I
AE
S
I
nte
rna
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
9
,
No
.
2
,
J
u
n
e
2020
,
p
p
.
2
12
~
2
2
0
I
SS
N:
2252
-
8938
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ai.
v
9
.i
2
.
p
p
2
12
-
2
2
0
212
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Senti
m
en
t
a
na
ly
sis
of inf
o
r
m
a
l Ma
l
a
y
t
w
eets w
ith
deep learning
O
ng
J
un
Ying
1
,
M
uh
a
mm
a
d
M
un
’
i
m
Ah
m
a
d
Z
a
bid
i
2
,
No
rha
f
iza
h Ra
m
li
3
,
Us
m
a
n Ul
la
h Shei
kh
4
1
De
p
a
rtme
n
t
o
f
De
v
ice
D
e
v
e
lo
p
m
e
n
t
G
ro
u
p
,
I
n
tel
P
e
n
a
n
g
,
M
a
lay
si
a
2,
3,
4
S
c
h
o
o
l
o
f
El
e
c
tri
c
a
l
En
g
in
e
e
ri
n
g
,
F
a
c
u
lt
y
o
f
En
g
in
e
e
rin
g
,
Un
iv
e
rsiti
T
e
k
n
o
lo
g
i
M
a
lay
sia
,
M
a
la
y
sia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
1
6
,
2
0
20
R
ev
i
s
ed
Mar
6
,
2
0
20
A
cc
ep
ted
A
pr
17
,
2
0
20
Tw
it
ter
is
a
n
o
n
l
in
e
m
icro
b
lo
g
g
in
g
a
n
d
s
o
c
ial
-
n
e
tw
o
rk
in
g
p
latf
o
rm
w
h
ich
a
ll
o
w
s
u
se
rs
to
w
rit
e
sh
o
rt
m
e
ss
a
g
e
s
c
a
ll
e
d
t
w
e
e
ts.
It
h
a
s
o
v
e
r
3
3
0
m
il
li
o
n
re
g
istere
d
u
se
rs
g
e
n
e
ra
ti
n
g
n
e
a
rl
y
2
5
0
m
il
li
o
n
tw
e
e
ts
p
e
r
d
a
y
.
As
M
a
la
y
is
th
e
n
a
ti
o
n
a
l
lan
g
u
a
g
e
in
M
a
lay
s
ia
,
th
e
re
is
a
sig
n
if
ica
n
t
n
u
m
b
e
r
o
f
u
se
rs
tw
e
e
ti
n
g
in
M
a
lay
.
Tw
e
e
ts
h
a
v
e
a
m
a
x
i
m
u
m
len
g
th
o
f
1
4
0
c
h
a
ra
c
ters
w
h
ich
f
o
rc
e
s
u
se
rs
to
sta
y
f
o
c
u
se
d
o
n
th
e
m
e
ss
a
g
e
th
e
y
w
ish
to
d
i
ss
e
m
in
a
te.
T
h
is
c
h
a
ra
c
teristic
m
a
k
e
s
twe
e
ts
a
n
in
tere
sti
n
g
su
b
jec
t
f
o
r
se
n
ti
m
e
n
t
a
n
a
ly
sis.
S
e
n
ti
m
e
n
t
a
n
a
l
y
sis
is
a
n
a
tu
ra
l
la
n
g
u
a
g
e
p
ro
c
e
ss
in
g
(N
L
P
)
tas
k
o
f
c
las
si
fy
in
g
w
h
e
th
e
r
a
t
we
e
t
h
a
s
a
p
o
siti
v
e
o
r
n
e
g
a
ti
v
e
se
n
ti
m
e
n
t.
Tw
e
e
ts
in
M
a
la
y
a
r
e
c
h
o
se
n
in
t
h
is
stu
d
y
a
s
li
m
it
e
d
re
se
a
rc
h
h
a
s
b
e
e
n
d
o
n
e
o
n
t
h
is
lan
g
u
a
g
e
.
In
th
is
w
o
rk
,
se
n
ti
m
e
n
t
a
n
a
l
y
sis
a
p
p
li
e
d
t
o
M
a
lay
t
w
e
e
ts
u
sin
g
th
e
d
e
e
p
lea
rn
in
g
m
o
d
e
l.
W
e
a
c
h
iev
e
d
7
7
.
5
9
%
a
c
c
u
ra
c
y
w
h
ich
e
x
c
e
e
d
s
si
m
il
a
r
w
o
rk
d
o
n
e
o
n
Ba
h
a
sa
In
d
o
n
e
sia
.
K
ey
w
o
r
d
s
:
B
ah
asa
I
n
d
o
n
es
ia
C
o
n
v
o
lu
tio
n
al
n
e
u
r
al
n
e
t
w
o
r
k
Ma
la
y
Natu
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
Sen
ti
m
e
n
t a
n
a
l
y
s
is
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
:
Mu
h
a
m
m
ad
Mu
n
’
i
m
A
h
m
ad
Z
ab
id
i,
Dep
ar
t
m
en
t o
f
E
lectr
o
n
ics a
n
d
C
o
m
p
u
ter
E
n
g
i
n
ee
r
i
n
g
,
Sch
o
o
l o
f
E
lectr
ical
E
n
g
i
n
ee
r
i
n
g
,
Facu
lt
y
o
f
E
n
g
i
n
ee
r
i
n
g
,
U
n
i
v
e
r
s
iti T
ek
n
o
lo
g
i M
ala
y
s
ia
,
8
1
3
1
0
U
T
M
J
o
h
o
r
,
Ma
lay
s
ia.
E
m
ail:
m
u
n
i
m
@
u
t
m
.
m
y
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
ad
v
an
ce
m
e
n
t
o
f
th
e
I
n
ter
n
et
ca
u
s
e
s
th
e
w
o
r
ld
t
o
f
ac
e
ev
er
-
g
r
o
w
i
n
g
v
o
lu
m
e
s
o
f
d
ata
in
v
ar
io
u
s
f
o
r
m
s
.
A
p
p
licatio
n
s
s
u
ch
as
s
o
cial
m
ed
ia,
r
e
v
ie
w
s
ite
s
,
f
o
r
u
m
s
,
b
lo
g
s
,
an
d
o
th
er
s
g
en
er
at
e
en
o
r
m
o
u
s
h
ea
p
s
o
f
d
ata
in
t
h
e
f
o
r
m
o
f
s
e
n
ti
m
en
t,
o
p
in
io
n
,
u
s
er
'
s
v
ie
w
s
,
ar
g
u
m
e
n
t
s
a
b
o
u
t
d
if
f
er
en
t
s
o
cial
ev
en
t
s
,
p
o
liti
cs,
p
r
o
d
u
cts
an
d
m
o
r
e.
T
h
e
s
en
ti
m
en
t
s
ex
p
r
es
s
ed
b
y
u
s
er
s
h
a
v
e
a
v
er
y
g
r
e
a
t
in
f
l
u
e
n
ce
o
n
th
e
r
ea
d
er
s
,
p
r
o
d
u
ct
v
en
d
o
r
s
,
an
d
ev
en
p
o
liti
cian
s
.
As
a
r
esu
lt,
m
a
n
y
co
m
p
a
n
ie
s
d
ep
en
d
o
n
I
n
ter
n
et
u
s
er
s
'
f
ee
d
b
ac
k
a
n
d
o
p
in
io
n
s
to
m
a
r
k
et
t
h
eir
p
r
o
d
u
cts
a
n
d
s
er
v
i
ce
s
o
n
s
o
cial
m
ed
ia
s
u
c
h
as
T
w
it
ter
,
Face
b
o
o
k
,
Sn
ap
ch
at,
an
d
I
n
s
ta
g
r
a
m
.
C
o
m
p
a
n
i
e
s
ar
e
i
n
ter
es
t
ed
in
k
n
o
w
i
n
g
w
h
at
u
s
er
s
t
h
i
n
k
ab
o
u
t
t
h
eir
s
er
v
ice
s
o
r
p
r
o
d
u
cts.
T
h
e
ch
an
g
e
s
in
t
h
es
e
b
u
s
in
e
s
s
m
o
d
els
h
ad
cr
ea
ted
a
h
u
g
e
b
u
s
i
n
es
s
o
p
p
o
r
tu
n
ity
f
o
r
th
e
s
en
t
i
m
e
n
t
a
n
al
y
s
is
o
f
t
h
e
d
ata
[
1
]
.
Ma
la
y
is
Ma
la
y
s
ia
's
n
atio
n
a
l
la
n
g
u
a
g
e.
I
n
Ma
la
y
s
ia
,
t
h
e
r
e
ar
e
a
s
i
g
n
i
f
ica
n
t
n
u
m
b
er
o
f
u
s
er
s
w
h
o
ar
e
u
s
i
n
g
Ma
la
y
to
ex
p
r
ess
t
h
eir
o
p
in
io
n
s
an
d
ar
g
u
m
e
n
t
s
o
n
s
o
cial
m
ed
ia.
T
h
er
e
is
a
h
u
g
e
b
u
s
in
e
s
s
o
p
p
o
r
tu
n
it
y
f
o
r
b
u
ild
in
g
a
Ma
la
y
s
e
n
ti
m
e
n
t
an
al
y
s
is
m
o
d
el.
Ho
w
e
v
er
,
v
er
y
l
i
m
ited
r
esear
ch
h
as
b
ee
n
attr
ib
u
ted
to
Ma
la
y
s
e
n
ti
m
en
t a
n
al
y
s
is
[
2
]
.
Data
o
n
s
o
cial
m
ed
ia
ar
e
m
o
s
tl
y
u
n
s
tr
u
ct
u
r
ed
.
T
h
e
lar
g
e
am
o
u
n
ts
o
f
u
n
s
tr
u
ct
u
r
ed
d
ata
m
ak
e
it
v
er
y
d
if
f
ic
u
lt
f
o
r
a
h
u
m
a
n
to
ex
tr
ac
t
an
d
s
u
m
m
ar
i
ze
th
e
o
p
in
io
n
c
o
n
tain
ed
th
er
ei
n
.
O
f
te
n
,
h
u
m
a
n
er
r
o
r
s
w
il
l
o
cc
u
r
th
u
s
r
ed
u
c
i
n
g
t
h
e
ac
c
u
r
ac
y
o
f
th
e
d
ata
an
a
l
y
s
is
.
A
u
to
m
ated
s
en
ti
m
e
n
t
an
al
y
s
is
w
i
ll
g
r
ea
tl
y
r
ed
u
ce
t
h
e
w
o
r
k
lo
ad
f
o
r
a
h
u
m
a
n
to
an
al
y
ze
t
h
e
d
ata
[3
-
4]
.
T
o
e
x
p
lo
it
th
i
s
o
p
p
o
r
tu
n
it
y
,
m
an
y
s
tar
t
-
u
p
s
ar
e
n
o
w
p
r
o
v
id
in
g
s
e
n
ti
m
e
n
t
an
a
l
y
s
is
s
er
v
ices
f
o
r
th
e
p
u
b
lic.
