I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
, pp.
4747
~
4756
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
13
.i
4
.pp
4747
-
4756
4747
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
S
e
n
t
i
m
e
n
t
an
al
ys
i
s
of
st
u
d
e
n
t
’
s
c
om
m
e
n
t
s
u
si
n
g
l
on
g s
h
or
t
-
t
e
r
m
m
e
m
or
y
w
i
t
h
m
u
l
t
i
-
h
e
ad
at
t
e
n
t
i
on
B
h
avan
a
P
.
B
h
agat
1
,
S
h
e
e
t
al
S.
D
h
an
d
e
-
D
an
d
ge
2
,
S
an
d
e
e
p
R
.
R
au
t
3
1
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
E
ngi
ne
e
r
i
ng
,
G
ove
r
nm
e
nt
P
ol
yt
e
c
hni
c
,
Y
a
va
t
m
a
l
,
I
n
di
a
2
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd
E
ngi
ne
e
r
i
ng
,
S
i
pna
C
ol
l
e
ge
of
E
ngi
ne
e
r
i
ng
a
nd
T
e
c
hnol
ogy,
A
m
r
a
va
t
i
,
I
ndi
a
3
IB
C
om
put
e
r
S
c
i
e
nc
e
,
S
ha
ngha
i
Q
i
ba
o
D
w
i
ght
H
i
gh
S
c
hool
,
S
ha
ngha
i
,
C
hi
na
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
F
e
b 11,
2024
R
e
vi
s
e
d
A
pr
9, 2024
A
c
c
e
pt
e
d
J
un 14, 2024
Classroom
teaching
is
a
viable
and
effective
approach
for
enhancing
s
tudent
learning
and
promoting
engagement
in
the
education
al
process.
The
o
pinions
of
students
play
a
vital
role
in
the
evaluatio
n
of
teacher
s.
This
paper
p
resents
a
comprehens
ive
overview
of
sentiment
analysis
techniques
based
on
recent
research
and
subsequently
explores
machine
learning,
i.e.,
e
n
semble
classifi
ers
,
deep
learning
,
long
short
-
term
memory
(LSTM)
,
convol
utional
neural
network
(CNN),
LSTM
with
single
a
ttention,
LSTM
with
mult
i
-
head
attenti
on
,
and
feature
extractio
n
techniques
(TFidfVector
and
Word2V
ec),
in
the
context
of
sentiment
analysis
over
student
opinion
datasets,
i.
e.,
the
Vietnamese
student
feedback
co
rpus,
as
well
as
data
collected
from
a
final
-
year
student'
s
comment
in
2023.
Further
,
the
Vietnamese
student
fe
edback
corpus
is
translated
to
English
and
pre
-
processed
with
the
pr
oposed
framework,
which
yields
interesting
facts
about
the
capabilit
ie
s
and
deficiencies
of
different
methods.
In
this
paper,
we
conducted
exper
iments
with
ensemble
classifi
ers,
LSTM
and
CNN,
LSTM
with
single
a
ttenti
on
,
and
LSTM
with
multi
-
head
attention
.
We
conclude
that
LSTM
with
mult
i
-
head
attenti
on
produces
an
accuracy
result
of
95.57%,
which
outperfo
rm
as
compare
to
other
three
baseline
methods
and
earlier
studies.
K
e
y
w
o
r
d
s
:
C
on
vol
ut
i
on
a
l
ne
ur
a
l
ne
two
r
k
D
e
e
p
l
e
a
r
ni
ng
L
e
a
r
n
in
g
l
ong
s
hor
t
-
t
e
r
m
m
e
m
or
y
O
pi
ni
on
m
in
in
g
S
e
nt
im
e
nt
a
na
ly
s
is
This
is
an
open
access
article
under
the
CC
BY
-
SA
license.
C
or
r
e
s
pon
di
n
g
A
u
th
or
:
B
ha
va
na
P
.
B
ha
g
a
t
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
E
ngi
ne
e
r
in
g
,
G
ove
r
nm
e
nt
P
ol
yt
e
c
hni
c
Y
a
va
tm
a
l
D
ha
m
a
nga
on R
oa
d, P
im
pa
lg
a
on, Y
a
va
tm
a
l,
M
a
h
a
r
a
s
ht
r
a
4450
01, I
ndi
a
E
m
a
il
:
bha
va
na
.bha
ga
t3
0475@
gm
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
C
la
s
s
r
oom
te
a
c
hi
ng
is
a
vi
a
bl
e
a
nd
e
f
f
ic
ie
nt
a
ppr
oa
c
h
f
or
e
nha
nc
in
g
s
tu
de
nt
le
a
r
ni
ng
a
nd
pr
om
ot
in
g
e
nga
ge
m
e
nt
in
th
e
e
duc
a
ti
ona
l
pr
oc
e
s
s
.
T
he
m
a
in
obj
e
c
ti
ve
of
te
a
c
hi
ng
is
to
im
pr
ove
s
tu
de
nt
s
le
a
r
ni
ng
a
bi
li
ti
e
s
a
nd
pr
om
ot
e
e
f
f
e
c
ti
ve
knowle
dge
in
th
e
m
.
T
e
a
c
hi
ng
e
f
f
e
c
ti
ve
ne
s
s
a
lwa
y
s
de
pe
nds
on
te
a
c
h
e
r
s
’
s
ubj
e
c
t
knowle
dge
a
nd
p
e
da
gogi
c
a
l
a
ppr
oa
c
h
to
te
a
c
hi
ng.
S
a
nge
e
th
a
a
n
d
P
r
a
bha
[
1]
s
ta
te
d
th
a
t
t
he
qua
li
ty
of
te
a
c
hi
ng
r
e
li
e
s
on
pe
da
gogi
c
a
l
s
ki
ll
s
,
de
f
in
e
d
obj
e
c
ti
ve
s
,
c
la
r
it
y,
a
nd
s
ubj
e
c
t
e
xpe
r
ti
s
e
.
E
va
lu
a
ti
ng
te
a
c
hi
ng
e
f
f
e
c
ti
ve
n
e
s
s
can
be
m
e
a
s
ur
e
d
by
c
ol
le
c
ti
ng
s
tu
de
nt
f
e
e
dba
c
k,
a
id
in
g
in
th
e
r
e
f
in
e
m
e
nt
of
s
tr
a
te
gi
e
s
f
or
im
pr
ovi
ng
te
a
c
hi
ng
e
xc
e
ll
e
nc
e
.
S
tu
de
nt
’
s
opi
ni
ons
a
r
e
a
lwa
y
s
de
e
p
a
nd
uni
que
.
T
h
e
y
gi
ve
an
in
-
de
pt
h
vi
e
w
of
how
th
e
ir
te
a
c
he
r
s
e
nc
our
a
ge
a
nd
e
duc
a
te
.
T
h
e
te
a
c
h
e
r
s
ge
t
oppor
tu
ni
ti
e
s
to
f
e
e
l
a
nd
va
lu
e
th
e
im
por
ta
nc
e
of
te
a
c
hi
ng
f
r
om
s
tu
de
nt
’
s
opi
ni
ons
.
S
tu
de
nt
f
e
e
dba
c
k
he
lp
s
th
e
te
a
c
he
r
s
f
or
m
ul
a
te
be
tt
e
r
de
c
is
io
ns
on
how
to
im
pr
ove
th
e
qua
li
ty
of
te
a
c
hi
ng.
T
hus
,
s
tu
de
nt
s
a
r
e
one
of
th
e
m
o
s
t
im
por
ta
nt
pa
r
ti
c
ip
a
nt
s
in
e
ve
r
y
e
duc
a
ti
ona
l
in
s
ti
tu
te
.
R
e
s
e
a
r
c
h
e
r
s
ha
v
e
de
v
e
lo
pe
d
a
num
be
r
of
m
e
th
od
s
a
nd
to
ol
s
to
m
e
a
s
ur
e
s
tu
de
nt
opi
ni
ons
.
In
th
is
e
r
a
,
s
e
nt
im
e
nt
a
na
ly
s
is
is
one
of
an
e
m
e
r
gi
ng
a
r
e
a
in
th
e
f
ie
ld
s
of
na
tu
r
a
l
la
ngua
ge
pr
oc
e
s
s
in
g
(
N
L
P
)
a
nd
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
.
It
in
vol
ve
s
th
e
us
e
of
c
om
put
a
ti
ona
l
m
e
th
ods
to
de
te
r
m
in
e
a
nd
id
e
nt
if
y
th
e
s
e
nt
im
e
nt
s
or
e
m
ot
io
na
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4747
-
4756
4748
a
tt
it
ude
s
e
xpr
e
s
s
e
d
in
te
xt
,
e
it
he
r
in
s
tu
de
nt
’
s
f
e
e
dba
c
k,
ne
w
s
a
r
ti
c
le
s
,
s
oc
ia
l
m
e
di
a
pos
t
s
,
c
us
to
m
e
r
r
e
vi
e
w
s
,
a
nd
ot
he
r
f
or
m
s
of
w
r
it
te
n
c
om
m
e
nt
s
.
It
is
us
e
d
as
a
to
ol
to
a
na
ly
z
e
a
nd
e
va
lu
a
te
us
e
r
s
’
c
om
m
e
nt
s
a
ut
om
a
ti
c
a
ll
y.
A
c
c
or
di
ng
to
Z
hou
a
nd
Y
e
[
2]
,
s
e
nt
im
e
nt
a
n
a
ly
s
i
s
ha
s
be
c
om
e
on
e
of
th
e
m
o
s
t
s
ig
ni
f
ic
a
nt
to
pi
c
s
in
e
duc
a
ti
on
r
e
s
e
a
r
c
h,
a
nd
th
e
r
e
a
r
e
s
e
ve
r
a
l
im
pl
ic
a
ti
ons
a
nd
r
e
s
e
a
r
c
h
is
s
u
e
s
di
s
c
u
s
s
e
d
in
th
e
r
e
s
e
a
r
c
h
r
e
la
te
d
to
th
e
us
e
of
s
e
nt
im
e
nt
a
na
ly
s
is
in
e
duc
a
ti
on.
T
hi
s
s
tu
dy
a
dd
r
e
s
s
e
s
th
e
us
e
of
m
a
c
hi
ne
le
a
r
ni
ng
a
nd
de
e
p
le
a
r
ni
ng
te
c
hnol
ogi
e
s
in
e
duc
a
ti
on
to
c
la
s
s
if
y
s
e
nt
im
e
nt
s
f
r
om
s
tu
de
nt
c
om
m
e
nt
s
us
in
g
f
our
di
f
f
e
r
e
nt
c
la
s
s
if
ic
a
ti
on
m
ode
l
s
a
nd
c
om
pa
r
e
s
th
e
ir
pe
r
f
or
m
a
nc
e
w
it
h
e
xi
s
ti
ng
s
tu
di
e
s
.
F
ur
th
e
r
m
or
e
,
th
is
s
tu
dy
a
im
s
to
e
nc
our
a
ge
r
e
s
e
a
r
c
he
r
s
to
f
oc
us
on
de
e
p
le
a
r
ni
ng
a
ppr
oa
c
he
s
to
im
pr
ove
th
e
pe
r
f
or
m
a
nc
e
of
c
la
s
s
if
ie
r
m
ode
l
s
f
or
s
e
nt
im
e
nt
c
la
s
s
if
ic
a
ti
on.
S
e
c
ti
on
2
w
il
l
br
ie
f
ly
e
xpl
a
in
r
e
la
te
d
w
or
k.
S
e
c
ti
on
3
be
g
in
s
w
it
h
an
out
li
ne
of
th
e
de
ta
il
e
d
m
e
th
odol
ogy
of
th
e
f
our
a
ppr
oa
c
he
s
us
e
d
in
our
pr
opos
e
d
w
or
k.
S
e
c
ti
on
4
pr
e
s
e
nt
s
e
xp
e
r
im
e
nt
a
l
r
e
s
ul
ts
a
nd
pe
r
f
or
m
a
nc
e
e
va
lu
a
ti
ons
a
nd
c
om
pa
r
is
on
s
w
it
h
c
om
pe
ti
ng
m
e
t
hods
.
S
e
c
ti
on
5
w
il
l
di
s
c
us
s
th
e
c
onc
lu
s
io
n
a
nd
f
ut
ur
e
s
c
ope
.
2.
R
E
L
A
T
E
D
WO
R
K
In
th
is
w
or
k,
we
f
oc
us
on
s
e
nt
im
e
nt
a
n
a
ly
s
is
of
s
tu
d
e
nt
c
om
m
e
nt
s
in
th
e
e
duc
a
ti
on
dom
a
in
.
S
om
e
s
tu
di
e
s
[
3
]
,
[
5]
-
[
8
]
us
e
d
a
le
xi
c
on
-
ba
s
e
d
a
ppr
oa
c
h;
s
om
e
s
tu
di
e
s
[
5]
,
[
6]
,
[
8]
–
[
10
]
di
d
not
e
va
lu
a
te
r
e
s
ul
ts
us
in
g
e
va
lu
a
ti
on
m
e
tr
ic
s
.
A
ls
o,
th
e
r
e
is
a
tr
e
nd
to
u
s
e
hybr
id
m
ode
l
s
to
a
c
hi
e
ve
be
tt
e
r
r
e
s
ul
ts
in
th
e
s
e
nt
im
e
nt
a
n
a
ly
s
is
pr
oc
e
s
s
.
S
om
e
m
ode
ls
do
not
pr
ovi
de
gr
a
phi
c
a
l
r
e
s
ul
ts
f
or
s
e
nt
i
m
e
nt
a
na
ly
s
i
s
,
nor
do
th
e
y
pr
ovi
de
a
ny
f
a
c
t
s
or
s
tu
di
e
s
to
e
va
lu
a
te
th
e
ir
vi
s
ua
l
us
a
bi
li
ty
.
M
a
ny
r
e
s
e
a
r
c
he
r
s
in
[
11]
–
[
15]
f
oc
us
on
e
-
le
a
r
ni
ng.
L
im
i
te
d
s
tu
di
e
s
us
e
th
e
pr
im
a
r
y
c
or
pus
,
m
a
c
hi
ne
le
a
r
ni
ng
a
nd
de
e
p
le
a
r
ni
ng
m
e
th
ods
,
a
lo
ng
w
it
h
e
va
lu
a
ti
on
m
e
tr
ic
s
[
16]
.
S
e
ve
r
a
l
r
e
s
e
a
r
c
he
r
s
in
[4
],
[
9]
,
[
10]
,
[
17
]
–
[
25]
e
m
pl
oye
d
m
a
c
hi
ne
le
a
r
ni
ng
-
ba
s
e
d
m
e
th
odol
ogi
e
s
.
T
he
m
ode
l
is
in
it
ia
ll
y
tr
a
in
e
d
w
it
h
known
in
put
s
a
nd
out
put
s
,
in
or
de
r
to
f
unc
ti
on
w
it
h
la
te
r
unknown
da
ta
.
As
s
ta
te
d
in
B
ha
ga
t
a
nd
D
a
ndge
[
16]
,
s
om
e
a
lg
or
it
hm
s
,
pe
r
f
or
m
m
or
e
a
c
c
ur
a
te
ly
th
a
n
ot
he
r
s
w
he
n
u
s
e
d
w
it
h
la
r
ge
c
or
por
a
.
In
s
om
e
s
tu
di
e
s
[
21]
,
[
26
]
–
[
29]
hybr
id
m
e
th
ods
w
e
r
e
us
e
d
to
im
pr
ove
th
e
a
c
c
ur
a
c
y
of
th
e
ir
w
or
k.
