I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
p
u
t
er
Science
Vo
l.
23
,
No
.
1
,
J
u
ly
2021
,
p
p
.
345
~
3
5
3
I
SS
N:
2
5
0
2
-
4
7
5
2
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ee
cs.v
23
.i
1
.
pp
345
-
3
5
3
345
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
Senti
m
en
t
cla
ss
ifi
ca
tion o
f
user's
re
v
ie
w
s o
n drugs
bas
ed on
g
lo
ba
l vectors
f
o
r w
o
rd repre
sen
tat
io
n and
bidire
ctio
na
l long
sho
rt
-
ter
m
m
e
m
o
ry
rec
urre
nt
neur
a
l net
w
o
rk
H
a
da
b K
ha
lid
O
ba
y
es
1
,
F
ira
s
Sa
ba
h Al
-
T
ura
ihi
2
,
K
ha
ldo
o
n H
.
Alhu
s
s
a
y
ni
3
1
Co
ll
e
g
e
o
f
e
d
u
c
a
ti
o
n
f
o
r
h
u
m
a
n
it
ies
stu
d
ies
,
Un
iv
e
rsit
y
o
f
Ba
b
y
lo
n
,
Ba
b
y
lo
n
,
Ira
q
2
Co
ll
e
g
e
o
f
in
f
o
rm
a
ti
o
n
tec
h
n
o
l
o
g
y
,
Un
iv
e
rsit
y
o
f
Ba
b
y
lo
n
,
Ba
b
y
l
o
n
,
Ira
q
3
Co
m
p
u
ter Cen
ter,
Un
iv
e
rsity
o
f
Ba
b
y
lo
n
,
Ba
b
y
lo
n
,
Ira
q
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
2
9
,
2
0
2
0
R
ev
i
s
ed
Ma
r
6
,
2
0
2
1
A
cc
ep
ted
J
u
l 1
,
2
0
2
1
T
h
e
p
ro
c
e
ss
o
f
p
ro
d
u
c
t
d
e
v
e
lo
p
m
e
n
t
in
th
e
h
e
a
lt
h
se
c
to
r,
e
sp
e
c
iall
y
p
h
a
rm
a
c
e
u
ti
c
a
ls,
g
o
e
s
th
ro
u
g
h
a
se
ries
o
f
p
re
c
ise
p
ro
c
e
d
u
re
s
d
u
e
t
o
it
s
d
irec
t
re
lev
a
n
c
e
to
h
u
m
a
n
li
f
e
.
T
h
e
o
p
in
io
n
o
f
p
a
ti
e
n
ts
o
n
a
p
a
rti
c
u
lar
d
ru
g
c
a
n
b
e
re
li
e
d
u
p
o
n
i
n
th
is
d
e
v
e
lo
p
m
e
n
t
p
ro
c
e
ss
,
a
s
th
e
p
a
ti
e
n
ts
c
o
n
v
e
y
th
e
ir
e
x
p
e
rien
c
e
w
it
h
th
e
d
ru
g
s
th
ro
u
g
h
th
e
ir
o
p
i
n
io
n
.
T
h
e
so
c
ial
m
e
d
ia
f
ield
p
ro
v
id
e
s
m
a
n
y
d
a
tas
e
ts
r
e
late
d
to
d
ru
g
s
th
r
o
u
g
h
k
n
o
w
in
g
th
e
u
se
r'
s
ra
ti
n
g
a
n
d
o
p
i
n
io
n
o
n
a
d
ru
g
a
f
ter
u
si
n
g
it
.
In
th
is
w
o
rk
,
a
d
a
tas
e
t
is
u
se
d
th
a
t
in
c
lu
d
e
s
th
e
u
se
r’s
ra
ti
n
g
a
n
d
re
v
ie
w
o
n
th
e
d
ru
g
,
f
o
r
th
e
p
u
rp
o
se
o
f
c
las
si
fy
in
g
th
e
u
se
r’s
o
p
in
io
n
s
(re
v
ie
w
s)
w
h
e
th
e
r
th
e
y
a
r
e
p
o
siti
v
e
o
r
n
e
g
a
ti
v
e
.
T
h
e
p
ro
p
o
se
d
m
e
th
o
d
in
c
lu
d
e
s
tw
o
p
h
a
se
s.
T
h
e
f
irst
p
h
a
se
is
to
u
se
th
e
g
lo
b
a
l
v
e
c
to
rs
f
o
r
w
o
rd
re
p
re
se
n
tatio
n
m
o
d
e
l
f
o
r
c
o
n
v
e
rti
n
g
tex
ts
in
to
th
e
v
e
c
to
r
o
f
e
m
b
e
d
d
e
d
w
o
rd
s.
A
s
f
o
r
th
e
se
c
o
n
d
sta
g
e
,
th
e
d
e
e
p
n
e
u
r
a
l
n
e
tw
o
rk
(
b
id
irec
ti
o
n
a
l
lo
n
g
sh
o
rt
-
term
m
e
m
o
r
y
)
is
e
m
p
lo
y
e
d
in
th
e
c
las
si
f
ica
ti
o
n
o
f
re
v
ie
w
s.
T
h
e
u
se
r
'
s
r
a
ti
n
g
is
u
se
d
a
s
a
g
ro
u
n
d
tru
th
i
n
e
v
a
l
u
a
ti
n
g
th
e
c
las
si
f
ica
ti
o
n
re
su
lt
s.
T
h
e
p
ro
p
o
s
e
d
m
e
th
o
d
p
re
se
n
ts
e
n
c
o
u
ra
g
in
g
re
su
lt
s,
a
s
th
e
c
las
si
f
ica
ti
o
n
re
su
lt
s
a
re
e
v
a
lu
a
ted
th
ro
u
g
h
th
re
e
c
rit
e
ria,
n
a
m
e
l
y
p
re
c
isio
n
,
re
c
a
ll
a
n
d
F
-
sc
o
re
,
w
h
o
se
o
b
tai
n
e
d
v
a
lu
e
s
e
q
u
a
l
(
0
.
9
5
4
3
,
0
.
9
5
9
7
a
n
d
0
.
9
5
5
8
),
re
sp
e
c
ti
v
e
ly
.
K
ey
w
o
r
d
s
:
B
id
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
Glo
b
al
v
ec
to
r
s
an
d
p
atien
t
o
p
in
io
n
class
if
icatio
n
R
ev
ie
w
s
a
n
al
y
s
is
Sen
ti
m
e
n
t a
n
a
l
y
s
is
T
h
is
is
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC
BY
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Had
ab
Kh
alid
Ob
a
y
es
Dep
ar
t
m
en
t o
f
g
eo
g
r
ap
h
y
C
o
lleg
e
o
f
e
d
u
ca
tio
n
f
o
r
h
u
m
a
n
itie
s
s
t
u
d
ies
Un
i
v
er
s
it
y
o
f
B
ab
y
lo
n
,
B
ab
y
l
o
n
,
I
r
aq
E
m
ail:
h
ed
h
ab
s
a
@
g
m
ai
l.c
o
m
,
h
ed
h
ab
@
u
o
b
ab
y
lo
n
.
ed
u
.
iq
1.
I
NT
RO
D
UCT
I
O
N
Sin
ce
t
h
e
tu
r
n
o
f
th
e
ce
n
t
u
r
y
,
s
en
ti
m
e
n
t
an
al
y
s
is
d
ev
elo
p
ed
in
s
u
ch
a
w
a
y
t
h
at
it
h
a
s
b
ec
o
m
e
a
m
o
n
g
th
e
m
o
s
t
ac
ti
v
el
y
r
esear
ch
ed
ar
ea
s
o
f
n
at
u
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
i
n
g
(
N
L
P
)
,
p
ar
ticu
lar
l
y
d
u
e
to
th
e
co
i
n
cid
in
g
g
r
o
w
t
h
i
n
s
o
cial
m
ed
ia
co
m
m
u
n
ica
tio
n
s
u
c
h
as
r
ev
i
e
w
s
,
f
o
r
u
m
d
is
c
u
s
s
io
n
,
a
n
d
(
m
icr
o
)
b
l
o
g
g
i
n
g
s
er
v
ice
s
,
as
w
ell
a
s
th
e
e
n
o
r
m
o
u
s
a
m
o
u
n
t
o
f
o
p
in
io
n
ated
d
ata
av
ailab
le
i
n
d
ig
ital
f
o
r
m
.
Ot
h
er
f
ield
s
t
h
at
o
v
er
lap
w
it
h
th
is
d
is
cip
lin
e
ar
e
d
ata
m
i
n
i
n
g
,
w
e
b
m
i
n
in
g
,
te
x
t
m
i
n
i
n
g
,
a
n
d
i
n
f
o
r
m
atio
n
r
etr
iev
al
[
1
]
.
NL
P
ca
n
b
e
d
ef
i
n
ed
as
a
n
u
m
b
er
o
f
co
m
p
u
tatio
n
a
l
tec
h
n
iq
u
es
u
s
ed
to
a
n
al
y
ze
n
atu
r
al
lan
g
u
a
g
e
tex
t
s
s
o
as
to
e
n
ab
le
co
m
p
u
ter
s
o
f
u
n
d
er
s
ta
n
d
in
g
h
u
m
a
n
la
n
g
u
a
g
e,
an
d
th
e
s
e
n
ti
m
e
n
t
clas
s
i
f
ica
tio
n
o
f
o
n
l
in
e
r
ev
ie
w
d
ata
is
c
o
n
s
id
er
ed
to
b
e
an
ess
e
n
tial p
r
o
ce
s
s
an
d
o
f
h
ig
h
i
m
p
o
r
ta
n
ce
.
T
h
e
d
is
cip
lin
e
o
f
o
p
in
io
n
m
i
n
in
g
is
f
o
u
n
d
at
t
h
e
o
v
er
lap
b
et
w
ee
n
ea
c
h
o
f
i
n
f
o
r
m
atio
n
r
et
r
iev
al
a
n
d
co
m
p
u
tatio
n
al
li
n
g
u
is
tic
s
[
2
]
,
f
u
r
th
er
e
x
te
n
d
in
g
to
f
ie
ld
s
o
f
m
an
a
g
e
m
en
t
a
n
d
s
o
cial
s
cien
ce
s
,
in
cl
u
d
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
23
,
No
.
1
,
J
u
ly
2
0
2
1
:
345
-
3
5
3
346
f
i
n
an
ce
s
,
p
o
liti
cs,
co
m
m
u
n
icat
io
n
,
an
d
t
h
e
h
ea
lth
s
ec
to
r
,
as it
ten
d
s
to
b
e
i
m
p
o
r
ta
n
t
n
o
t o
n
l
y
to
b
u
s
in
e
s
s
b
u
t to
th
e
s
o
ciet
y
as
a
w
h
o
le.
I
t
s
r
ap
id
g
r
o
w
th
a
n
d
d
ev
elo
p
m
e
n
t
c
an
b
e
tr
ac
ed
b
ac
k
to
th
e
n
o
tio
n
t
h
at
o
p
in
io
n
s
ar
e
th
e
ce
n
tr
e
ar
o
u
n
d
w
h
ic
h
al
m
o
s
t
all
h
u
m
a
n
ac
ti
v
it
ies
r
e
v
o
lv
e
,
as
th
e
s
e
es
s
en
tiall
y
i
n
f
lu
e
n
ce
o
u
r
b
eh
av
io
r
.
