th
e
ty
p
e
o
f
m
is
class
if
icatio
n
th
at
o
cc
u
r
r
ed
.
T
o
ev
alu
ate
p
er
f
o
r
m
a
n
ce
im
p
r
o
v
em
en
ts
,
t
h
e
p
r
o
p
o
s
ed
h
y
b
r
id
m
o
d
el
was
co
m
p
a
r
ed
with
s
tan
d
alo
n
e
B
iG
R
U
an
d
B
iLST
M
b
aselin
es.
W
h
ile
B
iG
R
U
ex
ce
lled
i
n
tr
ain
in
g
ef
f
icien
cy
an
d
B
i
L
STM
in
m
o
d
elin
g
lo
n
g
-
ter
m
co
n
tex
t,
th
e
I
n
d
o
B
E
R
T
-
b
ased
B
iGR
U
-
B
iLST
M
h
y
b
r
id
ac
h
iev
ed
s
u
p
er
io
r
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
an
d
s
tab
ilit
y
b
y
co
m
b
in
i
n
g
th
eir
s
tr
en
g
th
s
.
T
h
e
co
m
p
a
r
ativ
e
r
esu
lts
ar
e
s
u
m
m
ar
ized
in
T
ab
le
4
.
T
h
e
ev
alu
atio
n
r
esu
lts
s
h
o
w
th
at
th
e
I
n
d
o
B
E
R
T
-
b
ased
B
iG
R
U
–
B
iLST
M
h
y
b
r
id
m
o
d
el
p
r
o
v
id
es
th
e
m
o
s
t
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
ac
r
o
s
s
all
ev
alu
atio
n
m
etr
ics.
S
p
ec
if
ically
,
it
ac
h
iev
ed
an
ac
c
u
r
ac
y
o
f
9
8
.
7
3
%,
a
r
ec
all
o
f
9
9
.
0
1
%,
a
p
r
ec
is
io
n
o
f
9
8
.
0
4
%,
an
d
an
F1
-
s
co
r
e
o
f
9
8
.
9
8
%.
I
n
c
o
m
p
ar
is
o
n
,
B
iGR
U
an
d
B
iLST
M
m
o
d
els
ac
h
iev
e
d
ac
c
u
r
ac
ies
o
f
9
5
.
5
4
%
an
d
9
5
.
9
4
%,
r
esp
ec
tiv
ely
,
with
l
o
wer
v
alu
es
a
cr
o
s
s
o
th
er
m
etr
ics.
T
h
is
ad
v
an
tag
e
was
co
n
s
is
ten
tly
o
b
s
er
v
ed
ac
r
o
s
s
all
1
0
f
o
l
d
s
,
in
d
icatin
g
s
tr
o
n
g
g
e
n
er
ali
za
tio
n
an
d
s
tab
ilit
y
.
C
at
eg
o
r
y
-
wis
e
r
esu
lts
(
p
o
liti
cs,
s
p
o
r
ts
,
an
d
tech
n
o
lo
g
y
)
also
r
em
ain
co
n
s
is
ten
tly
h
ig
h
with
o
n
ly
m
in
o
r
v
ar
iatio
n
s
(
s
ee
T
ab
le
5
)
,
in
d
icatin
g
r
o
b
u
s
tn
ess
ac
r
o
s
s
h
eter
o
g
en
eo
u
s
co
n
ten
t
t
y
p
es.
I
n
ter
esti
n
g
ly
,
d
esp
ite
its
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Hyb
r
id
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
fo
r
I
n
d
o
n
esia
n
h
o
a
x
d
etec
tio
n
:
a
c
o
mp
a
r
a
tive
…
(
S
iti Mu
jila
h
w
a
ti
)
329
h
ig
h
er
ar
ch
itectu
r
al
co
m
p
le
x
ity
,
th
e
h
y
b
r
i
d
m
o
d
el
also
ac
h
iev
ed
th
e
f
astes
t
tr
ain
in
g
tim
e
(
1
7
2
.
7
2
s
)
,
o
u
tp
er
f
o
r
m
in
g
B
iGR
U
(
2
3
1
.
0
4
s
)
an
d
B
iLST
M
(
2
8
0
.
9
4
s
)
.
T
h
ese
f
in
d
in
g
s
s
u
g
g
est
th
at
th
e
in
teg
r
atio
n
o
f
B
iG
R
U
an
d
B
iLST
M
ef
f
ec
tiv
ely
lev
er
ag
es
th
e
s
eq
u
en
t
ial
m
o
d
elin
g
s
tr
en
g
th
s
o
f
b
o
th
ar
ch
itectu
r
es,
wh
er
e
B
iG
R
U
co
n
tr
ib
u
tes
to
f
aster
co
n
v
er
g
e
n
ce
,
an
d
B
iLST
M
ca
p
tu
r
es
lo
n
g
er
-
ter
m
d
ep
en
d
en
cies.