L
a
s
tl
y,
it
is
s
e
e
n
th
a
t
s
e
v
e
r
a
l
r
e
c
e
nt
s
tu
di
e
s
[
1]
,
[
12]
,
[
30]
–
[
34]
us
e
d
de
e
p
le
a
r
ni
ng
m
ode
ls
as
one
of
th
e
dom
in
a
nt
m
ode
ls
in
th
e
e
duc
a
ti
on
dom
a
in
w
it
h
hi
ghe
r
pe
r
f
or
m
a
nc
e
.
M
a
ny
s
tu
di
e
s
u
s
e
d
th
e
s
tu
de
nt
f
e
e
dba
c
k
da
ta
,
w
hi
c
h
is
a
va
il
a
bl
e
onl
in
e
to
e
xpl
or
e
a
nd
is
hum
a
n
-
a
nnot
a
t
e
d
in
V
ie
tn
a
m
e
s
e
,
as
in
N
guye
n
et
al
.
[
30]
.
T
he
pr
opos
e
d
s
y
s
te
m
im
pl
e
m
e
nt
s
m
a
c
hi
ne
l
e
a
r
ni
ng
a
nd
d
e
e
p
le
a
r
ni
ng
m
ode
ls
a
nd
c
om
pa
r
e
s
th
e
ir
r
e
s
ul
ts
,
to
r
e
s
ol
ve
th
e
pr
obl
e
m
of
te
xt
c
la
s
s
if
ic
a
ti
on
a
nd
c
om
pa
r
e
th
e
ir
a
c
c
ur
a
c
y.
A
ls
o
us
e
d
to
id
e
nt
if
y
th
e
be
s
t
m
ode
l
ba
s
e
d
on
e
va
lu
a
ti
on
m
e
tr
ic
s
f
or
s
tu
de
nt
’
s
c
om
m
e
nt
s
.
T
he
c
om
pa
r
a
ti
ve
a
na
ly
s
i
s
s
ta
te
s
th
a
t
th
e
de
e
p
le
a
r
ni
ng
m
e
th
ods
a
c
hi
e
ve
be
tt
e
r
th
a
n
th
e
r
e
s
t
of
th
e
th
r
e
e
ba
s
e
li
ne
m
ode
ls
.
T
he
pr
opos
e
d
s
ys
te
m
u
s
e
s
two
da
ta
s
e
ts
f
or
tr
a
in
in
g
a
nd
te
s
ti
ng,
i.
e
.,
th
e
V
ie
tn
a
m
e
s
e
s
tu
d
e
nt
f
e
e
dba
c
k
c
or
pus
a
nd
pr
im
a
r
y
da
ta
s
e
t
c
ol
le
c
te
d
f
r
om
a
f
in
a
l
-
ye
a
r
s
tu
de
nt
'
s
c
om
m
e
nt
.
T
he
e
duc
a
ti
on
f
ie
ld
is
e
x
pe
r
ie
nc
in
g
a
s
ig
ni
f
ic
a
nt
a
nd
pr
of
ound
c
ha
nge
.
W
it
hi
n
th
is
s
e
c
to
r
,
s
e
nt
im
e
nt
a
na
ly
s
is
hol
ds
gr
e
a
t
s
ig
ni
f
ic
a
nc
e
f
or
ke
y
s
ta
ke
hol
de
r
s
,
s
uc
h
as
s
tu
de
nt
s
a
nd
e
duc
a
ti
ona
l
in
s
ti
tu
ti
ons
.
A
v
a
r
ie
ty
of
m
a
c
hi
ne
le
a
r
ni
ng
a
nd
d
e
e
p
le
a
r
ni
ng
te
c
hni
que
s
w
il
l
be
e
m
pl
oye
d
to
e
va
lu
a
te
s
tu
de
nt
’
s
f
e
e
db
a
c
k.
A
c
c
or
di
ng
to
S
a
nge
e
th
a
a
nd
P
r
a
bha
[
1]
,
de
e
p
le
a
r
ni
ng
is
popula
r
c
ur
r
e
nt
ly
due
to
its
a
ut
om
a
ti
c
le
a
r
ni
ng
c
a
pa
bi
li
ti
e
s
a
nd
a
ls
o
it
de
a
ls
w
it
h
lo
ng
s
e
que
nc
e
s
r
a
th
e
r
th
a
n
s
hor
t
s
e
que
nc
e
s
of
s
e
nt
e
nc
e
s
.
It
is
obs
e
r
ve
d
th
a
t
in
pr
e
vi
ous
w
or
k,
a
V
ie
tn
a
m
e
s
e
s
tu
de
nt
f
e
e
dba
c
k
c
or
pus
c
ons
is
ti
ng
of
16,175
f
e
e
dba
c
k
s
e
nt
e
nc
e
s
w
a
s
us
e
d
w
it
h
th
e
lo
ng s
hor
t
-
te
r
m
m
e
m
or
y (
L
S
T
M
)
w
it
h
m
ul
ti
-
he
a
d a
tt
e
nt
io
n
m
ode
l
w
he
r
e
a
s
in
our
pr
opos
e
d
m
ode
ls
,
bot
h
m
a
c
hi
ne
le
a
r
ni
ng
a
nd
de
e
p
le
a
r
ni
ng
m
ode
ls
w
e
r
e
im
pl
e
m
e
nt
e
d,
a
nd
te
s
ti
ng
w
a
s
pe
r
f
or
m
e
d
on
bot
h
th
e
V
ie
tn
a
m
e
s
e
s
tu
de
nt
f
e
e
dba
c
k
da
ta
s
e
t
as
w
e
ll
as
un
s
e
e
n
da
ta
c
ol
le
c
te
d
f
r
om
f
in
a
l
ye
a
r
s
tu
de
nt
f
e
e
dba
c
k.
A
ls
o,
c
om
pa
r
a
ti
ve
a
na
ly
s
i
s
of
d
e
e
p
le
a
r
ni
ng
a
nd
m
a
c
hi
ne
le
a
r
ni
ng
m
e
th
ods
is
c
ondu
c
te
d
in
our
e
xpe
r
im
e
nt
s
to
id
e
nt
if
y
th
e
be
s
t
m
e
th
od
s
in
te
r
m
s
of
a
c
c
ur
a
c
y
pe
r
f
or
m
a
nc
e
.
In
th
is
w
or
k,
we
f
ir
s
t
c
onve
r
te
d
th
e
V
ie
tn
a
m
e
s
e
da
ta
s
e
t
to
E
ngl
is
h,
c
ont
a
in
in
g
21,561
s
e
nt
e
nc
e
s
a
nd
tr
a
in
e
d
it
us
in
g
a
m
a
c
hi
ne
le
a
r
ni
ng
a
nd
de
e
p
le
a
r
ni
ng
m
ode
l.
We
ut
il
iz
e
bot
h
th
e
V
ie
tn
a
m
e
s
e
a
nd
th
e
pr
i
m
a
r
y
da
ta
s
e
t,
w
hi
c
h
c
ont
a
in
520
c
om
m
e
nt
s
f
r
om
f
in
a
l
-
ye
a
r
s
tu
de
nt
s
ga
th
e
r
e
d
in
2023
f
or
m
ode
l
te
s
ti
ng.
3.
M
E
T
H
O
D
O
L
O
G
Y
T
he
m
o
de
l
s
pr
opo
s
e
d
in
th
i
s
s
t
udy
in
c
lu
de
s
e
v
e
r
a
l
s
te
p
s
f
or
s
e
nt
i
m
e
nt
a
n
a
ly
s
is
.
F
ir
s
t,
pr
e
pr
o
c
e
s
s
in
g
a
n
d
va
r
io
u
s
w
or
d
e
m
be
d
di
ng
te
c
hni
qu
e
s
w
e
r
e
us
e
d
to
g
e
ne
r
a
te
a
ve
c
to
r
r
e
pr
e
s
e
nt
a
ti
on.
T
h
e
n,
th
e
s
e
v
e
c
to
r
s
a
r
e
us
e
d
f
or
tr
a
in
i
ng
di
f
f
e
r
e
nt
c
la
s
s
if
ie
r
m
od
e
ls
f
or
s
e
nt
im
e
nt
a
n
a
ly
s
is
.
F
in
a
ll
y,
th
e
r
e
s
ul
ts
obt
a
in
e
d
w
it
h
th
e
pr
opo
s
e
d
m
ode
l
a
r
e
c
om
p
a
r
e
d
w
it
h
th
o
s
e
o
bt
a
in
e
d
w
it
h
th
e
ot
h
e
r
th
r
e
e
ba
s
e
li
ne
m
o
de
l
s
in
th
i
s
s
tu
dy
a
n
d
pr
e
vi
o
us
s
tu
di
e
s
.
F
ig
ur
e
1
il
lu
s
tr
a
te
s
th
e
ov
e
r
a
ll
m
e
th
odo
lo
gy
of
th
e
pr
op
os
e
d
s
e
nt
i
m
e
nt
c
la
s
s
if
ic
a
ti
on
m
od
e
l
s
,
w
hi
c
h
us
e
d
if
f
e
r
e
nt
w
or
d
e
m
be
ddi
ng
t
e
c
hn
iq
ue
s
a
nd
e
va
l
ua
ti
o
n
m
e
tr
i
c
s
to
m
e
a
s
ur
e
t
he
pe
r
f
or
m
a
n
c
e
of
th
e
s
y
s
te
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
Se
nt
ime
nt
analy
s
is
of
s
tu
d
e
nt
’
s
c
om
m
e
nt
s
us
in
g l
ong s
ho
r
t
-
te
r
m
m
e
m
or
y
w
it
h
…
(
B
hav
ana P
. B
hagat
)
4749
F
ig
ur
e
1.
T
he
m
e
th
odol
ogy
us
e
d
f
or
p
r
opos
e
d
m
ode
ls
3.1.
D
at
as
e
t
E
va
lu
a
te
d
m
ode
l'
s
pe
r
f
or
m
a
nc
e
on
two
da
ta
c
or
pus
e
s
,
t
he
V
ie
tn
a
m
e
s
e
s
tu
d
e
nt
f
e
e
dba
c
k
c
or
pus
by
[
30]
a
nd
pr
i
m
a
r
y
da
ta
c
ol
le
c
te
d
on
our
ow
n
f
r
om
a
f
in
a
l
-
ye
a
r
s
tu
de
nt
'
s
c
om
m
e
nt
.
In
our
m
ode
l,
21,561
r
ow
s
of
c
om
m
e
nt
s
w
e
r
e
us
e
d
f
or
s
e
nt
im
e
nt
a
na
ly
s
is
a
nd
520
s
tu
de
nt
’
s
c
om
m
e
nt
s
f
or
th
e
pr
e
di
c
ti
on
of
uns
e
e
n
da
ta
.
We
a
ll
oc
a
te
75%
of
each
da
ta
s
e
t
c
la
s
s
f
or
tr
a
in
in
g
a
nd
25%
f
or
te
s
ti
ng.
3.2.
P
r
e
p
r
oc
e
s
s
in
g
an
d
w
or
d
e
m
b
e
d
d
in
gs
On
th
is
w
or
k,
th
e
s
tu
de
nt
f
e
e
dba
c
k
da
ta
s
e
t
is
us
e
d.
It
ha
s
s
om
e
pr
im
a
l
da
ta
,
s
uc
h
as
a
bbr
e
vi
a
ti
on
s
,
s
pe
ll
in
g
m
is
ta
ke
s
,
e
m
ot
io
ns
,
s
ig
n
s
,
r
e
pe
a
te
d
a
nd
unw
a
nt
e
d
c
ha
r
a
c
te
r
s
,
num
be
r
s
,
a
nd
s
ym
bol
s
.
To
c
le
a
n
th
e
te
xt
da
ta
,
a
te
xt
pr
e
pr
oc
e
s
s
in
g
m
e
th
od
is
u
s
e
d.
S
te
m
m
in
g
is
us
e
d
f
or
r
e
duc
in
g
a
w
or
d
to
its
s
te
m
,
w
hi
c
h
a
f
f
ix
e
s
to
pr
e
f
ix
e
s
a
nd
s
uf
f
ix
e
s
or
to
th
e
r
oot
s
of
w
o
r
ds
c
a
ll
e
d
"
le
m
m
a
s
"
.
It
is
us
e
d
to
nor
m
a
li
z
e
th
e
te
xt
a
nd
m
a
ke
it
e
a
s
ie
r
to
pr
oc
e
s
s
.
L
e
m
m
a
ti
z
a
ti
on
a
n
a
ly
z
e
s
th
e
c
ont
e
xt
a
nd
c
onv
e
r
ts
th
e
w
or
d
to
its
m
e
a
ni
ngf
ul
f
or
m
,
w
hi
c
h
is
c
a
ll
e
d
a
le
m
m
a
.
T
he
ba
g
of
w
or
ds
m
ode
l,
w
hi
c
h
is
c
a
ll
e
d
B
oW
,
is
us
e
d
f
or
e
xt
r
a
c
ti
ng
f
e
a
tu
r
e
s
f
r
om
th
e
te
xt
to
be
us
e
d
in
m
ode
li
ng.
A
to
ke
ni
z
e
r
is
us
e
d
to
di
vi
de
pa
r
a
gr
a
phs
a
n
d
s
e
nt
e
nc
e
s
in
to
s
m
a
ll
e
r
pa
r
ts
th
a
t
can
be
e
a
s
il
y
a
s
s
ig
ne
d
m
e
a
ni
ng.
D
e
e
p
le
a
r
ni
ng
a
lg
or
it
hm
s
a
c
c
e
pt
a
ve
c
to
r
r
e
pr
e
s
e
nt
a
ti
on
of
num
be
r
s
a
s
in
put
.
W
or
d
e
m
be
ddi
ng
is
a
te
c
hni
que
f
or
pe
r
f
or
m
in
g
th
e
r
e
pr
e
s
e
nt
a
ti
on
of
w
or
ds
in
a
v
e
c
to
r
s
pa
c
e
.
I
t
is
a
te
c
hni
que
f
or
tr
a
ns
la
ti
ng
w
or
ds
in
to
a
m
a
th
e
m
a
ti
c
a
l
dom
a
in
w
he
r
e
num
be
r
s
c
a
pt
ur
e
th
e
s
e
m
a
nt
ic
s
of
th
e
w
or
ds
.
A
c
c
or
di
ng
to
P
a
c
ol
a
nd P
a
la
oa
g
[
4]
, t
he
pr
oc
e
s
s
i
s
done
u
s
in
g m
il
li
ons
of
l
e
a
r
ni
ng a
nd ma
c
hi
ne
phr
a
s
e
s
. W
or
d2ve
c
i
s
b
a
s
e
d
on
a
n
un
s
upe
r
vi
s
e
d
m
a
c
hi
ne
le
a
r
ni
ng
m
e
th
od
ba
s
e
d
on
a
la
r
ge
a
m
ount
of
te
xt
to
a
c
hi
e
ve
s
ynt
a
c
ti
c
a
nd
s
e
m
a
nt
ic
r
e
pr
e
s
e
nt
a
ti
on
in
[
35]
.
I
t
is
us
e
d
w
it
h
th
e
L
S
T
M
a
nd
c
onvolut
io
n
a
l
ne
ur
a
l
ne
twor
k
(
C
N
N
)
m
ode
ls
,
a
nd
th
e
m
os
t
c
om
m
on
e
nc
odi
ng
m
e
th
od
s
uc
h
a
s
T
F
I
D
F
ve
c
to
r
iz
e
r
,
ha
s
be
e
n
u
s
e
d
w
it
h
th
e
r
e
s
t
of
our
th
r
e
e
pr
opos
e
d
m
ode
ls
.