L
i
k
e
wis
e,
m
a
n
y
b
i
g
co
r
p
o
r
atio
n
s
ar
e
also
b
u
ild
in
g
th
e
ir
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
S
en
timen
t a
n
a
lysi
s
o
f in
fo
r
ma
l Ma
la
y
tw
ee
ts
w
i
th
d
ee
p
lea
r
n
in
g
(
On
g
Ju
n
Yin
g
)
213
in
-
h
o
u
s
e
ca
p
ab
ilit
ies
in
s
e
n
ti
m
en
t
an
al
y
s
is
.
T
h
ese
p
r
ac
tical
ap
p
licatio
n
s
an
d
i
n
d
u
s
tr
ial
in
ter
ests
h
av
e
p
r
o
v
id
ed
a
s
tr
o
n
g
m
o
tiv
a
tio
n
f
o
r
r
esear
ch
in
s
en
t
i
m
e
n
t a
n
al
y
s
i
s
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
2
.
1
.
Na
t
ura
l la
ng
ua
g
e
pro
ce
s
s
ing
Natu
r
al
L
a
n
g
u
a
g
e
P
r
o
ce
s
s
in
g
is
a
f
ield
th
at
in
ter
s
ec
t
s
o
f
ar
tific
ial
in
telli
g
e
n
ce
,
co
m
p
u
ter
s
c
ien
ce
,
an
d
lin
g
u
i
s
tic
s
.
I
t
r
ef
er
s
to
an
i
n
t
er
ac
tio
n
b
et
w
ee
n
m
ac
h
i
n
es
a
n
d
h
u
m
an
s
i
n
n
at
u
r
al
la
n
g
u
a
g
e
[
5
]
.
T
h
e
u
lti
m
ate
g
o
al
o
f
N
L
P
is
to
en
ab
le
co
m
p
u
ter
s
to
u
n
d
er
s
ta
n
d
lan
g
u
ag
e
as
w
ell
a
s
h
u
m
a
n
s
d
o
.
NL
P
is
d
iv
id
ed
in
to
th
r
ee
ca
teg
o
r
ies,
w
h
ic
h
ar
e:
Natu
r
al
L
an
g
u
ag
e
U
n
d
er
s
ta
n
d
in
g
: Co
m
p
u
ter
s
'
ab
ilit
y
to
u
n
d
er
s
tan
d
w
h
at
w
e
s
a
y
.
Natu
r
al
L
an
g
u
ag
e
Ge
n
er
atio
n
:
T
h
e
g
en
er
atio
n
o
f
n
at
u
r
al
lan
g
u
a
g
e
b
y
a
co
m
p
u
ter
.
Sp
ee
ch
R
ec
o
g
n
it
io
n
: T
h
e
tr
an
s
latio
n
o
f
s
p
o
k
en
la
n
g
u
a
g
e
i
n
t
o
tex
t.
NL
P
h
ad
b
ee
n
w
id
el
y
ap
p
lied
to
d
if
f
er
en
t
ap
p
licatio
n
s
,
s
u
c
h
as
p
r
ed
ictiv
e
t
y
p
in
g
,
s
en
ti
m
en
t
an
al
y
s
is
,
s
p
ell
ch
ec
k
in
g
,
s
p
a
m
d
etec
t
io
n
,
an
d
o
th
er
s
.
2
.
2
.
Senti
m
ent
a
na
ly
s
is
Sen
ti
m
e
n
t
an
al
y
s
is
(
S
A
)
is
als
o
k
n
o
w
n
as
o
p
in
io
n
m
in
i
n
g
[
6
]
.
I
t
r
ef
er
s
to
b
u
ild
a
s
y
s
te
m
to
id
en
tif
y
an
d
ex
tr
ac
t o
p
in
io
n
s
f
r
o
m
th
e
tex
t.
I
t o
f
te
n
ab
le
to
ex
tr
ac
t th
e
attr
ib
u
te
o
f
th
e
e
x
p
r
ess
io
n
,
w
h
ic
h
in
cl
u
d
es:
Su
b
j
ec
t: T
o
p
ic
th
at
is
talk
ed
a
b
o
u
t.
P
o
lar
ity
: P
o
s
itiv
e
o
r
n
e
g
ati
v
e
o
p
in
io
n
.
Op
in
io
n
h
o
ld
er
: E
n
tit
y
th
at
e
x
p
r
ess
es a
n
o
p
in
io
n
.
Sen
ti
m
e
n
t
a
n
al
y
s
is
g
e
n
er
ates
g
r
ea
t
in
ter
e
s
t
m
ain
l
y
d
u
e
t
o
h
av
i
n
g
m
a
n
y
p
r
ac
tical
ap
p
licatio
n
s
.
T
h
e
co
n
s
tan
t
e
x
p
an
s
io
n
o
f
th
e
I
n
ter
n
et
p
r
o
d
u
ce
s
lar
g
e
v
o
lu
m
es
o
f
te
x
t
u
al
d
ata
ex
p
r
ess
i
n
g
o
p
i
n
io
n
s
.
Wi
th
s
en
ti
m
en
t
a
n
al
y
s
i
s
,
th
i
s
m
ess
y
an
d
u
n
s
tr
u
ct
u
r
ed
in
f
o
r
m
atio
n
w
il
l
b
e
au
to
m
atica
l
l
y
tr
an
s
f
o
r
m
ed
i
n
to
s
tr
u
ct
u
r
ed
d
ata.
T
h
ese
d
ata
c
o
n
tain
th
e
p
u
b
lic
o
p
i
n
io
n
ab
o
u
t
s
er
v
ice
s
,
p
r
o
d
u
cts,
b
r
an
d
s
an
d
d
if
f
er
en
t
to
p
ics
th
at
p
eo
p
le
co
u
ld
ex
p
r
ess
o
p
in
io
n
s
ab
o
u
t.
T
h
e
d
ata
is
v
er
y
v
alu
ab
le
f
o
r
co
m
m
er
ci
al
ap
p
licatio
n
s
li
k
e
m
ar
k
et
in
g
a
n
al
y
s
i
s
,
p
r
o
d
u
ct
f
ee
d
b
ac
k
,
cu
s
to
m
er
s
er
v
ices,
p
u
b
lic
r
elatio
n
s
h
ip
,
a
n
d
o
th
er
s
.
I
n
s
h
o
r
t,
t
h
e
S
A
s
y
s
te
m
h
e
lp
s
co
m
p
a
n
ies
to
m
ak
e
s
e
n
s
e
o
f
th
i
s
s
ea
o
f
u
n
s
tr
u
ctu
r
ed
te
x
t
b
y
a
u
to
m
at
in
g
b
u
s
i
n
es
s
p
r
o
ce
s
s
es,
g
etti
n
g
ac
t
io
n
ab
le
in
s
i
g
h
t
s
,
an
d
s
av
in
g
ti
m
e
o
n
m
an
u
al
d
ata
p
r
o
ce
s
s
in
g
.
T
h
is
r
esear
ch
f
o
c
u
s
e
s
s
o
lel
y
o
n
th
e
p
o
lar
ity
e
x
tr
ac
tio
n
a
s
p
ec
t o
f
S
A
.
T
h
er
e
ar
e
s
ev
er
al
w
a
y
s
o
r
ap
p
r
o
ac
h
es
to
i
m
p
le
m
en
t
th
e
s
en
ti
m
en
t
an
al
y
s
is
s
y
s
te
m
.
As
s
h
o
w
n
i
n
Fig
u
r
e
1
,
t
h
ese
ap
p
r
o
ac
h
es
ar
e
o
f
ten
b
e
ca
teg
o
r
ized
in
to
t
w
o
m
aj
o
r
ca
teg
o
r
ies,
w
h
ic
h
a
r
e
m
ac
h
i
n
e
lear
n
i
n
g
ap
p
r
o
ac
h
es a
n
d
lex
ico
n
-
b
ased
ap
p
r
o
ac
h
es [
7
]
.
Ma
ch
i
n
e
L
ea
r
n
in
g
A
p
p
r
o
ac
h
es:
L
ea
r
n
t
h
e
f
ea
t
u
r
es
f
r
o
m
lab
elled
d
ata,
th
en
ap
p
ly
it
to
a
s
et
o
f
u
n
k
n
o
w
n
d
ata.
L
e
x
ico
n
B
ased
A
p
p
r
o
ac
h
es:
W
o
r
k
o
n
th
e
d
ictio
n
ar
y
w
it
h
a
li
s
t
o
f
w
o
r
d
s
/s
e
n
te
n
ce
s
w
it
h
d
i
f
f
er
en
t
p
o
lar
ity
,
o
u
tp
u
t p
o
lar
it
y
w
ill b
e
s
o
m
e
o
f
t
h
e
p
o
lar
ities
o
n
d
if
f
er
en
t
w
o
r
d
s
/
s
en
ten
ce
s
[
8
]
.
Fig
u
r
e
1
.
Sen
t
i
m
e
n
t c
las
s
if
icat
io
n
tech
n
iq
u
e
s
Ma
ch
i
n
e
lear
n
in
g
ap
p
r
o
ac
h
es
r
el
y
o
n
u
s
i
n
g
alg
o
r
it
h
m
s
to
a
u
to
m
at
ic
all
y
e
x
tr
ac
t
t
h
e
f
ea
t
u
r
e
s
f
r
o
m
a
d
ata
s
et.
Af
ter
t
h
e
m
o
d
els
h
a
v
e
b
ee
n
s
u
f
f
icie
n
t
ly
tr
ai
n
ed
,
it
w
i
ll
b
e
f
ed
w
it
h
u
n
k
n
o
w
n
tex
t
an
d
t
h
en
r
etu
r
n
s
w
it
h
t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
2
,
J
u
n
e
20
20
:
2
12
–
2
20
214
co
r
r
esp
o
n
d
in
g
p
o
lar
it
y
(
p
o
s
itiv
e,
n
eg
a
tiv
e,
o
r
n
eu
tr
al)
.
T
h
e
s
en
ti
m
e
n
t
an
al
y
s
is
tas
k
is
u
s
u
all
y
ca
teg
o
r
ize
d
in
NL
P
u
n
d
er
th
e
tex
t c
la
s
s
i
f
ica
ti
o
n
s
eg
m
e
n
t a
n
d
t
h
er
ef
o
r
e
is
a
t
y
p
e
o
f
class
if
ier
.
2
.
3
.
Da
t
a
p
re
-
pro
ce
s
s
ing
Data
p
r
e
-
p
r
o
ce
s
s
i
n
g
r
ef
er
s
to
th
e
s
tep
s
th
at
n
ee
d
to
b
e
e
m
p
lo
y
ed
b
ef
o
r
e
it
ca
n
b
e
tr
an
s
f
o
r
m
ed
i
n
to
th
e
n
u
m
er
ical
f
ea
tu
r
e
t
h
at
ca
n
w
o
r
k
w
it
h
t
h
e
m
ac
h
i
n
e
lea
r
n
in
g
al
g
o
r
ith
m
[
9
]
.
