I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
8
,
No
.
4
,
A
u
g
u
s
t
201
8
,
p
p
.
2
3
5
8
~
2
3
6
6
I
SS
N:
2088
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
8
i
4
.
p
p
2
3
5
8
-
2366
2358
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e
.
co
m/
jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JE
CE
Co
nceptual S
enti
m
en
t
A
na
ly
sis
Mo
del
K
ra
nti
Vit
ha
l G
ha
g
1
,
K
et
a
n
Sh
a
h
2
1
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
De
p
a
rt
m
e
n
t,
M
ET
’s S
A
KEC,
M
u
m
b
a
i
Un
iv
e
rsit
y
,
In
d
ia
2
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
De
p
a
rt
m
e
n
t,
S
V
KM’s
NMIM
S
M
P
S
T
M
E
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
ei
v
ed
J
an
14,
201
8
R
ev
i
s
ed
A
p
r
2
0
,
2
0
1
8
A
cc
ep
ted
A
p
r
2
7
,
2
0
1
8
Ba
g
-
of
-
w
o
rd
s
a
p
p
ro
a
c
h
is
p
o
p
u
la
rly
u
se
d
f
o
r
S
e
n
ti
m
e
n
t
a
n
a
l
y
sis.
I
t
m
a
p
s
th
e
term
s
in
th
e
re
v
ie
w
s
to
ter
m
-
d
o
c
u
m
e
n
t
v
e
c
to
rs
a
n
d
th
u
s d
isru
p
ts
th
e
s
y
n
tac
ti
c
stru
c
tu
re
o
f
se
n
ten
c
e
s
in
t
h
e
re
v
iew
s.
A
s
so
c
iatio
n
a
m
o
n
g
th
e
ter
m
s
o
r
t
h
e
se
m
a
n
ti
c
stru
c
tu
re
o
f
se
n
ten
c
e
s
is
a
lso
n
o
t
p
re
se
rv
e
d
.
T
h
is
re
se
a
rc
h
w
o
r
k
f
o
c
u
se
s
o
n
c
las
sify
in
g
th
e
se
n
ti
m
e
n
ts
b
y
c
o
n
sid
e
rin
g
th
e
sy
n
tac
ti
c
a
n
d
se
m
a
n
ti
c
stru
c
tu
re
o
f
th
e
se
n
ten
c
e
s
in
th
e
re
v
ie
w
.
T
o
i
m
p
ro
v
e
a
c
c
u
ra
c
y
,
se
n
ti
m
e
n
t
c
las
sif
i
e
rs b
a
se
d
o
n
re
lativ
e
f
re
q
u
e
n
c
y
,
a
v
e
ra
g
e
f
re
q
u
e
n
c
y
a
n
d
ter
m
f
re
q
u
e
n
c
y
in
v
e
rse
d
o
c
u
m
e
n
t
f
re
q
u
e
n
c
y
w
e
r
e
p
ro
p
o
se
d
.
T
o
h
a
n
d
le
term
s
w
it
h
a
p
o
stro
p
h
e
,
p
re
p
ro
c
e
ss
in
g
tec
h
n
i
q
u
e
s
w
e
re
e
x
ten
d
e
d
.
T
o
f
o
c
u
s
o
n
o
p
i
n
io
n
a
ted
c
o
n
ten
ts,
s
u
b
jec
ti
v
it
y
e
x
tra
c
ti
o
n
w
a
s
p
e
r
f
o
r
m
e
d
a
t
p
h
ra
se
lev
e
l.
Ex
p
e
rime
n
ts
we
re
p
e
r
f
o
r
m
e
d
o
n
P
a
n
g
&
L
e
e
s,
K
a
g
g
le’s
a
n
d
UCI
’s
d
a
tas
e
t
.
Clas
sif
ier
s
w
e
r
e
a
lso
e
v
a
lu
a
ted
o
n
th
e
UCI’s
P
r
o
d
u
c
t
a
n
d
Re
sta
u
r
a
n
t
d
a
tas
e
t.
S
e
n
ti
m
e
n
t
Clas
si
f
ica
ti
o
n
a
c
c
u
ra
c
y
i
m
p
ro
v
e
d
f
ro
m
6
7
.
9
%
f
o
r
a
c
o
m
p
a
ra
b
le
term
we
ig
h
in
g
tec
h
n
iq
u
e
,
De
lt
a
T
F
IDF,
u
p
to
7
7
.
2
%
f
o
r
p
ro
p
o
se
d
c
las
si
f
iers
.
In
c
e
p
ti
o
n
o
f
th
e
p
r
o
p
o
se
d
c
o
n
c
e
p
t
b
a
se
d
a
p
p
ro
a
c
h
,
s
u
b
jec
ti
v
it
y
e
x
trac
ti
o
n
a
n
d
e
x
ten
sio
n
s
to
p
re
p
r
o
c
e
ss
in
g
tec
h
n
iq
u
e
s,
im
p
ro
v
e
d
th
e
a
c
c
u
ra
c
y
to
9
3
.
9
%
.
K
ey
w
o
r
d
:
C
o
n
ce
p
t
-
b
a
s
ed
ap
p
r
o
ac
h
Mo
v
ie
r
ev
ie
w
s
P
o
lar
ity
clas
s
i
f
icatio
n
Sen
ti
m
e
n
t a
n
a
l
y
s
is
T
ex
t m
in
i
n
g
Co
p
y
rig
h
t
©
2
0
1
8
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Kr
an
ti Vit
h
al
G
h
ag
,
IT
Dep
ar
tm
e
n
t
,
ME
T
’
s
S
A
KE
C
,
Mu
m
b
ai
U
n
i
v
er
s
it
y
,
W
.
T
.
P
at
il M
ar
g
,
C
h
e
m
b
u
r
,
Mu
m
b
ai,
Ma
h
ar
as
h
tr
a,
I
n
d
ia
4
0
0
0
8
8
.
E
m
ail:
k
r
an
ti.
g
h
ag
@
s
a
k
ec
.
ac
.
i
n
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
w
eb
w
h
ich
is
m
a
s
s
i
v
el
y
i
n
cr
ea
s
i
n
g
r
eso
u
r
ce
o
f
in
f
o
r
m
a
tio
n
h
as
c
h
an
g
ed
f
r
o
m
r
ea
d
o
n
l
y
to
r
ea
d
w
r
ite.
L
ar
g
e
a
m
o
u
n
t
o
f
o
p
i
n
io
n
ated
d
ata
i
s
g
en
er
ated
b
y
o
n
l
in
e
ac
t
iv
i
ties
lik
e
s
o
cial
m
ed
ia,
b
lo
g
g
i
n
g
,
s
u
r
v
e
y
s
an
d
f
o
r
u
m
s
[
1
]
.
E
lectr
o
n
ic
W
o
r
d
o
f
Mo
u
th
(
eW
o
M)
h
as
b
e
co
m
e
m
o
r
e
p
o
p
u
lar
th
a
n
t
h
e
t
r
ad
itio
n
al
W
o
r
d
o
f
Mo
u
th
(
W
o
M)
p
u
b
licit
y
[
2
]
.
Au
to
m
a
ticall
y
p
r
o
ce
s
s
in
g
s
e
n
t
i
m
e
n
t d
ata
n
ee
d
s
to
b
e
h
an
d
le
d
s
y
s
te
m
atica
l
l
y
.
Sen
ti
m
e
n
t
An
al
y
s
i
s
i
n
v
o
lv
e
s
ex
tr
ac
ti
n
g
,
p
r
ep
r
o
ce
s
s
in
g
,
u
n
d
er
s
tan
d
i
n
g
,
clas
s
if
y
i
n
g
&
p
r
esen
tin
g
s
en
ti
m
e
n
ts
e
x
p
r
ess
ed
b
y
t
h
e
u
s
er
s
[
3
]
.
Sen
ti
m
e
n
t
a
n
al
y
s
i
s
g
en
er
all
y
i
n
v
o
l
v
es
cla
s
s
i
f
y
i
n
g
th
e
p
o
lar
it
y
o
f
a
p
iece
o
f
tex
t
as
p
o
s
itiv
e,
n
eg
a
tiv
e
o
r
n
eu
tr
al
.
I
t
also
in
v
o
lv
e
s
s
u
b
j
ec
tiv
it
y
ex
tr
ac
tio
n
[
4
]
,
in
ten
s
it
y
p
r
ed
ictio
n
[
5
]
an
d
em
o
tio
n
class
i
f
icatio
n
[
6
]
.
Sen
ti
m
e
n
t
a
n
al
y
s
is
w
as
also
p
er
f
o
r
m
ed
at
t
h
e
ter
m
,
s
en
ten
ce
,
p
ar
ag
r
ap
h
,
d
o
cu
m
en
t
lev
el
a
n
d
w
a
s
als
o
ex
ten
d
ed
to
asp
ec
t
lev
el
[
7
]
,
[
8
]
.
Sen
ti
m
en
t
A
n
al
y
s
is
w
a
s
p
er
f
o
r
m
ed
o
n
lan
g
u
a
g
es
lik
e
E
n
g
li
s
h
,
C
h
i
n
e
s
e,
A
r
ab
ic
an
d
Vie
tn
a
m
ese
la
n
g
u
a
g
e.
T
h
er
e
ar
e
v
er
y
f
e
w
m
u
ltil
in
g
u
al
s
en
t
i
m
e
n
t
class
i
f
ier
s
[
9
]
.
Sen
ti
m
e
n
t
a
n
al
y
s
i
s
tec
h
n
iq
u
e
s
ca
n
b
e
b
r
o
ad
ly
cla
s
s
i
f
ied
a
s
s
u
p
er
v
is
ed
l
ea
r
n
i
n
g
an
d
u
n
s
u
p
er
v
i
s
ed
lear
n
in
g
tec
h
n
iq
u
es
[
1
0
]
.
Ma
n
y
u
n
s
u
p
er
v
i
s
ed
lear
n
i
n
g
tec
h
n
iq
u
es
u
s
e
ex
is
ti
n
g
lex
ica
l
r
eso
u
r
ce
s
li
k
e
W
o
r
d
Net
[
1
1
]
an
d
lan
g
u
a
g
e
s
p
ec
i
f
ic
s
e
n
ti
m
en
t
i
n
f
o
r
m
atio
n
li
k
e
s
e
n
ti
m
en
t
s
ee
d
w
o
r
d
s
,
t
h
eir
S
y
n
o
n
y
m
s
a
n
d
an
to
n
y
m
s
to
co
n
s
tr
u
ct
a
n
d
u
p
d
ate
s
en
ti
m
en
t le
x
ico
n
s
[
5
]
,
[
1
2
]
.
Un
s
u
p
e
r
v
is
ed
lear
n
i
n
g
tec
h
n
iq
u
es a
s
s
i
g
n
ed
a
g
en
er
al
ized
p
o
lar
ity
a
n
d
w
eig
h
t
to
a
ter
m
an
d
t
h
u
s
f
ail
to
ca
p
tu
r
e
i
ts
d
o
m
ai
n
s
p
ec
if
ic
co
n
te
x
t.
Su
p
er
v
i
s
ed
lear
n
i
n
g
tech
n
iq
u
es
u
s
e
a
s
et
o
f
r
ev
ie
w
s
,
tag
g
ed
p
o
s
iti
v
e
o
r
n
e
g
ati
v
e
to
tr
ain
t
h
e
clas
s
i
f
ier
s
.
P
r
ep
r
o
ce
s
s
in
g
ta
s
k
s
s
u
c
h
as
s
to
p
w
o
r
d
s
r
e
m
o
v
a
l
[
1
3
]
,
p
u
n
ct
u
atio
n
s
r
e
m
o
v
al
ar
e
p
e
r
f
o
r
m
ed
o
n
t
h
ese
d
o
cu
m
e
n
t
s
.
T
er
m
s
ar
e
t
h
e
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
C
o
n
ce
p
tu
a
l S
e
n
timen
t A
n
a
lysi
s
Mo
d
el
(
K
r
a
n
ti V
ith
a
l G
h
a
g
)
2359
class
i
f
ied
o
n
th
e
b
asis
o
f
t
h
eir
d
o
m
i
n
a
n
c
y
i
n
p
o
s
itiv
el
y
ta
g
g
ed
d
o
cu
m
en
ts
v
er
s
u
s
n
eg
ati
v
el
y
ta
g
g
ed
d
o
cu
m
en
ts
.
T
h
is
le
x
ico
n
is
t
h
e
n
u
s
ed
f
o
r
class
i
f
y
i
n
g
t
h
e
r
ev
i
e
w
s
u
s
in
g
b
ag
-
of
-
w
o
r
d
s
(
B
OW
)
ap
p
r
o
ac
h
[
1
4
]
.
Su
p
er
v
i
s
ed
as
w
ell
a
s
u
n
s
u
p
e
r
v
is
ed
tec
h
n
iq
u
es
u
s
ed
b
ag
-
of
-
w
o
r
d
s
ap
p
r
o
ac
h
.
P
r
o
ce
s
s
in
g
s
tr
u
ct
u
r
ed
v
ec
to
r
s
w
a
s
s
y
s
te
m
atic
co
m
p
ar
ed
to
u
n
s
tr
u
ct
u
r
ed
,
b
u
t
p
o
s
i
tio
n
al
i
n
f
o
r
m
atio
n
ca
r
r
ied
b
y
ter
m
w
a
s
i
g
n
o
r
ed
.
S
y
n
tactic
s
tr
u
ct
u
r
e
o
f
a
s
en
te
n
ce
p
la
y
s
an
i
m
p
o
r
tan
t
r
o
le
i
n
s
en
ti
m
e
n
t
a
n
al
y
s
i
s
[
1
5
]
.
