I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
25
,
p
p
.
949
~
957
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
1
.
pp
9
4
9
-
9
5
7
949
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Enha
ncing
senti
ment ana
ly
sis
t
hr
o
ug
h deep
la
y
er i
ntegra
tion
with
lo
ng
sho
rt
-
te
rm memo
ry
n
e
tw
o
rks
P
a
rul D
ub
ey
1
,
P
us
hk
a
r
Dub
ey
2
,
H
it
esh
G
eha
ni
3
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
S
y
mb
i
o
s
i
s I
n
st
i
t
u
t
e
o
f
Te
c
h
n
o
l
o
g
y
,
N
a
g
p
u
r
C
a
mp
u
s,
S
y
m
b
i
o
si
s
I
n
t
e
r
n
a
t
i
o
n
a
l
(
D
e
e
me
d
t
o
b
e
U
n
i
v
e
r
si
t
y
)
,
P
u
n
e
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
M
a
n
a
g
e
me
n
t
,
P
a
n
d
i
t
S
u
n
d
a
r
l
a
l
S
h
a
r
ma
(
O
p
e
n
)
U
n
i
v
e
r
si
t
y
C
h
h
a
t
t
i
sg
a
r
h
,
B
i
l
a
s
p
u
r
,
I
n
d
i
a
3
S
c
h
o
o
l
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
S
h
r
i
R
a
m
d
e
o
b
a
b
a
C
o
l
l
e
g
e
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
M
a
n
a
g
e
m
e
n
t
,
R
a
m
d
e
o
b
a
b
a
U
n
i
v
e
r
s
i
t
y
,
N
a
g
p
u
r
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 1
2
,
2
0
2
4
R
ev
is
ed
Au
g
2
7
,
2
0
2
4
Acc
ep
ted
Sep
3
,
2
0
2
4
T
h
i
s
i
n
v
o
l
v
e
s
s
t
u
d
y
i
n
g
o
n
e
o
f
t
h
e
m
o
s
t
i
m
p
o
r
t
a
n
t
p
a
r
t
s
o
f
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
s
s
i
n
g
(
N
L
P
)
:
s
e
n
t
i
m
e
n
t
,
o
r
w
h
e
t
h
e
r
a
t
h
i
n
g
t
h
a
t
m
a
k
e
s
a
se
n
t
e
n
c
e
i
s
n
e
u
t
r
a
l
,
p
o
s
i
t
i
v
e
,
o
r
n
e
g
a
t
i
v
e
.
T
h
i
s
p
a
p
e
r
p
r
e
s
e
n
ts
a
n
e
n
h
a
n
c
e
d
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
(
L
S
T
M
)
n
e
t
w
o
r
k
f
o
r
t
h
e
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
i
s
ta
s
k
u
s
i
n
g
a
n
a
d
d
i
t
i
o
n
a
l
d
e
e
p
l
a
y
e
r
t
o
c
a
p
t
u
r
e
s
u
b
l
e
v
e
l
p
a
t
t
e
r
n
s
f
r
o
m
t
h
e
w
o
r
d
i
n
p
u
t
.
S
o
,
t
h
e
p
r
o
c
e
s
s
t
h
a
t
w
e
f
o
l
l
o
w
e
d
i
n
o
u
r
a
p
p
r
o
a
c
h
i
s
t
h
a
t
w
e
c
l
e
a
n
e
d
t
h
e
d
a
t
a
,
p
r
e
p
r
o
c
e
s
se
d
i
t
,
b
u
i
l
t
t
h
e
m
o
d
e
l
,
t
r
a
i
n
e
d
t
h
e
m
o
d
e
l
,
a
n
d
f
i
n
a
l
l
y
t
e
s
t
e
d
i
t
.
T
h
e
n
o
v
e
l
t
y
h
e
r
e
l
i
e
s
i
n
t
h
e
a
d
d
i
t
i
o
n
a
l
l
a
y
e
r
i
n
t
h
e
a
r
c
h
i
t
e
c
t
u
r
e
o
f
L
S
T
M
m
o
d
e
l
,
w
h
i
c
h
imp
r
o
v
e
s
t
h
e
m
o
d
e
l
p
e
r
f
o
r
m
a
n
c
e
.
W
e
a
d
d
e
d
a
d
e
e
p
l
a
y
e
r
w
i
t
h
t
h
e
i
n
t
e
n
t
i
o
n
o
f
i
m
p
r
o
v
i
n
g
a
c
c
u
ra
c
y
a
n
d
g
e
n
e
r
a
l
i
z
i
n
g
t
h
e
m
o
d
e
l
.
T
h
e
r
e
s
u
l
t
s
o
f
t
h
e
e
x
p
e
r
i
m
e
n
t
a
re
a
n
a
l
y
z
e
d
u
s
i
n
g
r
e
c
a
l
l
,
F
1
-
s
c
o
r
e
,
a
n
d
a
c
c
u
ra
c
y
,
w
h
i
c
h
i
n
t
u
r
n
s
h
o
w
t
h
a
t
t
h
e
d
e
e
p
-
l
a
y
e
re
d
L
S
T
M
m
o
d
e
l
g
i
v
e
s
u
s
a
b
e
t
te
r
p
r
e
d
i
c
t
i
o
n
.
T
h
e
L
S
TM
m
o
d
e
l
o
u
t
p
e
r
f
o
r
m
e
d
t
h
e
b
a
s
e
l
i
n
e
i
n
t
e
r
m
s
o
f
a
c
c
u
r
a
c
y
,
r
e
c
a
l
l
,
a
n
d
f
1
-
s
c
o
r
e
.
T
h
e
d
e
e
p
l
a
y
e
r
'
s
f
o
r
e
c
a
s
t
a
c
c
u
ra
c
y
i
n
c
re
a
se
d
d
r
a
m
a
t
ic
a
l
l
y
o
n
c
e
i
t
wa
s
tr
a
i
n
e
d
to
c
a
p
t
u
r
e
i
n
t
r
i
c
a
t
e
s
e
q
u
e
n
c
e
s
.
H
o
we
v
e
r
,
t
h
e
i
m
p
r
o
v
e
d
m
o
d
e
l
o
v
e
r
f
i
t
t
e
d
,
n
e
c
e
ss
i
t
a
t
i
n
g
a
d
d
i
t
i
o
n
a
l
r
e
g
u
l
a
r
i
z
a
t
i
o
n
a
n
d
h
y
p
e
r
p
a
r
a
m
e
te
r
a
d
j
u
s
t
m
e
n
t
.
I
n
t
h
i
s
p
a
p
e
r
,
w
e
h
a
v
e
d
i
s
c
u
s
se
d
t
h
e
a
d
v
a
n
t
a
g
e
s
a
n
d
d
i
s
a
d
v
a
n
t
a
g
e
s
o
f
u
s
i
n
g
d
e
e
p
l
a
y
e
r
s
i
n
L
S
T
M
n
e
t
w
o
r
k
s
a
n
d
t
h
e
i
r
a
p
p
l
i
c
a
t
i
o
n
t
o
d
e
v
e
l
o
p
i
n
g
m
o
d
e
l
s
f
o
r
d
e
e
p
l
e
a
r
n
i
n
g
w
i
t
h
b
e
t
t
e
r
-
p
e
r
f
o
rm
i
n
g
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
i
s
.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
L
o
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
Natu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
Sen
tim
en
t a
n
aly
s
is
T
ex
t c
lass
if
icatio
n
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
:
Par
u
l D
u
b
ey
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
Sy
m
b
io
s
is
I
n
s
titu
te
o
f
T
ec
h
n
o
l
o
g
y
,
Nag
p
u
r
C
am
p
u
s
,
Sy
m
b
io
s
is
I
n
ter
n
atio
n
al
(
Dee
m
ed
to
b
e
Un
iv
er
s
ity
)
Pu
n
e,
I
n
d
ia
E
m
ail: d
u
b
ey
p
ar
u
l2
9
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Sen
tim
en
t
an
aly
s
is
(
also
k
n
o
wn
as
o
p
in
io
n
m
i
n
in
g
o
r
em
o
tio
n
ar
tific
ial
in
tellig
en
ce
(
AI
)
)
i
s
a
p
ar
t
o
f
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P
)
th
at
f
o
cu
s
es
o
n
an
al
y
zin
g
an
d
d
ete
r
m
in
in
g
th
e
s
en
ti
m
en
ts
o
f
tex
t
d
ata.
B
ec
au
s
e
o
f
th
e
in
cr
ea
s
e
in
u
s
er
-
g
en
er
ated
in
f
o
r
m
atio
n
in
th
e
d
ig
ital
s
p
ac
e,
an
d
s
o
cial
n
etwo
r
k
in
g
s
ites
in
p
ar
ticu
lar
,
s
en
tim
en
t
an
aly
s
is
h
as
b
ee
n
a
v
alu
ab
le
to
o
l
f
o
r
b
o
th
r
esear
ch
er
s
a
n
d
b
u
s
in
ess
es.
I
n
th
is
in
v
esti
g
atio
n
,
L
STM
n
etwo
r
k
s
an
d
o
t
h
er
r
ec
u
r
r
e
n
t
n
e
u
r
al
n
etwo
r
k
s
(
R
NNs)
will
b
e
an
al
y
ze
d
in
th
e
ca
s
e
o
f
s
en
tim
en
t
an
aly
s
is
.
Sen
tim
en
t
an
aly
s
is
is
ar
g
u
ab
ly
o
n
e
o
f
th
e
m
o
s
t
ex
citin
g
ar
ea
s
o
f
r
esear
c
h
in
NL
P
in
th
at
it
allo
ws
u
s
to
au
to
m
atica
lly
d
eter
m
in
e
wh
eth
er
th
e
te
x
t
is
n
eu
tr
al,
p
o
s
itiv
e,
o
r
n
eg
ativ
e.
W
h
ile
tr
ad
itio
n
al
m
eth
o
d
s
u
s
ed
d
if
f
e
r
en
t
m
ac
h
i
n
e
lear
n
in
g
alg
o
r
ith
m
s
,
d
ee
p
l
ea
r
n
in
g
h
as
im
p
r
o
v
ed
s
en
tim
e
n
t
an
aly
s
is
m
o
d
els
in
s
u
ch
a
way
th
at
th
ey
ar
e
n
o
w
m
u
c
h
m
o
r
e
ac
cu
r
ate
an
d
r
o
b
u
s
t.
T
h
e
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
25
:
9
4
9
-
957
950
n
etwo
r
k
,
with
its
ab
ilit
y
to
h
a
n
d
le
lo
n
g
-
ter
m
d
ep
e
n
d
en
cies in
tex
tu
al
d
ata,
h
as m
ad
e
s
ig
n
if
i
ca
n
t im
p
r
o
v
em
en
ts
in
lan
g
u
ag
e
m
o
d
elin
g
a
n
d
h
a
s
b
ec
o
m
e
o
n
e
o
f
th
e
b
etter
d
e
ep
lear
n
in
g
n
etwo
r
k
ar
ch
itect
u
r
es
f
o
r
p
r
o
ce
s
s
in
g
s
eq
u
en
tial d
ata
in
s
tead
o
f
t
r
ad
i
tio
n
al
r
ec
u
r
r
e
n
t n
eu
r
al
n
etwo
r
k
s
(
R
NNs).
On
e
p
o
ten
tial w
ay
to
in
cr
ea
s
e
p
er
f
o
r
m
an
ce
f
u
r
th
er
with
L
ST
M
n
etwo
r
k
s
is
to
h
av
e
a
d
ee
p
e
r
lay
er
.
I
n
m
an
y
ca
s
es,
d
ee
p
er
L
STM
ar
ch
itectu
r
es
ar
e
u
s
ed
to
im
p
r
o
v
e
th
e
m
o
d
el'
s
ab
ilit
y
to
g
en
er
a
lize
to
u
n
s
ee
n
d
ata
an
d
lear
n
m
o
r
e
c
o
m
p
lex
d
im
e
n
s
io
n
al
p
atter
n
s
an
d
in
ter
ac
tiv
ities
in
th
e
d
ata.
T
h
e
ad
d
itio
n
al
lay
er
co
u
ld
h
elp
in
cr
ea
s
e
k
ey
p
er
f
o
r
m
an
ce
m
e
asu
r
es
s
u
ch
as
F1
-
s
co
r
e,
r
ec
all,
an
d
ac
cu
r
ac
y
,
as
it
will
m
a
k
e
th
e
m
o
d
el
b
etter
eq
u
ip
p
e
d
to
u
n
d
er
s
tan
d
c
o
m
p
l
ex
co
n
n
ec
tio
n
s
b
etwe
en
co
n
ten
t
in
th
e
tex
t.
W
e
ar
e
g
o
in
g
to
m
er
g
e
L
STM
an
d
d
ee
p
lay
er
s
to
in
tr
o
d
u
ce
m
o
r
e
r
eliab
le
p
r
ed
ictio
n
s
ab
o
u
t
th
e
s
en
tim
en
t
o
f
o
u
r
m
o
d
el.
