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.
7
,
No
.
3
,
Sep
tem
b
er
201
8
,
p
p
.
11
9
~
124
I
SS
N:
2252
-
8938
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ai.
v
7
.
i3
.
p
p
11
9
-
1
2
4
119
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
JA
I
Ana
ly
sing
Ev
ent
-
Rela
ted
S
enti
m
e
n
ts on So
cia
l Medi
a
w
ith
Neura
l Ne
tw
o
rk
s
P
.
Sa
nthi P
riy
a
1
,
T
.
Venk
a
t
e
s
w
a
ra
Ra
o
2
1
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter S
c
ien
c
e
a
n
d
E
n
g
in
e
e
rin
g
,
S
ri
S
a
iram
Co
ll
e
g
e
o
f
En
g
in
e
e
rin
g
,
Be
n
g
u
lu
r
u
,
In
d
ia
2
Co
m
p
u
ter S
c
ien
c
e
a
n
d
E
n
g
in
e
e
ri
n
g
,
P
VP
S
id
d
h
a
rth
a
En
g
in
e
e
rin
g
Co
ll
e
g
e
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
r
1
1
,
2
0
1
8
R
ev
i
s
ed
Ma
y
10
,
2
0
1
8
A
cc
ep
ted
J
u
n
20
,
2
0
1
8
S
e
n
ti
m
e
n
t
a
n
a
l
y
sis
is
p
e
rf
o
r
m
e
d
to
d
e
term
in
e
th
e
p
o
lari
ty
o
f
o
p
in
io
n
o
n
a
su
b
jec
t.
It
h
a
s
b
e
e
n
a
p
p
li
e
d
to
te
x
t
c
o
rp
o
ra
su
c
h
a
s
m
o
v
ie
re
v
ie
ws
,
f
in
a
n
c
ial
d
o
c
u
m
e
n
ts
to
g
lea
n
in
f
o
r
m
a
ti
o
n
a
b
o
u
t
o
v
e
ra
ll
-
se
n
ti
m
e
n
t
a
n
c
p
ro
d
u
c
e
a
c
ti
o
n
a
b
le
d
a
ta.
Re
c
e
n
t
e
v
e
n
ts
h
a
v
e
d
e
m
o
n
stra
ted
th
a
t
p
o
ll
i
n
g
c
a
n
b
e
so
m
e
ti
m
e
s
u
n
re
li
a
b
le.
P
e
o
p
le
c
a
n
b
e
d
iff
icu
lt
to
a
c
c
e
s
s
th
ro
u
g
h
c
o
n
v
e
n
ti
o
n
a
l
p
o
ll
i
n
g
m
e
th
o
d
s a
n
d
les
s th
a
n
f
ra
n
k
in
p
o
ll
s.
I
n
th
e
e
ra
o
f
so
c
ial
m
e
d
ia,
v
o
ters
a
re
li
k
e
l
y
to
m
o
re
f
re
e
l
y
e
x
p
re
s
s
th
e
ir
o
p
in
i
o
n
o
n
so
c
ial
m
e
d
ia
f
o
ru
m
s
a
b
o
u
t
d
iv
isiv
e
e
v
e
n
ts
e
sp
e
c
iall
y
in
m
e
d
ia
w
h
e
re
a
n
o
n
y
m
it
y
e
x
ists.
A
n
a
l
y
z
in
g
th
e
p
re
v
a
il
in
g
o
p
i
n
io
n
o
n
t
h
e
se
f
o
ru
m
s
c
a
n
in
d
ica
te
if
th
e
re
a
re
a
n
y
d
e
f
icie
n
c
ies
in
p
o
ll
in
g
a
n
d
c
a
n
b
e
a
v
a
lu
a
b
le
a
d
d
it
i
o
n
t
o
c
o
n
v
e
n
t
io
n
a
l
p
o
ll
in
g
.
W
e
a
n
a
ly
z
e
d
tex
t
c
o
rp
o
ra
f
ro
m
Re
d
d
it
f
o
ru
m
s
d
isc
u
ss
in
g
th
e
re
c
e
n
t
re
fe
re
n
d
u
m
in
Brit
a
in
to
e
x
it
f
ro
m
th
e
EU
(
k
n
o
w
n
a
s
Br
e
x
it
).
Bre
x
it
wa
s
a
n
im
p
o
rtan
t
w
o
rld
e
v
e
n
t
a
n
d
w
a
s
v
e
r
y
d
iv
isiv
e
in
th
e
r
u
n
-
u
p
a
n
d
p
o
st
v
o
te.
W
e
a
n
a
l
y
z
e
d
se
n
ti
m
e
n
t
in
t
w
o
w
a
y
s:
In
it
iall
y
w
e
tri
e
d
to
g
a
u
g
e
p
o
sit
iv
e
,
n
e
g
a
ti
v
e
,
a
n
d
n
e
u
tral
se
n
t
im
e
n
ts.
In
t
h
e
se
c
o
n
d
a
n
a
ly
sis,
w
e
f
u
rth
e
r
sp
li
t
t
h
e
se
se
n
ti
m
e
n
ts
in
to
six
d
if
f
e
re
n
t
p
o
lariti
e
s
b
a
se
d
o
n
t
h
e
d
irec
ti
o
n
a
li
ty
o
f
th
e
p
o
siti
v
e
a
n
d
n
e
g
a
ti
v
e
se
n
ti
m
e
n
ts
(f
o
r
o
r
a
g
a
in
s
t
Bre
x
it
).
Ou
r
tec
h
n
iq
u
e
u
t
li
li
z
e
d
p
a
ra
g
ra
p
h
ve
c
to
rs
(Do
c
2
V
e
c
)
t
o
c
o
n
stru
c
t
f
e
a
tu
re
v
e
c
to
rs
f
o
r
se
n
ti
m
e
n
t
a
n
a
l
y
sis
w
it
h
a
M
u
lt
il
a
y
e
r
P
e
rc
e
p
tro
n
c
las
sif
ier.
W
e
f
o
u
n
d
t
h
a
t
th
e
se
c
o
n
d
a
n
a
ly
s
is
y
ield
e
d
o
v
e
ra
ll
b
e
tt
e
r
re
su
lt
s;
a
lt
h
o
u
g
h
,
o
u
r
c
las
sif
ier
d
id
n
’t
p
e
rf
o
rm
a
s
w
e
ll
in
c
las
si
fy
in
g
p
o
siti
v
e
se
n
ti
m
e
n
t
s.
W
e
d
e
m
o
n
stra
te
th
a
t
i
t
is
p
o
ss
ib
le
g
lea
n
v
a
lu
a
b
le
in
f
o
r
m
a
ti
o
n
f
ro
m
c
o
m
p
li
c
a
ted
a
n
d
d
iv
e
rse
c
o
rp
o
ra
su
c
h
a
s
m
u
lt
i
-
p
a
ra
g
ra
p
h
c
o
m
m
e
n
ts
f
ro
m
re
d
d
it
w
it
h
se
n
ti
m
e
n
t
a
n
a
l
y
sis.
K
ey
w
o
r
d
:
Do
c2
Vec
E
U
r
ef
er
en
d
u
m
Mu
lti
-
la
y
er
p
er
ce
p
tr
o
n
s
Neu
r
al
n
et
w
o
r
k
s
P
ar
ag
r
ap
h
v
ec
to
r
s
Sen
ti
m
e
n
t a
n
a
l
y
s
is
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
:
P
.
San
th
i P
r
i
y
a,
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
i
n
ee
r
in
g
,
Sri
Sair
a
m
C
o
lleg
e
o
f
E
n
g
in
ee
r
in
g
,
B
en
g
u
lu
r
u
,
I
n
d
ia
.
E
m
ail:
s
h
a
n
tip
r
i
y
a.
p
@
g
m
ail.
c
o
m
1.
I
NT
RO
D
UCT
I
O
N
Sen
ti
m
e
n
t
a
n
al
y
s
i
s
i
s
a
q
u
a
n
tit
ativ
e
a
u
to
m
ated
o
p
in
io
n
m
i
n
i
n
g
tech
n
iq
u
e
w
h
ich
atte
m
p
ts
t
o
q
u
an
ti
f
y
th
e
s
en
t
i
m
e
n
t
(
p
o
s
itiv
e
/n
e
g
ati
v
e/n
e
u
tr
al)
a
ttach
ed
to
te
x
t
co
n
ten
t.