T
er
m
-
d
o
c
u
m
e
n
t
v
ec
to
r
s
p
r
o
v
id
ed
th
e
co
u
n
t
o
f
ter
m
i
n
d
o
cu
m
e
n
t
b
u
t
t
h
e
ter
m
as
an
in
d
i
v
id
u
al
m
a
y
ca
r
r
y
a
d
if
f
e
r
en
t
m
ea
n
in
g
t
h
a
n
w
h
a
t it
m
ea
n
t a
s
g
r
o
u
p
o
f
w
o
r
d
s
in
s
e
n
te
n
ce
.
Se
m
an
tic
s
tr
u
c
tu
r
e
of
w
o
r
d
s
i
n
s
e
n
ten
ce
co
n
t
r
ib
u
tes to
th
e
ac
tu
al
m
ea
n
in
g
&
th
u
s
m
a
y
ad
j
o
in
f
o
r
s
en
ti
m
e
n
t
a
n
al
y
s
is
al
s
o
.
T
h
is
in
f
o
r
m
atio
n
w
as
n
o
t
h
an
d
le
d
w
h
e
n
r
ep
r
esen
ti
n
g
d
o
cu
m
en
ts
as
ter
m
v
ec
to
r
s
.
B
ag
-
of
-
w
o
r
d
s
b
a
s
ed
ap
p
r
o
ac
h
es
t
h
u
s
f
ai
l
to
ca
p
tu
r
e
s
y
n
t
ac
tic
an
d
s
e
m
a
n
tic
s
tr
u
ct
u
r
e
o
f
r
ev
ie
w
s
.
A
lt
h
o
u
g
h
s
u
p
er
v
is
ed
tec
h
n
iq
u
es
s
u
c
h
as
SVM,
NN
&
u
n
s
u
p
er
v
i
s
ed
ap
p
r
o
ac
h
es
s
u
ch
a
s
lex
ico
n
-
b
ased
ap
p
r
o
ac
h
es
ar
e
p
o
p
u
lar
,
in
tellig
e
n
t
s
y
s
te
m
s
s
u
c
h
as
co
n
ce
p
t
-
b
ased
ap
p
r
o
ac
h
es
is
n
ee
d
o
f
h
o
u
r
[
1
]
.
Sen
ti
m
e
n
t
C
la
s
s
i
f
ier
s
,
p
r
ep
r
o
ce
s
s
in
g
tec
h
n
iq
u
es
&
d
ed
u
ctio
n
m
eth
o
d
o
lo
g
ies
ar
e
th
e
co
r
e
co
m
p
o
n
e
n
t
s
o
f
Su
p
er
v
is
ed
Se
n
ti
m
en
t
An
al
y
s
is
.
P
an
g
,
L
ee
,
&
Vait
h
y
an
a
th
a
n
s
tated
th
at
s
en
ti
m
e
n
t
a
n
al
y
s
i
s
n
ee
d
s
to
b
e
h
an
d
led
i
n
a
m
o
r
e
s
o
p
h
is
t
icate
d
w
a
y
t
h
an
tr
ad
itio
n
al
te
x
t
ca
te
g
o
r
izatio
n
[
1
6
]
.
T
h
ey
ar
e
p
io
n
ee
r
s
f
o
r
ex
tr
ac
ti
n
g
,
tr
a
n
s
f
o
r
m
i
n
g
&
tag
g
in
g
p
o
p
u
lar
m
o
v
ie
r
ev
ie
w
d
atase
t
[
1
7
]
.
Do
m
ai
n
s
p
ec
i
f
i
c
s
en
ti
m
e
n
t
a
n
al
y
s
i
s
m
o
d
el
s
w
er
e
d
e
s
ig
n
ed
f
o
r
v
ar
i
o
u
s
d
o
m
ai
n
s
s
u
c
h
a
s
m
o
v
ie,
r
estau
r
a
n
t,
m
o
b
ile,
b
o
o
k
s
a
n
d
DVD’
s
.
T
h
e
m
o
v
i
e
d
o
m
ai
n
w
a
s
r
elati
v
el
y
d
i
f
f
icu
l
t to
class
i
f
y
[
1
8
]
.
T
o
p
r
an
k
ed
i
n
d
ex
ter
m
s
w
er
e
n
o
t th
e
to
p
r
an
k
ed
s
e
n
ti
m
e
n
tall
y
p
o
lar
ized
ter
m
s
.
So
t
er
m
s
w
e
r
e
class
if
ied
o
n
th
e
b
as
is
o
f
ter
m
p
r
esen
ce
d
is
tr
ib
u
tio
n
ac
r
o
s
s
p
o
s
itiv
el
y
a
n
d
n
eg
at
iv
el
y
ta
g
g
ed
d
o
cu
m
e
n
ts
[
1
9
]
.
F
r
eq
u
en
c
y
d
is
tr
ib
u
t
io
n
w
a
s
n
o
t
co
n
s
id
er
ed
.
C
las
s
i
f
i
er
s
s
u
c
h
a
s
Nai
v
e
B
ay
e
s
,
SVM
a
n
d
R
a
n
d
o
m
Fo
r
est C
la
s
s
i
f
ier
s
w
er
e
a
m
o
n
g
p
o
p
u
lar
tech
n
iq
u
e
s
f
o
r
s
e
n
ti
m
e
n
t
class
i
f
icatio
n
[
2
0]
.
T
h
ese
class
i
f
ier
s
cla
s
s
i
f
ied
a
t
er
m
as p
o
s
iti
v
e
i
f
it
w
a
s
m
o
r
e
f
r
eq
u
en
t in
p
o
s
itiv
e
l
y
ta
g
g
ed
d
o
cu
m
e
n
ts
,
n
e
g
ati
v
e
if
m
o
r
e
f
r
eq
u
e
n
t
i
n
n
e
g
ati
v
el
y
tag
g
ed
d
o
cu
m
e
n
ts
a
n
d
n
e
u
tr
a
l
if
eq
u
a
ll
y
d
is
tr
ib
u
ted
.
Ne
u
tr
al
ter
m
s
m
i
g
h
t
n
o
t
h
av
e
e
x
ac
tl
y
eq
u
al
o
cc
u
r
r
en
ce
in
p
o
s
itiv
el
y
tag
g
ed
d
o
cu
m
e
n
ts
an
d
n
e
g
ati
v
el
y
ta
g
g
ed
d
o
cu
m
en
ts
.
E
v
e
n
i
f
t
h
e
ter
m
w
as
d
is
tr
ib
u
ted
w
it
h
a
s
lig
h
t
d
if
f
er
en
ce
,
t
h
e
ter
m
w
as
class
i
f
ied
.
A
ct
u
all
y
t
h
e
t
er
m
m
ig
h
t
b
e
n
e
u
tr
al.
C
las
s
i
f
ier
s
also
s
u
f
f
er
ed
d
u
e
t
o
h
ig
h
d
i
m
e
n
s
io
n
alit
y
if
p
r
ep
r
o
ce
s
s
in
g
w
as
n
o
t p
er
f
o
r
m
ed
.
As
t
h
e
tr
ain
i
n
g
d
ata
h
a
d
g
r
o
w
n
e
x
p
o
n
e
n
tiall
y
,
d
i
m
e
n
s
io
n
o
f
th
e
in
p
u
t
s
p
ac
e
h
a
d
also
s
p
lu
r
g
ed
.
Sen
ti
m
e
n
t
a
n
al
y
s
i
s
b
ei
n
g
a
s
p
ec
ialized
d
o
m
ain
o
f
tex
t
m
in
i
n
g
b
en
e
f
itted
a
f
ter
te
x
t
p
r
ep
r
o
ce
s
s
in
g
[
2
1
]
.
P
r
e
-
p
r
o
ce
s
s
in
g
i
n
cl
u
d
es
to
k
e
n
izi
n
g
,
eli
m
in
ati
n
g
tag
s
,
s
to
p
w
o
r
d
s
r
e
m
o
v
al,
d
is
ca
r
d
in
g
p
u
n
ct
u
atio
n
s
&
s
y
m
b
o
ls
,
s
te
m
m
i
n
g
&
le
m
m
atiza
tio
n
[
2
2
]
.
E
n
s
e
m
b
le
ap
p
r
o
ac
h
es
w
er
e
u
s
ed
to
id
en
tify
a
n
ap
p
r
o
p
r
i
a
te
co
m
b
i
n
atio
n
o
f
p
r
ep
r
o
ce
s
s
in
g
[2
3
]
.
P
u
n
ct
u
atio
n
m
ar
k
s
ar
e
i
m
p
o
r
tan
t
w
h
e
n
wr
itin
g
i
n
E
n
g
l
is
h
o
r
an
y
n
at
u
r
a
l la
n
g
u
ag
e
,
b
u
t
ar
e
h
ar
d
l
y
o
f
a
n
y
u
s
e
f
o
r
co
m
p
u
t
atio
n
al
l
in
g
u
is
tic
s
[
2
4
]
.
A
p
o
s
t
r
o
p
h
e
w
a
s
d
i
s
ca
r
d
ed
.
T
h
e
ter
m
s
w
it
h
ap
o
s
tr
o
p
h
e
w
o
u
ld
th
e
n
f
o
r
m
a
n
e
w
d
i
m
e
n
s
io
n
,
d
ef
ea
ti
n
g
t
h
e
p
u
r
p
o
s
e
o
f
d
i
m
e
n
s
io
n
a
lit
y
r
ed
u
c
tio
n
.
Fo
r
ex
a
m
p
le:
Af
ter
ap
o
s
tr
o
p
h
e
r
em
o
v
al
ter
m
d
id
n
’
t
w
a
s
m
ap
p
ed
to
d
id
n
t
.
T
h
e
ter
m
d
id
n
t
n
o
t
b
ein
g
a
w
o
r
d
in
E
n
g
l
is
h
,
m
ap
p
ed
to
a
n
e
w
d
i
m
en
s
io
n
.
T
h
er
e
w
as
n
’
t
a
n
y
s
ta
n
d
ar
d
s
to
p
w
o
r
d
s
lis
t
[
1
3
]
.
T
h
e
av
ailab
le
s
to
p
w
o
r
d
s
lis
t
d
id
n
’
t
i
n
cl
u
d
e
d
o
m
ai
n
-
s
p
ec
i
f
ic
s
to
p
w
o
r
d
s
.
D
is
ca
r
d
in
g
n
e
u
tr
al
ter
m
s
o
r
s
en
t
i
m
e
n
t sto
p
w
o
r
d
s
w
a
s
also
a
h
e
r
cu
lean
ta
s
k
.
T
r
ain
in
g
s
et
w
a
s
en
r
ic
h
ed
b
y
ad
d
in
g
a
n
to
n
y
m
s
o
f
t
h
e
ta
g
g
ed
r
ev
ie
w
s
i
n
ter
m
-
d
o
c
u
m
en
t
v
ec
to
r
s
o
f
o
p
p
o
s
ite
o
r
ien
tatio
n
[
2
5
]
.
N
eg
atio
n
h
an
d
li
n
g
w
as
n
o
t
p
er
f
o
r
m
ed
.
I
m
p
licit
a
n
d
ex
p
licit
n
eg
at
io
n
m
o
d
i
f
ier
s
w
er
e
h
an
d
led
u
s
i
n
g
a
t
h
r
ee
s
ta
g
e
m
o
d
el
[
2
6
]
.
T
h
e
p
o
la
r
ity
s
h
if
ter
ter
m
m
ig
h
t
n
o
t
m
o
d
i
f
y
t
h
e
t
er
m
ex
ac
tl
y
n
e
x
t
to
it.
T
h
e
o
th
er
d
r
aw
b
ac
k
w
a
s
th
e
av
ai
lab
ilit
y
o
f
ex
ac
t
an
to
n
y
m
.
S
co
p
e
o
f
lin
g
u
is
tic
n
e
g
ati
o
n
w
a
s
d
eter
m
i
n
ed
b
y
u
s
in
g
Dep
e
n
d
en
c
y
A
n
a
l
y
s
i
s
[
2
7
]
.
S
en
ten
ce
s
w
er
e
r
ep
r
esen
ted
in
f
o
r
m
o
f
p
ar
s
e
tr
ee
s
u
s
i
n
g
Sta
n
f
o
r
d
P
ar
s
er
[
2
8
]
.
I
n
itial
s
co
r
e
o
f
f
ea
t
u
r
e
d
escr
ip
to
r
s
w
er
e
b
ased
o
n
Se
n
tiW
o
r
d
Net.
9
3
.
7
5
%
o
f
th
e
s
y
n
o
n
y
m
o
u
s
s
ets
in
Sen
tiW
o
r
d
Net
ar
e
ig
n
o
r
ed
as t
h
e
y
h
av
e
a
s
tr
o
n
g
er
o
b
j
ec
tiv
e
t
en
d
en
c
y
th
a
n
p
o
s
iti
v
e
an
d
n
e
g
ativ
e
[
2
9
]
.
P
r
o
ce
s
s
in
g
s
tr
u
ct
u
r
ed
v
ec
to
r
s
w
a
s
ea
s
ier
an
d
s
y
s
te
m
a
tic
as
co
m
p
ar
ed
to
u
n
s
tr
u
c
tu
r
ed
tex
t
r
ev
ie
w
s
,
b
u
t
th
e
p
o
s
itio
n
al
in
f
o
r
m
atio
n
ca
r
r
ied
b
y
a
ter
m
w
as
n
o
t
co
n
s
id
er
ed
.
T
h
e
s
y
n
tactic
s
tr
u
ct
u
r
e
o
f
t
h
e
s
e
n
te
n
ce
p
lay
s
a
n
i
m
p
o
r
tan
t
r
o
le
i
n
s
e
n
ti
m
e
n
t
a
n
al
y
s
i
s
[
1
5
]
.