T
a
b
le
1
s
h
o
ws
th
e
li
s
t
o
f
p
r
ev
io
u
s
wo
r
k
th
at
h
as b
ee
n
d
o
n
e.
T
ab
le
1
.
C
o
m
p
a
r
is
o
n
o
f
l
iter
atu
r
e
r
ev
iewe
d
S
t
u
d
y
D
a
t
a
s
e
t
S
o
u
r
c
e
Te
c
h
n
i
q
u
e
s
K
e
y
F
i
n
d
i
n
g
s
C
o
m
p
a
r
i
so
n
w
i
t
h
p
r
o
p
o
se
d
m
o
d
e
l
S
r
i
n
i
v
a
s
e
t
a
l
.
[
1
]
Tw
i
t
t
e
r
(
K
a
g
g
l
e
)
C
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
(
C
N
N
)
,
LST
M
,
s
i
m
p
l
e
n
e
u
r
a
l
n
e
t
w
o
r
k
LSTM
h
a
d
t
h
e
h
i
g
h
e
s
t
a
c
c
u
r
a
c
y
w
i
t
h
8
7
%
a
c
c
u
r
a
c
y
Tr
a
i
n
i
n
g
a
c
c
u
r
a
c
y
b
e
t
t
e
r
t
h
a
n
p
r
e
v
i
o
u
s
mo
d
e
l
s
-
9
8
.
3
4
%
M
u
h
a
mm
a
d
e
t
a
l
.
[
2
]
I
n
d
o
n
e
si
a
n
h
o
t
e
l
r
e
v
i
e
w
s
W
o
r
d
2
V
e
c
,
LS
TM
H
i
g
h
e
s
t
a
c
c
u
r
a
c
y
w
i
t
h
W
o
r
d
2
V
e
c
+
LST
M
-
8
5
.
9
6
%
M
a
h
a
d
e
v
a
sw
a
my
a
n
d
S
w
a
t
h
i
[
3
]
A
maz
o
n
p
r
o
d
u
c
t
r
e
v
i
e
w
s
B
i
d
i
r
e
c
t
i
o
n
a
l
LSTM
B
e
t
t
e
r
p
r
e
d
i
c
t
i
o
n
s
w
i
t
h
b
i
d
i
r
e
c
t
i
o
n
a
l
LST
M
G
a
n
d
h
i
e
t
a
l
.
[
4
]
Tw
i
t
t
e
r
(
I
M
D
B
)
C
N
N
,
LST
M
I
mp
r
o
v
e
d
d
e
t
e
c
t
i
o
n
o
f
t
w
e
e
t
sen
t
i
m
e
n
t
a
n
d
r
e
v
i
e
w
s
-
8
7
.
7
4
%
(
t
w
e
e
t
s)
,
8
8
.
0
2
%
(
r
e
v
i
e
w
s)
B
e
h
e
r
a
e
t
a
l
.
[
5
]
S
o
c
i
a
l
me
d
i
a
r
e
v
i
e
w
s
C
o
n
v
o
l
u
t
i
o
n
a
l
LST
M
(
C
o
-
LSTM
)
B
e
t
t
e
r
o
u
t
c
o
m
e
s
i
n
so
c
i
a
l
b
i
g
d
a
t
a
s
e
n
t
i
me
n
t
Ji
n
e
t
a
l
.
[
6
]
S
t
o
c
k
mark
e
t
d
a
t
a
S
e
n
t
i
me
n
t
a
n
a
l
y
si
s
,
LST
M
,
A
t
t
e
n
t
i
o
n
mec
h
a
n
i
sm
En
h
a
n
c
e
d
st
o
c
k
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
a
c
c
u
r
a
c
y
An
ar
ch
itectu
r
al
d
e
p
ictio
n
is
u
s
ed
to
f
u
r
th
e
r
d
is
cu
s
s
th
e
im
p
lem
en
ted
p
r
o
ce
d
u
r
e
.
T
h
e
f
o
llo
win
g
s
ec
tio
n
ex
p
lain
s
d
ataset
d
escr
ip
tio
n
an
d
d
ata
g
ath
e
r
in
g
.
A
d
etailed
ex
p
lan
atio
n
o
f
th
e
m
o
d
el'
s
co
n
s
tr
u
ctio
n
,
tr
ain
in
g
,
an
d
h
y
p
er
p
a
r
am
eter
t
wea
k
in
g
is
p
r
o
v
id
ed
af
ter
th
e
f
in
d
in
g
s
.
Ultim
ately
,
we
wr
ap
u
p
th
e
p
a
p
er
with
im
p
r
o
v
e
d
p
er
f
o
r
m
a
n
ce
wh
en
c
o
m
p
ar
ed
to
ea
r
lier
m
o
d
els.
2.
M
E
T
H
O
D
Ou
r
s
en
tim
en
t
an
aly
s
is
m
eth
o
d
in
co
r
p
o
r
ates
a
n
u
m
b
er
o
f
c
r
itical
s
tep
s
,
in
clu
d
in
g
d
ata
p
r
ed
ictio
n
an
d
m
o
d
el
ev
alu
atio
n
.
I
t
is
an
i
m
p
r
o
v
ed
v
er
s
io
n
o
f
L
STM
n
etw
o
r
k
s
with
an
e
x
tr
a
d
ee
p
lay
er
.
As
s
u
ch
,
o
u
r
i
n
itial
task
will
b
e
to
am
ass
a
g
ig
an
t
ic
d
atab
ase
o
f
r
ev
iews
m
ad
e
b
y
u
s
er
s
an
d
an
n
o
tated
with
t
h
eir
s
en
tim
en
t.
T
h
e
tex
t
d
ata
in
th
is
d
ataset
h
as
b
ee
n
p
r
e
-
p
r
o
ce
s
s
ed
b
y
clea
n
in
g
t
h
e
in
itial
tex
t,
u
s
in
g
s
tem
m
in
g
o
r
lem
m
atiza
tio
n
,
an
d
r
em
o
v
i
n
g
s
to
p
wo
r
d
s
;
h
en
ce
,
it
is
in
a
to
k
en
ized
an
d
s
ta
n
d
ar
d
ized
f
o
r
m
[
7
]
–
[
9
]
.
Af
ter
th
is
p
h
ase,
th
e
tex
t
is
r
ea
d
y
f
o
r
an
al
y
s
is
[
1
0
]
.
T
h
e
n
ex
t
s
tep
is
to
c
o
n
v
e
r
t
th
is
p
r
e
-
p
r
o
ce
s
s
ed
tex
t
in
to
n
u
m
e
r
ical
r
ep
r
esen
tatio
n
s
u
s
in
g
tex
t
v
ec
to
r
izatio
n
m
eth
o
d
s
,
e.
g
.
,
w
o
r
d
em
b
ed
d
i
n
g
s
(
e
.
g
.
,
Glo
Ve)
.
W
e
h
ad
t
o
d
o
th
i
s
to
en
s
u
r
e
th
at
o
u
r
m
o
d
el
wo
r
k
s
well
with
tex
t
d
ata.
W
e
ca
n
s
ee
th
at
s
o
m
e
o
f
th
e
s
tr
u
ctu
r
e,
wh
en
we
d
o
o
u
r
o
wn
m
o
d
el,
is
a
m
u
lti
-
lay
er
a
r
ch
itectu
r
e.
T
h
e
em
b
ed
d
in
g
lay
e
r
is
th
e
f
ir
s
t
lay
er
,
wh
ich
co
n
v
er
ts
th
e
in
p
u
t
tex
t
in
to
w
o
r
d
em
b
ed
d
in
g
s
.
T
h
e
n
e
x
t
two
lay
er
s
,
n
am
ed
“
L
STM
,
”
wh
ich
h
er
e
s
tan
d
s
f
o
r
l
o
n
g
s
h
o
r
t
-
te
r
m
m
em
o
r
y
[
1
1
]
,
[
1
2
]
,
a
r
e
s
p
ec
ialized
to
d
etec
t
s
eq
u
en
tial
d
ep
en
d
en
cies
p
r
esen
t
i
n
tex
t
d
ata
[
1
3
]
,
[
1
4
]
.
Her
e,
“
d
ee
p
,
”
w
h
ich
s
its
ab
o
v
e
d
en
s
e,
r
ec
o
g
n
izes
m
o
r
e
co
m
p
lex
p
atter
n
s
an
d
in
ter
ac
t
io
n
s
.
A
b
in
ar
y
class
if
icatio
n
is
d
o
n
e
in
th
e
o
u
tp
u
t
lay
er
v
ia
a
s
in
g
le
n
eu
r
o
n
with
s
ig
m
o
id
ac
tiv
atio
n
,
a
n
d
ea
c
h
l
ay
er
is
a
d
en
s
e
lay
er
.
T
h
e
m
o
d
el
is
tr
ain
ed
o
n
th
e
tr
ain
in
g
d
ataset.
Fo
r
th
is
,
to
k
ee
p
f
r
o
m
o
v
er
f
itti
n
g
,
we
tu
n
e
th
e
weig
h
ts
in
th
is
s
tep
an
d
m
o
n
ito
r
th
e
lo
s
s
in
tr
ain
in
g
an
d
v
alid
atio
n
,
a
ll
th
is
with
th
e
ai
d
o
f
o
u
r
f
r
ien
d
Ad
am
o
p
tim
izer
.
I
n
th
e
en
d
,
we
will
ev
alu
ate
th
e
ef
f
icien
cy
o
f
th
e
m
o
d
el
b
y
d
er
iv
in
g
t
h
e
F1
-
s
co
r
e,
ac
c
u
r
ac
y
,
p
r
ec
is
io
n
,
a
n
d
r
ec
all.
T
h
e
m
o
d
el
is
th
en
v
alid
ated
o
n
a
s
ep
ar
ate
v
al
id
atio
n
d
ataset
to
en
s
u
r
e
it
is
f
u
n
ctio
n
al
an
d
g
en
er
aliza
b
le.
T
h
e
a
r
ch
itectu
r
e
o
f
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
ca
n
b
e
s
ee
n
in
Fig
u
r
e
1
.
T
o
th
e
b
est
o
f
o
u
r
k
n
o
wled
g
e
,
o
u
r
m
o
d
el
is
th
e
f
ir
s
t
to
in
co
r
p
o
r
ate
a
d
ee
p
lay
er
at
th
e
L
STM
lev
el,
an
d
it
o
f
f
e
r
s
an
e
f
f
icien
t
s
o
lu
tio
n
to
s
en
tim
en
t
an
aly
s
is
task
s
wi
th
th
e
ab
ilit
y
to
g
en
er
alize
a
n
d
p
r
ed
ict
m
o
r
e
ac
c
u
r
ately
.
Fr
o
m
d
ata
p
r
ep
a
r
atio
n
th
r
o
u
g
h
m
o
d
el
v
alid
atio
n
,
th
is
ex
ten
s
iv
e
ap
p
r
o
ac
h
en
s
u
r
es
ea
ch
s
tep
is
tak
en
with
th
e
u
tm
o
s
t
r
ig
o
r
to
o
b
tain
th
e
b
est r
esu
lts
.
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
-
8
7
0
8
E
n
h
a
n
cin
g
s
en
timen
t a
n
a
lysi
s
th
r
o
u
g
h
d
ee
p
la
ye
r
in
teg
r
a
ti
o
n
w
ith
…
(
P
a
r
u
l D
u
b
ey
)
951
Fig
u
r
e
1
.
Ar
c
h
itectu
r
e
o
f
p
r
o
p
o
s
ed
m
o
d
el
As
u
s
er
r
ev
iews
with
ass
o
ci
ated
s
en
tim
en
t
lab
els
m
ak
e
u
p
th
e
d
ataset
u
s
ed
f
o
r
th
is
s
en
tim
en
t
an
aly
s
is
p
r
o
ject,
wh
ich
is
d
er
i
v
ed
f
r
o
m
Kag
g
le
.
T
h
e
d
ataset
is
p
r
esen
ted
h
er
e
in
g
r
ea
t d
etai
l:
a.
Stru
ctu
r
e
o
f
t
h
e
d
ataset: M
o
s
t o
f
ten
,
th
e
d
ataset
r
eq
u
ir
es th
e
f
o
llo
win
g
co
l
u
m
n
s
:
−
R
ev
iew
tex
t:
T
h
e
r
e
v
iew
tex
t
its
elf
is
in
th
is
co
lu
m
n
.
T
h
i
s
co
lu
m
n
c
o
n
tain
s
s
tr
in
g
s
,
i.e
.
,
u
s
er
te
x
t,
ex
p
r
ess
in
g
an
o
p
in
io
n
o
r
c
o
m
m
en
t a
b
o
u
t p
r
o
d
u
cts,
s
er
v
ices,
o
r
ex
p
er
ien
ce
s
.
−
Sen
tim
en
t
L
ab
el
-
E
v
er
y
r
e
v
iew
h
as
its
s
en
tim
en
t
lab
els
in
th
is
co
lu
m
n
.
