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
.
3
,
J
u
n
e
20
25
,
p
p
.
3332
~
3
3
4
5
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
3
.
pp
3
3
3
2
-
3
3
4
5
3332
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Ara
bic
o
ff
ensiv
e t
ex
t
clas
sifica
tion
using
emo
jis
:
In
cl
uding
emo
ji
da
ta in
Ara
bic
na
tural la
ng
u
a
g
e process
ing
Am
a
l A
lba
la
wi
1,
2
,
Wa
el
M
.
S.
Ya
f
o
o
z
1
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
,
C
o
l
l
e
g
e
o
f
C
o
mp
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
,
Ta
i
b
a
h
U
n
i
v
e
r
si
t
y
,
M
a
d
i
n
a
h
,
S
a
u
d
i
A
r
a
b
i
a
2
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
C
o
l
l
e
g
e
o
f
C
o
m
p
u
t
e
r
a
n
d
C
y
b
e
r
S
c
i
e
n
c
e
s,
U
n
i
v
e
r
si
t
y
o
f
P
r
i
n
c
e
M
u
g
r
i
n
,
M
a
d
i
n
a
h
,
S
a
u
d
i
A
r
a
b
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ap
r
2
1
,
2
0
2
4
R
ev
is
ed
J
an
1
5
,
2
0
2
5
Acc
ep
ted
Ma
r
3
,
2
0
2
5
In
th
e
d
ig
i
tal
so
c
ial
m
e
d
ia
e
c
o
sy
st
e
m
,
c
o
n
tro
ll
in
g
o
ffe
n
siv
e
lan
g
u
a
g
e
re
q
u
ires
a
d
v
a
n
c
e
d
a
lg
o
r
it
h
m
ic
to
o
ls.
Th
is
stu
d
y
e
x
a
m
in
e
s
th
e
in
fl
u
e
n
c
e
o
f
in
c
l
u
d
i
n
g
e
m
o
ji
s
tran
sla
ti
o
n
in
th
e
tex
t
p
r
e
p
ro
c
e
ss
in
g
sta
g
e
o
f
t
h
e
c
las
sifica
ti
o
n
o
f
o
ffe
n
siv
e
Ara
b
ic
tex
t
.
A
n
o
v
e
l
d
a
t
a
se
t
o
f
1
0
,
0
0
0
Ara
b
ic
twe
e
ts
wa
s
d
e
v
e
lo
p
e
d
,
with
ri
g
o
r
o
u
s
a
n
n
o
tatio
n
s
t
o
c
las
sify
c
o
n
ten
t
a
s
o
ffe
n
si
v
e
o
r
n
o
n
-
o
ffe
n
siv
e
.
Th
e
d
a
tas
e
t
wa
s
m
e
ti
c
u
lo
u
sl
y
a
n
n
o
tate
d
a
n
d
v
a
li
d
a
ted
u
sin
g
C
o
h
e
n
'
s
k
a
p
p
a
(CK)
a
n
d
Krip
p
e
n
d
o
rff'
s
Alp
h
a
(α
)
t
o
e
n
s
u
re
c
o
n
siste
n
c
y
a
n
d
a
c
c
u
ra
c
y
.
S
e
v
e
ra
l
e
x
p
e
rime
n
ts
e
v
a
lu
a
ted
th
e
d
a
tas
e
t
wit
h
t
h
e
m
o
st
c
o
m
m
o
n
tex
t
c
las
sifica
ti
o
n
m
o
d
e
ls:
se
v
e
n
m
a
c
h
in
e
lea
rn
in
g
(
M
L)
c
las
sifiers
a
n
d
t
h
re
e
d
e
e
p
lea
rn
in
g
(DL)
m
o
d
e
ls.
Two
e
x
p
e
ri
m
e
n
tal
se
ts
we
re
c
o
n
d
u
c
ted
:
o
n
e
with
e
m
o
j
i
tran
sla
ti
o
n
in
p
re
p
ro
c
e
ss
in
g
to
e
n
rich
tex
t
in
p
u
t
a
n
d
a
n
o
t
h
e
r
wit
h
o
u
t
e
m
o
ji
tran
sla
ti
o
n
t
o
d
irec
tl
y
a
ss
e
ss
th
e
i
m
p
a
c
t
o
f
e
m
o
ji
s
o
n
c
las
sifica
ti
o
n
a
c
c
u
ra
c
y
.
Th
e
fin
d
in
g
s
i
n
d
ica
te
th
a
t
e
m
o
ji
s
sig
n
if
ica
n
tl
y
a
ffe
c
t
tex
t
c
las
sifica
ti
o
n
m
o
d
e
ls,
with
a
d
v
a
n
c
e
d
DL
m
o
d
e
ls
sh
o
wi
n
g
h
ig
h
e
r
se
n
siti
v
it
y
to
c
o
n
tex
tu
a
l
n
u
a
n
c
e
s
c
o
n
v
e
y
e
d
b
y
e
m
o
ji
s
c
o
m
p
a
re
d
to
tra
d
it
i
o
n
a
l
M
L
c
las
sifiers
.
Th
i
s
re
se
a
rc
h
h
ig
h
l
ig
h
ts
t
h
e
d
u
a
l
ro
le
o
f
e
m
o
ji
s,
wh
ic
h
a
re
o
f
ten
li
n
k
e
d
t
o
p
o
siti
v
e
e
m
o
ti
o
n
s
a
n
d
o
ffe
n
siv
e
c
o
n
tex
ts,
a
d
d
i
n
g
c
o
m
p
lex
it
y
t
o
d
i
g
it
a
l
c
o
m
m
u
n
ica
ti
o
n
.
It
c
o
n
tri
b
u
tes
to
th
e
d
e
v
e
lo
p
m
e
n
t
o
f
m
o
re
a
c
c
u
ra
te
a
n
d
c
o
n
tex
t
-
se
n
sit
iv
e
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
(NL
P
)
t
o
o
ls.
K
ey
w
o
r
d
s
:
Ar
ab
ic
tex
t c
lass
if
icatio
n
Dee
p
lear
n
in
g
e
m
o
jis
an
aly
s
is
Ma
ch
in
e
lear
n
in
g
Natu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
Of
f
en
s
iv
e
lan
g
u
a
g
e
d
etec
tio
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
:
Am
al
Alb
alaw
i
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
C
o
lleg
e
o
f
C
o
m
p
u
ter
Sci
en
ce
an
d
E
n
g
in
ee
r
in
g
,
T
ai
b
ah
Un
iv
er
s
ity
Ma
d
in
ah
,
Sau
d
i A
r
a
b
ia
E
m
ail:
Am
alM
o
h
Alb
alaw
i@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Du
e
to
th
e
g
r
o
wth
o
f
s
o
cial
n
e
two
r
k
s
in
r
ec
en
t
y
ea
r
s
,
p
e
o
p
le
ar
e
n
o
w
m
o
r
e
co
n
n
ec
ted
t
o
o
n
e
an
o
th
er
,
allo
win
g
f
o
r
la
r
g
e
-
s
ca
le,
r
e
al
-
tim
e
c
o
m
m
u
n
i
c
a
t
i
o
n
o
n
a
w
o
r
l
d
w
i
d
e
s
c
al
e
[
1
]
,
[
2
]
.
T
h
i
s
u
n
r
e
s
t
r
i
c
te
d
c
o
m
m
u
n
i
c
a
t
i
o
n
h
a
s
f
a
c
i
li
t
a
te
d
t
h
e
e
x
c
h
a
n
g
e
o
f
d
i
v
e
r
s
e
i
d
e
a
s
,
em
o
t
i
o
n
s
,
a
n
d
i
n
f
o
r
m
a
t
i
o
n
.
H
o
w
e
v
e
r
,
t
h
e
a
n
o
n
y
m
i
t
y
o
f
f
e
r
e
d
b
y
t
h
e
s
e
p
l
a
t
f
o
r
m
s
a
l
s
o
r
a
i
s
e
s
c
o
n
c
e
r
n
s
a
b
o
u
t
p
o
t
e
n
t
i
al
m
i
s
u
s
e
a
n
d
t
h
e
s
p
r
e
a
d
o
f
o
f
f
e
n
s
i
v
e
l
a
n
g
u
a
g
e
[
2
]
.
T
h
i
s
h
as
l
e
d
t
o
a
n
i
n
c
r
e
as
e
i
n
h
a
t
e
s
p
e
e
c
h
a
n
d
c
y
b
e
r
b
u
l
l
y
i
n
g
o
n
s
o
c
i
a
l
m
e
d
i
a
p
l
a
t
f
o
r
m
s
[
3
]
.
T
o
a
d
d
r
e
s
s
t
h
is
p
r
o
b
l
e
m
,
s
o
m
e
n
a
ti
o
n
s
h
a
v
e
t
a
k
e
n
m
e
a
s
u
r
es
t
o
p
r
o
h
i
b
it
t
h
e
p
r
o
li
f
e
r
a
t
i
o
n
o
f
h
a
te
s
p
ee
c
h
o
n
s
o
cia
l
m
e
d
i
a
p
l
a
t
f
o
r
m
s
.
I
n
2
0
0
7
,
S
a
u
d
i
A
r
a
b
i
a
i
m
p
l
e
m
e
n
t
e
d
l
e
g
a
l
m
e
as
u
r
e
s
t
o
r
e
g
u
l
ate
i
n
f
o
r
m
a
t
i
o
n
m
e
d
i
u
m
s
a
n
d
t
ec
h
n
o
l
o
g
i
e
s
u
s
e
d
f
o
r
i
n
s
u
l
ti
n
g
,
d
e
f
a
m
i
n
g
,
a
n
d
s
la
n
d
e
r
i
n
g
o
t
h
e
r
s
,
p
a
r
t
ic
u
l
a
r
l
y
w
h
e
n
s
u
c
h
a
c
ti
o
n
s
h
a
r
m
i
n
d
i
v
i
d
u
a
l
s
o
r
v
i
o
l
at
e
t
h
eir
p
r
i
v
a
c
y
.
T
h
i
s
f
r
a
m
e
w
o
r
k
a
d
d
r
e
s
s
e
s
t
h
e
u
s
e
o
f
c
o
n
t
e
m
p
o
r
a
r
y
i
n
f
o
r
m
a
t
i
o
n
d
is
s
e
m
i
n
a
ti
o
n
t
ec
h
n
o
l
o
g
i
e
s
i
n
t
h
es
e
c
o
n
t
e
x
t
s
[
4
]
.
I
n
an
o
th
er
e
x
a
m
p
le,
th
e
Netwo
r
k
E
n
f
o
r
ce
m
en
t
Act
was
en
ac
ted
in
G
er
m
an
y
in
2
0
1
7
[
5
]
.
Mo
r
eo
v
er
,
o
f
f
en
s
iv
e
lan
g
u
ag
e
h
as
b
ee
n
th
e
tar
g
et
o
f
o
n
g
o
in
g
leg
is
lativ
e
ch
a
n
g
es,
a
n
d
cu
tt
in
g
-
ed
g
e
tech
n
ical
s
o
lu
tio
n
s
ar
e
b
ein
g
in
v
esti
g
ate
d
to
h
elp
s
o
cial
m
ed
ia
p
latf
o
r
m
s
an
d
o
th
er
o
r
g
an
izatio
n
s
t
o
en
f
o
r
ce
th
ese
laws
[
6
]
.
T
h
e
liter
atu
r
e
o
n
o
f
f
en
s
iv
e
lan
g
u
ag
e
d
etec
tio
n
in
Ar
ab
i
c
tex
ts
in
clu
d
es
a
wid
e
r
an
g
e
o
f
s
tu
d
ies,
ea
ch
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
A
r
a
b
ic
o
ffen
s
ive
text
cla
s
s
ific
a
tio
n
u
s
in
g
em
o
jis
:
I
n
clu
d
i
n
g
emo
ji d
a
ta
in
A
r
a
b
ic
…
(
A
ma
l
A
lb
a
la
w
i
)
3333
wh
ich
co
n
tr
ib
u
tes
u
n
iq
u
e
in
s
ig
h
ts
o
n
th
e
p
r
ep
r
o
ce
s
s
in
g
,
an
a
ly
s
is
,
an
d
class
if
icatio
n
o
f
th
is
co
n
ten
t
o
n
v
ar
i
o
u
s
d
ig
ital
p
latf
o
r
m
s
.