Un
lik
e
p
r
io
r
s
tu
d
ies
th
at
u
tili
ze
d
B
iLST
M
[
2
0
]
o
r
B
iGR
U
[
1
9
]
in
i
s
o
latio
n
,
th
is
s
tu
d
y
d
e
m
o
n
s
tr
a
tes
th
at
co
m
b
in
in
g
b
o
th
with
in
th
e
I
n
d
o
B
E
R
T
f
r
am
ewo
r
k
o
f
f
er
s
a
m
o
r
e
b
alan
ce
d
an
d
ef
f
icie
n
t
s
o
lu
tio
n
f
o
r
h
o
ax
d
etec
tio
n
in
I
n
d
o
n
esian
lan
g
u
a
g
e
.
T
ab
le
3
.
R
esu
lt
ev
alu
atio
n
p
er
f
o
r
m
an
ce
o
f
th
e
h
y
b
r
id
m
o
d
el
Ev
a
l
u
a
t
i
o
n
V
a
l
u
e
%
A
c
c
u
r
a
c
y
9
8
.
7
3
R
e
c
a
l
l
9
9
.
0
1
P
r
e
c
i
s
i
o
n
9
8
.
0
4
F1
-
sc
o
r
e
9
8
.
9
8
Fig
u
r
e
7
.
Gr
a
p
h
ic
tr
ain
in
g
ac
c
u
r
ac
y
,
v
alid
atio
n
,
a
n
d
lo
s
s
Fig
u
r
e
8
.
C
o
n
f
u
s
io
n
m
atr
i
x
h
o
ax
d
etec
tio
n
T
ab
le
4
.
C
o
m
p
a
r
is
o
n
o
f
th
e
h
y
b
r
id
m
o
d
el
with
th
e
b
aselin
e
M
o
d
e
l
-
b
a
se
d
I
n
d
o
B
ER
T
A
c
c
u
r
a
c
y
%
R
e
c
a
l
l
%
P
r
e
c
i
s
i
o
n
%
F1
-
sc
o
r
e
%
Ti
me
/
se
c
o
n
d
B
i
G
R
U
[
1
9
]
9
5
.
5
4
9
6
.
1
5
9
5
.
0
1
9
5
.
5
8
2
3
1
.
0
4
B
i
LST
M
[
2
0
]
9
5
.
9
4
9
6
.
6
6
9
5
.
3
0
9
5
.
9
7
2
8
0
.
9
4
B
i
G
R
U
-
B
i
LS
TM
h
y
b
r
i
d
9
8
.
7
3
9
9
.
0
1
9
8
.
0
4
9
8
.
9
8
1
7
2
.
7
2
T
ab
le
5
.
Per
f
o
r
m
an
ce
o
f
th
e
h
y
b
r
id
m
o
d
el
ac
r
o
s
s
ca
teg
o
r
ies
C
a
t
e
g
o
r
y
A
c
c
u
r
a
c
y
%
R
e
c
a
l
l
%
P
r
e
c
i
s
i
o
n
%
F1
-
sc
o
r
e
%
P
o
l
i
t
i
c
s
9
9
.
16
9
9
.1
6
9
8
.
8
9
9
9
.
02
Te
c
h
n
o
l
o
g
y
9
8
.9
8
9
9
.
44
9
8
.
4
8
9
8
.9
6
S
p
o
r
t
s
9
9
.
04
9
9
.
86
98
.
6
2
9
9
.
24
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
2
0
2
6
:
322
-
3
3
2
330
Ov
er
all,
th
ese
r
esu
lts
p
r
o
v
e
th
at
th
e
I
n
d
o
B
E
R
T
-
b
ased
h
y
b
r
id
B
iGR
U
-
B
iLST
M
ap
p
r
o
ac
h
is
an
ef
f
ec
tiv
e
s
tr
ateg
y
to
im
p
r
o
v
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
tex
t
class
if
icatio
n
s
y
s
tem
s
,
esp
ec
ially
in
t
h
e
co
n
tex
t
o
f
n
ews
o
r
in
f
o
r
m
atio
n
d
etec
tio
n
b
as
ed
o
n
th
e
I
n
d
o
n
esian
lan
g
u
a
g
e.
T
h
e
s
u
p
er
io
r
ity
in
all
asp
ec
ts
o
f
ev
alu
atio
n
m
etr
ics
an
d
tim
e
ef
f
icie
n
cy
m
ak
es
th
is
m
o
d
el
v
er
y
f
ea
s
ib
le
to
b
e
ap
p
lied
i
n
r
ea
l
s
ce
n
a
r
io
s
with
lar
g
e
an
d
co
m
p
lex
d
ata
v
o
lu
m
es.