Ma
c
h
in
e
lear
n
in
g
ap
p
r
o
ac
h
es
[
1
]
lo
o
k
f
o
r
w
o
r
d
s
o
r
p
h
r
ases
th
at
d
en
o
te
t
h
e
s
e
n
ti
m
en
t
p
o
lar
it
y
o
f
th
e
te
x
t.
A
tr
ain
i
n
g
s
et
o
f
ex
a
m
p
le
s
is
u
s
ed
to
ass
i
g
n
a
p
o
lar
ity
a
n
d
w
eig
h
t
to
w
o
r
d
s
w
h
ic
h
ca
n
la
ter
b
e
u
s
ed
to
p
r
ed
ict
th
e
s
e
n
ti
m
e
n
t
o
f
a
w
id
er
s
et
n
o
t
i
n
cl
u
d
ed
in
th
e
tr
ai
n
i
n
g
s
et.
T
r
ain
in
g
ca
n
b
e
s
u
p
er
v
is
ed
b
y
ex
p
lic
it
lab
ellin
g
o
f
p
o
lar
it
y
m
an
u
all
y
o
r
b
y
i
m
p
licit
y
d
eter
m
in
i
n
g
p
o
lar
it
y
f
r
o
m
f
ea
t
u
r
es
s
u
c
h
as
e
m
o
j
is
.
Sen
ti
m
e
n
t
an
al
y
s
i
s
an
d
p
r
ed
ictio
n
h
av
e
b
ee
n
e
m
p
lo
y
ed
i
n
v
ar
io
u
s
d
o
m
ai
n
s
s
u
cc
ess
f
u
ll
y
.
Sen
t
i
m
e
n
t
p
r
ed
ictio
n
h
a
s
b
ee
n
u
s
ed
to
g
e
n
er
ate
s
to
c
k
r
ec
o
m
m
en
d
atio
n
s
b
ased
o
n
r
ea
ctio
n
s
to
n
e
w
s
an
d
b
lo
g
s
[
5
]
,
ass
i
g
n
e
m
o
tio
n
al
p
o
lar
ity
to
n
e
w
s
ar
ticle
s
[
6
]
an
d
p
r
e
d
ict
th
e
r
ea
ctio
n
o
f
s
p
ec
if
ic
a
u
d
ien
ce
s
/co
m
m
u
n
i
ti
es to
n
e
w
s
[
1
1
]
.
Ma
r
k
ets
all
o
v
er
th
e
w
o
r
ld
r
es
p
o
n
d
to
m
aj
o
r
ev
en
t
s
n
o
t
o
n
l
y
p
o
s
t
th
e
ev
e
n
t
b
u
t
a
ls
o
in
th
e
r
u
n
-
u
p
to
ev
en
t
s
.
P
r
ices
i
n
s
to
ck
m
ar
k
et
s
,
r
ea
l
-
e
s
tate
m
ar
k
et
s
a
n
d
cu
r
r
en
c
y
m
ar
k
ets
all
f
l
u
ct
u
ate
ac
c
o
r
d
in
g
to
p
er
ce
iv
ed
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
7
,
No
.
3
,
Sep
tem
b
e
r
20
1
8
:
11
9
–
12
4
120
o
u
tco
m
e
o
f
ev
e
n
t
s
.
T
h
e
ab
ilit
y
to
p
r
ed
ict
h
o
w
t
h
ese
ev
e
n
ts
w
ill
p
an
o
u
t
b
y
g
a
u
g
i
n
g
s
e
n
t
i
m
e
n
t
w
o
u
ld
b
e
an
i
m
m
en
s
el
y
u
s
e
f
u
l
to
o
l
f
o
r
t
ak
in
g
p
r
o
f
itab
le
p
o
s
itio
n
s
in
th
ese
m
ar
k
e
ts
.
P
o
lls
d
o
n
’
t
al
w
a
y
s
r
ef
lect
th
e
s
en
ti
m
e
n
t
ac
c
u
r
atel
y
s
i
n
ce
p
e
o
p
le
ca
n
b
e
d
if
f
icu
lt
to
ac
ce
s
s
o
r
less
t
h
a
n
f
r
a
n
k
in
p
o
ll
s
[
7
]
;
th
er
ef
o
r
e,
it
is
w
o
r
th
w
h
i
le
i
n
v
e
s
ti
g
ati
n
g
if
s
o
cial
m
ed
ia
a
n
d
esp
ec
iall
y
,
a
n
o
n
y
m
o
u
s
s
o
cial
m
ed
ia
ca
n
b
e
a
g
o
o
d
p
r
ed
icto
r
o
f
s
en
ti
m
e
n
t.
A
n
o
n
y
m
i
t
y
a
f
f
o
r
d
s
a
d
eg
r
ee
o
f
p
r
o
tectio
n
to
th
e
co
m
m
e
n
ter
s
o
th
at
t
h
e
y
ca
n
ex
p
r
ess
th
e
ir
o
p
in
io
n
m
o
r
e
f
r
ee
l
y
.
T
h
is
is
e
s
p
ec
iall
y
r
elev
a
n
t
w
h
e
n
t
h
e
e
v
en
t
is
h
i
g
h
l
y
d
i
v
i
s
iv
e
an
d
a
h
i
g
h
s
o
cial
co
s
t
i
s
attac
h
ed
to
ex
p
r
ess
i
n
g
o
n
e’
s
o
p
in
io
n
f
r
an
k
l
y
.
R
ed
d
it
is
an
en
ter
tain
m
e
n
t,
n
e
w
s
a
n
d
g
e
n
er
al
p
u
r
p
o
s
e
s
o
cial
n
et
w
o
r
k
in
g
s
ite
w
h
er
e
m
e
m
b
er
s
o
f
th
e
co
m
m
u
n
it
y
s
u
b
m
it
eit
h
er
d
ir
ec
t
li
n
k
to
w
eb
s
ites
o
r
te
x
t
p
o
s
ts
.
R
ed
d
it
i
s
o
r
g
a
n
is
ed
b
y
to
p
ics
o
f
in
ter
e
s
t
i
n
to
s
u
b
-
co
m
m
u
n
it
ies
ca
lled
s
u
b
r
ed
d
its
.
T
h
e
m
e
m
b
er
s
p
o
s
ti
n
g
in
a
s
u
b
-
r
ed
d
it
ar
e
o
n
l
y
i
d
en
tifie
d
b
y
th
e
ir
u
s
er
n
a
m
e
s
th
er
e
f
o
r
e
af
f
o
r
d
in
g
a
d
eg
r
ee
o
f
an
o
n
y
m
it
y
o
n
R
ed
d
it.
Oth
er
m
e
m
b
er
s
o
f
t
h
e
co
m
m
u
n
it
y
ca
n
co
m
m
e
n
t
o
n
t
h
ese
p
o
s
t
s
.
T
h
ese
co
m
m
en
t
s
ar
e
th
e
n
u
p
-
v
o
ted
o
r
d
o
w
n
-
v
o
ted
r
ef
lect
in
g
th
e
v
ie
w
s
o
r
t
h
e
s
en
ti
m
e
n
ts
o
f
t
h
e
m
e
m
b
er
s
o
f
th
e
co
m
m
u
n
it
y
.
T
h
e
co
m
m
e
n
t
v
o
tes
ca
n
b
e
u
s
ed
to
m
ea
s
u
r
e
th
e
s
e
n
ti
m
en
t
in
th
at
p
ar
ticu
lar
s
u
b
-
co
m
m
u
n
it
y
w
h
ic
h
m
a
y
d
i
f
f
er
f
r
o
m
s
en
t
i
m
en
ts
i
n
o
th
er
s
u
b
r
ed
d
its
.
R
e
d
d
it
co
m
m
e
n
ts
ca
n
b
e
m
u
lti
-
p
ar
ag
r
ap
h
a
n
d
ca
n
a
ls
o
co
n
tai
n
l
in
k
s
,
e
m
o
j
is
a
n
d
i
m
a
g
es.
T
h
e
co
m
p
lex
n
at
u
r
e
o
f
t
h
ese
co
m
m
en
ts
co
n
tain
i
n
g
d
if
f
er
en
t
m
ed
ia,
r
h
eto
r
ical
f
lo
u
r
is
h
es a
n
d
s
ar
ca
s
m
ca
n
b
e
ch
al
len
g
i
n
g
f
o
r
s
en
ti
m
en
t a
n
al
y
s
is
.