T
h
e
w
a
y
i
n
w
h
ic
h
o
th
er
s
v
ie
w
t
h
e
w
o
r
ld
a
n
d
cr
ea
te
a
n
e
v
al
u
ati
o
n
o
f
it
ar
e,
to
a
ce
r
tai
n
e
x
ten
t
d
eter
m
i
n
ed
b
y
o
u
r
b
elief
s
a
n
d
p
er
s
p
ec
tiv
es
u
p
o
n
r
ea
lit
y
.
T
h
er
ef
o
r
e,
w
e
o
f
te
n
t
en
d
to
lo
o
k
f
o
r
o
th
er
's
o
p
in
io
n
s
b
ef
o
r
e
m
ak
i
n
g
a
ce
r
tain
d
ec
is
io
n
,
a
n
d
th
e
latter
is
tr
u
e
f
o
r
b
o
th
in
d
iv
id
u
als as
w
ell
a
s
o
r
g
an
iza
tio
n
s
[
3
]
.
A
co
n
s
id
er
ab
le
n
u
m
b
er
o
f
to
o
ls
h
a
v
e
b
ee
n
d
ev
elo
p
ed
f
o
r
o
p
in
io
n
m
in
i
n
g
to
b
e
u
s
ed
i
n
an
al
y
z
in
g
co
n
s
u
m
er
o
p
in
io
n
s
i
n
m
o
s
t
m
aj
o
r
b
u
s
in
es
s
en
v
ir
o
n
m
e
n
ts
,
s
u
ch
a
s
tr
a
v
elli
n
g
,
ac
co
m
m
o
d
a
tio
n
,
co
n
s
u
m
ab
les,
m
ater
ials
,
p
r
o
d
u
cts,
an
d
s
er
v
i
ce
s
[
4
]
.
Ho
w
ev
er
,
t
h
e
h
ea
lt
h
s
ec
to
r
h
as
w
it
n
ess
ed
r
elati
v
el
y
le
s
s
d
ev
elo
p
m
en
t
.
Dr
u
g
s
u
r
v
eilla
n
ce
,
f
o
r
ex
a
m
p
l
e,
is
co
n
s
id
er
ed
to
b
e
an
ess
en
tial
f
ac
to
r
in
ter
m
s
o
f
d
r
u
g
s
af
el
y
s
o
o
n
af
ter
t
h
e
r
elea
s
e
o
f
a
ce
r
tain
d
r
u
g
to
p
u
b
lic
u
s
e.
Si
n
ce
d
r
u
g
tr
ials
ar
e
u
s
u
al
l
y
co
n
d
u
c
ted
w
it
h
a
r
estricte
d
n
u
m
b
er
o
f
s
u
b
j
ec
ts
,
th
e
ch
a
n
ce
o
f
d
etec
ti
n
g
a
n
y
u
n
co
m
m
o
n
ad
v
er
s
e
ef
f
ec
t te
n
d
s
to
b
e
m
i
n
i
m
a
l [
2
]
.
A
t
p
r
ese
n
t
ti
m
e,
th
e
s
a
f
et
y
o
f
p
h
ar
m
ac
eu
t
ical
p
r
o
d
u
cts
is
d
eter
m
i
n
ed
b
y
ce
r
tai
n
clin
ical
tr
ials
an
d
s
p
ec
if
ied
te
s
ti
n
g
p
r
o
to
co
ls
[
5
]
.
T
h
ese
r
esear
ch
es
ar
e
o
f
te
n
c
o
n
d
u
cted
u
n
d
er
s
ta
n
d
ar
d
co
n
d
itio
n
s
,
r
e
s
tr
icted
b
y
ti
m
e
a
n
d
n
u
m
b
er
o
f
test
s
u
b
j
ec
ts
.
T
h
is
co
u
ld
p
o
ten
tiall
y
lead
to
ca
s
es
w
h
er
e
d
is
cr
ep
an
cies
i
n
s
elec
ti
n
g
p
atien
ts
,
as
w
ell
as
tr
ea
t
m
e
n
t
co
n
d
itio
n
,
co
u
ld
s
ig
n
i
f
ica
n
tl
y
af
f
ec
t
th
e
d
r
u
g
'
s
ef
f
icie
n
c
y
a
n
d
th
e
p
o
s
s
ib
le
r
is
k
s
o
f
ad
v
er
s
e
d
r
u
g
r
ea
ctio
n
s
(
A
D
R
s
)
[
6
]
.
Dee
p
lear
n
in
g
m
et
h
o
d
s
co
u
ld
b
e
d
escr
ib
ed
as
b
ein
g
m
et
h
o
d
s
o
f
s
ev
er
al
r
ep
r
ese
n
tatio
n
lev
els,
w
h
ic
h
ar
e
o
b
tain
ed
b
y
m
ea
n
s
o
f
t
h
e
co
m
p
o
s
i
tio
n
o
f
s
i
m
p
le
y
et
n
o
n
-
lin
ea
r
m
o
d
u
le
s
,
ea
ch
o
f
w
h
i
ch
h
a
s
t
h
e
ab
ilit
y
o
f
tr
an
s
f
o
r
m
i
n
g
a
r
ep
r
ese
n
tatio
n
f
r
o
m
a
p
ar
ticu
lar
le
v
el
(
s
tar
t
in
g
w
i
th
t
h
e
r
a
w
in
p
u
t)
to
a
h
ig
h
er
r
ep
r
esen
tatio
n
at
a
lev
el
o
f
a
litt
le
m
o
r
e
ab
s
tr
ac
tn
es
s
.
C
o
m
p
o
s
i
n
g
a
s
u
f
f
ici
en
t
n
u
m
b
er
o
f
s
u
c
h
tr
an
s
f
o
r
m
atio
n
s
en
ab
le
s
th
e
lear
n
in
g
o
f
f
u
n
ctio
n
s
t
h
at
ar
e
o
f
co
n
s
id
er
ab
le
co
m
p
le
x
it
y
[
7
]
.
Fo
r
th
e
an
al
y
s
is
o
f
p
atie
n
t
'
s
o
p
i
n
io
n
o
n
th
e
ef
f
ec
tiv
e
n
e
s
s
o
f
d
r
u
g
s
,
th
is
s
tu
d
y
p
r
o
p
o
s
es
th
e
in
te
g
r
atio
n
o
f
G
lo
Ve,
an
ef
f
ic
ien
t
N
L
P
tr
an
s
f
er
lear
n
i
n
g
m
o
d
el,
w
it
h
B
iLST
M,
b
ein
g
a
ty
p
ical
s
tate
-
of
-
t
h
e
-
ar
t
b
i
-
d
i
r
ec
tio
n
al
m
o
d
el.
T
h
e
v
ec
to
r
r
ep
r
esen
tatio
n
o
f
a
w
o
r
d
is
o
f
g
r
ea
t
u
s
e
i
n
tex
t
cla
s
s
i
f
y
in
g
,
clu
s
ter
i
n
g
,
an
d
r
etr
ie
v
in
g
i
n
f
o
r
m
atio
n
.
T
h
er
e
ar
e
a
n
u
m
b
er
o
f
b
en
e
f
it
s
o
b
s
er
v
ed
w
h
en
co
m
p
ar
in
g
w
o
r
d
em
b
ed
d
in
g
m
et
h
o
d
s
w
it
h
b
ag
-
of
-
w
o
r
d
s
r
ep
r
ese
n
tatio
n
s
.
T
h
e
B
iL
ST
M
tak
es
in
to
co
n
s
id
er
atio
n
ef
f
icie
n
t
a
m
o
u
n
ts
o
f
co
n
te
x
t
s
b
o
th
b
ef
o
r
e
an
d
af
ter
a
w
o
r
d
,
th
er
eb
y
eli
m
i
n
ati
n
g
t
h
e
is
s
u
es
o
f
co
n
tex
t li
m
itatio
n
t
h
at
i
s
f
o
u
n
d
i
n
f
ee
d
-
f
o
r
w
ar
d
m
o
d
els.
2.
RE
S
E
ARCH
M
E
T
H
O
D
2
.
1
.
G
lo
ba
l v
ec
t
o
rs f
o
r
w
o
r
d r
epre
s
ent
a
t
io
n
I
n
th
i
s
w
o
r
k
,
w
e
u
s
ed
t
h
e
p
r
e
-
tr
ain
ed
e
m
b
ed
d
in
g
o
f
Glo
Ve
[
8
]
,
[
9
]
w
i
th
a
3
0
0
-
d
i
m
e
n
s
io
n
a
l,
w
h
ich
i
s
p
air
ed
w
it
h
a
n
n
-
g
r
a
m
e
m
b
ed
d
in
g
.
Fo
r
th
e
e
m
b
ed
d
in
g
o
f
w
o
r
d
s
w
i
th
i
n
a
v
ec
to
r
s
p
ac
e,
th
e
G
lo
Ve
m
o
d
el
is
u
s
ed
i
n
t
h
e
ex
a
m
i
n
atio
n
o
f
t
h
e
Xij
p
r
esen
ce
m
atr
i
x
i
n
a
lar
g
e
te
x
t
b
lo
ck
r
ep
r
esen
ti
n
g
t
h
e
u
s
er
'
s
r
e
v
ie
w
s
o
n
d
r
u
g
s
.
N
u
m
er
ic
v
ec
to
r
s
t
h
at
r
ep
r
esen
t
u
s
er
r
ev
ie
w
s
ar
e
o
b
tain
ed
.
Glo
Ve
ca
n
b
e
d
escr
ib
ed
as
an
u
n
s
u
p
er
v
is
ed
lear
n
in
g
al
g
o
r
ith
m
u
s
ed
to
o
b
tain
v
ec
to
r
r
ep
r
esen
tat
io
n
s
f
o
r
w
o
r
d
s
[
1
0
]
.
T
h
e
tr
ain
in
g
o
f
th
e
m
ai
n
id
ea
is
s
tated
as f
o
llo
w
s
:
(
)
(
1
)
w
h
er
e
r
ep
r
esen
t
th
e
tr
ain
ed
v
ec
to
r
s
,
an
d
in
d
icate
th
e
s
ca
lar
b
ias
ter
m
s
r
elate
d
to
th
e
w
o
r
d
s
i a
n
d
j
,
r
esp
ec
tiv
el
y
.
T
h
e
ess
en
t
ial
asp
ec
ts
o
f
tr
ain
i
n
g
p
r
o
ce
s
s
es in
Glo
Ve
i
n
cl
u
d
e:
a)
A
w
ei
g
h
t
in
g
f
u
n
ct
io
n
f
to
eli
m
i
n
ate
v
er
y
co
m
m
o
n
l
y
o
cc
u
r
r
in
g
w
o
r
d
s
(
li
k
e
s
to
p
w
o
r
d
s
)
,
a
s
t
h
ese
ten
d
to
ad
d
n
o
is
e
an
d
ar
e
n
o
t o
v
er
w
ei
g
h
ted
;
b)
R
ar
e
w
o
r
d
s
ar
e
n
o
t o
v
er
w
ei
g
h
ted
;
c)
T
h
e
co
-
o
cc
u
r
r
en
ce
s
tr
e
n
g
t
h
,
wh
en
e
v
er
m
o
d
eled
as
a
d
i
s
ta
n
c
e,
n
ee
d
s
to
u
n
d
er
g
o
s
m
o
o
th
i
n
g
b
y
m
ea
n
s
o
f
a
lo
g
f
u
n
ct
io
n
.