T
e
r
m
-
d
o
cu
m
e
n
t
v
ec
to
r
s
p
r
o
v
id
ed
t
h
e
c
o
u
n
t
o
f
t
h
e
ter
m
i
n
d
o
cu
m
en
t
b
u
t
t
h
e
ter
m
m
i
g
h
t
b
e
ass
o
ciate
d
w
it
h
o
n
e
o
r
m
o
r
e
ter
m
s
in
t
h
e
d
o
cu
m
en
t.
T
h
is
i
n
f
o
r
m
atio
n
w
as
n
o
t
r
ec
o
r
d
ed
w
h
e
n
r
ep
r
ese
n
ti
n
g
ter
m
s
a
s
d
o
cu
m
e
n
ts
.
T
h
e
ter
m
a
s
a
n
i
n
d
iv
id
u
al
m
a
y
ca
r
r
y
a
d
if
f
er
en
t
m
ea
n
i
n
g
th
an
w
h
at
it
m
ea
n
t
a
s
g
r
o
u
p
o
f
w
o
r
d
s
in
t
h
e
s
e
n
te
n
ce
s
tr
u
ct
u
r
e.
T
h
e
s
e
m
a
n
tic
s
tr
u
ct
u
r
es
o
f
th
e
w
o
r
d
s
i
n
th
e
s
en
te
n
ce
co
n
tr
ib
u
te
to
th
e
ac
t
u
al
m
ea
n
in
g
a
n
d
t
h
u
s
m
a
y
ad
j
o
in
f
o
r
s
en
ti
m
en
t
an
a
l
y
s
is
als
o
.
T
h
e
b
ag
-
of
-
w
o
r
d
s
b
ased
ap
p
r
o
ac
h
es f
ailed
to
ca
p
tu
r
e
th
e
s
y
n
tactic
a
n
d
s
e
m
a
n
ti
c
s
tr
u
ct
u
r
e
o
f
r
ev
ie
w
s
.
A
lt
h
o
u
g
h
s
u
p
er
v
i
s
ed
tec
h
n
iq
u
es
s
u
c
h
as
SVM,
NN
a
n
d
u
n
s
u
p
er
v
i
s
ed
ap
p
r
o
ac
h
es
s
u
c
h
a
s
le
x
ico
n
b
ased
ap
p
r
o
ac
h
es
ar
e
v
er
y
p
o
p
u
lar
,
in
te
lli
g
en
t
s
y
s
te
m
s
s
u
c
h
as
co
n
ce
p
t
b
a
s
ed
ap
p
r
o
ac
h
es
is
t
h
e
n
ee
d
o
f
th
e
h
o
u
r
[
1
]
.
1
.
1
.
Co
ntr
ibu
t
io
n
P
r
o
p
o
s
ed
m
o
d
el
atte
m
p
ts
to
g
o
b
e
y
o
n
d
tr
ad
itio
n
al
B
a
g
s
-
of
-
w
o
r
d
s
ap
p
r
o
ac
h
.
I
t
co
n
s
id
er
s
s
y
n
tac
tic
an
d
t
h
e
s
e
m
a
n
tic
s
tr
u
ctu
r
e
o
f
s
en
te
n
ce
s
i
n
r
ev
ie
w
s
.
T
er
m
s
i
n
a
r
ev
ie
w
w
er
e
cla
s
s
i
f
ied
a
s
p
o
s
itiv
e
o
r
n
e
g
ativ
e
b
ased
o
n
th
eir
p
o
s
itio
n
&
a
s
s
o
ciatio
n
w
it
h
o
th
er
ter
m
s
i
n
t
h
e
r
ev
ie
w
.
T
h
e
ass
o
ciatio
n
a
m
o
n
g
th
e
ter
m
s
an
d
t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2088
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
4
,
A
u
g
u
s
t
201
8
:
2
3
5
8
–
2
3
6
6
2360
p
o
s
itio
n
o
f
th
e
ter
m
i
n
th
e
s
e
n
ten
ce
w
as
co
n
s
id
er
ed
.
C
lass
i
f
i
er
s
w
er
e
d
esig
n
ed
to
h
an
d
le
n
eu
tr
al
ter
m
s
,
o
n
th
e
b
asis
o
f
r
elati
v
e
d
o
m
in
a
n
ce
.
E
x
is
ti
n
g
p
r
ep
r
o
ce
s
s
in
g
tec
h
n
iq
u
es
w
er
e
ex
ten
d
ed
to
h
an
d
le
ter
m
s
w
it
h
ap
o
s
tr
o
p
h
e.
R
u
le
s
w
er
e
s
et
to
h
an
d
le
ter
m
s
w
i
th
ap
o
s
tr
o
p
h
e
b
ef
o
r
e
d
is
ca
r
d
in
g
p
u
n
ct
u
atio
n
s
.
A
s
et
o
f
d
o
m
ai
n
s
p
ec
if
ic
r
u
le
s
f
o
r
th
e
m
o
v
ie
d
o
m
a
in
w
er
e
p
r
o
p
o
s
ed
in
th
e
d
ed
u
ctio
n
p
h
ase.
S
u
b
j
ec
tiv
it
y
e
x
tr
ac
tio
n
w
as
d
o
n
e
at
p
h
r
ase
lev
el.
R
ea
s
ea
r
ch
m
eth
o
d
f
o
r
p
r
o
p
o
s
ed
C
S
A
M
m
o
d
el
i
s
d
escr
ib
e
d
in
S
ec
ti
o
n
2
.
R
es
u
lt
s
ar
e
r
ep
o
r
ted
&
an
al
y
ze
d
i
n
Sectio
n
3
.
C
o
n
cl
u
s
io
n
&
f
u
t
u
r
e
s
co
p
e
ar
e
p
u
t f
o
r
th
in
S
ec
tio
n
4
.
2.
P
RO
P
O
SE
D
M
E
T
H
O
D
Fig
u
r
e
1
p
r
esen
ts
t
h
e
p
r
o
p
o
s
ed
C
o
n
ce
p
tu
al
S
en
t
i
m
e
n
t
An
al
y
s
is
Mo
d
el
(
C
S
A
M)
.
Fig
u
r
e
1
.
P
r
o
p
o
s
ed
c
o
n
ce
p
tu
al
s
en
ti
m
e
n
t a
n
a
l
y
s
is
m
o
d
el
2
.
1
.
P
re
pro
ce
s
s
ing
ph
a
s
e
T
h
e
w
o
r
d
s
w
i
th
ap
o
s
tr
o
p
h
e
s
u
ch
as
is
n
’
t
a
n
d
t
h
at
’
s
w
er
e
a
n
o
v
er
h
ea
d
i
n
ter
m
-
d
o
c
u
m
e
n
t
m
atr
i
x
,
a
s
illu
s
tr
ated
w
ith
a
n
ex
a
m
p
le
i
n
T
ab
le
1
.
Un
lik
e
tr
ad
itio
n
al
p
r
ep
r
o
ce
s
s
in
g
tec
h
n
iq
u
e
w
h
er
e
p
u
n
ct
u
a
tio
n
s
y
m
b
o
l
s
ar
e
d
is
ca
r
d
ed
,
a
s
et
o
f
r
u
les
to
h
an
d
le
w
o
r
d
s
w
it
h
ap
o
s
tr
o
p
h
e,
ar
e
p
r
o
p
o
s
ed
in
T
ab
le
2
.
A
f
ter
p
r
o
p
o
s
ed
ex
ten
s
io
n
o
f
h
a
n
d
li
n
g
ter
m
s
w
it
h
ap
o
s
tr
o
p
h
e,
tr
ad
itio
n
al
p
r
ep
r
o
ce
s
s
in
g
ta
s
k
o
f
d
is
ca
r
d
in
g
p
u
n
ctu
a
tio
n
s
an
d
o
th
er
s
y
m
b
o
ls
w
er
e
p
er
f
o
r
m
ed
.
T
ab
le
1
.
E
x
am
p
le
o
f
T
er
m
-
Do
cu
m
e
n
t M
atr
i
x
en
tr
ies
f
o
r
w
o
r
d
“
is
n
’
t”
b
ef
o
r
e
an
d
af
ter
h
a
n
d
lin
g
ap
o
s
tr
o
p
h
e
B
e
f
o
r
e
H
a
n
d
l
i
n
g
T
e
r
ms w
i
t
h
A
p
o
st
r
o
p
h
e
A
f
t
e
r
H
a
n
d
l
i
n
g
T
e
r
ms w
i
t
h
A
p
o
st
r
o
p
h
e
D
o
c
u
me
n
t
s
T
e
r
ms
D
o
c
u
me
n
t
s
T
e
r
ms
Is
n
o
t
i
sn
’
t
Is
n
o
t
D
o
c
u
me
n
t
1
1
3
-
D
o
c
u
me
n
t
1
1
3
D
o
c
u
me
n
t
2
-
1
1
D
o
c
u
me
n
t
2
1
2
D
o
c
u
me
n
t
3
1
1
-
D
o
c
u
me
n
t
3
1
1
T
ab
le
2
.
Set o
f
r
u
les f
o
r
Han
d
l
in
g
ter
m
s
w
ith
A
p
o
s
tr
o
p
h
e
No
R
u
l
e
Ex
a
mp
l
e
1
n
’
t
_
n
o
t
w
a
sn
’
t
w
a
s n
o
t
2
’s
_
i
s
t
h
a
t
’
s
t
h
a
t
i
s
3
’re
_
a
r
e
y
o
u
’
r
e
yo
u
a
r
e
4
’
v
e
_
h
a
v
e
t
h
e
y
’
v
e
t
h
e
y
h
a
v
e
5
’m
_
a
m
I
’
m
I
a
m
6
’d
_
w
o
u
l
d
t
h
e
y
’
d
t
h
e
y
w
o
u
l
d
7
’
l
l
_
w
i
l
l
y
o
u
’
l
l
y
o
u
w
i
l
l
8
’
e
m
_
t
h
e
m
mak
e
’
e
m
mak
e
t
h
e
m
9
i
n
’
_
i
n
g
f
r
i
n
g
g
i
n
’
f
r
i
g
g
i
n
g
N
o
t
e
:
S
y
mb
o
l
“
_
”
i
n
d
i
c
a
t
e
s s
p
a
c
e
2
.
2
.
P
o
la
rit
y
cla
s
s
if
ica
t
io
n pha
s
e
P
r
ep
r
o
ce
s
s
ed
ter
m
-
d
o
cu
m
e
n
t
m
atr
ices
w
er
e
p
r
o
v
id
ed
as
in
p
u
t
to
p
r
o
p
o
s
ed
Sen
ti
m
e
n
t
C
l
ass
i
f
ier
.
A
ter
m
w
as
clas
s
i
f
ied
as
p
o
s
itiv
e
if
it
w
a
s
d
o
m
i
n
an
t
i
n
p
o
s
itiv
el
y
tag
g
ed
r
ev
ie
w
s
&
v
ice
v
er
s
a
.
Do
m
in
a
n
c
y
o
f
a
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
C
o
n
ce
p
tu
a
l S
e
n
timen
t A
n
a
lysi
s
Mo
d
el
(
K
r
a
n
ti V
ith
a
l G
h
a
g
)
2361
ter
m
in
r
ev
ie
w
s
w
as
d
eter
m
i
n
ed
b
y
p
r
o
p
o
s
ed
C
lass
i
f
ier
s
.
Neg
atio
n
h
an
d
li
n
g
w
as
p
er
f
o
r
m
ed
at
co
n
ce
p
tu
a
l
lev
el.
W
eig
h
ted
R
e
lati
v
e
T
er
m
Fre
q
u
en
c
y
Se
n
ti
m
e
n
t C
lass
if
ier
(
W
R
T
FS
C
)
class
i
f
ied
ter
m
a
s
p
o
s
itiv
e
i
f
its
f
r
eq
u
en
c
y
i
n
p
o
s
iti
v
el
y
ta
g
g
ed
d
o
cu
m
en
t
s
w
as lar
g
er
th
a
n
in
n
eg
at
iv
el
y
ta
g
g
ed
d
o
cu
m
en
ts
&
v
ice
-
v
er
s
a
W
eig
h
ted
Av
er
ag
e
R
elati
v
e
T
er
m
Fre
q
u
e
n
c
y
Se
n
ti
m
e
n
t
C
las
s
i
f
ier
(
W
AR
T
FS
C
)
w
as
p
r
o
p
o
s
ed
to
o
v
er
co
m
e
th
e
d
r
a
w
b
ac
k
o
f
W
R
T
FS
C
.
As
W
R
T
FS
C
w
as
b
a
s
ed
o
n
f
r
eq
u
en
c
y
co
u
n
t,
it
f
ai
ls
to
h
an
d
le
b
iased
d
ata.
B
iased
d
ata
t
y
p
icall
y
i
n
c
lu
d
es
r
e
v
ie
w
s
w
it
h
i
m
p
o
r
tan
t
t
er
m
s
r
ep
ea
ted
to
o
m
a
n
y
ti
m
es
.
T
o
o
v
er
co
m
e
t
h
is
d
r
a
w
b
ac
k
,
a
ter
m
w
as
cla
s
s
i
f
ied
as
p
o
s
itiv
e
if
it
s
av
er
a
g
e
f
r
eq
u
en
c
y
i
n
p
o
s
itiv
el
y
ta
g
g
ed
d
o
cu
m
e
n
t
s
w
a
s
lar
g
er
th
a
n
its
a
v
er
ag
e
f
r
eq
u
en
c
y
in
n
eg
a
tiv
e
l
y
ta
g
g
ed
d
o
cu
m
en
ts
a
n
d
v
ice
-
v
er
s
a.