T
h
e
r
a
w
tex
t
co
m
es
w
it
h
a
v
ar
iet
y
o
f
f
o
r
m
s
,
f
o
r
ex
a
m
p
le,
in
d
i
v
id
u
al
w
o
r
d
s
,
to
s
e
n
t
en
ce
s
,
m
u
ltip
le
p
ar
a
g
r
ap
h
s
,
ev
en
w
it
h
s
p
ec
ial
ch
ar
ac
ter
s
an
d
co
n
tai
n
s
a
lo
t o
f
r
ed
u
n
d
a
n
t in
f
o
r
m
a
tio
n
.
A
ll
t
h
e
s
e
r
a
w
te
x
t
s
n
ee
d
to
b
e
clea
n
ed
u
p
an
d
tr
i
m
d
o
w
n
to
th
e
r
o
o
t
f
o
r
m
at
t
h
at
ca
n
w
o
r
k
w
ell
w
it
h
th
e
m
ac
h
i
n
e
lear
n
in
g
al
g
o
r
it
h
m
.
G
o
o
d
tex
t
p
r
e
-
p
r
o
ce
s
s
i
n
g
e
n
h
a
n
ce
s
t
h
e
cla
s
s
i
f
icatio
n
ac
c
u
r
ac
y
a
n
d
a
s
w
ell
as
tr
ain
i
n
g
e
f
f
icie
n
c
y
[
1
0
]
.
T
h
e
d
if
f
er
en
t te
x
t p
r
e
-
p
r
o
ce
s
s
i
n
g
tec
h
n
iq
u
es t
h
at
ar
e
o
f
te
n
u
s
ed
ar
e
s
tated
b
elo
w
:
T
o
k
en
izer
: T
o
cr
ea
te
a
to
k
en
f
o
r
ea
ch
w
o
r
d
in
th
e
s
en
te
n
ce
s
/
d
o
cu
m
en
t.
C
ap
italizatio
n
: C
h
an
g
i
n
g
all
t
h
e
w
o
r
d
s
to
lo
w
er
ca
s
e
o
r
u
p
p
er
ca
s
e.
A
b
b
r
ev
iatio
n
R
ep
lace
m
en
t: E
x
p
an
d
th
e
ab
b
r
ev
iatio
n
to
th
e
o
r
ig
in
al
w
o
r
d
s
.
Sto
p
w
o
r
d
r
em
o
v
al
: Re
m
o
v
e
t
h
e
m
ea
n
i
n
g
less
w
o
r
d
s
.
Fo
r
ex
a
m
p
le:
’
is
’
,
’
t
h
e
’
,
’
ar
e
’
.
Ste
m
m
i
n
g
:
P
r
o
ce
s
s
o
f
r
e
m
o
v
in
g
t
h
e
e
n
d
in
g
s
o
f
t
h
e
w
o
r
d
.
Fo
r
ex
a
m
p
le:
c
h
a
n
g
i
n
g
'
s
tu
d
ies
'
to
'st
u
d
y
',
ch
an
g
i
n
g
'st
u
d
y
i
n
g
'
to
'
s
tu
d
y
'
.
L
e
m
m
atiza
t
io
n
:
P
r
o
ce
s
s
o
f
g
r
o
u
p
in
g
t
h
e
d
if
f
er
e
n
t
i
n
f
lecte
d
f
o
r
m
s
o
f
a
w
o
r
d
.
Fo
r
ex
a
m
p
le:
ch
a
n
g
i
n
g
'
s
tu
d
ie
s
'
an
d
'
s
t
u
d
y
i
n
g
'
to
'st
u
d
y
'.
Sp
ellin
g
C
o
r
r
ec
tio
n
:
R
ec
o
v
er
t
h
e
w
o
r
d
s
f
r
o
m
t
y
p
o
s
.
P
ar
t o
f
s
p
ee
ch
tag
g
in
g
: ta
g
to
k
en
f
o
r
th
e
t
y
p
e
o
f
t
h
e
w
o
r
d
.
Fo
r
ex
a
m
p
le
No
u
n
s
,
V
er
b
s
,
A
d
j
ec
tiv
es.
R
e
m
o
v
e
n
u
m
b
er
s
: Re
m
o
v
e
t
h
e
n
u
m
er
ical
e
x
p
r
ess
io
n
.
Na
m
ed
en
ti
t
y
r
ec
o
g
n
itio
n
:
lab
el
th
e
w
o
r
d
s
w
i
th
r
esp
ec
ti
v
e
e
n
tit
ie
s
.
Fo
r
ex
a
m
p
le,
n
a
m
e,
lo
ca
tio
n
,
d
ates,
ad
d
r
ess
.
No
t
all
p
r
e
-
p
r
o
ce
s
s
in
g
tech
n
i
q
u
es
ar
e
s
u
i
tab
le
f
o
r
ev
er
y
s
ce
n
ar
io
.
T
h
e
tech
n
iq
u
e
s
ar
e
ch
o
s
en
o
n
a
ca
s
e
b
y
ca
s
e
b
asi
s.
2
.
4
.
T
ex
t
t
o
nu
m
er
ica
l r
epre
s
ent
a
t
io
n
T
h
e
co
m
p
u
ter
o
n
l
y
u
n
d
er
s
ta
n
d
s
th
e
d
ata
w
h
ic
h
i
s
a
n
u
m
er
ical
f
o
r
m
at,
s
o
d
o
th
e
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
.
B
e
f
o
r
e
in
p
u
tt
in
g
t
h
e
d
ata
f
o
r
th
e
m
ac
h
i
n
e
lear
n
i
n
g
al
g
o
r
it
h
m
to
p
r
o
ce
s
s
,
w
e
n
ee
d
to
co
n
v
er
t
th
e
d
ata
f
r
o
m
te
x
t
u
al
f
o
r
m
at
to
n
u
m
er
ical
f
o
r
m
at.
T
h
is
p
r
o
ce
s
s
i
s
also
k
n
o
w
n
as
te
x
t
v
ec
to
r
iza
tio
n
.
T
h
e
co
m
m
o
n
tech
n
iq
u
es
f
o
r
tex
t
v
ec
to
r
izati
o
n
ar
e:
B
ag
Of
W
o
r
d
s
(
B
o
W
)
:
B
o
W
is
a
b
asic
an
d
s
tr
aig
h
t
f
o
r
w
ar
d
tech
n
iq
u
e.
W
ith
th
i
s
m
e
th
o
d
,
th
e
o
r
d
er
an
d
g
r
a
m
m
ar
o
f
t
h
e
w
o
r
d
s
ar
e
d
is
ca
r
d
ed
.
I
t
o
n
l
y
tell
s
w
h
e
th
er
a
w
o
r
d
is
p
r
esen
t i
n
t
h
e
d
o
cu
m
e
n
t o
r
n
o
t.
E
ac
h
w
o
r
d
o
f
t
h
e
e
n
tire
d
ata
s
et
i
s
co
r
r
esp
o
n
d
in
g
to
a
co
l
u
m
n
,
if
th
e
p
ar
tic
u
lar
ex
is
t
in
th
e
in
p
u
t
s
e
n
te
n
ce
s
,
v
ec
to
r
r
ep
r
esen
tatio
n
o
f
t
h
i
s
s
e
n
ten
ce
h
as a
1
in
t
h
e
co
r
r
esp
o
n
d
in
g
co
lu
m
n
f
o
r
th
i
s
w
o
r
d
[
1
1
]
.
T
er
m
Fre
q
u
e
n
c
y
–
I
n
v
er
s
e
Do
cu
m
e
n
t
Fre
q
u
e
n
c
y
(
TF
-
I
DF
)
:
E
ac
h
w
o
r
d
w
i
th
i
n
a
n
i
n
p
u
t
s
en
te
n
ce
is
r
ep
lace
d
w
it
h
its
T
F
-
ID
F
s
co
r
e
an
d
cr
ea
ted
a
v
ec
to
r
o
u
t
o
f
th
e
s
e
s
co
r
es
f
o
r
ea
ch
in
p
u
t
s
en
te
n
ce
.
T
h
e
w
h
o
le
id
ea
o
f
th
i
s
m
ea
s
u
r
e
is
to
g
iv
e
m
o
r
e
i
m
p
o
r
tan
ce
to
t
h
e
ter
m
s
th
at
ar
e
m
o
r
e
s
p
ec
if
i
c
to
a
ce
r
tain
class
(
T
F)
an
d
r
ed
u
ce
th
e
i
m
p
o
r
tan
ce
o
f
th
e
ter
m
th
at
i
s
v
er
y
f
r
eq
u
en
t i
n
t
h
e
en
t
ir
e
co
r
p
u
s
(
I
DF)
[
1
2
]
.
W
o
r
d
E
m
b
ed
d
in
g
s
:
A
w
a
y
o
f
s
tatis
t
icall
y
e
x
tr
ac
ti
n
g
t
h
e
m
ea
n
i
n
g
o
f
a
w
o
r
d
f
r
o
m
th
e
tex
t
an
d
r
ep
r
esen
tin
g
it
w
it
h
a
s
et
o
f
n
u
m
b
er
s
.
T
o
b
e
ab
le
to
u
n
d
er
s
tan
d
th
e
m
ea
n
i
n
g
o
f
a
w
o
r
d
,
th
e
s
e
m
o
d
els
u
s
e
co
n
tex
t
u
al
s
i
m
ilar
itie
s
.
W
it
h
th
i
s
te
c
h
n
iq
u
e,
w
o
r
d
s
w
it
h
s
i
m
ilar
m
ea
n
i
n
g
s
te
n
d
to
h
av
e
a
s
i
m
ilar
r
ep
r
esen
tatio
n
.
T
h
e
w
o
r
d
2
v
ec
m
et
h
o
d
is
co
m
m
o
n
l
y
u
s
ed
f
o
r
w
o
r
d
e
m
b
ed
d
in
g
s
[
1
3
]
.
2
.
5
.
Cla
s
s
if
ica
t
io
n
a
lg
o
rit
h
m
s
Ma
n
y
cla
s
s
i
f
icatio
n
alg
o
r
it
h
m
s
ex
is
t
i
n
m
ac
h
i
n
e
lear
n
i
n
g
.
F
o
r
s
en
ti
m
e
n
t
an
a
l
y
s
is
,
s
tat
is
tic
al
m
o
d
el
s
lik
e
S
u
p
p
o
r
t V
ec
to
r
Ma
ch
in
e
(
SVM)
,
Na
ï
v
e
B
a
y
e
s
(
NB
)
o
r
n
eu
r
al
n
et
w
o
r
k
s
ar
e
co
m
m
o
n
l
y
u
s
ed
.
Naïv
e
B
a
y
es
(
NB
)
:
Use
B
ay
e
s
’
s
T
h
eo
r
em
(
p
r
o
b
ab
ilis
tic
alg
o
r
it
h
m
s
)
to
p
r
ed
ict
th
e
ca
teg
o
r
y
o
f
a
tex
t
[
1
4
]
.
Su
p
p
o
r
t
Vec
to
r
Ma
ch
i
n
e
(
SV
M
)
:
T
ex
t
is
r
ep
r
esen
ted
a
s
a
p
o
in
t
i
n
a
m
u
ltid
i
m
en
s
io
n
al
s
p
a
ce
.
T
h
e
p
o
in
ts
w
il
l
b
e
m
ap
p
ed
to
d
if
f
er
en
t
ca
teg
o
r
ies.