I
n
t
h
is
p
ap
er
,
w
e
p
er
f
o
r
m
ed
f
i
n
e
-
g
r
ai
n
ed
s
e
n
ti
m
e
n
t
a
n
al
y
s
i
s
to
d
eter
m
in
e
if
w
e
co
u
ld
class
i
f
y
s
en
ti
m
e
n
t
r
elate
d
to
a
p
ar
ticu
l
ar
ev
en
t.
W
e
tr
ied
to
d
eter
m
i
n
e
if
co
m
m
e
n
ter
s
v
ie
w
ed
a
ce
r
tain
ev
e
n
t
p
o
s
iti
v
el
y
o
r
n
eg
ati
v
el
y
a
n
d
s
a
m
p
led
f
r
o
m
m
u
lt
ip
le
s
u
b
-
r
ed
d
its
.
Fo
r
th
e
ev
e
n
t,
w
e
ch
o
s
e
t
h
e
r
ef
er
e
n
d
u
m
co
n
d
u
cted
i
n
B
r
itain
as
to
w
h
et
h
er
B
r
itain
s
h
o
u
ld
leav
e
t
h
e
E
u
r
o
p
ea
n
u
n
io
n
(
k
n
o
w
n
as
B
r
ex
it)
as
th
is
w
a
s
a
m
aj
o
r
h
is
to
r
ical
ev
e
n
t
w
it
h
a
m
p
le
d
i
s
cu
s
s
io
n
b
e
f
o
r
e
an
d
af
ter
th
e
ev
en
t.
T
h
is
a
f
f
o
r
d
ed
u
s
th
e
p
o
s
s
ib
ilit
y
to
co
llect
an
d
an
al
y
ze
lar
g
e
a
m
o
u
n
ts
o
f
d
ata
2.
DATAS
E
T
I
n
th
i
s
p
ap
er
,
w
e
ex
tr
ac
ted
c
o
m
m
e
n
t
s
f
r
o
m
p
o
s
ts
i
n
t
w
o
s
p
ec
if
ic
s
u
b
r
ed
d
its
co
n
ce
n
tr
at
ed
o
n
th
e
Un
ited
K
in
g
o
m
-
r
/
u
n
ited
k
in
g
d
o
m
an
d
r
/
u
k
p
o
liti
c
s
.
T
h
ese
t
w
o
s
u
b
r
ed
d
its
h
a
v
e
a
lar
g
e
n
u
m
b
er
o
f
s
u
b
s
cr
ib
er
s
-
1
3
4
,
9
1
7
an
d
5
6
,
4
9
3
r
esp
ec
tiv
el
y
.
W
e
f
ir
s
t
f
il
ter
ed
p
o
s
ts
t
h
at
h
ad
t
h
e
k
e
y
w
o
r
d
s
”b
r
ex
it”
o
r
”r
ef
er
en
d
u
m
”
i
n
th
e
titl
e.
P
o
s
ts
w
it
h
ti
m
e
s
ta
m
p
s
r
an
g
i
n
g
f
r
o
m
J
u
n
e
1
,
2
0
1
6
-
J
u
l
y
3
1
,
2
0
1
6
w
er
e
in
cl
u
d
ed
in
th
e
s
t
u
d
y
.
T
h
is
ti
m
e
p
er
io
d
co
v
er
ed
th
e
p
r
e
an
d
p
o
s
t
r
ef
er
en
d
u
m
(
co
n
d
u
cted
o
n
J
u
n
e
2
3
,
2
0
1
6
)
p
e
r
io
d
.
A
ll
p
o
s
ts
an
d
co
m
m
e
n
t
s
w
er
e
ex
tr
ac
ted
u
s
in
g
th
e
R
ed
d
it A
P
I
p
latf
o
r
m
i
m
p
le
m
e
n
ted
in
P
y
t
h
o
n
2
.
7
.
1
2
4
,
7
8
8
co
m
m
en
t
s
w
er
e
ex
tr
ac
ted
to
tall
y
f
r
o
m
t
h
e
p
o
s
t
s
ac
r
o
s
s
all
ep
o
ch
o
f
t
h
e
t
w
o
s
u
b
r
ed
d
its
.
W
e
p
ick
ed
a
s
u
b
s
et
o
f
ap
p
r
o
x
i
m
at
el
y
3
0
%
o
f
t
h
e
co
m
m
e
n
t
s
f
o
r
th
e
tr
ai
n
in
g
a
n
d
test
s
et
(
2
4
,
4
2
3
co
m
m
en
ts
)
.
W
e
d
iv
id
ed
th
e
co
m
m
e
n
t
s
in
to
f
i
v
e
r
o
u
g
h
l
y
7
-
1
5
d
ay
ep
o
ch
s
:
1
-
1
7
J
u
n
e,
1
7
-
2
3
J
u
n
e,
2
4
J
u
n
e
-
1
J
u
l
y
,
1
-
1
6
J
u
ly
an
d
1
6
-
3
1
J
u
l
y
.
T
h
e
co
m
m
en
t
s
w
er
e
s
e
lecte
d
r
an
d
o
m
l
y
f
r
o
m
ea
c
h
ep
o
ch
in
p
r
o
p
o
r
tio
n
to
th
e
to
tal
n
u
m
b
er
o
f
co
m
m
e
n
t
s
u
s
in
g
s
tr
ati
f
ied
r
an
d
o
m
s
a
m
p
li
n
g
.
T
h
is
w
a
s
d
o
n
e
to
en
s
u
r
e
t
h
at
w
e
d
id
n
’
t
ac
cid
en
tl
y
o
v
er
l
y
s
a
m
p
l
e
f
r
o
m
a
ce
r
tai
n
ti
m
e
p
er
io
d
.
W
e
also
s
elec
ted
p
o
s
ts
f
r
o
m
o
t
h
er
s
u
b
r
ed
d
its
s
u
ch
a
s
r
/
w
o
r
ld
n
e
w
s
a
n
d
r
/p
o
liti
cs
th
at
h
ad
”B
r
ex
it”
o
r
”E
U
r
ef
er
en
d
u
m
”
p
o
s
ted
w
it
h
in
t
h
e
ti
m
esp
an
u
n
d
er
co
n
s
id
er
atio
n
(
1
5
4
5
co
m
m
e
n
ts
)
T
h
e
co
m
m
e
n
t
s
f
r
o
m
t
h
ese
p
o
s
ts
wer
e
ex
tr
ac
ted
an
d
ad
d
e
d
to
th
e
s
u
b
s
et.
T
h
ese
co
m
m
e
n
t
s
w
er
e
th
en
an
n
o
tated
b
y
t
w
o
h
u
m
a
n
an
n
o
tater
s
.
W
e
o
n
l
y
k
ep
t t
h
o
s
e
an
n
o
tatio
n
s
o
n
wh
ich
b
o
th
t
h
e
a
n
n
o
tater
s
a
g
r
ee
d
.
T
h
e
d
ata
w
as
p
r
ep
r
o
ce
s
s
ed
b
y
r
e
m
o
v
i
n
g
ex
tr
a
w
h
ites
p
ac
e
s
,
r
em
o
v
i
n
g
all
p
u
n
ct
u
atio
n
a
n
d
all
tex
t
w
a
s
co
n
v
er
ted
to
lo
w
er
ca
s
e.
W
e
th
en
p
er
f
o
r
m
ed
an
al
y
s
is
o
n
t
w
o
le
v
els.
T
h
e
tr
ain
in
g
s
et
w
a
s
an
n
o
tated
o
n
o
n
e
lev
el
as
p
o
s
itiv
e,
n
eg
at
iv
e
a
n
d
n
eu
tr
al.
T
h
ese
p
o
lar
ities
w
er
e
t
h
en
f
u
r
t
h
er
s
p
l
it
in
to
s
i
x
d
if
f
er
en
t
s
e
n
ti
m
e
n
t
p
o
lar
ities
:
p
o
s
itiv
e
to
w
ar
d
s
leav
i
n
g
th
e
E
U
(
P
L
)
,
p
o
s
itiv
e
to
w
ar
d
s
r
e
m
ai
n
i
n
g
i
n
th
e
E
U
(
P
R
)
,
n
eg
ativ
e
to
w
a
r
d
s
leav
in
g
th
e
E
U
(
NR
)
,
n
eg
at
iv
e
to
w
ar
d
s
s
ta
y
i
n
g
in
t
h
e
E
U
(
N
L
)
,
n
eg
at
iv
e
to
w
ar
d
s
r
e
m
a
in
s
u
p
p
o
r
ter
s
(
NL
R
)
an
d
n
e
u
tr
al.