W
eig
h
ted
Se
n
t
i
m
e
n
t
T
er
m
Fre
q
u
en
c
y
I
n
v
er
s
e
Do
cu
m
e
n
t
F
r
eq
u
en
c
y
(
W
Se
n
ti
-
T
FID
F)
w
o
r
k
s
o
n
t
h
e
p
r
in
cip
le
o
f
r
elativ
e
T
er
m
Fr
eq
u
en
c
y
I
n
v
er
s
e
Do
cu
m
e
n
t
Fre
q
u
en
c
y
(
T
FID
F)
o
f
a
ter
m
ac
r
o
s
s
p
o
s
iti
v
el
y
tag
g
ed
d
o
cu
m
en
ts
a
n
d
n
eg
at
i
v
el
y
ta
g
g
ed
d
o
cu
m
e
n
ts
.
A
ter
m
w
as
cla
s
s
i
f
ied
as
p
o
s
it
i
v
e
i
f
its
T
FID
F
ac
r
o
s
s
p
o
s
itiv
el
y
tag
g
ed
d
o
cu
m
e
n
t
s
w
a
s
lar
g
er
th
a
n
it
s
T
FID
F a
cr
o
s
s
n
e
g
ati
v
el
y
ta
g
g
ed
d
o
cu
m
e
n
ts
a
n
d
v
ice
-
v
er
s
a
.
T
er
m
w
o
u
ld
b
e
clas
s
i
f
ied
a
s
n
e
u
tr
al
ter
m
i
f
it
s
d
o
m
i
n
a
n
ce
i
n
p
o
s
iti
v
el
y
&
n
e
g
ati
v
el
y
ta
g
g
ed
d
o
cu
m
en
ts
w
as
eq
u
al.
C
o
n
v
er
s
el
y
ev
e
n
i
f
d
o
m
i
n
an
ce
v
ar
ied
s
li
g
h
tl
y
,
ter
m
w
o
u
ld
b
e
c
lass
i
f
ied
as
p
o
s
iti
v
e
o
r
n
eg
at
iv
e.
I
d
ea
ll
y
ter
m
s
h
o
u
ld
h
av
e
b
ee
n
cla
s
s
i
f
ied
as
n
e
u
tr
a
l.
T
o
av
o
id
b
iased
class
if
ica
ti
o
n
,
a
w
i
n
d
o
w
w
as
p
r
o
v
id
ed
d
ef
in
ed
b
y
±
Neu
tr
al
T
e
r
m
W
in
d
o
w
B
o
u
n
d
ar
y
(
NT
W
B
)
as
s
h
o
w
n
in
F
i
g
u
r
e
2
f
o
r
h
an
d
lin
g
n
eu
tr
al
ter
m
s
.
P
o
l
a
r
i
t
y
t
=
N
e
g
a
t
i
v
e
P
o
l
a
r
i
t
y
t
=
P
o
l
a
r
i
t
y
t
=
P
o
si
t
i
v
e
N
e
u
t
r
a
l
-
N
T
W
B
0
N
T
W
B
Fig
u
r
e
2
.
Sen
ti
m
e
n
t Cl
a
s
s
i
f
ica
tio
n
b
ased
o
n
P
o
lar
ity
v
alu
e
with
w
i
n
d
o
w
f
o
r
n
e
u
tr
al
w
o
r
d
s
.
T
h
at
is
if
t
h
e
P
o
lar
it
y
v
al
u
e
o
f
a
ter
m
w
a
s
b
et
w
ee
n
–
NT
W
B
an
d
NT
W
B
,
th
e
ter
m
w
a
s
c
lass
i
f
ied
as
n
eu
tr
al.
I
f
P
o
lar
it
y
t
v
alu
e
w
as
g
r
ea
ter
th
a
n
NT
W
B
th
en
t
h
e
ter
m
w
as
clas
s
if
ied
as
p
o
s
iti
v
e.
C
o
n
v
er
s
el
y
,
i
f
P
o
lar
ity
t
v
al
u
e
w
as
les
s
er
t
h
an
−N
T
W
B
th
en
th
e
ter
m
w
as
cl
ass
i
f
ied
as
n
e
g
ati
v
e.
Mo
r
e
w
o
r
d
s
w
er
e
cla
s
s
i
f
ie
d
as
n
eu
tr
al
i
f
NT
W
B
w
a
s
lar
g
e
r
,
r
esu
lti
n
g
to
less
er
n
u
m
b
er
o
f
o
p
in
io
n
ated
w
o
r
d
s
.
C
o
n
v
er
s
el
y
i
f
NT
W
B
v
alu
e
w
a
s
a
s
m
al
ler
v
al
u
e
n
e
u
tr
al
w
o
r
d
s
w
o
u
ld
n
o
t
b
e
ap
p
r
o
p
r
i
atel
y
id
e
n
ti
f
ied
.
E
i
th
er
co
n
d
it
io
n
w
o
u
ld
a
f
f
ec
t
t
h
e
ac
cu
r
ac
y
o
f
t
h
e
clas
s
i
f
ier
s
.
So
an
o
p
ti
m
al
NT
W
B
v
alu
e
f
o
r
ea
ch
Sen
ti
m
en
t
C
las
s
i
f
i
ca
tio
n
Mo
d
el
w
a
s
ex
p
er
i
m
e
n
tall
y
d
eter
m
i
n
ed
to
m
ax
i
m
ize
ac
cu
r
ac
y
.
T
ab
le
3
.
P
r
o
p
o
s
ed
an
d
tr
ad
itio
n
al
Se
n
ti
m
en
t Cl
a
s
s
i
f
ier
Mo
d
els
S
r
.
N
o
.
S
e
n
t
i
me
n
t
C
l
a
ss
i
f
i
e
r
C
l
a
ssi
f
i
c
a
t
i
o
n
C
r
i
t
e
r
i
a
f
o
r
t
e
r
m
B
a
se
d
o
n
1
T
S
C
T
r
a
d
i
t
i
o
n
a
l
S
e
n
t
i
me
n
t
C
l
a
ssi
f
i
e
r
[
1
6
]
c
t
d
c
t
d
N
P
M
a
x
,
F
r
e
q
u
e
n
c
y
C
o
u
n
t
2
D
e
l
t
a
-
TFI
D
F
D
e
l
t
a
-
TF
I
D
F
S
e
n
t
i
me
n
t
C
l
a
ssi
f
i
e
r
[
1
9
]
t
t
c
td
c
td
N
P
N
P
2
l
o
g
R
e
l
a
t
i
v
e
P
r
e
se
n
c
e
C
o
u
n
t
3
P
r
o
p
o
se
d
WR
TF
S
C
W
e
i
g
h
t
e
d
R
e
l
a
t
i
v
e
T
e
r
m
F
r
e
q
u
e
n
c
y
S
e
n
t
i
me
n
t
C
l
a
ss
i
f
i
e
r
001
.
0
001
.
0
l
o
g
c
td
c
td
e
N
P
R
e
l
a
t
i
v
e
F
r
e
q
u
e
n
c
y
C
o
u
n
t
4
P
r
o
p
o
se
d
W
A
R
TF
S
C
W
e
i
g
h
t
e
d
A
v
e
r
a
g
e
R
e
l
a
t
i
v
e
T
e
r
m
F
r
e
q
u
e
n
c
y
S
e
n
t
i
me
n
t
C
l
a
ss
i
f
i
e
r
001
.
0
001
.
0
l
og
t
c
t
d
t
c
t
d
e
N
N
P
P
R
e
l
a
t
i
v
e
A
v
e
r
a
g
e
F
r
e
q
u
e
n
c
y
C
o
u
n
t
5
P
r
o
p
o
se
d
W
S
e
n
t
i
-
TFI
D
F
W
e
i
g
h
t
e
d
S
e
n
t
i
me
n
t
T
e
r
m F
r
e
q
u
e
n
c
y
I
n
v
e
r
se
D
o
c
u
me
n
t
F
r
e
q
u
e
n
c
y
S
e
n
t
i
m
e
n
t
C
l
a
ssi
f
i
e
r
001
.
0
l
o
g
001
.
0
l
o
g
l
o
g
N
N
N
P
P
P
t
e
c
t
d
t
e
c
t
d
e
R
e
l
a
t
i
v
e
TFI
D
F
v
a
l
u
e
s o
f
t
e
r
ms
w
h
e
r
e
,
P
c
t
d
=
F
r
e
q
u
e
n
c
y
o
f
t
e
r
m t
i
n
p
o
si
t
i
v
e
l
y
t
a
g
g
e
d
d
o
c
u
me
n
t
s.
P
t
=
c
o
u
n
t
o
f
p
o
s
i
t
i
v
e
l
y
t
a
g
g
e
d
d
o
c
u
me
n
t
s w
i
t
h
t
e
r
m
t
.
P
=
T
o
t
a
l
N
u
mb
e
r
o
f
p
o
si
t
i
v
e
l
y
t
a
g
g
e
d
d
o
c
u
me
n
t
s.
N
c
t
d
=
F
r
e
q
u
e
n
c
y
o
f
t
e
r
m t
i
n
n
e
g
a
t
i
v
e
l
y
t
a
g
g
e
d
d
o
c
u
me
n
t
s
.
N
t
=
c
o
u
n
t
o
f
n
e
g
a
t
i
v
e
l
y
t
a
g
g
e
d
d
o
c
u
me
n
t
s w
i
t
h
t
e
r
m
t
.
N
=
T
o
t
a
l
N
u
mb
e
r
o
f
n
e
g
a
t
i
v
e
l
y
t
a
g
g
e
d
d
o
c
u
me
n
t
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2088
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
4
,
A
u
g
u
s
t
201
8
:
2
3
5
8
–
2
3
6
6
2362
NT
W
B
v
alu
es
f
o
r
a
ll
3
p
r
o
p
o
s
ed
class
i
f
ier
s
w
er
e
ex
p
er
i
m
e
n
tall
y
d
eter
m
i
n
ed
.
T
h
e
class
if
ied
ter
m
s
w
er
e
u
s
ed
in
C
o
n
ce
p
t
b
ased
Ded
u
ctio
n
P
h
a
s
e.
T
h
e
m
at
h
e
m
atica
l
m
o
d
els
o
f
t
h
e
p
r
o
p
o
s
ed
class
i
f
ier
an
d
t
h
o
s
e
w
h
ic
h
w
er
e
ex
p
er
i
m
e
n
ted
f
o
r
co
m
p
ar
ati
v
e
an
al
y
s
is
ar
e
s
u
m
m
ar
ized
in
T
ab
le
3.
2
.
3
.
Co
ncept
ba
s
ed
ded
uct
io
n ph
a
s
e
B
ag
-
of
-
w
o
r
d
s
ap
p
r
o
ac
h
ass
u
m
es
t
h
at
p
o
s
itio
n
s
o
f
w
o
r
d
s
i
n
r
ev
ie
w
s
ar
e
n
’
t
i
m
p
o
r
tan
t
.
I
t
also
f
ails
to
ca
p
tu
r
e
ass
o
ciatio
n
w
it
h
i
n
te
r
m
s
.
T
r
ad
itio
n
al
ap
p
r
o
ac
h
es
th
u
s
ig
n
o
r
e
s
y
n
tactic
&
s
e
m
an
tic
s
tr
u
ctu
r
e
o
f
r
ev
ie
w
s
.
C
o
n
s
id
er
th
e
r
ev
ie
w
s
:
R
e
v
ie
w
1
:
B
o
r
in
g
a
n
d
n
o
t in
teres
tin
g
.
R
ev
ie
w
2
:
I
n
teres
tin
g
a
n
d
n
o
t
b
o
r
in
g
.
A
lt
h
o
u
g
h
b
o
th
o
f
ab
o
v
e
m
e
n
ti
o
n
ed
r
ev
ie
w
s
m
ea
n
ex
ac
tl
y
t
h
e
o
p
p
o
s
ite,
s
tatis
tical
m
eth
o
d
w
o
u
ld
tr
ea
t
th
e
m
i
n
th
e
s
a
m
e
w
a
y
.
B
o
th
t
h
e
r
ev
ie
w
s
w
il
l
b
e
r
ep
r
esen
te
d
as
v
ec
to
r
s
.
E
ac
h
w
o
r
d
i
n
th
e
r
ev
ie
w
w
o
u
ld
b
e
r
ec
o
r
d
e
d
as a
n
ele
m
e
n
t i
n
th
e
t
er
m
-
d
o
cu
m
e
n
t
v
ec
to
r
w
i
th
its
f
r
eq
u
en
c
y
.
B
o
t
h
v
ec
to
r
s
w
i
ll b
e
id
en
tical,
t
h
o
u
g
h
th
e
y
ar
e
o
f
d
i
f
f
er
e
n
t
o
r
ien
tat
io
n
.
T
h
u
s
s
tati
s
tical
cla
s
s
i
f
ie
r
s
f
ail
to
ca
p
tu
r
e
th
e
ir
co
r
r
ec
t
o
r
ien
tatio
n
.
Th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
es
h
a
n
d
le
th
e
is
s
u
e
s
r
elate
d
to
s
y
n
tacti
c
&
s
e
m
a
n
tic
s
tr
u
ct
u
r
e
o
f
s
e
n
ten
ce
s
.
L
an
g
u
ag
e
g
r
a
m
m
ar
ad
d
s
m
ea
n
i
n
g
to
t
h
e
te
x
t.
Gr
a
m
m
atica
l
s
tr
u
ctu
r
e
w
as
e
x
p
lo
r
ed
u
s
i
n
g
Sta
n
f
o
r
d
P
ar
s
er
[
3
0
]
.
I
t
r
etu
r
n
ed
d
ep
en
d
en
cies
a
m
o
n
g
ter
m
s
i
n
r
ev
ie
w
s
.
T
h
e
p
r
o
p
o
s
ed
co
n
ce
p
t
b
ased
ap
p
r
o
ac
h
th
en
id
e
n
ti
f
ied
th
a
t
n
eg
at
io
n
m
o
d
i
f
ier
not
is
ass
o
ciate
d
w
it
h
ter
m
b
o
r
in
g
in
f
ir
s
t
r
ev
ie
w
t
h
u
s
it
m
o
d
i
f
ies
t
h
e
o
r
ien
tatio
n
o
f
ter
m
b
o
r
in
g
.