Ne
w
te
x
t
is
m
ap
p
ed
o
n
to
th
e
s
a
m
e
s
p
ac
e
an
d
p
r
ed
icted
to
b
elo
n
g
to
ca
teg
o
r
ies b
ased
o
n
th
e
r
eg
i
o
n
th
at
lie
s
[
1
5
]
.
A
r
ti
f
icial
Neu
r
al
Net
w
o
r
k
s
(
ANN)
an
d
Dee
p
L
ea
r
n
in
g
(
D
L
)
:
Usi
n
g
a
d
iv
er
s
e
s
et
o
f
al
g
o
r
it
h
m
s
to
i
m
itate
h
o
w
th
e
ac
t
u
al
h
u
m
a
n
b
r
ain
wo
r
k
s
b
y
e
m
p
lo
y
in
g
ar
ti
f
icial
n
eu
r
al
n
et
w
o
r
k
s
to
p
r
o
ce
s
s
d
ata
[
1
6
]
.
C
o
n
v
o
lu
tio
n
al
Ne
u
r
al
Ne
t
w
o
r
k
(
C
NN)
s
u
ch
as
s
h
o
w
n
i
n
F
ig
u
r
e
2
i
s
a
t
y
p
e
o
f
ANN
t
h
at
i
s
co
m
m
o
n
l
y
u
s
e
d
in
i
m
a
g
e
r
ec
o
g
n
itio
n
.
I
n
it
iall
y
,
it
w
a
s
d
esi
g
n
ed
to
p
r
o
ce
s
s
t
h
e
p
ix
e
l
d
ata,
b
u
t
r
ec
en
tl
y
C
NN
also
h
as b
ee
n
ap
p
lied
to
tex
t c
l
ass
i
f
icatio
n
tas
k
s
a
n
d
ac
h
ie
v
e
v
er
y
g
o
o
d
p
er
f
o
r
m
a
n
ce
[
1
7
-
1
9
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
S
en
timen
t a
n
a
lysi
s
o
f in
fo
r
ma
l Ma
la
y
tw
ee
ts
w
i
th
d
ee
p
lea
r
n
in
g
(
On
g
Ju
n
Yin
g
)
215
T
h
e
ar
ch
itectu
r
e
o
f
C
N
N
p
r
ett
y
m
u
c
h
li
k
e
a
m
u
lti
-
la
y
er
p
er
ce
p
tr
o
n
th
at
h
a
s
b
ee
n
d
esig
n
ed
f
o
r
r
ed
u
ce
d
p
r
o
ce
s
s
in
g
r
eq
u
ir
e
m
e
n
ts
.
I
n
g
e
n
er
al,
C
N
N
w
i
ll
co
n
s
is
t
o
f
a
n
in
p
u
t
la
y
er
,
a
h
id
d
en
la
y
er
,
an
d
o
u
tp
u
t
la
y
er
s
.
T
h
e
h
id
d
en
la
y
er
in
cl
u
d
es
m
u
ltip
le
co
n
v
o
lu
tio
n
al
la
y
er
s
,
n
o
r
m
a
lizatio
n
la
y
er
,
p
o
o
lin
g
la
y
er
a
n
d
f
u
ll
y
co
n
n
ec
ted
l
a
y
er
[
16
]
.
C
NN
h
as
f
o
r
w
ar
d
an
d
r
ev
er
s
e
tr
an
s
m
is
s
io
n
.
I
n
f
o
r
w
ar
d
tr
an
s
m
i
s
s
io
n
,
th
e
d
ata
in
p
u
t
s
tr
u
ct
u
r
es
u
n
d
er
g
o
m
u
ltip
le
l
a
y
er
s
o
f
p
r
o
ce
s
s
es
an
d
o
u
tp
u
t
s
tr
u
c
tu
r
es
i
n
th
e
o
u
tp
u
t
la
y
er
,
an
ac
tiv
atio
n
f
u
n
ctio
n
is
n
ee
d
ed
in
e
v
er
y
la
y
er
.
I
n
r
ev
er
s
e
tr
a
n
s
m
i
s
s
io
n
,
t
h
e
p
r
o
b
ab
ilit
y
o
f
er
r
o
r
is
ca
lcu
lated
u
s
i
n
g
a
g
i
v
e
n
m
o
d
el
r
es
u
lt
an
d
th
e
f
o
r
w
ar
d
tr
an
s
m
is
s
io
n
r
es
u
lt,
an
d
its
tr
an
s
f
er
to
th
e
er
r
o
r
f
u
n
ctio
n
b
ac
k
to
ea
ch
r
esp
ec
tiv
e
la
y
er
,
last
l
y
th
e
g
r
ad
ie
n
t
d
es
ce
n
t
tec
h
n
iq
u
e
is
ap
p
lied
to
t
u
n
e
th
e
b
ias
p
ar
a
m
eter
a
n
d
n
et
w
o
r
k
w
ei
g
h
ts
to
o
b
tain
b
etter
ac
cu
r
ac
y
.
Fig
u
r
e
2
.
C
o
n
v
o
l
u
tio
n
al
Neu
r
a
l N
et
w
o
r
k
[
1
6
]
.
T
r
ain
in
g
o
f
C
NN
is
d
o
n
e
i
n
t
w
o
s
tag
e
s
w
h
ic
h
ar
e
th
e
tr
ain
in
g
s
ta
g
e
a
n
d
test
i
n
g
s
tag
e.
D
u
r
in
g
t
h
e
tr
ain
i
n
g
s
tag
e,
a
s
et
o
f
lab
elle
d
d
ata
n
ee
d
s
to
b
e
p
r
o
v
id
ed
.
T
h
is
s
tag
e
w
il
l
in
v
o
lv
e
t
h
e
co
n
v
o
l
u
tio
n
p
r
o
ce
s
s
,
s
u
b
s
a
m
p
li
n
g
a
n
d
m
u
ltip
l
y
i
n
g
th
e
en
d
o
u
tco
m
e
o
f
t
h
e
s
u
b
s
a
m
p
li
n
g
w
i
th
ce
r
tai
n
ar
tifi
cial
n
eu
r
al
w
e
ig
h
ts
d
u
r
in
g
th
e
tr
ai
n
i
n
g
p
r
o
ce
s
s
.
W
h
ile
t
h
e
test
i
n
g
s
ta
g
e
is
to
ap
p
ly
in
p
u
t d
ata
an
d
test
o
n
t
h
e
tr
a
in
ed
m
o
d
el
[
1
0
]
.
3.
M
E
T
H
O
DO
L
O
G
Y
Fig
u
r
e
3
s
h
o
w
s
t
h
e
m
e
th
o
d
o
lo
g
y
u
s
ed
f
o
r
th
i
s
w
o
r
k
.
Firs
t,
th
e
d
ataset
is
p
r
e
p
ar
ed
.
I
t
th
en
g
o
e
s
th
r
o
u
g
h
p
r
e
-
p
r
o
ce
s
s
i
n
g
,
lab
ellin
g
an
d
co
n
v
er
s
io
n
to
n
u
m
er
ical
r
ep
r
esen
ta
tio
n
.
Af
ter
th
e
s
e
s
tep
s
ar
e
d
o
n
e
th
e
d
ata
is
r
ea
d
y
u
s
e
to
tr
ain
an
d
test
o
n
th
e
C
NN
m
o
d
el.
L
ast
l
y
,
t
h
e
p
er
f
o
r
m
an
c
e
o
f
th
e
s
e
n
ti
m
en
t
an
al
y
s
is
ta
s
k
is
an
a
l
y
ze
d
.
Fig
u
r
e
3
.
Data
p
r
o
ce
s
s
in
g
3
.
1
.
P
r
o
po
s
ed
CNN
a
rc
hite
ct
ure
T
h
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
s
h
o
w
n
i
n
Fi
g
u
r
e
4
an
d
T
ab
le
1
is
th
e
C
NN
ar
ch
itectu
r
e
ad
o
p
ted
f
r
o
m
[
2
0
]
w
it
h
t
h
e
ad
d
itio
n
o
f
a
d
r
o
p
o
u
t la
y
er
to
r
ed
u
ce
th
e
o
v
er
f
itti
n
g
p
r
o
b
lem
.
1.
E
ac
h
w
o
r
d
is
m
ap
p
ed
to
a
w
o
r
d
v
ec
to
r
r
ep
r
esen
tatio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
2
,
J
u
n
e
20
20
:
2
12
–
2
20
216
T
h
e
en
tire
t
w
ee
t
ca
n
b
e
m
ap
p
ed
to
a
m
atr
ix
o
f
s
ize
s
×
d
,
w
h
er
e
s
=
n
u
m
b
er
o
f
w
o
r
d
s
an
d
d
=
d
i
m
en
s
io
n
o
f
t
h
e
e
m
b
ed
d
i
n
g
s
p
ac
e
2.
Z
er
o
p
ad
d
in
g
T
o
m
a
k
e
s
u
r
e
all
t
w
ee
ts
h
a
v
e
t
h
e
s
a
m
e
m
a
tr
ix
d
i
m
en
s
io
n
.
3.
Ma
x
p
o
o
lin
g
la
y
er
E
x
tr
ac
ts
t
h
e
m
o
s
t i
m
p
o
r
tan
t
f
e
atu
r
e
f
o
r
ea
ch
co
n
v
o
lu
tio
n
.
C
o
m
b
i
n
e
all
th
e
c
m
a
x
o
f
ea
ch
f
ilter
i
n
to
o
n
e
v
ec
to
r
.
4.
Dr
o
p
o
u
t la
y
er
E
x
p
licitl
y
alter
in
g
t
h
e
n
et
w
o
r
k
ar
ch
itectu
r
e
at
tr
ai
n
in
g
ti
m
e
5.
So
f
t
m
ax
la
y
er
G
iv
e
o
u
t
f
i
n
al
clas
s
i
f
icatio
n
p
r
o
b
ab
ilit
ies
Fig
u
r
e
4
.
P
r
o
p
o
s
ed
C
NN
ar
ch
itectu
r
e
T
ab
le
1
.
B
aselin
e
C
N
N
co
n
f
ig
u
r
atio
n
.
P
a
r
a
me
t
e
r
V
a
l
u
e
F
i
l
t
e
r
R
e
g
i
o
n
2
,
3
,
4
F
e
a
t
u
r
e
M
a
p
1
0
0
P
o
o
l
i
n
g
L
a
y
e
r
M
a
x
P
o
o
l
i
n
g
A
c
t
i
v
a
t
i
o
n
F
u
n
c
t
i
o
n
R
e
L
u
D
r
o
p
O
u
t
L
a
y
e
r
R
a
t
e
0
.
5
3
.
2
.
Da
t
a
p
re
-
pro
ce
s
s
ing
T
h
e
r
a
w
tex
t
co
m
e
s
in
a
v
ar
iet
y
o
f
f
o
r
m
s
,
f
o
r
ex
a
m
p
le,
i
n
d
iv
id
u
al
w
o
r
d
s
,
to
s
en
ten
ce
s
,
m
u
ltip
l
e
p
ar
ag
r
ap
h
s
,
ev
en
w
i
th
s
p
ec
ial
ch
ar
ac
ter
s
an
d
co
n
tai
n
s
a
lo
t
o
f
r
ed
u
n
d
an
t
i
n
f
o
r
m
atio
n
.