Af
ter
r
e
m
o
v
i
n
g
b
lan
k
p
ar
ag
r
a
p
h
s
a
n
d
s
o
o
n
,
w
e
e
n
d
ed
u
p
w
it
h
2
0
7
5
0
co
m
m
e
n
t
s
i
n
t
h
e
t
r
ain
in
g
s
et
an
d
5
1
8
7
co
m
m
e
n
t
s
i
n
th
e
t
est
s
et
f
o
r
th
e
f
ir
s
t
an
al
y
s
is
af
ter
a
8
0
:2
0
s
p
l
it
o
f
t
h
e
o
r
ig
in
a
l
s
et
a
n
d
2
0
7
7
4
co
m
m
e
n
t
s
i
n
t
h
e
tr
ai
n
i
n
g
s
et
an
d
5
1
6
2
co
m
m
en
ts
i
n
test
s
et
f
o
r
t
h
e
s
ec
o
n
d
an
al
y
s
is
.
T
h
e
d
if
f
er
e
n
ce
i
n
t
h
e
n
u
m
b
er
o
f
co
m
m
e
n
ts
i
s
d
u
e
t
o
th
e
f
ac
t
t
h
at
t
h
e
an
al
y
s
is
w
a
s
r
u
n
s
ep
ar
atel
y
o
n
t
h
e
t
w
o
s
e
ts
b
ec
au
s
e
t
h
e
p
ick
s
w
er
e
r
an
d
o
m
ized
a
n
d
th
e
r
atio
s
p
lits
v
ar
ied
s
li
g
h
tl
y
.
3.
AL
G
O
RI
T
H
M
Ma
ch
i
n
e
lear
n
in
g
al
g
o
r
it
h
m
s
f
o
r
ap
p
licatio
n
s
li
k
e
s
e
n
ti
m
e
n
t
an
al
y
s
is
r
eq
u
ir
e
te
x
t
to
b
e
tr
an
s
f
o
r
m
ed
in
to
a
f
ix
ed
-
le
n
g
th
v
ec
to
r
s
u
itab
le
f
o
r
p
r
o
ce
s
s
in
g
.
T
h
e
co
m
m
o
n
b
ag
-
of
-
w
o
r
d
s
(
B
o
W
)
ap
p
r
o
ac
h
w
h
ic
h
r
ep
r
esen
ts
te
x
t/d
o
cu
m
e
n
ts
in
t
er
m
s
o
f
a
v
ec
to
r
o
f
w
o
r
d
/to
k
en
f
r
eq
u
en
c
y
h
as
s
o
m
e
d
r
a
w
b
ac
k
s
-
w
o
r
d
o
r
d
er
is
lo
s
t
an
d
s
o
m
e
o
f
th
e
s
e
m
a
n
tic
s
as
s
o
ciate
d
w
it
h
t
h
e
w
o
r
d
i
n
tex
t
i
s
al
s
o
lo
s
t.
Vec
to
r
r
ep
r
esen
tatio
n
s
o
f
w
o
r
d
s
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
A
n
a
lysi
n
g
E
ve
n
t
-
R
el
a
ted
S
e
n
timen
ts
o
n
S
o
cia
l Med
i
a
w
ith
N
eu
r
a
l Netw
o
r
k
s
(
P
.
S
a
n
th
i P
r
iya
)
121
h
av
e
atte
m
p
ted
to
r
em
ed
y
s
o
m
e
o
f
th
e
s
e
d
r
a
w
b
ac
k
s
.
I
n
W
o
r
d
2
Vec
,
v
ec
to
r
r
e
p
r
esen
tatio
n
s
ar
e
co
m
p
u
ted
f
o
r
ea
ch
w
o
r
d
.
I
ts
in
p
u
t is a
te
x
t c
o
r
p
u
s
an
d
its
o
u
tp
u
t is a
s
et
o
f
f
ea
t
u
r
e
v
ec
to
r
s
f
o
r
w
o
r
d
s
i
n
t
h
at
co
r
p
u
s
.
W
o
r
d
2
v
ec
is
a
t
w
o
-
la
y
er
n
eu
r
al
n
et
th
at
p
r
o
ce
s
s
es
tex
t.
I
ts
i
n
p
u
t
is
a
tex
t
co
r
p
u
s
an
d
i
ts
o
u
tp
u
t
is
a
s
et
o
f
v
ec
to
r
s
: f
ea
t
u
r
e
v
ec
to
r
s
f
o
r
w
o
r
d
s
in
th
at
co
r
p
u
s
.
I
n
W
o
r
d
2
Vec
,
a
d
is
tr
ib
u
ted
r
ep
r
ese
n
tatio
n
o
f
a
w
o
r
d
i
s
u
s
ed
.
Giv
e
n
a
f
ea
tu
r
e
v
ec
to
r
w
it
h
s
e
v
er
al
h
u
n
d
r
ed
d
im
e
n
s
i
o
n
s
,
ea
ch
w
o
r
d
is
r
ep
r
esen
ted
b
y
a
d
is
tr
ib
u
tio
n
o
f
w
ei
g
h
ts
ac
r
o
s
s
th
o
s
e
ele
m
en
t
s
a
n
d
ea
ch
ele
m
en
t
i
n
t
h
e
v
e
cto
r
co
n
tr
ib
u
tes
to
t
h
e
d
e
f
in
it
io
n
o
f
m
an
y
w
o
r
d
s
.
T
h
e
n
eu
r
al
n
et
w
o
r
k
b
ased
wo
r
d
v
ec
to
r
s
ar
e
u
s
u
all
y
tr
ai
n
ed
u
s
i
n
g
s
to
ch
as
tic
g
r
ad
ien
t
d
escen
t
w
h
er
e
t
h
e
g
r
ad
ien
t is o
b
tai
n
ed
v
ia
b
ac
k
p
r
o
p
ag
atio
n
.
A
m
o
r
e
r
ef
i
n
ed
ap
p
r
o
ac
h
i
n
v
o
lv
e
s
p
ar
ag
r
ap
h
v
ec
to
r
s
w
h
ich
w
o
r
k
a
t
th
e
lev
e
l
o
f
s
e
n
ten
ce
s
o
r
d
o
cu
m
en
ts
.
T
r
ain
in
g
w
o
r
d
v
ec
to
r
s
o
cc
u
r
s
as
n
o
r
m
al,
e
x
ce
p
t
th
at
a
n
ad
d
itio
n
al
v
ec
t
o
r
r
ep
r
esen
tin
g
th
e
p
ar
ag
r
ap
h
is
ad
d
ed
to
th
e
task
w
h
e
n
ev
er
t
h
e
s
a
m
p
led
w
i
n
d
o
w
co
m
e
s
f
r
o
m
th
a
t p
ar
ag
r
ap
h
.
I
n
th
e
P
ar
ag
r
ap
h
Vec
to
r
(
Do
c2
Vec
)
[
9
]
f
r
am
e
wo
r
k
,
ev
er
y
p
ar
ag
r
ap
h
is
m
ap
p
ed
to
a
u
n
iq
u
e
v
ec
to
r
an
d
ev
e
r
y
w
o
r
d
is
also
m
ap
p
ed
to
a
u
n
iq
u
e
v
ec
t
o
r
lik
e
i
n
W
o
r
d
2
Vec
.
B
o
th
th
e
p
ar
ag
r
ap
h
v
ec
to
r
s
an
d
w
o
r
d
v
ec
to
r
s
ar
e
tr
ain
ed
u
s
i
n
g
s
to
ch
a
s
tic
g
r
ad
ien
t
d
es
ce
n
t
an
d
t
h
e
g
r
ad
ien
t
is
o
b
tain
ed
v
ia
b
ac
k
p
r
o
p
ag
atio
n
.
T
h
e
o
b
tain
ed
f
ea
tu
r
e
v
ec
to
r
s
ca
n
b
e
u
s
ed
a
s
i
n
p
u
t
s
to
co
n
v
e
n
tio
n
al
m
ac
h
i
n
e
lear
n
in
g
alg
o
r
it
h
m
s
s
u
c
h
a
s
clas
s
i
f
i
er
s
,
etc.