W
h
ile
class
i
f
y
in
g
r
ev
ie
w
,
th
e
p
o
lar
it
y
o
f
ter
m
b
o
r
in
g
w
a
s
r
ev
er
s
ed
&
s
u
m
m
ed
w
i
th
t
h
e
t
er
m
in
teres
tin
g
,
to
co
m
p
u
te
t
h
e
co
r
r
ec
t
o
r
ien
tatio
n
o
f
r
ev
ie
w
1
.
Si
m
ilar
l
y
r
e
v
ie
w
2
co
u
ld
also
b
e
ap
p
r
o
p
r
iately
class
i
f
ied
.
I
n
s
tead
o
f
s
u
m
m
i
n
g
p
o
lar
ities
o
f
ter
m
in
r
ev
ie
w
s
,
t
h
e
y
w
er
e
ad
d
ed
b
y
co
n
s
id
er
in
g
s
y
n
tactica
l
s
tr
u
ct
u
r
e
o
f
th
e
ter
m
s
i
n
a
r
ev
i
e
w
.
C
o
n
s
id
er
an
o
th
er
r
ev
ie
w
:
A
v
is
u
all
y
f
las
h
y
b
u
t
n
ar
r
ati
v
el
y
o
p
aq
u
e
v
ap
id
ex
er
cise
in
s
t
y
le
&
m
y
s
ti
f
icatio
n
.
A
clas
s
i
f
ier
th
at
f
o
llo
w
s
B
o
W
ap
p
r
o
ac
h
w
o
u
ld
s
u
m
t
h
e
w
ei
g
h
ts
o
f
all
ter
m
s
to
co
m
p
u
te
r
ev
ie
w
p
o
lar
ity
.
Actu
a
ll
y
t
h
e
co
n
j
u
n
c
tio
n
“
b
u
t”
n
o
ti
f
ie
s
t
h
at
o
n
l
y
la
ter
p
ar
t
o
f
s
e
n
te
n
ce
s
h
o
u
ld
b
e
co
n
s
id
er
ed
&
p
r
io
r
s
h
o
u
ld
b
e
ig
n
o
r
ed
.
T
r
ad
itio
n
al
ap
p
r
o
ac
h
es
f
ail
to
h
an
d
le
t
h
ese
s
e
m
a
n
tic
r
elatio
n
s
h
ip
s
.
Ad
j
ec
tiv
es
&
ad
v
er
b
s
also
m
o
d
if
y
th
e
i
n
te
n
s
it
y
o
f
ass
o
ciate
d
ter
m
s
.
T
h
e
y
s
h
o
u
l
d
n
o
t
b
e
s
im
p
l
y
s
u
m
m
ed
as
in
B
o
W
ap
p
r
o
ac
h
.
S
e
m
a
n
tic
s
tr
u
ctu
r
e
o
f
r
ev
ie
w
w
as
s
y
s
te
m
atica
ll
y
h
an
d
led
in
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
.
Gr
am
m
atica
l
r
u
le
s
w
er
e
ex
p
lo
r
ed
to
h
an
d
le
s
e
m
an
tic
s
tr
u
ctu
r
e
o
f
r
ev
ie
w
s
.
Sta
n
f
o
r
d
P
ar
s
er
r
etu
r
n
s
d
ep
en
d
en
cie
s
a
m
o
n
g
t
h
e
ter
m
s
i
n
r
ev
ie
w
.
P
o
lar
ities
o
f
th
e
ter
m
w
er
e
ag
g
r
eg
a
ted
b
y
co
n
s
id
er
in
g
t
h
e
as
s
o
ciatio
n
o
f
t
h
e
ter
m
s
i
n
a
r
ev
ie
w
w
it
h
o
th
er
ter
m
s
i
n
t
h
e
s
a
m
e
r
e
v
ie
w
.
Dep
e
n
d
en
cie
s
w
er
e
tr
ea
te
d
to
ca
p
tu
r
e
th
e
co
r
r
ec
t
s
e
n
ti
m
en
t
i
n
t
h
e
r
ev
ie
w
.
C
o
n
j
u
n
c
tio
n
s
,
ad
v
er
b
m
o
d
i
f
ier
s
an
d
n
e
g
atio
n
m
o
d
if
ier
d
ep
en
d
en
cies
w
er
e
tr
ea
ted
in
d
iv
i
d
u
a
ll
y
.
Se
m
a
n
tic
s
tr
u
ct
u
r
e
o
f
ex
a
m
p
l
e
r
ev
ie
w
w
a
s
h
an
d
led
as
in
F
i
g
u
r
e
3
.
Dete
r
m
i
n
an
t
A
w
as
cl
ass
i
f
ied
as
s
to
p
w
o
r
d
s
.
S
ub
-
tr
ee
b
ef
o
r
e
c
o
n
j
u
n
ctio
n
B
UT
&
co
n
j
u
n
c
ti
o
n
B
UT
w
as
d
i
s
ca
r
d
ed
.
Po
lar
ities
o
f
ter
m
s
i
n
ad
j
ec
tiv
e
p
h
r
ase,
w
it
h
ad
v
er
b
,
N
A
R
R
A
TIV
E
LY
&
ad
j
ec
ti
v
e,
OP
A
QUE
w
a
s
m
u
ltip
lie
d
.
T
h
is
b
r
an
ch
w
a
s
s
u
m
m
ed
w
it
h
n
o
u
n
E
X
E
R
C
I
S
E
an
d
ad
d
ed
to
p
o
lar
it
y
o
f
r
ig
h
t
s
u
b
-
tr
ee
.
I
n
r
i
g
h
t
s
u
b
-
tr
ee
IN
w
a
s
cla
s
s
i
f
ied
a
s
s
to
p
w
o
r
d
s
.
C
o
n
j
u
n
ct
io
n
A
N
D
w
a
s
i
g
n
o
r
ed
.
P
o
lar
ities
o
f
as
s
o
ciate
d
n
o
u
n
p
h
r
ases
S
TY
LE
&
MY
S
TIF
I
C
A
TIO
N
w
er
e
ad
d
ed
.
Fig
u
r
e
3
.
Gr
a
m
m
at
ical
d
ep
en
d
en
cies
f
o
r
th
e
ex
a
m
p
le
r
ev
ie
w
u
s
i
n
g
Sta
n
f
o
r
d
P
ar
s
er
A
p
ar
t
f
r
o
m
h
a
n
d
li
n
g
s
y
n
tact
ic
an
d
s
e
m
an
t
ic
s
tr
u
ctu
r
e
o
f
r
ev
ie
w
s
,
Se
n
tiW
o
r
d
Net
an
d
Do
m
ain
Sp
ec
if
ic
R
u
le
s
w
er
e
also
i
n
co
r
p
o
r
ate
d
.
T
o
h
an
d
le
r
ar
e
ca
s
e
o
f
th
e
n
e
w
ter
m
s
t
h
at
ar
e
n
o
t
p
r
esen
t
in
tr
ai
n
in
g
s
et,
Sen
tiW
o
r
d
Net
w
a
s
u
s
ed
.
I
f
a
ter
m
w
as
p
r
ese
n
t
in
le
x
ico
n
as
w
ell
as
Se
n
tiW
o
r
d
Net,
th
e
p
o
lar
it
y
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
C
o
n
ce
p
tu
a
l S
e
n
timen
t A
n
a
lysi
s
Mo
d
el
(
K
r
a
n
ti V
ith
a
l G
h
a
g
)
2363
d
o
m
ai
n
s
p
ec
if
ic
lex
ico
n
w
a
s
u
t
ilized
.
A
lt
h
o
u
g
h
o
n
l
y
n
ea
r
l
y
7
%
o
f
th
e
w
o
r
d
s
i
n
Sen
tiW
o
r
d
Net
ar
e
o
p
in
io
n
ated
,
s
o
m
eti
m
e
s
th
e
s
e
w
er
e
also
u
s
ef
u
l
f
o
r
Sen
t
i
m
e
n
t
C
la
s
s
i
f
icatio
n
.
P
o
lar
ity
o
f
th
e
ter
m
s
i
n
co
n
s
tr
u
cted
d
o
m
ai
n
-
s
p
ec
i
f
ic
le
x
ico
n
w
a
s
n
o
r
m
al
ized
to
ad
ap
t
w
it
h
t
h
e
Sen
tiW
o
r
d
Net.
Do
m
ain
s
p
ec
if
ic
r
u
les
f
o
r
Mo
v
ie
d
o
m
a
in
w
er
e
g
e
n
er
a
ted
an
d
ex
p
an
d
ed
to
im
p
r
o
v
e
o
v
er
all
Sen
ti
m
e
n
t
C
las
s
i
f
icatio
n
ac
c
u
r
ac
y
.
R
e
v
ie
w
er
s
g
e
n
er
all
y
r
e
m
ar
k
s
i
n
a
r
ev
ie
w
w
it
h
t
h
eir
f
in
al
o
p
in
io
n
ab
o
u
t t
h
e
m
o
v
ie.
T
h
e
o
p
in
io
n
ex
p
r
ess
ed
ab
o
u
t
m
o
v
ie
co
u
ld
b
e
p
o
s
itiv
e
o
r
n
eg
ativ
e.
T
w
o
s
ets
o
f
r
u
l
e
b
ased
class
if
ier
s
w
er
e
d
esi
g
n
ed
to
class
if
y
t
h
e
s
en
ti
m
e
n
ts
.
I
n
f
ir
s
t
s
ets,
v
ar
io
u
s
r
u
les
th
at
ar
e
r
elate
d
to
ter
m
Mo
v
ie
an
d
all
its
s
y
n
o
n
y
m
s
to
w
o
r
d
s
w
ith
p
o
s
itiv
e
o
r
ien
ta
tio
n
li
k
e
g
o
o
d
,
g
r
ea
t
an
d
h
ap
p
y
w
er
e
lis
ted
.
Si
m
i
lar
ter
m
s
w
er
e
ex
p
lo
r
ed
u
s
i
n
g
W
o
r
d
Net
[
1
1
]
.
I
f
a
r
ev
ie
w
s
ati
s
f
ie
d
an
y
o
f
th
e
co
n
d
itio
n
i
n
t
h
is
it
w
o
u
l
d
b
e
class
if
ied
a
s
p
o
s
itiv
e.
T
h
e
s
ec
o
n
d
s
et
w
as d
esig
n
ed
to
class
i
f
y
a
r
e
v
ie
w
as
n
eg
ati
v
e.
T
h
e
o
b
j
ec
t
p
ar
t
o
f
tex
t
r
ev
ie
w
,
w
h
ich
i
s
th
e
v
er
b
p
h
r
ase
o
f
s
en
te
n
ce
,
ca
r
r
ies
m
o
r
e
o
p
in
io
n
ated
in
f
o
r
m
atio
n
a
s
co
m
p
ar
ed
to
s
u
b
j
e
ct
p
ar
t o
r
th
e
n
o
u
n
p
h
r
ase
o
f
th
e
s
a
m
e
s
en
ten
ce
.
C
o
n
s
id
er
th
e
r
ev
ie
w
.
Th
e
mo
vi
e
w
a
s
g
o
o
d
.
Th
e
mo
vie
s
u
b
j
ec
t p
ar
t
in
f
o
r
m
ati
v
e
w
a
s
g
o
o
d
o
b
j
ec
t p
a
r
t
o
p
in
io
n
ated
C
o
n
s
id
er
in
g
t
h
i
s
s
u
b
j
ec
tiv
e
&
o
b
j
ec
tiv
e
co
m
p
o
s
it
io
n
o
f
a
s
tate
m
e
n
t,
s
u
b
j
ec
tiv
it
y
e
x
tr
ac
t
io
n
w
as
p
er
f
o
r
m
ed
at
p
h
r
ase
lev
e
l.
T
h
e
n
o
u
n
p
h
r
ase
w
as
ig
n
o
r
ed
&
t
h
e
p
o
lar
it
y
f
o
r
th
e
r
est
o
f
th
e
r
e
v
ie
w
w
a
s
co
m
p
u
ted
.
No
u
n
P
h
r
ase
(
NP
)
,
THE
MOVI
E
w
as
i
g
n
o
r
ed
.
T
h
e
p
o
lar
ity
w
as
co
m
p
u
ted
f
o
r
th
e
Ver
b
P
h
r
ase
(
VP
)
W
A
S
GOOD
.
2
.
4
.
E
x
peri
m
e
nta
l s
et
up
P
r
o
p
o
s
ed
C
o
n
ce
p
tu
al
Sen
ti
m
e
n
t
A
n
al
y
s
is
Mo
d
el
(
C
S
A
M)
m
o
d
el
w
a
s
ev
al
u
ated
o
n
P
an
g
&
L
ee
’
s
,
Kag
g
le’
s
&
UC
I
’
s
m
o
v
ie
d
ata
s
ets.
P
r
o
p
o
s
ed
class
if
ier
,
b
ein
g
d
o
m
ai
n
in
d
ep
en
d
e
n
t,
w
er
e
e
v
alu
a
ted
on
m
o
v
ie
as
w
ell
as
r
esta
u
r
an
t
s
a
n
d
p
r
o
d
u
ct
d
ataset.
P
an
g
an
d
L
ee
’
s
Mo
v
ie
R
ev
ie
w
Data
s
e
t
co
n
tai
n
s
2000
p
o
s
itiv
el
y
&
n
eg
at
iv
el
y
ta
g
g
ed
tex
t
d
o
cu
m
en
ts
[
1
7
]
.