I
n
o
u
r
ca
s
e,
th
e
d
ata
is
r
ef
er
r
ed
to
as
r
a
w
t
w
ee
ts
.
All
t
h
ese
r
a
w
te
x
ts
n
ee
d
to
b
e
clea
n
ed
u
p
to
f
it
w
i
th
t
h
e
m
ac
h
in
e
lear
n
in
g
al
g
o
r
ith
m
w
ell.
T
h
e
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
s
tep
s
p
r
o
p
o
s
ed
ar
e
as f
o
llo
w
s
:
1.
Sp
l
it
h
as
h
ta
g
to
w
o
r
d
s
2.
C
ap
italizatio
n
3.
R
e
m
o
v
e
n
u
m
b
er
,
h
t
m
l ta
g
,
u
r
l
4.
Ste
m
m
i
n
g
,
l
e
m
m
atiz
in
g
(
n
o
r
m
aliza
tio
n
)
5.
Sto
p
w
o
r
d
r
em
o
v
al
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
S
en
timen
t a
n
a
lysi
s
o
f in
fo
r
ma
l Ma
la
y
tw
ee
ts
w
i
th
d
ee
p
lea
r
n
in
g
(
On
g
Ju
n
Yin
g
)
217
6.
T
o
k
en
izatio
n
7.
Data
a
n
n
o
tat
io
n
/lab
elli
n
g
(
p
o
s
i
tiv
e
,
n
e
g
ati
v
e
,
n
e
u
tr
al
)
3
.
3
.
T
ex
t
t
o
nu
m
er
ica
l r
epre
s
ent
a
t
io
n
B
ef
o
r
e
in
p
u
t,
t
h
e
d
ata
f
o
r
t
h
e
m
ac
h
in
e
lear
n
i
n
g
al
g
o
r
ith
m
to
p
r
o
ce
s
s
,
th
e
d
ata
i
n
th
e
tex
tu
a
l
f
o
r
m
at
i
s
co
n
v
er
ted
to
n
u
m
er
ical
f
o
r
m
at
u
s
i
n
g
W
o
r
d
2
Vec
.
W
o
r
d
2
V
e
c
is
a
t
w
o
-
la
y
er
n
eu
r
al
n
et
t
h
at
p
r
o
ce
s
s
es
te
x
t.
I
ts
in
p
u
t
is
a
tex
t
co
r
p
u
s
an
d
its
o
u
tp
u
t
is
a
s
et
o
f
v
ec
t
o
r
s
:
f
ea
tu
r
e
v
ec
to
r
s
f
o
r
w
o
r
d
s
in
th
at
co
r
p
u
s
.
W
h
ile
W
o
r
d
2
v
e
c
is
n
o
t
a
d
ee
p
n
eu
r
al
n
et
w
o
r
k
,
it
t
u
r
n
s
te
x
t
i
n
to
a
n
u
m
er
ical
f
o
r
m
th
at
d
ee
p
n
et
s
ca
n
u
n
d
er
s
ta
n
d
.
T
h
e
p
u
r
p
o
s
e
an
d
u
s
ef
u
l
n
e
s
s
o
f
W
o
r
d
2
v
e
c
a
r
e
to
g
r
o
u
p
t
h
e
v
ec
to
r
s
o
f
s
i
m
ilar
w
o
r
d
s
i
n
v
ec
to
r
s
p
ac
e.
T
h
at
is
,
it
d
etec
ts
s
i
m
i
l
ar
ities
m
at
h
e
m
atica
ll
y
.
W
o
r
d
2
v
e
c
cr
ea
tes
v
ec
to
r
s
th
at
ar
e
d
is
tr
ib
u
ted
n
u
m
er
ical
r
ep
r
esen
tatio
n
s
o
f
w
o
r
d
f
ea
tu
r
es,
f
ea
t
u
r
es s
u
ch
a
s
th
e
co
n
tex
t o
f
in
d
i
v
id
u
al
w
o
r
d
s
.
3
.
4
.
T
ra
ini
ng
s
t
ra
t
eg
y
Af
ter
t
h
e
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
s
tep
an
d
m
o
d
el
ar
e
d
ef
in
ed
,
t
h
e
n
e
x
t
s
tep
to
d
eter
m
i
n
e
is
t
h
e
tr
ain
i
n
g
s
tr
ateg
ie
s
to
tr
ain
t
h
e
m
o
d
el.
A
Gr
ap
h
ics
P
r
o
ce
s
s
in
g
U
n
it
(
GP
U)
is
ch
o
s
en
to
tr
ain
t
h
e
m
o
d
el.
Usi
n
g
a
GP
U
s
av
e
s
s
u
b
s
ta
n
tial
a
m
o
u
n
ts
o
f
t
i
m
e
s
i
n
ce
a
GP
U
ca
n
s
u
p
p
o
r
t
a
h
i
g
h
er
n
u
m
b
er
o
f
p
ar
allel
ta
s
k
s
co
m
p
ar
e
to
th
e
C
en
tr
al
P
r
o
ce
s
s
i
n
g
U
n
it
(
C
P
U)
d
u
e
to
th
e
h
i
g
h
er
n
u
m
b
er
o
f
co
r
es.
A
t
t
h
e
s
o
f
t
w
ar
e
le
v
el,
Ker
as
p
ac
k
ag
e
i
m
p
le
m
en
t
s
t
h
e
tr
ain
i
n
g
p
r
o
ce
d
u
r
e
[
2
1
]
Nex
t,
an
o
t
h
er
i
m
p
o
r
tan
t
ite
m
o
n
d
ee
p
lear
n
in
g
is
t
h
e
d
ata
s
et
u
s
ed
to
tr
ain
t
h
e
m
o
d
el.
I
n
th
is
s
t
u
d
y
,
t
w
o
d
if
f
er
en
t
d
ata
s
ets
w
er
e
u
s
ed
s
ep
ar
atel
y
to
tr
ain
th
e
E
n
g
li
s
h
S
A
m
o
d
el
an
d
Ma
la
y
'
s
S
A
m
o
d
el.
Fo
r
th
e
E
n
g
l
is
h
S
A
m
o
d
el,
th
e
s
en
ti
m
e
n
t1
4
0
d
ata
s
et
w
as
u
s
ed
[
22
]
.
T
h
is
d
ata
s
et
co
n
tai
n
s
ab
o
u
t
1
6
m
illi
o
n
E
n
g
l
is
h
t
w
ee
t
s
w
i
th
p
o
s
iti
v
e
an
d
n
e
g
ati
v
e
lab
el
s
.
I
t
i
s
w
id
el
y
u
s
ed
i
n
r
e
s
ea
r
ch
p
u
r
p
o
s
es
.
Fo
r
th
e
Ma
la
y
S
A
m
o
d
el,
t
h
e
M
AL
A
Y
A
d
ata
s
et
w
a
s
u
s
ed
[
23
]
.
I
t
co
n
tain
s
ab
o
u
t
6
m
i
llio
n
Ma
la
y
T
w
ee
t
s
w
it
h
p
o
s
itiv
e
a
n
d
n
eg
at
iv
e
lab
el
s
.
Fo
r
ea
ch
o
f
th
e
d
ata
s
et,
th
e
tr
ain
i
n
g
i
s
co
n
f
i
g
u
r
ed
to
s
et
8
0
%
o
f
th
e
d
ata
to
b
ec
o
m
e
th
e
tr
ain
i
n
g
d
ata
w
h
ile
t
h
e
2
0
%
th
at
lef
t
w
ill
b
e
s
ep
ar
ate
h
alf
b
ec
o
m
e
v
ali
d
ate
d
ata
an
d
an
o
th
er
h
alf
b
ec
o
m
e
te
s
ti
n
g
d
ata.
B
esid
es,
th
e
tr
ain
i
n
g
is
co
n
f
i
g
u
r
ed
to
u
s
e
1
0
0
ep
o
ch
s
.
4.
RE
SU
L
T
S
A
ND
AN
AL
Y
SI
S
4
.
1
.
T
ex
t
t
o
nu
m
er
ica
l r
epre
s
ent
a
t
io
n r
esu
lt
T
h
is
s
ec
tio
n
r
ep
o
r
ts
t
h
e
r
es
u
lt
s
f
r
o
m
t
h
e
W
o
r
d
2
Vec
n
u
m
er
ic
co
n
v
er
s
io
n
.
A
f
ter
th
e
W
o
r
d
2
Vec
m
o
d
e
l
is
tr
ai
n
ed
,
th
e
w
ei
g
h
t
o
f
t
h
e
w
o
r
d
s
(
n
u
m
er
ic
f
o
r
m
at)
w
ill
b
e
f
i
x
ed
an
d
w
o
u
ld
n
o
t
b
e
c
h
an
g
ed
.
T
h
e
tr
ai
n
ed
W
o
r
d
2
Vec
m
o
d
el
b
ec
o
m
es
o
n
e
o
f
t
h
e
la
y
er
s
i
n
t
h
e
f
i
n
al
C
NN
s
en
t
i
m
e
n
t
a
n
al
y
s
i
s
m
o
d
el
.
T
h
e
m
ai
n
p
u
r
p
o
s
e
an
d
u
s
e
f
u
ln
e
s
s
o
f
W
o
r
d
2
v
e
c
a
r
e
to
g
r
o
u
p
th
e
v
ec
to
r
s
o
f
s
im
i
lar
w
o
r
d
s
in
v
ec
to
r
s
p
ac
e.
A
s
s
h
o
w
n
i
n
Fi
g
u
r
e
5
,
W
o
r
d
2
Vec
s
h
o
w
s
th
e
m
at
h
e
m
atica
l si
m
ilar
itie
s
f
o
r
th
e
to
p
2
0
m
o
s
t si
m
ilar
ter
m
s
f
o
r
d
if
f
er
en
t in
p
u
t
w
o
r
d
s
.
Fig
u
r
e
5
.
W
o
r
d
2
Vec
s
i
m
ilar
iti
es r
esu
lt
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
2
,
J
u
n
e
20
20
:
2
12
–
2
20
218
4
.
2
.
CNN
m
o
del hy
per
-
pa
ra
m
et
er
t
un
ing
T
h
e
h
y
p
er
-
p
ar
a
m
eter
s
ar
e
t
h
e
s
etti
n
g
s
ca
n
b
e
tu
n
ed
t
o
o
p
tim
ize
t
h
e
d
ee
p
lear
n
in
g
m
o
d
el.
I
n
th
i
s
p
r
o
j
ec
t
th
er
e
ar
e
f
e
w
h
y
p
er
-
p
ar
a
m
eter
h
ad
b
ee
n
d
eter
m
i
n
ed
an
d
tu
n
ed
.
T
h
ese
ar
e
th
e
f
ilter
r
eg
io
n
s
ize,
f
ea
t
u
r
e
m
ap
s
a
n
d
d
r
o
p
o
u
t
lay
er
r
ate.