3.3.
C
la
s
s
if
i
c
at
io
n
t
e
c
h
n
iq
u
e
s
In
th
is
pr
opos
e
d
w
or
k,
f
our
di
f
f
e
r
e
nt
s
e
nt
im
e
nt
c
la
s
s
if
ic
a
ti
on
a
ppr
oa
c
he
s
w
e
r
e
us
e
d
to
c
la
s
s
if
y
th
e
s
e
nt
im
e
nt
s
of
s
tu
de
nt
s
f
r
om
s
tu
de
nt
’
s
c
om
m
e
nt
s
.
V
a
r
io
us
w
or
d
e
m
be
ddi
ngs
te
c
hni
que
s
w
e
r
e
us
e
d
f
or
f
e
a
tu
r
e
e
xt
r
a
c
ti
on.
T
he
s
e
m
ode
ls
w
e
r
e
bui
lt
,
in
c
lu
di
ng
m
a
c
hi
ne
le
a
r
ni
ng
s
uc
h
a
s
e
ns
e
m
bl
e
c
la
s
s
if
ie
r
a
nd
de
e
p
le
a
r
ni
ng
s
uc
h a
s
C
N
N
a
nd L
S
T
M
, L
S
T
M
w
it
h a
tt
e
nt
io
n l
a
ye
r
,
a
nd L
S
T
M
w
it
h
m
ul
ti
-
he
a
d a
tt
e
nt
io
n l
a
ye
r
.
3.3.1
.
E
n
s
e
m
b
le
c
la
s
s
i
f
ie
r
T
he
f
ir
s
t
a
ppr
oa
c
h
us
e
s
an
e
n
s
e
m
bl
e
c
la
s
s
if
ie
r
.
It
is
a
de
c
i
s
io
n
tr
e
e
e
ns
e
m
bl
e
m
a
de
out
of
a
r
a
ndom
s
e
le
c
ti
on
of
de
c
is
io
n
tr
e
e
s
.
It
is
a
tr
e
e
pr
e
di
c
to
r
c
om
bi
na
ti
on
in
w
hi
c
h
each
tr
e
e
is
d
e
pe
nde
nt
on
th
e
va
lu
e
s
of
a
r
a
ndom
ve
c
to
r
s
a
m
pl
e
d
in
de
pe
nd
e
nt
ly
a
nd
w
it
h
th
e
s
a
m
e
di
s
tr
ib
ut
io
n
f
or
a
ll
tr
e
e
s
in
th
e
f
or
e
s
t.
In
th
is
,
a
d
e
c
is
io
n
tr
e
e
ha
s
be
e
n
u
s
e
d
as
a
ba
s
ic
l
e
a
r
ne
r
.
E
a
c
h
tr
e
e
w
a
s
c
on
s
tr
uc
te
d
us
in
g
boot
s
tr
a
p
s
a
m
pl
e
s
f
r
om
th
e
tr
a
in
in
g
da
ta
.
A
r
a
ndom
f
e
a
tu
r
e
s
e
le
c
ti
on
pr
ovi
de
s
va
r
ia
ti
on
a
m
ong
th
e
ba
s
e
le
a
r
ne
r
s
.
B
a
ggi
ng
is
u
s
e
d
to
r
e
duc
e
th
e
v
a
r
ia
ti
on
of
a
lg
or
it
hm
s
w
it
h
a
hi
gh
va
r
ia
nc
e
.
A
noi
s
e
or
an
out
li
e
r
in
a
s
i
ngl
e
tr
e
e
c
la
s
s
if
ie
r
can
a
f
f
e
c
t
th
e
c
la
s
s
if
ic
a
ti
on
m
ode
l'
s
pe
r
f
or
m
a
nc
e
;
how
e
ve
r
,
r
a
ndom
f
o
r
e
s
t
m
e
r
ge
d
va
r
io
us
w
e
a
k
le
a
r
ne
r
s
to
bui
ld
a
s
tr
ong
le
a
r
ne
r
th
a
t
can
ha
ndl
e
noi
s
e
or
out
li
e
r
s
due
to
th
e
r
a
ndomne
s
s
it
can
gi
v
e
.
It
is
a
ls
o
c
a
pa
bl
e
of
ha
ndl
in
g
la
r
ge
da
ta
s
e
ts
w
it
h
a
la
r
ge
num
be
r
of
a
tt
r
ib
ut
e
s
.
M
or
e
c
r
uc
ia
ll
y,
it
r
e
ta
in
s
a
c
c
ur
a
c
y
e
ve
n
w
he
n
da
ta
is
un
a
va
il
a
bl
e
[
36]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4747
-
4756
4750
3.3.2
.
C
on
vol
u
t
io
n
al
n
e
u
r
al
n
e
t
w
or
k
an
d
lo
n
g s
h
or
t
-
t
e
r
m
m
e
m
or
y
T
he
s
e
c
ond
a
ppr
oa
c
h
us
e
s
C
N
N
a
nd
L
S
T
M
.
C
N
N
s
a
r
e
a
s
pe
c
ia
l
ty
pe
of
m
ul
ti
la
ye
r
ne
ur
a
l
ne
twor
k
th
a
t
le
a
r
ns
f
r
om
da
ta
.
C
N
N
s
a
r
e
th
e
s
a
m
e
as
n
e
ur
a
l
ne
twor
ks
,
ha
vi
ng
w
e
ig
ht
s
,
bi
a
s
e
s
,
a
nd
in
put
w
it
h
w
hi
c
h
th
e
y
m
a
ke
a
s
c
a
la
r
pr
oduc
t
a
nd
a
ppl
y
an
a
c
ti
va
ti
on
f
unc
ti
on.
D
ue
to
th
e
in
c
r
e
a
s
e
d
num
be
r
of
hi
dde
n
la
ye
r
s
,
th
e
r
e
is
a
ne
e
d
f
or
m
or
e
c
om
put
a
ti
ona
l
r
e
s
our
c
e
s
;
he
nc
e
,
ove
r
f
it
ti
ng
a
r
is
e
s
.
T
hus
,
C
N
N
s
he
lp
to
ove
r
c
om
e
pr
obl
e
m
s
by
di
vi
di
ng
m
ode
ls
in
to
s
m
a
ll
pa
r
ts
of
in
f
or
m
a
ti
on
a
nd
c
om
bi
ni
ng
th
is
in
f
or
m
a
ti
on
in
th
e
de
e
pe
s
t
la
ye
r
s
of
th
e
ne
ur
a
l
ne
twor
k.
T
he
y
a
r
e
ve
r
y
us
e
f
ul
f
or
id
e
nt
if
y
in
g
c
la
s
s
e
s
,
obj
e
c
ts
,
a
nd
gr
oups
in
im
a
ge
s
by
lo
oki
ng
f
or
s
ha
pe
s
in
th
e
m
.
T
hi
s
f
r
a
m
e
w
or
k
ha
s
c
onvolut
io
na
l,
pool
in
g,
a
nd
c
om
pl
e
te
ly
c
onne
c
te
d
la
ye
r
s
.
A
de
e
p
le
a
r
ni
ng
m
ode
l
ne
e
d
s
a
lo
t
of
c
om
put
a
ti
ona
l
po
w
e
r
a
nd
da
ta
to
tr
a
in
.
As
a
r
e
s
ul
t,
C
N
N
s
w
e
r
e
s
tr
ic
tl
y
c
ons
tr
a
in
e
d
to
li
m
it
e
d
s
e
c
to
r
s
a
nd
w
e
r
e
una
bl
e
to
e
nt
e
r
th
e
m
a
c
hi
ne
le
a
r
ni
ng
s
pa
c
e
[
37]
.
L
S
T
M
is
one
of
th
e
m
os
t
c
om
m
on
ly
us
e
d
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
(
R
N
N
)
ne
twor
ks
c
a
pa
bl
e
of
le
a
r
ni
ng
lo
ng
-
te
r
m
de
pe
nde
nc
ie
s
in
c
la
s
s
if
ic
a
ti
on
m
ode
ls
.
T
he
c
or
r
e
c
t
r
e
s
ul
t
is
a
c
hi
e
ve
d
w
h
e
n
L
S
T
M
is
e
xt
e
nde
d
w
it
h
ot
he
r
m
ode
ls
,
a
nd
it
m
a
y
s
e
r
ve
as
th
e
ne
w
ba
s
e
f
or
a
ddi
ti
ona
l
m
ode
ls
.
T
he
va
ni
s
hi
ng
gr
a
di
e
nt
pr
obl
e
m
,
w
hi
c
h
a
f
f
e
c
ts
ne
ur
a
l
ne
twor
ks
,
m
a
ke
s
it
c
ha
ll
e
ngi
ng
to
c
om
pr
e
he
nd
th
e
pa
r
a
m
e
te
r
s
of
th
e
pr
e
vi
ous
la
ye
r
.
A
L
S
T
M
in
c
om
bi
na
ti
on
w
it
h
o
th
e
r
m
ode
ls
he
lp
s
to
s
ol
ve
th
is
is
s
ue
.
In
th
e
L
S
T
M
a
r
c
hi
te
c
tu
r
e
,
as
s
ta
te
d
in
[
1]
,
[
30
]
,
[
38
]
th
e
m
e
m
or
y
c
e
ll
s
s
to
r
e
a
nd
ge
t
in
f
or
m
a
ti
on
ove
r
lo
ng
de
pe
nde
nc
ie
s
,
c
on
s
is
t
of
th
r
e
e
ga
te
s
c
a
ll
e
d
th
e
in
put
,
f
or
ge
t,
a
nd
out
put
ga
te
s
.
T
he
in
put
ga
te
i
t
s
ig
ni
f
ie
s
th
e
ne
w
in
f
or
m
a
ti
on
in
th
e
c
e
ll
us
in
g
in
put
x
t,
th
e
pr
e
vi
ous
hi
dde
n
la
ye
r
,
r
e
c
ur
r
e
nt
w
e
ig
ht
,
a
nd
bi
a
s
.
A
f
o
r
ge
t
ga
te
f
t
de
te
r
m
in
e
s
th
e
e
xt
e
nt
to
w
hi
c
h
th
e
pr
e
vi
ous
m
e
m
or
y
c
e
ll
is
f
or
got
te
n.
T
he
in
f
or
m
a
ti
on
obt
a
in
e
d
f
r
om
th
e
c
ur
r
e
nt
in
put
x
t
,
is
gi
ve
n
by
th
e
out
pu
t
ga
te
u
t.
T
he
in
put
ga
te
i
t
,
c
hoos
e
s
w
hi
c
h
in
f
or
m
a
ti
on
f
r
om
out
put
ga
te
u
t
to
s
to
r
e
in
m
e
m
or
y
c
e
ll
c
t
w
it
h
a
c
ondi
ti
on
f
or
ge
t
ga
te
f
t
a
nd
f
in
a
ll
y
th
e
out
put
o
t
of
th
e
L
S
T
M
de
pe
nds
on
o
t
a
nd
h
t
.
=
(
(
)
+
(
)
ℎ
−
1
+
(
)
)
(
1)
=
(
(
)
+
(
)
ℎ
−
1
+
(
)
)
(
2)
=
ℎ
(
(
)
+
(
)
ℎ
−
1
+
(
)
)
(
3)
=
.
+
.
−
1
(
4)
=
(
(
)
+
(
)
ℎ
−
1
+
(
)
)
(
5)
3.
3.3
.
L
on
g s
h
or
t
-
t
e
r
m
m
e
m
o
r
y
w
it
h
at
t
e
n
t
io
n
la
ye
r
A
not
he
r
a
ppr
oa
c
h
u
s
e
d
is
L
S
T
M
w
it
h
a
s
in
gl
e
a
tt
e
nt
io
n
la
ye
r
.
It
c
om
bi
ne
s
th
e
s
in
gl
e
a
tt
e
nt
io
n
la
y
e
r
w
it
h
th
e
L
S
T
M
.
It
is
e
m
pl
oye
d
to
br
in
g
a
tt
e
nt
io
n
to
th
e
lo
ng
s
e
que
nc
e
'
s
r
e
f
e
r
r
a
l
w
or
d.
By
f
oc
us
in
g
on
im
por
ta
nt
in
f
or
m
a
ti
on
th
a
t
r
e
la
te
s
to
th
e
ta
r
ge
t
va
lu
e
,
th
e
a
tt
e
nt
io
n
la
ye
r
ha
s
us
e
f
ul
in
f
or
m
a
ti
on
a
nd
e
nha
nc
e
s
m
ode
l
pe
r
f
or
m
a
nc
e
.
A
lt
hough
it
c
onc
e
nt
r
a
te
s
on
th
e
de
s
ir
e
d
va
lu
e
,
it
doe
s
n'
t
e
xc
lu
de
a
ny
e
s
s
e
nt
ia
l
in
f
or
m
a
ti
on.
In
c
ont
r
a
s
t
to
th
e
ot
he
r
a
ppr
oa
c
he
s
,
it
can
ha
ndl
e
w
or
d
s
e
que
n
c
e
s
of
di
f
f
e
r
e
nt
le
ngt
hs
.
T
he
a
tt
e
nt
io
n
m
ode
l
e
m
pha
s
iz
e
s
s
e
m
a
nt
ic
e
l
e
m
e
nt
s
w
hi
le
a
ls
o
c
a
pt
ur
in
g
th
e
s
ig
ni
f
ic
a
nc
e
of
e
a
c
h
c
ont
e
xt
w
or
d.
T
h
e
y
s
ugge
s
te
d
an
a
tt
e
nt
io
n
-
ba
s
e
d
L
S
T
M
th
a
t
out
pe
r
f
or
m
s
th
e
pr
io
r
m
ode
l
in
te
r
m
s
of
out
put
.
3.
3.4
.
L
on
g s
h
or
t
-
t
e
r
m
m
e
m
o
r
y
w
it
h
m
u
lt
i
-
h
e
ad
at
t
e
n
t
io
n
l
aye
r
F
in
a
ll
y,
our
s
ta
te
-
of
-
a
r
t
-
a
ppr
oa
c
h
to
s
e
nt
im
e
nt
a
na
ly
s
is
is
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
la
ye
r
.
As
s
ta
te
d
by
[
1]
,
a
tt
e
nt
io
n
m
e
c
ha
ni
s
m
s
ha
ve
r
e
c
e
nt
ly
pr
ove
n
to
be
be
ne
f
ic
ia
l
in
a
va
r
ie
ty
of
N
L
P
a
ppl
ic
a
ti
ons
.
V
a
s
w
a
ni
et
al
.
[
39]
s
t
a
te
d
th
a
t
th
e
a
tt
e
nt
io
n
m
e
c
ha
ni
s
m
s
u
c
c
e
s
s
f
ul
ly
ha
ndl
e
s
s
e
que
n
c
e
d
a
ta
.
In
th
is
w
or
k,
to
in
c
r
e
a
s
e
pe
r
f
or
m
a
nc
e
a
nd
a
c
c
ur
a
c
y
,
we
u
s
e
d
L
S
T
M
w
it
h
a
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
m
e
c
ha
ni
s
m
,
w
hi
c
h
in
te
gr
a
te
s
m
ul
ti
pl
e
he
a
ds
.
It
le
a
r
ns
a
va
r
ie
ty
of
uni
que
f
e
a
tu
r
e
s
.