T
h
e
f
in
al
lo
s
s
f
u
n
ctio
n
f
o
r
G
lo
Ve
[
1
1
]
m
o
d
el
w
ill
b
e
as f
o
llo
w
s
:
∑
(
)
(
2
)
w
h
er
e
V
is
a
co
m
p
lete
v
o
ca
b
u
lar
y
,
an
d
an
d
f
(
x
)
=1
o
th
er
w
is
e.
2
.
2
.
L
o
ng
-
s
ho
rt
-
t
er
m
-
m
e
mo
ry
deep
neura
l net
w
o
rk
(
L
ST
M
-
DNN)
L
ST
M
is
a
d
is
ti
n
ct
ca
s
e
o
f
R
NNs.
I
t
is
d
esi
g
n
ed
to
co
p
e
w
ith
t
h
e
p
r
o
b
lem
o
f
g
r
ad
ien
ts
t
h
at
ex
p
lo
d
e
o
r
v
an
is
h
[
1
2
]
.
T
h
e
b
asic
id
e
a
b
eh
in
d
t
h
e
L
ST
M
is
th
e
e
x
i
s
ten
ce
o
f
a
m
e
m
o
r
y
ce
l
l
an
d
a
n
u
m
b
er
o
f
g
ates.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4752
S
en
timen
t a
n
a
lysi
s
fo
r
d
r
u
g
s
r
ev
iew
s
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ha
d
a
b
K
h
a
lid
Ob
a
ye
s
)
347
T
h
es
e
m
e
m
o
r
y
ce
ll
s
an
d
g
ates
ar
e
ad
d
ed
to
ea
ch
n
eu
r
o
n
in
th
e
n
et
w
o
r
k
[
1
3
]
.
T
h
e
p
r
i
n
cip
l
e
o
f
L
ST
M
w
o
r
k
i
s
tr
an
s
m
itti
n
g
th
e
i
m
p
o
r
tan
t
i
n
f
o
r
m
at
io
n
i
n
a
r
eliab
le
w
a
y
o
v
er
a
n
u
m
b
er
o
f
ti
m
e
s
tep
s
o
n
t
o
th
e
n
ex
t
ti
m
e
s
tep
.
L
ST
M
ce
ll
w
h
er
eb
y
t
h
e
g
a
tes
in
th
e
L
ST
M
m
o
d
el
ar
e
u
s
e
d
w
it
h
ad
d
itio
n
al
p
ar
a
m
eter
s
ca
n
b
e
ap
p
lied
to
a
m
e
m
o
r
y
cir
c
u
it
u
s
ed
t
o
m
e
m
o
r
ize
o
r
s
to
r
e
th
e
in
f
o
r
m
a
tio
n
f
o
r
a
lo
n
g
ter
m
f
r
o
m
t
h
e
r
ec
u
r
r
en
t
la
y
er
[
1
4
]
.
T
h
e
in
p
u
t o
f
R
NN
m
o
d
el
is
a
s
eq
u
en
ce
{x
1
,
x
2
, ..., x
n
}
th
at
u
s
es t
h
e
f
o
llo
w
i
n
g
r
ec
u
r
r
en
ce
.
(
3
)
w
h
er
e:
T
h
e
g
ates
ar
r
iv
e
to
th
e
r
ec
u
r
r
en
t
f
u
n
ctio
n
to
co
p
e
w
ith
v
an
is
h
in
g
o
r
ex
p
lo
d
in
g
p
r
o
b
lem
s
.
T
h
e
L
ST
M
ce
lls
ar
e
ap
p
lied
as f
o
llo
w
s
:
[
]
(
4
)
(
[
]
)
(
5
)
[
]
(
6
)
͂
]
(
7
)
͂
(
8
)
(
9
)
w
h
er
e:
i
t
is
th
e
i
n
p
u
t
g
ate.
f
t
is
t
h
e
f
o
r
g
e
te
g
ate.
o
t
_
t
is
th
e
o
u
tp
u
t
g
ate
W
h
i
l
e
W
an
d
b
r
e
p
r
e
s
e
n
t
t
h
e
p
a
r
am
et
e
r
s
o
f
L
S
T
M
,
a
n
d
͂
t
h
e
p
r
e
s
e
n
t
c
e
l
l
s
t
a
te
a
n
d
th
e
n
e
w
c
a
n
d
i
d
at
e
v
alu
e
s
f
o
r
c
e
l
l
s
t
a
te
r
e
s
p
e
c
t
iv
ely
.
T
h
e
s
i
g
m
o
i
d
f
u
n
c
t
i
o
n
is
u
s
e
d
t
h
r
ee
t
im
es
i
n
(
)
,
r
e
s
p
e
c
t
iv
ely
.
T
h
e
o
u
t
p
u
t
o
f
th
e
s
e
g
a
t
es
is
b
e
tw
e
en
0
an
d
1
a
s
in
(
4
)
t
o
(
6
)
.
T
h
e
d
e
c
i
s
i
o
n
o
f
t
h
e
th
r
e
e
g
a
t
es
d
e
p
e
n
d
s
o
n
th
e
c
u
r
r
en
t in
p
u
t
an
d
t
h
e
p
r
e
v
i
o
u
s
o
u
t
p
u
t
.
T
h
e
f
o
r
g
e
t
g
at
e
i
s
r
e
s
p
o
n
s
i
b
l
e
f
o
r
al
lo
w
i
n
g
th
e
am
o
u
n
t
o
f
p
r
e
v
i
o
u
s
s
t
a
t
e
t
o
p
a
s
s
.
T
h
e
i
n
p
u
t
g
at
e
d
e
c
i
d
e
s
w
h
et
h
e
r
th
e
r
e
c
en
t
d
at
a
o
f
t
h
e
i
n
p
u
t
d
em
an
d
s
an
u
p
d
a
t
e
o
r
a
p
p
en
d
t
o
th
e
c
e
ll
s
t
at
e
[
1
5
]
.
T
h
e
o
u
t
p
u
t
g
a
t
e
d
e
t
e
r
m
i
n
es
th
e
d
a
ta
t
o
b
e
o
u
t
p
u
t
d
e
p
e
n
d
in
g
o
n
t
h
e
c
e
l
l
s
t
a
te
.
T
h
e
s
e
tw
o
g
at
e
s
f
u
n
c
ti
o
n
s
i
d
e
b
y
s
i
d
e
t
o
le
a
r
n
an
d
s
to
r
e
i
n
f
o
r
m
a
ti
o
n
r
e
l
a
t
e
d
t
o
th
e
l
o
n
g
an
d
s
h
o
r
t
-
t
e
r
m
s
e
r
i
es
[
1
3
]
.
T
h
e
c
e
l
l
s
t
at
e
is
th
e
c
o
r
e
o
f
th
e
L
S
T
M
w
o
r
k
,
as
i
t
is
r
eg
a
r
d
e
d
t
o
b
e
a
ty
p
e
o
f
c
o
n
v
ey
o
r
b
e
l
t
[
1
6
]
.
2
.3
.
B
idi
re
ct
io
na
l LS
T
M
s
B
id
ir
ec
tio
n
al
L
ST
Ms
ex
ten
d
f
r
o
m
tr
ad
itio
n
al
L
ST
Ms
[
1
7
]
,
an
d
th
e
y
co
u
ld
e
n
h
a
n
ce
th
e
m
o
d
el
p
er
f
o
r
m
a
n
ce
i
n
ter
m
s
o
f
s
eq
u
en
ce
clas
s
if
icatio
n
is
s
u
e
s
.
I
n
s
i
tu
at
io
n
s
w
h
er
ei
n
al
l
ti
m
e
s
tep
s
o
f
t
h
e
i
n
p
u
t
s
eq
u
en
ce
h
a
v
e
b
ee
n
p
r
o
v
id
ed
,
th
e
B
id
ir
ec
tio
n
al
L
ST
Ms
tr
a
in
s
t
w
o
L
ST
Ms
[
1
8
]
r
ath
er
th
an
j
u
s
t
o
n
e
o
n
t
h
e
in
p
u
t
s
eq
u
e
n
ce
:
th
e
f
ir
s
t
o
n
e
o
n
th
e
i
n
p
u
t
s
eq
u
e
n
ce
w
i
ll
b
e
as
-
is
,
w
h
er
ea
s
th
e
o
th
er
i
s
o
n
a
r
ev
er
s
ed
co
p
y
o
f
th
e
i
n
p
u
t
s
eq
u
e
n
ce
.
Gi
v
en
th
at
th
e
s
e
p
r
esen
t
a
n
e
x
tr
a
co
n
t
ex
t
to
t
h
e
n
et
w
o
r
k
,
t
h
e
y
w
ill
lead
to
a
r
elativ
el
y
f
aster
a
n
d
m
o
r
e
co
m
p
r
eh
en
s
i
v
e
lear
n
in
g
o
n
t
h
e
p
r
o
b
lem
i
n
c
ase
[
1
9
]
.
Bi
-
L
ST
Ms
h
a
v
e
b
ee
n
f
o
u
n
d
to
b
e
p
ar
ticu
lar
o
f
u
s
e
i
n
ca
s
es
th
at
r
eq
u
ir
e
a
co
n
tex
t
f
o
r
th
e
i
n
p
u
t
,
in
cl
u
d
in
g
s
en
ti
m
en
t
cla
s
s
i
f
ica
tio
n
.
I
n
u
n
id
ir
ec
tio
n
al
L
ST
M,
th
e
in
f
o
r
m
atio
n
s
ee
m
s
to
f
lo
w
b
ac
k
w
ar
d
s
,
th
u
s
f
r
o
m
b
ac
k
to
f
r
o
n
t.
I
n
co
n
tr
ast
,
in
B
i
-
d
ir
ec
tio
n
al
L
ST
M
th
e
i
n
f
o
r
m
atio
n
f
lo
w
s
b
o
th
b
ac
k
war
d
s
an
d
f
o
r
w
ar
d
s
,
b
y
m
ea
n
s
o
f
t
w
o
h
id
d
en
s
tate
s
,
h
en
ce
t
h
e
B
i
-
L
ST
M'
s
w
id
e
r
co
m
p
r
eh
e
n
s
io
n
o
n
th
e
co
n
te
x
t
[
2
0
]
.
B
i
-
L
ST
Ms
h
av
e
b
ee
n
ap
p
lied
i
n
e
s
ca
lat
in
g
c
h
u
n
k
s
o
f
i
n
p
u
t
i
n
f
o
r
m
a
tio
n
t
h
at
ar
e
o
f
u
s
e
to
t
h
e
n
et
w
o
r
k
.
T
h
er
e
ar
e
s
tr
u
ct
u
r
es
o
f
R
NN
w
it
h
L
ST
M
an
d
R
NN
w
it
h
B
i
-
L
ST
M
[
2
1
]
.