B
asically
,
th
e
m
o
o
d
is
ju
s
t
a
y
es/n
o
with
tag
s
.
T
h
e
lab
els
c
an
also
b
e
m
ath
em
atica
l.
I
n
p
n
eu
m
atica
l
ter
m
s
,
1
is
to
b
e
r
ewa
r
d
ed
as
p
o
s
itiv
e
an
d
0
is
to
b
e
n
e
g
ativ
e.
b.
Sam
p
le
d
ata:
Her
e’
s
a
s
m
all
s
am
p
le
to
illu
s
tr
ate
th
e
s
tr
u
ctu
r
e
o
f
th
e
d
ataset
in
T
ab
le
1
.
T
ab
le
1
.
Stru
ctu
r
e
o
f
d
ataset
R
e
v
i
e
w
t
e
x
t
S
e
n
t
i
me
n
t
l
a
b
e
l
“
I
l
o
v
e
t
h
i
s
p
r
o
d
u
c
t
!
I
t
w
o
r
k
s
g
r
e
a
t
a
n
d
i
s
v
e
r
y
e
a
sy
t
o
u
se
.
”
1
“
Te
r
r
i
b
l
e
e
x
p
e
r
i
e
n
c
e
.
T
h
e
i
t
e
m
a
r
r
i
v
e
d
b
r
o
k
e
n
a
n
d
t
h
e
c
u
s
t
o
mer
s
e
r
v
i
c
e
w
a
s u
n
h
e
l
p
f
u
l
.
”
0
“
Ju
s
t
o
k
a
y
,
n
o
t
h
i
n
g
sp
e
c
i
a
l
b
u
t
n
o
t
b
a
d
e
i
t
h
e
r
.
”
0
“
F
a
n
t
a
s
t
i
c
!
E
x
c
e
e
d
e
d
m
y
e
x
p
e
c
t
a
t
i
o
n
s
i
n
e
v
e
r
y
w
a
y
.
”
1
“
N
o
t
w
o
r
t
h
t
h
e
mo
n
e
y
.
P
o
o
r
q
u
a
l
i
t
y
a
n
d
b
a
d
p
e
r
f
o
r
ma
n
c
e
.
”
0
c.
Data
c
h
ar
ac
ter
is
tics
−
Div
er
s
ity
o
f
r
ev
iews:
b
etwe
e
n
th
e
r
ev
iews
o
f
a
b
r
o
a
d
r
a
n
g
e
o
f
item
s
an
d
s
er
v
ices,
th
er
e
is
a
d
iv
e
r
s
e
d
ataset
th
at
m
ay
allo
w
th
e
m
o
d
el
to
g
en
e
r
alize
ac
r
o
s
s
s
ettin
g
s
.
−
L
en
g
th
v
a
r
iab
ilit
y
:
th
er
e
is
a
lo
t
o
f
v
ar
ia
b
ilit
y
in
h
o
w
lo
n
g
t
h
e
r
ev
iews
ar
e,
f
r
o
m
s
h
o
r
t
co
m
m
en
ts
to
in
-
d
ep
th
an
al
y
s
is
.
T
h
is
u
n
ce
r
tain
t
y
is
p
ar
tly
ad
d
r
ess
ed
d
u
r
in
g
p
r
ep
r
o
ce
s
s
in
g
b
y
u
s
in
g
p
a
d
d
in
g
.
−
Sen
tim
en
t
d
is
tr
ib
u
tio
n
:
th
e
u
s
er
n
ee
d
s
to
v
er
if
y
th
e
s
en
tim
en
t
lab
el
d
is
tr
ib
u
tio
n
to
m
ak
e
th
e
d
ataset
b
alan
ce
d
.
I
f
th
e
r
e
ar
e
in
d
ee
d
n
o
b
alan
ce
d
d
ata,
o
v
er
-
s
am
p
lin
g
,
u
n
d
er
-
s
am
p
lin
g
,
o
r
b
al
an
cin
g
class
weig
h
ts
in
th
e
m
o
d
el
t
r
ain
in
g
will b
e
r
eq
u
ir
ed
.
d.
I
m
p
o
r
ta
n
ce
o
f
p
r
ep
r
o
ce
s
s
in
g
T
ex
t
clea
n
in
g
in
v
o
lv
es
r
em
o
v
in
g
r
ed
u
n
d
an
t
p
u
n
ctu
atio
n
,
HT
ML
tag
s
,
an
d
s
p
ec
ial
ch
a
r
ac
ter
s
to
r
etain
th
e
co
r
e
co
n
ten
t
[
1
5
]
,
[
1
6
]
.
T
o
k
en
izatio
n
s
p
lits
th
e
tex
t
in
to
in
d
i
v
id
u
al
wo
r
d
s
o
r
t
o
k
en
s
,
f
o
ll
o
wed
b
y
s
to
p
wo
r
d
s
r
em
o
v
al
to
elim
in
ate
co
m
m
o
n
wo
r
d
s
lik
e
“
an
d
”
an
d
“
th
e.
”
Stem
m
in
g
o
r
l
em
m
atiza
tio
n
th
en
s
im
p
lifie
s
wo
r
d
s
to
th
eir
r
o
o
t
f
o
r
m
s
,
wh
ile
p
ad
d
in
g
en
s
u
r
es
u
n
if
o
r
m
s
eq
u
en
ce
len
g
t
h
s
b
y
ad
d
in
g
ze
r
o
s
wh
er
e
n
ec
ess
ar
y
[
1
7
]
–
[
1
9
]
.
e.
Use in
m
o
d
el
tr
ain
in
g
T
h
e
p
r
e
-
p
r
o
ce
s
s
ed
d
ata
s
et
is
s
ep
ar
ated
in
to
a
tr
ain
in
g
s
et
an
d
a
test
in
g
s
et.
Fo
r
tr
ain
in
g
th
e
L
STM
m
o
d
el,
we
u
s
e
th
e
tr
ain
in
g
s
et
an
d
ev
alu
ate
it
o
n
th
e
test
in
g
s
et
[
2
0
]
.
T
h
ese
lab
els
ar
e
th
e
s
en
tim
en
t
lab
els,
wh
ich
ar
e
th
e
tar
g
et
v
ar
ia
b
les th
at
th
e
m
o
d
el
lea
r
n
s
to
u
s
e
to
ca
teg
o
r
ize
r
ev
iews in
to
s
en
tim
en
t c
lass
es.
3.
M
O
DE
L
DE
VE
L
O
P
M
E
N
T
T
h
e
ce
n
ter
p
iece
o
f
o
u
r
m
eth
o
d
is
d
esig
n
in
g
an
L
STM
-
b
a
s
ed
n
eu
r
al
n
etwo
r
k
m
o
d
el
f
o
r
s
en
tim
en
t
an
aly
s
is
.
T
o
tr
an
s
f
o
r
m
r
aw
tex
t
in
p
u
t
in
to
v
alu
ab
le
p
r
ed
ictio
n
s
as
d
ef
in
ed
in
th
e
task
s
,
th
is
m
o
d
el
ar
ch
itectu
r
e
d
ep
en
d
s
o
n
m
a
n
y
k
ey
p
a
r
ts
.
T
h
e
f
o
llo
win
g
s
ec
tio
n
f
u
r
t
h
er
ex
p
lain
s
th
e
ar
ch
itectu
r
e
o
f
th
e
m
eth
o
d
in
d
etail.
T
h
ey
g
iv
e
u
s
a
clea
r
p
ictu
r
e
o
f
th
e
m
o
d
el.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
25
:
9
4
9
-
957
952
3
.
1
.
E
m
bedd
ing
l
a
y
er
Firstl
y
,
th
e
em
b
ed
d
in
g
lay
er
o
f
th
e
m
o
d
el
is
its
in
p
u
t
lay
e
r
[
2
1
]
.
I
ts
jo
b
is
to
g
e
n
er
ate
d
en
s
e
wo
r
d
em
b
ed
d
in
g
s
f
o
r
t
h
e
tex
t
in
p
u
t.
T
h
is
cr
ea
tes
a
d
en
s
e
v
ec
to
r
o
f
a
f
ix
ed
s
ize
f
o
r
ea
ch
wo
r
d
i
n
t
h
e
in
p
u
t
s
eq
u
en
ce
,
allo
win
g
th
e
m
o
d
el
to
en
co
d
e
s
em
an
tic
in
f
o
r
m
atio
n
ab
o
u
t
th
e
wo
r
d
s
as
we
ll
as
th
eir
r
elatio
n
s
h
ip
s
to
o
n
e
an
o
th
er
.
T
h
u
s
,
in
s
tead
o
f
[
1
,
2
,
3
,
4
]
,
f
o
r
ex
am
p
le
,
we
g
et
[
0
.
1
,
0
.
2
]
,
[
-
0
.
2
5
,
.
.
.
]
,
[
0
.
0
5
,
0
.
1
,
.
.
.
]
,
[
0
.
3
,
0
.
4
,
.
.
.
]
in
s
tead
o
f
wo
r
d
in
d
ex
es.
Mo
r
eo
v
er
,
g
i
v
en
th
at
we
u
s
e
p
r
e
-
tr
ain
ed
em
b
e
d
d
in
g
s
lik
e
Glo
V
e,
we
ca
n
in
cr
ea
s
e
tex
t u
n
d
er
s
tan
d
in
g
in
th
e
m
o
d
el
b
y
lev
er
a
g
in
g
p
r
e
-
ex
is
tin
g
k
n
o
wled
g
e
a
b
o
u
t w
o
r
d
co
n
n
ec
t
io
n
s
.
3
.
2
.
L
ST
M
l
a
y
er
T
h
e
co
r
e
o
f
th
e
m
o
d
el
ar
e
th
e
L
STM
lay
er
s
,
wh
ich
ar
e
d
esig
n
ed
to
lear
n
th
e
tem
p
o
r
al
d
e
p
en
d
en
cies
in
th
e
tex
t
in
p
u
t.
L
STM
lay
er
s
wer
e
d
esig
n
e
d
s
p
ec
if
ically
to
h
an
d
le
th
e
v
an
is
h
in
g
g
r
ad
ie
n
t
is
s
u
e
th
at
im
p
ac
ts
co
n
v
en
tio
n
al
R
NNs
(
as
well
as
m
o
d
el
lo
n
g
-
r
an
g
e
d
ep
e
n
d
en
cies)
b
y
en
a
b
lin
g
th
e
n
et
wo
r
k
to
r
em
e
m
b
er
in
f
o
r
m
atio
n
f
o
r
lo
n
g
p
e
r
io
d
s
o
f
tim
e
a
n
d
to
m
o
d
er
ate
th
e
m
o
v
em
en
t
o
f
d
ata
th
r
o
u
g
h
th
e
n
etwo
r
k
[
2
2
]
.
T
h
e
r
e
ar
e
s
u
b
-
f
u
n
ctio
n
s
with
in
th
em
th
at
let
th
em
s
to
r
e
d
ata,
ca
lle
d
m
em
o
r
y
ce
lls
,
an
d
cir
cu
late
d
ata,
ca
lled
g
ates
[
2
3
]
.
B
y
s
tack
in
g
L
STM
la
y
er
s
,
it
is
p
o
s
s
ib
le
to
g
r
ea
tly
en
h
an
ce
th
e
m
o
d
el'
s
ca
p
ab
ilit
y
to
lear
n
in
tr
icate
p
atter
n
s
[
2
4
]
,
[
2
5
]
.
3
.
3
.
Dee
p
l
a
y
er
MA
DE
-
L
STM
with
d
ee
p
lay
er
:
we
ad
d
a
th
ick
d
ee
p
la
y
er
,
wh
ich
en
ab
les
lear
n
in
g
m
o
r
e
co
m
p
lex
p
atter
n
s
an
d
in
ter
ac
tio
n
s
.
I
ts
p
r
im
ar
y
f
u
n
ctio
n
is
to
h
elp
th
e
m
o
d
el
lear
n
d
if
f
ic
u
lt
p
atter
n
s
an
d
co
r
r
elatio
n
s
th
at
wer
e
u
n
n
o
ticed
b
y
t
h
e
L
S
T
M
lay
er
s
o
p
er
atin
g
alo
n
e.
W
e
ca
n
h
av
e
th
e
in
tu
itio
n
th
at
w
ith
th
is
ex
tr
a
lay
er
,
th
e
m
o
d
el
is
m
o
r
e
a
b
le
to
lear
n
th
e
s
tr
u
ctu
r
e
o
f
th
e
in
p
u
t
d
at
a
an
d
to
g
e
n
er
alize
.
T
h
is
p
r
o
v
i
d
es
th
e
m
o
d
el
with
an
u
n
s
u
r
p
ass
ed
ab
ilit
y
to
s
h
ap
e
th
e
in
co
m
in
g
d
ata
in
m
o
r
e
c
o
m
p
licated
way
s
,
allo
win
g
th
e
m
o
d
el
to
m
in
g
le
an
d
in
ter
p
r
et
in
f
o
r
m
atio
n
in
a
d
ee
p
er
s
en
s
e.