A
m
ode
r
n
a
r
c
hi
te
c
tu
r
e
f
or
ne
ur
a
l
m
a
c
hi
ne
tr
a
ns
la
ti
on
is
m
ul
ti
he
a
d
s
e
lf
-
a
tt
e
nt
io
n.
K
e
ys
,
va
lu
e
s
,
a
nd
qu
e
r
ie
s
w
it
h
n
d
im
e
ns
io
ns
a
r
e
m
a
n
a
ge
d
by
m
a
ny
he
a
ds
.
An
a
tt
e
nt
io
n
f
unc
ti
on
ta
ke
s
a
qu
e
r
y
Q,
a
s
e
t
of
ke
ys
a
nd
va
lu
e
s
<
K,
V>
to
obt
a
in
an
out
put
O.
T
hi
s
te
c
hni
que
is
of
te
n
c
a
ll
e
d
s
c
a
le
d
dot
-
pr
oduc
t
a
tt
e
nt
io
n
.
It
is
a
s
e
t
of
m
ul
ti
pl
e
he
a
ds
th
a
t
to
ge
th
e
r
le
a
r
n
va
r
io
us
de
pi
c
ti
ons
at
e
ve
r
y
pos
it
io
n
in
th
e
s
e
que
nc
e
,
as
m
e
nt
io
ne
d
in
[
40]
.
F
ig
ur
e
2
il
l
us
tr
a
te
s
a
r
c
hi
te
c
tu
r
e
of
pr
opos
e
d
s
ys
te
m
u
s
in
g
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
w
it
h
L
S
T
M
f
or
s
e
nt
im
e
nt
a
na
ly
s
is
.
3.4.
C
om
p
ar
at
iv
e
an
al
ys
is
T
he
c
om
pa
r
a
ti
ve
a
n
a
ly
s
is
s
ta
te
s
th
a
t
de
e
p
le
a
r
ni
ng
m
ode
ls
out
p
e
r
f
or
m
s
a
s
c
om
pa
r
e
to
r
e
s
t
of
s
tu
di
e
s
in
e
a
r
li
e
r
w
or
ks
.
T
he
m
ode
l
w
a
s
tr
a
in
e
d
a
nd
te
s
t
e
d
w
it
h
f
our
di
f
f
e
r
e
nt
m
a
c
hi
ne
le
a
r
ni
ng
a
nd
de
e
p
le
a
r
ni
ng
a
lg
or
it
hm
s
a
lo
ng
w
it
h
w
or
d
e
m
be
ddi
ng
te
c
hni
que
s
.
T
h
e
r
e
s
ul
ts
of
two
da
ta
s
e
ts
w
e
r
e
r
e
por
te
d
u
s
in
g
th
e
c
onf
us
io
n
m
a
tr
ix
:
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F1
s
c
or
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
Se
nt
ime
nt
analy
s
is
of
s
tu
d
e
nt
’
s
c
om
m
e
nt
s
us
in
g l
ong s
ho
r
t
-
te
r
m
m
e
m
or
y
w
it
h
…
(
B
hav
ana P
. B
hagat
)
4751
F
ig
ur
e
2.
A
r
c
hi
te
c
tu
r
e
of
p
r
opos
e
d
s
ys
te
m
us
in
g
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
w
it
h
L
S
T
M
f
or
s
e
nt
im
e
nt
a
na
ly
s
is
4.
E
X
P
E
R
I
M
E
N
T
A
L
R
E
S
U
L
T
S
AND
P
E
R
F
O
R
M
A
N
C
E
E
V
A
L
U
A
T
I
O
N
In
or
de
r
to
f
in
d
th
e
be
s
t
m
ode
l,
it
is
im
po
r
ta
nt
to
e
s
ta
bl
is
h
a
s
e
r
ie
s
of
s
te
ps
th
a
t
a
ll
ow
us
to
e
xpl
or
e
di
f
f
e
r
e
nt
m
ode
l
c
on
f
ig
ur
a
ti
ons
,
s
uc
h
as
e
ns
e
m
bl
e
te
c
hni
que
s
,
f
ul
ly
c
onne
c
te
d
ne
twor
ks
,
de
ns
e
a
nd
dr
opout
la
ye
r
s
,
w
e
ig
ht
in
it
ia
li
z
a
ti
on
a
nd
us
e
of
a
c
ti
va
ti
on
f
unc
ti
ons
,
l
e
a
r
ni
ng
r
a
te
w
it
h
opt
im
iz
e
r
s
,
as
w
e
ll
as
ot
he
r
hype
r
-
pa
r
a
m
e
te
r
s
.
It
is
a
l
s
o
im
por
ta
nt
to
de
c
id
e
ba
t
c
h
s
i
z
e
,
th
e
num
be
r
of
e
poc
hs
,
a
nd
th
e
e
va
lu
a
ti
on
m
e
tr
ic
s
us
e
d.
F
ir
s
t,
we
pr
e
-
pr
oc
e
s
s
e
d
a
nd
to
ke
ni
z
e
d
th
e
s
tu
de
nt
’
s
da
ta
s
e
t
,
w
hi
c
h
c
ont
a
in
s
pos
it
iv
e
,
ne
g
a
ti
ve
,
a
nd
ne
ut
r
a
l
s
e
nt
im
e
nt
s
,
u
s
in
g
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
te
c
hni
que
s
(
i.
e
.,
T
F
I
D
F
V
e
c
to
r
a
nd
W
or
d2V
e
c
)
to
pr
ovi
de
th
e
to
ke
ni
z
e
d
f
or
m
of
th
e
c
om
m
e
nt
s
f
or
ou
r
pr
opos
e
d
a
ppr
oa
c
he
s
,
w
hi
c
h
a
r
e
gi
ve
n
in
T
a
bl
e
1.
W
or
d2V
e
c
a
nd
T
F
I
D
F
V
e
c
to
r
iz
e
r
w
e
r
e
u
s
e
d
to
tr
a
n
s
f
or
m
th
e
to
ka
ni
z
e
d
c
om
m
e
n
ts
in
to
ve
c
to
r
s
.
F
in
a
ll
y,
th
e
m
ode
l
s
w
e
r
e
tr
a
in
e
d
us
in
g
m
e
th
ods
s
uc
h
as
e
n
s
e
m
bl
e
c
la
s
s
if
ie
r
s
,
L
S
T
M
a
nd
C
N
N
,
L
S
T
M
w
it
h
s
in
gl
e
-
he
a
d
a
tt
e
nt
io
n
a
nd
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n.
O
nc
e
th
e
tr
a
in
in
g
a
nd
te
s
ti
ng
of
s
tu
d
e
nt
c
om
m
e
nt
s
a
r
e
c
om
pl
e
te
d,
we
c
om
pa
r
e
our
th
r
e
e
ba
s
s
li
ne
m
ode
ls
to
our
pr
opos
e
d
m
ode
l,
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
a
nd
e
va
lu
a
t
e
pe
r
f
or
m
a
nc
e
u
s
in
g
a
c
c
ur
a
c
y,
pr
e
c
is
io
n, r
e
c
a
ll
,
a
nd
F1
-
s
c
or
e
.
T
a
bl
e
1.
N
um
be
r
of
s
e
nt
im
e
nt
la
be
l
s
in
da
ta
s
e
t
S
r
.
no
S
e
nt
i
m
e
nt
l
a
be
l
s
0
l
i
ve
l
y
2
1
A
l
t
hough
my
t
e
s
t
s
c
or
e
is
not
hi
gh
be
c
a
u
s
e
I
d...
1
2
G
r
e
a
t
w
or
k!
2
3
T
e
a
c
he
r
s
a
r
e
ve
r
y
e
nt
hus
i
a
s
t
i
c
in
t
e
a
c
hi
ng.
2
4
T
o
s
pe
nd
time
gr
ow
i
ng
up
to
l
e
c
t
ur
e
f
or
s
t
ude
nt
0
4
.1.
E
n
s
e
m
b
le
c
la
s
s
i
f
ie
r
In
th
is
a
ppr
oa
c
h,
a
hybr
id
c
la
s
s
if
ie
r
,
w
hi
c
h
is
a
c
om
bi
na
ti
on
of
m
ul
ti
pl
e
ba
s
e
c
l
a
s
s
if
ie
r
s
us
in
g
vot
in
g
e
ns
e
m
bl
e
s
,
is
us
e
d.
F
iv
e
di
f
f
e
r
e
nt
ba
s
e
c
la
s
s
if
ie
r
s
,
each
w
it
h
a
s
pe
c
if
ic
a
lg
or
it
hm
,
w
e
r
e
u
s
e
d.
F
ir
s
tl
y,
two
r
a
ndom f
or
e
s
t
c
la
s
s
if
ie
r
s
w
it
h
de
c
is
io
n
tr
e
e
s
w
e
r
e
us
e
d.
L
a
te
r
,
t
w
o
A
da
B
oos
t
c
la
s
s
if
ie
r
s
w
it
h
de
c
is
io
n
tr
e
e
s
as
w
e
a
k
le
a
r
ne
r
s
w
e
r
e
us
e
d,
a
nd
f
in
a
ll
y,
one
gr
a
di
e
nt
boos
ti
ng
c
la
s
s
if
ie
r
w
it
h
de
c
is
io
n
tr
e
e
s
w
a
s
us
e
d.
F
or
m
ul
ti
-
c
la
s
s
c
la
s
s
if
ic
a
ti
on
pr
obl
e
m
s
,
th
e
'
O
ne
V
s
R
e
s
tC
la
s
s
if
ie
r
'
is
us
e
d
to
pe
r
f
or
m
m
ul
ti
-
la
be
l
c
la
s
s
if
ic
a
ti
on,
w
he
r
e
each
c
la
s
s
if
ie
r
is
tr
a
in
e
d
to
ha
ndl
e
a
s
pe
c
if
ic
c
la
s
s
,
tr
e
a
ti
ng
a
ll
ot
he
r
c
la
s
s
e
s
as
a
s
in
gl
e
"
r
e
s
t"
c
la
s
s
.
T
h
e
`
V
ot
in
gC
la
s
s
if
ie
r
`
is
an
e
ns
e
m
bl
e
m
e
th
od
th
a
t
c
om
bi
ne
s
th
e
pr
e
di
c
ti
ons
of
th
e
ba
s
e
c
la
s
s
if
ie
r
s
us
in
g
m
a
jo
r
it
y
vot
in
g
(
'
s
of
t'
vot
in
g)
.
T
he
c
la
s
s
la
be
ls
w
it
h
th
e
hi
ghe
s
t
c
om
bi
ne
d
pr
oba
bi
li
ti
e
s
a
m
ong
a
ll
ba
s
e
c
la
s
s
if
ie
r
s
a
r
e
s
e
le
c
te
d
as
th
e
f
in
a
l
pr
e
di
c
ti
on.
4.2
.
L
on
g s
h
or
t
-
t
e
r
m
m
e
m
o
r
y
+
co
n
vol
u
t
io
n
al
n
e
u
r
al
n
e
t
w
or
k
In
th
is
a
ppr
oa
c
h,
th
e
m
ode
l
is
e
va
lu
a
te
d
us
in
g
w
or
d
e
m
be
ddi
ng
,
pr
e
-
tr
a
in
e
d
f
in
e
-
gr
a
in
e
d
W
or
d2ve
c
,
L
S
T
M
a
nd
1
-
D
C
N
N
.
T
w
o
L
S
T
M
la
y
e
r
s
w
it
h
100
uni
ts
each
w
e
r
e
a
dde
d
w
it
h
a
dr
opout
of
0.3
in
or
de
r
to
pr
e
ve
nt
ove
r
f
it
ti
ng.
A
1D
c
onvolut
io
na
l
la
ye
r
(
`
C
onv1D`
)
w
it
h
100
f
il
te
r
s
a
nd
a
f
il
te
r
s
iz
e
of
5
w
a
s
a
dde
d
ne
xt
,
w
hi
c
h
pe
r
f
or
m
e
d
1D
c
onvolut
io
n
ope
r
a
ti
ons
on
th
e
in
pu
t
s
e
que
nc
e
.
T
he
a
c
ti
va
ti
on
f
unc
ti
on
r
e
lu
is
us
e
d
a
f
te
r
th
e
c
onvolut
io
n
la
ye
r
.
A
'
G
lo
ba
lM
a
xP
ool
1D
'
la
ye
r
ta
k
e
s
th
e
m
a
xi
m
um
va
lu
e
ove
r
th
e
te
m
por
a
l
di
m
e
n
s
io
n
a
nd
r
e
duc
e
s
th
e
di
m
e
ns
io
na
li
ty
of
th
e
da
ta
,
ke
e
pi
ng
th
e
m
os
t
i
m
por
ta
nt
f
e
a
tu
r
e
s
.
A
de
ns
e
la
ye
r
w
it
h
16
un
it
s
a
ppl
ie
s
m
a
tr
ix
m
ul
ti
pl
ic
a
ti
on
to
th
e
in
put
da
ta
a
nd
a
ppl
ie
s
th
e
r
e
lu
a
c
ti
va
ti
on
f
unc
ti
on
e
le
m
e
nt
-
w
is
e
.
A
f
in
a
l
de
ns
e
la
ye
r
a
nd
a
S
of
tM
a
x
f
unc
ti
on
a
r
e
r
e
s
pons
ib
le
f
or
pr
e
di
c
ti
ng
th
e
s
e
nt
i
m
e
nt
c
la
s
s
pr
oba
bi
li
ti
e
s
.
4.3
.
L
on
g s
h
or
t
-
t
e
r
m
m
e
m
o
r
y
+
s
in
gl
e
h
e
ad
at
t
e
n
t
io
n
In
th
is
a
ppr
oa
c
h,
th
e
m
ode
l
is
e
va
lu
a
te
d
us
in
g
L
S
T
M
w
it
h
a
s
i
ngl
e
he
a
d
a
tt
e
nt
io
n
la
ye
r
.
An
L
S
T
M
la
ye
r
w
it
h
64
uni
ts
a
nd
r
e
tu
r
n
s
e
que
n
c
e
s
w
il
l
r
e
tu
r
n
th
e
f
ul
l
s
e
que
nc
e
out
put
r
a
th
e
r
th
a
n
ju
s
t
th
e
f
in
a
l
out
put
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4747
-
4756
4752
T
he
a
tt
e
nt
io
n
la
ye
r
he
lp
s
th
e
m
ode
l
f
oc
us
on
im
por
ta
nt
pa
r
ts
of
th
e
in
put
s
e
que
nc
e
dur
in
g
t
r
a
in
in
g.
N
e
xt
,
a
dr
opout
la
ye
r
w
it
h
a
dr
opout
r
a
te
of
0.5
is
a
dde
d,
w
hi
c
h
he
lp
s
to
pr
e
ve
nt
ove
r
f
it
ti
ng.
F
in
a
ll
y,
a
de
ns
e
la
ye
r
,
i.
e
.,
an
out
put
la
ye
r
w
it
h
3
un
it
s
a
nd
a
s
of
tm
a
x
a
c
ti
va
ti
on
f
unc
ti
on,
is
a
dde
d,
w
hi
c
h
is
r
e
s
pons
ib
le
f
or
pr
e
di
c
ti
ng
th
e
s
e
nt
im
e
nt
c
la
s
s
pr
oba
bi
li
ti
e
s
.
A
ls
o,
th
e
c
a
te
gor
ic
a
l
c
r
os
s
e
nt
r
opy
-
ba
s
e
d
lo
s
s
f
unc
ti
on,
A
da
m
as
an
opt
im
iz
e
r
,
a
nd
an
a
c
c
ur
a
c
y
m
e
tr
ic
f
or
e
va
lu
a
ti
on
a
r
e
us
e
d
in
th
e
m
ode
l.