T
ab
les
2
-
6
r
ep
o
r
ts
th
e
v
ar
io
u
s
ad
j
u
s
t
m
e
n
ts
m
ad
e
to
th
e
ar
ch
itec
tu
r
e
d
u
r
in
g
tr
ain
i
n
g
.
T
ab
le
2
.
R
esu
lts
o
f
d
if
f
er
en
ce
s
in
g
le
f
ilter
r
eg
io
n
s
ize
s
I
t
e
r
a
t
i
o
n
R
e
g
i
o
n
S
i
z
e
A
c
c
u
r
a
c
y
T1
1
76.
00
%
T2
3
77.
30
%
T3
5
76.
93
%
T4
7
76.
52
%
T5
10
76.
42
%
T6
15
76.
68
%
T7
20
76.
91
%
T8
30
76.
58
%
T
ab
le
3
.
R
esu
lts
o
f
d
if
f
er
en
t
m
u
lti
-
f
ilter
r
eg
io
n
s
ize
s
I
t
e
r
a
t
i
o
n
M
u
l
t
i
-
R
e
g
i
o
n
S
i
z
e
A
c
c
u
r
a
c
y
T1
2
,
3
,
4
76.
56
%
T2
3
,
4
,
5
76.
85
%
T3
4
,
5
,
6
76.
74
%
T4
5
,
6
,
7
76.
09
%
T5
7
,
8
,
9
76.
59
%
T6
3
,
4
,
5
,
6
76.
85
%
T7
6
,
7
,
8
,
9
76.
80
%
T8
3
,
3
,
3
77.
37
%
T9
3
,
3
,
3
,
3
76.
48
%
T
ab
le
4
.
R
esu
lts
o
f
d
i
f
f
er
en
ce
f
ea
tu
r
e
m
ap
s
I
t
e
r
a
t
i
o
n
F
e
a
t
u
r
e
M
a
p
s
A
c
c
u
r
a
c
y
T1
10
7
6
.
3
5
%
T2
50
7
7
.
1
2
%
T3
1
0
0
7
7
.
3
7
%
T4
2
0
0
7
7
.
5
9
%
T5
4
0
0
7
7
.
1
5
%
T6
6
0
0
7
7
.
5
6
%
T7
1
0
0
0
7
7
.
2
6
%
T8
2
0
0
0
7
7
.
3
7
%
T
ab
le
5
.
R
esu
lt o
f
d
i
f
f
er
e
n
t d
r
o
p
o
u
t r
ate
I
t
e
r
a
t
i
o
n
R
e
g
i
o
n
S
i
z
e
A
c
c
u
r
a
c
y
T1
0
.
1
7
6
.
1
7
%
T2
0
.
2
7
7
.
3
7
%
T3
0
.
3
7
7
.
1
7
%
T4
0
.
4
7
7
.
3
4
%
T5
0
.
5
7
7
.
5
9
%
T6
0
.
6
7
7
.
4
0
%
T7
0
.
7
7
6
.
9
9
%
T8
0
.
8
7
6
.
9
1
%
T8
0
.
9
7
7
.
4
4
%
Af
ter
g
o
in
g
t
h
r
o
u
g
h
m
an
y
ite
r
atio
n
s
o
f
t
h
e
h
y
p
er
-
p
ar
a
m
ete
r
tu
n
i
n
g
.
T
h
e
b
est
C
NN
m
o
d
e
l
co
n
f
i
g
u
r
atio
n
t
h
at
o
b
tain
ed
ar
e
s
h
o
w
n
i
n
T
ab
le
6
.
T
ab
le
6
.
B
est co
n
f
i
g
u
r
at
io
n
af
t
er
h
y
p
er
-
p
ar
a
m
eter
t
u
n
in
g
F
e
a
t
u
r
e
P
a
r
a
me
t
e
r
F
i
l
t
e
r
R
e
g
i
o
n
3
,
3
,
3
F
e
a
t
u
r
e
M
a
p
2
0
0
P
o
o
l
i
n
g
L
a
y
e
r
M
a
x
P
o
o
l
i
n
g
A
c
t
i
v
a
t
i
o
n
F
u
n
c
t
i
o
n
R
e
L
u
D
r
o
p
O
u
t
L
a
y
e
r
R
a
t
e
0
.
5
4
.
3
.
M
o
del per
f
o
r
m
a
nce
a
nd
a
cc
ura
cy
Af
ter
t
u
n
i
n
g
all
t
h
e
p
ar
a
m
ete
r
s
,
th
e
E
n
g
li
s
h
an
d
Ma
la
y
Se
n
ti
m
en
t
An
al
y
s
i
s
C
NN
m
o
d
el
s
w
er
e
b
e
tr
ain
ed
an
d
t
h
e
f
in
al
ac
c
u
r
ac
y
r
es
u
lts
w
er
e
o
b
tai
n
ed
.
T
h
e
r
esu
lt
s
o
f
o
u
r
m
o
d
els
a
g
ain
s
t
o
th
er
m
et
h
o
d
s
ar
e
lis
ted
in
T
ab
le
7
.
T
ab
le
7
.
A
cc
u
r
ac
y
co
m
p
ar
is
o
n
w
it
h
p
u
b
li
s
h
ed
w
o
r
k
s
.
M
o
d
e
l
A
c
c
u
r
a
c
y
S
V
M
[
2
5
]
7
5
.
8
6
%
N
a
i
v
e
B
a
y
e
s
[
2
5
]
7
7
.
4
5
%
I
n
d
o
n
e
si
a
n
C
N
N
w
i
t
h
o
u
t
N
o
r
mal
i
z
e
r
[
2
5
]
6
9
.
9
2
%
I
n
d
o
n
e
si
a
n
C
N
N
w
i
t
h
N
o
r
mal
i
z
e
r
[
2
5
]
6
5
.
4
5
%
I
n
d
o
n
e
si
a
n
L
S
T
M
w
i
t
h
N
o
r
mal
i
z
e
r
[
2
5
]
7
3
.
2
2
%
T
h
i
s w
o
r
k
En
g
l
i
s
h
SA
C
N
N
8
1
.
8
7
%
T
h
i
s w
o
r
k
M
a
l
a
y
SA
C
N
N
7
7
.
5
9
%
V
er
y
li
m
i
ted
w
o
r
k
o
n
Ma
la
y
s
en
ti
m
e
n
t
a
n
al
y
s
is
h
a
s
b
ee
n
d
o
n
e
u
s
i
n
g
d
ee
p
lear
n
i
n
g
ap
p
r
o
ac
h
es.
T
h
er
ef
o
r
e,
th
e
co
m
p
ar
is
o
n
ca
n
o
n
l
y
b
e
d
o
n
e
ag
ai
n
s
t
a
s
i
m
ilar
w
o
r
k
in
B
ah
asa
I
n
d
o
n
es
ia
b
y
[
24
]
.
T
h
er
e
is
an
o
th
er
w
o
r
k
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
S
en
timen
t a
n
a
lysi
s
o
f in
fo
r
ma
l Ma
la
y
tw
ee
ts
w
i
th
d
ee
p
lea
r
n
in
g
(
On
g
Ju
n
Yin
g
)
219
d
o
n
e
in
B
ah
a
s
a
I
n
d
o
n
es
ia
w
h
i
ch
ac
h
iev
ed
1
0
0
%
b
u
t
t
h
e
d
at
aset
w
as
to
o
s
m
all
(
le
s
s
t
h
a
n
3
0
0
s
am
p
le
s
)
[
2
5
]
.
A
lt
h
o
u
g
h
t
h
er
e
ar
e
s
o
m
e
li
m
i
tatio
n
s
w
er
e
d
eter
m
i
n
ed
in
th
i
s
co
m
p
ar
is
o
n
,
b
u
t
it
ca
n
s
er
v
e
as
an
in
d
icato
r
to
p
r
o
v
id
e
a
g
en
er
al
v
ie
w
o
n
t
h
e
cu
r
r
en
t sta
n
d
i
n
g
o
f
t
h
i
s
p
r
o
j
ec
t
.
T
h
e
ac
cu
r
ac
y
o
b
tain
ed
f
r
o
m
o
u
r
m
o
d
el
s
i
s
co
m
p
ar
ab
le
a
n
d
s
lig
h
tl
y
b
et
ter
t
h
an
[
2
4
]
.
Fro
m
th
e
tab
le
,
SVM
a
n
d
Naï
v
e
B
a
y
e
s
ar
e
t
h
e
m
ac
h
i
n
e
lear
n
in
g
ap
p
r
o
ac
h
es
w
it
h
ac
c
u
r
ac
ies
clo
s
e
to
th
e
r
esu
l
ts
o
b
tai
n
ed
i
n
th
is
w
o
r
k
.
Ho
w
ev
er
,
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
es
ha
v
e
a
s
ig
n
i
f
ica
n
t
ad
v
a
n
ta
g
e
d
u
e
to
its
d
ata
-
d
r
i
v
en
ch
ar
ac
ter
is
tic
m
ea
n
in
g
t
h
at
t
h
e
m
o
r
e
d
ata
is
f
ed
t
h
e
h
ig
h
er
its
ac
cu
r
ac
y
.
O
n
th
e
o
t
h
er
h
an
d
,
o
th
er
m
ac
h
i
n
e
lear
n
in
g
al
g
o
r
ith
m
s
w
i
ll h
ar
d
l
y
i
m
p
r
o
v
e
t
h
eir
ac
cu
r
ac
y
af
ter
r
ea
ch
in
g
a
ce
r
tain
v
alu
e
.
5.
CO
NCLU
SI
O
N
I
n
co
n
cl
u
s
io
n
,
t
h
e
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
h
a
s
b
ee
n
s
u
c
ce
s
s
f
u
l
l
y
ap
p
lied
in
t
h
e
ta
s
k
o
f
Ma
la
y
s
en
ti
m
e
n
t
a
n
al
y
s
i
s
.
T
h
e
d
ev
elo
p
ed
s
en
ti
m
e
n
t
a
n
al
y
s
i
s
m
o
d
el
w
as
ab
le
to
class
i
f
y
t
h
e
t
w
ee
t
tex
t
in
to
t
w
o
s
en
ti
m
e
n
t
ca
teg
o
r
ies
w
h
ich
ar
e
p
o
s
itiv
e
an
d
n
e
g
ati
v
e.
T
h
e
d
ee
p
lear
n
in
g
ar
c
h
itect
u
r
e
t
h
a
t
w
as
u
s
ed
to
b
u
i
ld
th
e
Ma
la
y
Se
n
ti
m
en
t
A
n
a
l
y
s
i
s
m
o
d
el
w
as
b
ased
o
n
t
h
e
C
o
n
v
o
l
u
tio
n
a
l
Neu
r
al
Ne
t
w
o
r
k
(
C
NN)
.
T
o
f
ee
d
th
e
tex
t
d
ata
to
t
h
e
C
o
n
v
o
lu
t
io
n
Neu
r
al
Net
w
o
r
k
,
th
e
W
o
r
d
2
V
ec
w
o
r
d
e
m
b
ed
d
in
g
m
o
d
el
w
a
s
b
u
ilt
f
r
o
m
s
cr
atc
h
to
co
n
v
er
t
th
e
tex
t
i
n
p
u
t
to
n
u
m
er
ical
r
ep
r
esen
tatio
n
.