Kag
g
le
’
s
B
a
g
o
f
W
o
r
d
s
m
ee
ts
B
a
g
o
f
P
o
p
co
r
n
s
d
ataset
co
n
tain
s
2
5
0
0
0
p
o
s
itiv
el
y
&
n
e
g
ati
v
e
l
y
tag
g
ed
r
ev
ie
w
s
[
3
1
]
.
Du
e
to
co
m
p
u
tatio
n
a
l
li
m
i
tatio
n
ex
p
er
i
m
en
t
s
w
er
e
p
er
f
o
r
m
ed
o
n
Kag
g
le
’
s
s
u
b
s
e
t
o
f
1
8
0
0
p
o
s
itiv
el
y
&
n
e
g
ati
v
el
y
ta
g
g
ed
r
ev
ie
w
s
.
UC
I
’
s
S
en
ti
m
en
t
L
ab
elled
Mo
v
ie,
R
esta
u
r
an
t
&
P
r
o
d
u
ct
d
atasets
in
d
i
v
id
u
a
ll
y
co
n
s
i
s
ts
o
f
1000
r
ev
ie
w
s
ta
g
g
ed
p
o
s
iti
v
e
&
n
eg
a
tiv
e
[
3
2
]
.
I
f
t
h
e
lab
el
o
f
th
e
r
ev
ie
w
a
n
d
th
e
cla
s
s
i
f
icat
io
n
o
u
tco
m
e
w
a
s
s
a
m
e,
th
e
n
i
t
co
n
tr
ib
u
ted
to
ac
cu
r
ac
y
.
Oth
er
w
i
s
e
t
h
e
o
u
tco
m
e
ad
d
ed
to
er
r
o
r
r
ate.
A
cc
u
r
ac
y
w
a
s
co
m
p
u
ted
u
s
in
g
,
1
0
Fo
ld
C
r
o
s
s
Valid
atio
n
(
1
0
f
o
ld
C
V)
[
2
1
]
.
I
n
s
tead
o
f
e
v
al
u
at
in
g
t
h
e
C
o
n
ce
p
tu
al
Se
n
ti
m
e
n
t
A
n
al
y
s
is
Mo
d
el
(
C
S
A
M)
a
s
a
w
h
o
le,
it
w
as
ev
alu
a
ted
in
i
n
cr
e
m
e
n
ta
l
m
an
n
er
in
e
v
er
y
ex
p
er
i
m
e
n
t.
E
x
p
er
i
m
e
n
t
1
w
as
p
er
f
o
r
m
ed
to
d
eter
m
i
n
e
th
e
o
p
ti
m
al
v
alu
es
o
f
Neu
tr
al
T
er
m
W
in
d
o
w
B
o
u
n
d
ar
y
(
NT
W
B
)
p
ar
am
eter
f
o
r
p
r
o
p
o
s
ed
W
R
T
FS
C
,
W
AR
T
FS
C
an
d
W
Sen
t
iT
FID
F
class
if
ier
s
.
T
h
e
NT
W
B
v
alu
e
w
a
s
v
ar
ied
b
et
w
ee
n
,
0
to
±
2
in
s
tep
o
f
0
.
1
.
A
cc
u
r
ac
y
w
as
c
o
m
p
u
ted
at
ea
ch
s
tep
.
M
in
i
m
u
m
v
a
lu
e
f
o
r
NT
W
B
p
ar
am
eter
t
h
at
y
ield
ed
m
ax
i
m
u
m
ac
c
u
r
ac
y
w
a
s
s
et
f
o
r
th
e
r
esp
ec
tiv
e
clas
s
i
f
ier
s
i
n
all
f
u
r
t
h
er
ex
p
er
i
m
en
t
s
.
E
x
p
er
i
m
e
n
t
2
w
as
p
er
f
o
r
m
e
d
to
ev
alu
ate
p
r
o
p
o
s
ed
class
if
ier
s
.
A
cc
u
r
ac
y
w
as
co
m
p
u
t
ed
f
o
r
th
e
T
r
a
d
itio
n
al
Sen
ti
m
e
n
t
C
la
s
s
i
f
i
er
(
T
SC
)
,
a
co
m
p
ar
ab
le
ter
m
w
ei
g
h
ti
n
g
cla
s
s
i
f
ier
(
Delta
-
T
FID
F)
&
p
r
o
p
o
s
ed
W
R
T
FS
C
,
W
AR
T
FS
C
an
d
W
Sen
ti
-
T
FID
F
class
if
ier
s
.
E
x
p
er
i
m
e
n
t
3
w
as
p
er
f
o
r
m
ed
to
ev
alu
ate
p
r
o
p
o
s
ed
ex
ten
s
io
n
s
to
e
x
is
t
in
g
p
r
ep
r
o
ce
s
s
in
g
t
ec
h
n
iq
u
es.
T
r
a
d
itio
n
al
p
r
ep
r
o
ce
s
s
in
g
tec
h
n
iq
u
es
o
f
d
is
ca
r
d
in
g
p
u
n
ct
u
a
tio
n
s
&
o
th
er
s
y
m
b
o
ls
w
er
e
ap
p
lied
o
n
p
r
o
p
o
s
ed
class
i
f
ier
s
.
T
h
e
p
r
o
p
o
s
ed
ex
te
n
s
io
n
to
h
an
d
le
ter
m
s
w
it
h
ap
o
s
tr
o
p
h
e
w
a
s
ap
p
lied
.
A
cc
u
r
a
c
y
w
as
co
m
p
u
ted
f
o
r
th
e
tr
ad
itio
n
al
p
r
ep
r
o
ce
s
s
in
g
t
ec
h
n
iq
u
es a
n
d
th
e
p
r
o
p
o
s
ed
ex
ten
s
io
n
to
th
e
e
x
is
ti
n
g
p
r
ep
r
o
c
ess
i
n
g
tec
h
n
iq
u
e
s
.
E
x
p
er
i
m
e
n
t 4
w
a
s
p
er
f
o
r
m
ed
t
o
ev
alu
ate
P
r
o
p
o
s
ed
C
o
n
ce
p
tu
al
Sen
ti
m
e
n
t
An
al
y
s
is
Mo
d
el
.
E
x
p
er
i
m
e
n
t
5
w
a
s
p
er
f
o
r
m
ed
t
o
ev
alu
a
te
P
r
o
p
o
s
ed
C
S
A
M
with
p
h
r
ase
le
v
el
Su
b
j
ec
tiv
it
y
E
x
tr
ac
tio
n
.
On
l
y
th
e
o
b
j
ec
t p
ar
t o
f
th
e
s
en
ten
ce
s
i
n
t
h
e
d
ataset
w
as p
r
o
v
i
d
ed
as in
p
u
t to
th
e
C
S
A
M
m
o
d
el.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
NS
R
es
u
lts
o
f
ex
p
er
i
m
en
t 1
,
f
o
r
P
an
g
a
n
d
L
ee
’
s
Mo
v
ie
R
e
v
ie
w
d
ataset
ar
e
p
r
esen
ted
in
F
i
g
u
r
e
4
.
Fig
u
r
e
4
. N
T
W
B
p
ar
am
eter
d
eter
m
i
n
atio
n
f
o
r
p
r
o
p
o
s
ed
W
R
T
FS
C
,
W
A
R
T
FS
C
a
n
d
W
Sen
t
iT
FID
F c
lass
if
ier
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2088
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
4
,
A
u
g
u
s
t
201
8
:
2
3
5
8
–
2
3
6
6
2364
E
x
p
er
i
m
e
n
ted
NT
W
B
v
alu
es
ar
e
r
ep
r
esen
ted
o
n
x
ax
i
s
.
A
cc
u
r
ac
y
at
s
p
ec
i
f
ic
NT
W
B
v
al
u
e
i
s
r
ep
r
esen
ted
o
n
y
a
x
is
.
A
cc
u
r
ac
y
o
f
ea
ch
f
o
ld
is
r
ep
r
esen
ted
in
d
if
f
er
en
t
co
lo
u
r
.
I
t
ca
n
b
e
o
b
s
er
v
ed
f
r
o
m
f
i
g
u
r
e
4
,
w
h
en
NT
W
B
=0
,
ac
cu
r
ac
y
w
a
s
n
ea
r
l
y
5
0
%.
A
s
NT
W
B
p
ar
am
eter
w
as
in
cr
e
m
e
n
ted
,
ac
cu
r
ac
y
i
m
p
r
o
v
ed
u
n
t
il
a
ce
r
tai
n
v
al
u
e.
T
h
is
NT
W
B
v
alu
e
w
as
s
elec
ted
as
t
h
e
o
p
ti
m
al
NT
W
B
v
alu
e.
I
t
ca
n
b
e
o
b
s
er
v
ed
f
r
o
m
f
i
g
u
r
e
4
t
h
at
m
a
x
i
m
u
m
ac
cu
r
ac
y
f
o
r
W
R
T
FS
C
cla
s
s
i
f
ier
w
a
s
ac
h
iev
ed
at
NT
W
B
v
alu
e
o
f
1
.
4
.
Si
m
ilar
l
y
it
ca
n
al
s
o
b
e
n
o
ted
th
at
o
p
ti
m
a
l
NT
W
B
v
alu
es
f
o
r
W
AR
T
FS
C
cla
s
s
i
f
ier
an
d
W
Se
n
tiT
FID
F
ar
e
1
.
1
an
d
0
.
4
r
esp
ec
tiv
el
y
.
T
h
ese
ar
e
m
ar
k
e
d
u
s
in
g
a
b
lu
e
b
lo
ck
.
W
in
d
o
w
b
o
u
n
d
ar
y
v
al
u
es
co
n
ce
p
t
w
as
n
o
t
ap
p
licab
le
f
o
r
T
r
a
d
itio
n
al
Sen
ti
m
e
n
t C
lass
if
i
er
(
T
SC
)
as it is n
o
t b
ased
o
n
r
elativ
e
o
r
r
atio
b
ased
m
at
h
e
m
a
tical
m
o
d
el.
Fig
u
r
e
5
.
A
cc
u
r
ac
y
Gr
ap
h
f
o
r
ev
alu
a
tin
g
P
er
f
o
r
m
a
n
ce
o
f
P
r
o
p
o
s
ed
C
lass
i
f
ier
s
o
n
m
e
n
tio
n
e
d
d
atasets
R
es
u
lts
f
o
r
ex
p
er
i
m
en
t
2
ar
e
p
r
esen
ted
in
F
ig
u
r
e
5
.
C
lass
i
f
ier
s
p
er
f
o
r
m
a
n
ce
w
as
ev
alu
a
ted
.
C
las
s
i
f
ier
s
b
ein
g
i
n
d
ep
en
d
en
t
o
f
d
o
m
ai
n
w
er
e
e
v
alu
a
ted
o
n
d
ataset
s
f
r
o
m
m
o
v
ie
a
s
w
e
ll
as
p
r
o
d
u
ct
an
d
r
estau
r
an
t
d
o
m
ai
n
s
.
I
t
w
as
o
b
s
er
v
ed
th
at
p
r
o
p
o
s
ed
cl
ass
i
f
ier
s
o
u
tp
er
f
o
r
m
t
h
e
tr
a
d
itio
n
al
class
i
f
ier
s
.
I
r
r
esp
ec
tiv
e
o
f
th
e
i
n
p
u
t d
atas
et
th
e
p
r
o
p
o
s
ed
class
if
ier
s
ac
h
i
ev
e
ac
cu
r
ac
y
h
i
g
h
er
th
a
n
tr
ad
itio
n
al
clas
s
i
f
ier
s
.
Fig
u
r
e
6
.
A
cc
u
r
ac
y
g
r
ap
h
f
o
r
ev
alu
a
tin
g
p
er
f
o
r
m
a
n
ce
af
ter
p
r
ep
r
o
ce
s
s
in
g
d
atase
t
Fig
u
r
e
7
.
A
cc
u
r
ac
y
g
r
ap
h
f
o
r
ev
alu
a
tin
g
p
er
f
o
r
m
a
n
ce
o
f
p
r
o
p
o
s
ed
c
o
n
ce
p
tu
al
s
en
ti
m
en
t a
n
al
y
s
is
m
o
d
el
R
es
u
lts
f
o
r
ex
p
er
i
m
e
n
t
3
&
4
ar
e
p
r
esen
ted
in
Fig
u
r
e
6
an
d
Fig
u
r
e
7
r
esp
ec
tiv
el
y
.
A
s
lig
h
t
i
m
p
r
o
v
e
m
en
t
i
n
ac
c
u
r
ac
y
d
u
e
to
p
r
o
p
o
s
ed
ex
ten
s
io
n
s
to
ex
i
s
tin
g
p
r
ep
r
o
ce
s
s
in
g
tec
h
n
iq
u
e
s
ca
n
b
e
o
b
s
er
v
ed
.
I
t
ca
n
b
e
o
b
s
er
v
ed
th
a
t
t
h
e
p
r
o
p
o
s
ed
C
o
n
ce
p
tu
al
Se
n
ti
m
en
t
A
n
a
l
y
s
is
m
o
d
el
s
(
C
S
A
M)
s
h
o
w
s
r
e
m
ar
k
ab
le
i
m
p
r
o
v
e
m
en
t
i
n
ac
c
u
r
ac
y
.
R
es
u
lts
f
o
r
ex
p
er
i
m
e
n
t
5
ar
e
p
r
esen
ted
in
F
i
g
u
r
e
8
.
T
h
e
P
r
o
p
o
s
ed
C
o
n
ce
p
tu
al
Sen
t
i
m
e
n
t
An
al
y
s
i
s
m
o
d
el
s
h
o
w
s
a
r
e
m
ar
k
ab
le
i
m
p
r
o
v
e
m
en
t
i
n
ac
c
u
r
ac
y
a
f
ter
S
u
b
j
ec
tiv
it
y
E
x
tr
ac
tio
n
at
t
h
e
p
h
r
ase
lev
e
l
.
Fig
u
r
e
8
.