4.4.
L
on
g s
h
or
t
-
t
e
r
m
m
e
m
o
r
y
+
m
u
lt
i
-
h
e
ad
a
t
t
e
n
t
io
n
In
th
is
pr
opos
e
d
a
ppr
oa
c
h,
th
e
m
ode
l
is
e
va
lu
a
te
d
us
in
g
L
S
T
M
w
it
h
a
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
la
ye
r
.
F
i
g
ur
e
3
s
h
o
w
s
o
ur
e
x
p
e
r
i
m
e
n
t
a
l
s
e
t
u
p
to
c
l
a
s
s
i
f
y
s
e
n
ti
m
e
n
t
s
of
s
t
u
de
n
t
’
s
f
e
e
d
b
a
c
k
i
n
to
p
o
s
it
i
v
e
,
n
e
u
tr
a
l
a
n
d
n
e
g
a
t
iv
e
u
s
i
n
g
m
u
lt
i
h
e
a
d
a
t
t
e
n
ti
o
n
w
it
h
L
S
T
M
.
An
e
m
b
e
d
di
n
g
l
a
y
e
r
is
u
s
e
d
to
c
o
n
v
e
r
t
i
n
t
e
g
e
r
-
e
n
c
o
d
e
d
in
p
u
t
d
a
t
a
i
n
to
d
e
n
s
e
v
e
c
t
o
r
s
of
a
f
i
x
e
d
s
i
z
e
.
It
h
a
s
1
00
i
n
p
ut
d
i
m
e
n
s
i
o
n
s
(
vo
c
a
b
s
i
z
e
)
a
nd
p
r
o
d
u
c
e
s
20
-
d
im
e
n
s
i
o
n
a
l
e
m
b
e
d
di
n
g
s
.
N
e
x
t
is
a
m
u
l
t
i
-
h
e
a
d
a
tt
e
n
t
io
n
l
a
y
e
r
w
i
th
h
e
a
d
s
e
q
u
a
l
to
3
a
n
d
an
e
m
b
e
d
d
in
g
d
i
m
e
n
s
i
on
e
q
u
a
l
to
128.
A
f
e
e
d
f
o
r
w
a
r
d
n
e
ur
a
l
n
e
t
w
o
r
k
(
F
F
N
)
is
i
m
p
l
e
m
e
nt
e
d
as
a
s
e
q
u
e
n
ti
a
l
m
o
d
e
l
c
o
n
s
i
s
ti
n
g
of
t
w
o
d
e
n
s
e
l
a
y
e
r
s
.
T
h
e
f
i
r
s
t
l
a
y
e
r
u
s
e
s
t
h
e
Re
L
u
a
c
t
i
v
a
t
io
n
f
u
n
c
ti
o
n
,
a
n
d
t
h
e
s
e
c
o
n
d
l
a
y
e
r
o
u
t
p
ut
s
a
t
e
n
s
or
w
i
t
h
d
i
m
e
n
s
i
o
n
s
e
q
u
a
l
to
128.
T
h
i
s
FFN
is
u
s
e
d
f
o
r
f
u
r
th
e
r
p
r
o
c
e
s
s
i
n
g
a
f
t
e
r
m
u
lt
i
-
h
e
a
d
a
t
t
e
nt
i
o
n.
N
e
xt
,
l
a
y
e
r
no
r
m
a
l
i
z
a
ti
o
n
is
a
p
p
l
i
e
d
to
t
h
e
ou
t
p
ut
s
of
t
h
e
m
u
lt
i
-
h
e
a
d
a
t
t
e
n
ti
on
a
n
d
f
e
e
d
f
or
w
a
r
d
l
a
y
e
r
s
.
It
h
e
l
p
s
to
s
t
a
bi
l
i
z
e
tr
a
i
ni
n
g
by
n
or
m
a
li
z
i
ng
t
h
e
a
c
t
i
v
a
t
io
n
s
w
i
t
hi
n
each
l
a
y
e
r
.
D
r
o
p
ou
t
is
a
r
e
g
u
l
a
r
i
z
a
ti
o
n
m
e
t
h
o
d
u
s
e
d
d
ur
i
n
g
t
r
a
i
n
i
ng
to
p
r
e
v
e
n
t
ov
e
r
f
i
t
ti
n
g
,
w
i
t
h
a
dr
o
po
u
t
r
a
t
e
of
0.
5
.
4.5.
E
val
u
at
io
n
m
e
t
r
ic
s
A
c
c
or
di
n
g
t
o
L
iu
[
4
1]
,
t
h
e
t
e
s
t
i
n
di
c
a
t
or
s
s
e
le
c
t
e
d
f
o
r
o
ur
pr
op
os
e
d
m
od
e
l
s
w
e
r
e
a
c
c
ur
a
c
y
,
pr
e
c
i
s
io
n,
r
e
c
a
l
l,
a
nd
F1
-
s
c
o
r
e
.
We
pe
r
f
or
m
e
d
e
x
p
e
r
im
e
nt
s
a
nd
c
om
p
a
r
e
d
t
he
pr
o
p
o
s
e
d
L
S
T
M
w
i
th
t
he
m
ul
ti
-
h
e
a
d
a
tt
e
nt
io
n
m
o
d
e
l
w
it
h
o
th
e
r
b
a
s
e
li
n
e
m
o
d
e
l
s
.
T
h
e
pr
o
po
s
e
d
L
S
T
M
w
it
h
th
e
m
u
lt
i
-
h
e
a
d
a
tt
e
nt
io
n
c
l
a
s
s
if
i
c
a
t
io
n
m
o
d
e
l
is
c
o
m
pa
r
e
d
w
i
th
o
th
e
r
b
a
s
e
l
in
e
m
o
d
e
l
s
as
g
iv
e
n:
i)
a
n
e
n
s
e
m
b
le
c
l
a
s
s
i
f
i
e
r
is
a
d
e
c
i
s
i
o
n
t
r
e
e
e
n
s
e
m
bl
e
m
a
d
e
ou
t
of
a
r
a
nd
om
s
e
l
e
c
t
io
n
of
d
e
c
i
s
io
n
t
r
e
e
s
;
ii)
L
S
T
M
+
C
N
N
w
it
h
m
or
e
g
a
t
e
s
to
c
on
tr
ol
l
on
g
s
e
q
u
e
nc
e
s
of
i
nf
or
m
a
ti
on
,
a
l
o
ng
w
it
h
w
or
d
e
m
b
e
d
di
n
g
in
W
or
d2
v
e
c
;
i
ii
)
L
S
T
M
+
a
tt
e
nt
io
n
is
a
c
om
bi
n
a
t
io
n
of
L
S
T
M
a
n
d
s
in
gl
e
a
tt
e
nt
io
n
l
a
y
e
r
u
s
e
d
to
f
o
c
u
s
t
h
e
r
e
f
e
r
r
a
l
w
or
d
in
l
on
g
s
e
qu
e
n
c
e
s
;
a
n
d
iv
)
L
S
T
M
w
it
h
m
ul
ti
-
h
e
a
d
a
tt
e
nt
io
n
is
a
c
o
n
c
a
t
e
n
a
ti
on
of
e
m
be
d
di
ng
+
m
ul
ti
-
h
e
a
d
a
tt
e
nt
io
n
+
L
S
T
M
.
It
c
o
m
b
in
e
s
m
u
lt
i
p
le
h
e
a
d
s
to
i
n
c
r
e
a
s
e
a
c
c
u
r
a
c
y
p
e
r
f
o
r
m
a
n
c
e
.
It
c
o
n
ta
in
s
m
or
e
l
a
y
e
r
s
t
h
a
n
o
th
e
r
b
a
s
e
m
o
d
e
l
s
.
T
w
o
dr
op
o
ut
a
n
d
f
l
a
t
t
e
n
la
y
e
r
s
w
e
r
e
u
s
e
d
to
a
vo
id
ov
e
r
f
i
tt
in
g.
4.6.
E
xp
e
r
im
e
n
t
al
r
e
s
u
lt
s
In
th
is
s
e
c
ti
on,
T
a
bl
e
s
2
to
5
di
s
pl
a
y
a
c
onf
us
io
n
m
a
tr
ix
a
nd
c
la
s
s
if
ic
a
ti
on
r
e
por
t
f
or
a
ll
f
our
m
ode
ls
.
T
a
bl
e
6
s
ho
w
s
th
e
e
v
a
lu
a
ti
on
m
e
tr
ic
s
,
i.
e
.,
te
s
ti
ng
a
c
c
ur
a
c
y
w
it
h
r
e
a
s
ona
bl
e
pe
r
f
or
m
a
nc
e
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F1
-
s
c
or
e
f
or
a
ll
m
ode
ls
.
It
is
s
e
e
n
th
a
t
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
pe
r
f
or
m
s
be
tt
e
r
th
a
n
th
e
r
e
s
t
of
th
e
th
r
e
e
m
ode
ls
in
te
r
m
s
of
a
c
c
ur
a
c
y,
F1
-
s
c
or
e
,
pr
e
c
is
io
n
,
a
nd
r
e
c
a
ll
of
95
.
56%
,
95.57%
,
95.44%
a
nd
95.57%
r
e
s
pe
c
ti
ve
ly
,
f
ol
lo
w
e
d
by
L
S
T
M
+
a
tt
e
nt
io
n
,
e
n
s
e
m
bl
e
c
la
s
s
if
ie
r
,
a
nd
L
S
T
M
+
C
N
N
.
T
a
bl
e
2. E
ns
e
m
bl
e
c
l
a
s
s
if
ie
r
c
onf
us
io
n m
a
tr
ix
A
c
t
ua
l
P
r
e
di
c
t
e
d
N
e
ut
r
a
l
N
e
ga
t
i
ve
P
os
i
t
i
ve
N
e
ut
r
a
l
1,523
26
38
N
e
ga
t
i
ve
0
1,761
337
P
os
i
t
i
ve
4
51
1,651
T
a
bl
e
3. L
S
T
M
+
C
N
N
m
ode
l
c
onf
us
io
n m
a
tr
ix
A
c
t
ua
l
P
r
e
di
c
t
e
d
N
e
ut
r
a
l
N
e
ga
t
i
ve
P
os
i
t
i
ve
N
e
ut
r
a
l
1,788
75
145
N
e
ga
t
i
ve
228
1,621
207
P
os
i
t
i
ve
205
125
2,075
T
a
bl
e
4. L
S
T
M
+
s
in
gl
e
-
he
a
d
a
tt
e
nt
io
n
m
ode
l
c
onf
us
io
n m
a
tr
ix
A
c
t
ua
l
P
r
e
di
c
t
e
d
N
e
ut
r
a
l
N
e
ga
t
i
ve
P
os
i
t
i
ve
N
e
ut
r
a
l
1,694
8
156
N
e
ga
t
i
ve
91
1,431
71
P
os
i
t
i
ve
85
18
1,837
T
a
bl
e
5.
L
S
T
M
+
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
m
ode
l
c
onf
us
io
n
m
a
tr
ix
A
c
t
ua
l
P
r
e
di
c
t
e
d
N
e
ut
r
a
l
N
e
ga
t
i
ve
P
os
i
t
i
ve
N
e
ut
r
a
l
5,218
5
83
N
e
ga
t
i
ve
224
4,589
245
P
os
i
t
i
ve
127
33
5,646
T
a
bl
e
6.
E
va
lu
a
ti
on
m
e
tr
ic
s
f
or
a
ll
m
ode
ls
M
e
t
r
i
c
e
s
/
m
ode
l
s
L
S
T
M
+
C
N
N
L
S
T
M
+
A
T
T
E
ns
e
m
bl
e
c
l
a
s
s
i
f
i
e
r
L
S
T
M
+
m
ul
t
i
-
he
a
d
a
t
t
e
nt
i
on
A
c
c
ur
a
c
y
(%)
84.778
92.04
87.18
95.56
F1
-
S
c
or
e
0.8477
0.9204
0.9156
0.9557
P
r
e
c
i
s
i
on
0.8472
0.9190
0.94
0.9544
R
e
c
a
l
l
0.8477
0.9204
0.87
0.9557
A
l
s
o,
pr
e
c
i
s
i
on,
r
e
c
a
l
l,
a
nd
t
he
F
1
-
s
c
o
r
e
c
om
put
e
d
f
o
r
a
ll
m
od
e
l
s
w
e
r
e
s
ho
w
n
i
n
T
a
bl
e
s
7
t
o
9
.
I
t
i
s
s
e
e
n
th
a
t
th
e
pe
r
f
or
m
a
n
c
e
on
te
s
t
d
a
t
a
u
s
i
ng
pr
e
c
i
s
i
on
f
or
p
o
s
it
i
ve
s
e
n
ti
m
e
n
ts
p
e
r
f
or
m
b
e
tt
e
r
i
n
e
n
s
e
m
bl
e
c
l
a
s
s
if
i
e
r
a
nd
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
Se
nt
ime
nt
analy
s
is
of
s
tu
d
e
nt
’
s
c
om
m
e
nt
s
us
in
g l
ong s
ho
r
t
-
te
r
m
m
e
m
or
y
w
it
h
…
(
B
hav
ana P
. B
hagat
)
4753
th
e
p
e
r
f
or
m
a
nc
e
on
t
e
s
t
da
ta
u
s
i
ng
r
e
c
a
ll
a
nd
F
1
-
s
c
or
e
f
or
ne
ut
r
a
l
s
e
nt
im
e
nt
s
p
e
r
f
or
m
s
b
e
tt
e
r
in
e
n
s
e
m
bl
e
c
l
a
s
s
if
i
e
r
a
s
c
o
m
p
a
r
e
d
to
t
he
r
e
s
t
of
t
hr
e
e
pr
opo
s
e
d
m
od
e
l
s
.
W
e
te
s
t
e
d
t
h
e
m
o
de
l
on
u
ns
e
e
n
d
a
t
a
c
ol
l
e
c
te
d
f
r
o
m
th
e
52
0
c
om
m
e
nt
s
of
f
i
na
l
-
y
e
a
r
s
t
ud
e
nt
s
e
nr
ol
le
d i
n
202
3
w
hi
c
h
a
l
s
o
pe
r
f
or
m
e
d
w
e
ll
a
s
s
h
ow
n i
n
T
a
b
le
10
.
T
a
bl
e
7.
P
e
r
f
or
m
a
nc
e
on
te
s
t
da
ta
u
s
in
g
P
r
e
c
is
io
n
M
ode
l
P
r
e
c
i
s
i
on
N
e
ut
r
a
l
N
e
ga
t
i
ve
P
os
i
t
i
ve
L
S
T
M
+C
N
N
0.79
0.89
0.86
L
S
T
M
+A
T
T
0.91
0.90
0.95
E
ns
e
m
bl
e
c
l
a
s
s
i
f
i
e
r
0.96
0.84
0.97
T
a
bl
e
8.
P
e
r
f
or
m
a
nc
e
on
te
s
t
da
ta
u
s
in
g
R
e
c
a
ll
M
ode
l
R
e
c
a
l
l
N
e
ut
r
a
l
N
e
ga
t
i
ve
P
os
i
t
i
ve
L
S
T
M
+C
N
N
0.84
0.85
0.86
L
S
T
M
+A
T
T
0.91
0.94
0.92
E
ns
e
m
bl
e
c
l
a
s
s
i
f
i
e
r
0.98
0.89
0.88
T
a
bl
e
9.