T
h
e
DHL
m
o
d
el
is
a
n
o
n
-
lin
ea
r
h
id
d
en
lay
e
r
m
o
d
el
b
ec
au
s
e
o
f
th
e
R
ec
tifie
d
L
in
ea
r
U
n
it
ac
tiv
atio
n
f
u
n
ctio
n
,
wh
ich
h
elp
s
th
e
DHL
m
o
d
el
ac
co
m
m
o
d
at
e
th
e
n
o
n
-
lin
ea
r
it
y
.
T
h
e
m
o
d
el
is
ab
le
t
o
lear
n
n
o
n
-
lin
ea
r
r
elatio
n
s
h
ip
s
am
o
n
g
s
t
th
e
d
ata,
w
h
ich
h
el
p
s
in
th
e
ac
cu
r
ate
r
ep
r
esen
tatio
n
o
f
h
ig
h
-
d
eg
r
ee
r
elatio
n
s
h
ip
s
ex
is
tin
g
in
t
h
e
d
ata.
No
w
we
ca
n
d
o
th
is
u
s
in
g
th
e
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U)
ac
tiv
atio
n
f
u
n
ctio
n
b
ec
au
s
e
th
is
en
ab
les
o
u
r
m
o
d
el
to
tr
ain
with
o
u
t
r
u
n
n
in
g
in
to
p
r
o
b
lem
s
s
u
ch
as
th
e
v
an
is
h
in
g
g
r
a
d
ien
t
p
r
o
b
lem
.
T
h
is
ad
d
s
a
lev
el
o
f
d
e
p
th
,
an
d
n
o
w
th
e
m
o
d
el
b
ased
o
n
L
STM
lear
n
s
f
r
o
m
d
ata
in
a
b
etter
way
,
w
h
ich
is
wh
y
it
p
er
f
o
r
m
s
b
ette
r
in
task
s
lik
e
s
en
tim
en
t
an
al
y
s
is
.
B
es
id
es
L
STM
lay
er
s
,
th
e
ad
d
itio
n
al
lay
er
s
u
p
p
lem
en
ts
th
e
L
STM
lay
e
r
s
b
y
h
elp
in
g
to
d
is
co
v
er
th
e
co
r
r
elatio
n
s
an
d
p
atter
n
s
th
at
ar
e
g
en
er
ally
lef
t
u
n
n
o
tic
ed
an
d
,
as
a
r
esu
lt,
a
r
e
ca
p
ab
l
e
o
f
g
en
e
r
atin
g
a
m
o
r
e
s
tab
le
as
well
as
ac
cu
r
ate
tab
le
-
b
ased
r
ep
r
esen
tatio
n
.
3
.
4
.
Dense
l
a
y
er
s
T
h
i
s
is
f
o
l
l
o
w
e
d
b
y
d
e
n
s
e
l
a
y
e
r
s
a
p
p
l
i
e
d
t
o
t
h
e
o
u
t
p
u
t
a
f
t
e
r
t
h
e
L
S
T
M
l
a
y
e
r
s
.
O
n
e
o
f
t
h
e
m
a
in
t
a
s
k
s
f
o
r
t
h
e
s
e
f
u
ll
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
s
i
s
t
o
t
a
k
e
t
h
e
o
u
t
p
u
t
o
f
t
h
e
L
S
T
M
a
n
d
p
u
t
i
t
i
n
t
o
a
f
o
r
m
a
t
t
h
at
m
a
t
c
h
e
s
u
p
w
it
h
t
h
e
f
i
n
a
l
c
la
s
s
i
f
i
ca
t
i
o
n
.
S
i
n
c
e
a
ll
n
e
u
r
o
n
s
i
n
a
g
i
v
e
n
l
a
y
e
r
a
r
e
c
o
n
n
e
c
t
e
d
t
o
al
l
t
h
e
n
e
u
r
o
n
s
i
n
th
e
p
r
e
v
i
o
u
s
la
y
e
r
,
d
e
n
s
e
l
a
y
e
r
s
a
r
e
a
b
l
e
t
o
p
r
o
c
ess
v
e
r
y
c
o
m
p
l
e
x
d
a
t
a
.
R
e
L
U
a
ct
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
s
a
r
e
f
r
e
q
u
e
n
t
ly
u
s
e
d
t
o
i
n
t
r
o
d
u
c
e
non
-
l
i
n
e
a
r
i
t
y
i
n
t
h
es
e
m
i
d
d
l
e
l
ay
e
r
s
,
a
s
t
h
i
s
h
el
p
s
t
h
e
m
o
d
el
l
ea
r
n
m
o
r
e
c
o
m
p
l
e
x
c
o
r
r
e
l
a
t
i
o
n
s
i
n
t
h
e
d
a
t
a
.
3
.
5
.
O
utput
l
a
y
er
Fin
ally
,
th
e
m
o
d
el
in
clu
d
es
an
o
u
tp
u
t
l
ay
er
,
th
e
h
ea
r
t
o
f
th
e
m
o
d
el,
in
wh
ich
with
b
in
ar
y
class
if
icatio
n
th
e
m
o
d
el
ca
n
s
ay
if
th
o
s
e
f
ee
lin
g
s
ar
e
p
o
s
itiv
e
o
r
n
e
g
ativ
e.
T
h
is
lay
er
co
m
p
r
is
es
a
n
eu
r
o
n
with
th
e
s
ig
m
o
id
ac
tiv
atio
n
f
u
n
ctio
n
o
u
tp
u
ttin
g
a
v
al
u
e
b
etwe
en
0
an
d
1
.
T
h
is
n
u
m
b
er
r
ep
r
esen
ts
th
e
ch
an
ce
o
f
th
e
p
o
s
itiv
e
class
,
u
s
in
g
a
co
m
m
o
n
th
r
esh
o
ld
o
f
0
.
5
to
p
r
ed
ict
th
e
class
lab
el.
An
y
th
in
g
g
r
ea
t
er
th
an
h
alf
m
ea
n
s
th
at
we
ar
e
f
ee
lin
g
p
o
s
itiv
e
em
o
tio
n
s
an
d
an
y
th
in
g
less
th
an
h
alf
m
ea
n
s
th
at
we
a
r
e
f
ee
lin
g
n
eg
ativ
e
em
o
tio
n
s
.
4.
RE
SU
L
T
AND
CO
M
P
A
RIS
O
N
Pre
cisi
o
n
lev
el,
th
e
p
r
o
p
o
r
tio
n
o
f
p
o
s
itiv
e
p
r
e
d
ictio
n
s
m
ad
e
b
y
th
e
m
o
d
el
th
at
wer
e
ac
tu
all
y
co
r
r
ec
t.
T
h
e
p
r
ec
is
io
n
o
f
th
e
b
asic
L
STM
m
o
d
el
is
0
.
8
,
an
d
th
e
L
STM
with
d
ee
p
lay
e
r
h
as
a
h
ig
h
er
p
r
ec
is
io
n
o
f
0
.
8
5
in
th
is
co
m
p
ar
is
o
n
asp
ec
t.
T
h
is
m
ea
n
s
th
at
th
e
im
p
r
o
v
e
d
m
o
d
el
is
ca
p
ab
le
o
f
m
o
r
e
ac
c
u
r
ately
d
etec
tin
g
th
e
p
o
s
itiv
es,
th
u
s
r
ed
u
cin
g
th
e
n
u
m
b
er
o
f
f
alse p
o
s
itiv
es.
T
h
e
b
ase
L
STM
m
o
d
el
h
as
a
r
ec
all
o
f
0
.
8
5
;
with
d
ee
p
lay
er
ed
L
STM
,
it
r
ea
ch
es
0
.
9
1
.
T
h
is
is
an
in
d
icatio
n
th
at
th
e
d
ee
p
lay
er
h
elp
s
th
e
m
o
d
el
b
etter
d
is
cr
im
in
ate
p
o
s
itiv
e
ca
s
es,
w
h
ich
r
ed
u
ce
s
f
alse
n
eg
ativ
es.
T
h
e
F1
-
s
co
r
e,
wh
ich
b
alan
ce
s
p
r
ec
is
io
n
a
n
d
r
ec
all
in
to
a
s
in
g
le
m
etr
ic,
g
iv
es
an
av
er
ag
e
d
r
ep
r
esen
tatio
n
o
f
h
o
w
well
t
h
e
m
o
d
el
p
r
e
d
icts
wh
en
clas
s
if
y
in
g
n
o
n
-
m
ain
m
ater
ial.
T
h
e
F1
-
s
co
r
e
f
o
r
th
e
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
-
8
7
0
8
E
n
h
a
n
cin
g
s
en
timen
t a
n
a
lysi
s
th
r
o
u
g
h
d
ee
p
la
ye
r
in
teg
r
a
ti
o
n
w
ith
…
(
P
a
r
u
l D
u
b
ey
)
953
B
asic L
STM
Mo
d
el
is
0
.
8
0
,
an
d
th
e
F1
-
s
co
r
e
f
o
r
th
e
L
STM
with
d
ee
p
l
ay
er
is
0
.
8
6
.
T
h
is
m
ea
n
s
th
e
im
p
r
o
v
ed
m
o
d
el
k
ee
p
s
a
v
er
y
g
o
o
d
b
alan
ce
b
etwe
en
p
r
ec
is
io
n
an
d
r
ec
all,
wh
ich
r
esu
lts
in
en
h
an
ce
d
g
e
n
er
al
p
er
f
o
r
m
an
ce
.
W
ith
a
d
ee
p
la
y
er
o
n
th
e
L
STM
,
ad
d
in
g
it
to
th
e
m
o
d
el
b
r
i
n
g
s
a
s
u
b
s
tan
tial
in
cr
ea
s
e
in
p
er
f
o
r
m
an
ce
f
o
r
all
m
etr
ics.
Sam
e
f
o
r
b
asic
L
STM
o
n
l
y
;
its
d
ee
p
L
STM
lay
er
h
as
b
etter
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e.
T
h
is
in
tu
r
n
m
ea
n
s
th
at
th
e
ca
p
a
b
le
m
o
d
el
is
p
er
f
o
r
m
in
g
well;
it
will
r
ep
o
r
t f
ewe
r
tr
u
e
p
o
s
itiv
es
an
d
h
ig
h
er
p
r
e
d
ictio
n
s
f
o
r
n
o
n
-
a
cc
ep
ted
s
tu
d
e
n
ts
.
So
,
f
o
r
ap
p
licatio
n
s
th
at
n
ee
d
h
ig
h
p
r
ec
is
io
n
,
r
ec
all,
a
n
d
b
alan
ce
d
p
er
f
o
r
m
a
n
ce
,
L
STM
with
d
ee
p
lay
er
is
a
g
o
o
d
o
p
t
io
n
.
Fig
u
r
e
2
s
h
o
ws
th
e
co
m
p
ar
is
o
n
r
esu
lt
b
ased
o
n
p
r
ec
is
io
n
r
ec
all
an
d
t
h
e
F
1
-
s
co
r
e.
Fig
u
r
e
2
.
C
o
m
p
a
r
is
o
n
f
o
r
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
4
.
1
.
T
ra
ini
ng
a
nd
v
a
lid
a
t
io
n
lo
s
s
I
n
th
e
L
STM
m
o
d
el,
t
h
e
tr
ain
lo
s
s
is
d
ec
r
ea
s
in
g
o
v
er
th
e
e
p
o
ch
s
,
wh
ic
h
in
d
icate
s
th
at
th
e
m
o
d
el
is
lear
n
in
g
f
r
o
m
tr
ain
i
n
g
d
ata
v
er
y
ef
f
ec
tiv
ely
.
T
h
is
co
n
s
is
ten
t d
ec
lin
e
s
h
o
ws th
at
th
e
m
o
d
el
is
ab
le
to
r
ed
u
ce
t
h
e
tr
ain
in
g
er
r
o
r
.
B
u
t
th
e
v
alid
atio
n
lo
s
s
d
ec
r
ea
s
es
an
d
th
e
n
s
tay
s
p
u
t
in
th
e
ea
r
ly
m
i
d
d
le;
it
ev
en
s
tar
ts
in
cr
ea
s
in
g
to
war
d
s
th
e
en
d
,
b
u
t
v
er
y
s
lig
h
tly
.
T
h
is
ca
n
b
e
a
n
in
d
icatio
n
th
at
th
e
m
o
d
el
is
b
eg
in
n
in
g
to
o
v
er
f
it
th
e
tr
ain
in
g
d
ata
d
u
r
in
g
tr
ain
i
n
g
.
I
n
s
im
p
le
wo
r
d
s
,
o
v
e
r
f
itti
n
g
ar
is
es
wh
en
th
e
m
o
d
el
d
et
er
m
in
es
th
e
tr
ain
in
g
d
ata
s
o
well
th
at
it
ca
tch
es
th
e
n
o
is
e
an
d
p
atter
n
s
th
at
ar
e
p
ec
u
liar
to
th
e
tr
ain
i
n
g
d
ata
a
n
d
will
n
o
t
tr
ac
k
th
e
n
ew,
u
n
s
ee
n
d
ata.