A
s
ig
n
if
ica
n
t
lim
itatio
n
h
as
b
ee
n
th
e
lac
k
o
f
r
o
b
u
s
t
p
u
b
lic
d
atasets
f
o
r
Ar
ab
ic,
wh
ich
ar
e
ess
en
tial
f
o
r
ef
f
ec
tiv
e
co
m
p
u
t
atio
n
al
an
aly
s
is
[
6
]
.
T
h
is
g
a
p
h
ig
h
lig
h
ts
th
e
cr
itical
n
ee
d
f
o
r
r
o
b
u
s
t
d
atasets
th
at
ca
n
b
etter
r
e
p
r
esen
t
th
e
lin
g
u
is
tic
co
m
p
lex
ities
o
f
Ar
ab
ic.
Sev
er
al
s
tu
d
ies
h
av
e
p
r
o
p
o
s
e
d
d
iv
er
s
e
d
atasets
,
s
tar
tin
g
with
s
m
all
co
llectio
n
s
,
s
u
ch
as
th
e
1
,
0
0
0
twee
ts
an
a
ly
ze
d
b
y
AlGh
am
d
i
a
n
d
Kh
an
[
7
]
to
lar
g
e
r
s
ets,
s
u
ch
as
th
e
1
5
,
0
5
0
-
co
m
m
en
t
d
ataset
u
s
ed
b
y
Alak
r
o
t
et
a
l.
[
8
]
.
T
h
ese
d
atasets
co
v
er
m
u
ltip
le
lan
g
u
a
g
es,
in
clu
d
in
g
Ar
a
b
ic,
an
d
o
r
ig
in
ate
f
r
o
m
d
i
v
er
s
e
s
o
u
r
ce
s
s
u
ch
as
T
witter
,
Yo
u
T
u
b
e,
a
n
d
o
th
er
s
o
cial
m
ed
i
a
p
latf
o
r
m
s
,
r
ef
lectin
g
th
e
g
lo
b
a
l,
m
u
ltil
in
g
u
al
ch
allen
g
es o
f
m
o
d
er
atin
g
o
f
f
en
s
iv
e
co
n
ten
t A
l
Gh
am
d
i a
n
d
Kh
an
[
7
]
an
d
Alak
r
o
t
et
a
l.
[
8
]
.
T
h
e
p
r
ep
r
o
ce
s
s
in
g
m
eth
o
d
s
u
s
ed
v
ar
y
wid
ely
in
th
ese
s
tu
d
ies,
wi
th
tech
n
iq
u
es
s
u
ch
as
r
o
o
t
-
b
ased
s
tem
m
in
g
,
n
-
g
r
a
m
m
o
d
els,
lig
h
t
s
tem
m
in
g
,
a
n
d
lem
m
atiza
tio
n
co
m
m
o
n
ly
u
s
ed
to
o
p
tim
ize
tex
t
d
ata
f
o
r
s
u
b
s
eq
u
en
t
an
aly
s
is
A
l
-
Saif
an
d
Al
-
Do
s
s
ar
i
[
9
]
an
d
Fo
u
n
ta
et
a
l.
[
1
0
]
.
T
h
is
in
itial st
ep
is
ess
en
tial
f
o
r
ef
f
ec
tiv
e
m
o
d
el
tr
ain
in
g
a
n
d
th
e
ac
cu
r
ate
class
if
icatio
n
o
f
co
n
ten
t.
I
n
ter
m
s
o
f
m
o
d
el
d
ep
lo
y
m
en
t,
s
ev
er
al
wo
r
d
r
ep
r
esen
tatio
n
tech
n
o
lo
g
ies
wer
e
u
s
ed
,
in
clu
d
in
g
Glo
Ve,
f
a
s
tTe
x
t,
b
ag
o
f
wo
r
d
s
(
B
o
W
)
a
n
d
ter
m
f
r
eq
u
e
n
cy
-
in
v
er
s
e
d
o
cu
m
e
n
t
f
r
eq
u
en
cy
(
T
F
-
I
DF)
,
wh
ich
f
ac
ilit
ate
th
e
t
r
an
s
f
er
o
f
tex
t
in
to
f
o
r
m
ats
th
a
t
ca
n
b
e
ef
f
icien
tly
p
r
o
ce
s
s
ed
b
y
m
ac
h
in
e
lear
n
in
g
(
ML
)
class
if
ier
s
an
d
d
ee
p
lear
n
in
g
(
DL
)
m
o
d
els
.
T
h
e
u
lt
im
ate
g
o
al
o
f
th
ese
r
esear
ch
ef
f
o
r
ts
is
to
cr
ea
te
au
to
m
ated
s
y
s
tem
s
th
at
ar
e
ab
le
to
ef
f
ec
tiv
ely
d
etec
t
a
n
d
class
if
y
v
a
r
io
u
s
f
o
r
m
s
o
f
o
f
f
en
s
iv
e
c
o
n
ten
t,
s
u
c
h
as
cy
b
er
b
u
lly
in
g
,
h
ate
s
p
ee
ch
,
a
n
d
o
th
er
s
u
s
p
icio
u
s
m
ess
ag
es.
So
m
e
s
tu
d
ies
s
u
g
g
est
en
h
an
cin
g
class
if
ier
p
er
f
o
r
m
a
n
ce
b
y
a
d
o
p
tin
g
DL
tech
n
i
q
u
e
s
o
r
ex
p
a
n
d
in
g
th
e
tr
ain
i
n
g
d
ataset,
wh
ich
is
ev
id
en
ce
o
f
o
n
g
o
in
g
in
n
o
v
atio
n
s
in
th
is
f
ield
Fk
ih
et
a
l.
[
1
1
]
.
A
v
ar
iety
o
f
co
m
p
u
tatio
n
al
tech
n
iq
u
es
h
av
e
b
ee
n
u
s
ed
f
o
r
o
f
f
e
n
s
iv
e
lan
g
u
ag
e
d
et
ec
tio
n
an
d
class
if
icatio
n
.
Am
o
n
g
th
ese
ap
p
r
o
ac
h
es,
b
o
th
tr
a
d
itio
n
al
ML
alg
o
r
ith
m
s
an
d
ad
v
a
n
ce
d
DL
tec
h
n
iq
u
es,
in
clu
d
in
g
t
r
an
s
f
o
r
m
e
r
m
o
d
els,
h
av
e
p
r
o
v
e
n
to
b
e
o
f
g
r
ea
t
b
e
n
ef
it.
S
tu
d
ies
s
u
ch
as
A
l
-
Saif
an
d
Al
-
Do
s
s
ar
i
[
9
]
an
d
AlGh
am
d
i
a
n
d
K
h
an
[
7
]
l
ev
er
ag
ed
tr
ad
itio
n
al
ML
m
eth
o
d
s
s
u
ch
as
s
u
p
p
o
r
t
v
ec
to
r
m
a
ch
in
es
(
SVM)
an
d
d
ec
is
io
n
tr
ee
s
to
class
if
y
d
if
f
e
r
en
t
ty
p
es
o
f
Ar
ab
ic
twee
ts
in
t
o
ca
teg
o
r
ies
s
u
ch
as
h
ate
s
p
ee
ch
an
d
c
y
b
er
c
r
im
e.
T
h
ese
ap
p
r
o
ac
h
es
h
ig
h
lig
h
t
th
e
ad
ap
tab
ilit
y
o
f
ML
in
d
ea
lin
g
with
th
e
s
u
b
tle
co
m
p
lex
ities
o
f
d
ig
ital
co
m
m
u
n
icatio
n
d
ata
[
7
]
,
[
9
]
.
On
th
e
DL
f
r
o
n
t,
Fo
n
ta
et
a
l.
[
1
0
]
u
s
ed
a
u
n
if
ied
ar
ch
itect
u
r
e
th
at
in
teg
r
ates
ch
ar
ac
ter
-
lev
el
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
an
d
wo
r
d
-
lev
el
r
ec
u
r
r
e
n
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs)
to
r
o
b
u
s
tly
d
etec
t
o
f
f
en
s
iv
e
co
n
t
en
t
o
n
v
a
r
io
u
s
s
o
cial
m
ed
ia
p
l
atf
o
r
m
s
,
d
em
o
n
s
tr
atin
g
th
e
d
e
p
th
with
wh
ich
DL
m
o
d
els
ca
n
ex
tr
ac
t
an
d
lear
n
f
r
o
m
tex
t
d
ata.
Sig
n
if
ican
t
p
r
o
g
r
ess
in
th
e
ap
p
licatio
n
o
f
D
L
in
t
h
is
f
ield
was
h
ig
h
lig
h
ted
b
y
th
e
wo
r
k
o
f
Al
-
Sh
aa
lan
a
n
d
Al
-
Kh
alif
a
[
1
2
]
,
wh
o
u
s
ed
th
e
b
id
ir
ec
tio
n
al
en
c
o
d
er
r
ep
r
esen
tatio
n
s
f
r
o
m
tr
an
s
f
o
r
m
er
s
(
B
E
R
T
)
m
o
d
el,
a
s
tate
-
of
-
th
e
-
ar
t
tr
an
s
f
o
r
m
er
m
o
d
el.
B
E
R
T
,
wh
ich
s
tan
d
s
f
o
r
b
id
ir
ec
tio
n
al
en
co
d
in
g
r
ep
r
esen
tatio
n
tr
a
n
s
f
o
r
m
er
s
,
ex
ce
ls
at
u
n
d
er
s
tan
d
in
g
th
e
f
u
ll
co
n
tex
t
o
f
a
s
en
ten
c
e
b
y
lo
o
k
i
n
g
at
th
e
tex
t
s
u
r
r
o
u
n
d
in
g
ea
c
h
wo
r
d
.
T
h
is
is
p
ar
ticu
lar
ly
u
s
ef
u
l
f
o
r
d
e
tectin
g
h
ate
s
p
ee
ch
b
ec
a
u
s
e
th
e
co
n
tex
t
in
wh
ich
wo
r
d
s
ar
e
u
s
ed
ca
n
f
u
n
d
am
e
n
t
ally
ch
an
g
e
th
eir
m
ea
n
in
g
[
1
3
]
.
T
h
eir
ap
p
r
o
ac
h
e
x
em
p
lifie
s
th
e
s
tate
-
of
-
th
e
-
ar
t
ca
p
ab
ilit
ies
o
f
tr
an
s
f
o
r
m
er
m
o
d
els
in
ca
p
tu
r
in
g
co
m
p
lex
l
in
g
u
is
tic
n
u
an
ce
s
an
d
im
p
r
o
v
in
g
class
if
icatio
n
ac
cu
r
ac
y
.
T
h
e
wo
r
k
o
f
Mu
b
ar
ak
et
a
l.
[
1
4
]
p
r
esen
ts
an
in
n
o
v
ativ
e
a
p
p
r
o
ac
h
to
u
s
in
g
em
o
jis
as
a
n
ch
o
r
s
to
ef
f
ec
tiv
ely
a
g
g
r
e
g
ate
lar
g
e
am
o
u
n
ts
o
f
o
f
f
en
s
iv
e
twee
ts
an
d
h
ate
s
p
ee
ch
in
A
r
ab
ic.
T
h
is
m
eth
o
d
s
ig
n
if
ican
tly
im
p
r
o
v
es
th
e
ef
f
icien
cy
o
f
d
at
a
co
llectio
n
b
y
f
o
c
u
s
in
g
o
n
t
wee
ts
th
at
co
n
tain
s
p
ec
if
ic
em
o
jis
ass
o
ciate
d
with
o
f
f
en
s
iv
e
c
o
n
ten
t.
H
o
wev
er
,
it
is
im
p
o
r
tan
t
to
n
o
te
th
at
em
o
ji
s
ca
n
h
av
e
d
u
al
m
ea
n
in
g
s
,
a
n
d
th
eir
in
ter
p
r
etatio
n
ca
n
lar
g
el
y
d
e
p
en
d
o
n
th
e
co
n
tex
t
in
wh
ich
th
ey
ar
e
u
s
ed
.