P
ar
ag
r
ap
h
v
ec
to
r
s
i
n
h
er
it
a
n
i
m
p
o
r
tan
t
p
r
o
p
er
ty
o
f
th
e
w
o
r
d
v
ec
to
r
s
:
th
e
s
e
m
a
n
tic
s
o
f
t
h
e
w
o
r
d
s
.
i.e
.
it
g
r
o
u
p
s
t
h
e
v
ec
to
r
s
o
f
s
i
m
ilar
w
o
r
d
s
to
g
et
h
er
in
v
ec
to
r
s
p
ac
e.
T
h
e
s
ec
o
n
d
ad
v
an
t
ag
e
o
f
th
e
p
ar
ag
r
ap
h
v
ec
to
r
s
is
th
at
t
h
e
y
ta
k
e
in
to
co
n
s
id
er
atio
n
t
h
e
w
o
r
d
o
r
d
er
an
d
r
etain
s
o
m
e
i
n
f
o
r
m
atio
n
a
b
o
u
t
th
e
co
n
te
x
t.
W
e
co
n
s
id
er
ed
th
is
ap
p
r
o
ac
h
to
co
n
s
tr
u
ct
i
n
g
f
ea
t
u
r
e
v
ec
to
r
s
m
o
s
t
s
u
i
tab
le
f
o
r
o
u
r
a
n
al
y
s
i
s
an
d
tr
an
s
f
o
r
m
ed
o
u
r
co
r
p
u
s
in
to
f
ea
tu
r
e
v
ec
to
r
s
u
s
i
n
g
t
h
e
g
e
n
s
i
m
i
m
p
le
m
e
n
ta
tio
n
o
f
t
h
e
Do
c2
Vec
al
g
o
r
ith
m
[
4
]
.
Af
ter
tr
an
s
f
o
r
m
atio
n
,
ea
ch
co
m
m
e
n
t
w
as
r
ep
r
esen
ted
b
y
a
f
ix
ed
-
le
n
g
th
f
ea
tu
r
e
v
ec
to
r
w
it
h
a
d
i
m
e
n
s
io
n
alit
y
o
f
1
0
0
.
4.
M
E
T
H
O
D
T
h
e
f
ea
tu
r
e
v
ec
to
r
s
o
b
tain
ed
f
r
o
m
t
h
e
Do
c2
Vec
p
r
o
ce
s
s
wer
e
s
tan
d
ar
d
ized
b
y
r
e
m
o
v
i
n
g
th
e
m
ea
n
an
d
s
ca
li
n
g
to
u
n
i
t v
ar
ia
n
ce
.
A
Mu
lti
L
a
y
er
P
er
ce
p
tr
o
n
C
las
s
if
ier
(
ML
P
)
[
1
0
]
w
a
s
th
e
n
tr
ai
n
ed
o
n
th
e
s
e
s
ca
led
an
d
ce
n
ter
ed
f
ea
t
u
r
e
v
ec
to
r
s
o
f
th
e
tr
ai
n
i
n
g
s
et.
W
e
f
o
u
n
d
th
e
b
est
p
ar
am
eter
s
t
h
r
o
u
g
h
a
g
r
id
s
ea
r
ch
alg
o
r
it
h
m
w
h
ic
h
p
er
f
o
r
m
s
e
x
h
a
u
s
tiv
e
s
e
ar
ch
es
o
v
er
s
p
ec
if
ied
p
ar
a
m
et
er
s
.
A
k
-
f
o
ld
(
5
-
f
o
ld
)
cr
o
s
s
v
alid
atio
n
ap
p
r
o
ac
h
w
a
s
u
s
ed
to
tu
n
e
t
h
e
clas
s
i
f
ier
an
d
p
ick
p
ar
a
m
eter
s
w
h
ic
h
y
ie
ld
ed
th
e
b
est ac
cu
r
ac
y
.
W
e
th
en
r
a
n
a
n
ev
a
lu
atio
n
(
te
s
t)
s
et
t
h
r
o
u
g
h
th
e
o
p
ti
m
ized
class
i
f
ier
to
e
v
alu
a
te
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
class
i
f
ier
o
n
h
i
th
er
to
u
n
s
e
en
d
ata.
W
e
test
ed
d
if
f
er
en
t
a
lg
o
r
ith
m
s
a
n
d
ac
tiv
atio
n
f
u
n
c
tio
n
s
f
o
r
th
e
M
L
P
class
i
f
ier
,
f
o
r
ex
a
m
p
le,
L
-
B
F
GS,
A
d
a
m
an
d
Sto
ch
a
s
tic
g
r
a
d
ien
t d
esce
n
t (
SG
D)
.
Fo
r
t
h
e
f
ir
s
t c
ase
o
f
a
s
s
i
g
n
in
g
th
r
ee
s
e
n
ti
m
e
n
t
p
o
lar
ities
,
w
e
o
b
tain
ed
th
e
b
e
s
t
r
es
u
lt
s
w
it
h
th
e
L
-
B
FGS
alg
o
r
it
h
m
[
2
]
an
d
a
r
ec
tif
ied
li
n
ea
r
u
n
i
t
ac
tiv
atio
n
f
u
n
c
tio
n
.
Fo
r
th
e
s
ec
o
n
d
ca
s
e,
w
e
o
b
tain
ed
th
e
b
e
s
t
r
esu
lt
s
w
ith
t
h
e
A
d
a
m
a
lg
o
r
it
h
m
a
n
d
a
h
y
p
er
b
o
lic
tan
ac
ti
v
atio
n
f
u
n
c
t
io
n
[
8
]
.
Up
o
n
an
n
o
tatio
n
in
to
t
h
e
d
if
f
er
en
t
p
o
lar
ities
,
w
e
d
is
co
v
er
e
d
th
at
o
u
r
d
ataset
h
ad
m
o
r
e
ex
a
m
p
le
s
o
f
n
eg
at
iv
e
a
n
d
n
e
u
tr
al
s
en
ti
m
en
ts
th
a
n
p
o
s
iti
v
e
s
en
ti
m
en
t
s
.
Fo
r
th
e
f
ir
s
t
a
n
al
y
s
i
s
t
y
p
e,
w
e
o
b
tain
ed
th
e
tr
ai
n
i
n
g
an
d
test
s
et
b
y
p
ic
k
i
n
g
p
r
o
p
o
r
tio
n
atel
y
f
r
o
m
d
i
f
f
er
e
n
t
cla
s
s
es
u
s
i
n
g
s
tr
ati
f
ied
s
a
m
p
l
in
g
.
[
1
0
]
.
I
n
th
e
s
ec
o
n
d
t
y
p
e
o
f
an
al
y
s
i
s
w
h
er
e
w
e
w
e
r
e
class
if
y
i
n
g
th
e
d
ata
in
to
s
i
x
d
if
f
er
en
t
cla
s
s
es,
t
h
e
NR
a
n
d
n
eg
ati
v
e
p
o
lar
it
y
w
a
s
u
n
d
er
s
a
m
p
led
to
7
0
%
o
f
th
e
clas
s
in
o
r
d
er
to
b
alan
ce
th
e
tr
ain
i
n
g
s
et
a
n
d
n
o
t
b
ias
t
h
e
class
i
f
ier
to
w
ar
d
s
n
eg
at
iv
e
a
n
d
n
e
u
tr
al
s
e
n
ti
m
en
t
s
.
5.
RE
SU
L
T
S
Fo
r
o
u
r
f
ir
s
t
a
n
al
y
s
is
,
w
e
at
te
m
p
ted
class
if
ica
tio
n
i
n
to
t
h
r
e
e
s
en
t
i
m
e
n
t
p
o
lar
ities
:
p
o
s
iti
v
e,
n
eg
at
iv
e
an
d
n
eu
tr
al.
W
e
tu
n
ed
o
u
r
p
ar
a
m
eter
s
b
y
co
n
d
u
cti
n
g
a
g
r
id
s
ea
r
ch
w
ith
i
n
a
cr
o
s
s
-
v
alid
atio
n
lo
o
p
u
s
in
g
t
h
e
n
ested
cr
o
s
s
-
v
alid
atio
n
p
ar
ad
ig
m
[
3
]
.