E
s
s
en
tiall
y
,
t
h
e
B
R
NN
u
n
d
er
g
o
es
a
p
r
o
ce
d
u
r
e
w
h
er
eb
y
t
h
e
n
eu
r
o
n
s
o
f
a
n
o
r
m
al
R
NN
ar
e
b
r
o
k
en
to
b
id
ir
ec
tio
n
al
w
a
y
s
.
T
h
e
f
ir
s
t
i
s
f
o
r
th
e
b
ac
k
w
ar
d
s
ta
tes
(
n
eg
ati
v
e
ti
m
e
d
ir
ec
tio
n
)
,
w
h
e
r
ea
s
th
e
s
ec
o
n
d
is
f
o
r
th
e
f
o
r
w
ar
d
s
tate
s
(
p
o
s
itiv
e
t
i
m
e
d
ir
ec
tio
n
)
.
T
h
er
e
is
n
o
co
n
n
ec
tio
n
b
et
w
ee
n
t
h
e
in
p
u
t
o
f
th
e
r
ev
er
s
e
d
ir
ec
tio
n
s
tates
an
d
th
e
t
w
o
f
o
r
m
er
s
tates
’
r
es
u
lts
.
T
h
e
B
iL
ST
M
is
s
tr
u
c
t
u
r
ed
as i
llu
s
tr
ated
i
n
F
ig
u
r
e
1.
A
s
f
o
r
t
h
e
d
ir
ec
tio
n
s
,
th
e
i
n
p
u
t
d
ata
r
eg
ar
d
i
n
g
th
e
p
ast a
n
d
f
u
t
u
r
e
o
f
th
e
cu
r
r
en
t ti
m
e
f
r
a
m
e
co
u
ld
b
e
ap
p
lied
,
m
ea
n
w
h
i
le
th
e
s
tan
d
ar
d
R
NN
n
ee
d
s
t
h
e
d
ela
y
s
to
i
n
c
lu
d
e
f
u
t
u
r
e
d
ata.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
23
,
No
.
1
,
J
u
ly
2
0
2
1
:
345
-
3
5
3
348
Fig
u
r
e
1
.
B
iL
ST
M
3.
RE
L
E
A
T
E
D
WO
RK
Qu
et.
a
l.
in
tr
o
d
u
ce
d
a
b
ag
-
of
-
o
p
in
io
n
s
r
ev
ie
w
te
x
t
r
ep
r
esen
t
atio
n
m
o
d
el
t
h
at
d
i
f
f
er
s
f
r
o
m
alter
n
ati
v
e
m
o
d
el
s
o
f
t
h
i
s
t
y
p
e,
t
h
r
o
u
g
h
w
h
ic
h
t
h
e
s
e
n
ti
m
en
t
p
h
r
ase
s
h
av
e
b
ee
n
o
b
tain
ed
.
V
ie
w
p
o
i
n
t
s
ar
e
co
m
p
o
s
ed
o
f
a
tr
ip
le
asp
ec
t,
n
a
m
el
y
a
s
e
n
ti
m
en
t
w
o
r
d
,
a
m
o
d
i
f
ier
,
an
d
a
n
eg
ati
v
e
w
o
r
d
.
T
o
ex
e
m
p
li
f
y
,
in
th
e
p
h
r
ase
“n
o
t
q
u
ite
w
e
ll”,
“w
e
ll”
is
a
s
en
t
i
m
en
t
w
o
r
d
,
w
h
er
ea
s
“
q
u
ite”
i
s
a
m
o
d
i
f
ier
,
an
d
“n
o
t”
is
n
e
g
ati
v
e
w
o
r
d
.
As
f
o
r
s
en
ti
m
e
n
tal
cla
s
s
i
f
ica
tio
n
s
wh
ich
o
n
l
y
id
en
tify
t
h
e
te
n
d
e
n
c
y
o
f
a
p
h
r
ase,
b
ein
g
p
o
s
it
iv
el
y
o
r
n
e
g
ati
v
el
y
,
m
o
d
i
f
ier
s
ar
e
n
o
t
v
er
y
s
i
g
n
i
f
i
ca
n
t,
m
ea
n
w
h
i
le
f
o
r
R
R
P
,
ea
ch
o
f
th
e
m
o
d
i
f
ier
a
n
d
th
e
n
e
g
ati
v
e
w
o
r
d
ar
e
o
f
i
m
p
o
r
tan
ce
.
W
ith
a
lab
eled
d
o
m
ai
n
-
in
d
ep
en
d
e
n
t
co
r
p
u
s
(
f
r
o
m
m
u
ltip
le
d
o
m
ain
s
)
,
a
co
n
s
tr
ain
ed
r
id
g
e
r
eg
r
ess
io
n
m
o
d
el
w
ill
b
e
f
ir
s
t
ap
p
lied
f
o
r
lear
n
in
g
e
v
er
y
v
i
e
w
p
o
in
t
f
ea
t
u
r
e
(
s
e
n
ti
m
en
t
in
ten
s
i
t
y
a
n
d
s
co
r
e)
,
a
f
t
e
r
w
h
i
c
h
t
h
e
t
r
a
d
i
t
i
o
n
a
l
u
n
i
g
r
a
m
s
f
e
a
t
u
r
e
i
s
e
x
t
e
n
d
e
d
b
y
m
e
a
n
s
o
f
v
i
e
w
p
o
i
n
t
f
e
a
t
u
r
e
s
f
o
r
a
c
h
i
e
v
i
n
g
R
R
P
[
2
2
]
.
T
h
er
e
ar
e
a
n
u
m
b
er
o
f
r
ese
ar
ch
es
t
h
at
m
a
k
e
u
s
e
o
f
o
p
in
io
n
m
i
n
i
n
g
w
i
th
i
n
t
h
e
m
ed
ical
f
ield
,
p
ar
ticu
lar
l
y
b
y
m
ea
n
s
o
f
m
ac
h
in
e
lear
n
in
g
tec
h
n
iq
u
es,
g
r
e
av
es,
r
a
m
ir
ez
-
ca
n
o
,
m
i
llett,
d
ar
zi,
an
d
d
o
n
ald
s
o
n
(
2
0
1
3
)
co
n
d
u
cted
an
an
al
y
s
is
o
f
p
atien
t
’
s
o
p
in
io
n
s
o
n
v
ar
io
u
s
f
ac
to
r
s
o
f
h
o
s
p
ital
p
er
f
o
r
m
an
ce
s
i
n
th
e
U
n
ited
Kin
g
d
o
m
.
T
h
e
au
th
o
r
s
ap
p
lied
f
o
u
r
d
if
f
er
e
n
t
m
et
h
o
d
s
,
to
s
ee
w
h
ic
h
g
a
v
e
th
e
q
u
ic
k
e
s
t
an
d
m
o
s
t
ac
cu
r
at
e
r
esu
lt
s
:
n
aï
v
e
b
a
y
e
s
m
u
lti
n
o
m
ials
(
NB
M)
d
ec
is
io
n
tr
ee
s
,
b
ag
g
i
n
g
,
a
n
d
s
u
p
p
o
r
t
v
ec
to
r
m
a
ch
in
e
s
.
T
h
er
e
w
a
s
8
1
%,
8
4
%,
an
d
8
9
%
ag
r
ee
m
en
t
b
et
w
ee
n
q
u
a
n
titat
iv
e
r
atin
g
s
o
f
ca
r
e
an
d
th
o
s
e
d
e
r
iv
ed
f
r
o
m
f
r
ee
-
te
x
t
co
m
m
e
n
t
s
u
s
in
g
s
e
n
ti
m
e
n
t
an
al
y
s
i
s
f
o
r
clea
n
lin
e
s
s
,
b
ein
g
tr
ea
ted
w
it
h
d
ig
n
it
y
,
an
d
o
v
er
al
l
r
ec
o
m
m
e
n
d
atio
n
o
f
h
o
s
p
ital r
esp
ec
ti
v
el
y
(
k
ap
p
a
s
co
r
es: .
4
0
–
.
7
4
,
P
<.
0
0
1
f
o
r
a
ll)
[
2
3
]
.
Nik
f
ar
j
am
et
a
l.
u
s
ed
ass
o
ciati
o
n
r
u
le
m
i
n
i
n
g
s
o
as
to
id
en
tify
p
atter
n
s
,
i.e
.
co
m
b
in
a
tio
n
o
f
ter
m
s
,
o
r
co
n
d
itio
n
al
r
an
d
o
m
f
ie
ld
s
(
C
R
Fs
)
,
f
o
r
th
e
e
x
tr
ac
tio
n
o
f
m
en
t
io
n
s
o
f
ADR
s
.
Si
n
ce
it
i
s
i
m
p
l
icitl
y
a
s
s
u
m
ed
th
at
a
p
atien
t
'
s
p
o
s
t
o
n
A
D
R
s
t
y
p
icall
y
e
x
p
r
ess
es
n
e
g
ati
v
el
y
te
n
d
in
g
s
e
n
ti
m
e
n
ts
,
e
x
a
m
in
ed
t
h
e
i
m
p
ac
t
o
f
u
s
in
g
s
en
ti
m
e
n
t a
n
al
y
s
is
f
ea
t
u
r
es to
en
r
ich
a
le
x
ico
n
-
b
ased
A
DR
i
d
en
tify
i
n
g
m
et
h
o
d
[
2
4
]
:
a)
Go
p
alak
r
is
h
n
an
et
a
l.
an
aly
ze
d
th
e
ex
ten
t
to
w
h
ich
p
atien
ts
w
er
e
s
atis
f
ied
w
ith
a
p
ar
ticu
lar
d
r
u
g
,
th
r
o
u
g
h
th
e
u
s
e
o
f
a
s
u
p
er
v
is
ed
lear
n
in
g
s
en
tim
en
t
an
aly
s
is
ap
p
r
o
ac
h
.
T
h
er
e
ar
e
th
r
ee
lev
els
o
f
p
o
lar
ity
class
if
i
ed
in
th
is
s
tu
d
y
,
d
r
aw
in
g
a
co
m
p
ar
is
o
n
b
etw
ee
n
SVM
an
d
n
eu
r
al
n
etw
o
r
k
b
ased
m
eth
o
d
s
[
2
]
.
b)
A.
Nav
in
d
g
i
an
d
et
a
l
.
(
2
0
1
6
)
m
ad
e
u
s
e
o
f
a
s
u
b
j
ec
tiv
ity
lex
ico
n
as
w
ell
as
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
to
an
aly
ze
th
e
s
en
tim
e
n
t
in
p
o
s
ts
o
n
f
o
r
u
m
s
th
at
d
is
c
u
s
s
th
e
to
p
ic
o
f
h
ea
r
in
g
lo
s
s
.
W
e
r
ep
o
r
t
ex
p
er
im
en
ts
o
n
a
s
en
tim
en
t
-
lab
eled
co
r
p
u
s
o
f
p
o
s
ts
tak
en
f
r
o
m
a
m
ed
ical
s
u
p
p
o
r
t
f
o
r
u
m
.
th
ey
ar
g
u
e
th
at
n
o
t
o
n
ly
is
a
m
o
r
e
f
in
e
-
g
r
ain
ed
ap
p
r
o
ac
h
to
tex
t
an
aly
s
is
im
p
o
r
tan
t
b
u
t
s
im
u
ltan
eo
u
s
ly
r
ec
o
g
n
izin
g
th
e
s
o
cial
f
u
n
ctio
n
b
eh
in
d
af
f
ec
tiv
e
ex
p
r
ess
io
n
s
en
ab
le
a
m
o
r
e
ac
cu
r
ate
an
d
v
alu
ab
le
lev
el
o
f
u
n
d
er
s
tan
d
in
g
[
2
5
]
.