A
cc
u
r
ac
y
g
r
ap
h
f
o
r
ev
alu
a
tin
g
p
er
f
o
r
m
a
n
c
e
a
f
ter
s
u
b
j
ec
tiv
it
y
e
x
tr
ac
tio
n
R
es
u
lts
ar
e
d
is
cu
s
s
ed
h
er
e
co
n
s
id
er
in
g
t
h
e
o
u
tco
m
e
s
o
f
P
a
n
g
a
n
d
L
ee
’
s
Mo
v
ie
R
e
v
ie
w
D
ataset
as
it
h
as
b
ee
n
p
o
p
u
lar
l
y
u
s
ed
f
o
r
a
n
al
y
s
i
s
.
P
r
o
p
o
s
ed
clas
s
if
ier
W
R
T
FS
C
s
ta
n
d
s
f
ir
s
t
w
it
h
a
n
ac
cu
r
ac
y
o
f
7
7
.
2
%.
Fig
u
r
e
6
r
ep
r
esen
ts
th
e
ev
a
l
u
atio
n
w
it
h
tr
ad
itio
n
al
p
r
ep
r
o
ce
s
s
in
g
a
n
d
p
r
o
p
o
s
ed
p
r
ep
r
o
ce
s
s
in
g
e
x
te
n
s
io
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
C
o
n
ce
p
tu
a
l S
e
n
timen
t A
n
a
lysi
s
Mo
d
el
(
K
r
a
n
ti V
ith
a
l G
h
a
g
)
2
365
W
Sen
tiT
FID
F
o
u
tp
er
f
o
r
m
ed
w
it
h
an
ac
cu
r
ac
y
o
f
7
8
%.
Fi
g
u
r
e
7
an
d
Fi
g
u
r
e
8
r
ep
r
esen
ts
t
h
e
ac
c
u
r
ac
y
o
f
p
r
o
p
o
s
ed
C
o
n
ce
p
tu
al
Se
n
ti
m
e
n
t
An
al
y
s
i
s
m
o
d
el
a
n
d
p
r
o
p
o
s
ed
Mo
d
if
ied
C
o
n
ce
p
tu
al
Sen
ti
m
en
t
Mo
d
el
w
h
er
e
W
A
R
T
FS
C
class
if
ier
h
ad
m
a
x
i
m
u
m
ac
c
u
r
ac
y
o
f
8
8
.
5
%
&
9
3
.
9
% r
esp
ec
tiv
el
y
.
4.
CO
NCLU
SI
O
N
/S
A
ND
F
UT
URE SCO
P
E
P
r
o
p
o
s
ed
W
R
T
FS
C
,
W
A
R
T
F
SC
&
W
Sen
tiT
FID
F
clas
s
i
f
ie
r
s
s
h
o
w
ed
an
i
m
p
r
o
v
e
m
e
n
t
i
n
s
en
ti
m
e
n
t
class
i
f
icatio
n
ac
cu
r
ac
y
f
r
o
m
6
7
.
9
%
to
7
7
.
2
%.
P
r
e
p
r
o
ce
s
s
in
g
tec
h
n
iq
u
es
w
er
e
e
x
te
n
d
ed
to
h
an
d
le
ter
m
w
it
h
ap
o
s
tr
o
p
h
e,
alo
n
g
w
it
h
tr
ad
iti
o
n
al
w
a
y
o
f
d
is
ca
r
d
in
g
p
u
n
ct
u
atio
n
s
an
d
o
t
h
er
s
y
m
b
o
ls
.
A
s
lig
h
t
i
m
p
r
o
v
e
m
en
t
in
ac
c
u
r
ac
y
f
r
o
m
7
7
.
2
%
to
7
8
%
w
as o
b
s
er
v
ed
d
u
e
to
p
r
ep
r
o
ce
s
s
i
ng
ex
te
n
s
io
n
s
.
B
ag
-
of
-
w
o
r
d
s
ap
p
r
o
ac
h
f
ail
ed
to
h
an
d
le
s
y
n
tact
ic
&
s
e
m
a
n
tic
s
tr
u
ctu
r
e
o
f
r
ev
ie
w
s
.
As
class
i
f
ier
s
ar
e
b
ased
o
n
p
r
o
b
ab
ilis
tic
m
o
d
els,
ad
ap
tin
g
cla
s
s
i
f
ier
s
f
o
r
h
a
n
d
l
in
g
n
at
u
r
al
la
n
g
u
a
g
e
s
p
ec
i
f
ic
atio
n
s
w
a
s
an
in
tr
icate
tas
k
.
P
r
o
p
o
s
ed
c
o
n
ce
p
t
b
as
ed
ap
p
r
o
ac
h
es
h
an
d
led
s
y
n
tactic
&
s
e
m
a
n
tic
s
tr
u
c
tu
r
e
o
f
s
e
n
ten
ce
.
R
e
m
ar
k
ab
le
leap
in
ac
c
u
r
ac
y
f
r
o
m
78%
to
8
8
.
5
%
w
as
o
b
s
er
v
ed
d
u
e
to
p
r
o
p
o
s
ed
C
o
n
ce
p
tu
al
Sen
ti
m
en
t
An
al
y
s
is
m
o
d
el.
Su
b
j
ec
tiv
it
y
ex
tr
ac
tio
n
w
a
s
p
er
f
o
r
m
ed
at
th
e
p
h
r
ase
le
v
el,
r
esu
l
tin
g
to
an
i
m
p
r
o
v
e
m
e
n
t in
ac
c
u
r
ac
y
f
r
o
m
8
8
.
5
%
to
9
3
.
9
%.
T
h
e
ac
cu
r
ac
ies
o
f
s
u
r
v
e
y
ed
tech
n
iq
u
es
t
h
at
u
s
ed
P
an
g
a
n
d
L
ee
’
s
Mo
v
ie
R
ev
ie
w
Dat
aset
w
er
e
b
et
w
ee
n
7
6
.
3
7
%
an
d
92.
7
%.
A
lt
h
o
u
g
h
t
h
ese
ac
c
u
r
ac
ies
c
an
n
o
t
b
e
d
ir
ec
tl
y
co
m
p
ar
ed
a
s
t
h
e
e
x
p
er
i
m
e
n
tal
s
etu
p
m
a
y
v
ar
y
,
t
h
e
p
r
o
p
o
s
ed
C
o
n
ce
p
tu
a
l
Sen
ti
m
en
t
An
al
y
s
i
s
Mo
d
el
p
er
f
o
r
m
s
b
etter
th
a
n
ex
is
t
in
g
tec
h
n
iq
u
es
w
it
h
a
r
e
m
ar
k
ab
le
ac
cu
r
ac
y
o
f
9
3
.
9
%.
C
S
A
M
m
o
d
el
f
o
llo
w
s
t
h
e
co
n
ce
p
t
b
ased
ap
p
r
o
ac
h
at
w
o
r
d
,
p
h
r
ase
an
d
s
en
ten
ce
le
v
el.
Mo
d
els
th
at
f
o
cu
s
o
n
th
e
co
n
ce
p
t
b
ased
ap
p
r
o
ac
h
es
,
at
i
n
ter
s
tate
m
e
n
t
lev
el
an
d
in
ter
d
o
cu
m
en
t
o
r
r
ev
ie
w
lev
e
l
ca
n
b
e
d
esig
n
ed
.
RE
F
E
R
E
NC
E
S
[1
]
Ra
v
i
K,
Ra
v
i
V
,
“
A
su
rv
e
y
o
n
o
p
i
n
io
n
m
in
in
g
a
n
d
se
n
ti
m
e
n
t
a
n
a
ly
sis:
T
a
s
k
s,
a
p
p
ro
a
c
h
e
s
a
n
d
a
p
p
l
ica
ti
o
n
s”
Kn
o
wled
g
e
-
Ba
se
d
S
y
ste
ms
,
2
0
1
5
,
8
9
,
p
p
.
1
4
-
4
6
.
[2
]
Na
ir
L
R,
S
h
e
tt
y
S
D,
S
h
e
tt
y
S
D,
“
S
trea
m
in
g
Big
Da
ta
A
n
a
ly
sis
f
o
r
Re
a
l
-
T
i
m
e
S
e
n
ti
m
e
n
t
b
a
se
d
T
a
r
g
e
ted
A
d
v
e
rti
sin
g
,
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
E
lec
trica
l
a
n
d
Co
mp
u
ter
En
g
i
n
e
e
rin
g
(
IJ
ECE
)
,
2
0
1
7
Ja
n
,
v
o
l.
7
,
n
o
.
1
,
4
0
2
.
[3
]
P
a
n
g
B,
L
e
e
L
J,
Op
in
io
n
m
in
in
g
a
n
d
se
n
ti
m
e
n
t
a
n
a
ly
sis,
s.n
.
;
2
0
1
0
.
[4
]
K
KP
,
S
N.
In
sig
h
ts
to
P
r
o
b
l
e
m
s,
“
Re
se
a
rc
h
T
re
n
d
a
n
d
P
r
o
g
re
ss
in
T
e
c
h
n
iq
u
e
s
o
f
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis
”,
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
E
lec
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
2
0
1
7
Ja
n
,
v
o
l
.
7
,
n
o
.
5
,
2
0
1
8
.
[5
]
Yu
L
C,
W
u
JL
,
Ch
a
n
g
P
C,
Ch
u
HS,
“
Us
in
g
a
c
o
n
tex
tu
a
l
e
n
tr
o
p
y
m
o
d
e
l
to
e
x
p
a
n
d
e
m
o
ti
o
n
w
o
rd
s
a
n
d
t
h
e
ir
in
ten
sity
f
o
r
th
e
se
n
ti
m
e
n
t
c
las
si
f
i
c
a
ti
o
n
o
f
sto
c
k
m
a
rk
e
t
n
e
w
s
”
,
Kn
o
wled
g
e
-
B
a
se
d
S
y
ste
ms
,
2
0
1
3
,
4
1
,
p
p
.
89
-
9
7
.
[6
]
Zh
a
n
g
P
,
W
a
n
g
S
,
L
i
D,
“
Cro
ss
-
li
n
g
u
a
l
se
n
t
im
e
n
t
c
las
si
f
ica
ti
o
n
:
S
im
il
a
rit
y
d
isc
o
v
e
r
y
p
lu
s
trai
n
i
n
g
d
a
ta
a
d
ju
stm
e
n
t”,
Kn
o
wled
g
e
-
Ba
se
d
S
y
ste
ms
,
2
0
1
6
,
1
0
7
,
p
p
.
1
2
9
-
4
1
.
[7
]
F
e
ld
m
a
n
R,
“
T
e
c
h
n
iq
u
e
s &
a
p
p
li
c
a
ti
o
n
s f
o
r
se
n
ti
m
e
n
t
a
n
a
ly
sis
”
,
Co
mm
u
n
ic
a
ti
o
n
s o
f
ACM
,
2
0
1
3
,
v
o
l.
5
6
,
n
o
.
4
,
8
2
.
[8
]
Ka
sth
u
riara
c
h
c
h
y
BH,
Zo
y
sa
K
D,
P
re
m
a
ra
tn
e
H,
“
En
h
a
n
c
e
d
b
a
g
-
of
-
w
o
rd
s
m
o
d
e
l
f
o
r
p
h
ra
se
-
l
e
v
e
l
s
e
n
ti
m
e
n
t
a
n
a
ly
sis
”
,
1
4
th
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
s in
ICT
f
o
r E
me
rg
in
g
Re
g
io
n
s (
ICT
e
r)
,
2
0
1
4
.
[9
]
Krc
a
d
in
a
c
U,
P
a
sq
u
ier
P
,
Jo
v
a
n
o
v
ic
J,
De
v
e
d
z
ic
V
.
S
y
n
e
sk
e
tch
,
“
A
n
Op
e
n
S
o
u
rc
e
L
ib
ra
ry
f
o
r
S
e
n
ten
c
e
-
Ba
se
d
Em
o
ti
o
n
Re
c
o
g
n
it
i
o
n
”
,
I
EE
E
T
ra
n
sa
c
ti
o
n
s o
n
Af
fec
ti
v
e
Co
mp
u
t
in
g
,
2
0
1
3
,
v
o
l.
4
,
n
o
.
3
,
p
p
.
3
1
2
-
25.
[1
0
]
G
h
a
g
K,
S
h
a
h
K
,
“
Co
m
p
a
ra
ti
v
e
a
n
a
ly
sis
o
f
th
e
tec
h
n
iq
u
e
s
f
o
r
S
e
n
t
im
e
n
t
A
n
a
l
y
sis
”,
2
0
1
3
I
n
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
A
d
v
a
n
c
e
s in
T
e
c
h
n
o
l
o
g
y
a
n
d
E
n
g
i
n
e
e
rin
g
(
ICAT
E)
,
2
0
1
3
.
[1
1
]
M
il
ler
GA
,
“
W
o
r
d
Ne
t:
a
lex
ica
l
d
a
tab
a
se
f
o
r
En
g
li
sh
”
,
Co
mm
u
n
i
c
a
ti
o
n
s
o
f
t
h
e
ACM
,
1
9
9
5
Ja
n
,
v
o
l.
3
8
,
n
o
.
1
1
,
pp.
39
-
4
1
.
[1
2
]
Ne
v
iaro
u
sk
a
y
a
A
,
P
re
n
d
in
g
e
r
H,
Ish
iz
u
k
a
M
.
S
e
n
ti
F
u
l
,
“
A
Lex
ico
n
f
o
r
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis
”
,
IEE
E
T
ra
n
sa
c
ti
o
n
s o
n
Af
fec
ti
v
e
Co
mp
u
t
in
g
,
2
0
1
1
,
v
o
l.
2
,
n
o
.
1
,
p
p
.
22
-
3
6
.