P
e
r
f
or
m
a
nc
e
on
te
s
t
da
ta
u
s
in
g
F1
-
S
c
or
e
M
ode
l
F1
-
s
c
or
e
N
e
ut
r
a
l
N
e
ga
t
i
ve
P
os
i
t
i
ve
L
S
T
M
+C
N
N
0.89
0.81
0.85
L
S
T
M
+A
T
T
0.91
0.98
0.89
E
ns
e
m
bl
e
c
l
a
s
s
i
f
i
e
r
f
or
e
s
t
1
0.96
0.81
T
a
bl
e
10.
F
e
w
s
a
m
pl
e
pr
e
di
c
ti
ons
a
nd
a
c
tu
a
l
va
lu
e
s
on
uns
e
e
n
da
ta
I
nput
t
e
xt
A
c
t
ua
l
s
e
nt
i
m
e
nt
s
P
r
e
di
c
t
e
d
s
e
nt
i
m
e
nt
s
L
e
c
t
ur
e
r
s
a
r
e
e
a
s
y
to
unde
r
s
t
a
nd
a
nd
e
nt
hu
s
i
a
s
t
i
c
P
os
i
t
i
ve
P
os
i
t
i
ve
I
f
e
e
l
pr
e
s
s
ur
e
d
a
nd
unc
om
f
or
t
a
bl
e
to
t
a
l
k
to
you
a
s
k
you
but
you
don
t
a
ns
w
e
r
N
e
ga
t
i
ve
N
e
ga
t
i
ve
T
he
t
e
a
c
he
r
f
oc
us
e
s
on
t
he
r
i
ght
f
oc
us
P
os
i
t
i
ve
N
e
ut
r
a
l
A
s
f
or
vi
s
ua
l
ba
s
i
c
i
don
t
t
hi
nk
it
s
ne
c
e
s
s
a
r
y
so
i
s
hor
t
e
n
t
he
t
e
a
c
hi
ng
time
i
ns
t
e
a
d
of
s
pe
ndi
ng
ne
a
r
l
y
ha
l
f
a
s
e
m
e
s
t
e
r
to
gi
ve
an
ove
r
vi
e
w
of
vi
s
u
a
l
ba
s
i
c
N
e
ga
t
i
ve
N
e
ga
t
i
ve
A
ppl
y
i
m
a
ge
s
vi
de
o
c
l
i
ps
in
t
he
t
e
a
c
hi
ng
pr
oc
e
s
s
P
os
i
t
i
ve
N
e
ga
t
i
ve
S
om
e
t
i
m
e
s
a
bi
t
f
a
s
t
N
e
ga
t
i
ve
N
e
ga
t
i
ve
F
ig
ur
e
3
(
a
)
to
3
(
d
)
s
how
s
th
e
r
e
c
e
iv
e
r
ope
r
a
ti
ng
c
h
a
r
a
c
te
r
is
ti
c
c
ur
ve
(
R
O
C
)
c
ur
ve
s
f
or
each
c
la
s
s
f
or
a
ll
m
ode
ls
,
a
nd
th
e
a
r
e
a
unde
r
each
c
ur
ve
(
A
U
C
)
in
di
c
a
te
s
th
e
c
la
s
s
if
ie
r
'
s
pe
r
f
or
m
a
nc
e
f
or
th
a
t
c
la
s
s
.
It
s
how
s
th
e
c
om
put
a
ti
on
of
th
e
f
a
ls
e
po
s
it
iv
e
r
a
te
,
tr
ue
pos
it
iv
e
r
a
te
,
a
nd
t
hr
e
s
hol
d
f
or
each
c
la
s
s
.
T
he
f
unc
ti
on
is
a
ppl
ie
d
to
th
e
tr
ue
la
be
ls
a
nd
pr
e
di
c
te
d
pr
oba
bi
li
ti
e
s
f
or
each
c
la
s
s
a
nd
f
in
a
ll
y
c
om
put
e
s
th
e
a
r
e
a
unde
r
th
e
R
O
C
c
ur
ve
f
or
each
c
la
s
s
.
A
hi
gh
e
r
AUC
va
lu
e
in
di
c
a
t
e
s
th
a
t
m
ode
ls
a
r
e
p
e
r
f
or
m
in
g
be
tt
e
r
c
om
pa
r
e
d
to
e
a
r
li
e
r
w
or
ks
.
(
a
)
(
b)
(
c
)
(
d)
F
ig
ur
e
3.
R
O
C
c
ur
ve
f
or
(
a)
e
ns
e
m
bl
e
c
la
s
s
if
ie
r
,
(
b)
L
S
T
M
a
nd
C
N
N
,
(
c
)
L
S
T
M
w
it
h
s
in
gl
e
-
he
a
d
a
tt
e
nt
io
n,
a
nd
(
d)
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4747
-
4756
4754
4.7.
D
is
c
u
s
s
io
n
an
d
c
om
p
ar
i
s
on
m
od
e
ls
P
r
e
vi
ous
r
e
s
e
a
r
c
he
r
s
us
e
d
va
r
io
us
le
xi
c
on
-
ba
s
e
d
,
m
a
c
hi
n
e
le
a
r
ni
ng
,
a
nd
de
e
p
le
a
r
ni
ng
te
c
hni
que
s
to
id
e
nt
if
y
s
tu
de
nt
’
s
s
e
nt
im
e
nt
s
ba
s
e
d
on
th
e
ir
r
e
s
pons
e
s
.
A
ls
o,
s
om
e
of
th
e
e
a
r
li
e
r
s
tu
di
e
s
di
d
not
pr
ovi
de
e
va
lu
a
ti
on
m
e
tr
ic
s
f
or
pe
r
f
or
m
a
nc
e
m
e
a
s
ur
e
m
e
nt
.
T
he
s
e
e
a
r
li
e
r
w
or
ks
s
e
r
ve
d
as
b
e
nc
hm
a
r
ks
a
g
a
in
s
t
w
hi
c
h
we
c
om
pa
r
e
d
th
e
r
e
s
ul
ts
of
our
pr
o
pos
e
d
m
ode
ls
.
In
pr
e
vi
ous
w
or
k
by
[
1]
,
[
30]
us
e
d
th
e
U
I
T
-
V
S
F
C
c
or
pus
a
nd
c
la
s
s
if
ic
a
ti
on
m
e
th
ods
L
S
T
M
a
nd
DT
-
L
S
T
M
w
it
h
S
V
M
m
ode
ls
w
it
h
a
c
c
ur
a
c
y
of
90.74%
a
nd
F
U
S
I
O
N
m
ode
ls
w
it
h
a
c
c
ur
a
c
y
of
91.96%
,
r
e
s
p
e
c
ti
ve
ly
,
w
he
r
e
a
s
our
pr
opos
e
d
m
ode
l
us
e
d
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
m
ode
ls
w
it
h
an
a
c
c
ur
a
c
y
of
95.75%
.
A
ls
o,
our
pr
opos
e
d
m
ode
l
w
a
s
te
s
te
d
on
520
uns
e
e
n
c
om
m
e
nt
s
c
ol
le
c
te
d
f
r
om
th
e
f
e
e
dba
c
k
of
f
in
a
l
-
ye
a
r
s
tu
de
nt
s
in
2023,
w
hi
c
h
p
e
r
f
or
m
e
d
w
e
ll
on
uns
e
e
n
d
a
ta
.
T
a
bl
e
11
s
ho
w
s
a
c
om
pa
r
is
on
of
our
pr
opos
e
d
m
ode
l
to
th
e
m
ode
ls
f
r
om
pr
e
v
io
us
s
tu
di
e
s
in
w
hi
c
h
va
r
io
us
de
e
p
le
a
r
ni
ng
a
ppr
oa
c
he
s
w
e
r
e
us
e
d
on
s
tu
de
nt
da
ta
s
e
ts
as
one
of
th
e
dom
in
a
nt
m
ode
ls
us
e
d
in
th
e
e
duc
a
ti
on
dom
a
in
f
o
r
be
tt
e
r
pe
r
f
or
m
a
nc
e
.
It
de
m
ons
tr
a
te
s
th
a
t
our
m
ode
l
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
out
pe
r
f
or
m
s
e
xi
s
ti
ng
c
la
s
s
if
ie
r
s
in
e
a
r
li
e
r
w
or
ks
as
w
e
ll
as
our
pr
opos
e
d
r
e
s
t
of
th
e
th
r
e
e
be
nc
hm
a
r
k
m
e
th
ods
.
T
a
bl
e
11.
C
om
pa
r
i
s
on
of
our
m
ode
l
s
to
e
a
r
l
i
e
r
s
t
udi
e
s
us
i
ng
de
e
p
l
e
a
r
ni
ng
a
ppr
oa
c
he
s
S
t
udy/
ye
a
r
M
e
t
hod/
a
l
gor
i
t
hm
D
a
t
a
c
or
pus
P
e
r
f
or
m
a
nc
e
A
c
c
ur
a
c
y
(%)
P
r
e
c
i
s
i
on
(%)
R
e
c
a
l
l
(%)
F1
-
S
c
or
e
(%)
S
i
ndhu
e
t
al
.
[
32]
/
2016
L
S
T
M
m
ode
l
f
or
l
a
ye
r
1
a
nd
l
a
ye
r
2
5,000 c
om
m
e
nt
s
93.0
88.0
85.0
86.0
N
guye
n
e
t
al
.
[
30]
/
2018
L
S
T
M
a
nd
D
T
-
L
S
T
M
(
w
i
t
h
S
V
M
)
M
ode
l
s
U
I
T
-
V
S
F
C
c
or
pus
16,175
s
e
nt
e
nc
e
s
90.74
NA
NA
90.2
C
a
ba
da
e
t
al
.
[
31]
/
2018
C
N
N
+L
S
T
M
ye
l
p
-
147672,
S
e
nt
i
T
e
xt
-
10834,
E
duS
e
r
e
-
4300
s
t
ude
nt
’
s
c
om
m
e
nt
s
S
e
nt
i
T
e
xt
-
88.26 a
nd
E
duE
R
A
S
-
90.30
NA
NA
NA
S
a
nge
e
t
ha
a
nd
P
r
a
bha
[
1]
/
2020
L
S
T
M
,
L
S
T
M
+A
T
T
,
m
ui
t
i
he
a
d
A
T
T
a
nd
F
U
S
I
O
N
(
m
ul
t
i
-
he
a
d
A
T
T
+
E
m
be
ddi
ng
+
L
S
T
M
)
16,175
s
t
ude
nt
s
f
e
e
dba
c
k
s
e
nt
e
nc
e
s
91.96
75.60
87.51
76.90
M
a
bunda
e
t
al
.
[
27]
/
2021
S
uppor
t
ve
c
t
or
m
a
c
hi
ne
s
,
m
ul
t
i
nom
i
a
l
m
a
i
ve
B
a
ye
s
,
r
a
ndom
f
or
e
s
t
s
,
k
-
ne
a
r
e
s
t
ne
i
ghbour
s
a
nd ne
ur
a
l
ne
t
w
or
ks
S
t
ude
nt
s
f
e
e
dba
c
k
f
r
om
K
a
ggl
e
da
t
a
s
e
t
185 r
e
c
or
ds
84
NA
NA
NA
O
ur
pr
opos
e
d
m
ode
l
s
E
ns
e
m
bl
e
c
l
a
s
s
i
f
i
e
r
s
T
he
V
i
e
t
na
m
e
s
e
21,561
s
t
ude
nt
f
e
e
dba
c
k
c
or
pus
, a
s
w
e
l
l
a
s
520
s
t
ude
nt
da
t
a
c
ol
l
e
c
t
e
d
f
r
om
a
f
i
na
l
-
ye
a
r
s
t
ude
nt
'
s
c
om
m
e
nt
i
n 2023
87.18
94.0
87.0
91.0
L
S
T
M
a
nd C
N
N
84.77
84.0
84.0
84.0
L
S
T
M
w
i
t
h a
t
t
e
nt
i
on
92.04
91.0
92.0
92.0
L
S
T
M
w
i
t
h m
ul
t
i
-
he
a
d a
t
t
e
nt
i
on
95.57
95.44
95.57
95.56
*N
A
–
N
ot
a
va
i
l
a
bl
e
5.
C
O
N
C
L
U
S
I
O
N
We
in
tr
oduc
e
d
th
e
U
I
T
-
V
S
F
C
c
or
pus
a
nd
uns
e
e
n
f
in
a
l
-
ye
a
r
s
tu
de
nt
’
s
c
om
m
e
nt
s
f
or
th
e
a
na
ly
s
is
of
e
duc
a
ti
ona
l
s
e
nt
im
e
nt
.
We
pr
opos
e
d
th
e
m
e
th
ods
us
in
g
a
m
a
c
hi
ne
le
a
r
ni
ng
hybr
id
m
ode
l
,
i.
e
.,
e
ns
e
m
bl
e
c
la
s
s
if
ie
r
,
a
nd
de
e
p
l
e
a
r
ni
ng
m
ode
ls
,
i.
e
.
,
L
S
T
M
a
nd
C
N
N
m
o
de
ls
,
a
lo
ng
w
it
h
w
or
d2ve
c
e
m
be
ddi
ng,
L
S
T
M
w
it
h
s
in
gl
e
a
tt
e
nt
io
n,
a
nd
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n.
A
c
c
or
di
ng
to
th
e
f
in
di
ngs
of
our
e
xpe
r
im
e
nt
s
,
it
w
a
s
s
e
e
n
th
a
t
our
pr
opos
e
d
m
ode
l
,
L
S
T
M
w
it
h
m
ul
ti
-
he
a
d
a
tt
e
nt
io
n
w
it
h
e
m
be
ddi
ng
la
ye
r
,
a
c
hi
e
ve
s
be
tt
e
r
r
e
s
ul
ts
as
pe
r
th
e
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
c
onduc
te
d
us
in
g
ot
he
r
ba
s
e
li
ne
m
e
th
ods
as
w
e
ll
as
e
a
r
li
e
r
m
ode
ls
.
O
ur
m
ode
l
pe
r
f
or
m
s
be
tt
e
r
f
or
lo
ng
s
e
que
nc
e
s
of
s
e
nt
e
nc
e
s
a
nd
a
l
s
o
y
ie
ld
s
a
tt
e
nt
io
n
to
th
e
r
e
f
e
r
r
a
l
w
or
ds
.
M
or
e
ove
r
,
by
e
m
pl
oyi
ng
L
S
T
M
w
it
h
a
M
ul
ti
-
he
a
d
a
tt
e
nt
io
n
la
ye
r
,
we
a
c
hi
e
ve
th
e
m
os
t
opt
im
a
l
r
e
s
ul
ts
w
it
h
te
s
ti
ng
a
c
c
ur
a
c
y,
F1
-
s
c
or
e
,
pr
e
c
i
s
io
n
a
nd
r
e
c
a
ll
p
e
r
f
or
m
a
nc
e
of
95.57%
,
95.56%
,
95.44%
,
a
nd
95.57%
r
e
s
pe
c
ti
ve
ly
,
w
hi
c
h
out
pe
r
f
or
m
th
e
r
e
s
t
of
th
e
th
r
e
e
ba
s
e
li
ne
m
od
e
ls
a
nd
a
l
s
o
pr
oduc
e
be
tt
e
r
r
e
s
ul
ts
in
te
r
m
s
of
a
c
c
ur
a
c
y
,
pr
e
c
is
io
n
,
r
e
c
a
ll
,
a
nd
F1
-
s
c
or
e
as
pr
opos
e
d
by
e
a
r
li
e
r
s
tu
di
e
s
.