4.
T
H
E
P
RO
P
O
SE
D
SYS
T
E
M
T
h
e
cu
r
r
en
t
s
ec
tio
n
p
r
esen
ts
a
d
etailed
ex
p
lan
atio
n
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
tem
.
Fig
u
r
e
2
s
h
o
w
s
th
e
s
tr
u
ctu
r
e
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
tem
in
th
is
w
o
r
k
.
T
h
e
d
ataset
u
s
ed
in
th
is
w
o
r
k
is
o
b
tain
ed
f
r
o
m
th
e
Kag
g
le
w
eb
s
ite
f
o
r
r
etr
iev
in
g
u
s
er
r
ev
iew
s
an
d
r
atin
g
s
o
n
d
r
u
g
ex
p
er
ien
ce
,
n
am
ely
th
e
UC
I
ML
Dr
u
g
R
ev
iew
d
ataset.
I
t
in
v
o
lv
es
u
s
er
r
ev
iew
s
o
n
s
p
ec
if
ic
d
r
u
g
s
,
as
w
ell
as
an
y
r
elate
d
co
n
d
itio
n
s
an
d
a
1
0
-
s
tar
u
s
er
r
atin
g
th
at
r
ef
lects
a
co
m
p
r
eh
en
s
iv
e
v
iew
o
n
th
e
u
s
er
'
s
s
atis
f
ac
tio
n
.
T
h
e
em
o
tio
n
d
ataset
co
n
s
is
ts
o
f
2
1
5
,
0
6
3
in
s
tan
ce
s
,
an
d
th
er
e
ar
e
6
attr
ib
u
tes
in
th
e
d
ataset.
T
h
e
to
tal
n
u
m
b
er
o
f
in
d
iv
id
u
al
d
r
u
g
s
w
ith
in
th
e
d
ataset
is
ab
o
u
t 6
3
4
5
d
r
u
g
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4752
S
en
timen
t a
n
a
lysi
s
fo
r
d
r
u
g
s
r
ev
iew
s
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ha
d
a
b
K
h
a
lid
Ob
a
ye
s
)
349
T
h
e
d
ataset
attr
ib
u
tes
ar
e:
a)
Dr
u
g
n
am
e
(
ca
teg
o
r
ical)
: n
am
e
o
f
d
r
u
g
b)
C
o
n
d
itio
n
(
ca
teg
o
r
ical)
: n
am
e
o
f
co
n
d
itio
n
c)
R
ev
iew
(
tex
t)
:
p
atien
t r
ev
iew
d)
R
atin
g
(
n
u
m
er
ical)
:
1
0
s
tar
p
atien
t r
atin
g
e)
Date
(
d
ate)
: d
ate
o
f
r
ev
iew
en
tr
y
f)
Usef
u
l
co
u
n
t
(
n
u
m
er
ical)
:
n
u
m
b
er
o
f
u
s
er
s
w
h
o
f
o
u
n
d
r
ev
iew
u
s
ef
u
l
T
h
e
d
ataset
w
as
ex
p
lo
r
ed
to
f
in
d
an
y
im
p
o
r
tan
t
s
tatis
tics
th
at
h
elp
in
u
n
d
er
s
tan
d
in
g
an
d
an
aly
zin
g
th
e
co
n
ten
t
o
f
th
e
d
ata
s
et.
T
h
e
d
ataset
co
n
tain
s
2
1
5
,
0
6
3
in
s
tan
ce
s
o
f
r
ev
iew
er
s
an
d
6
3
4
5
d
r
u
g
s
.
Su
p
er
v
is
ed
m
ac
h
in
e
lear
n
in
g
ap
p
r
o
ac
h
es
r
eq
u
ir
e
lab
eled
d
ata
to
class
if
y
o
p
in
io
n
s
in
r
ev
iew
s
.
T
h
e
tex
t
r
ev
iew
s
(
p
atien
t
r
ev
iew
)
ar
e
ass
ig
n
ed
a
lab
el
ac
co
r
d
in
g
to
th
e
r
atin
g
(
1
0
s
tar
p
atien
t
r
atin
g
)
.
T
h
e
tex
t
r
ev
iew
co
n
s
is
ts
o
f
a
w
r
itten
tex
t
th
at
ex
p
lain
s
th
e
r
atin
g
s
co
r
e
o
f
th
e
r
ev
iew
.
A
s
th
e
r
ev
iew
s
ar
e
co
n
d
u
cted
b
y
d
if
f
er
en
t
p
er
s
o
n
s
w
ith
v
ar
y
in
g
b
ac
k
g
r
o
u
n
d
s
,
th
e
in
ten
d
ed
m
ea
n
in
g
o
f
a
1
0
s
tar
r
ev
iew
w
ill
d
if
f
er
f
r
o
m
o
n
e
u
s
er
to
an
o
th
er
.
T
h
e
p
atien
t
r
ev
iew
is
th
en
ass
ig
n
ed
an
o
p
in
io
n
class
ac
co
r
d
in
g
to
th
e
r
atin
g
in
th
e
d
ata
f
o
r
m
at.
T
h
e
ass
ig
n
e
d
r
ev
iew
class
is
p
o
s
itiv
e
w
h
en
ev
er
th
e
r
atin
g
is
>=
5
,
o
th
er
w
is
e,
if
th
e
r
atin
g
is
<5
th
en
th
e
class
lab
el
w
ill
b
e
n
eg
ativ
e.
Fig
u
r
e
3
s
h
o
w
s
th
e
co
u
n
t
o
f
th
e
r
atin
g
v
alu
e.
Fig
u
r
e
2
.
P
r
o
p
o
s
es s
y
s
te
m
Fig
u
r
e
3
.
T
h
e
co
u
n
t o
f
r
atin
g
v
alu
e
i
n
t
h
e
d
ataset
a)
T
h
e
tw
o
r
o
w
s
o
f
th
e
d
ataset
d
ea
lt w
ith
ar
e
th
e
u
s
er
r
ev
iew
s
an
d
r
atin
g
.
b)
T
o
g
et
th
e
w
o
r
d
s
v
ec
to
r
em
b
ed
d
in
g
,
w
e
em
p
lo
y
ed
th
e
Glo
Ve
m
o
d
el,
w
ith
th
e
u
s
er
r
ev
iew
s
as
its
en
tr
y
,
an
d
af
ter
its
ap
p
licatio
n
th
is
m
o
d
el
th
e
w
o
r
d
s
v
ec
to
r
em
b
ed
d
in
g
is
o
b
tain
ed
.
T
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
23
,
No
.
1
,
J
u
ly
2
0
2
1
:
345
-
3
5
3
350
p
r
e
-
tr
ain
ed
em
b
ed
d
in
g
o
f
GL
o
Ve
w
o
r
d
is
u
s
ed
,
w
h
ich
is
p
air
ed
w
ith
n
-
g
r
am
em
b
ed
d
in
g
ch
ar
ac
ter
is
tics
,
an
d
is
r
etain
ed
th
r
o
u
g
h
o
u
t
th
e
tr
ain
in
g
.
c)
T
o
p
er
f
o
r
m
th
e
p
r
o
ce
s
s
o
f
class
if
y
in
g
u
s
er
r
ev
iew
er
s
in
to
tw
o
class
es,
w
h
ich
ar
e
p
o
s
itiv
e
r
ev
iew
er
s
an
d
n
eg
ativ
e,
th
e
B
iL
ST
M
Dee
p
Netw
o
r
k
is
em
p
lo
y
ed
.
T
h
e
em
o
tio
n
d
ataset
co
n
s
is
ts
o
f
2
1
5
,
0
6
3
in
s
tan
ce
s
,
s
p
lit
to
1
6
1
,
2
9
7
as
tr
ain
in
g
d
ata
an
d
5
3
,
7
6
6
as
test
in
g
d
ata.
T
h
e
r
atin
g
p
er
r
ev
iew
lab
eled
as in
teg
er
n
u
m
b
er
s
b
etw
ee
n
1
an
d
1
0
.
d)
T
h
e
u
n
its
o
f
th
e
B
I
-
L
ST
Ms
ar
e
d
escr
ib
ed
in
1
2
8
h
id
d
en
d
im
en
s
io
n
s
,
an
d
th
e
in
itial
lear
n
in
g
r
ate
is
tr
ain
ed
u
s
in
g
an
A
DA
M
o
p
tim
izer
.
Fo
r
th
e
em
b
ed
d
in
g
lay
er
,
w
e
d
ef
in
ed
th
e
d
r
o
p
-
o
u
t
r
ate.
T
h
e
b
atch
tr
ain
in
g
w
as
p
er
f
o
r
m
ed
w
ith
a
b
atch
s
ize
o
f
1
2
8
.
e)
R
NNs
w
ill
b
e
u
s
ed
d
u
e
o
f
s
ec
r
et
s
tates,
a
s
th
ey
r
ec
all
p
r
io
r
k
n
o
w
led
g
e
an
d
lin
k
it
to
th
e
cu
r
r
en
t
m
is
s
io
n
.
L
o
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
n
etw
o
r
k
s
(
L
ST
M)
ar
e
an
R
NN
s
u
b
s
et,
s
p
ec
ialized
in
co
llectin
g
in
f
o
r
m
atio
n
f
o
r
lo
n
g
er
p
er
io
d
s
o
f
tim
e.
I
n
ad
d
itio
n
,
a
b
id
ir
ec
tio
n
al
L
ST
M
h
o
ld
s
co
n
tex
tu
al
d
etails in
b
o
th
d
ir
ec
tio
n
s
,
w
h
ich
is
v
er
y
u
s
ef
u
l
w
h
en
class
if
y
in
g
tex
t.
5.
E
XP
E
R
I
M
E
NT
A
L
RE
SUL
T
Fo
r
th
e
e
v
al
u
atio
n
o
f
t
h
e
e
f
f
ici
en
c
y
o
f
t
h
i
s
clas
s
i
f
icatio
n
m
o
d
el,
th
e
clas
s
m
u
s
t
b
e
co
m
p
ar
ed
w
it
h
t
h
e
g
r
o
u
n
d
lin
e.
T
h
er
e
ar
e
m
u
lt
ip
le
cr
iter
ia
to
r
ea
lize
cla
s
s
i
f
icatio
n
e
f
f
ec
t
iv
e
n
es
s
.
E
x
p
er
im
en
tal
r
es
u
lt
s
ar
e
co
n
d
u
cted
o
n
t
h
e
g
r
o
u
n
d
-
tr
u
t
h
,
b
ased
o
n
th
e
u
s
ed
d
ataset.