T
h
e
h
y
p
er
p
ar
a
m
eter
s
w
er
e
tu
n
ed
i
n
t
h
e
i
n
n
er
cr
o
s
s
-
v
alid
atio
n
lo
o
p
(
s
tr
atif
ied
K
-
f
o
ld
w
it
h
3
s
p
lits
)
w
h
i
le
th
e
g
e
n
er
aliza
tio
n
er
r
o
r
w
as
m
ea
s
u
r
ed
o
v
er
s
ev
er
al
d
ataset
s
p
lits
in
t
h
e
o
u
ter
cr
o
s
s
-
v
alid
atio
n
lo
o
p
(
s
tr
atif
ied
K
-
f
o
ld
w
ith
3
s
p
lit
s
)
.
T
h
e
p
ar
am
eter
s
th
a
t
w
er
e
v
ar
i
ed
w
er
e
th
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
i
n
t
h
e
h
id
d
en
la
y
er
an
d
th
e
L
2
-
r
eg
u
lar
izat
io
n
p
ar
a
m
eter
(
alp
h
a)
a
n
d
w
e
u
s
ed
F1
-
w
e
ig
h
ted
s
co
r
e
in
o
r
d
er
to
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
.
Ou
r
h
id
d
en
la
y
er
p
ar
a
m
eter
s
w
er
e:
5
0
,
1
0
0
o
r
2
0
0
n
eu
r
o
n
s
in
a
s
i
n
g
le
h
id
d
en
la
y
er
;
an
d
o
u
r
alp
h
a
p
a
r
a
m
eter
s
w
er
e:
0
.
0
0
0
1
,
0
.
0
0
1
,
0
.
0
1
,
0
.
1
.
T
h
e
cr
o
s
s
v
alid
atio
n
p
r
o
ce
s
s
y
ield
ed
a
m
ea
n
F1
-
s
co
r
e
o
f
0
.
5
1
8
(
th
is
co
r
r
esp
o
n
d
ed
to
a
m
ea
n
clas
s
i
f
icatio
n
ac
cu
r
ac
y
o
f
6
0
%
e
s
ti
m
ated
i
n
a
s
ep
ar
ate
r
u
n
w
i
th
th
e
s
a
m
e
tr
ain
in
g
s
e
t
)
an
d
a
co
m
b
in
at
io
n
o
f
h
id
d
e
n
la
y
er
p
ar
a
m
eter
o
f
5
0
n
e
u
r
o
n
s
a
n
d
an
a
lp
h
a
o
f
0
.
0
0
0
1
y
ield
ed
th
e
b
est
m
ea
n
F1
-
s
co
r
e
o
f
0
.
5
2
7
as sh
o
w
n
i
n
Fig
u
r
e
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
7
,
No
.
3
,
Sep
tem
b
e
r
20
1
8
:
11
9
–
12
4
122
Fig
u
r
e
1
.
R
esu
lts
o
f
Gr
id
-
Sear
ch
A
l
g
o
r
ith
m
f
o
r
A
n
a
l
y
s
is
1
.
Me
an
F1
-
s
co
r
e
v
s
L
2
-
R
e
g
u
lar
izatio
n
;
L
eg
e
n
d
:
Hid
d
en
L
a
y
er
(
HL
)
N
u
m
b
er
o
f
Neu
r
o
n
s
F
i
g
u
r
e 2
.
a)
C
o
n
f
u
s
io
n
Ma
tr
ix
f
o
r
An
al
y
s
is
F
i
g
u
r
e 2
.
b
)
No
r
m
alized
C
o
n
f
u
s
i
o
n
Ma
tr
ix
f
o
r
An
al
y
s
is
F
i
g
u
r
e 2
.
C
o
n
f
u
s
io
n
Ma
tr
ices
f
o
r
An
al
y
s
is
1
W
e
th
en
r
an
t
h
e
te
s
t
s
et
t
h
r
o
u
g
h
t
h
e
o
p
ti
m
ized
,
tu
n
ed
clas
s
if
e
r
an
d
o
b
tain
ed
an
ac
cu
r
ac
y
o
f
5
5
.
6
%.
A
co
n
f
u
s
io
n
m
a
tr
ix
s
h
o
w
ed
th
at
o
u
r
class
i
f
ier
w
as
esp
ec
ial
l
y
p
r
o
n
e
to
m
is
cla
s
s
i
f
y
in
g
p
o
s
iti
v
e
s
a
m
p
le
s
a
s
s
h
o
w
n
in
Fig
u
r
e
2
.
W
e
t
h
en
s
o
u
g
h
t
t
o
i
m
p
r
o
v
e
o
u
r
m
o
d
el
b
y
f
u
r
t
h
er
s
p
litt
i
n
g
th
e
s
e
n
ti
m
e
n
t
p
o
la
r
ities
.
T
h
e
s
e
n
ti
m
e
n
t
p
o
lar
ities
w
er
e
s
p
lit
in
to
:
s
i
x
d
if
f
er
e
n
t
s
e
n
ti
m
en
t
p
o
lar
ities
:
p
o
s
itiv
e
to
w
ar
d
s
leav
in
g
th
e
E
U
(
P
L
)
,
p
o
s
itiv
e
to
w
ar
d
s
r
e
m
ai
n
in
g
in
t
h
e
E
U
(
P
R
)
,
n
eg
ati
v
e
to
w
ar
d
s
leav
in
g
t
h
e
E
U
(
NR
)
,
n
e
g
ati
v
e
to
war
d
s
s
ta
y
in
g
in
th
e
E
U
(
NL
)
,
n
eg
at
iv
e
to
w
ar
d
s
r
e
m
ai
n
s
u
p
p
o
r
ter
s
(
NL
R
)
an
d
n
eu
tr
al
as
d
escr
ib
ed
p
r
ev
io
u
s
l
y
.
W
e
o
p
tim
ized
t
h
e
class
i
f
er
u
s
i
n
g
t
h
e
s
a
m
e
p
ar
am
eter
s
p
ac
e
d
escr
ib
ed
ab
o
v
e.
T
h
e
cr
o
s
s
v
al
i
d
atio
n
p
r
o
ce
s
s
y
ield
ed
a
m
ea
n
F1
-
s
co
r
e
o
f
0
.
5
4
9
(
th
is
co
r
r
esp
o
n
d
ed
to
a
m
ea
n
cla
s
s
i
f
icatio
n
a
cc
u
r
ac
y
o
f
6
1
.
6
%
esti
m
ated
in
a
s
ep
ar
ate
r
u
n
w
it
h
th
e
s
a
m
e
tr
ain
i
n
g
s
et)
an
d
a
co
m
b
i
n
atio
n
o
f
h
id
d
en
la
y
er
p
ar
am
eter
o
f
1
0
0
n
e
u
r
o
n
s
a
n
d
an
alp
h
a
o
f
0
.
0
0
1
y
ield
ed
t
h
e
b
es
t
m
ea
n
F1
-
s
co
r
e
o
f
0
.
5
5
9
as
s
h
o
w
n
i
n
Fig
u
r
e
3
.
T
h
e
o
p
ti
m
ized
clas
s
if
ier
s
h
o
w
ed
i
m
p
r
o
v
ed
class
i
f
icatio
n
ac
cu
r
ac
y
o
f
6
0
.
2
% o
n
th
e
te
s
t set a
s
co
m
p
ar
ed
to
th
e
f
ir
s
t a
n
al
y
s
i
s
as s
h
o
w
n
i
n
Fi
g
u
r
e
4.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
A
n
a
lysi
n
g
E
ve
n
t
-
R
el
a
ted
S
e
n
timen
ts
o
n
S
o
cia
l Med
i
a
w
ith
N
eu
r
a
l Netw
o
r
k
s
(
P
.
S
a
n
th
i P
r
iya
)
123
Fig
u
r
e
3
.
R
esu
lts
o
f
Gr
id
-
Sear
ch
A
l
g
o
r
ith
m
f
o
r
A
n
a
l
y
s
is
2
.
Me
an
F1
s
co
r
e
v
s
L
2
-
r
e
g
u
lar
i
za
tio
n
;
L
eg
e
n
d
:
Hid
d
en
L
a
y
er
(
HL
)
N
u
m
b
er
o
f
Neu
r
o
n
s
Fig
u
r
e
4
.
a)
C
o
n
f
u
s
io
n
Ma
tr
ix
f
o
r
An
al
y
s
i
s
2
Fig
u
r
e
4
.
b
)
No
r
m
alize
d
C
o
n
f
u
s
io
n
Ma
tr
ix
f
o
r
An
al
y
s
i
F
i
g
u
r
e 4
.