Fo
r
e
x
am
p
le,
em
o
jis
lik
e
“r
o
llin
g
o
n
th
e
f
lo
o
r
lau
g
h
in
g
”
o
r
“h
ea
r
t”
,
wh
ich
ar
e
ty
p
ically
ass
o
ciate
d
with
p
o
s
itiv
e
em
o
tio
n
s
,
ca
n
also
b
e
u
s
ed
in
a
s
ar
ca
s
tic
m
an
n
er
o
r
in
a
way
t
h
at
im
p
lies
a
s
p
ec
if
ic
m
ea
n
i
n
g
.
T
h
is
d
u
al
u
s
e
h
ig
h
lig
h
ts
th
e
c
o
m
p
lex
ity
o
f
em
o
jis
u
s
e
in
d
ig
ital c
o
m
m
u
n
icatio
n
.
I
n
lig
h
t
o
f
th
ese
ch
allen
g
es,
th
is
p
ap
er
aim
s
to
ad
v
an
ce
th
e
f
ield
o
f
o
f
f
en
s
iv
e
lan
g
u
a
g
e
d
etec
tio
n
in
Ar
ab
ic
tex
t.
T
h
is
r
esear
ch
ev
alu
ates
th
e
ef
f
ec
tiv
en
ess
o
f
c
u
r
r
en
t
m
eth
o
d
o
lo
g
ies
in
ML
class
if
ier
s
an
d
DL
m
o
d
els f
o
r
i
d
en
tify
in
g
o
f
f
en
s
i
v
e
co
n
ten
t
o
n
th
e
X
(
T
witter
)
p
latf
o
r
m
.
Ad
d
itio
n
ally
,
e
x
p
lo
r
e
th
e
tr
an
s
f
o
r
m
atio
n
o
f
em
o
jis
in
to
tex
t
an
d
its
im
p
ac
t
o
n
d
etec
tio
n
ac
cu
r
ac
y
.
B
y
co
m
p
ar
in
g
tr
a
d
itio
n
al
ML
m
o
d
els
with
ad
v
a
n
ce
d
DL
m
o
d
els,
it
p
r
o
v
id
es
s
tr
ateg
ic
in
s
ig
h
ts
f
o
r
im
p
r
o
v
in
g
co
n
te
n
t
m
o
d
er
ati
o
n
.
T
h
e
co
n
tr
ib
u
tio
n
o
f
t
h
is
p
ap
er
ca
n
b
e
s
u
m
m
ar
ized
as f
o
llo
ws:
a.
A
n
o
v
el
in
v
esti
g
atio
n
o
f
h
o
w
th
e
in
clu
s
io
n
o
f
em
o
jis
im
p
ac
ts
th
e
p
er
f
o
r
m
an
ce
o
f
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P)
m
o
d
els
in
class
if
y
in
g
o
b
jectio
n
ab
le
co
n
te
n
t.
B
y
in
co
r
p
o
r
atin
g
em
o
jis
in
to
tex
t
an
aly
s
is
,
th
e
r
esear
ch
ex
p
lo
r
es
h
o
w
th
ese
s
y
m
b
o
ls
m
o
d
if
y
th
e
in
ter
p
r
et
iv
e
d
y
n
am
ics
o
f
co
n
ten
t
m
o
d
e
r
atio
n
s
y
s
tem
s
,
u
n
d
er
s
co
r
i
n
g
th
e
c
o
m
p
lex
r
o
le
o
f
em
o
jis
in
d
ig
ital c
o
m
m
u
n
i
ca
tio
n
.
b.
Dev
elo
p
a
n
o
v
el
Ar
ab
ic
d
atas
et
co
llected
f
r
o
m
th
e
X
(
T
wit
ter
)
p
latf
o
r
m
c
o
m
p
r
is
ed
o
f
1
0
,
0
0
0
an
n
o
tated
twee
ts
o
f
o
f
f
en
s
iv
e
co
n
ten
t b
y
th
r
ee
Ar
ab
ic
-
n
ativ
e
s
p
ea
k
er
s
.
c.
E
v
alu
ate
th
e
ef
f
ec
tiv
en
ess
o
f
t
r
ad
itio
n
al
ML
class
if
ier
s
an
d
DL
m
o
d
els o
n
th
e
p
r
o
p
o
s
ed
d
ataset
to
m
ea
s
u
r
e
th
eir
p
er
f
o
r
m
an
ce
an
d
a
b
ilit
y
to
d
etec
t o
f
f
en
s
iv
e
tex
t.
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
.
3
,
J
u
n
e
20
25
:
3
3
3
2
-
3
3
4
5
3334
d.
T
h
e
r
esear
ch
estab
lis
h
es
r
ig
o
r
o
u
s
s
tan
d
ar
d
s
th
at
n
o
t
o
n
ly
ad
v
an
ce
th
e
u
n
d
er
s
tan
d
in
g
o
f
co
n
ten
t
r
ec
o
g
n
itio
n
tech
n
iq
u
es
b
u
t
also
p
av
e
th
e
way
f
o
r
th
e
d
ev
elo
p
m
e
n
t
o
f
th
ese
s
y
s
tem
s
in
d
ea
lin
g
wit
h
co
m
p
lex
an
d
cu
ltu
r
ally
v
ar
y
in
g
te
x
ts
.
e.
A
co
m
p
r
eh
e
n
s
iv
e
co
m
p
a
r
is
o
n
b
etwe
en
tr
ad
itio
n
al
ML
m
eth
o
d
s
an
d
DL
tech
n
iq
u
es
in
class
if
y
in
g
o
f
f
en
s
iv
e
co
n
ten
t.
T
h
is
an
al
y
s
is
p
r
o
v
i
d
es
s
tr
ateg
ic
in
s
ig
h
ts
th
at
ar
e
im
p
o
r
tan
t
f
o
r
im
p
r
o
v
i
n
g
co
n
t
en
t
m
o
d
er
atio
n
f
r
am
ewo
r
k
s
an
d
g
u
id
in
g
t
h
e
d
ev
elo
p
m
en
t
o
f
m
o
r
e
ef
f
ec
tiv
e
an
d
ac
cu
r
ate
m
o
d
e
r
atio
n
to
o
ls
.
T
h
e
r
est
o
f
th
is
p
ap
er
is
s
tr
u
ctu
r
ed
as
f
o
llo
ws.
Sectio
n
2
ex
p
lain
th
e
m
eth
o
d
s
an
d
f
r
am
e
wo
r
k
o
f
th
e
s
tu
d
y
.
T
h
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
ar
e
s
h
o
w
n
in
s
ec
tio
n
3
.
Sec
tio
n
4
co
n
clu
d
es th
e
p
ap
er
.
2.
M
E
T
H
O
DS A
ND
M
A
T
E
R
I
AL
S
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
lo
g
y
f
o
r
t
h
is
s
tu
d
y
,
em
p
h
asizin
g
t
h
e
m
o
d
el
d
e
s
ig
n
s
tag
e
as
s
h
o
wn
in
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
h
as
f
iv
e
k
ey
p
h
ases
:
d
ata
co
llectio
n
,
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
,
m
o
d
e
l
co
n
s
tr
u
ctio
n
,
m
o
d
el
p
er
f
o
r
m
a
n
ce
ev
alu
atio
n
,
an
d
co
m
p
ar
is
o
n
o
f
t
h
e
r
esu
lts
.
T
h
e
m
eth
o
d
o
l
o
g
y
is
d
iv
i
d
ed
in
t
o
two
d
if
f
er
en
t
p
ath
s
d
u
r
in
g
th
e
d
ata
p
r
ep
r
o
ce
s
s
in
g
p
h
ase
to
r
ig
o
r
o
u
s
ly
ev
alu
ate
th
e
im
p
ac
t
o
f
em
o
jis
o
n
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
T
h
e
f
ir
s
t
p
ath
in
v
o
lv
es
p
r
o
ce
s
s
in
g
d
ata
u
s
in
g
em
o
jis
,
tr
an
s
lated
in
to
th
eir
tex
tu
al
m
ea
n
in
g
s
,
with
th
e
ai
m
o
f
p
r
e
s
er
v
in
g
an
d
an
aly
zi
n
g
th
e
em
o
tio
n
s
an
d
i
n
f
o
r
m
atio
n
th
e
y
c
o
n
v
ey
.
On
th
e
o
th
er
h
an
d
,
th
e
s
ec
o
n
d
p
ath
,
in
v
o
lv
e
s
p
r
ep
r
o
ce
s
s
in
g
th
e
d
ata
with
all
em
o
jis
r
em
o
v
ed
,
to
s
er
v
e
as
a
ch
ec
k
to
ev
alu
ate
th
e
n
ec
ess
ity
an
d
im
p
ac
t o
f
in
clu
d
in
g
em
o
jis
o
n
t
h
e
m
o
d
el'
s
p
r
ed
ictiv
e
ac
c
u
r
ac
y
.
Af
ter
b
u
ild
in
g
t
h
e
m
o
d
el
an
d
ev
alu
atin
g
it u
n
d
er
th
ese
two
co
n
d
itio
n
s
,
th
e
m
eth
o
d
o
lo
g
y
cu
lm
in
ates in
a
co
m
p
ar
ativ
e
an
aly
s
is
o
f
th
e
r
esu
lt
s
.
T
h
is
co
m
p
ar
is
o
n
cr
itically
ev
alu
ates
wh
eth
er
th
e
m
o
d
el’
s
p
er
f
o
r
m
an
ce
is
im
p
r
o
v
e
d
o
r
wo
r
s
en
ed
b
y
th
e
in
clu
s
io
n
o
f
em
o
jis
,
p
r
o
v
id
i
n
g
a
clea
r
in
d
icatio
n
o
f
th
e
im
p
o
r
tan
ce
o
f
em
o
jis
in
tex
t c
lass
if
icatio
n
task
s
wi
th
in
NL
P.
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
2
.
1
.
Da
t
a
s
et
cr
ea
t
io
n
ph
a
s
e
T
h
e
n
o
v
el
d
ataset
p
r
o
p
o
s
ed
in
th
is
s
tu
d
y
was
co
llected
f
r
o
m
th
e
X
(
T
witter
)
API
.
Fu
r
th
e
r
m
o
r
e,
th
e
d
ataset
was
m
an
u
ally
lab
elled
b
y
th
r
ee
Ar
ab
ic
n
ativ
e
s
p
ea
k
er
s
.
T
h
e
d
ataset
s
ize
is
1
0
,
0
0
0
T
wee
ts
.
T
ab
le
1
p
r
o
v
id
es a
s
am
p
le
lis
t o
f
k
ey
t
er
m
s
u
s
ed
in
th
e
d
ata
co
llectio
n
p
r
o
ce
s
s
,
h
ig
h
lig
h
tin
g
th
e
v
ar
io
u
s
lin
g
u
is
tic
cu
es
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
A
r
a
b
ic
o
ffen
s
ive
text
cla
s
s
ific
a
tio
n
u
s
in
g
em
o
jis
:
I
n
clu
d
i
n
g
emo
ji d
a
ta
in
A
r
a
b
ic
…
(
A
ma
l
A
lb
a
la
w
i
)
3335
an
d
e
x
p
r
ess
io
n
s
th
at
ch
ar
ac
ter
i
ze
o
f
f
e
n
s
iv
e
lan
g
u
ag
e
in
Ar
ab
i
c.
T
ab
le
2
p
r
o
v
id
es
r
e
p
r
esen
tat
iv
e
ex
am
p
les
f
r
o
m
ea
ch
ca
teg
o
r
y
.
T
h
is
tab
le
n
o
t
o
n
ly
s
er
v
es
as
a
g
u
id
e
to
u
n
d
er
s
tan
d
in
g
th
e
ty
p
es
o
f
co
n
ten
t
in
clu
d
ed
with
in
ea
c
h
ca
teg
o
r
y
b
u
t
also
em
p
h
asizes
th
e
c
o
m
p
r
e
h
en
s
iv
e
s
co
p
e
o
f
th
e
d
ataset
in
co
v
er
in
g
a
wid
e
r
an
g
e
o
f
o
f
f
e
n
s
iv
e
ex
p
r
ess
io
n
s
.