Desp
ite
its
h
ig
h
p
er
f
o
r
m
an
ce
,
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
m
o
d
el
h
as
s
ev
er
al
lim
itatio
n
s
.
I
t
s
d
u
al
-
lay
er
s
eq
u
en
tial
ar
ch
itect
u
r
e
in
cr
ea
s
es
th
e
n
u
m
b
er
o
f
p
ar
am
ete
r
s
an
d
m
em
o
r
y
u
s
ag
e
co
m
p
ar
ed
to
s
in
g
le
m
o
d
els.
Mo
r
eo
v
e
r
,
th
e
m
o
d
e
l
h
as
o
n
ly
b
ee
n
ev
alu
ated
o
n
a
b
in
ar
y
-
class
d
ataset,
an
d
its
ef
f
ec
tiv
en
ess
o
n
m
u
lti
-
class
o
r
m
u
lti
-
lab
el
h
o
ax
d
etec
tio
n
r
em
ain
s
t
o
b
e
i
n
v
esti
g
ated
.
Fu
tu
r
e
wo
r
k
s
h
o
u
ld
also
ev
alu
ate
in
f
er
en
ce
s
p
ee
d
an
d
d
ep
lo
y
ab
i
lity
o
n
lo
w
-
r
eso
u
r
ce
o
r
ed
g
e
d
ev
ices to
en
s
u
r
e
r
ea
l
-
tim
e
f
ea
s
ib
ilit
y
.
T
h
e
f
in
d
i
n
g
s
o
f
th
is
s
tu
d
y
h
a
v
e
s
ev
er
al
im
p
o
r
tan
t
im
p
licatio
n
s
.
First,
th
ey
v
alid
ate
t
h
e
s
y
n
er
g
is
tic
ef
f
ec
t
o
f
c
o
m
b
in
in
g
B
iGR
U
an
d
B
iLST
M
o
n
to
p
o
f
I
n
d
o
B
E
R
T
,
r
esu
ltin
g
in
im
p
r
o
v
e
d
clas
s
if
icatio
n
ac
cu
r
ac
y
an
d
co
m
p
u
tatio
n
al
e
f
f
icien
cy
.
T
h
is
d
ir
ec
tly
ad
d
r
ess
es
a
k
ey
tr
ad
e
-
o
f
f
in
p
r
io
r
s
tu
d
ies,
wh
er
e
eith
er
m
o
d
el
ef
f
ec
tiv
en
ess
o
r
t
r
ain
in
g
ef
f
ici
en
cy
was
o
f
ten
p
r
io
r
itized
,
r
ar
ely
b
o
th
.
Un
lik
e
p
r
e
v
io
u
s
wo
r
k
s
th
at
r
elied
s
o
lel
y
o
n
eith
er
T
r
an
s
f
o
r
m
e
r
-
b
ased
o
r
R
NN
-
b
ased
m
o
d
els,
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
f
r
am
ewo
r
k
lev
er
ag
es
th
e
s
tr
en
g
th
s
o
f
b
o
t
h
,
m
ak
i
n
g
it
a
p
r
ac
tical
an
d
h
i
g
h
-
p
e
r
f
o
r
m
in
g
s
o
lu
tio
n
.
R
esid
u
al
er
r
o
r
s
ar
e
c
o
n
ce
n
tr
ated
o
n
b
o
r
d
er
lin
e
ca
s
es
s
u
ch
as
s
atir
e,
clic
k
b
ait,
o
r
p
ar
tial
ly
tr
u
e
cl
aim
s
,
s
u
g
g
esti
n
g
th
e
v
alu
e
o
f
in
co
r
p
o
r
atin
g
s
ty
lis
tic/p
r
ag
m
atic
cu
es in
f
u
tu
r
e
wo
r
k
.
T
h
e
m
o
d
el'
s
ab
ilit
y
to
ac
h
iev
e
f
aster
co
n
v
er
g
e
n
ce
an
d
h
ig
h
er
ac
cu
r
ac
y
d
em
o
n
s
tr
ates
its
s
u
itab
ilit
y
f
o
r
r
ea
l
-
wo
r
ld
d
ep
l
o
y
m
en
ts
,
esp
ec
ially
o
n
m
o
b
ile
o
r
we
b
p
latf
o
r
m
s
wh
er
e
c
o
m
p
u
tatio
n
al
r
eso
u
r
ce
s
a
r
e
lim
ited
.
T
h
is
is
p
a
r
ticu
lar
ly
s
ig
n
if
ican
t
i
n
th
e
I
n
d
o
n
esian
co
n
tex
t,
wh
e
r
e
th
e
lan
g
u
ag
e'
s
r
ich
m
o
r
p
h
o
lo
g
y
an
d
co
m
p
lex
s
y
n
tax
p
o
s
e
c
h
allen
g
es
th
at
th
is
h
y
b
r
id
a
p
p
r
o
ac
h
ca
n
ef
f
ec
tiv
ely
ad
d
r
ess
.