We
a
ls
o
te
s
te
d
th
e
m
od
e
l
on
uns
e
e
n
da
t
a
c
ol
le
c
te
d
f
r
om
th
e
520
c
om
m
e
nt
s
of
f
in
a
l
-
ye
a
r
s
tu
de
nt
s
.
T
he
m
ode
l
pe
r
f
or
m
s
w
e
ll
on
un
s
e
e
n
da
ta
.
In
th
e
f
ut
ur
e
,
we
a
r
e
goi
ng
to
pe
r
f
or
m
e
xpe
r
im
e
nt
s
u
s
in
g
va
r
ia
nt
s
of
th
e
L
S
T
M
,
L
S
T
M
w
it
h
bi
di
r
e
c
ti
ona
l
e
n
c
ode
r
r
e
pr
e
s
e
nt
a
ti
ons
f
r
om
tr
a
ns
f
or
m
e
r
s
(
B
E
R
T
)
,
a
nd
ot
he
r
de
e
p
le
a
r
ni
ng
m
ode
ls
.
A
ls
o,
c
ol
le
c
t
m
or
e
da
ta
to
pe
r
f
or
m
s
e
nt
im
e
nt
a
na
ly
s
is
on
th
e
r
e
a
l
-
w
or
ld
pr
obl
e
m
.
R
E
F
E
R
E
N
C
E
S
[
1]
K.
S
a
nge
e
t
ha
a
nd
D.
P
r
a
bha
,
“
S
e
nt
i
m
e
nt
a
na
l
ys
i
s
of
s
t
ude
nt
f
e
e
dba
c
k
us
i
ng
m
u
l
t
i
-
he
a
d
a
t
t
e
nt
i
on
f
us
i
on
m
ode
l
of
w
o
r
d
a
nd
c
ont
e
xt
e
m
be
ddi
ng
f
or
L
S
T
M
,”
J
ou
r
nal
of
A
m
bi
e
nt
I
nt
e
l
l
i
ge
nc
e
and
H
um
ani
z
e
d
C
o
m
put
i
ng
,
vol
.
12,
no.
3,
pp.
4117
–
4126,
2021,
doi
:
10.1007/
s
12652
-
020
-
01791
-
9.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
Se
nt
ime
nt
analy
s
is
of
s
tu
d
e
nt
’
s
c
om
m
e
nt
s
us
in
g l
ong s
ho
r
t
-
te
r
m
m
e
m
or
y
w
it
h
…
(
B
hav
ana P
. B
hagat
)
4755
[
2]
J.
Z
hou
a
nd
J.
Y
e
,
“
S
e
nt
i
m
e
nt
a
na
l
ys
i
s
in
e
du
c
a
t
i
on
r
e
s
e
a
r
c
h:
a
r
e
vi
e
w
of
j
our
na
l
publ
i
c
a
t
i
ons
,”
I
nt
e
r
ac
t
i
v
e
L
e
ar
ni
ng
E
nv
i
r
onm
e
nt
s
,
vol
.
31,
no.
3,
pp.
1252
–
1264,
A
pr
.
2023,
doi
:
10.1080/
10494820.2020.182698
5.
[
3]
M.
A
t
i
f
,
“
A
n
e
nha
nc
e
d
f
r
a
m
e
w
or
k
f
or
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
of
s
t
ude
nt
s
’
s
ur
ve
y
s
:
A
r
a
b
O
pe
n
U
ni
ve
r
s
i
t
y
bus
i
n
e
s
s
pr
ogr
a
m
c
our
s
e
s
c
a
s
e
s
t
udy,”
B
us
i
ne
s
s
and
E
c
onom
i
c
s
J
ou
r
nal
,
vol
.
9,
no.
1,
2018,
doi
:
10.4172/
2151
-
6219.1000337.
[
4]
C
. A
. P
a
c
ol
a
nd
T
. D
. P
a
l
a
oa
g,
“
E
nha
nc
i
ng
s
e
nt
i
m
e
nt
a
n
a
l
ys
i
s
of
t
e
xt
ua
l
f
e
e
db
a
c
k i
n t
he
s
t
ude
nt
-
f
a
c
ul
t
y e
v
a
l
ua
t
i
on us
i
ng
m
a
c
hi
n
e
l
e
a
r
ni
ng
t
e
c
hni
que
s
,”
E
ur
ope
an
J
ou
r
nal
of
E
ngi
ne
e
r
i
ng
Sc
i
e
nc
e
and
T
e
c
hnol
ogy
,
vol
.
4,
no.
1,
pp.
27
–
34,
2021,
doi
:
10.33422/
e
j
e
s
t
.v4i
1.604.
[
5]
K.
Z.
A
ung
a
nd
N.
N.
M
yo,
“
S
e
nt
i
m
e
nt
a
n
a
l
ys
i
s
of
s
t
ude
nt
s
’
c
om
m
e
nt
us
i
ng
l
e
xi
c
on
-
ba
s
e
d
a
ppr
oa
c
h,
”
P
r
oc
e
e
di
ngs
-
16t
h
I
E
E
E
/
A
C
I
S
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
C
om
put
e
r
and
I
nf
or
m
at
i
on
Sc
i
e
nc
e
,
I
C
I
S
2017
,
pp.
149
–
154,
2017,
doi
:
10.1109/
I
C
I
S
.2017.7959985.
[
6]
H.
N
e
w
m
a
n
a
nd
D.
J
oyne
r
,
“
S
e
nt
i
m
e
nt
a
na
l
ys
i
s
of
s
t
ude
nt
e
va
l
ua
t
i
ons
of
t
e
a
c
hi
ng,”
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
in
E
duc
at
i
on:
19t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
,
A
I
E
D
2018
,
pp.
246
–
250,
2018,
doi
:
10.1007/
978
-
3
-
319
-
93846
-
2_45.
[
7]
Q.
R
a
j
put
,
S.
H
a
i
de
r
,
a
nd
S.
G
ha
ni
,
“
L
e
xi
c
on
-
ba
s
e
d
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
of
t
e
a
c
he
r
s
’
e
va
l
ua
t
i
on,”
A
ppl
i
e
d
C
om
put
at
i
onal
I
nt
e
l
l
i
ge
nc
e
and
Sof
t
C
om
put
i
ng
,
pp.
1
–
12,
2016,
doi
:
10.1155/
2016/
2385429.
[
8]
S.
R
a
ni
a
nd
P.
K
um
a
r
,
“A
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
s
ys
t
e
m
to
i
m
pr
ove
t
e
a
c
hi
ng
a
nd
l
e
a
r
ni
ng,”
C
om
put
e
r
,
vol
.
50,
no.
5,
pp.
36
–
43,
2017,
doi
:
10.1109/
M
C
.2017.133.
[
9]
F.
F.
B
a
l
a
ha
di
a
,
M.
C.
G.
F
e
r
na
ndo,
a
nd
I.
C.
J
ua
na
t
a
s
,
“
T
e
a
c
he
r
’
s
pe
r
f
or
m
a
nc
e
e
va
l
ua
t
i
on
t
ool
us
i
ng
opi
ni
on
m
i
ni
ng
w
i
t
h
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
,”
in
P
r
oc
e
e
di
ngs
-
2016
I
E
E
E
R
e
gi
on
10
Sy
m
pos
i
um
,
T
E
N
SY
M
P
2016
,
2016,
pp.
95
–
98,
doi
:
10.1109/
T
E
N
C
O
N
S
pr
i
ng.2016.7519384.
[
10]
M.
S
i
va
kum
a
r
a
nd
U.
S.
R
e
ddy,
“
A
s
pe
c
t
ba
s
e
d
s
e
nt
i
m
e
nt
a
na
l
y
s
i
s
of
s
t
ude
n
t
s
opi
ni
on
us
i
ng
m
a
c
hi
ne
l
e
a
r
ni
ng
t
e
c
hni
que
s
,
”
in
P
r
oc
e
e
di
ngs
of
t
he
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
I
nv
e
nt
i
v
e
C
om
put
i
ng
and
I
nf
or
m
at
i
c
s
,
I
C
I
C
I
2017
,
2018,
pp.
726
–
731,
doi
:
10.1109/
I
C
I
C
I
.2017.8365231.
[
11]
N.
C
a
pua
no,
S.
C
a
b
a
l
l
é
,
J.
C
one
s
a
,
a
nd
A.
G
r
e
c
o,
“
A
t
t
e
nt
i
on
-
ba
s
e
d
hi
e
r
a
r
c
hi
c
a
l
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
t
w
or
ks
f
or
M
O
O
C
f
or
um
pos
t
s
a
na
l
ys
i
s
,”
J
ou
r
nal
of
A
m
bi
e
nt
I
nt
e
l
l
i
ge
nc
e
and
H
um
ani
z
e
d
C
om
put
i
ng
,
vol
.
12,
no.
11,
pp.
9977
–
9989,
2021,
doi
:
10.1007/
s
12652
-
020
-
02747
-
9.
[
12]
Z.
K
a
s
t
r
a
t
i
,
A.
S.
I
m
r
a
n,
a
nd
A.
K
ur
t
i
,
“
W
e
a
kl
y
s
up
e
r
vi
s
e
d
f
r
a
m
e
w
or
k
f
or
a
s
pe
c
t
-
ba
s
e
d
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
on
s
t
ude
nt
s
’
r
e
vi
e
w
s
of
M
O
O
C
s
,
”
I
E
E
E
A
c
c
e
s
s
,
vol
.
8,
pp.
106799
–
106810,
2020,
doi
:
10.1109/
A
C
C
E
S
S
.2020.3000739.
[
13]
A.
O
na
n,
“
S
e
nt
i
m
e
nt
a
na
l
ys
i
s
on
m
a
s
s
i
ve
op
e
n
onl
i
ne
c
our
s
e
e
v
a
l
ua
t
i
ons
:
a
t
e
xt
m
i
ni
ng
a
nd
de
e
p
l
e
a
r
ni
ng
a
ppr
oa
c
h,”
C
om
put
e
r
A
ppl
i
c
at
i
ons
in
E
ngi
ne
e
r
i
ng
E
duc
at
i
on
,
vol
.
29,
no.
3,
pp.
572
–
589,
2021,
doi
:
10.1002/
c
a
e
.22253.
[
14]
M.
E
l
ba
w
a
b
a
nd
R.
H
e
nr
i
que
s
,
“
M
a
c
hi
ne
l
e
a
r
ni
ng
a
ppl
i
e
d
to
s
t
ude
nt
a
t
t
e
nt
i
ve
ne
s
s
de
t
e
c
t
i
on:
us
i
ng
e
m
ot
i
ona
l
a
nd
non
-
e
m
ot
i
ona
l
m
e
a
s
ur
e
s
,”
E
duc
at
i
on
and
I
nf
or
m
at
i
on
T
e
c
hnol
ogi
e
s
,
vol
.
28,
no.
12,
pp.
15717
–
15737,
2023,
doi
:
10.1007/
s
10639
-
023
-
11814
-
5.
[
15]
A.
B
a
qa
c
h
a
nd
A.
B
a
t
t
ou,
“A
ne
w
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
m
ode
l
to
c
l
a
s
s
i
f
y
s
t
ude
nt
s
’
r
e
vi
e
w
s
on
M
O
O
C
s
,”
E
duc
at
i
on
and
I
nf
or
m
at
i
on
T
e
c
hnol
ogi
e
s
,
pp.
1
–
28,
2024,
doi
:
10.1007/
s
10639
-
024
-
12526
-
0.
[
16]
B.
P.
B
ha
ga
t
a
nd
S.
S.
D
.
-
D
a
ndg
e
,
“A
c
om
pa
r
i
s
on
of
di
f
f
e
r
e
nt
m
a
c
hi
ne
l
e
a
r
n
i
ng
t
e
c
hni
que
s
f
or
s
e
nt
i
m
e
nt
a
na
l
y
s
i
s
in
e
duc
a
t
i
on
dom
a
i
n,”
C
ogni
t
i
v
e
Sc
i
e
nc
e
and
T
e
c
hnol
ogy
,
pp.
441
–
450,
2023,
doi
:
10.1007/
978
-
981
-
19
-
8086
-
2_43.
[
17]
N.
A
l
t
r
a
bs
he
h,
M.
C
oc
e
a
,
a
nd
S.
F
a
l
l
a
hkha
i
r
,
“
S
e
nt
i
m
e
nt
a
na
l
ys
i
s
:
t
ow
a
r
ds
a
t
ool
f
or
a
na
l
ys
i
ng
r
e
a
l
-
time
s
t
ude
nt
s
f
e
e
dba
c
k,”
P
r
oc
e
e
di
ngs
-
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
T
ool
s
w
i
t
h
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
,
I
C
T
A
I
,
vol
.
2014
-
D
e
c
e
m
,
pp.
419
–
423,
2014,
doi
:
10.1109/
I
C
T
A
I
.2014.70.
[
1
8]
G.
S.
C
ha
u
ha
n
,
P.
A
gr
a
w
a
l
,
a
n
d
Y.
K.
M
e
e
na
,
“
A
s
pe
c
t
-
ba
s
e
d
s
e
n
t
i
m
e
n
t
a
na
l
ys
i
s
of
s
t
u
de
n
t
s
’
f
e
e
dba
c
k
to
i
m
p
r
ov
e
t
e
a
c
hi
ng
–
l
e
a
r
ni
n
g
pr
oc
e
s
s
,”
S
m
ar
t
I
nn
ov
a
t
i
on
,
S
y
s
t
e
m
s
a
nd
T
e
c
hn
ol
ogi
e
s
,
vo
l
.
10
7,
pp
.
259
–
2
66,
2
0
19,
do
i
:
1
0.1
00
7/
978
-
9
81
-
13
-
17
47
-
7
_2
5.
[
19]
G.
G.
E
s
pa
r
z
a
et
al
.
,
“A
s
e
nt
i
m
e
nt
a
n
a
l
ys
i
s
m
ode
l
to
a
na
l
yz
e
s
t
ude
nt
s
’
r
e
vi
e
w
s
of
t
e
a
c
he
r
pe
r
f
or
m
a
nc
e
us
i
ng
s
uppor
t
ve
c
t
or
m
a
c
hi
ne
s
,”
A
dv
anc
e
s
in
I
nt
e
l
l
i
ge
nt
Sy
s
t
e
m
s
and
C
om
put
i
ng
,
vol
.
620,
pp.
157
–
164,
2018,
doi
:
10.1007/
978
-
3
-
319
-
62410
-
5_19.
[
20]
G.
G
ut
i
é
r
r
e
z
,
J.
P
onc
e
,
A.
O
c
hoa
,
a
nd
M.
Á
l
va
r
e
z
,
“
A
na
l
yz
i
ng
s
t
ude
nt
s
r
e
vi
e
w
s
of
t
e
a
c
he
r
pe
r
f
or
m
a
nc
e
us
i
ng
s
uppor
t
ve
c
t
or
m
a
c
hi
ne
s
by
a
pr
opos
e
d
m
ode
l
,”
C
om
m
uni
c
at
i
ons
in
C
om
put
e
r
and
I
nf
or
m
at
i
on
Sc
i
e
nc
e
,
vol
.
820,
pp.
113
–
122,
2018,
doi
:
10.1007/
978
-
3
-
319
-
76261
-
6_9.
[
21]
J.
A.
P.
L
a
l
a
t
a
,
B.
G
e
r
a
r
do,
a
nd
R.
M
e
di
na
,
“A
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
m
ode
l
f
or
f
a
c
ul
t
y
c
om
m
e
nt
e
va
l
ua
t
i
on
us
i
ng
e
ns
e
m
bl
e
m
a
c
hi
ne
l
e
a
r
ni
ng
a
l
gor
i
t
hm
s
,”
in
A
C
M
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
P
r
oc
e
e
di
ng
Se
r
i
e
s
,
201
9,
pp.