T
h
e
m
o
d
el
h
as
a
ls
o
b
ee
n
g
o
i
n
g
th
r
o
u
g
h
t
h
e
h
y
p
er
-
p
ar
am
eter
t
u
n
in
g
to
ac
h
iev
e
i
ts
o
p
ti
m
al
p
er
f
o
r
m
an
c
e.
T
h
e
Ma
la
y
s
e
n
ti
m
e
n
t
a
n
al
y
s
is
C
N
N
m
o
d
el
h
a
s
b
ee
n
s
u
cc
e
s
s
f
u
ll
y
b
u
ilt an
d
b
ee
n
v
al
id
ated
to
ac
h
iev
ed
ac
cu
r
ac
y
u
p
to
7
7
.
5
9
%.
T
h
er
e
ar
e
s
o
m
e
p
o
ten
tial
e
n
h
an
ce
m
e
n
t
a
n
d
m
o
d
i
f
icatio
n
ca
n
b
e
d
o
n
e
o
n
th
e
c
u
r
r
en
t
d
esi
g
n
.
F
ir
s
tl
y
,
th
e
c
u
r
r
en
t
d
esi
g
n
u
s
es
s
tatic
w
o
r
d
e
m
b
ed
d
in
g
,
w
h
ich
m
ea
n
s
a
f
ter
t
h
e
w
o
r
d
e
m
b
ed
d
in
g
m
o
d
el
w
as
tr
ai
n
ed
,
th
e
w
ei
g
h
t
f
o
r
th
e
w
o
r
d
s
(
n
u
m
er
ical
r
ep
r
esen
ta
tio
n
)
w
o
u
ld
b
e
f
i
x
ed
.
D
y
n
a
m
ic
w
o
r
d
e
m
b
ed
d
in
g
i
s
a
p
o
s
s
ib
l
e
en
h
a
n
ce
m
en
t
.
Usi
n
g
d
y
n
a
m
ic
w
o
r
d
e
m
b
ed
d
in
g
,
th
e
w
ei
g
h
t
o
f
th
e
w
o
r
d
em
b
ed
d
in
g
m
o
d
el
ca
n
b
e
tu
n
ed
w
h
il
e
o
n
th
e
C
NN
tr
ain
in
g
s
tate.
L
astl
y
,
t
h
e
d
if
f
er
e
n
t
tex
t
p
r
e
-
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
ca
n
ex
p
l
o
r
e
to
en
h
an
ce
t
h
e
m
o
d
el.
Fo
r
tex
t
clas
s
if
icatio
n
,
tex
t
-
p
r
o
ce
s
s
i
n
g
is
eq
u
al
ly
i
m
p
o
r
ta
n
t
w
it
h
th
e
d
ee
p
lear
n
in
g
m
o
d
el
b
u
ild
in
g
.
A
lit
tle
m
o
d
i
f
icatio
n
d
o
n
e
o
n
t
h
is
s
tag
e
w
i
ll h
i
g
h
l
y
in
f
l
u
e
n
ce
th
e
o
u
tco
m
e.
ACK
NO
WL
E
D
G
E
M
E
NT
S
T
h
is
w
o
r
k
i
s
s
u
p
p
o
r
ted
b
y
Un
i
v
er
s
iti T
ek
n
o
lo
g
i M
ala
y
s
ia
(
U
T
M)
.
T
h
is
w
o
r
k
is
f
i
n
a
n
ciall
y
s
u
p
p
o
r
ted
b
y
t
h
e
Ger
an
U
n
i
v
er
s
iti
P
en
y
e
lid
ik
an
(
GUP
)
1
5
J
3
5
.
T
h
e
au
th
o
r
s
w
o
u
ld
lik
e
to
ex
p
r
ess
t
h
e
ir
d
ee
p
est
g
r
atitu
d
e
to
th
e
R
esear
c
h
Ma
n
a
g
e
m
en
t Cen
ter
(
R
M
C
)
,
UT
M
f
o
r
m
a
k
i
n
g
t
h
i
s
r
esear
ch
a
s
u
cc
es
s
.
RE
F
E
R
E
NC
E
S
[1
]
G
a
o
,
S
h
a
n
sh
a
n
,
Jin
x
in
g
Ha
o
,
a
n
d
Yu
F
u
.
"
T
h
e
a
p
p
li
c
a
ti
o
n
a
n
d
c
o
m
p
a
riso
n
o
f
w
e
b
se
r
v
ice
s
f
o
r
se
n
ti
m
e
n
t
a
n
a
l
y
sis in
to
u
ri
sm
.
"
1
2
th
In
ter
n
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
S
e
rv
ice
S
y
ste
ms
a
n
d
S
e
rv
ice
M
a
n
a
g
e
me
n
t
(
ICS
S
S
M
)
.
IE
EE
,
2
0
1
5
.
[2
]
Ha
n
d
a
y
a
n
i,
D.,
Ba
k
a
r,
N.
S
.
A
.
A
.
a
n
d
Ya
a
c
o
b
,
H.
S
e
n
ti
me
n
t
An
a
lys
is
fo
r
M
a
l
a
y
L
a
n
g
u
a
g
e
:
S
y
ste
ma
ti
c
L
it
e
ra
t
u
re
Rev
iew
.
2
0
1
8
In
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
In
f
o
rm
a
ti
o
n
a
n
d
C
o
m
m
u
n
ica
ti
o
n
T
e
c
h
n
o
lo
g
y
f
o
r
th
e
M
u
slim
W
o
rld
(ICT
4
M
).
IEE
E.
2
0
1
8
.
3
0
5
-
3
1
0
.
[3
]
E.
Ca
m
b
ria,
B.
S
c
h
u
ll
e
r,
Y.
X
i
a
,
a
n
d
C.
Ha
v
a
si,
“
Ne
w
a
v
e
n
u
e
s
in
o
p
i
n
i
o
n
m
in
in
g
a
n
d
se
n
ti
m
e
n
t
a
n
a
ly
sis.”
In
telli
g
e
n
t
sy
ste
ms
,
IEE
E
,
v
o
l.
2
8
,
n
o
.
2
,
p
p
.
1
5
-
2
1
,
M
a
rc
h
2
0
1
3
.
[4
]
B.
P
a
n
g
,
L
il
li
a
n
L
e
e
,
a
n
d
S
.
V
a
it
h
y
a
n
a
th
a
n
.
"
T
h
u
m
b
s
u
p
?
:
s
e
n
ti
m
e
n
t
c
las
si
f
ic
a
ti
o
n
u
sin
g
m
a
c
h
in
e
lea
rn
in
g
tec
h
n
iq
u
e
s."
Pro
c
.
ACL
-
0
2
c
o
n
f
e
re
n
c
e
o
n
Emp
irica
l
me
th
o
d
s
in
n
a
t
u
ra
l
la
n
g
u
a
g
e
p
ro
c
e
ss
in
g
,
Asso
c
ia
ti
o
n
f
o
r
Co
mp
u
t
a
ti
o
n
a
l
L
in
g
u
isti
c
s,
v
o
l
1
0
.
2
0
0
2
.
[5
]
J.
-
Y.
Nie
,
D.
Z
h
a
o
a
n
d
Y.
F
e
n
g
.
Na
tu
ra
l
L
a
n
g
u
a
g
e
Pr
o
c
e
ss
in
g
a
n
d
Ch
i
n
e
se
Co
mp
u
ti
n
g
.
S
p
ri
n
g
e
r.
2
0
1
4
.
[6
]
B.
P
a
n
g
a
n
d
L
.
L
e
e
.
"
Op
in
io
n
m
in
in
g
a
n
d
se
n
ti
m
e
n
t
a
n
a
l
y
sis."
Fo
u
n
d
a
ti
o
n
s
a
n
d
T
re
n
d
s®
in
I
n
fo
rm
a
ti
o
n
Retrie
v
a
l,
v
o
l.
2
:
No
.
1
-
2
,
p
p
.
1
-
1
3
5
,
2
0
0
8
.
[7
]
V
.
K
o
rd
e
a
n
d
C
.
N.
M
a
h
e
n
d
e
r,
“
T
e
x
t
c
las
si
f
ica
ti
o
n
a
n
d
c
las
sif
ier
s:
A
su
rv
e
y
.
”
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
Arti
fi
c
i
a
l
In
telli
g
e
n
c
e
&
Ap
p
li
c
a
ti
o
n
s
,
v
o
l.
3
,
n
o
.
2
,
p
.
8
5
,
2
0
1
2
.
[8
]
M
.
T
a
b
o
a
d
a
,
J.
Bro
o
k
e
,
M
.
T
o
f
il
o
sk
i,
K.
V
o
ll
,
a
n
d
M
.
S
ted
e
.
L
e
x
i
c
o
n
-
b
a
se
d
m
e
th
o
d
s
f
o
r
se
n
ti
m
e
n
t
a
n
a
ly
sis.
Co
mp
u
t
a
ti
o
n
a
l
li
n
g
u
isti
c
s
,
v
o
l.
3
2
,
n
o
2
,
p
p
.
2
6
7
-
3
0
7
,
2
0
1
1
.
[9
]
J.
Ya
o
.
“
A
u
to
m
a
ted
S
e
n
ti
m
e
n
t
A
n
a
l
y
si
s
o
f
T
e
x
t
Da
ta
w
it
h
NL
TK.
”
J
o
u
rn
a
l
o
f
P
h
y
sic
s:
Co
n
fer
e
n
c
e
S
e
rie
s,
IOP
Pu
b
li
s
h
in
g
,
v
o
l.
1
1
8
7
,
n
o
.
5
,
2
0
1
9
.
[1
0
]
A
.
Ku
lk
a
rn
i
a
n
d
A
.
S
h
iv
a
n
a
n
d
a
.
Na
tu
ra
l
L
a
n
g
u
a
g
e
Pro
c
e
ss
in
g
Rec
ip
e
s:
Un
lo
c
k
in
g
T
e
x
t
Da
ta
wit
h
M
a
c
h
in
e
L
e
a
rn
in
g
a
n
d
De
e
p
L
e
a
r
n
in
g
Us
i
n
g
Pyt
h
o
n
.
A
p
re
ss
;
2
0
1
9
.
[1
1
]
ZS
Ha
rris,
“
Distrib
u
ti
o
n
a
l
str
u
c
tu
re
”
,
W
o
rd
,
v
o
l.
1
0
,
n
o
.
2
-
3
,
p
p
.
1
4
6
-
1
6
2
,
1
9
5
4
.
[1
2
]
J.
Ra
m
o
s.
“
Us
in
g
tf
-
id
f
to
d
e
ter
m
in
e
w
o
rd
re
le
v
a
n
c
e
in
d
o
c
u
m
e
n
t
q
u
e
ries
,
”
Pro
c
.
Fi
rs
t
in
str
u
c
ti
o
n
a
l
c
o
n
fer
e
n
c
e
o
n
ma
c
h
in
e
le
a
rn
in
g
,
v
o
l.
2
4
2
,
p
p
.
1
3
3
-
1
4
2
,
2
0
0
3
.
[1
3
]
T
.