B
ey
o
n
d
tech
n
ical
p
er
f
o
r
m
an
ce
,
th
is
s
u
itab
ilit
y
also
ex
ten
d
s
to
r
ea
l
-
wo
r
ld
ap
p
l
icatio
n
s
s
u
ch
as
s
o
cial
m
ed
ia
m
o
d
er
atio
n
,
m
ed
ia
v
er
if
icatio
n
,
a
n
d
d
ig
ital
liter
ac
y
in
itiativ
es,
r
ein
f
o
r
cin
g
th
e
r
o
le
o
f
AI
in
s
tr
en
g
t
h
en
in
g
p
u
b
lic
tr
u
s
t
in
in
f
o
r
m
atio
n
s
y
s
tem
s
.
T
h
e
h
y
b
r
id
m
o
d
el
ar
ch
itectu
r
e
p
r
esen
t
s
p
r
o
m
is
in
g
av
en
u
es
f
o
r
e
x
p
a
n
s
io
n
in
to
d
iv
er
s
e
NL
P
ap
p
licatio
n
s
,
in
clu
d
in
g
r
u
m
o
r
d
etec
tio
n
[
2
1
]
,
c
y
b
er
b
u
l
ly
in
g
,
a
n
d
o
f
f
e
n
s
iv
e
co
n
te
n
t
d
etec
tio
n
[
2
2
]
–
[
2
5
]
,
an
d
m
en
tal
h
ea
lth
class
if
icati
o
n
f
r
o
m
s
o
cial
m
ed
ia
[
2
6
]
,
[
2
7
]
.
T
h
ese
task
s
h
av
e
s
ee
n
g
r
o
win
g
in
ter
est
d
u
e
to
th
eir
s
o
cial
im
p
ac
t
an
d
d
ata
c
o
m
p
lex
ity
.
Fu
r
th
e
r
wo
r
k
co
u
l
d
ex
p
lo
r
e
m
o
d
el
co
m
p
r
ess
io
n
an
d
q
u
an
tizatio
n
t
o
en
h
an
ce
r
ea
l
-
tim
e
d
ep
lo
y
m
en
t
in
co
n
s
tr
ain
ed
en
v
ir
o
n
m
en
ts
.
4.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
in
tr
o
d
u
ce
d
a
h
y
b
r
id
d
ee
p
lear
n
in
g
m
o
d
el
in
teg
r
atin
g
I
n
d
o
B
E
R
T
with
B
iGR
U
an
d
B
iLST
M
f
o
r
I
n
d
o
n
esian
h
o
ax
d
etec
tio
n
.
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
ac
h
ie
v
ed
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
(
ac
cu
r
ac
y
9
8
.
7
3
%
an
d
F1
-
s
co
r
e
9
8
.
9
8
%)
co
m
p
a
r
ed
to
s
tan
d
alo
n
e
m
o
d
els,
d
em
o
n
s
tr
at
in
g
th
e
b
en
ef
it
o
f
co
m
b
in
in
g
co
n
te
x
tu
al
an
d
s
e
q
u
en
tial
lear
n
in
g
.
T
h
e
s
y
n
er
g
y
b
etwe
en
B
iGR
U’
s
ef
f
icien
cy
an
d
B
iLST
M’
s
ab
ilit
y
to
ca
p
tu
r
e
lo
n
g
-
ter
m
d
e
p
en
d
en
cies
o
f
f
e
r
s
a
b
alan
ce
d
s
o
lu
tio
n
f
o
r
tex
t
class
if
icatio
n
i
n
m
o
r
p
h
o
lo
g
icall
y
r
ich
lan
g
u
ag
es
s
u
ch
as
I
n
d
o
n
esian
lan
g
u
ag
e
.
Desp
ite
th
ese
s
tr
en
g
th
s
,
lim
itatio
n
s
r
em
ain
r
eg
ar
d
in
g
d
ataset
d
iv
er
s
ity
,
m
o
d
el
c
o
m
p
lex
ity
,
an
d
r
elia
n
ce
o
n
a
s
in
g
le
p
r
e
-
tr
ain
ed
em
b
ed
d
in
g
.
Fu
t
u
r
e
r
esear
ch
s
h
o
u
ld
a
d
d
r
ess
th
ese
asp
ec
ts
an
d
ex
p
lo
r
e
b
r
o
ad
er
ap
p
licatio
n
s
s
u
ch
as
r
u
m
o
r
d
etec
tio
n
,
o
f
f
en
s
iv
e
co
n
ten
t
f
ilter
in
g
,
a
n
d
r
ea
l
-
tim
e
an
aly
tics
.