C
o
n
f
u
s
io
n
Ma
tr
ices
f
o
r
An
al
y
s
is
2
W
e
f
o
u
n
d
th
a
t
th
e
clas
s
i
f
ier
s
t
ill
s
tr
u
g
g
led
w
i
th
ac
c
u
r
atel
y
c
lass
i
f
y
in
g
p
o
s
iti
v
e
s
e
n
ti
m
en
ts
as
b
ef
o
r
e
alth
o
u
g
h
t
h
e
o
v
er
all
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
cla
s
s
i
f
ier
i
m
p
r
o
v
ed
w
it
h
th
e
s
ec
o
n
d
ap
p
r
o
ac
h
.
W
e
b
eliev
e
th
a
t
t
h
is
i
s
d
u
e
to
th
e
n
at
u
r
e
o
f
t
h
e
d
at
a
its
el
f
an
d
t
h
at
o
v
er
al
l
p
o
s
i
tiv
e
s
tate
m
en
t
s
m
i
g
h
t
d
is
p
la
y
a
h
ig
h
d
eg
r
ee
o
f
v
ar
iab
ilit
y
a
n
d
in
cl
u
d
e
n
e
s
ted
n
eg
at
iv
e
s
tate
m
en
t
s
as
w
e
o
b
s
er
v
ed
w
h
ile
a
n
n
o
tati
n
g
th
e
d
at
a.
6.
CO
NCLU
SI
O
N
Ou
r
s
t
u
d
y
w
as
co
n
d
u
cted
to
ass
es
s
i
f
f
i
n
e
-
g
r
ai
n
ed
s
en
ti
m
e
n
t
a
n
al
y
s
is
co
u
ld
b
e
p
er
f
o
r
m
ed
w
it
h
a
d
ataset
ex
tr
ac
ted
f
r
o
m
a
co
m
p
lex
ec
o
s
y
s
te
m
lik
e
R
ed
d
it.
W
e
p
er
f
o
r
m
ed
s
e
n
ti
m
en
t
a
n
a
l
y
s
i
s
an
d
as
s
es
s
ed
o
v
er
all
s
en
t
i
m
e
n
t
co
n
ta
in
ed
i
n
m
u
lti
-
s
en
te
n
ce
an
d
m
u
lti
-
p
ar
a
g
r
ap
h
b
lo
ck
s
o
f
co
m
m
e
n
ts
.
Fo
r
th
is
r
ea
s
o
n
,
w
e
u
s
ed
n
e
w
ap
p
r
o
ac
h
es
to
co
n
s
tr
u
cti
n
g
f
ea
t
u
r
e
v
ec
to
r
s
s
u
c
h
a
s
p
ar
ag
r
ap
h
v
ec
to
r
s
th
a
t
ca
n
b
e
ap
p
lied
to
v
ar
iab
le
-
le
n
g
th
p
iece
s
o
f
tex
t.
W
e
atte
m
p
ted
to
d
eter
m
i
n
e
if
w
e
co
u
ld
ac
c
u
r
atel
y
cla
s
s
i
f
y
i
f
co
m
m
e
n
ter
s
w
er
e
o
v
er
all
p
o
s
iti
v
el
y
o
r
n
e
g
ati
v
el
y
v
ie
w
i
n
g
a
p
ar
tic
u
lar
e
v
en
t.
W
e
th
e
n
atte
m
p
ted
to
d
eter
m
i
n
e
t
h
e
d
ir
ec
tio
n
a
lit
y
o
f
t
h
e
co
m
m
e
n
t
’
s
p
o
s
itiv
e
o
r
n
eg
ati
v
e
s
en
ti
m
en
t
w
it
h
m
o
r
e
f
i
n
e
-
g
r
ain
ed
a
n
al
y
s
i
s
,
f
o
r
ex
a
m
p
le,
i
f
a
co
m
m
e
n
ter
w
as
p
o
s
iti
v
el
y
v
i
e
w
i
n
g
leav
i
n
g
o
r
r
e
m
ai
n
i
n
g
i
n
th
e
E
U.
W
e
f
o
u
n
d
th
at
o
u
r
m
o
d
el
w
as
ab
le
to
class
i
f
y
o
v
er
all
p
o
s
itiv
e
a
n
d
n
eg
at
iv
e
s
tate
m
en
t
s
b
u
t
also
b
e
ab
le
to
d
is
tin
g
u
i
s
h
t
h
e
d
i
r
ec
tio
n
alit
y
o
f
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
7
,
No
.
3
,
Sep
tem
b
e
r
20
1
8
:
11
9
–
12
4
124
s
tate
m
en
t
s
.
W
e
f
o
u
n
d
t
h
at
t
h
er
e
w
er
e
m
an
if
e
s
t
d
i
f
f
er
e
n
ce
s
in
o
u
r
c
lass
if
ier
’
s
ab
ilit
y
to
clas
s
i
f
y
s
e
n
ti
m
e
n
ts
.
Ou
r
class
if
er
w
as
i
n
b
o
th
m
o
d
es
o
f
an
al
y
s
i
s
b
etter
at
class
i
f
y
in
g
n
e
g
ati
v
e
an
d
n
e
u
tr
al
s
e
n
ti
m
en
ts
t
h
an
p
o
s
iti
v
e
s
en
ti
m
e
n
ts
a
n
d
s
h
o
w
ed
a
h
i
g
h
d
eg
r
ee
o
f
co
n
f
u
s
io
n
b
et
w
ee
n
p
o
s
itiv
e
an
d
n
eg
ati
v
e
s
tate
m
e
n
ts
.
W
e
b
eliev
e
t
h
at
th
is
e
f
f
ec
t
m
ig
h
t
b
e
d
u
e
to
a
f
e
w
i
s
s
u
es:
t
h
e
p
r
ep
o
n
d
er
an
ce
o
f
n
e
g
ativ
e
s
tate
m
en
ts
i
n
th
e
o
v
er
all
d
ataset
an
d
th
e
i
n
ter
lea
v
in
g
o
f
n
eg
at
iv
e
s
tate
m
en
t
s
w
i
th
i
n
o
v
er
all
p
o
s
iti
v
e
s
ta
te
m
e
n
t
s
as
o
b
s
er
v
ed
an
ec
d
o
tall
y
w
h
ile
an
n
o
tati
n
g
t
h
e
d
ataset.
W
e
a
tt
e
m
p
ted
to
o
f
f
s
et
th
e
f
ir
s
t
i
s
s
u
e
b
y
u
n
d
er
s
a
m
p
lin
g
t
h
e
m
aj
o
r
it
y
cla
s
s
e
s
a
n
d
th
i
s
s
h
o
w
ed
s
o
m
e
i
m
p
r
o
v
e
m
en
t
i
n
th
e
class
i
f
ica
tio
n
ac
cu
r
ac
y
.
T
o
ad
d
r
ess
th
e
s
ec
o
n
d
is
s
u
e,
in
f
u
tu
r
e
r
esear
ch
,
w
e
w
il
l
atte
m
p
t
f
u
r
t
h
er
g
r
ad
ed
class
i
f
icatio
n
,
f
o
r
ex
a
m
p
le,
s
o
m
e
w
h
at
p
o
s
iti
v
e,
s
o
m
e
w
h
at
n
e
g
a
tiv
e,
v
er
y
p
o
s
iti
v
e,
v
er
y
n
e
g
ati
v
e,
to
a
s
s
e
s
s
i
f
t
h
i
s
i
m
p
r
o
v
es
o
u
r
cla
s
s
i
f
icatio
n
ac
cu
r
ac
y
.
W
e
h
a
v
e
e
s
tab
lis
h
e
d
th
at
s
ati
s
f
ac
to
r
y
class
i
f
icatio
n
i
s
i
n
d
ee
d
p
o
s
s
ib
le
w
it
h
a
co
m
p
lex
d
ataset
s
u
ch
as
o
u
r
s
an
d
th
i
s
m
o
d
el
ca
n
b
e
u
s
ed
to
s
o
cial
m
ed
ia
to
ass
es
s
s
e
n
ti
m
e
n
ts
as
an
ad
d
itio
n
to
p
o
llin
g
d
ata.