O
n
t
h
e
o
t
h
e
r
h
a
n
d
,
t
h
e
t
r
a
i
n
i
n
g
l
o
s
s
f
o
r
t
h
e
L
S
T
M
w
i
t
h
d
ee
p
l
a
y
e
r
i
s
w
a
y
s
h
o
r
t
e
r
t
h
a
n
t
h
a
t
o
f
t
h
e
s
i
m
p
l
is
ti
c
L
S
T
M
m
o
d
e
l
.
S
u
c
h
a
l
a
r
g
e
d
ec
r
e
a
s
e
i
n
t
h
e
t
r
ai
n
i
n
g
l
o
s
s
i
n
d
i
c
at
e
s
t
h
a
t
t
h
e
d
e
e
p
m
o
d
e
l
i
n
f
a
c
t
l
e
a
r
n
s
t
h
e
t
r
a
i
n
i
n
g
d
a
t
a
e
x
t
r
e
m
e
l
y
w
e
l
l
,
a
ll
o
w
i
n
g
i
t
t
o
g
e
n
e
r
a
l
iz
e
f
u
r
t
h
e
r
a
n
d
p
i
c
k
u
p
c
o
m
p
l
e
x
p
a
t
t
e
r
n
s
a
n
d
r
e
l
a
t
i
o
n
s
h
i
p
s
.
Se
c
o
n
d
l
y
,
w
e
n
o
t
i
c
e
t
h
a
t
t
h
e
v
a
l
i
d
at
i
o
n
l
o
s
s
is
r
e
l
a
t
i
v
el
y
h
i
g
h
a
n
d
d
o
e
s
n
o
t
s
h
o
w
a
l
a
r
g
e
d
e
c
r
e
as
e.
T
h
e
m
o
d
e
l
w
i
t
h
t
h
e
m
o
r
e
c
o
m
p
l
i
c
a
t
e
d
d
e
e
p
la
y
e
r
i
s
l
e
a
r
n
i
n
g
p
a
t
t
e
r
n
s
t
h
at
a
r
e
f
o
u
n
d
w
i
t
h
i
n
th
e
t
r
a
i
n
i
n
g
d
a
t
a
,
b
u
t
t
h
e
s
e
p
a
t
t
e
r
n
s
a
r
e
n
o
t
g
e
n
e
r
a
l
(
t
h
e
y
'
r
e
o
n
l
y
p
r
e
s
e
n
t
i
n
t
h
e
t
r
a
i
n
i
n
g
d
a
t
a
)
.
S
o
,
t
h
e
m
o
d
e
l
i
s
o
v
e
r
f
i
t
t
i
n
g
t
o
t
h
e
t
r
a
i
n
i
n
g
d
a
t
a
a
n
d
t
h
e
r
e
f
o
r
e
n
o
t
g
e
n
e
r
a
l
i
zi
n
g
as
we
l
l
t
o
t
h
e
v
a
l
i
d
a
t
i
o
n
d
a
t
a
t
h
is
ti
m
e
.
T
h
e
h
i
g
h
v
a
l
i
d
a
ti
o
n
l
o
s
s
,
e
v
e
n
t
h
o
u
g
h
i
t
g
i
v
e
s
a
r
e
al
l
y
g
o
o
d
p
e
r
f
o
r
m
a
n
c
e
o
n
t
h
e
t
r
a
i
n
i
n
g
s
e
t
,
i
n
d
ic
a
t
es
b
a
d
g
e
n
e
r
a
l
i
za
t
i
o
n
.
4
.
2
.
T
ra
in v
s
v
a
lid
a
t
i
o
n a
cc
ura
cy
T
r
ain
in
g
ac
cu
r
ac
y
f
o
r
th
e
L
S
T
M
m
o
d
el
in
cr
ea
s
es
at
a
s
tea
d
y
r
ate
ac
r
o
s
s
th
e
ep
o
c
h
s
,
wh
ich
s
u
g
g
ests
th
at
it
is
lear
n
in
g
co
n
s
is
ten
tly
.
T
h
e
ac
cu
r
ac
y
in
cr
ea
s
in
g
s
lo
w
ly
s
h
o
ws
th
at
th
e
m
o
d
el
is
p
r
e
d
ictin
g
th
e
co
r
r
ec
t
s
en
tim
en
t
f
r
o
m
th
e
tr
ain
in
g
d
ata.
T
h
e
v
alid
atio
n
ac
cu
r
ac
y
is
a
to
u
ch
lo
wer
th
an
o
u
r
tr
a
in
in
g
ac
cu
r
ac
y
b
u
t
f
lu
ctu
ates
m
u
ch
less
o
v
er
th
e
co
u
r
s
e
o
f
th
e
ep
o
ch
s
.
T
h
is
co
n
s
is
ten
cy
m
ea
n
s
th
at
th
e
m
o
d
el
is
g
en
er
alizin
g
well
an
d
co
n
tin
u
in
g
to
p
er
f
o
r
m
r
eliab
ly
o
n
u
n
s
ee
n
d
ata.
T
h
e
L
STM
with
d
ee
p
lay
er
q
u
ick
ly
g
ets
p
er
f
ec
t
tr
ain
in
g
ac
cu
r
ac
y
an
d
n
ev
e
r
m
o
v
es.
T
h
is
h
ig
h
s
co
r
e
s
u
g
g
ests
th
e
m
o
d
el
is
o
v
er
f
it
tin
g
,
an
d
it
is
v
er
y
g
o
o
d
at
l
ea
r
n
in
g
t
h
e
tr
ain
in
g
d
ata
an
d
in
f
er
r
in
g
co
m
p
lex
p
atter
n
s
.
B
u
t it
d
o
es n
o
t
s
ee
m
ajo
r
im
p
r
o
v
e
m
en
t in
th
e
v
alid
atio
n
ac
cu
r
ac
y
,
an
d
it is
a
b
it
wo
r
s
e
th
an
th
e
b
asic
m
o
d
el,
as
s
h
o
wn
b
elo
w.
T
h
e
f
ac
t
th
at
tr
ain
in
g
an
d
v
ali
d
atio
n
ac
cu
r
ac
y
ar
e
q
u
ite
d
i
s
tan
t
r
ea
f
f
ir
m
s
o
u
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
25
:
9
4
9
-
957
954
s
u
s
p
icio
n
o
f
o
v
e
r
f
itti
n
g
.
Su
ch
a
m
o
d
el
wo
u
ld
f
it
th
e
tr
ain
in
g
d
ata
v
er
y
well
b
u
t
wo
u
ld
f
a
il
to
r
ea
ch
th
e
s
am
e
lev
el
o
f
p
r
ed
ictio
n
o
n
th
e
v
alid
atio
n
s
et,
wh
ich
m
ea
n
s
it h
as p
o
o
r
g
en
er
aliza
tio
n
ca
p
ab
ilit
y
.
T
h
e
L
STM
m
o
d
el
with
a
n
ex
t
r
a
-
d
ee
p
la
y
er
g
iv
es
a
q
u
an
tita
tiv
ely
s
tr
o
n
g
er
p
er
f
o
r
m
an
ce
o
n
tr
ain
in
g
b
u
t
lack
s
g
en
er
alizin
g
ab
ilit
y
.
T
h
is
is
ev
id
en
ce
d
b
y
th
e
lar
g
e
v
alid
atio
n
lo
s
s
an
d
th
e
u
n
u
s
u
al
s
tab
ilit
y
o
f
th
e
v
alid
atio
n
ac
cu
r
ac
y
.
T
h
is
o
v
e
r
f
itti
n
g
s
ee
n
in
th
e
im
p
r
o
v
ed
m
o
d
el
r
ev
ea
ls
th
at
th
e
m
o
d
el
is
ab
le
to
ca
p
tu
r
e
d
etailed
in
f
o
r
m
atio
n
f
r
o
m
t
h
e
tr
ain
in
g
d
ata,
b
u
t
th
is
ca
n
n
o
t
b
e
g
en
er
alize
d
to
u
n
s
ee
n
d
a
ta.
R
eg
u
lar
izatio
n
:
Ad
d
itio
n
al
m
ea
s
u
r
es
to
im
p
r
o
v
e
th
e
g
en
e
r
aliza
tio
n
p
er
f
o
r
m
a
n
ce
o
f
th
e
en
h
an
ce
d
m
o
d
el
co
u
ld
b
e
n
ee
d
ed
,
m
a
y
it
b
e
d
r
o
p
o
u
t,
ea
r
ly
s
to
p
p
in
g
,
o
r
a
m
o
r
e
ag
g
r
ess
iv
e
d
ata
au
g
m
en
tatio
n
.
T
h
ese
tech
n
iq
u
es
ad
d
co
n
s
tr
ain
ts
th
at
h
elp
a
m
o
d
el
b
y
tr
ain
in
g
to
p
r
ev
en
t
o
v
er
f
itti
n
g
,
wh
ich
m
ay
in
tu
r
n
m
ak
e
a
m
o
d
el
m
o
r
e
g
en
er
aliza
b
le
to
n
ew
d
ata.
Fig
u
r
e
3
.
d
ep
icts
th
e
tr
ai
n
in
g
an
d
v
alid
atio
n
lo
s
s
/acc
u
r
ac
y
o
f
b
o
th
m
o
d
els.
Fig
u
r
e
3
.
T
r
ain
in
g
a
n
d
v
alid
atio
n
lo
s
s
an
d
ac
cu
r
ac
y
4
.
3
.
Co
rr
el
a
t
io
n
plo
t
a
na
l
y
s
is
I
n
n
eu
r
al
n
etwo
r
k
tr
ain
in
g
,
it
is
ess
en
tial
to
an
aly
ze
th
e
r
elatio
n
s
h
ip
b
etwe
en
tr
ain
in
g
an
d
v
alid
atio
n
m
etr
ics
to
ass
ess
m
o
d
el
p
e
r
f
o
r
m
an
ce
an
d
g
e
n
er
aliza
tio
n
c
ap
ab
ilit
ies.
C
o
r
r
elatio
n
p
lo
ts
h
elp
v
is
u
alize
h
o
w
tr
ain
in
g
lo
s
s
,
v
alid
atio
n
l
o
s
s
,
an
d
ac
cu
r
ac
y
m
etr
ics
ev
o
l
v
e
o
v
er
tim
e,
p
r
o
v
id
i
n
g
in
s
ig
h
ts
in
to
m
o
d
el
b
eh
av
io
r
.
T
h
is
co
m
p
ar
is
o
n
is
p
ar
ticu
la
r
ly
im
p
o
r
ta
n
t
wh
en
ev
al
u
atin
g
d
if
f
er
en
t
ar
ch
itectu
r
es,
s
u
ch
a
s
a
s
tan
d
ar
d
L
STM
m
o
d
el
v
er
s
u
s
an
L
STM
with
an
a
d
d
itio
n
al
d
ee
p
lay
er
,
to
d
eter
m
in
e
wh
ich
m
o
d
el
ac
h
i
ev
es
b
etter
o
v
er
all
p
er
f
o
r
m
an
ce
an
d
g
e
n
er
alize
s
m
o
r
e
ef
f
ec
tiv
el
y
o
n
u
n
s
ee
n
d
a
ta.
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
-
8
7
0
8
E
n
h
a
n
cin
g
s
en
timen
t a
n
a
lysi
s
th
r
o
u
g
h
d
ee
p
la
ye
r
in
teg
r
a
ti
o
n
w
ith
…
(
P
a
r
u
l D
u
b
ey
)
955
a.
L
STM
m
o
d
el
co
r
r
elatio
n
m
atr
ix
−
T
r
ain
in
g
L
o
s
s
v
s
.
Valid
atio
n
L
o
s
s
:
T
h
er
e
is
a
p
o
s
itiv
e
co
r
r
e
latio
n
b
etwe
en
tr
ain
in
g
an
d
v
a
lid
atio
n
lo
s
s
,
wh
ich
m
ea
n
s
th
at
as
th
e
tr
ain
in
g
lo
s
s
d
ec
r
ea
s
es,
th
e
v
alid
a
tio
n
lo
s
s
d
ec
r
ea
s
es
a
s
wel
l.
I
t
m
ea
n
s
th
e
m
o
d
el
is
g
en
er
alizin
g
well.
−
T
r
ain
in
g
a
n
d
v
alid
atio
n
ac
cu
r
ac
y
:
T
r
ain
in
g
ac
cu
r
ac
y
is
in
d
ir
ec
t
p
r
o
p
o
r
tio
n
to
v
ali
d
atio
n
ac
cu
r
ac
y
.