Ov
er
all,
th
is
wo
r
k
p
r
o
v
id
es
a
p
r
ac
tical
an
d
ef
f
ec
tiv
e
f
r
am
ewo
r
k
with
p
o
ten
tial
to
s
u
p
p
o
r
t
m
ed
ia
m
o
n
ito
r
in
g
,
f
ac
t
-
ch
ec
k
in
g
,
an
d
d
ig
ital
liter
ac
y
in
itiativ
es.
T
h
ese
f
in
d
in
g
s
h
ig
h
lig
h
t
n
o
t
o
n
ly
th
e
m
o
d
el’
s
tech
n
ical
m
er
its
b
u
t
also
its
p
o
ten
tial
f
o
r
r
ea
l
-
t
im
e
f
ilter
in
g
,
f
ac
t
-
ch
ec
k
in
g
s
u
p
p
o
r
t,
an
d
civ
ic
ed
u
ca
tio
n
,
t
h
er
eb
y
co
n
tr
i
b
u
tin
g
to
b
r
o
ad
er
s
o
cial
im
p
ac
t.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
d
id
n
o
t
r
ec
eiv
e
an
y
s
p
ec
if
ic
g
r
an
t
f
r
o
m
f
u
n
d
i
n
g
ag
en
cies
in
th
e
p
u
b
lic,
co
m
m
er
cial,
o
r
not
-
f
o
r
-
p
r
o
f
it secto
r
s
.
All r
esear
ch
ex
p
e
n
s
es we
r
e
f
u
lly
s
u
p
p
o
r
ted
b
y
th
e
au
t
h
o
r
s
th
r
o
u
g
h
p
er
s
o
n
al
f
u
n
d
in
g
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Sit
i M
u
jilah
wati
✓
✓
✓
✓
✓
✓
✓
✓
✓
M
oh
.
R
o
s
id
i Z
am
r
o
n
i
✓
✓
✓
✓
✓
✓
✓
✓
Mif
tah
u
s
Sh
o
lih
in
✓
✓
✓
✓
✓
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Hyb
r
id
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
fo
r
I
n
d
o
n
esia
n
h
o
a
x
d
etec
tio
n
:
a
c
o
mp
a
r
a
tive
…
(
S
iti Mu
jila
h
w
a
ti
)
331
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
e
au
th
o
r
s
d
ec
lar
e
th
at
t
h
ey
h
av
e
n
o
k
n
o
wn
co
n
f
licts
o
f
in
ter
est,
wh
eth
er
f
in
an
cial
o
r
n
o
n
-
f
in
an
cial
(
in
clu
d
in
g
p
o
liti
ca
l,
p
er
s
o
n
al,
r
elig
io
u
s
,
id
eo
l
o
g
ical,
ac
a
d
em
ic,
o
r
in
tellectu
al)
,
th
at
co
u
l
d
h
av
e
in
f
lu
en
ce
d
o
r
b
ee
n
p
er
ce
i
v
ed
to
i
n
f
lu
en
ce
th
e
r
esu
lts
an
d
in
ter
p
r
etatio
n
o
f
th
e
r
esear
ch
r
e
p
o
r
ted
in
th
is
p
a
p
er
.
I
NF
O
RM
E
D
CO
NS
E
N
T
T
h
e
d
ata
an
aly
ze
d
in
th
is
s
tu
d
y
wer
e
o
b
tain
ed
f
r
o
m
p
u
b
licly
av
ailab
le
s
o
u
r
ce
s
.
No
p
er
s
o
n
ally
id
en
tifia
b
le
in
f
o
r
m
atio
n
was c
o
llected
o
r
d
is
clo
s
ed
; th
er
e
f
o
r
e,
in
f
o
r
m
ed
co
n
s
en
t w
as n
o
t
r
e
q
u
ir
ed
.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
e
d
ata
u
s
ed
in
th
is
s
tu
d
y
w
er
e
o
b
tain
ed
f
r
o
m
p
u
b
licly
av
ailab
le
s
o
u
r
ce
s
an
d
d
id
n
o
t
in
v
o
lv
e
d
ir
ec
t
in
ter
ac
tio
n
with
h
u
m
an
s
u
b
je
cts
o
r
th
e
co
llectio
n
o
f
p
er
s
o
n
ally
id
en
tifia
b
le
i
n
f
o
r
m
atio
n
.
T
h
er
ef
o
r
e,
et
h
ical
ap
p
r
o
v
al
was n
o
t
r
eq
u
ir
ed
.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
t
h
at
s
u
p
p
o
r
t
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
a
r
e
o
p
en
l
y
av
ailab
le
o
n
Kag
g
le
at
th
e
f
o
llo
win
g
r
ep
o
s
ito
r
y
:
h
o
ax
n
ews
I
n
d
o
n
esia
d
ataset
(
h
ttp
s
://www.
k
ag
g
le.
co
m
/d
atasets
/v
ijay
an
d
ik
a/h
o
ax
-
n
ews
-
in
d
o
n
esia)
.