M
ik
o
lo
v
,
K.
Ch
e
n
,
G
.
Co
rra
d
o
a
n
d
J.
De
a
n
,
“
Ef
f
i
c
ien
t
e
stim
a
ti
o
n
o
f
w
o
rd
re
p
re
se
n
tatio
n
s
in
v
e
c
to
r
sp
a
c
e
,
”
a
rXiv
p
re
p
rin
t
a
rXiv:1
3
0
1
.
3
7
8
1
,
2
0
1
3
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
2
,
J
u
n
e
20
20
:
2
12
–
2
20
220
[1
4
]
G
.
I.
W
e
b
b
,
J.R.
B
o
u
g
h
t
o
n
,
a
n
d
Z
.
W
a
n
g
.
“
No
t
so
n
a
iv
e
Ba
y
e
s:
a
g
g
re
g
a
ti
n
g
o
n
e
-
d
e
p
e
n
d
e
n
c
e
e
stim
a
to
rs,”
M
a
c
h
i
n
e
lea
rn
in
g
,
v
o
l
.
5
8
,
n
o
.
1
,
p
p
.
5
-
2
4
,
2
0
0
5
.
[1
5
]
V
.
V
a
p
n
ik
,
T
h
e
n
a
tu
re
o
f
st
a
ti
stic
a
l
lea
r
n
in
g
t
h
e
o
ry
.
S
p
rin
g
e
r
sc
ien
c
e
&
b
u
sin
e
ss
m
e
d
ia,
2
0
1
3
.
[1
6
]
M
.
M
.
L
o
p
e
z
a
n
d
J.
Ka
li
ta.
“
De
e
p
L
e
a
rn
in
g
a
p
p
li
e
d
t
o
NL
P
,
”
a
rXiv
p
re
p
rin
t
a
rXiv:1
7
0
3
.
0
3
0
9
1
.
2
0
1
7
.
[1
7
]
Y.
Kim
.
“
Co
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
s
f
o
r
se
n
ten
c
e
c
las
si
f
ica
ti
o
n
,
”
a
rXiv p
re
p
ri
n
t
a
rX
iv: 1
4
0
8
.
5
8
8
2
,
2
0
1
4
.
[1
8
]
N
.
Ka
lch
b
re
n
n
e
r,
E.
G
re
fe
n
ste
tt
e
a
n
d
P
.
B
lu
n
so
m
.
“
A
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
r
a
l
n
e
tw
o
rk
f
o
r
m
o
d
e
ll
in
g
se
n
ten
c
e
s,”
a
rXiv
p
re
p
rin
t
a
rXiv:
1
4
0
4
.
2
1
8
8
,
2
0
1
4
.
[1
9
]
S
.
L
iao
,
J.
W
a
n
g
,
R.
Yu
,
K.
S
a
t
o
a
n
d
Z.
C
h
e
n
g
.
“
CNN
f
o
r
situ
a
ti
o
n
s
u
n
d
e
rsta
n
d
i
n
g
b
a
se
d
o
n
se
n
ti
m
e
n
t
a
n
a
ly
si
s
o
f
tw
it
ter d
a
ta,”
Pro
c
e
d
ia
c
o
m
p
u
ter
sc
ien
c
e
,
v
o
l.
1
1
1
,
p
p
.
3
7
6
-
3
8
1
,
2
0
1
7
.
[2
0
]
Y.
Zh
a
n
g
a
n
d
B.
W
a
ll
a
c
e
.
“
A
se
n
siti
v
it
y
a
n
a
l
y
sis
o
f
(a
n
d
p
ra
c
ti
ti
o
n
e
rs
g
u
id
e
to
)
c
o
n
v
o
lu
t
io
n
a
l
n
e
u
r
a
l
n
e
tw
o
rk
s
f
o
r
se
n
ten
c
e
c
las
si
f
ica
ti
o
n
,
”
a
rXiv
p
re
p
rin
t
a
rXiv: 1
5
1
0
.
0
3
8
2
0
,
2
0
1
5
.
[2
1
]
F
.
Ch
o
ll
e
t,
Ke
ra
s
,
G
it
Hu
b
re
p
o
sit
o
ry
,
h
tt
p
s://
g
it
h
u
b
.
c
o
m
/fch
o
ll
e
t/
k
e
ra
s,
2
0
1
5
.
[2
2
]
P
.
Ch
ik
e
rsa
l,
S
.
P
o
ria
a
n
d
E.
Ca
m
b
ria.
“
S
e
N
T
U:
se
n
ti
m
e
n
t
a
n
a
l
y
sis
o
f
twe
e
ts
b
y
c
o
m
b
in
in
g
a
ru
le
-
b
a
se
d
c
las
si
f
ier
w
it
h
su
p
e
rv
ise
d
lea
rn
in
g
,
”
Pr
o
c
.
9
th
In
ter
n
a
ti
o
n
a
l
W
o
rk
sh
o
p
o
n
S
e
ma
n
ti
c
Eva
l
u
a
ti
o
n
(
S
e
mEv
a
l
2
0
1
5
)
p
p
.
6
4
7
-
6
5
1
,
2
0
1
5
.
[2
3
]
H.
Zo
lk
e
p
li
.
M
a
la
y
a
,
Na
tu
ra
l
-
L
a
n
g
u
a
g
e
-
T
o
o
lkit
li
b
ra
ry
fo
r
b
a
h
a
s
a
M
a
la
y
sia
,
p
o
we
re
d
b
y
De
e
p
L
e
a
rn
in
g
T
e
n
so
rflo
w
.
G
it
Hu
b
re
p
o
sito
ry
,
h
tt
p
s://
g
it
h
u
b
.
c
o
m
/h
u
se
in
z
o
l
0
5
/m
a
la
y
a
,
2
0
1
8
.
[2
4
]
T
.
A
.
L
e
,
D.
M
o
e
lj
a
d
i,
Y.
M
iu
ra
a
n
d
T
.
Oh
k
u
m
a
.
“
S
e
n
ti
m
e
n
t
A
n
a
ly
sis
f
o
r
L
o
w
Re
so
u
rc
e
L
a
n
g
u
a
g
e
s:
A
S
tu
d
y
o
n
In
f
o
rm
a
l
In
d
o
n
e
sia
n
T
we
e
ts,”
Pr
o
c
.
o
f
th
e
1
2
th
W
o
rk
sh
o
p
o
n
Asi
a
n
L
a
n
g
u
a
g
e
Res
o
u
rc
e
s
(
AL
R1
2
)
,
p
p
.
1
2
3
-
1
3
1
,
2
0
1
6
.
[2
5
]
W
.
C.
F
.
M
a
riel,
S
.
M
a
riy
a
h
,
a
n
d
S
.
P
ra
m
a
n
a
.
“
S
e
n
ti
m
e
n
t
a
n
a
ly
sis:
a
c
o
m
p
a
riso
n
o
f
d
e
e
p
lea
rn
i
n
g
n
e
u
ra
l
n
e
tw
o
rk
a
lg
o
rit
h
m
w
it
h
S
VM
a
n
d
n
a
ïv
e
B
a
y
e
s
f
o
r
In
d
o
n
e
sia
n
tex
t,
”
J
o
u
r
n
a
l
o
f
P
h
y
sic
s:
Co
n
fer
e
n
c
e
S
e
rie
s.
IOP
Pu
b
l
ish
i
n
g
.
v
o
l.
9
7
1
,
2
0
1
8
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
On
g
Ju
n
Yin
g
re
c
e
i
v
e
d
h
is
M
.
En
g
(2
0
1
9
)
in
Co
m
p
u
ter
a
n
d
M
icro
e
lec
tro
n
ics
S
y
ste
m
f
ro
m
Un
iv
e
rsiti
T
e
k
n
o
lo
g
i
M
a
la
y
sia
.
His
re
se
a
rc
h
w
o
rk
is
m
a
in
l
y
o
n
m
a
c
h
in
e
lea
rn
in
g
,
a
rti
f
icia
l
in
telli
g
e
n
c
e
a
n
d
Na
tu
ra
l
L
a
n
g
u
a
g
e
P
ro
c
e
ss
in
g
.
He
jo
in
e
d
In
tel
Co
rp
i
n
2
0
1
7
a
n
d
c
u
rre
n
tl
y
a
S
y
st
e
m
On
Ch
ip
De
sig
n
E
n
g
in
e
e
r.
M
u
h
a
m
m
a
d
M
u
n
’im
A
h
m
a
d
Zab
id
i
re
c
e
c
iv
e
d
h
is
M
S
i
n
Co
m
p
u
te
r
En
g
in
e
e
rin
g
(1
9
8
8
)
f
ro
m
th
e
Un
iv
e
rsit
y
o
f
Brid
g
e
p
o
rt,
USA
.
His
re
se
a
r
c
h
w
o
rk
is
m
a
in
l
y
o
n
a
rti
f
icia
l
in
telli
g
e
n
c
e
,
d
ig
it
a
l
sig
n
a
l
p
ro
c
e
ss
in
g
a
n
d
e
m
b
e
d
d
e
d
s
y
ste
m
s
d
e
sig
n
.
He
is
c
u
rre
n
tl
y
a
S
e
n
i
o
r
L
e
c
tu
re
r
a
t
Un
iv
e
rsiti
T
e
k
n
o
lo
g
i
M
a
lay
sia
No
rh
a
f
iza
h
Ra
m
li
re
c
e
i
v
e
d
h
e
r
M
EE
d
e
g
re
e
in
El
e
c
tri
c
a
l
-
El
e
c
tro
n
ics
En
g
in
e
e
rin
g
f
ro
m
Un
iv
e
rsiti
T
e
k
n
o
lo
g
i
M
a
lay
sia
in
1
9
9
2
.
He
r
re
se
a
rc
h
in
tere
st i
s
m
a
i
n
ly
in
Bio
m
e
d
ica
l
a
n
d
Dig
it
a
l
El
e
c
tro
n
ics
.
S
h
e
is cu
rre
n
tl
y
a
S
e
n
io
r
L
e
c
tu
re
r
a
t
Un
iv
e
rsiti
T
e
k
n
o
l
o
g
i
M
a
lay
si
a
Us
m
a
n
Ullah
S
h
e
ik
h
re
c
e
iv
e
d
h
i
s
P
h
D
d
e
g
re
e
(2
0
0
9
)
in
im
a
g
e
p
ro
c
e
ss
in
g
a
n
d
c
o
m
p
u
ter
v
isio
n
f
ro
m
Un
iv
e
r
siti
Tek
n
o
lo
g
i
M
a
la
y
sia
.
His
re
s
e
a
rc
h
w
o
rk
is
m
a
in
ly
o
n
c
o
m
p
u
ter
v
isio
n
,
m
a
c
h
in
e
lea
rn
in
g
a
n
d
e
m
b
e
d
d
e
d
sy
ste
m
s
d
e
sig
n
.
He
is
c
u
rre
n
tl
y
a
S
e
n
io
r
L
e
c
tu
re
r
a
t
Un
iv
e
rsiti
T
e
k
n
o
lo
g
i
M
a
la
y
sia
.
Evaluation Warning : The document was created with Spire.PDF for Python.