R
em
in
d
y
o
u
r
s
elf
th
at
wh
ate
v
e
r
is
d
o
n
e
to
im
p
r
o
v
e
tr
ai
n
in
g
a
cc
u
r
ac
y
(
th
e
h
o
r
izo
n
tal
ax
is
)
a
ls
o
ten
d
s
to
im
p
r
o
v
e
v
alid
atio
n
ac
c
u
r
ac
y
(
t
h
e
v
er
tical
ax
is
)
,
wh
ich
is
a
s
ig
n
o
f
g
o
o
d
g
e
n
er
aliza
tio
n
.
b.
L
STM
with
d
ee
p
lay
er
: c
o
r
r
el
atio
n
m
atr
ix
−
T
r
ain
in
g
lo
s
s
v
s
.
v
ali
d
atio
n
l
o
s
s
:
Sti
ll
in
d
icate
s
a
h
ig
h
p
o
s
itiv
e
r
elatio
n
s
h
ip
,
b
u
t
p
o
s
s
ib
l
y
lo
wer
th
an
th
e
b
asic
L
STM
m
o
d
el.
W
h
il
e
th
e
tr
ain
in
g
lo
s
s
d
im
in
is
h
es
co
n
s
id
er
ab
ly
,
th
e
v
alid
atio
n
l
o
s
s
g
ets
lo
w
(
v
er
y
h
ig
h
co
m
p
ar
ed
to
th
e
tr
ain
in
g
lo
s
s
)
,
w
h
ich
m
a
y
i
n
d
ic
ate
th
at
th
e
tr
ain
in
g
lo
s
s
is
o
v
er
f
itti
n
g
,
b
u
t
th
e
v
alid
atio
n
lo
s
s
is
n
o
t.
−
T
r
ain
in
g
ac
cu
r
ac
y
v
s
.
v
alid
ati
o
n
ac
cu
r
ac
y
:
th
e
co
r
r
elatio
n
b
etwe
en
tr
ain
in
g
an
d
v
alid
atio
n
ac
cu
r
ac
y
is
s
tr
o
n
g
an
d
h
as
a
s
lig
h
tly
d
if
f
er
en
t
b
e
h
av
io
r
co
m
p
ar
ed
to
t
h
e
b
asic
L
STM
m
o
d
el.
W
h
en
th
e
tr
ain
i
n
g
ac
cu
r
ac
y
is
h
ig
h
an
d
we
h
av
e
d
if
f
er
en
ce
s
m
o
r
e
lik
e
th
is
,
it su
g
g
ests
th
at
th
e
m
o
d
el
is
o
v
er
f
itti
n
g
.
Fro
m
th
e
co
r
r
elatio
n
p
l
o
ts
,
th
e
v
ar
iatio
n
in
h
o
w
m
u
ch
th
e
m
o
d
els
g
en
e
r
alize
f
r
o
m
t
h
e
tr
ain
in
g
d
ata
to
th
e
u
n
s
ee
n
v
alid
atio
n
d
ata
c
an
b
e
s
ee
n
.
T
h
e
s
im
p
le
L
STM
m
o
d
el
s
ee
m
s
to
h
av
e
r
ea
s
o
n
ab
le
g
en
er
aliza
tio
n
,
with
s
tab
le
b
eh
av
io
r
ac
r
o
s
s
th
e
tr
ain
in
g
a
n
d
v
alid
atio
n
m
etr
ics.
T
h
e
L
STM
m
o
d
el
with
a
d
ee
p
la
y
er
s
u
f
f
e
r
s
f
r
o
m
o
v
er
f
itti
n
g
,
wh
ich
is
wh
en
th
e
m
o
d
el
p
e
r
f
o
r
m
an
ce
is
am
az
in
g
o
n
tr
ain
d
ata,
b
u
t
wh
en
we
v
alid
ate
th
is
m
o
d
el,
it
is
s
o
m
y
s
tiq
u
e
b
ec
a
u
s
e
th
e
m
o
d
el
j
u
s
t
lear
n
s
ex
is
tin
g
d
ata
with
o
u
t
lear
n
in
g
t
h
e
p
atter
n
co
r
r
elatio
n
in
th
e
d
ata.
W
ith
th
ese
less
o
n
s
,
m
o
r
e
wo
r
k
ca
n
b
e
d
o
n
e
to
im
p
r
o
v
e
th
e
d
esig
n
o
f
m
o
d
els
an
d
th
e
r
e
g
u
lar
izatio
n
tech
n
iq
u
es
to
f
u
r
th
er
im
p
r
o
v
e
g
en
er
aliza
tio
n
p
e
r
f
o
r
m
an
ce
.
Fig
u
r
e
4
s
h
o
ws
th
e
c
o
r
r
elatio
n
p
lo
t
an
aly
s
is
f
o
r
b
o
th
m
o
d
els.
T
ab
le
3
co
m
p
a
r
e
s
th
e
p
er
f
o
r
m
an
ce
s
o
f
L
STM
with
th
e
d
ee
p
lay
e
r
m
o
d
el
.
Fig
u
r
e
4
.
C
o
r
r
elatio
n
p
lo
t a
n
al
y
s
is
-
co
m
p
ar
is
o
n
Tab
le 3
.
Co
m
p
a
riso
n
o
f
LS
T
M
a
n
d
LS
T
M
wit
h
d
e
e
p
lay
e
r
m
o
d
e
l
M
e
t
r
i
c
LSTM
m
o
d
e
l
LSTM
w
i
t
h
d
e
e
p
l
a
y
e
r
Tr
a
i
n
i
n
g
l
o
ss
0
.
1
6
7
6
0
.
0
5
2
4
V
a
l
i
d
a
t
i
o
n
l
o
ss
0
.
3
4
1
9
0
.
6
6
0
4
Tr
a
i
n
i
n
g
a
c
c
u
r
a
c
y
0
.
9
3
6
7
0
.
9
8
3
4
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
0
.
8
7
6
4
0
.
8
6
7
5
P
r
e
c
i
s
i
o
n
0
.
8
0
.
8
5
R
e
c
a
l
l
0
.
8
5
0
.
9
1
F1
-
s
c
o
r
e
0
.
8
0
.
8
6
5.
CO
NCLU
SI
O
N
T
h
is
ar
ticle
lo
o
k
ed
at
th
e
s
en
tim
en
t
an
aly
s
is
o
f
ad
d
in
g
a
d
ee
p
lay
er
to
an
L
STM
n
et
wo
r
k
.
Ou
r
ex
ten
s
iv
e
ev
alu
atio
n
,
wh
ich
r
an
g
ed
f
r
o
m
d
ata
p
r
ep
r
o
ce
s
s
in
g
to
m
o
d
el
ev
al
u
atio
n
,
p
r
o
v
ed
th
at
th
e
d
ee
p
-
lay
er
L
STM
m
o
d
el
h
a
d
b
etter
ac
c
u
r
ac
y
,
r
ec
all,
a
n
d
F1
-
s
co
r
e
co
m
p
ar
ed
t
o
th
e
b
asic
m
o
d
el.
T
h
e
d
ee
p
lay
er
was
tr
ain
ed
to
r
ec
o
r
d
co
m
p
lex
s
eq
u
en
ce
s
an
d
s
ig
n
if
ican
tly
i
m
p
r
o
v
e
d
p
r
ed
ictiv
e
ac
c
u
r
ac
y
.
Nev
er
th
eless
,
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
25
:
9
4
9
-
957
956
u
p
g
r
a
d
ed
m
o
d
el
o
v
er
f
itted
,
w
h
ich
ca
lled
f
o
r
e
v
en
m
o
r
e
r
e
g
u
lar
izatio
n
a
n
d
h
y
p
er
p
ar
a
m
eter
tu
n
i
n
g
.
Ho
wev
er
,
a
co
r
e
lev
el
o
f
e
n
h
an
ce
d
s
en
ti
m
en
t
an
aly
s
is
.
I
n
s
u
m
m
ar
y
,
th
e
m
ain
co
n
tr
ib
u
tio
n
o
f
th
is
wo
r
k
is
u
s
in
g
a
d
ee
p
-
lay
er
ed
L
STM
n
etwo
r
k
to
i
n
c
r
ea
s
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
m
o
d
el.
I
n
f
u
tu
r
e
wo
r
k
,
th
ese
is
s
u
es,
o
v
er
f
itti
n
g
,
an
d
s
o
m
e
ef
f
ec
tiv
e
m
o
d
el
o
p
tim
izatio
n
will
b
e
a
p
p
lied
t
o
t
h
e
h
ig
h
-
s
co
r
e
g
am
e
f
o
r
b
etter
g
en
er
aliza
tio
n
an
d
ac
cu
r
ac
y
.
T
h
is
s
tu
d
y
b
e
n
ef
its
d
ee
p
lear
n
in
g
m
o
d
els f
o
r
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
.
RE
F
E
R
E
NC
E
S
[
1
]
A
.
C
.
M
.
V
.
S
r
i
n
i
v
a
s,
C
.
S
a
t
y
a
n
a
r
a
y
a
n
a
,
C
.
D
i
v
a
k
a
r
,
a
n
d
K
.
P
.
S
i
r
i
sh
a
,
“
S
e
n
t
i
me
n
t
a
n
a
l
y
s
i
s
u
si
n
g
n
e
u
r
a
l
n
e
t
w
o
r
k
a
n
d
LST
M
,
”
I
O
P
C
o
n
f
e
re
n
c
e
S
e
r
i
e
s:
M
a
t
e
ri
a
l
s
S
c
i
e
n
c
e
a
n
d
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
0
7
4
,
n
o
.
1
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
8
8
/
1
7
5
7
-
8
9
9
x
/
1
0
7
4
/
1
/
0
1
2
0
0
7
.
[
2
]
P
.
F
.
M
u
h
a
m
m
a
d
,
R
.
K
u
s
u
m
a
n
i
n
g
r
u
m
,
a
n
d
A
.
W
i
b
o
w
o
,
“
S
e
n
t
i
m
e
n
t
a
n
a
l
y
s
i
s
u
s
i
n
g
W
o
r
d
2
V
e
c
a
n
d
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
(
L
S
T
M
)
f
o
r
I
n
d
o
n
e
s
i
a
n
h
o
t
e
l
r
e
v
i
e
w
s
,
”
P
r
o
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
1
7
9
,
p
p
.
7
2
8
–
7
3
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s
.
2
0
2
1
.
0
1
.
0
6
1
.
[
3
]
U
.
B
.
M
a
h
a
d
e
v
a
sw
a
m
y
a
n
d
P
.
S
w
a
t
h
i
,
“
S
e
n
t
i
me
n
t
a
n
a
l
y
s
i
s
u
s
i
n
g
b
i
d
i
r
e
c
t
i
o
n
a
l
LST
M
n
e
t
w
o
r
k
,
”
Pro
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
2
1
8
,
p
p
.
4
5
–
5
6
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.
2
0
2
2
.
1
2
.
4
0
0
.
[
4
]
U
.
D
.
G
a
n
d
h
i
,
P
.
M
.
K
u
m
a
r
,
G
.
C
.
B
a
b
u
,
a
n
d
G
.
K
a
r
t
h
i
c
k
,
“
S
e
n
t
i
me
n
t
a
n
a
l
y
si
s
o
n
t
w
i
t
t
e
r
d
a
t
a
b
y
u
s
i
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
(
C
N
N
)
a
n
d
l
o
n
g
sh
o
r
t
-
t
e
r
m
mem
o
r
y
(
LS
TM
)
,
”
Wi
r
e
l
e
s
s
P
e
rso
n
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
s
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
2
7
7
-
021
-
0
8
5
8
0
-
3.
[
5
]
R
.
K
.
B
e
h
e
r
a
,
M
.
J
e
n
a
,
S
.
K
.
R
a
t
h
,
a
n
d
S
.
M
i
sr
a
,
“
Co
-
LST
M
:
c
o
n
v
o
l
u
t
i
o
n
a
l
LST
M
mo
d
e
l
f
o
r
se
n
t
i
me
n
t
a
n
a
l
y
s
i
s
i
n
s
o
c
i
a
l
b
i
g
d
a
t
a
,
”
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
a
n
d
M
a
n
a
g
e
m
e
n
t
,
v
o
l
.
5
8
,
n
o
.
1
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
p
m.
2
0
2
0
.
1
0
2
4
3
5
.
[
6
]
Z.
J
i
n
,
Y
.
Y
a
n
g
,
a
n
d
Y
.
Li
u
,
“
S
t
o
c
k
c
l
o
s
i
n
g
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
b
a
s
e
d
o
n
se
n
t
i
me
n
t
a
n
a
l
y
si
s
a
n
d
LST
M
,
”
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
2
,
n
o
.
1
3
,
p
p
.
9
7
1
3
–
9
7
2
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s0
0
5
2
1
-
0
1
9
-
0
4
5
0
4
-
2.
[
7
]
B
.
Li
n
d
e
ma
n
n
,
T.
M
ü
l
l
e
r
,
H
.
V
i
e
t
z
,
N
.
Jaz
d
i
,
a
n
d
M
.