T
h
r
o
u
g
h
t
h
is
d
eta
iled
ca
teg
o
r
izatio
n
a
n
d
p
r
o
v
is
io
n
o
f
e
x
am
p
les,
t
h
e
d
ataset
r
e
p
r
esen
ts
a
v
al
u
ab
le
r
eso
u
r
ce
f
o
r
r
esear
ch
e
r
s
aim
in
g
to
ex
p
lo
r
e
th
e
d
y
n
am
ics
o
f
o
f
f
en
s
iv
e
lan
g
u
ag
e
in
Ar
ab
ic
s
o
cial
m
ed
ia,
p
r
o
v
id
i
n
g
in
s
ig
h
ts
in
to
t
h
e
lin
g
u
is
tic
s
tr
ateg
ies u
s
ed
in
d
if
f
er
en
t f
o
r
m
s
o
f
o
f
f
en
s
iv
e
co
m
m
u
n
icatio
n
.
T
ab
le
1
.
E
x
am
p
le
f
o
r
k
e
y
wo
r
d
s
u
s
ed
to
co
llect
twee
ts
f
o
r
th
e
d
ataset
f
o
r
ea
ch
t
y
p
e
o
f
twee
t
C
a
t
e
g
o
r
y
K
e
y
w
o
r
d
i
n
A
r
a
b
i
c
Tr
a
n
s
l
a
t
i
o
n
o
f
t
h
e
K
e
y
w
o
r
d
i
n
En
g
l
i
s
h
H
a
t
e
sp
e
e
c
h
علب
ا
،
ر
ق
ح
ت
س
ا
ر
ق
ت
ح
ا
هر
ك
أ
،
هر
ش
ح
،
هن
و
خ
h
a
t
e
,
d
e
sp
i
se
,
d
e
t
e
st
,
sw
a
l
l
o
w
,
b
u
g
,
t
r
a
i
t
o
r
H
a
r
a
ssm
e
n
t
ر
ي
ق
ح
،
ر
ا
م
ح
،
قم
ح
أ
،
ي
ب
غ
S
t
u
p
i
d
,
i
d
i
o
t
,
d
o
n
k
e
y
,
d
e
s
p
i
c
a
b
l
e
Ex
p
l
i
c
i
t
c
o
n
t
e
n
t
ك
غ
ا
م
د
لوس
غ
م
،
خ
ب
طم
لا
ك
ن
ا
ك
م
Y
o
u
r
p
l
a
c
e
i
s i
n
t
h
e
k
i
t
c
h
e
n
,
y
o
u
a
r
e
b
r
a
i
n
w
a
s
h
e
d
"
R
e
l
i
g
i
o
u
s
i
n
s
u
l
t
ي
ب
ا
هو ي
ن
س
ي
ع
ي
ش
S
h
i
a
,
S
u
n
n
i
,
W
a
h
h
a
b
i
O
f
f
e
n
si
v
e
l
a
n
g
u
a
g
e
r
e
l
a
t
i
n
g
t
o
g
e
n
d
e
r
.
ةد
ئ
ا
ف
م
ه
لا
م
لا
ج
ر
لا
،
خ
ب
طم
لا
ك
ن
ا
ك
م
Y
o
u
r
p
l
a
c
e
i
s i
n
t
h
e
k
i
t
c
h
e
n
,
me
n
a
r
e
u
se
l
e
ss
O
f
f
e
n
si
v
e
l
a
n
g
u
a
g
e
r
e
l
a
t
i
n
g
t
o
e
t
h
n
i
c
i
t
y
o
r
r
a
c
e
.
ي
ر
ص
م
،
ي
ن
م
ي
،
ي
د
وع
س
،
ي
ن
ي
ب
لف
،
د
ب
ع
S
l
a
v
e
,
F
i
l
i
p
i
n
o
,
S
a
u
d
i
,
Y
a
m
a
n
i
,
Eg
y
p
t
i
a
n
B
o
d
y
s
h
a
m
i
n
g
،
ر
ي
ص
ق
،
ث
لث
م
ك
س
ار
،
ب
د
ت
ن
ا
لي
وط
F
a
t
,
t
r
i
a
n
g
l
e
h
e
a
d
,
u
g
l
y
,
s
h
o
r
t
,
t
a
l
l
C
l
a
s
si
sm
ر
ي
ق
ف
،
ي
م
د
خ
S
e
r
v
a
n
t
,
p
o
o
r
A
g
e
i
s
m
رزب
ب
ي
ا
ش
K
i
d
,
o
l
d
ma
n
A
b
l
e
i
sm
ئ
ب
ع
ض
ي
ر
م
،
ئ
ب
ع
ن
ي
قا
ع
م
لا
,
d
i
s
a
b
l
e
d
a
r
e
a
b
u
r
d
e
n
;
si
c
k
a
r
e
a
b
u
r
d
e
n
T
ab
le
2
.
E
x
am
p
le
f
o
r
twee
ts
f
r
o
m
ea
ch
ca
teg
o
r
y
t
h
at
ar
e
co
n
s
id
er
ed
o
f
f
e
n
s
iv
e
in
th
e
d
ataset
C
a
t
e
g
o
r
y
Tw
e
e
t
(
i
n
A
r
a
b
i
c
)
En
g
l
i
sh
t
r
a
n
sl
a
t
i
o
n
Ex
p
l
a
n
a
t
i
o
n
H
a
t
e
sp
e
e
c
h
"
لب
هأ
وا
ي
ب
غ
ت
ن
ا
ي
ن
ع
ي
"
M
e
a
n
s
y
o
u
a
r
e
st
u
p
i
d
o
r
d
u
m
b
.
D
i
r
e
c
t
i
n
s
u
l
t
H
a
r
a
ssm
e
n
t
"
.
.
.
لب
هأ
ا
ي
هي
قا
طا
ي
"
Y
o
u
a
r
e
a
d
u
mb
h
a
t
I
n
d
i
r
e
c
t
i
n
s
u
l
t
t
h
a
t
b
u
l
l
i
e
s a
s
p
e
c
i
f
i
c
f
o
o
t
b
a
l
l
t
e
a
m
R
a
c
e
"
.
.
.
قم
ح
أو
ي
ب
غ
د
وس
ا
وها
م
لك
"
Ev
e
r
y
t
h
i
n
g
b
l
a
c
k
i
s st
u
p
i
d
a
n
d
f
o
o
l
i
s
h
D
i
r
e
c
t
i
n
s
u
l
t
Et
h
n
i
c
i
t
y
"
ن
ب
لف
ن
م
م
ك
ي
ف
كرا
ب
.
.
.
"
M
a
y
G
o
d
d
o
e
s
n
o
t
b
l
e
s
s
y
o
u
,
F
i
l
i
p
i
n
o
s
I
n
d
i
r
e
c
t
i
n
s
u
l
t
f
o
r
a
sp
e
c
i
f
i
c
n
a
t
i
o
n
a
l
i
t
y
A
b
l
e
i
sm
ه
ن
ك
ا
س
م
ا
ي
ك
ر
ح
و
ا
ي
لق
ع
ن
ي
قا
ع
م
لا
"
.
.
ئ
ب
ع
ةش
ي
ع
م
شم
.
.
.
؟ء
ا
ي
ح
أ
م
هول
خ
ي
“
M
e
n
t
a
l
l
y
a
n
d
p
h
y
si
c
a
l
l
y
d
i
s
a
b
l
e
d
p
e
o
p
l
e
a
r
e
f
o
r
l
o
r
n
w
h
y
d
o
t
h
e
y
l
e
t
t
h
e
m
l
i
v
e
?
Th
e
y
a
r
e
a
b
u
r
d
e
n
.
D
i
r
e
c
t
i
n
s
u
l
t
S
e
x
t
u
a
l
"
.
.
.
ةط
*
*
*
ر
ش
لا
ةر
م
لا
ةب
ش
لم
ا
ع
"
Y
o
u
a
r
e
l
i
k
e
a
b
i
*
*
h
w
o
ma
n
D
i
r
e
c
t
i
n
s
u
l
t
R
e
l
i
g
i
o
u
s
h
a
t
e
ن
وك
ت
س
د
ي
ك
أ
ت
لك
ب
ي
ع
ي
ش
ن
وك
ت
ن
م
"
"
.
.
.
ر
ا
م
ح
I
f
y
o
u
a
r
e
S
h
i
a
,
y
o
u
a
r
e
d
e
f
i
n
i
t
e
l
y
a
d
o
n
k
e
y
D
i
r
e
c
t
i
n
s
u
l
t
G
e
n
d
e
r
"
خ
ب
طم
لا
ى
لع"
G
o
t
o
t
h
e
k
i
t
c
h
e
n
.
I
n
d
i
r
e
c
t
i
n
s
u
l
t
t
h
a
t
mea
n
s
w
o
m
e
n
c
a
n
’
t
d
o
a
n
y
t
h
i
n
g
o
t
h
e
r
t
h
a
n
c
o
o
k
i
n
g
B
o
d
y
S
h
a
m
i
n
g
"
.
.
.
ي
ب
غ
م
لا
ع
لا
ي
ف
ب
د
لك
ي
خ
ا"
Ev
e
r
y
f
a
t
p
e
r
s
o
n
i
s
st
u
p
i
d
…
D
i
r
e
c
t
i
n
s
u
l
t
C
l
a
s
si
sm
"
.
.
قز
ت
ر
م
و
لب
هأ
ر
ي
ق
فو
ي
ف
ح
ص
.
.
"
….
p
o
o
r
j
o
u
r
n
a
l
i
st
,
f
o
o
l
i
s
h
a
n
d
m
e
r
c
e
n
a
r
y
…
D
i
r
e
c
t
i
n
s
u
l
t
A
g
e
i
s
m
"
.
.
د
س
ا
ف
ب
ي
ا
ش
ب
ر
د
م
.
.
"
O
v
e
r
a
g
e
,
c
o
r
r
u
p
t
c
o
a
c
h
D
i
r
e
c
t
i
n
s
u
l
t
T
h
e
p
r
o
p
o
s
ed
d
ataset
with
1
0
,
0
0
0
twee
ts
was
m
an
u
ally
an
n
o
tated
b
y
th
r
e
e
h
u
m
an
n
ativ
e
Ar
ab
ic
s
p
ea
k
er
s
.
I
n
th
is
s
tu
d
y
,
ea
ch
t
wee
t w
as e
v
alu
ated
b
y
th
r
ee
a
n
n
o
tato
r
s
to
en
s
u
r
e
t
h
e
r
o
b
u
s
tn
ess
o
f
o
u
r
o
f
f
en
s
iv
e
lan
g
u
ag
e
d
etec
tio
n
d
ataset.
T
o
m
ea
s
u
r
e
in
ter
-
an
n
o
tat
o
r
a
g
r
ee
m
en
t,
we
ca
lcu
lated
th
e
C
K
[
1
5
]
f
o
r
ea
ch
p
air
,
r
esu
ltin
g
in
v
alu
es o
f
0
.
8
5
4
,
0
.
8
8
5
,
an
d
0
.
8
8
2
,
wh
ich
in
d
icat
e
a
h
ig
h
lev
el
o
f
co
n
s
en
s
u
s
am
o
n
g
th
e
an
n
o
tato
r
s
.
Ad
d
itio
n
ally
,
we
co
m
p
u
te
d
th
e
Kr
ip
p
en
d
o
r
f
f
'
s
Alp
h
a
[
1
6
]
,
o
b
tain
in
g
a
v
alu
e
o
f
0
.
8
7
4
,
f
u
r
th
er
co
n
f
ir
m
in
g
th
e
r
eliab
ilit
y
o
f
o
u
r
an
n
o
tatio
n
s
ac
r
o
s
s
all
th
r
ee
ev
alu
ato
r
s
.
T
h
is
s
tr
o
n
g
ag
r
ee
m
en
t
u
n
d
er
s
co
r
es
th
e
h
ig
h
q
u
ality
o
f
o
u
r
d
ataset,
wh
ich
is
ess
en
tial
f
o
r
d
ev
elo
p
in
g
ef
f
ec
tiv
e
ML
class
if
ier
s
f
o
r
NL
P.