68
–
73,
doi
:
10.1145/
3341620.3341638.
[
22]
C.
P
.
-
I
nw
ong
a
nd
W.
S.
R
ungw
or
a
w
ut
,
“
T
e
a
c
hi
ng
s
e
nt
i
-
l
e
xi
c
on
f
or
a
ut
om
a
t
e
d
s
e
nt
i
m
e
nt
pol
a
r
i
t
y
de
f
i
ni
t
i
on
in
t
e
a
c
hi
ng
e
va
l
ua
t
i
on,”
in
P
r
oc
e
e
di
ngs
-
2014
10t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
Se
m
ant
i
c
s
,
K
now
l
e
dge
and
G
r
i
ds
,
SK
G
2014
,
2014,
pp.
84
–
91,
doi
:
10.1109/
S
K
G
.2014.25.
[
23]
W.
K
a
ur
,
V.
B
a
l
a
kr
i
s
hna
n,
a
nd
B.
S
i
ngh,
“
I
m
p
r
ovi
ng
t
e
a
c
hi
ng
a
nd
l
e
a
r
ni
ng
e
xpe
r
i
e
nc
e
in
e
ngi
ne
e
r
i
ng
e
duc
a
t
i
on
us
i
ng
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
t
e
c
hni
que
s
,”
I
O
P
C
onf
e
r
e
nc
e
Se
r
i
e
s
:
M
at
e
r
i
al
s
Sc
i
e
nc
e
and
E
ngi
ne
e
r
i
ng
,
vol
.
834,
no.
1,
2020,
doi
:
10.1088/
1757
-
899X
/
834/
1/
012026.
[
24]
N.
K.
S
i
ngh,
D.
S.
T
om
a
r
,
a
nd
A.
K.
S
a
nga
i
a
h,
“
S
e
nt
i
m
e
nt
a
na
l
y
s
i
s
:
a
r
e
vi
e
w
a
n
d
c
om
pa
r
a
t
i
ve
a
na
l
y
s
i
s
ov
e
r
s
oc
i
a
l
m
e
di
a
,”
J
our
na
l
of
A
m
bi
e
nt
I
nt
e
l
l
i
ge
nc
e
and
H
um
ani
z
e
d
C
om
put
i
ng
,
vol
.
11,
no.
1,
pp.
97
–
117,
2020,
doi
:
10.1007/
s
12652
-
018
-
0862
-
8.
[
25]
P.
P.
K
um
a
r
,
P.
G
ur
una
dh,
a
nd
B.
M
a
r
a
m
,
“A
s
t
udy
on
s
e
nt
i
m
e
nt
a
l
a
na
l
ys
i
s
us
i
ng
m
a
c
hi
ne
l
e
a
r
ni
ng
a
ppr
oa
c
h,”
J
our
nal
of
A
dv
anc
e
d
R
e
s
e
ar
c
h
in
A
ppl
i
e
d
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
and
N
e
ur
al
N
e
t
w
or
k
,
vol
.
5,
no.
1,
p
p.
1
–
4,
2021.
[
26]
Y.
C
ha
ndr
a
a
nd
A.
J
a
na
,
“
S
e
nt
i
m
e
nt
a
na
l
y
s
i
s
u
s
i
ng
m
a
c
hi
n
e
l
e
a
r
ni
ng
a
nd
de
e
p
l
e
a
r
ni
ng,”
in
P
r
oc
e
e
di
ngs
of
t
he
7t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
C
om
put
i
ng
f
or
Sus
t
ai
nabl
e
G
l
obal
D
e
v
e
l
opm
e
nt
,
I
N
D
I
A
C
om
2020
,
2020,
pp.
1
–
4,
doi
:
10.23919/
I
N
D
I
A
C
om
49435.2020.9083703.
[
27]
J.
G.
K.
M
a
bunda
,
A.
J
a
dha
v,
a
nd
R.
A
j
oodha
,
“
S
e
nt
i
m
e
nt
a
na
l
ys
i
s
of
s
t
ude
nt
t
e
xt
ua
l
f
e
e
dba
c
k
to
i
m
pr
ove
t
e
a
c
hi
ng,
”
I
nt
e
r
di
s
c
i
pl
i
nar
y
R
e
s
e
a
r
c
h
in
T
e
c
hnol
ogy
and
M
anage
m
e
nt
,
pp.
643
–
651,
2021,
doi
:
10.1201/
9781003202240
-
100.
[
28]
Z.
N
a
s
i
m
,
Q.
R
a
j
put
,
a
nd
S.
H
a
i
de
r
,
“
S
e
nt
i
m
e
nt
a
na
l
ys
i
s
of
s
t
ude
nt
f
e
e
dba
c
k
us
i
ng
m
a
c
hi
ne
l
e
a
r
ni
ng
a
nd
l
e
xi
c
on
ba
s
e
d
a
ppr
oa
c
he
s
,
”
in
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
R
e
s
e
ar
c
h
and
I
nnov
at
i
on
in
I
nf
or
m
at
i
on
Sy
s
t
e
m
s
,
I
C
R
I
I
S
,
2017,
pp.
1
–
6,
doi
:
10.1109/
I
C
R
I
I
S
.2017.8002475.
[
29]
S.
Q
a
i
s
e
r
,
N.
Y
us
of
f
,
M.
R
e
m
l
i
,
a
nd
H.
K.
A
dl
i
,
“A
c
om
pa
r
i
s
on
of
m
a
c
hi
ne
l
e
a
r
ni
ng
t
e
c
hni
que
s
f
or
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
,”
T
ur
k
i
s
h
J
our
nal
of
C
om
put
e
r
and
M
at
he
m
at
i
c
s
E
duc
at
i
on
(
T
U
R
C
O
M
A
T
)
,
vol
.
12,
no.
3,
pp.
1738
–
1744,
2021,
doi
:
10.17762/
t
ur
c
om
a
t
.v12i
3.999.
[
30]
V.
D.
N
guye
n,
K.
V
.
N
guye
n,
a
nd
N.
L.
T.
N
guye
n,
“
V
a
r
i
a
nt
s
of
l
ong
s
hor
t
-
t
e
r
m
m
e
m
or
y
f
or
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
on
V
i
e
t
na
m
e
s
e
s
t
ude
nt
s
’
f
e
e
dba
c
k
c
or
pus
,”
P
r
oc
e
e
di
ngs
of
2018
10t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
K
now
l
e
dge
and
Sy
s
t
e
m
s
E
ngi
ne
e
r
i
ng,
K
SE
2018
,
pp.
306
–
311,
2018,
doi
:
10.1109/
K
S
E
.2018.8573351.
[
31]
R.
Z.
C
a
ba
da
,
M.
L.
B.
E
s
t
r
a
da
,
a
nd
R.
O.
B
u
s
t
i
l
l
os
,
“
M
i
ni
ng
of
e
duc
a
t
i
ona
l
opi
ni
ons
w
i
t
h
de
e
p
l
e
a
r
ni
ng,”
J
our
nal
of
U
ni
v
e
r
s
al
C
om
put
e
r
Sc
i
e
n
c
e
,
vol
.
24,
no.
11,
pp.
1604
–
1626,
2018.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4747
-
4756
4756
[
32]
I.
S
i
ndhu,
S.
M
.
D
a
udpot
a
,
K.
B
a
da
r
,
M.
B
a
kht
ya
r
,
J.
B
a
be
r
,
a
nd
M.
N
ur
un
na
bi
,
“
A
s
pe
c
t
-
ba
s
e
d
opi
ni
on
m
i
ni
ng
on
s
t
ude
nt
’
s
f
e
e
dba
c
k
f
or
f
a
c
ul
t
y
t
e
a
c
hi
ng
pe
r
f
or
m
a
nc
e
e
va
l
ua
t
i
on,”
I
E
E
E
A
c
c
e
s
s
,
vol
.
7,
pp.
108729
–
108741,
2019,
doi
:
10.1109/
A
C
C
E
S
S
.2019.2928872.
[
33]
A.
A
bdi
,
G.
S
e
dr
a
kya
n,
B.
V
e
l
dka
m
p,
J.
V.
H
i
l
l
e
ge
r
s
be
r
g,
a
nd
S.
M.
V.
D.
B
e
r
g,
“
S
t
ude
nt
s
f
e
e
dba
c
k
a
na
l
ys
i
s
m
ode
l
u
s
i
ng
de
e
p
l
e
a
r
ni
ng
-
ba
s
e
d
m
e
t
hod
a
nd
l
i
ngui
s
t
i
c
know
l
e
dge
f
or
i
n
t
e
l
l
i
ge
nt
e
duc
a
t
i
ona
l
s
ys
t
e
m
s
,”
Sof
t
C
om
put
i
ng
,
vol
.
27,
no.
19,
pp.
14073
–
14094,
2023,
doi
:
10.1007/
s
00500
-
023
-
07926
-
2.
[
34]
I.
A.
K
a
ndhr
o,
S.
W
a
s
i
,
K.
K
um
a
r
,
M.
R
i
nd,
a
nd
M.
A
m
e
e
n,
“
S
e
nt
i
m
e
nt
a
n
a
l
ys
i
s
of
s
t
ude
nt
s
’
c
om
m
e
nt
us
i
ng
l
ong
-
s
hor
t
t
e
r
m
m
ode
l
,”
I
ndi
an
J
our
nal
of
Sc
i
e
nc
e
and
T
e
c
hnol
ogy
,
vol
.
12,
no.
8,
pp.
1
–
16,
20
19,
doi
:
10.17485/
i
j
s
t
/
2019/
v12i
8/
141741.
[
35]
T.
M
i
kol
ov,
K.
C
he
n,
G.
C
or
r
a
do,
a
nd
J.
D
e
a
n,
“
E
f
f
i
c
i
e
nt
e
s
t
i
m
a
t
i
on
of
w
or
d
r
e
pr
e
s
e
nt
a
t
i
ons
in
ve
c
t
or
s
pa
c
e
,”
ar
X
i
v
-
C
om
put
e
r
Sc
i
e
nc
e
,
pp. 1
-
12,
2013
,
doi
:
10.48550/
a
r
X
i
v.1301.3781
.
[
36]
L.
B
r
e
i
m
a
n,
“
R
a
ndom
f
or
e
s
t
s
,”
in
M
ac
hi
ne
l
e
ar
ni
ng
,
vol
.
45,
pp.
5
–
32,
2001
, d
oi
:
10.1023/
A
:
1010933404324
.
[
37]
K.
O
’
s
he
a
a
nd
R.
N
a
s
h,
“
A
n
i
nt
r
oduc
t
i
on
to
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
ks
,”
ar
X
i
v
-
C
om
put
e
r
S
c
i
e
nc
e
,
pp.
1
-
11,
2015
,
doi
:
10.48550/
a
r
X
i
v.1511.08458
.
[
38]
S.
H
oc
hr
e
i
t
e
r
a
nd
J.
S
c
hm
i
dhube
r
,
“
L
ong
s
hor
t
-
t
e
r
m
m
e
m
o
r
y,”
N
e
ur
al
c
om
put
at
i
on
,
vol
.
9,
no.
8,
pp.
1735
–
1780,
1997
,
doi
:
10.1162/
ne
c
o.1997.9.8.1735
.
[
39]
A.
V
a
s
w
a
ni
et
al
.
,
“
A
t
t
e
nt
i
on
is
a
l
l
you
ne
e
d,”
31s
t
C
onf
e
r
e
nc
e
on
N
e
ur
al
I
n
f
or
m
at
i
on
P
r
oc
e
s
s
i
ng
Sy
s
t
e
m
s
(
N
I
P
S
2017)
,
L
ong
B
e
a
c
h, C
A
,
U
S
A
,
vol
.
30,
pp. 1
-
11,
2017.
[
40]
H.
A
bba
s
i
m
e
hr
a
nd
R.
P
a
ki
,
“
I
m
pr
ovi
ng
time
s
e
r
i
e
s
f
or
e
c
a
s
t
i
ng
us
i
ng
L
S
T
M
a
nd
a
t
t
e
nt
i
on
m
ode
l
s
,”
J
our
nal
of
A
m
bi
e
nt
I
nt
e
l
l
i
ge
nc
e
and
H
um
ani
z
e
d
C
om
put
i
ng
,
vol
.
13,
no.
1,
pp.
673
–
691,
2022,
doi
:
10.1007/
s
1
2652
-
020
-
02761
-
x.
[
41]
B.
L
i
u,
“
T
e
xt
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
ba
s
e
d
on
C
B
O
W
m
ode
l
a
nd
de
e
p
l
e
a
r
ni
ng
in
bi
g
da
t
a
e
nvi
r
onm
e
nt
,”
J
our
nal
of
A
m
bi
e
nt
I
nt
e
l
l
i
ge
nc
e
and
H
um
ani
z
e
d
C
om
put
i
ng
,
vol
.
11,
no.
2,
pp.
451
–
458,
2020,
doi
:
10.1007/
s
12652
-
018
-
1095
-
6.
B
I
O
G
R
A
P
H
I
E
S
OF
A
U
T
H
O
R
S
Bhavana
P.
Bhagat
working
as
Sr.
Lecturer
in
Department
of
Computer
Engineering
Government
Polytec
hnic
Yavatmal
(M.S
.
)
,
India.
She
is
Resear
ch
Scholar
in
Department
of
Computer
Scienc
e
and
Engineering
Sipna
Colle
ge
of
Engineering
and
Technology,
Amravati
,
India
and
affiliated
to
Sant
Gadge
Baba
Amra
vati
University
Amravati
(M.S
.
),
India.
She
has
26
years
of
teaching
experi
ence.
Her
area
of
s
pecialization
is
artificial
intelligence
,
machine
learning
,
and
natural
languge
processing
.
She
h
as
presented
papers
in
4
internationa
l
and
2
national
conferences.
She
can
be
contacted
at
email:
bhavana.bhagat30475@gmail.com
.
Sheeta
l
S.
Dhande
-
Dandge
working
as
Profes
sor
at
Departmen
t
of
Computer
Scienc
e
Engineering
and
Technology
Amravati.
She
has
Completed
P
h.
D
.
from
Profes
sor
Ram
Meghe
Institut
e
of
Resear
ch
and
Technology
Amravati
(M.S
.
),
India
affiliated
to
Sant
Gadge
Baba
Amravati
Univ
ersity
Amravati.
(M.S
.
),
India.
She
has
24
years
of
teaching
and
experience.
Her
area
of
specialization
is
artificial
intelligence
.
She
has
publis
hed
over
63
papers
in
internationa
l
journals
and
conferences
.
She
can
be
contacted
at
email:
sheetaldhande
dandge@gmail.com
.
Sandeep
R.
Raut
working
as
IB
Computer
Scienc
e
teacher
in
S
hanghai
Qibao
Dwight
High
Scho
ol,
Shangha
i,
China.
He
has
14
years
of
t
eaching
ex
perience.
He
participated
recently
in
“
Going
deeper
with
technology
in
education
and
social
media
"
(EARCOS)
”
,
and
“
Teaching
ESL
students
in
mainstream
classrooms
(TESMC)
”
,
which
f
ocuses
on
how
to
engage
and
support
students,
by
taking
into
considerat
ion
their
social,
devel
opmental,
and
language
needs.
He
can
be
contacted
at
email:
sraut@
qibaodwight.org
.
Evaluation Warning : The document was created with Spire.PDF for Python.