T
h
e
d
ataset
was a
cc
ess
ed
an
d
u
s
ed
in
ac
c
o
r
d
a
n
ce
with
Kag
g
le’
s
d
ata
u
s
ag
e
p
o
li
cies.
RE
F
E
R
E
NC
E
S
[
1
]
N
.
R
a
i
,
D
.
K
u
mar,
N
.
K
a
u
s
h
i
k
,
C
.
R
a
j
,
a
n
d
A
.
A
l
i
,
“
F
a
k
e
n
e
w
s
c
l
a
ssi
f
i
c
a
t
i
o
n
u
s
i
n
g
t
r
a
n
sf
o
r
mer
b
a
se
d
e
n
h
a
n
c
e
d
LST
M
a
n
d
B
ER
T,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
C
o
g
n
i
t
i
v
e
C
o
m
p
u
t
i
n
g
i
n
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
3
,
p
p
.
9
8
–
1
0
5
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
j
c
c
e
.
2
0
2
2
.
0
3
.
0
0
3
.
[
2
]
A
.
Z.
M
.
N
o
o
r
,
R
.
G
e
r
n
o
w
o
,
a
n
d
O
.
D
.
N
u
r
h
a
y
a
t
i
,
“
D
a
t
a
a
u
g
me
n
t
a
t
i
o
n
f
o
r
h
o
a
x
d
e
t
e
c
t
i
o
n
t
h
r
o
u
g
h
t
h
e
m
e
t
h
o
d
o
f
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
i
n
I
n
d
o
n
e
s
i
a
n
n
e
w
s,”
J
u
r
n
a
l
Pe
n
e
l
i
t
i
a
n
Pe
n
d
i
d
i
k
a
n
I
PA
,
v
o
l
.
9
,
n
o
.
7
,
p
p
.
5
0
7
8
–
5
0
8
4
,
2
0
2
3
,
d
o
i
:
1
0
.
2
9
3
0
3
/
j
p
p
i
p
a
.
v
9
i
7
.
4
2
1
4
.
[
3]
H
.
S
.
G
a
v
a
l
a
n
,
M
.
N
.
R
a
s
t
g
o
o
,
a
n
d
B
.
N
a
k
i
s
a
,
“
A
B
E
R
T
-
b
a
se
d
s
u
m
mariz
a
t
i
o
n
a
p
p
r
o
a
c
h
f
o
r
d
e
p
r
e
ss
i
o
n
d
e
t
e
c
t
i
o
n
,
”
2
0
2
4
,
a
rXi
v
:
2
4
0
9
.
0
8
4
8
3
.
[
4
]
A
.
R
a
h
m
a
w
a
t
i
,
A
.
A
l
a
ms
y
a
h
,
a
n
d
A
.
R
o
m
a
d
h
o
n
y
,
“
H
o
a
x
n
e
w
s
d
e
t
e
c
t
i
o
n
a
n
a
l
y
s
i
s
u
si
n
g
I
n
d
o
B
ER
T
d
e
e
p
l
e
a
r
n
i
n
g
m
e
t
h
o
d
o
l
o
g
y
,
”
i
n
2
0
2
2
1
0
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
f
o
rm
a
t
i
o
n
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
(
I
C
o
I
C
T
)
,
A
u
g
.
2
0
2
2
,
p
p
.
3
6
8
–
3
7
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
o
I
C
T
5
5
0
0
9
.
2
0
2
2
.
9
9
1
4
9
0
2
.
[
5
]
M
.
Y
.
R
i
d
h
o
a
n
d
E.
Y
u
l
i
a
n
t
i
,
“
F
r
o
m
t
e
x
t
t
o
t
r
u
t
h
:
l
e
v
e
r
a
g
i
n
g
I
n
d
o
B
ER
T
a
n
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
m
o
d
e
l
s
f
o
r
h
o
a
x
d
e
t
e
c
t
i
o
n
i
n
I
n
d
o
n
e
si
a
n
n
e
w
s,
”
J
u
r
n
a
l
I
l
m
i
a
h
T
e
k
n
i
k
El
e
k
t
r
o
K
o
m
p
u
t
e
r
d
a
n
I
n
f
o
rm
a
t
i
k
a
,
v
o
l
.
1
0
,
n
o
.
3
,
p
p
.
5
4
4
–
5
5
5
,
2
0
2
4
,
d
o
i
:
1
0
.
2
6
5
5
5
/
j
i
t
e
k
i
.
v
1
0
i
3
.
2
9
4
5
0
.
[
6
]
F
.
A
.
N
u
g
r
a
h
a
a
n
d
D
.
H
.