B
y
d
em
o
n
s
tr
atin
g
co
n
s
is
ten
t
an
n
o
tatio
n
s
,
o
u
r
r
esear
ch
lay
s
a
s
o
lid
f
o
u
n
d
atio
n
f
o
r
al
g
o
r
it
h
m
s
th
at
ac
c
u
r
ately
i
n
ter
p
r
et
co
m
p
lex
lan
g
u
a
g
e
n
u
an
ce
s
o
n
s
o
cial
m
e
d
ia.
T
h
e
an
n
o
tatio
n
p
r
o
ce
s
s
r
ep
r
esen
ts
a
cr
u
cial
s
tep
i
n
cr
ea
tin
g
a
g
o
l
d
-
s
tan
d
ar
d
co
r
p
u
s
.
T
o
en
s
u
r
e
th
e
i
n
te
g
r
ity
a
n
d
co
n
s
is
ten
cy
o
f
th
is
p
r
o
ce
s
s
,
all
a
n
n
o
tato
r
s
ad
h
er
ed
t
o
th
e
f
o
llo
win
g
g
u
id
elin
es
an
d
in
s
tr
u
ctio
n
s
:
a.
An
n
o
tato
r
s
wer
e
s
tr
o
n
g
ly
e
n
co
u
r
ag
ed
to
a
p
p
r
o
ac
h
th
e
an
n
o
tatio
n
p
r
o
ce
s
s
o
b
jectiv
ely
,
m
a
k
in
g
s
u
r
e
th
at
th
eir
b
ac
k
g
r
o
u
n
d
s
o
r
p
e
r
s
o
n
al
p
r
ej
u
d
ices,
s
u
ch
as
th
eir
cu
ltu
r
al
o
r
r
elig
io
u
s
b
elief
s
,
d
id
n
o
t
in
te
r
f
er
e
with
th
eir
o
b
jectiv
ity
.
T
h
e
ac
cu
r
ac
y
an
d
d
ep
en
d
a
b
i
lity
o
f
th
e
d
ataset
d
e
p
en
d
s
h
ea
v
ily
o
n
t
h
is
ad
h
er
en
ce
to
n
eu
tr
ality
.
b.
T
h
e
s
tan
d
ar
d
s
f
o
r
class
if
y
in
g
twee
ts
a
s
o
f
f
en
s
iv
e
wer
e
well
-
d
ef
in
ed
an
d
co
v
e
r
ed
a
b
r
o
a
d
s
p
ec
tr
u
m
o
f
to
p
ics,
in
clu
d
in
g
h
ate
s
p
ee
ch
,
cy
b
er
b
u
lly
in
g
,
r
ac
ial/eth
n
ic
d
is
cr
im
in
atio
n
,
d
is
ab
ilit
ies,
s
ex
u
al
o
r
ien
tatio
n
,
an
d
r
elig
io
n
.
An
n
o
tato
r
s
h
a
d
to
u
s
e
th
eir
th
o
r
o
u
g
h
k
n
o
wled
g
e
o
f
i
n
ap
p
r
o
p
r
iate
co
n
ten
t
to
id
e
n
tif
y
twee
ts
th
at
f
it
in
to
an
y
o
f
th
ese
ca
teg
o
r
ies.
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
.
3
,
J
u
n
e
20
25
:
3
3
3
2
-
3
3
4
5
3336
c.
An
n
o
tato
r
s
wer
e
tr
ain
ed
n
o
t
ju
s
t
to
id
en
tify
o
v
er
tly
o
f
f
e
n
s
iv
e
twee
ts
b
u
t
also
to
r
ec
o
g
n
ize
in
d
ir
ec
tly
o
f
f
en
s
iv
e
twee
ts
b
y
an
al
y
zin
g
t
h
e
o
v
e
r
t
an
d
co
v
er
t
m
e
an
in
g
s
c
o
n
tain
ed
with
in
th
e
twee
ts
.
T
h
i
s
m
eth
o
d
o
l
o
g
y
g
u
ar
an
tees
th
at
th
e
d
ataset
en
co
m
p
ass
es
th
e
en
tire
r
an
g
e
o
f
o
f
f
en
s
iv
e
lan
g
u
a
g
e,
in
clu
d
i
n
g
s
u
b
tleties
an
d
im
p
licatio
n
s
th
at
m
ig
h
t
n
o
t
b
e
q
u
ick
ly
d
is
ce
r
n
e
d
.
T
o
aid
th
e
a
n
n
o
tato
r
s
in
th
is
d
if
f
icu
lt
en
d
e
av
o
u
r
,
e
x
am
p
les
o
f
in
d
ir
ec
tly
o
f
f
e
n
s
iv
e
twee
ts
wer
e
g
iv
en
.
Fig
u
r
es
2
an
d
3
p
r
o
v
id
e
e
x
a
m
p
les
o
f
th
e
d
ataset,
s
h
o
win
g
twee
ts
with
q
u
o
tes
f
r
o
m
th
e
Qu
r
an
an
d
Ar
ab
ic
B
ib
le,
r
esp
ec
tiv
ely
.
W
e
co
n
s
id
er
ed
th
ese
twee
ts
a
s
n
o
n
-
o
f
f
en
s
iv
e
in
o
u
r
d
ataset,
wh
ich
ar
e
r
esp
ec
ted
in
th
eir
r
esp
ec
tiv
e
r
elig
io
u
s
co
n
tex
ts
.
T
h
is
class
if
i
ca
tio
n
d
em
o
n
s
tr
ates
o
u
r
d
e
d
icatio
n
t
o
co
n
tex
tu
al
s
en
s
itiv
ity
as
well
as
o
u
r
u
n
d
er
s
tan
d
in
g
o
f
th
e
wr
itin
g
s
’
cu
ltu
r
al
an
d
r
elig
i
o
u
s
r
elev
an
ce
.
B
y
p
r
esen
tin
g
t
h
ese
in
s
tan
ce
s
,
we
h
o
p
e
t
o
d
r
aw
atten
tio
n
to
th
e
ca
r
ef
u
l
p
r
o
ce
s
s
u
s
ed
to
class
if
y
co
n
ten
t,
m
a
k
in
g
s
u
r
e
th
at
in
f
o
r
m
atio
n
th
at
is
f
u
n
d
am
e
n
tal
to
c
u
ltu
r
al
a
n
d
r
e
lig
io
u
s
d
eb
ate
a
n
d
is
b
y
its
v
e
r
y
n
atu
r
e
r
esp
ec
tf
u
l,
ap
p
r
o
p
r
iately
lab
eled
as n
o
n
-
o
f
f
en
s
iv
e.
Mo
r
e
o
v
er
,
th
e
d
atas
et
co
n
s
is
ts
o
f
d
if
f
er
e
n
t
twee
ts
t
h
at
ar
e
ex
p
r
ess
ed
in
d
if
f
e
r
en
t A
r
ab
ic
d
ialec
ts
b
u
t
ha
v
e
s
im
ilar
m
ea
n
in
g
s
.
T
h
is
h
ig
h
lig
h
ts
th
e
in
tr
icac
y
an
d
n
u
an
ce
in
v
o
lv
ed
in
ac
cu
r
at
ely
class
if
y
in
g
an
d
in
ter
p
r
etin
g
twee
ts
an
d
s
er
v
es
as
an
ex
am
p
le
o
f
th
e
lin
g
u
is
tic
d
iv
er
s
ity
an
d
r
ich
tap
estry
o
f
d
ialec
ts
with
in
th
e
Ar
ab
ic
lan
g
u
a
g
e.
T
h
e
d
ataset'
s
th
o
r
o
u
g
h
c
o
v
er
ag
e
an
d
t
h
e
ca
r
ef
u
l
co
n
s
id
er
atio
n
g
iv
e
n
to
th
e
n
u
an
ce
s
o
f
r
eg
io
n
al
v
ar
iatio
n
s
an
d
co
llo
q
u
ial
id
io
m
s
co
m
m
o
n
in
s
o
cial
m
e
d
ia
d
is
co
u
r
s
e
ar
e
h
ig
h
lig
h
te
d
b
y
t
h
e
in
clu
s
io
n
o
f
s
u
ch
s
am
p
les.
T
h
is
is
a
cr
u
cial
f
ac
to
r
to
tak
e
in
t
o
ac
c
o
u
n
t
wh
en
cr
e
atin
g
s
o
p
h
is
tic
ated
,
a
n
d
cu
ltu
r
ally
awa
r
e
a
n
aly
tic
s
y
s
tem
s
th
at
ca
n
co
r
r
ec
tly
u
n
d
er
s
tan
d
an
d
ca
te
g
o
r
ize
co
n
ten
t
ac
r
o
s
s
th
e
v
ast Ar
ab
ic
-
s
p
ea
k
i
n
g
wo
r
ld
.
Fig
u
r
e
2
.
No
n
-
o
f
f
en
s
iv
e
twee
t
s
th
at
h
av
e
a
q
u
o
te
f
r
o
m
Qu
r
a
n
Fig
u
r
e
3
.
No
n
-
o
f
f
en
s
iv
e
twee
t
s
th
at
h
av
e
a
q
u
o
te
f
r
o
m
Ar
a
b
ic
B
ib
le
A
Fig
u
r
e
s
4
an
d
5
p
r
esen
t
th
e
m
o
s
t
f
r
eq
u
e
n
tly
u
s
ed
wo
r
d
s
in
th
e
o
f
f
en
s
iv
e
a
n
d
n
o
n
-
o
f
f
en
s
iv
e
ca
teg
o
r
ies o
f
th
e
d
ataset,
r
esp
ec
tiv
ely
.
Fig
u
r
e
4
h
i
g
h
lig
h
ts
c
o
m
m
o
n
wo
r
d
s
f
o
u
n
d
in
o
f
f
e
n
s
iv
e
twee
ts
,
s
h
ed
d
i
n
g
lig
h
t
o
n
s
p
ec
if
ic
ter
m
s
o
r
p
att
er
n
s
th
at
f
r
e
q
u
en
tly
ap
p
ea
r
in
o
b
jectio
n
a
b
le
c
o
n
ten
t.
Un
d
er
s
tan
d
in
g
t
h
ese
wo
r
d
d
is
tr
ib
u
tio
n
s
ca
n
en
h
a
n
ce
cl
ass
if
icatio
n
alg
o
r
ith
m
s
b
y
id
en
tify
in
g
k
ey
in
d
icato
r
s
o
f
o
f
f
en
s
iv
e
lan
g
u
ag
e.
Similar
ly
,
Fig
u
r
e
5
d
is
p
lay
s
t
h
e
m
o
s
t
u
s
ed
wo
r
d
s
in
th
e
n
o
n
-
o
f
f
en
s
iv
e
ca
teg
o
r
y
,
p
r
o
v
id
in
g
in
s
ig
h
t
i
n
to
th
e
v
o
ca
b
u
lar
y
ty
p
ically
f
o
u
n
d
in
n
eu
tr
al
o
r
a
p
p
r
o
p
r
iate
twee
t
s
.
B
y
co
m
p
ar
i
n
g
th
e
wo
r
d
d
i
s
tr
ib
u
tio
n
s
in
b
o
th
ca
teg
o
r
ies,
we
ca
n
b
etter
d
is
tin
g
u
is
h
o
f
f
e
n
s
iv
e
f
r
o
m
n
o
n
-
o
f
f
en
s
iv
e
co
n
te
n
t,
im
p
r
o
v
i
n
g
f
ilter
in
g
a
n
d
d
etec
tio
n
tech
n
iq
u
es.
Fig
u
r
e
4
.
Of
f
en
s
iv
e
d
ata
Fig
u
r
e
5
.
N
on
-
o
f
f
en
s
iv
e
d
ata
I
n
th
e
an
aly
s
is
s
h
o
wn
in
Fig
u
r
e
6
,
a
v
ar
iety
o
f
em
o
jis
ap
p
ea
r
with
in
t
h
e
d
ataset,
h
i
g
h
l
ig
h
tin
g
th
e
d
iv
er
s
ity
o
f
th
e
twee
ts
ex
am
in
ed
.
I
n
p
ar
ticu
lar
,
t
h
e
m
o
s
t
p
o
p
u
lar
em
o
jis
in
th
is
d
ataset
in
clu
d
e
s
ev
er
al
lau
g
h
in
g
em
o
jis
.