As
m
e
n
tio
n
ed
ea
r
lier
,
th
e
u
s
er
's
r
ev
ie
w
s
w
er
e
s
o
r
ted
o
n
t
h
e
d
atase
t
o
f
t
h
e
c
o
r
r
esp
o
n
d
in
g
r
ati
n
g
v
al
u
e,
a
n
d
th
i
s
e
n
ab
led
u
s
to
ad
o
p
t
r
atin
g
as
a
ca
te
g
o
r
y
o
n
th
e
b
as
is
o
f
w
h
ic
h
w
e
cla
s
s
i
f
y
th
e
u
s
er
'
s
r
ev
ie
w
s
.
I
n
t
h
is
w
o
r
k
,
t
h
r
ee
cr
iter
ia
ar
e
u
s
ed
n
a
m
e
l
y
p
r
ec
is
io
n
,
r
ec
all
an
d
f
-
s
co
r
e.
P
r
ec
is
io
n
o
r
co
n
f
id
en
ce
(
as
c
alled
in
d
ata
m
i
n
i
n
g
)
i
n
d
icate
s
th
e
a
m
o
u
n
t
o
f
p
r
ed
icted
p
o
s
itiv
e
ca
s
e
s
r
ep
r
esen
tin
g
co
r
r
ec
t r
ea
l p
o
s
it
iv
es.
(
1
0
)
w
h
er
e
TP
is
tr
u
e
p
o
s
itiv
e
a
n
d
PP
is
p
r
ed
icted
p
o
s
itiv
es
a
n
d
T
P
is
r
ea
l p
o
s
itiv
e.
R
ec
all
o
r
s
e
n
s
iti
v
it
y
ca
n
b
e
d
ef
in
ed
a
s
th
e
r
ate
b
et
w
ee
n
t
h
e
n
u
m
b
er
o
f
co
r
r
ec
t
d
etec
tio
n
s
f
o
r
an
o
m
alo
u
s
m
ea
s
u
r
e
m
e
n
t
s
to
th
e
to
tal
s
u
m
o
f
an
o
m
alo
u
s
m
e
asu
r
e
m
en
t
s
.
(
1
1
)
F
-
Me
a
s
u
r
e
p
r
o
v
id
es
a
s
i
n
g
le
s
co
r
e
th
at
b
ala
n
ce
s
b
o
th
t
h
e
co
n
ce
r
n
s
o
f
p
r
ec
is
io
n
an
d
r
ec
all
in
o
n
e
n
u
m
b
er
.
(
1
2
)
w
h
e
n
tr
ain
i
n
g
t
h
e
B
i
-
L
ST
M
d
ee
p
n
et
w
o
r
k
,
w
e
o
b
tain
ed
a
p
r
ec
is
io
n
v
alu
e
o
f
0
.
9
5
4
3
,
w
h
ile
R
ec
al
l
an
d
F
-
Me
asu
r
e
w
er
e
0
.
9
5
9
7
an
d
0
.
6
5
5
8
,
r
esp
ec
tiv
el
y
.
Fi
g
u
r
e
4
s
h
o
w
s
t
h
e
co
n
f
u
s
io
n
m
atr
i
x
t
o
th
e
c
lass
if
ica
tio
n
r
esu
lt.
Fo
r
th
e
p
u
r
p
o
s
e
o
f
ex
a
m
i
n
i
n
g
t
h
e
r
elati
v
e
p
er
f
o
r
m
a
n
ce
s
o
f
clas
s
i
f
y
in
g
m
et
h
o
d
s
i
n
t
h
e
co
n
tex
t
o
f
m
u
ltip
le
ac
c
u
r
ac
y
e
v
al
u
atio
n
m
ea
s
u
r
es,
t
h
e
r
e
s
u
lt
o
f
B
i
-
L
ST
M
h
a
s
b
ee
n
co
m
p
ar
ed
w
it
h
a
n
u
m
b
er
o
f
class
i
f
ier
s
t
h
at
ar
e
w
ell
k
n
o
wn
in
t
h
e
m
ac
h
in
e
lear
n
i
n
g
f
iel
d
.
T
ab
le
1
lis
ts
th
e
r
esu
lt
o
f
c
o
m
p
ar
is
o
n
.
T
ab
le
1
p
r
esen
ts
a
s
u
m
m
ar
y
o
f
t
h
e
v
al
u
e
f
o
r
ea
ch
ac
cu
r
ac
y
m
ea
s
u
r
e
th
at
h
as
b
ee
n
o
b
tai
n
ed
t
h
r
o
u
g
h
f
i
v
e
cla
s
s
i
f
icatio
n
tech
n
iq
u
es
ap
p
lied
to
th
e
u
s
e
r
r
ev
ie
w
er
d
ataset.
I
t
i
n
d
icate
s
th
a
t
t
h
e
r
es
u
lts
o
f
t
h
e
B
i
-
L
ST
M
ar
e
t
h
e
b
est
r
esu
lt
s
.
Fi
g
u
r
e
5
s
h
o
w
s
t
h
e
c
r
iter
ia
f
o
r
ev
al
u
ati
n
g
t
h
e
e
f
f
ic
ien
c
y
o
f
th
e
p
r
o
p
o
s
ed
w
o
r
k
u
s
ed
in
co
m
p
ar
is
o
n
w
it
h
f
o
u
r
o
th
er
clas
s
i
f
ier
s
,
as
th
e
r
esu
l
ts
s
h
o
w
th
e
s
u
p
er
io
r
it
y
o
f
t
h
e
p
r
o
p
o
s
ed
s
y
s
te
m
o
v
er
th
e
r
est
o
f
t
h
e
class
i
f
ier
s
.
T
ab
le
1
.
T
h
e
class
if
ier
s
r
esu
l
ts
C
l
a
ssi
f
i
e
r
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F
-
M
e
a
su
r
e
B
e
r
n
o
l
l
i
0
.
8
8
2
9
0
.
8
3
8
6
0
.
8
6
0
2
R
a
n
d
o
m fo
r
e
st
0
.
8
8
6
9
0
.
9
4
5
1
0
.
9
3
7
7
D
e
c
i
si
o
n
t
r
e
e
0
.
8
8
6
3
0
.
9
3
5
1
0
.
9
3
7
6
K
n
e
i
g
h
b
o
r
s
0
.
8
5
8
6
0
.
9
3
9
0
0
.
9
2
3
5
B
i
L
S
T
M
0
.
9
5
4
3
0
.
9
5
9
7
0
.
9
5
5
8
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4752
S
en
timen
t a
n
a
lysi
s
fo
r
d
r
u
g
s
r
ev
iew
s
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ha
d
a
b
K
h
a
lid
Ob
a
ye
s
)
351
Fig
u
r
e
4
.
T
h
e
co
n
f
u
s
io
n
m
a
tr
i
x
o
f
clas
s
i
f
icatio
n
r
es
u
lt
Fig
u
r
e
5
.
T
h
e
r
esu
lt o
f
5
class
i
f
ier
s
6.
CO
NCLU
SI
O
N
A
cc
o
r
d
in
g
to
th
e
ex
p
er
i
m
en
t,
th
e
i
n
te
g
r
ated
Glo
Ve
an
d
B
i
L
ST
M
m
o
d
els
h
av
e
ac
h
iev
ed
g
o
o
d
s
co
r
es
in
ter
m
s
o
f
F1
-
s
co
r
e,
P
r
ec
is
io
n
an
d
R
ec
a
ll.
T
h
is
s
h
o
w
s
a
p
r
o
m
in
e
n
t
i
n
s
p
ir
atio
n
f
o
r
n
e
w
id
ea
s
to
b
e
ap
p
lied
in
o
th
er
f
u
t
u
r
e
w
o
r
k
s
.
T
h
e
m
o
d
e
l
ac
h
ie
v
ed
a
s
u
f
f
icie
n
t
p
er
f
o
r
m
an
ce
co
n
s
id
er
in
g
t
h
e
co
s
t
o
f
m
a
s
s
i
v
e
co
m
p
u
ti
n
g
p
o
w
er
.
I
n
th
is
s
tu
d
y
,
th
e
u
s
er
's
r
ev
ie
w
er
s
ar
e
u
s
ed
as
a
lab
el
f
o
r
th
e
t
w
o
cla
s
s
es,
t
h
e
p
o
s
iti
v
e
an
d
t
h
e
n
eg
ati
v
e
class
,
th
er
eb
y
d
i
v
id
in
g
t
h
e
ten
-
s
co
r
es
i
n
to
t
w
o
-
p
ar
ts
o
n
l
y
.
I
n
f
u
t
u
r
e
w
o
r
k
s
,
it
is
p
o
s
s
ib
le
to
u
s
e
t
h
e
m
o
d
el
af
ter
i
m
p
l
y
in
g
s
e
v
er
al
i
m
p
r
o
v
e
m
e
n
t
s
in
o
r
d
er
to
p
r
e
d
ict
th
e
u
s
er
’
s
r
atin
g
t
h
r
o
u
g
h
t
h
eir
r
ev
ie
w
er
s
.
RE
F
E
R
E
NC
E
S
[1
]
B.
W
a
n
g
,
S
.
X
io
n
g
,
Y.
Hu
a
n
g
,
a
n
d
X
.
L
i,
“
Re
v
ie
w
ra
ti
n
g
p
re
d
ictio
n
b
a
se
d
o
n
u
se
r
c
o
n
tex
t
a
n
d
p
r
o
d
u
c
t
c
o
n
tex
t,
”
Ap
p
l
.
S
c
i.
,
v
o
l.
8
,
n
o
.
1
0
,
p
p
.
1
-
1
3
,
2
0
1
8
,
d
o
i:
1
0
.
3
3
9
0
/ap
p
8
1
0
1
8
4
9
.
[2
]
V
.
G
o
p
a
lak
rish
n
a
n
a
n
d
C.
Ra
m
a
sw
a
m
y
,
“
P
a
ti
e
n
t
o
p
i
n
io
n
m
in
in
g
to
a
n
a
ly
z
e
d
ru
g
s
sa
ti
s
f
a
c
ti
o
n
u
sin
g
su
p
e
rv
ise
d
lea
rn
in
g
,
”
J
.
Ap
p
l.
Res
.
T
e
c
h
n
o
l.
,
v
o
l.
1
5
,
n
o
.
4
,
p
p
.
3
1
1
-
3
1
9
,
2
0
1
7
,
d
o
i:
1
0
.
1
0
1
6
/j
.
jart.
2
0
1
7
.
0
2
.
0
0
5
.
[3
]
L
.
Zh
a
n
g
,
S
.
W
a
n
g
,
B.
L
iu
,
“
De
e
p
L
e
a
rn
in
g
f
o
r
S
e
n
ti
m
e
n
t
A
n
a
l
y
si
s:
A
S
u
rv
e
y
,
”
W
IRE
’s
d
a
t
a
mi
n
in
g
a
n
d
k
n
o
wled
g
e
d
isc
o
v
e
ry
.
,
v
o
l.
8
,
n
o
.
4
,
e
1
2
5
3
,
2
0
1
8
,
d
o
i:
1
0
.
1
0
0
2
/w
id
m
.
1
2
5
3
.
0
,
7
5
0
,
8
0
,
8
5
0
,
9
0
,9
5
1
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F
-
M
e
a
su
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
23
,
No
.
1
,
J
u
ly
2
0
2
1
:
345
-
3
5
3
352
[4
]
P
.
Ha
jek
,
A
.
Ba
ru
sh
k
a
,
a
n
d
M
.