F
u
d
h
o
l
i
,
“
C
O
V
I
D
-
1
9
mi
si
n
f
o
r
mat
i
o
n
d
e
t
e
c
t
i
o
n
i
n
I
n
d
o
n
e
si
a
n
t
w
e
e
t
s
u
s
i
n
g
B
ER
T
,
”
J
u
rn
a
l
Me
d
i
a
I
n
f
o
rm
a
t
i
k
a
Bu
d
i
d
a
rm
a
,
v
o
l
.
7
,
n
o
.
2
,
p
p
.
6
0
0
–
6
0
8
,
2
0
2
3
,
d
o
i
:
1
0
.
3
0
8
6
5
/
mi
b
.
v
7
i
2
.
5
6
6
8
.
[
7
]
R
.
Y
u
s
u
f
a
n
d
S
.
S
u
y
a
n
t
o
,
“
H
o
a
x
d
e
t
e
c
t
i
o
n
o
n
I
n
d
o
n
e
s
i
a
n
t
e
x
t
u
si
n
g
l
o
n
g
sh
o
r
t
-
t
e
r
m
mem
o
r
y
,
”
i
n
I
C
O
I
AC
T
2
0
2
2
-
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
n
f
o
rm
a
t
i
o
n
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
s
T
e
c
h
n
o
l
o
g
y
:
A
N
e
w
Wa
y
t
o
M
a
k
e
AI
U
se
f
u
l
f
o
r
Ev
e
ry
o
n
e
i
n
t
h
e
N
e
w
N
o
rm
a
l
Era
,
Pr
o
c
e
e
d
i
n
g
,
2
0
2
2
,
p
p
.
2
6
8
–
271
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
O
I
A
C
T
5
5
5
0
6
.
2
0
2
2
.
9
9
7
2
0
8
6
.
[
8
]
M
.
N
.
F
a
k
h
r
u
z
z
a
m
a
n
a
n
d
S
.
W
.
G
u
n
a
w
a
n
,
“
W
e
b
-
b
a
se
d
a
p
p
l
i
c
a
t
i
o
n
f
o
r
d
e
t
e
c
t
i
n
g
I
n
d
o
n
e
s
i
a
n
c
l
i
c
k
b
a
i
t
h
e
a
d
l
i
n
e
s
u
s
i
n
g
I
n
d
o
B
ER
T
,
”
2
0
2
1
,
a
rXi
v
:
2
1
0
2
.
1
0
6
0
1
.
[
9
]
J.
D
o
n
g
,
“
I
mp
r
o
v
i
n
g
t
h
e
B
i
LST
M
m
o
d
e
l
p
e
r
f
o
r
ma
n
c
e
f
o
r
t
w
e
e
t
se
n
t
i
me
n
t
a
n
a
l
y
si
s,
”
A
p
p
l
i
e
d
a
n
d
C
o
m
p
u
t
a
t
i
o
n
a
l
E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
6
,
n
o
.
1
,
p
p
.
1
1
0
6
–
1
1
1
7
,
Ju
n
.
2
0
2
3
,
d
o
i
:
1
0
.
5
4
2
5
4
/
2
7
5
5
-
2
7
2
1
/
6
/
2
0
2
3
0
4
1
5
.
[
1
0
]
A
.
A
b
d
u
l
l
a
h
,
“
C
l
a
ss
i
f
y
i
n
g
i
n
n
o
v
a
t
i
o
n
w
i
t
h
b
i
LST
M
a
n
d
b
i
G
R
U
:
a
s
t
u
d
y
o
f
l
a
b
o
r
mar
k
e
t
d
a
t
a
i
n
t
h
e
N
e
t
h
e
r
l
a
n
d
s,”
M.
S
.
T
h
e
s
i
s
,
D
e
p
a
r
t
me
n
t
o
f
C
o
g
n
i
t
i
v
e
S
c
i
e
n
c
e
a
n
d
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
T
i
l
b
u
r
g
U
n
i
v
e
r
si
t
y
,
Ti
l
b
u
r
g
,
T
h
e
N
e
t
h
e
r
l
a
n
d
s
,
2
0
2
1
.
[
1
1
]
T.
H
u
a
n
g
,
Z.
X
u
,
P
.
Y
u
,
J.
Y
i
,
a
n
d
X
.
X
u
,
“
A
h
y
b
r
i
d
t
r
a
n
sf
o
r
mer m
o
d
e
l
f
o
r
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
:
l
e
v
e
r
a
g
i
n
g
B
a
y
e
si
a
n
o
p
t
i
m
i
z
a
t
i
on
a
n
d
b
i
d
i
r
e
c
t
i
o
n
a
l
r
e
c
u
r
r
e
n
t
u
n
i
t
,
”
i
n
2
0
2
5
8
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
S
y
m
p
o
si
u
m
o
n
Bi
g
D
a
t
a
a
n
d
A
p
p
l
i
e
d
S
t
a
t
i
st
i
c
s
(
I
S
B
D
AS
)
,
F
e
b
.