[1
3
]
S
a
in
i
JR,
Ra
k
h
o
li
a
R
M
,
“
On
Co
n
ti
n
e
n
t
a
n
d
S
c
rip
t
-
W
ise
Div
isio
n
s
-
Ba
se
d
S
tatisti
c
a
l
M
e
a
su
re
s
f
o
r
S
t
o
p
-
w
o
rd
s
L
ists
o
f
In
tern
a
ti
o
n
a
l
L
a
n
g
u
a
g
e
s
”
,
Pro
c
e
d
ia
Co
m
p
u
ter
S
c
ien
c
e
,
2
0
1
6
,
v
o
l.
89
,
p
p
.
3
1
3
-
31
9.
[1
4
]
T
rip
a
th
y
A
,
A
g
ra
wa
l
A
,
Ra
th
S
K
,
“
Clas
sif
ic
a
ti
o
n
o
f
se
n
ti
m
e
n
t
re
v
iew
s
u
sin
g
n
-
g
ra
m
m
a
c
h
in
e
lea
rn
in
g
a
p
p
r
o
a
c
h
”
Exp
e
rt S
y
ste
ms
wit
h
A
p
p
l
ica
ti
o
n
s
,
2
0
1
6
,
v
o
l.
57
,
p
p
.
1
1
7
-
1
2
6
.
[1
5
]
S
o
c
h
e
r
R
,
“
De
e
p
Lea
rn
in
g
f
o
r
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis
-
In
v
it
e
d
T
a
l
k
”
,
Pro
c
e
e
d
in
g
s
o
f
th
e
7
t
h
W
o
rk
sh
o
p
o
n
Co
mp
u
t
a
ti
o
n
a
l
A
p
p
r
o
a
c
h
e
s t
o
S
u
b
jec
ti
v
it
y
,
S
e
n
t
ime
n
t
a
n
d
S
o
c
ia
l
M
e
d
ia
An
a
lys
is
,
2
0
1
6
.
[1
6
]
B.
P
a
n
g
,
L
.
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
?
,
”
Pro
c
e
e
d
in
g
s
o
f
th
e
ACL
-
0
2
c
o
n
fer
e
n
c
e
o
n
Emp
irica
l
me
th
o
d
s i
n
n
a
t
u
ra
l
la
n
g
u
a
g
e
p
ro
c
e
ss
in
g
-
EM
NL
P
'0
2
,
2
0
0
2
.
[1
7
]
P
a
n
g
B,
L
e
e
L
.
A
,
S
e
n
ti
m
e
n
tal
E
d
u
c
a
ti
o
n
”
,
Pro
c
e
e
d
in
g
s
o
f
t
h
e
4
2
n
d
An
n
u
a
l
M
e
e
ti
n
g
o
n
A
ss
o
c
ia
ti
o
n
fo
r
Co
mp
u
t
a
ti
o
n
a
l
L
in
g
u
isti
c
s
-
ACL
'
0
4
,
2
0
0
4
.
[1
8
]
A
ro
ra
P
,
V
irm
a
n
i
D,
Ku
lk
a
rn
i
P
S
,
“
A
n
A
p
p
ro
a
c
h
f
o
r
Big
Da
ta
to
Ev
o
lv
e
th
e
A
u
sp
icio
u
s
I
n
f
o
rm
a
ti
o
n
f
ro
m
Cro
ss
-
Do
m
a
in
s
”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
El
e
c
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
2
0
1
7
Ja
n
,
v
o
l.
7
,
n
o
.
2
,
9
6
7
.
[1
9
]
J
M
a
rti
n
e
a
u
J
,
F
i
n
i
n
T
,
Jo
s
h
i
A
,
P
a
tel
S
,
“
Im
p
ro
v
in
g
b
in
a
ry
c
las
si
f
ica
ti
o
n
o
n
tex
t
p
r
o
b
lem
s
u
sin
g
d
iff
e
re
n
ti
a
l
w
o
rd
f
e
a
tu
re
s
”
,
Pro
c
e
e
d
in
g
o
f
t
h
e
1
8
t
h
ACM
c
o
n
fer
e
n
c
e
o
n
I
n
f
o
rm
a
ti
o
n
a
n
d
k
n
o
w
led
g
e
m
a
n
a
g
e
me
n
t
-
CI
KM
'
0
9
.
2
0
0
9
.
[2
0
]
S
il
v
a
N,
Hru
sc
h
k
a
E,
Hru
sc
h
k
a
E.
Bio
c
o
m
Us
p
,
“
T
we
e
t
S
e
n
ti
m
e
n
t
A
n
a
ly
sis
w
it
h
A
d
a
p
ti
v
e
Bo
o
stin
g
En
se
m
b
le
”
,
Pro
c
e
e
d
in
g
s
o
f
t
h
e
8
th
I
n
ter
n
a
ti
o
n
a
l
W
o
rk
sh
o
p
o
n
S
e
ma
n
ti
c
Eva
lu
a
ti
o
n
(
S
e
mEv
a
l
2
0
1
4
)
,
2
0
1
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2088
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
4
,
A
u
g
u
s
t
201
8
:
2
3
5
8
–
2
3
6
6
2366
[2
1
]
Ha
n
J,
Ka
m
b
e
r
M
,
P
e
i
J
,
“
Da
ta
m
in
in
g
:
c
o
n
c
e
p
ts
a
n
d
tec
h
n
iq
u
e
s
”
,
El
se
v
ier/M
o
rg
a
n
Ka
u
fm
a
n
n
;
2
0
1
2
.
[2
2
]
Ha
d
d
i
E,
L
iu
X
,
S
h
i
Y
,
“
T
h
e
R
o
le
o
f
T
e
x
t
P
re
-
p
ro
c
e
ss
in
g
in
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis
”
,
Pro
c
e
d
ia
Co
mp
u
ter
S
c
ien
c
e
,
2
0
1
3
,
v
o
l.
17
,
p
p
.
26
-
32.
[2
3
]
L
o
c
h
ter
JV
,
Z
a
n
e
tt
i
RF
,
Re
ll
e
r
D,
A
l
m
e
id
a
TA
,
“
S
h
o
rt
tex
t
o
p
i
n
io
n
d
e
tec
ti
o
n
u
si
n
g
e
n
se
m
b
le
o
f
c
la
ss
i
f
iers
a
n
d
se
m
a
n
ti
c
in
d
e
x
in
g
”
,
Exp
e
rt S
y
ste
ms
wit
h
A
p
p
li
c
a
ti
o
n
s
,
2
0
1
6
,
v
o
l.
62
,
p
p
.
2
4
3
-
24
9.
[2
4
]
U
y
sa
l
A
K,
G
u
n
a
l
S
,
“
T
h
e
i
m
p
a
c
t
o
f
p
re
p
r
o
c
e
ss
in
g
o
n
tex
t
c
las
si
f
ic
a
ti
o
n
”
,
In
fo
rm
a
ti
o
n
Pro
c
e
ss
in
g
&
M
a
n
a
g
e
me
n
t
,
2
0
1
4
,
v
o
l.
50
,
n
o
.
1
,
p
p
.
1
0
4
-
1
1
2
.
[2
5
]
X
ia
R,
Xu
F
,
Z
o
n
g
C,
L
i
Q,
Qi
Y
,
L
i
T
,
“
Du
a
l
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis:
Co
n
si
d
e
ri
n
g
Tw
o
S
id
e
s
o
f
On
e
Re
v
ie
w
”
,
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Kn
o
wled
g
e
a
n
d
Da
ta
E
n
g
in
e
e
rin
g
,
2
0
1
5
Ja
n
,
v
o
l.
27
,
n
o
.
8
,
p
p
.
2
1
2
0
-
21
3
3
.
[2
6
]
X
ia
R,
X
u
F
,
Y
u
J,
Qi
Y,
Ca
m
b
ria
E
,
“
P
o
larity
sh
if
t
d
e
tec
ti
o
n
,
e
li
m
in
a
ti
o
n
a
n
d
e
n
se
m
b
le:
A
th
re
e
-
sta
g
e
m
o
d
e
l
f
o
r
d
o
c
u
m
e
n
t
-
lev
e
l
se
n
ti
m
e
n
t
a
n
a
ly
si
s
”
,
In
fo
rm
a
t
io
n
Pro
c
e
ss
in
g
&
M
a
n
a
g
e
me
n
t
,
2
0
1
6
,
v
o
l.
52
,
n
o
.
1
,
p
p
.
36
-
45.
[2
7
]
P
a
d
m
a
ja
S
,
F
a
ti
m
a
S
S
,
Ba
n
d
u
S
,
S
o
wm
y
a
B
,
“
Co
m
p
a
riso
n
o
f
th
e
sc
o
p
e
o
f
n
e
g
a
ti
o
n
in
o
n
li
n
e
n
e
w
s
a
rti
c
les
”
,
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
o
mp
u
ti
n
g
a
n
d
C
o
mm
u
n
ica
ti
o
n
T
e
c
h
n
o
lo
g
ies
,
2
0
1
4
.
[2
8
]
In
d
h
u
ja
K,
Re
g
h
u
R
P
C
,
“
F
u
z
z
y
lo
g
ic
b
a
s
e
d
se
n
ti
m
e
n
t
a
n
a
l
y
s
is
o
f
p
ro
d
u
c
t
re
v
iew
d
o
c
u
m
e
n
ts
”
,
2
0
1
4
Fi
rs
t
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
S
y
s
tem
s a
n
d
C
o
mm
u
n
i
c
a
ti
o
n
s (
ICCS
C)
,
2
0
1
4
.
[2
9
]
Hu
n
g
C,
L
in
HK
,
“
Us
in
g
Ob
jec
ti
v
e
W
o
rd
s
in
S
e
n
ti
W
o
rd
Ne
t
t
o
Im
p
ro
v
e
W
o
rd
-
of
-
M
o
u
th
S
e
n
ti
m
e
n
t
Clas
sif
ic
a
ti
o
n
”
,
IEE
E
In
telli
g
e
n
t
S
y
ste
ms
,
2
0
1
3
,
v
o
l.
28
,
n
o
.
2
,
p
p
.
47
-
5
4
.
[3
0
]
S
o
f
tw
a
r
e
>
S
tan
f
o
rd
P
a
rse
r
[
I
n
tern
e
t
]
.
T
h
e
S
tan
f
o
rd
Na
t
u
ra
l
L
a
n
g
u
a
g
e
P
ro
c
e
ss
in
g
G
ro
u
p
.
Av
a
il
a
b
le
f
ro
m
:
h
tt
p
s:/
/n
l
p
.
sta
n
f
o
rd
.
e
d
u
/s
o
f
twa
re
/l
e
x
-
p
a
rse
r.
sh
t
m
l
.
[3
1
]
M
a
a
s
A
,
Da
l
y
R,
P
h
a
m
P
,
Hu
a
n
g
D,
Ng
A
,
P
o
tt
s C
,
“
L
e
a
rn
in
g
w
o
r
d
v
e
c
to
rs
f
o
r
se
n
ti
m
e
n
t
a
n
a
l
y
sis
”
,
Pro
c
e
e
d
in
g
s o
f
th
e
4
9
t
h
A
n
n
u
a
l
M
e
e
ti
n
g
o
n
Asso
c
ia
ti
o
n
f
o
r
C
o
mp
u
ta
ti
o
n
a
l
L
i
n
g
u
i
sti
c
s:
Hu
ma
n
L
a
n
g
u
a
g
e
T
e
c
h
n
o
lo
g
ies
-
AC
L
'1
1
,
2
0
1
1
.
[3
2
]
Ko
tzia
s
D,
De
n
il
M
,
F
re
it
a
s
ND
,
S
m
y
th
P
,
“
F
r
o
m
G
ro
u
p
to
I
n
d
iv
i
d
u
a
l
L
a
b
e
ls
Us
in
g
De
e
p
F
e
a
tu
re
s
”
,
Pro
c
e
e
d
in
g
s
o
f
t
h
e
2
1
th
ACM
S
IGKD
D In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
K
n
o
w
led
g
e
Disc
o
v
e
ry
a
n
d
Da
t
a
M
in
i
n
g
-
K
DD
'
1
5
.
2
0
1
5
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
M
s.
K
r
a
n
ti
V
it
h
a
l
G
h
a
g
is
a
n
A
s
sista
n
t
P
ro
f
e
ss
o
r
a
t
In
f
o
r
m
a
ti
o
n
T
e
c
h
n
o
lo
g
y
De
p
a
rt
m
e
n
t
o
f
M
ET
’s
S
h
a
h
&
A
n
c
h
o
r
Ku
tch
h
i
E
n
g
i
n
e
e
rin
g
Co
ll
e
g
e
,
M
u
m
b
a
i,
In
d
ia.
S
h
e
i
s p
u
rsu
i
n
g
P
h
D i
n
Co
m
p
u
ter S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
f
ro
m
S
V
KM’s
NMIM
S
M
P
S
T
M
E,
M
u
m
b
a
i,
In
d
ia.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis a
n
d
Da
ta M
in
in
g
.
Dr
.
K
e
ta
n
S
h
a
h
h
a
s
a
P
h
D
i
n
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
lo
g
y
a
n
d
is
w
o
rk
in
g
a
s
a
P
ro
f
e
ss
o
r
a
t
S
VK
M
’s
NMIM
S
M
u
k
e
sh
P
a
tel
S
c
h
o
o
l
o
f
T
e
c
h
n
o
lo
g
y
M
a
n
a
g
e
m
e
n
t
&
En
g
in
e
e
rin
g
.
His
re
s
e
a
rc
h
in
te
re
sts
in
c
lu
d
e
Da
ta M
in
i
n
g
u
sin
g
p
a
ra
ll
e
l
a
p
p
ro
a
c
h
e
s
a
n
d
S
e
n
ti
m
e
n
t
A
n
a
ly
sis
.
Evaluation Warning : The document was created with Spire.PDF for Python.