M
u
n
k
,
“
Op
i
n
io
n
m
in
in
g
o
f
c
o
n
su
m
e
r
re
v
ie
w
s
u
sin
g
d
e
e
p
n
e
u
ra
l
n
e
tw
o
rk
s
w
it
h
w
o
rd
-
se
n
ti
m
e
n
t
a
ss
o
c
iatio
n
s
,
”
v
o
l.
5
8
3
,
p
p
.
4
1
9
-
4
2
9
,
IF
I
P
.
S
p
rin
g
e
r
In
ter
n
a
ti
o
n
a
l
P
u
b
li
s
h
in
g
,
2
0
2
0
.
I
n
:
M
a
g
lo
g
ian
n
is
I.
,
Ili
a
d
is
L
.
,
P
im
e
n
id
is
E.
(e
d
s)
A
rti
f
icia
l
In
telli
g
e
n
c
e
A
p
p
li
c
a
t
io
n
s
a
n
d
In
n
o
v
a
ti
o
n
s.
A
I
A
I
2
0
2
0
,
d
o
i:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
4
9
1
6
1
-
1
_
3
5
[5
]
S
.
V
ij
a
y
a
ra
g
h
a
v
a
n
a
n
d
D.
Ba
su
,
“
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis
in
Dru
g
Re
v
ie
w
s
u
sin
g
S
u
p
e
rv
ise
d
M
a
c
h
in
e
L
e
a
rn
in
g
A
l
g
o
rit
h
m
s,”
a
rXiv
,
2
0
2
0
.
[6
]
F
.
G
rä
ß
e
r,
H.
M
a
lb
e
rg
,
S
.
Ka
ll
u
m
a
d
i,
a
n
d
S
.
Zau
n
se
d
e
r,
“
A
sp
e
c
t
-
Ba
se
d
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis
o
f
Dru
g
R
e
v
ie
ws
A
p
p
l
y
in
g
Cro
ss
-
Do
m
a
in
a
n
d
Cro
ss
-
Da
ta
Lea
rn
in
g
,
”
DH
'
1
8
:
Pro
c
e
e
d
in
g
s
o
f
th
e
2
0
1
8
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Dig
it
a
l
He
a
lt
h
,
v
o
l.
2
0
1
8
-
A
p
ril
,
p
p
.
1
2
1
–
1
2
5
,
2
0
1
8
.
[7
]
N.
C.
Da
n
g
,
M
.
N.
M
o
re
n
o
-
G
a
rc
ía,
a
n
d
F
.
De
la P
rieta
,
“
S
e
n
ti
m
e
n
t
a
n
a
ly
sis
b
a
se
d
o
n
d
e
e
p
lea
rn
i
n
g
:
A
c
o
m
p
a
ra
ti
v
e
s
tu
d
y
,
”
El
e
c
tro
n
.
,
v
o
l
.
9
,
n
o
.
3
,
p
p
.
4
8
3
,
2
0
2
0
,
d
o
i:
1
0
.
3
3
9
0
/ele
c
tro
n
ics
9
0
3
0
4
8
3
.
[8
]
P
a
tel,
V.,
M
ish
ra
,
P
.
,
&
P
a
tn
i
,
J.
C.
(2
0
1
8
,
Ju
n
e
).
P
sy
He
a
l:
A
n
A
p
p
ro
a
c
h
to
Re
m
o
te
M
e
n
tal
He
a
lt
h
M
o
n
i
to
ri
n
g
S
y
st
e
m
.
In
2
0
1
8
I
n
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
A
d
v
a
n
c
e
s
in
Co
m
p
u
ti
n
g
a
n
d
C
o
m
m
u
n
ica
ti
o
n
En
g
in
e
e
rin
g
(ICA
CCE)
(p
p
.
3
8
4
-
3
9
3
).
IE
EE
.
[9
]
M
.
Ka
n
e
k
o
a
n
d
D.
B
o
ll
e
g
a
la,
“
Au
to
e
n
c
o
d
i
n
g
Im
p
ro
v
e
s
P
re
-
train
e
d
W
o
rd
Em
b
e
d
d
in
g
s,”
in
Pro
c
e
e
d
in
g
s
o
f
th
e
2
8
t
h
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
o
mp
u
t
a
ti
o
n
a
l
L
i
n
g
u
isti
c
s
,
Ba
rc
e
lo
n
a
,
2
0
2
0
,
p
p
.
1
6
9
9
-
1
7
1
3
.
[1
0
]
C.
D.
P
e
n
n
i
n
g
to
n
,
Je
ff
re
y
;
S
o
c
h
e
r,
Rich
a
rd
;
M
a
n
n
i
n
g
,
“
G
lo
V
e
:
G
lo
b
a
l
V
e
c
to
rs
f
o
r
W
o
rd
Re
p
re
se
n
tatio
n
,
”
i
n
Co
n
fer
e
n
c
e
o
n
Emp
iric
a
l
M
e
th
o
d
s in
Na
tu
ra
l
L
a
n
g
u
a
g
e
Pr
o
c
e
ss
in
g
(
EM
NL
P)
,
2
0
1
4
,
v
o
l.
1
9
,
n
o
.
5
,
p
p
.
1
5
3
2
-
1
5
4
3
.
[1
1
]
V
.
M
a
k
a
re
n
k
o
v
,
B.
S
h
a
p
ira,
a
n
d
L
.
Ro
k
a
c
h
,
“
L
a
n
g
u
a
g
e
M
o
d
e
ls
w
it
h
P
re
-
T
ra
in
e
d
(
G
lo
V
e
)
W
o
rd
E
m
b
e
d
d
in
g
s,
”
a
rXiv Co
mp
u
t.
L
a
n
g
.
,
2
0
1
6
.
[1
2
]
H.
K.
Ob
a
y
e
s,
N.
A
l
A
,
a
n
d
E.
A
l
-
sh
a
m
e
r
y
,
“
A
I
DS’
s Dru
g
s Qu
a
n
ti
f
ica
ti
o
n
a
n
d
S
u
rv
e
il
lan
c
e
Us
in
g
De
e
p
L
e
a
rn
in
g
,
”
v
o
l.
5
9
,
n
o
.
6
s,
p
p
.
2
8
2
-
2
9
0
,
2
0
1
9
.
[1
3
]
S
.
Bo
u
k
ti
f
,
A
.
F
iaz
,
A
.
Ou
n
i,
a
n
d
M
.
A
.
S
e
rh
a
n
i,
“
Op
ti
m
a
l
d
e
e
p
lea
rn
in
g
L
S
T
M
m
o
d
e
l
f
o
r
e
lec
tri
c
lo
a
d
f
o
re
c
a
stin
g
u
sin
g
f
e
a
tu
re
se
l
e
c
ti
o
n
a
n
d
g
e
n
e
ti
c
a
lg
o
rit
h
m
:
Co
m
p
a
riso
n
w
it
h
m
a
c
h
in
e
lea
rn
in
g
a
p
p
ro
a
c
h
e
s,”
E
n
e
rg
ies
,
v
o
l.
1
1
,
no
.
7
,
p
p
.
1
-
2
0
,
2
0
1
8
,
d
o
i:
1
0
.
3
3
9
0
/en
1
1
0
7
1
6
3
6
.
[1
4
]
C.
Zh
a
n
g
a
n
d
P
.
C.
W
o
o
d
lan
d
,
“
Hig
h
Ord
e
r
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
tw
o
rk
s
f
o
r
A
c
o
u
stic
M
o
d
e
ll
in
g
,
”
2
0
1
8
IE
E
E
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
Ac
o
u
stics
,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pro
c
e
ss
in
g
(
ICAS
S
P)
,
2
0
1
8
,
p
p
.
5
8
4
9
-
5
8
5
3
,
d
o
i
:
1
0
.
1
1
0
9
/ICA
S
S
P
.
2
0
1
8
.
8
4
6
1
6
0
8
.
[1
5
]
H.
K.
Ob
a
y
e
s,
N.
A
l
-
A
’
a
ra
ji
,
a
n
d
E.
A
l
-
S
h
a
m
e
r
y
,
“
Ex
a
m
in
a
ti
o
n
a
n
d
f
o
re
c
a
stin
g
o
f
d
ru
g
c
o
n
su
m
p
ti
o
n
b
a
se
d
o
n
re
c
u
rre
n
t
d
e
e
p
lea
rn
in
g
,
”
In
t.
J
.
Rec
e
n
t
T
e
c
h
n
o
l.
E
n
g
.
,
v
o
l.
8
,
n
o
.
2
S
p
e
c
ial
Iss
u
e
1
0
,
p
p
.
4
1
4
-
4
2
0
,
2
0
1
9
,
d
o
i
:
10.
3
5
9
4
0
/i
jrt
e
.
B
1
0
6
9
.
0
9
8
2
S
1
0
1
9
.
[1
6
]
J.
Ku
m
a
r,
R.
G
o
o
m
e
r,
a
n
d
A
.
K.
S
i
n
g
h
,
“
L
o
n
g
S
h
o
r
t
T
e
rm
M
e
m
o
r
y
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
tw
o
rk
(L
S
T
M
-
RNN
)
Ba
se
d
W
o
rk
lo
a
d
F
o
re
c
a
stin
g
M
o
d
e
l
F
o
r
Clo
u
d
Da
tac
e
n
ters
t,
”
Pro
c
e
d
ia
Co
m
p
u
t
.
S
c
i.
,
v
o
l.
1
2
5
,
n
o
.
Ja
n
u
a
ry
,
p
p
.
676
-
6
8
2
,
2
0
1
8
, d
o
i:
1
0
.
1
0
1
6
/j
.
p
ro
c
s.2
0
1
7
.
1
2
.
0
8
7
.
[1
7
]
A
.
P
o
g
iatz
is
a
n
d
G
.
S
a
m
a
k
o
v
i
ti
s,
“
Us
in
g
b
il
st
m
n
e
t
w
o
rk
s
f
o
r
c
o
n
tex
t
-
a
w
a
re
d
e
e
p
se
n
siti
v
it
y
lab
e
ll
in
g
o
n
c
o
n
v
e
rsa
ti
o
n
a
l
d
a
ta,”
A
p
p
l.
S
c
i
.
,
v
o
l.
1
0
,
n
o
.
2
4
,
p
p
.
1
–
1
7
,
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/ap
p
1
0
2
4
8
9
2
4
.
[1
8
]
B.
Y.
L
in
,
F
.
X
u
,
Z.
L
u
o
,
K
.
Z
h
u
,
“
M
u
lt
i
-
c
h
a
n
n
e
l
BiL
S
T
M
-
CRF
M
o
d
e
l
f
o
r
Em
e
rg
in
g
Na
m
e
d
En
ti
ty
Re
c
o
g
n
it
io
n
i
n
S
o
c
ial
M
e
d
ia,”
Pro
c
e
e
d
in
g
s
o
f
t
h
e
3
rd
W
o
rk
sh
o
p
o
n
No
isy
Us
e
r
-
g
e
n
e
ra
ted
T
e
x
t
,
C
o
p
e
n
h
a
g
e
n
,
2
0
1
7
,
p
p
.
1
6
0
-
1
6
5
,
d
o
i:
1
0
.
1
8
6
5
3
/v
1
/W
1
7
-
4
4
2
1
.