RE
F
E
R
E
NC
E
S
[
1
]
L
.
L
e
e
B.
P
a
n
g
.
,
S
.
V
a
it
h
y
a
n
a
th
a
n
.
“
T
h
u
mb
s
u
p
?
S
e
n
ti
me
n
t
c
la
s
sifi
c
a
ti
o
n
u
sin
g
ma
c
h
in
e
le
a
rn
in
g
tec
h
n
iq
u
e
s
.
”
A
C
L
-
0
2
c
o
n
f
e
re
n
c
e
o
n
Em
p
iri
c
a
l
m
e
th
o
d
s in
n
a
tu
ra
l
la
n
g
u
a
g
e
p
ro
c
e
ss
in
g
,
V
o
l
u
m
e
1
0
,
2
0
0
2
.
[
2
]
Rich
a
rd
H.
By
rd
,
P
e
i
h
u
a
n
g
L
u
,
J
o
rg
e
No
c
e
d
a
l.
,
Ciy
o
u
Zh
u
.
“
A
li
m
it
e
d
m
e
m
o
r
y
a
lg
o
rit
h
m
f
o
r
b
o
u
n
d
c
o
n
stra
in
e
d
o
p
ti
m
iza
ti
o
n
.
”
S
IAM
J
o
u
r
n
a
l
o
n
S
c
ien
ti
fi
c
Co
mp
u
ti
n
g
,
1
6
(5
):
1
1
9
0
-
1
2
0
8
,
1
9
9
5
.
[
3
]
G
.
C.
Ca
w
l
e
y
.
,
N.
L
.
C.
Talb
o
t.
“
On
o
v
e
r
-
f
it
ti
n
g
in
m
o
d
e
l
se
lec
ti
o
n
a
n
d
su
b
s
e
q
u
e
n
t
se
lec
ti
o
n
b
ias
in
p
e
rf
o
rm
a
n
c
e
e
v
a
lu
a
ti
o
n
.
”
J
.
M
a
c
h
.
L
e
a
r
n
.
Res
,
1
1
:
2
0
7
9
-
2
1
0
7
.
,
2
0
1
0
.
[
4
]
Re
h
u
re
k
,
R
.
,
S
o
jk
a
,
P
.
“
S
o
f
twa
re
Fra
me
wo
rk
fo
r
T
o
p
ic
M
o
d
e
ll
i
n
g
wit
h
L
a
rg
e
C
o
rp
o
r
a
.
”
P
r
o
c
e
e
d
in
g
s
o
f
th
e
L
REC
2
0
1
0
W
o
rk
sh
o
p
o
n
Ne
w
Ch
a
ll
e
n
g
e
s
f
o
r
NL
P
F
ra
m
e
w
o
rk
s,
p
a
g
e
s 4
5
-
5
0
,
M
a
y
2
0
1
0
.
[
5
]
J.
X
.
Yu
G
.
P
.
C.
F
u
n
g
.
,
W
.
L
a
m
.
“
S
to
c
k
P
re
d
ictio
n
:
In
teg
ra
ti
n
g
T
e
x
t
M
in
in
g
A
p
p
ro
a
c
h
u
si
n
g
Re
a
lt
ime
Ne
ws
.
”
Co
mp
u
t
a
ti
o
n
a
l
I
n
telli
g
e
n
c
e
fo
r F
i
n
a
n
c
i
a
l
E
n
g
in
e
e
rin
g
,
Ho
n
g
Ko
n
g
,
2
0
0
3
.
[
6
]
M
.
G
a
m
o
n
,
S
.
Ba
su
,
D.
Be
len
k
o
,
D.
F
ish
e
r,
M
.
Hu
rst.
,
A
.
C.
K¨o
n
ig
.
“
Bl
e
ws:
Us
in
g
Bl
o
g
s
to
Pr
o
v
id
e
Co
n
tex
t
fo
r
Ne
ws Arti
c
les
”
.
2
n
d
A
AA
I
Co
n
f
e
re
n
c
e
o
n
W
e
b
lo
g
s a
n
d
S
o
c
ial
M
e
d
ia,
S
e
a
tt
le,
W
a
sh
in
g
to
n
,
USA
,
2
0
0
8
.
[
7
]
R.
Jo
w
e
ll
.
,
B.
He
d
g
e
s
.
,
P
.
L
y
n
n
.
,
G
.
F
a
rra
n
t.
,
A
.
He
a
th
.
“
W
h
o
m
is
led
w
h
o
m
?
T
h
e
p
o
ll
s
a
n
d
t
h
e
v
o
t
e
rs
in
t
h
e
1
9
9
2
b
rit
ish
e
lec
ti
o
n
”
.
S
t
C
h
a
rle
s,
Il
li
n
o
is,
1
9
9
3
a
.
[
8
]
D.
Kin
g
m
a
.
,
J.
Ba
.
“
A
d
a
m
:
A
m
e
t
h
o
d
f
o
r
sto
c
h
a
stic
o
p
ti
m
iza
ti
o
n
”
.
a
rXiv,
1
4
1
2
.
6
9
8
0
,
2
0
1
4
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
S
a
n
th
i
P
riy
a
P
.
S
h
e
re
c
e
iv
e
d
a
B.
T
e
c
h
(Co
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
)
d
e
g
re
e
f
ro
m
Ja
w
a
h
a
rlal
Ne
h
ru
T
e
c
h
n
o
lo
g
ica
l
Un
iv
e
rsity
,
H
y
d
e
r
a
b
a
d
(JN
T
UH
),
In
d
ia
in
2
0
0
3
a
n
d
M
.
T
e
c
h
(c
o
m
p
u
ter
sc
ien
c
e
)
d
e
g
re
e
f
ro
m
JN
T
U
H
in
th
e
y
e
a
r
2
0
0
9
.
S
h
e
is
d
o
i
n
g
a
p
a
rt
-
ti
m
e
re
se
a
rc
h
in
Un
iv
e
rsit
y
Co
ll
e
g
e
o
f
En
g
in
e
e
r
in
g
a
n
d
T
e
c
h
n
o
lo
g
y
,
A
c
h
a
r
y
a
Na
g
a
rju
n
a
Un
iv
e
rsity
,
G
u
n
tu
r,
A
.
P
,
In
d
ia.
S
h
e
is
w
o
rk
in
g
a
s
a
n
A
ss
is
tan
t
P
r
o
f
e
ss
o
r
in
th
e
De
p
a
rtm
e
n
t
o
f
CS
E
in
S
ri
S
a
iram
Co
ll
e
g
e
o
f
En
g
in
e
e
rin
g
,
Be
n
g
u
l
u
ru
,
In
d
ia.
V
e
n
k
a
tes
wa
ra
Ra
o
T
.
H
e
re
c
e
i
v
e
d
a
B
a
c
h
e
lo
r
o
f
En
g
in
e
e
rin
g
d
e
g
re
e
in
El
e
c
tro
n
ics
a
n
d
Co
m
p
u
ter
En
g
in
e
e
rin
g
a
n
d
a
M
a
ste
rs
o
f
En
g
in
e
e
rin
g
in
Co
m
p
u
ter
S
c
ien
c
e
,
a
n
d
a
P
h
.
D
i
n
c
o
m
p
u
ter
sc
ien
c
e
a
n
d
e
n
g
in
e
e
rin
g
f
ro
m
W
a
y
n
e
S
tate
Un
iv
e
rsit
y
,
De
tro
it
,
USA
.
He
is
c
u
rre
n
tl
y
w
o
rk
in
g
a
s
P
ro
f
e
ss
o
r
in
KL
Un
i
v
e
rsit
y
,
V
a
d
d
e
sw
a
ra
m
,
G
u
n
tu
r
D
t,
In
d
ia.
He
h
a
s
m
o
re
th
a
n
3
2
y
e
a
rs o
f
e
x
p
e
rien
c
e
a
n
d
h
a
s p
u
b
li
sh
e
d
m
a
n
y
p
a
p
e
rs i
n
n
a
ti
o
n
a
l
a
n
d
in
tern
a
ti
o
n
a
l
c
o
n
f
e
re
n
c
e
s.
His
a
re
a
s o
f
in
tere
st are
m
u
lt
ico
re
a
n
d
p
a
ra
ll
e
l
p
r
o
g
ra
m
m
in
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.