W
e
y
r
i
c
h
,
“
A
s
u
r
v
e
y
o
n
l
o
n
g
sh
o
r
t
-
t
e
r
m
mem
o
r
y
n
e
t
w
o
r
k
s
f
o
r
t
i
me
seri
e
s
p
r
e
d
i
c
t
i
o
n
,
”
Pr
o
c
e
d
i
a
C
I
R
P
,
v
o
l
.
9
9
,
p
p
.
6
5
0
–
6
5
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
i
r
.
2
0
2
1
.
0
3
.
0
8
8
.
[
8
]
A
.
S
h
e
r
st
i
n
s
k
y
,
“
F
u
n
d
a
me
n
t
a
l
s
o
f
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
(
R
N
N
)
a
n
d
l
o
n
g
sh
o
r
t
-
t
e
r
m
mem
o
r
y
(
LST
M
)
n
e
t
w
o
r
k
,
”
Ph
y
si
c
a
D
:
N
o
n
l
i
n
e
a
r P
h
e
n
o
m
e
n
a
,
v
o
l
.
4
0
4
,
M
a
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
h
y
s
d
.
2
0
1
9
.
1
3
2
3
0
6
.
[
9
]
M
.
A
l
i
z
a
mi
r
e
t
a
l
.
,
“
I
mp
r
o
v
i
n
g
t
h
e
a
c
c
u
r
a
c
y
o
f
d
a
i
l
y
s
o
l
a
r
r
a
d
i
a
t
i
o
n
p
r
e
d
i
c
t
i
o
n
b
y
c
l
i
ma
t
i
c
d
a
t
a
u
si
n
g
a
n
e
f
f
i
c
i
e
n
t
h
y
b
r
i
d
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
:
L
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
mo
r
y
(
LST
M
)
n
e
t
w
o
r
k
c
o
u
p
l
e
d
w
i
t
h
w
a
v
e
l
e
t
t
r
a
n
sf
o
r
m,
”
E
n
g
i
n
e
e
r
i
n
g
Ap
p
l
i
c
a
t
i
o
n
s
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
2
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
a
p
p
a
i
.
2
0
2
3
.
1
0
6
1
9
9
.
[
1
0
]
Z.
G
a
o
e
t
a
l
.
,
“
A
n
o
v
e
l
mu
l
t
i
v
a
r
i
a
t
e
t
i
me
seri
e
s
p
r
e
d
i
c
t
i
o
n
o
f
c
r
u
c
i
a
l
w
a
t
e
r
q
u
a
l
i
t
y
p
a
r
a
me
t
e
r
s
w
i
t
h
l
o
n
g
sh
o
r
t
-
t
e
r
m
mem
o
r
y
(
LSTM
)
n
e
t
w
o
r
k
s,
”
J
o
u
rn
a
l
o
f
C
o
n
t
a
m
i
n
a
n
t
H
y
d
r
o
l
o
g
y
,
v
o
l
.
2
5
9
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
c
o
n
h
y
d
.
2
0
2
3
.
1
0
4
2
6
2
.
[
1
1
]
M
.
W
e
n
,
M
.
S
.
I
b
r
a
h
i
m,
A
.
H
.
M
e
d
a
,
G
.
Zh
a
n
g
,
a
n
d
J.
F
a
n
,
“
In
-
S
i
t
u
e
a
r
l
y
a
n
o
mal
y
d
e
t
e
c
t
i
o
n
a
n
d
r
e
m
a
i
n
i
n
g
u
sefu
l
l
i
f
e
t
i
me
p
r
e
d
i
c
t
i
o
n
f
o
r
h
i
g
h
-
p
o
w
e
r
w
h
i
t
e
LE
D
s
w
i
t
h
d
i
st
a
n
c
e
a
n
d
e
n
t
r
o
p
y
-
b
a
s
e
d
l
o
n
g
sh
o
r
t
-
t
e
r
m
mem
o
r
y
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
Ex
p
e
rt
S
y
st
e
m
s w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
3
8
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
3
.
1
2
1
8
3
2
.
[
1
2
]
S
.
C
h
a
e
,
W
.
N
.
S
t
r
e
e
t
,
N
.
R
a
mar
a
j
u
,
a
n
d
S
.
G
i
l
b
e
r
t
so
n
-
W
h
i
t
e
,
“
P
r
e
d
i
c
t
i
o
n
o
f
c
a
n
c
e
r
sy
mp
t
o
m
t
r
a
j
e
c
t
o
r
y
u
si
n
g
l
o
n
g
i
t
u
d
i
n
a
l
e
l
e
c
t
r
o
n
i
c
h
e
a
l
t
h
r
e
c
o
r
d
d
a
t
a
a
n
d
l
o
n
g
sh
o
r
t
-
t
e
r
m
m
e
m
o
r
y
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
J
C
O
C
l
i
n
i
c
a
l
C
a
n
c
e
r
I
n
f
o
rm
a
t
i
c
s
,
n
o
.
8
,
2
0
2
4
,
d
o
i
:
1
0
.
1
2
0
0
/
c
c
i
.
2
3
.
0
0
0
3
9
.
[
1
3
]
M
.
L
i
u
,
W
.
Zh
o
u
,
a
n
d
Z
.
X
u
,
“
H
e
s
i
t
a
n
t
f
u
z
z
y
l
o
n
g
sh
o
r
t
-
t
e
r
m
m
e
mo
r
y
n
e
t
w
o
r
k
a
n
d
i
t
s
a
p
p
l
i
c
a
t
i
o
n
i
n
t
h
e
i
n
t
e
l
l
i
g
e
n
t
b
u
i
l
d
i
n
g
sel
e
c
t
i
o
n
,
”
I
EE
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
F
u
z
zy
S
y
st
e
m
s
,
v
o
l
.
3
2
,
n
o
.
5
,
p
p
.
2
5
9
0
–
2
6
0
2
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
TFU
ZZ.
2
0
2
4
.
3
3
5
5
0
0
0
.
[
1
4
]
J.
S
h
a
n
,
X
.
Z
h
a
n
g
,
Y
.
Li
u
,
C
.
Z
h
a
n
g
,
a
n
d
J.
Z
h
o
u
,
“
D
e
f
o
r
m
a
t
i
o
n
p
r
e
d
i
c
t
i
o
n
o
f
l
a
r
g
e
-
sc
a
l
e
c
i
v
i
l
st
r
u
c
t
u
r
e
s
u
si
n
g
s
p
a
t
i
o
t
e
m
p
o
r
a
l
c
l
u
st
e
r
i
n
g
a
n
d
e
mp
i
r
i
c
a
l
m
o
d
e
d
e
c
o
mp
o
si
t
i
o
n
-
b
a
s
e
d
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
mo
r
y
n
e
t
w
o
r
k
,
”
A
u
t
o
m
a
t
i
o
n
i
n
C
o
n
s
t
ru
c
t
i
o
n
,
v
o
l
.
1
5
8
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
u
t
c
o
n
.
2
0
2
3
.
1
0
5
2
2
2
.
[
1
5
]
A
.
G
o
g
i
n
e
n
i
,
M
.
K
.
D
.
R
o
u
t
,
a
n
d
K
.
S
h
u
b
h
a
m,
“
E
v
a
l
u
a
t
i
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
f
o
r
p
r
e
d
i
c
t
i
n
g
c
o
m
p
r
e
ssi
v
e
st
r
e
n
g
t
h
o
f
c
o
n
c
r
e
t
e
w
i
t
h
mi
n
e
r
a
l
a
d
mi
x
t
u
r
e
u
s
i
n
g
l
o
n
g
s
h
o
r
t
-
t
e
r
m
mem
o
r
y
(
LST
M
)
T
e
c
h
n
i
q
u
e
,
”
Asi
a
n
J
o
u
r
n
a
l
o
f
C
i
v
i
l
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
2
5
,
n
o
.
2
,
p
p
.
1
9
2
1
–
1
9
3
3
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s4
2
1
0
7
-
023
-
0
0
8
8
5
-
x.
[
1
6
]
B
.
G
h
o
j
o
g
h
a
n
d
A
.
G
h
o
d
s
i
,
“
R
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s
a
n
d
l
o
n
g
s
h
o
r
t
-
t
e
r
m
mem
o
r
y
n
e
t
w
o
r
k
s
:
t
u
t
o
r
i
a
l
a
n
d
s
u
r
v
e
y
,
”
a
rXi
v
:
2
3
0
4
.
1
1
4
6
1
,
2
0
2
3
.
[
1
7
]
F
.
Li
u
,
J.
L
i
,
a
n
d
L.
W
a
n
g
,
“
PI
-
LSTM
:
p
h
y
s
i
c
s
-
i
n
f
o
r
m
e
d
l
o
n
g
sh
o
r
t
-
t
e
r
m
memo
r
y
n
e
t
w
o
r
k
f
o
r
st
r
u
c
t
u
r
a
l
r
e
sp
o
n
s
e
mo
d
e
l
i
n
g
,
”
En
g
i
n
e
e
ri
n
g
S
t
r
u
c
t
u
r
e
s
,
v
o
l
.
2
9
2
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
s
t
r
u
c
t
.
2
0
2
3
.
1
1
6
5
0
0
.
[
1
8
]
S
.
H
.
P
a
r
k
e
t
a
l
.
,
“
D
e
v
e
l
o
p
m
e
n
t
o
f
a
so
i
l
m
o
i
st
u
r
e
p
r
e
d
i
c
t
i
o
n
m
o
d
e
l
b
a
se
d
o
n
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
l
o
n
g
sh
o
r
t
-
t
e
r
m
m
e
mo
r
y
(
R
N
N
-
LSTM
)
i
n
so
y
b
e
a
n
c
u
l
t
i
v
a
t
i
o
n
,
”
S
e
n
so
rs
,
v
o
l
.
2
3
,
n
o
.
4
,
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
3
0
4
1
9
7
6
.
[
1
9
]
C
.
A
v
c
i
,
B
.
T
e
k
i
n
e
r
d
o
g
a
n
,
a
n
d
C
.
C
a
t
a
l
,
“
A
n
a
l
y
z
i
n
g
t
h
e
p
e
r
f
o
r
ma
n
c
e
o
f
l
o
n
g
sh
o
r
t
-
t
e
r
m
mem
o
r
y
a
r
c
h
i
t
e
c
t
u
r
e
s
f
o
r
ma
l
w
a
r
e
d
e
t
e
c
t
i
o
n
m
o
d
e
l
s
,
”
C
o
n
c
u
rr
e
n
c
y
a
n
d
C
o
m
p
u
t
a
t
i
o
n
:
Pr
a
c
t
i
c
e
a
n
d
Ex
p
e
ri
e
n
c
e
,
v
o
l
.
3
5
,
n
o
.
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
2
/
c
p
e
.
7
5
8
1
.
[
2
0
]
S
.
B
a
n
i
a
s
a
d
i
,
R
.
S
a
l
e
h
i
,
S
.
S
o
l
t
a
n
i
,
D
.
M
a
r
t
í
n
,
P
.
P
o
u
r
m
a
n
d
,
a
n
d
E.
G
h
a
f
o
u
r
i
a
n
,
“
O
p
t
i
m
i
z
i
n
g
l
o
n
g
s
h
o
r
t
-
t
e
r
m me
mo
r
y
n
e
t
w
o
r
k
f
o
r
a
i
r
p
o
l
l
u
t
i
o
n
p
r
e
d
i
c
t
i
o
n
u
si
n
g
a
n
o
v
e
l
b
i
n
a
r
y
c
h
i
m
p
o
p
t
i
m
i
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m,
”
El
e
c
t
ro
n
i
c
s
(
S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
2
,
n
o
.
1
8
,
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s1
2
1
8
3
9
8
5
.
[
2
1
]
L.
W
.
Tsa
i
a
n
d
A
.
A
l
i
p
o
u
r
,
“
P
h
y
s
i
c
s
-
i
n
f
o
r
m
e
d
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
mo
r
y
n
e
t
w
o
r
k
s
f
o
r
r
e
s
p
o
n
se
p
r
e
d
i
c
t
i
o
n
o
f
a
w
i
n
d
-
e
x
c
i
t
e
d
f
l
e
x
i
b
l
e
st
r
u
c
t
u
r
e
,
”
En
g
i
n
e
e
r
i
n
g
S
t
r
u
c
t
u
res
,
v
o
l
.
2
7
5
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
st
r
u
c
t
.
2
0
2
2
.
1
1
4
9
6
8
.
[
2
2
]
R
.
G
h
o
sh
,
S
.
P
h
a
d
i
k
a
r
,
N
.
D
e
b
,
N
.
S
i
n
h
a
,
P
.
D
a
s
,
a
n
d
E.