[1
9
]
A
.
Az
iz
S
h
a
rf
u
d
d
in
,
M
.
Na
f
is
Ti
h
a
m
i
a
n
d
M
.
S
a
if
u
l
Isla
m
,
“
A
De
e
p
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
t
w
o
rk
w
it
h
BiL
S
T
M
m
o
d
e
l
f
o
r
S
e
n
ti
m
e
n
t
Clas
sif
ic
a
ti
o
n
,
”
2
0
1
8
I
n
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
B
a
n
g
la
S
p
e
e
c
h
a
n
d
L
a
n
g
u
a
g
e
Pr
o
c
e
ss
in
g
(
ICBS
L
P)
,
2
0
1
8
,
p
p
.
1
-
4
,
d
o
i:
1
0
.
1
1
0
9
/ICBS
L
P
.
2
0
1
8
.
8
5
5
4
3
9
6
.
[2
0
]
G
.
G
u
a
n
a
n
d
M
.
Z
h
u
,
“
Ne
w
Re
s
e
a
rc
h
o
n
T
ra
n
sf
e
r
Lea
rn
in
g
M
o
d
e
l
o
f
Na
m
e
d
En
ti
ty
Re
c
o
g
n
it
io
n
,
”
J
.
Ph
y
s.
Co
n
f
.
S
e
r.
,
v
o
l.
1
2
6
7
,
n
o
.
1
,
2
0
1
9
.
[2
1
]
T
.
T
h
ireo
u
a
n
d
M
.
Re
c
z
k
o
,
“
Bid
irec
ti
o
n
a
l
L
o
n
g
S
h
o
rt
-
T
e
r
m
M
e
m
o
r
y
Ne
t
w
o
rk
s
f
o
r
P
re
d
icti
n
g
th
e
S
u
b
c
e
ll
u
lar
L
o
c
a
li
z
a
ti
o
n
o
f
Eu
k
a
ry
o
ti
c
P
r
o
tein
s,”
in
IEE
E
/A
CM
T
ra
n
sa
c
ti
o
n
s
o
n
Co
m
p
u
t
a
ti
o
n
a
l
Bi
o
l
o
g
y
a
n
d
Bi
o
in
fo
rm
a
t
ics
,
v
o
l.
4
,
n
o
.
3
,
p
p
.
4
4
1
-
4
4
6
,
Ju
ly
-
S
e
p
t.
2
0
0
7
,
d
o
i:
1
0
.
1
1
0
9
/t
c
b
b
.
2
0
0
7
.
1
0
1
5
.
[2
2
]
L
.
Qu
,
G
.
I
f
ri
m
,
a
n
d
G
.
Weik
u
m
,
“
T
h
e
b
a
g
-
of
-
o
p
in
i
o
n
s
m
e
th
o
d
f
o
r
re
v
ie
w
r
a
ti
n
g
p
re
d
ict
io
n
f
ro
m
sp
a
rs
e
te
x
t
p
a
tt
e
rn
s.,
”
Co
li
n
g
2
0
1
0
-
2
3
rd
In
t.
Co
n
f.
C
o
mp
u
t.
L
i
n
g
u
ist.
Pro
c
.
Co
n
f.
,
Be
ij
in
g
,
v
o
l
.
2
,
n
o
.
Ja
n
u
a
ry
,
p
p
.
9
1
3
–
9
2
1
,
2
0
1
0
.
[2
3
]
F
.
G
re
a
v
e
s,
D.
Ra
m
ir
e
z
-
Ca
n
o
,
C.
M
il
lett,
A
.
Da
rz
i,
a
n
d
L
.
Do
n
a
ld
so
n
,
“
Us
e
o
f
se
n
ti
m
e
n
t
a
n
a
l
y
si
s
f
o
r
c
a
p
tu
rin
g
p
a
ti
e
n
t
e
x
p
e
rien
c
e
f
ro
m
f
re
e
-
t
e
x
t
c
o
m
m
e
n
ts
p
o
ste
d
o
n
li
n
e
,
”
J
.
M
e
d
.
In
ter
n
e
t
Res
.
,
v
o
l.
1
5
,
n
o
.
1
1
,
2
0
1
3
,
p
p
.
1
-
9.
[2
4
]
A
.
N
i
k
f
a
r
j
a
m
,
A
.
S
a
r
k
e
r
,
K
.
O
’
C
o
n
n
o
r
,
R
.
G
i
n
n
,
a
n
d
G
.
G
o
n
z
a
l
e
z
,
“
P
h
a
r
m
a
c
o
v
i
g
i
l
a
n
c
e
f
r
o
m
s
o
c
i
a
l
m
e
d
i
a
:
M
i
n
i
n
g
a
d
v
e
r
s
e
d
r
u
g
r
e
a
c
t
i
o
n
m
e
n
t
i
o
n
s
u
s
i
n
g
s
e
q
u
e
n
c
e
l
a
b
e
l
i
n
g
w
i
t
h
w
o
r
d
e
m
b
e
d
d
i
n
g
c
l
u
s
t
e
r
f
e
a
t
u
r
e
s
,
”
J
.
A
m
.
M
e
d
.
I
n
f
o
r
m
a
t
i
c
s
A
s
s
o
c
.
,
v
o
l
.
2
2
,
n
o
.
3
,
p
p
.
6
7
1
–
6
8
1
,
2
0
1
5
,
d
o
i
:
1
0
.
1
0
9
3
/
j
a
m
i
a
/
o
c
u
0
4
1
.
[2
5
]
A
.
Na
v
in
d
g
i,
C.
Br
u
n
,
C.
Bo
u
l
a
rd
M
a
ss
o
n
,
a
n
d
S
.
No
w
so
n
,
“
S
tep
s
T
o
w
a
rd
A
u
to
m
a
ti
c
Un
d
e
rsta
n
d
in
g
o
f
th
e
F
u
n
c
ti
o
n
o
f
Aff
e
c
ti
v
e
L
a
n
g
u
a
g
e
in
S
u
p
p
o
r
t
G
ro
u
p
s,”
Pro
c
e
e
d
i
n
g
s
o
f
T
h
e
Fo
u
rth
I
n
ter
n
a
ti
o
n
a
l
W
o
rk
sh
o
p
o
n
Na
tu
ra
l
L
a
n
g
u
a
g
e
Pr
o
c
e
ss
in
g
f
o
r S
o
c
i
a
l
M
e
d
ia
,
A
u
stin
,
2
0
1
6
,
p
p
.
2
6
–
3
3
,
d
o
i:
1
0
.
1
8
6
5
3
/v
1
/W
1
6
-
6
2
0
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4752
S
en
timen
t a
n
a
lysi
s
fo
r
d
r
u
g
s
r
ev
iew
s
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ha
d
a
b
K
h
a
lid
Ob
a
ye
s
)
353
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
H
a
d
a
b
K
h
a
li
d
O
b
a
y
e
s
is
a
ss
is
tan
t
p
r
o
f
e
ss
o
r
a
t
Un
iv
e
rsit
y
o
f
Ba
b
y
lo
n
,
Ba
b
e
l,
Ira
q
.
S
h
e
re
c
e
iv
e
d
h
e
r
P
h
D d
e
g
re
e
f
ro
m
Un
iv
e
rsit
y
o
f
Ba
b
y
lo
n
,
Ira
q
,
in
2
0
2
0
.
S
h
e
p
u
b
li
sh
e
d
a
n
u
m
b
e
r
o
f
p
a
p
e
rs
in
in
tern
a
ti
o
n
a
l
jo
u
rn
a
ls
a
n
d
re
f
e
rre
d
c
o
n
fe
re
n
c
e
s.
He
r
c
u
rre
n
t
re
se
a
rc
h
a
re
a
s
a
re
,
A
I,
M
a
c
h
in
e
lea
rn
in
g
,
n
e
u
ra
l
n
e
tw
o
rk
s,
d
e
e
p
lea
rn
in
g
,
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
a
n
d
I
m
a
g
e
P
r
o
c
e
ss
in
g
Fi
r
a
s
S
a
b
a
h
Al
-
Tu
r
a
i
h
i
is
a
lec
tu
re
r
a
t
Un
iv
e
rsit
y
o
f
Ba
b
y
lo
n
,
Co
ll
e
g
e
o
f
In
f
o
r
m
a
ti
o
n
T
e
c
h
n
o
lo
g
y
,
Ne
t
w
o
rk
s
De
p
a
rt
m
e
n
ts.
He
re
c
e
iv
e
d
h
is
P
h
D
d
e
g
re
e
f
ro
m
Bru
n
e
l
Un
iv
e
rsit
y
L
o
n
d
o
n
,
U
n
it
e
d
Kin
g
d
o
m
.
He
re
c
e
iv
e
d
h
is
BS
c
a
n
d
M
S
c
d
e
g
re
e
s
in
C
o
m
p
u
ter
S
c
ien
c
e
.
Hi
s
re
se
a
rc
h
in
tere
sts c
o
v
e
r
c
o
m
m
u
n
ica
ti
o
n
n
e
tw
o
rk
s.
K
h
a
ld
o
o
n
H
a
sa
n
Alh
u
ss
a
y
n
i
re
c
e
iv
e
d
th
e
BS
c
d
e
g
re
e
s
in
Co
m
p
u
ter
S
c
ien
c
e
f
ro
m
th
e
Un
iv
e
rsit
y
o
f
B
a
b
y
lo
n
,
Ira
q
,
i
n
2
0
0
8
.
Af
ter
c
o
m
p
letin
g
h
is
BS
c
,
h
e
w
o
rk
e
d
a
s
a
p
ro
g
ra
m
m
e
r
a
t
th
e
De
p
a
rtm
e
n
t
o
f
Co
m
p
u
ter
Ce
n
ter,
th
e
Un
iv
e
rsity
o
f
Ba
b
y
lo
n
in
2
0
0
9
.
He
re
c
e
iv
e
d
th
e
M
S
c
d
e
g
re
e
s
in
Co
m
p
u
te
r
S
c
ien
c
e
a
n
d
Co
m
p
u
ter
E
n
g
in
e
e
rin
g
f
ro
m
th
e
T
u
la
S
tate
Un
iv
e
rsit
y
,
Ru
ss
ian
,
in
2
0
1
4
.
S
e
c
in
c
e
No
v
2
0
1
6
.
He
e
n
tere
d
th
e
T
o
m
sk
sta
t
e
u
n
iv
e
rsit
y
(
T
S
U),
De
p
a
rtme
n
t
o
f
c
o
m
p
u
ter
sc
ien
c
e
th
e
o
ry
a
s
a
P
h
.
D
.
stu
d
e
n
t
in
sp
e
c
ialt
y
M
a
th
e
m
a
ti
c
a
l
a
n
d
so
f
t
w
a
re
o
f
c
o
m
p
u
ters
,
c
o
m
p
lex
e
s
a
n
d
c
o
m
p
u
ter
n
e
tw
o
rk
s.
His
m
a
in
re
se
a
rc
h
in
tere
sts
a
re
m
a
c
h
in
e
lea
rn
in
g
,
n
e
u
ra
l
n
e
tw
o
rk
s,
d
e
e
p
lea
rn
in
g
,
d
ialo
g
u
e
s
y
s
tem
s
a
n
d
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.