2
0
2
5
,
p
p
.
6
9
6
–
701
,
d
o
i
:
1
0
.
1
1
0
9
/
I
S
B
D
A
S
6
4
7
6
2
.
2
0
2
5
.
1
1
1
1
7
0
5
8
.
[
1
2
]
N
.
N
o
o
r
,
N
.
Z.
M
.
S
a
f
a
r
,
a
n
d
C
.
S
u
p
r
i
y
a
n
t
o
,
“
S
e
g
me
n
t
a
t
i
o
n
w
o
r
d
t
o
i
m
p
r
o
v
e
p
e
r
f
o
r
ma
n
c
e
se
n
t
i
me
n
t
a
n
a
l
y
s
i
s
f
o
r
I
n
d
o
n
e
si
a
n
l
a
n
g
u
a
g
e
,
”
Fu
s
i
o
n
:
Pra
c
t
i
c
e
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
5
,
n
o
.
2
,
p
p
.
1
4
5
–
1
5
4
,
2
0
2
4
,
d
o
i
:
1
0
.
5
4
2
1
6
/
F
P
A
.
1
5
0
2
1
3
.
[
1
3
]
F
.
K
o
t
o
,
J.
H
.
L
a
u
,
a
n
d
T.
B
a
l
d
w
i
n
,
“
I
N
D
O
B
ER
TWEET
:
a
p
r
e
t
r
a
i
n
e
d
l
a
n
g
u
a
g
e
m
o
d
e
l
f
o
r
I
n
d
o
n
e
si
a
n
Tw
i
t
t
e
r
w
i
t
h
e
f
f
e
c
t
i
v
e
d
o
m
a
i
n
-
s
p
e
c
i
f
i
c
v
o
c
a
b
u
l
a
r
y
i
n
i
t
i
a
l
i
z
a
t
i
o
n
,
”
i
n
E
MN
L
P
2
0
2
1
-
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
2
1
C
o
n
f
e
re
n
c
e
o
n
Em
p
i
ri
c
a
l
Me
t
h
o
d
s
i
n
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
Pro
c
e
ss
i
n
g
,
2
0
2
1
,
p
p
.
1
0
6
6
0
–
1
0
6
6
8
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
2
0
2
1
.
e
m
n
l
p
-
mai
n
.
8
3
3
.
[
1
4
]
F
.
C
.
Z
e
g
a
r
r
a
,
J
.
V
.
-
M
a
c
h
u
c
a
,
a
n
d
A
.
M
.
C
o
r
o
n
a
d
o
,
“
A
c
o
m
p
a
r
a
t
i
v
e
s
t
u
d
y
o
f
C
N
N
,
L
S
T
M
,
B
i
L
S
T
M
,
a
n
d
G
R
U
a
r
c
h
i
t
e
c
t
u
r
e
s
f
o
r
t
o
o
l
w
e
a
r
p
r
e
d
i
c
t
i
o
n
i
n
m
i
l
l
i
n
g
p
r
o
c
e
s
s
e
s
,
”
J
o
u
r
n
a
l
o
f
M
a
c
h
i
n
e
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
2
3
,
n
o
.
4
,
p
p
.
1
2
2
–
1
3
6
,
2
0
2
3
,
d
o
i
:
1
0
.
3
6
8
9
7
/
j
m
e
/
1
7
4
0
1
9
.
[
1
5
]
B
.
K
h
a
l
d
i
,
F
.
H
a
r
r
o
u
,
A
.
D
a
i
r
i
,
a
n
d
Y
.
S
u
n
,
“
A
d
e
e
p
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
f
r
a
m
e
w
o
r
k
f
o
r
sw
a
r
m
m
o
t
i
o
n
s
p
e
e
d
p
r
e
d
i
c
t
i
o
n
,
”
J
o
u
rn
a
l
o
f
El
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
8
,
n
o
.
5
,
p
p
.
3
8
1
1
–
3
8
2
5
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
2
8
3
5
-
0
2
3
-
0
1
4
4
6
-
7.
[
1
6
]
X
.
Li
n
,
Z.
Q
u
a
n
,
Z
.
J.
W
a
n
g
,
H
.
H
u
a
n
g
,
a
n
d
X
.
Ze
n
g
,
“
A
n
o
v
e
l
m
o
l
e
c
u
l
a
r
r
e
p
r
e
s
e
n
t
a
t
i
o
n
w
i
t
h
B
i
G
R
U
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
l
e
a
r
n
i
n
g
a
t
o
m,”
Bri
e
f
i
n
g
s
i
n
B
i
o
i
n
f
o
r
m
a
t
i
c
s
,
v
o
l
.
2
1
,
n
o
.
6
,
p
p
.
2
0
9
9
–
2
1
1
1
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
9
3
/
b
i
b
/
b
b
z
1
2
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.