G
h
a
d
e
r
p
o
u
r
,
“
A
u
t
o
ma
t
i
c
e
y
e
b
l
i
n
k
a
n
d
mu
sc
u
l
a
r
a
r
t
i
f
a
c
t
d
e
t
e
c
t
i
o
n
a
n
d
r
e
mo
v
a
l
f
r
o
m
E
EG
si
g
n
a
l
s
u
s
i
n
g
k
-
n
e
a
r
e
s
t
n
e
i
g
h
b
o
r
c
l
a
ss
i
f
i
e
r
a
n
d
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
n
e
t
w
o
r
k
s,
”
I
EEE
S
e
n
so
rs
J
o
u
r
n
a
l
,
v
o
l
.
2
3
,
n
o
.
5
,
p
p
.
5
4
2
2
–
5
4
3
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
EN
.
2
0
2
3
.
3
2
3
7
3
8
3
.
[
2
3
]
H
.
H
.
Z
h
a
n
g
,
H
.
M
.
Y
a
o
,
L.
J
i
a
n
g
,
a
n
d
M
.
N
g
,
“
D
e
e
p
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
mo
r
y
n
e
t
w
o
r
k
s
-
b
a
s
e
d
so
l
v
i
n
g
m
e
t
h
o
d
f
o
r
t
h
e
F
D
TD
met
h
o
d
:
2
-
D
c
a
se
,
”
I
E
EE
Mi
c
r
o
w
a
v
e
a
n
d
W
i
re
l
e
ss
T
e
c
h
n
o
l
o
g
y
L
e
t
t
e
rs
,
v
o
l
.
3
3
,
n
o
.
5
,
p
p
.
4
9
9
–
5
0
2
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
l
mw
t
.
2
0
2
2
.
3
2
2
3
9
5
9
.
[
2
4
]
T
.
L
i
a
n
d
T
.
W
u
,
“
M
o
d
e
l
i
n
g
n
o
n
l
i
n
e
a
r
f
l
u
t
t
e
r
b
e
h
a
v
i
o
r
o
f
l
o
n
g
-
sp
a
n
b
r
i
d
g
e
s
u
si
n
g
k
n
o
w
l
e
d
g
e
-
e
n
h
a
n
c
e
d
l
o
n
g
sh
o
r
t
-
t
e
r
m
m
e
mo
r
y
n
e
t
w
o
r
k
,
”
C
o
m
p
u
t
e
r
-
Ai
d
e
d
C
i
v
i
l
a
n
d
I
n
f
r
a
s
t
r
u
c
t
u
r
e
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
3
8
,
n
o
.
1
1
,
p
p
.
1
5
0
4
–
1
5
1
9
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
1
1
/
mi
c
e
.
1
3
0
0
0
.
[
2
5
]
A
.
I
.
Jo
n
y
a
n
d
A
.
K
.
B
.
A
r
n
o
b
,
“
A
l
o
n
g
s
h
o
r
t
-
t
e
r
m
mem
o
r
y
b
a
se
d
a
p
p
r
o
a
c
h
f
o
r
d
e
t
e
c
t
i
n
g
c
y
b
e
r
a
t
t
a
c
k
s i
n
I
o
T
u
s
i
n
g
C
I
C
-
I
o
T2
0
2
3
d
a
t
a
se
t
,
”
J
o
u
rn
a
l
o
f
Ed
g
e
C
o
m
p
u
t
i
n
g
,
v
o
l
.
3
,
n
o
.
1
,
p
p
.
2
8
–
4
2
,
2
0
2
4
,
d
o
i
:
1
0
.
5
5
0
5
6
/
j
e
c
.
6
4
8
.
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
-
8
7
0
8
E
n
h
a
n
cin
g
s
en
timen
t a
n
a
lysi
s
th
r
o
u
g
h
d
ee
p
la
ye
r
in
teg
r
a
ti
o
n
w
ith
…
(
P
a
r
u
l D
u
b
ey
)
957
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Pa
r
u
l
Dub
e
y
is
c
u
rre
n
t
ly
w
o
rk
i
n
g
a
s
a
n
a
ss
istan
t
p
r
o
fe
ss
o
r
in
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ie
n
c
e
a
n
d
E
n
g
i
n
e
e
rin
g
,
S
y
m
b
io
sis
I
n
stit
u
te
o
f
Te
c
h
n
o
l
o
g
y
,
Na
g
p
u
r
Ca
m
p
u
s
,
S
y
m
b
i
o
sis
In
tern
a
ti
o
n
a
l
(De
e
m
e
d
Un
iv
e
rsit
y
),
P
u
n
e
,
I
n
d
ia
.
Au
t
h
o
rs
h
i
g
h
e
st
d
e
g
re
e
is
th
e
d
e
g
re
e
o
f
m
a
ste
r’s
in
c
o
m
p
u
ter
e
n
g
i
n
e
e
rin
g
.
S
h
e
is
a
lso
a
P
h
.
D.
R
e
se
a
rc
h
sc
h
o
lar
in
Co
m
p
u
te
r
S
c
ien
c
e
a
n
d
En
g
in
e
e
ri
n
g
a
t
C.
V
.
Ra
m
a
n
Un
iv
e
rsity
,
Bil
a
sp
u
r
,
C
h
h
a
tt
isg
a
r
h
,
I
n
d
ia.
S
h
e
h
a
s
g
o
o
d
a
c
a
d
e
m
ic
a
n
d
re
se
a
rc
h
e
x
p
e
rien
c
e
in
v
a
rio
u
s
a
re
a
s
o
f
CS
E
a
n
d
IT.
S
h
e
h
a
s
tea
c
h
in
g
e
x
p
e
rien
c
e
o
f
f
iv
e
y
e
a
rs.
S
h
e
is
i
n
d
u
lg
e
d
i
n
re
se
a
rc
h
a
c
ti
v
it
ies
wh
ich
in
c
lu
d
e
s
b
o
o
k
c
h
a
p
ters
,
b
o
o
k
s,
c
o
n
fe
re
n
c
e
p
a
p
e
rs
a
n
d
jo
u
rn
a
l
a
rti
c
les
a
s
we
ll
.
S
h
e
h
a
s
1
7
I
n
d
ian
p
u
b
li
s
h
e
d
p
a
ten
ts
.
S
h
e
h
o
l
d
s
a
ro
u
n
d
1
5
p
u
b
li
c
a
ti
o
n
s
wh
i
c
h
a
re
p
a
rt
o
f
c
o
n
fe
re
n
c
e
s,
S
c
o
p
u
s
a
n
d
o
th
e
r
jo
u
rn
a
ls
a
s
we
ll
.
S
h
e
c
a
n
b
e
c
o
n
tac
t
e
d
o
n
e
m
a
il
:
d
u
b
e
y
p
a
ru
l
2
9
@
g
m
a
il
.
c
o
m
.
Pu
shk
a
r
Dub
e
y
is
c
u
rre
n
tl
y
wo
r
k
in
g
a
s
a
ss
istan
t
p
ro
fe
ss
o
r
a
n
d
h
e
a
d
i
n
De
p
a
rtme
n
t
o
f
M
a
n
a
g
e
m
e
n
t
a
t
P
a
n
d
it
S
u
n
d
a
rlal
S
h
a
rm
a
(Op
e
n
)
Un
i
v
e
rsity
Ch
h
a
tt
isg
a
r
h
Bil
a
sp
u
r.
He
is
a
g
o
l
d
m
e
d
a
li
st
in
M
a
ste
r
o
f
B
u
sin
e
ss
Ad
m
i
n
istr
a
ti
o
n
(M
BA)
a
n
d
P
h
.
D.
i
n
h
u
m
a
n
re
so
u
rc
e
m
a
n
a
g
e
m
e
n
t
.
He
h
a
s
p
u
b
li
s
h
e
d
m
o
re
t
h
a
n
7
0
re
se
a
rc
h
p
a
p
e
rs
i
n
re
p
u
ted
jo
u
r
n
a
ls
su
c
h
a
s
Eme
ra
ld
,
Ta
y
lo
r
a
n
d
F
ra
n
c
is,
S
p
r
in
g
e
r.
His
1
2
a
rt
icle
s
a
re
in
d
e
x
e
d
i
n
S
c
o
p
u
s.
He
h
a
s a
lso
a
tt
e
n
d
e
d
se
v
e
ra
l
in
ter
n
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
s in
c
lu
d
i
n
g
II
M
lev
e
l
c
o
n
fe
re
n
c
e
s.
Ha
v
i
n
g
su
c
c
e
ss
fu
ll
y
g
u
id
e
d
6
P
h
.
D.
sc
h
o
lars
,
Dr.
Du
b
e
y
is
a
lso
a
c
c
re
d
it
e
d
with
0
7
p
u
b
li
sh
e
d
p
a
ten
ts.
He
h
a
s
a
lso
a
c
c
o
m
p
li
sh
e
d
0
5
re
s
e
a
rc
h
p
ro
jec
ts
in
c
lu
d
in
g
0
3
sp
o
n
so
re
d
b
y
I
n
d
ian
Co
u
n
c
il
o
f
S
o
c
ial
S
c
ie
n
c
e
Re
se
a
rc
h
(ICS
S
R
)
Ne
w
De
lh
i.
Ha
v
i
n
g
sp
e
c
ialize
d
in
sta
ti
stica
l
s
o
ftwa
re
s
fo
r
d
a
ta
a
n
a
ly
sis,
h
e
h
a
s
d
e
li
v
e
re
d
se
v
e
ra
l
lec
tu
re
s
o
n
S
P
S
S
,
AM
OS
a
n
d
o
th
e
rs.
He
is
a
ls
o
p
u
rsu
i
n
g
th
e
h
ig
h
e
st
a
c
a
d
e
m
ic
d
e
g
re
e
fo
r
re
se
a
rc
h
i.
e
.
,
d
o
c
to
r
o
f
letters
in
t
h
e
a
re
a
o
f
a
p
p
li
c
a
ti
o
n
o
f
S
h
r
ima
d
Bh
a
g
wa
d
G
e
e
ta
in
to
m
a
n
a
g
e
m
e
n
t
p
ra
c
ti
c
e
s
.
He
c
a
n
b
e
c
o
n
tac
ted
o
n
e
m
a
il
:
d
rd
u
b
e
y
p
k
a
g
@
g
m
a
il
.
c
o
m
.
H
ites
h
G
e
h
a
n
i
(P
h
.
D
.
,
e
lec
tro
n
ics
e
n
g
in
e
e
rin
g
),
a
ss
istan
t
p
r
o
f
e
ss
o
r
a
t
S
c
h
o
o
l
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
in
e
e
rin
g
,
S
h
ri
Ra
m
d
e
o
b
a
b
a
C
o
l
leg
e
o
f
En
g
in
e
e
rin
g
a
n
d
M
a
n
a
g
e
m
e
n
t,
Ra
m
d
e
o
b
a
b
a
U
n
iv
e
rsity
,
Na
g
p
u
r
,
In
d
ia
is
a
re
n
o
wn
e
d
a
c
a
d
e
m
ic
p
e
rso
n
a
li
t
y
a
n
d
h
a
s
a
c
h
iev
e
d
c
o
n
ti
n
u
e
d
s
u
c
c
e
ss
in
h
is
r
o
le
a
s
a
sc
h
o
las
ti
c
i
n
tell
e
c
tu
a
l
lea
d
e
r
in
d
iv
e
rsifie
d
a
re
a
s.
He
h
o
ld
s
a
p
a
ten
tab
le
p
ro
fil
e
with
b
ril
li
a
n
t
a
c
a
d
e
m
ic
a
c
h
iev
e
m
e
n
ts.
His
tu
to
rial
e
x
c
e
ll
e
n
c
e
,
e
sp
e
c
ially
th
e
v
a
rio
u
s
in
n
o
v
a
ti
v
e
tea
c
h
in
g
/l
e
a
rn
in
g
p
ra
c
ti
c
e
s
a
d
o
p
ted
b
y
h
im
m
a
rk
s
h
im
a
s
a
d
isti
n
g
u
ish
e
d
lu
m
i
n
a
ry
in
a
c
a
d
e
m
ic
c
ircle
s.
Re
c
ip
ien
t
o
f
Be
st
Tea
c
h
e
r
Aw
a
rd
,
NPT
EL
S
tar
a
n
d
P
h.
D
.
P
e
rsu
i
n
g
.
Hites
h
G
e
h
a
n
i
h
a
s
2
p
a
ten
ts
(P
u
b
li
sh
e
d
),
1
c
o
p
y
ri
g
h
t,
3
b
o
o
k
c
h
a
p
ters
,
6
re
se
a
rc
h
p
a
p
e
rs
(p
u
b
l
ish
e
d
i
n
i
n
tern
a
ti
o
n
a
l
j
o
u
r
n
a
ls,
c
o
n
fe
re
n
c
e
s).
He
h
a
s
wo
r
k
e
d
a
s
a
re
v
iew
e
r
on
m
a
n
y
in
tern
a
ti
o
n
a
l
jo
u
r
n
a
ls
a
n
d
c
o
n
fe
re
n
c
e
s.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
g
e
h
a
n
i
h
s
@rk
n
e
c
.
e
d
u
.
Evaluation Warning : The document was created with Spire.PDF for Python.