T
h
is
o
cc
u
r
r
en
ce
o
f
lau
g
h
in
g
em
o
jis
ca
n
lik
ely
b
e
attr
ib
u
ted
to
th
e
f
r
e
q
u
en
t a
p
p
ea
r
a
n
ce
o
f
th
ese
em
o
jis
m
u
ltip
le
tim
es
in
in
d
iv
id
u
al
twee
ts
.
Par
ticu
lar
em
o
jis
wh
ic
h
ar
e
lin
k
ed
to
p
o
s
itiv
e
f
ee
lin
g
s
,
s
u
ch
as
h
ea
r
ts
,
lau
g
h
in
g
,
an
d
s
m
ilin
g
em
o
jis
,
ar
e
am
o
n
g
th
o
s
e
th
at
a
p
p
ea
r
m
o
s
t
f
r
eq
u
en
tly
in
th
e
o
f
f
e
n
s
iv
e
tex
t.
T
h
is
r
ev
ea
ls
a
ce
r
tain
cu
ltu
r
al
d
if
f
er
e
n
ce
in
Ar
ab
ic
twee
ts
in
wh
ich
lau
g
h
in
g
em
o
jis
ar
e
o
f
ten
u
s
ed
in
s
ar
ca
s
tic
co
n
tex
ts
,
ev
en
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
A
r
a
b
ic
o
ffen
s
ive
text
cla
s
s
ific
a
tio
n
u
s
in
g
em
o
jis
:
I
n
clu
d
i
n
g
emo
ji d
a
ta
in
A
r
a
b
ic
…
(
A
ma
l
A
lb
a
la
w
i
)
3337
wh
en
th
e
u
n
d
er
ly
i
n
g
m
ess
ag
e
is
o
f
f
en
s
iv
e.
T
h
is
p
atter
n
in
d
icate
s
a
cu
ltu
r
ally
d
is
tin
ct
ap
p
licatio
n
o
f
em
o
jis
,
u
n
d
er
s
co
r
i
n
g
th
e
c
o
m
p
le
x
ity
o
f
in
ter
p
r
etin
g
em
o
jis
in
d
if
f
er
en
t
cu
ltu
r
al
b
ac
k
g
r
o
u
n
d
s
.
Alth
o
u
g
h
t
h
e
in
itial
co
llectio
n
o
f
th
e
p
r
o
p
o
s
ed
d
at
aset
d
id
n
o
t
f
o
cu
s
s
p
ec
if
ically
o
n
em
o
jis
,
th
eir
in
f
lu
e
n
ce
was
co
n
s
id
er
ed
in
th
e
class
if
icatio
n
p
r
o
ce
s
s
,
in
d
icatin
g
th
e
im
p
o
r
tan
ce
o
f
em
o
jis
in
co
n
v
e
y
in
g
s
u
b
tle
co
m
m
u
n
ic
atio
n
s
ig
n
als
with
in
in
ter
ac
tio
n
s
.
Fig
u
r
e
6
.
Mo
s
t u
s
ed
e
m
o
jis
in
th
e
d
ataset
2
.
2
.
Da
t
a
pre
-
pro
ce
s
s
ing
ph
a
s
e
C
lean
in
g
th
e
d
ataset
is
a
cr
u
ci
al
s
tep
wh
en
d
ea
lin
g
with
NL
P
task
s
.
T
h
is
s
tu
d
y
u
s
ed
two
m
eth
o
d
s
to
clea
n
th
e
d
ataset.
T
h
e
m
ain
d
if
f
er
en
ce
in
th
ese
m
eth
o
d
s
is
th
at
th
e
f
ir
s
t
m
eth
o
d
tr
an
s
lated
th
e
em
o
jis
to
in
clu
d
e
th
e
m
ea
n
in
g
o
f
th
e
em
o
jis
in
t
h
e
twee
t
as
f
u
r
th
er
d
is
cu
s
s
ed
in
th
e
f
o
llo
win
g
s
ec
tio
n
.
T
h
e
s
ec
o
n
d
m
eth
o
d
u
s
ed
th
e
s
am
e
clea
n
in
g
p
r
o
ce
s
s
b
u
t
r
em
o
v
ed
all
th
e
em
o
jis
.
2
.
2
.
1
.
Da
t
a
clea
nin
g
T
h
is
s
ec
tio
n
ex
p
lain
s
th
e
d
ata
clea
n
in
g
p
r
o
ce
s
s
im
p
lem
en
ted
in
b
o
th
p
ath
s
;
with
em
o
ji a
cr
o
n
y
m
s
an
d
with
o
u
t
em
o
ji
ac
r
o
n
y
m
s
.
I
n
th
e
f
ir
s
t
p
ath
,
th
e
clea
n
in
g
p
r
o
ce
s
s
s
p
li
t
th
e
h
ash
tag
s
;
r
em
o
v
ed
UR
L
s
;
r
em
o
v
ed
u
s
er
n
am
es;
r
em
o
v
e
d
th
e
n
u
m
b
er
s
,
s
p
ec
ial
ch
ar
ac
ter
s
an
d
d
i
ac
r
itics
(
T
ash
k
ee
l)
;
tr
an
s
lated
em
o
jis
’
m
ea
n
in
g
s
;
r
em
o
v
ed
E
n
g
lis
h
te
x
t; a
n
d
r
e
m
o
v
ed
s
to
p
wo
r
d
s
.
T
h
is
s
tu
d
y
lev
er
ag
ed
a
s
et
o
f
em
o
jis
id
en
tifie
d
b
y
Mu
b
ar
ak
et
a
l.
[
1
4
]
as
co
m
m
o
n
in
o
f
f
en
s
iv
e
co
m
m
u
n
icatio
n
.
T
h
is
s
elec
tio
n
h
as
b
ee
n
ca
r
ef
u
lly
cu
r
ated
t
o
ca
p
tu
r
e
t
h
e
ty
p
es
o
f
em
o
jis
th
at
ar
e
m
o
s
t
o
f
te
n
u
s
ed
to
ex
p
r
ess
n
eg
ativ
e
em
o
tio
n
s
s
u
ch
as
d
is
r
esp
ec
t,
a
n
g
e
r
,
o
r
d
is
g
u
s
t.
T
h
ese
o
f
ten
in
cl
u
d
e
v
ar
io
u
s
an
im
al
s
y
m
b
o
ls
an
d
o
t
h
e
r
ty
p
es o
f
s
y
m
b
o
lis
m
u
s
ed
m
etap
h
o
r
ically
o
r
d
ir
ec
tly
t
o
in
s
u
lt in
d
iv
i
d
u
al
s
.
On
th
is
b
asis
,
th
is
r
esear
ch
n
o
t
o
n
ly
in
clu
d
ed
th
ese
well
-
k
n
o
wn
o
f
f
en
s
iv
e
em
o
jis
b
u
t
also
ex
p
an
d
ed
th
e
s
et
to
in
clu
d
e
em
o
jis
g
en
er
ally
co
n
s
id
er
ed
n
o
n
-
o
f
f
e
n
s
iv
e.
T
h
is
b
r
o
ad
er
s
p
ec
tr
u
m
o
f
em
o
jis
was
in
clu
d
ed
to
im
p
r
o
v
e
th
e
ac
cu
r
ac
y
o
f
o
u
r
an
al
y
s
is
an
d
in
cr
ea
s
e
th
e
r
o
b
u
s
tn
ess
o
f
o
u
r
d
ataset.
T
h
e
p
r
ep
r
o
ce
s
s
in
g
p
h
ase
em
b
ar
k
ed
o
n
tr
a
n
s
latin
g
th
e
s
em
an
tic
co
n
ten
t
o
f
ea
ch
em
o
ji
in
to
tex
tu
al
r
ep
r
esen
tatio
n
s
.
T
h
is
p
r
o
ce
s
s
in
v
o
lv
ed
in
ter
p
r
etin
g
th
e
m
ea
n
in
g
s
o
f
em
o
jis
with
in
twee
ts
to
en
s
u
r
e
th
at
th
eir
co
n
tex
tu
al
s
ig
n
if
ican
ce
was
p
r
eser
v
ed
an
d
ac
cu
r
ate
ly
r
ef
lect
ed
in
tex
t
-
b
ased
class
if
icatio
n
task
s
.
I
n
d
o
in
g
s
o
,
we
aim
ed
t
o
ca
p
tu
r
e
th
e
s
u
b
tle
way
s
in
wh
ich
em
o
jis
co
n
tr
ib
u
te
to
th
e
to
n
e
an
d
i
n
ten
t
o
f
o
n
li
n
e
co
m
m
u
n
icatio
n
,
th
u
s
en
r
ich
in
g
m
o
d
els’
ab
ilit
y
to
id
en
tify
an
d
class
if
y
twee
t
s
m
o
r
e
ef
f
ec
tiv
ely
.
T
h
is
m
eth
o
d
is
d
etailed
b
elo
w
an
d
is
illu
s
tr
ated
th
r
o
u
g
h
T
ab
l
e
3
,
wh
ich
s
h
o
ws
an
ex
am
p
le
o
f
h
o
w
em
o
jis
ca
n
b
e
tr
an
s
f
o
r
m
ed
in
to
th
eir
tex
tu
al
co
u
n
ter
p
a
r
ts
to
en
r
ich
co
n
tex
t
u
al
an
aly
s
is
in
twee
t c
lass
if
ica
tio
n
.
I
n
th
e
co
n
s
tr
u
ctio
n
o
f
th
e
p
r
ep
r
o
ce
s
s
in
g
p
ip
elin
e
f
o
r
t
h
is
s
tu
d
y
,
a
s
ec
o
n
d
clea
n
in
g
p
r
o
ce
s
s
is
m
eticu
lo
u
s
ly
d
esig
n
ed
f
o
c
u
s
in
g
o
n
th
e
co
r
e
lin
g
u
is
tic
co
m
p
o
n
e
n
ts
o
f
th
e
an
aly
s
is
.
T
h
is
p
r
o
ce
s
s
in
clu
d
es
s
p
litt
in
g
h
ash
tag
s
an
d
r
em
o
v
in
g
UR
L
s
,
u
s
er
n
am
es,
n
u
m
b
er
s
,
s
p
ec
ial
ch
ar
ac
ter
s
,
d
iacr
itics
,
em
o
jis
,
E
n
g
lis
h
tex
t,
an
d
s
to
p
wo
r
d
s
,
th
u
s
f
o
cu
s
in
g
ex
clu
s
iv
ely
o
n
Ar
ab
ic
tex
t
elem
en
ts
.
T
h
is
s
tr
ateg
y
d
if
f
er
s
f
r
o
m
th
e
i
n
itial
ap
p
r
o
ac
h
b
y
ex
clu
d
i
n
g
em
o
jis
to
all
o
w
f
o
r
a
clea
r
co
m
p
ar
is
o
n
with
th
e
f
ir
s
t
clea
n
u
p
,
wh
i
ch
in
cl
u
d
ed
em
o
jis
.
T
h
is
en
s
u
r
es
th
at
th
e
d
ataset
r
em
ain
s
f
o
cu
s
ed
o
n
ly
o
n
Ar
ab
i
c
tex
t,
allo
win
g
a
d
ir
ec
t
ass
ess
m
en
t
o
f
th
e
im
p
ac
t
o
f
d
if
f
e
r
en
t p
r
ep
r
o
ce
s
s
in
g
s
tr
ateg
ies o
n
th
e
r
esu
lts
o
f
o
u
r
a
n
a
ly
s
is
.
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
.
3
,
J
u
n
e
20
25
:
3
3
3
2
-
3
3
4
5
3338
T
ab
le
3
.
E
x
am
p
le
f
o
r
clea
n
in
g
twee
ts
p
r
o
ce
s
s
with
em
o
jis
C
l
e
a
n
i
n
g
st
e
p
s
Te
x
t
a
f
t
e
r
A
p
p
l
i
n
g
t
h
e
s
t
e
p
O
r
i
g
i
n
a
l
t
e
x
t
@
A
d
a
m
ا
ً
ن
ا
#
ُ
ب
ِ
ح
ُ
أ
_
ب
ِ
ع
ّ
ل
ل
ا
لا
و
g
a
mi
n
g
ةب
ع
لل
ا
ي
ذ
ه ي
ف
<
3
w
w
w
.
g
a
m
e
x
.
c
o
m
S
p
l
i
t
h
a
s
h
t
a
g
s
@
A
d
a
m
ُ
ب
ِ
ح
ُ
أ
ا
ً
ن
ا
ب
ِ
ع
ّ
ل
ل
ا
لاو
g
a
mi
n
g
ةب
ع
لل
ا
ي
ذ
ه
ي
ف
<
3
w
w
w
.
g
a
m
e
x
.
c
o
m
R
e
m
o
v
e
d
U
R
L
s
@
A
d
a
m
ا
ً
ن
ا
ب
ِ
ع
ّ
ل
ل
ا
ُ
ب
ِ
ح
ُ
أ
لاو
g
a
mi
n
g
ةب
ع
لل
ا
ي
ذ
ه
ي
ف
<
3
R
e
m
o
v
e
d
u
sern
a
mes
.
ب
ِ
ع
ّ
ل
ل
ا
ُ
ب
ِ
ح
ُ
أ
ا
ً
ن
ا
لا
و
g
a
mi
n
g
ة
ب
ع
لل
ا
ي
ذ
ه ي
ف
<
3
R
e
m
o
v
e
d
n
u
m
b
e
r
s
ب
ِ
ع
ّ
ل
ل
ا
ُ
ب
ِ
ح
ُ
أ
ا
ً
ن
ا
لا
و
g
a
mi
n
g
ة
ب
ع
لل
ا
ي
ذ
ه ي
ف
<
R
e
m
o
v
e
d
s
p
e
c
i
a
l
c
h
a
r
a
c
t
e
r
s
ب
ِ
ع
ّ
ل
ل
ا
ُ
ب
ِ
ح
ُ
أ
ا
ً
ن
ا
لا
و
g
a
mi
n
g
ي
ف
ة
ب
ع
لل
ا
ي
ذ
ه
R
e
m
o
v
e
d
d
i
a
c
r
i
t
i
c
s
(
Ta
s
h
k
e
e
l
)
ب
ع
ل
لا
ب
ح
أ
ا
ن
ا
لاو
g
a
mi
n
g
ةب
ع
لل
ا
ي
ذ
ه ي
ف
Tr
a
n
s
l
a
t
e
e
m
o
j
i
s me
a
n
i
n
g
لا
و
ب
ح
ب
ع
ل
لا
ب
ح
أ
ا
ن
ا
g
a
mi
n
g
ةب
ع
لل
ا
ي
ذ
ه ي
ف
R
e
m
o
v
e
d
E
n
g
l
i
s
h
ب
ح
ب
ع
لل
ا
ب
ح
أ
ا
ن
ا
لا
و
ةب
ع
لل
ا
ي
ذ
ه ي
ف
R
e
m
o
v
e
s
t
o
p
w
o
r
d
ب
ح
ب
ع
لل
ا
ب
ح
أ
ةب
ع
لل
ا
C
l
e
a
n
Tw
e
e
t
ب
ع
لل
ا
ب
ح
أ
ةب
ع
لل
ا
ب
ح
2
.
2
.
2
.
T
o
k
eniza
t
io
n
T
h
is
s
ec
tio
n
ex
p
lain
s
th
e
to
k
e
n
izatio
n
s
tep
u
s
ed
in
th
is
s
tu
d
y
.
I
n
th
e
in
itial
s
tag
es
o
f
p
r
e
p
r
o
ce
s
s
in
g
,
en
co
d
in
g
was
ap
p
lied
to
s
p
lit
th
e
tex
t
in
to
its
co
m
p
o
n
e
n
t
co
d
es,
m
ain
ly
to
allo
w
f
o
r
m
o
r
e
d
etailed
an
aly
s
is
.
T
h
e
wo
r
d
t
o
k
en
ize
f
u
n
ctio
n
f
r
o
m
t
h
e
n
atu
r
al
lan
g
u
ag
e
T
o
o
lk
it
(
NL
T
K)
,
a
well
-
r
ec
o
g
n
ized
i
n
s
tr
u
m
en
t
i
n
th
e
f
ield
o
f
NL
P,
was
u
s
ed
.
T
h
is
s
tep
is
n
ec
ess
ar
y
to
co
n
v
er
t
r
aw
te
x
t
d
ata
i
n
to
a
s
tr
u
ctu
r
ed
f
o
r
m
th
at
ca
n
b
e
ea
s
ily
an
aly
ze
d
an
d
p
r
o
ce
s
s
ed
in
th
e
f
o
llo
win
g
s
tag
es.
2
.
2
.
3
.
Ste
mm
ing
Fo
llo
win
g
th
e
to
k
en
izatio
n
,
s
tem
m
in
g
was
ap
p
lied
t
o
r
etu
r
n
wo
r
d
s
to
th
eir
r
o
o
t
f
o
r
m
,
c
o
m
b
in
in
g
v
ar
iatio
n
s
o
f
th
e
s
am
e
wo
r
d
in
to
o
n
e
r
ep
r
esen
tativ
e
f
o
r
m
.
T
h
e
Ar
ab
lig
h
t
s
tem
m
er
f
r
o
m
th
e
NL
T
K
lib
r
ar
y
was
u
s
ed
,
wh
ich
is
s
p
ec
if
ically
d
esi
g
n
ed
to
h
an
d
le
th
e
m
o
r
p
h
o
lo
g
i
ca
l
r
ich
n
ess
o
f
th
e
Ar
a
b
ic
lan
g
u
ag
e.
T
h
is
em
is
s
io
n
p
r
o
ce
s
s
is
in
teg
r
al
to
r
ed
u
cin
g
th
e
co
m
p
lex
ity
o
f
th
e
d
ata
s
et
an
d
im
p
r
o
v
in
g
th
e
ef
f
icie
n
cy
o
f
th
e
f
ea
t
u
r
e
ex
tr
ac
tio
n
p
r
o
ce
s
s
.
2
.
2
.
4
.
F
ea
t
ure
e
x
t
ra
ct
i
o
n pha
s
e
T
h
e
n
e
x
t
cr
u
cial
s
tag
e
i
n
th
e
p
r
ep
r
o
ce
s
s
in
g
was
f
ea
tu
r
e
ex
t
r
ac
tio
n
.
Var
i
o
u
s
ty
p
es
o
f
f
ea
t
u
r
es
wer
e
ex
p
lo
r
ed
,
lead
in
g
to
a
th
o
r
o
u
g
h
s
ea
r
ch
to
id
en
tify
th
e
m
o
s
t
ef
f
ec
tiv
e
co
m
b
in
atio
n
.
T
h
e
r
esu
lts
in
d
icate
d
th
at
th
e
o
p
tim
al
p
er
f
o
r
m
an
ce
was
ac
h
iev
ed
b
y
e
x
clu
s
i
v
ely
u
tili
zin
g
T
F
-
I
DF.
T
F
-
I
DF
is
a
s
tatis
t
ic
al
m
ea
s
u
r
e
u
s
ed
to
ev
alu
ate
th
e
im
p
o
r
tan
ce
o
f
a
wo
r
d
in
a
d
o
cu
m
en
t
r
elativ
e
to
a
s
et
o
f
d
o
cu
m
e
n
ts
o
r
te
x
ts
.
T
h
is
tech
n
iq
u
e
h
ig
h
lig
h
ts
wo
r
d
s
th
at
ar
e
r
ep
e
ated
with
in
a
d
o
cu
m
e
n
t
b
u
t
n
o
t
ac
r
o
s
s
d
o
cu
m
e
n
ts
,
id
e
n
tify
i
n
g
s
ev
er
al
f
ea
tu
r
es
u
s
ef
u
l f
o
r
ML
class
if
ier
s
.
2
.
2
.
5
.
Da
t
a
s
et
s
plit
An
ess
en
tial
p
ar
t
o
f
p
r
ep
ar
i
n
g
a
d
ataset
f
o
r
th
e
m
o
d
elin
g
p
r
o
ce
s
s
is
d
iv
id
in
g
th
e
d
ata
in
to
tr
ain
in
g
an
d
test
s
et
s
f
o
r
all
m
o
d
els
u
s
ed
in
th
is
s
tu
d
y
.
Allo
ca
tin
g
7
0
%
o
f
t
h
e
d
ata
to
tr
ain
i
n
g
en
s
u
r
es
th
at
m
o
d
els
lear
n
f
r
o
m
a
s
u
b
s
tan
tial
p
o
r
tio
n
o
f
th
e
d
at
aset
,
ca
p
tu
r
in
g
im
p
o
r
ta
n
t
p
atter
n
s
an
d
f
ea
tu
r
es.
T
h
e
r
em
ai
n
in
g
3
0
%
is
u
s
ed
f
o
r
test
in
g
,
allo
win
g
an
in
d
e
p
en
d
en
t
ev
alu
atio
n
o
f
th
e
m
o
d
els'
p
r
ed
ictiv
e
p
er
f
o
r
m
a
n
ce
an
d
g
en
er
aliza
tio
n
ab
ilit
y
.
T
h
is
s
ep
ar
atio
n
o
f
d
ata
in
to
tr
ain
in
g
an
d
test
in
g
s
ets
allo
ws
f
o
r
an
ef
f
ec
tiv
e
ass
ess
m
en
t
o
f
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
o
n
u
n
s
ee
n
in
p
u
ts
.
2
.
3
.
M
o
del
co
ns
t
ruct
io
n pha
s
e
T
h
e
m
o
d
el
co
n
s
tr
u
ctio
n
p
h
ase
o
f
th
is
s
tu
d
y
to
o
k
a
co
m
p
r
eh
en
s
iv
e
ap
p
r
o
ac
h
,
in
teg
r
atin
g
a
v
ar
iety
o
f
m
o
d
els
in
clu
d
in
g
ML
class
if
ier
s
an
d
DL
m
o
d
els.
T
h
is
s
tr
ateg
ic
s
elec
tio
n
o
f
m
o
d
els
aim
s
to
ac
cu
r
ately
e
v
alu
ate
an
d
co
m
p
a
r
e
th
e
ef
f
ec
tiv
en
es
s
o
f
d
if
f
er
en
t
co
m
p
u
tatio
n
al
tech
n
iq
u
es
in
p
r
o
ce
s
s
in
g
an
d
class
if
y
in
g
Ar
ab
ic
twee
ts
,
esp
ec
ially
in
th
e
c
o
n
te
x
t
o
f
d
etec
tin
g
o
f
f
en
s
iv
e
lan
g
u
ag
e.
T
h
e
f
o
llo
win
g
s
ec
tio
n
s
p
r
o
v
id
e
an
in
-
d
e
p
th
ex
p
lo
r
atio
n
o
f
th
e
m
o
d
els
u
s
ed
in
th
is
s
tu
d
y
.
T
h
e
co
m
p
u
tati
o
n
al
ex
p
er
im
e
n
ts
wer
e
co
n
d
u
ct
ed
u
s
in
g
a
s
y
s
tem
with
an
I
n
tel(
R
)
C
o
r
e(
T
M)
i5
-
1
0
3
5
G4
C
PU
r
u
n
n
in
g
at
1
.
1
0
GHz
.
T
h
e
s
y
s
tem
h
ad
8
.
0
0
GB
o
f
R
AM
in
s
talled
an
d
was r
u
n
n
in
g
a
6
4
-
b
it v
er
s
i
o
n
o
f
W
in
d
o
ws 1
0
o
n
a
x
6
4
b
ased
p
r
o
ce
s
s
o
r
.
2
.
3
.
1
.
M
a
chine
lea
rning
m
o
dels
:
T
h
e
s
t
u
d
y
i
n
i
t
i
al
l
y
u
s
e
d
t
r
a
d
i
t
io
n
a
l
M
L
c
l
a
s
s
i
f
i
e
r
s
f
o
r
t
h
e
i
r
e
f
f
i
c
i
e
n
c
y
,
i
n
t
e
r
p
r
e
t
a
b
i
l
it
y
,
a
n
d
l
o
n
g
-
t
e
r
m
s
u
c
c
es
s
i
n
v
a
r
i
o
u
s
N
L
P
t
as
k
s
.
L
o
g
i
s
t
i
c
r
e
g
r
ess
i
o
n
(
L
R
)
i
s
a
s
u
p
e
r
v
i
s
e
d
le
a
r
n
i
n
g
c
l
as
s
i
f
i
c
at
i
o
n
a
l
g
o
r
i
t
h
m
t
h
at
m
o
d
e
l
s
t
h
e
p
r
o
b
a
b
i
li
t
y
o
f
a
n
e
v
e
n
t
u
s
i
n
g
a
l
o
g
i
s
t
i
c
f
u
n
c
ti
o
n
b
as
e
d
o
n
i
n
p
u
t
f
e
at
u
r
e
s
[
1
7
]
.
S
u
p
p
o
r
t
v
e
c
t
o
r
c
l
as
s
i
f
i
e
r
(
S
VC
)
is
a
s
u
p
e
r
v
i
s
e
d
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
u
s
e
d
f
o
r
b
o
t
h
c
l
as
s
i
f
i
ca
t
i
o
n
a
n
d
r
e
g
r
es
s
i
o
n
c
h
a
ll
e
n
g
e
s
.
I
t
w
o
r
k
s
b
y
f
i
n
d
i
n
g
t
h
e
o
p
t
i
m
a
l
h
y
p
e
r
p
l
a
n
e
t
h
at
b
es
t
s
e
p
a
r
a
te
s
t
h
e
d
a
ta
p
o
i
n
ts
i
n
t
o
d
i
f
f
e
r
e
n
t
c
la
s
s
es
.
N
ai
v
e
B
a
y
es
(
N
B
)
is
c
h
o
s
e
n
f
o
r
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
A
r
a
b
ic
o
ffen
s
ive
text
cla
s
s
ific
a
tio
n
u
s
in
g
em
o
jis
:
I
n
clu
d
i
n
g
emo
ji d
a
ta
in
A
r
a
b
ic
…
(
A
ma
l
A
lb
a
la
w
i
)
3339
i
t
s
s
i
m
p
l
i
c
it
y
a
n
d
s
p
ee
d
i
n
m
a
k
i
n
g
p
r
e
d
i
c
t
i
o
n
s
;
it
as
s
u
m
es
f
e
a
tu
r
e
i
n
d
e
p
e
n
d
e
n
c
e
,
u
s
i
n
g
it
t
o
p
r
e
d
i
c
t
t
h
e
c
l
as
s
w
it
h
t
h
e
h
i
g
h
e
s
t
l
i
k
el
i
h
o
o
d
.
NB
is
c
o
m
m
o
n
l
y
a
p
p
l
i
e
d
i
n
te
x
t
c
ate
g
o
r
i
z
a
t
i
o
n
,
e
m
p
l
o
y
i
n
g
b
i
n
a
r
y
o
r
f
r
e
q
u
e
n
c
y
-
b
a
s
e
d
v
e
c
t
o
r
s
t
o
r
e
p
r
e
s
e
n
t
d
a
t
a
[
1
8
]
–
[
2
0
]
.
R
a
n
d
o
m
f
o
r
e
s
t
(
R
F
)
a
n
d
d
e
c
i
s
i
o
n
t
r
ee
s
(
D
T
)
a
r
e
cl
u
s
t
e
r
in
g
m
e
t
h
o
d
s
k
n
o
w
n
f
o
r
t
h
e
i
r
r
o
b
u
s
t
n
e
s
s
a
n
d
a
b
i
l
i
t
y
t
o
h
a
n
d
l
e
n
o
n
-
l
i
n
e
a
r
d
a
t
a
.
R
F
is
a
n
e
n
s
e
m
b
l
e
l
e
a
r
n
i
n
g
m
e
t
h
o
d
t
h
a
t
b
u
i
l
d
s
m
u
lt
i
p
l
e
d
e
c
i
s
i
o
n
t
r
e
es
a
n
d
a
g
g
r
e
g
a
t
e
s
th
e
i
r
o
u
t
p
u
t
s
t
o
i
m
p
r
o
v
e
t
h
e
a
c
c
u
r
a
c
y
a
n
d
s
t
a
b
i
l
i
t
y
o
f
c
l
a
s
s
i
f
ic
a
t
i
o
n
o
r
r
e
g
r
e
s
s
i
o
n
m
o
d
e
l
s
[
2
1
]
D
T
i
s
a
s
u
p
e
r
v
i
s
ed
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
u
s
e
d
f
o
r
c
l
a
s
s
i
f
i
ca
t
i
o
n
a
n
d
r
e
g
r
e
s
s
i
o
n
t
a
s
k
s
.
I
t
s
e
g
m
e
n
ts
d
a
ta
i
n
t
o
s
u
b
s
e
t
s
b
a
s
e
d
o
n
f
e
at
u
r
e
v
a
l
u
e
s
,
c
o
n
s
t
r
u
c
ti
n
g
a
m
o
d
e
l
t
h
a
t
r
e
s
e
m
b
l
e
s
a
t
r
e
e
s
t
r
u
c
t
u
r
e
wi
t
h
l
a
b
e
ls
at
t
h
e
l
e
a
f
n
o
d
e
s
[
1
7
]
.
X
G
B
o
o
s
t
(
X
GB
)
is
a
n
e
f
f
i
c
i
e
n
t
a
n
d
s
c
a
l
a
b
l
e
i
m
p
l
em
e
n
t
a
t
i
o
n
o
f
t
h
e
g
r
a
d
i
e
n
t
b
o
o
s
t
i
n
g
f
r
a
m
e
w
o
r
k
t
h
a
t
p
r
o
v
i
d
e
s
f
a
s
t
a
n
d
a
c
c
u
r
a
t
e
m
e
t
h
o
d
s
f
o
r
r
e
g
r
e
s
s
i
o
n
,
a
n
d
c
l
a
s
s
i
f
i
c
a
t
i
o
n
t
as
k
s
[
2
2
]
.
L
i
g
h
t
g
r
a
d
i
e
n
t
b
o
o
s
t
i
n
g
m
a
c
h
i
n
e
(
L
i
g
h
t
G
B
M
)
is
a
f
a
s
t
,
d
i
s
t
r
i
b
u
t
e
d
g
r
a
d
i
e
n
t
b
o
o
s
t
i
n
g
f
r
a
m
ewo
r
k
t
h
a
t
u
s
e
s
t
r
e
e
-
b
a
s
e
d
l
e
a
r
n
in
g
a
l
g
o
r
i
t
h
m
s
;
it
is
o
p
t
i
m
i
z
e
d
f
o
r
l
a
r
g
e
d
a
t
as
e
ts
a
n
d
h
i
g
h
l
y
e
f
f
i
c
i
e
n
t
i
n
c
o
m
p
u
t
at
io
n
a
l
p
e
r
f
o
r
m
a
n
c
e
[
2
3
]
.
2
.
3
.
2
.
Dee
p
lea
rning
m
o
dels
Ad
v
an
ce
s
in
DL
h
av
e
v
astl
y
i
m
p
r
o
v
e
d
th
e
ap
p
licatio
n
s
o
f
NL
P,
lead
in
g
to
th
e
i
n
clu
s
io
n
o
f
s
ev
er
a
l
DL
m
o
d
els.
L
o
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
an
d
b
i
d
ir
ec
tio
n
al
g
ated
r
ec
u
r
r
en
t
u
n
it
n
etw
o
r
k
(
B
iGR
U)
m
o
d
els
ar
e
ab
le
to
ca
p
tu
r
e
lo
n
g
-
ter
m
d
ep
en
d
e
n
cies
in
s
eq
u
en
tial
d
ata,
a
co
m
m
o
n
f
ea
tu
r
e
o
f
tex
t
[
2
4
]
.
B
id
ir
ec
tio
n
al
L
STM
s
:
An
ex
ten
s
io
n
o
f
tr
ad
itio
n
al
L
STM
s
th
at
in
clu
d
e
f
o
r
war
d
a
n
d
b
ac
k
war
d
p
ass
es
to
b
etter
u
n
d
er
s
tan
d
co
n
tex
t
an
d
s
em
an
tics
in
tex
tu
al
d
ata.
B
iGR
U
p
r
o
ce
s
s
es
d
ata
in
b
o
th
f
o
r
war
d
an
d
b
a
ck
war
d
d
ir
ec
tio
n
s
,
en
h
an
cin
g
s
eq
u
en
ce
m
o
d
elin
g
b
y
in
c
o
r
p
o
r
atin
g
th
e
p
ast an
d
f
u
tu
r
e
c
o
n
tex
t
[
2
5
]
.
2
.
4
.
M
o
del e
v
a
lua
t
io
n pha
s
e
T
h
e
m
o
d
el
ev
alu
atio
n
p
h
ase
is
a
cr
u
cial
p
h
ase
o
f
th
e
s
tu
d
y
an
d
ai
m
s
to
r
ig
o
r
o
u
s
ly
e
v
a
lu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
g
en
e
r
ated
m
o
d
els
in
d
if
f
er
e
n
t
d
im
en
s
i
o
n
s
.
T
h
is
ev
alu
atio
n
n
o
t
o
n
l
y
s
h
ed
s
lig
h
t
o
n
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
m
o
d
els
in
class
if
y
in
g
Ar
ab
ic
twee
ts
b
u
t
also
ex
am
i
n
es
th
e
im
p
ac
t
o
f
i
n
clu
d
in
g
em
o
jis
o
n
th
eir
p
r
ed
ictiv
e
ca
p
ab
ilit
ies.
T
o
en
s
u
r
e
a
co
m
p
r
e
h
en
s
iv
e
ev
a
lu
atio
n
o
f
th
e
m
o
d
els,
th
e
f
o
llo
win
g
m
etr
ics
an
d
asp
ec
ts
wer
e
co
n
s
id
er
ed
:
Pre
c
is
io
n
,
R
ec
all,
F1
-
Sco
r
e,
ar
ea
u
n
d
er
th
e
r
ec
eiv
er
o
p
e
r
atin
g
ch
ar
ac
ter
is
tic
cu
r
v
e
(
AUC
-
R
O
C
)
,
M
o
d
el
co
m
p
lex
ity
,
in
ter
p
r
etab
ilit
y
,
an
d
c
o
n
f
u
s
io
n
m
at
r
ix
.
T
h
e
co
m
p
ar
ativ
e
an
aly
s
is
f
o
cu
s
es
o
n
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
m
o
d
e
ls
in
two
d
if
f
er
en
t
s
ce
n
ar
io
s
:
with
em
o
jis
tr
an
s
lat
io
n
an
d
with
o
u
t
em
o
jis
as
d
escr
ib
ed
in
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
.
T
h
e
f
o
r
m
u
la
f
o
r
class
if
icatio
n
ac
cu
r
ac
y
is
t
h
e
r
atio
o
f
co
r
r
ec
t
p
r
ed
ictio
n
s
to
to
tal
p
r
ed
ictio
n
s
as sh
o
wn
b
elo
w
th
e
f
o
r
m
u
las f
o
r
ac
c
u
r
ac
y
,
F
-
S
co
r
e,
an
d
p
r
ec
is
io
n
[
2
6
]
,
[2
7
]
.
=
(
+
)
(
+
+
+
)
(
1
)
F1
-
S
co
r
e=
2
+
(
2
)
Pre
cisi
o
n
=
(
+
)
(
3
)
R
ec
all
=
(
+
)
(
4
)
w
h
er
e
in
(
1
)
-
(
4
)
.
T
P
=
tr
u
e
p
o
s
itiv
es
co
r
r
ec
tly
class
if
ied
as
p
o
s
itiv
e.
T
N
=
t
r
u
e
n
eg
ativ
es
:
T
h
ese
ar
e
in
s
tan
ce
s
co
r
r
ec
tly
class
if
ied
as n
eg
ativ
e.
FP
=
f
al
s
e
p
o
s
itiv
es
: T
h
ese
ar
e
in
s
tan
ce
s
in
co
r
r
ec
tly
class
i
f
ied
as p
o
s
itiv
e.
F
N
=
f
alse
n
eg
ativ
es
:
T
h
ese
ar
e
in
s
tan
ce
s
in
co
r
r
ec
tly
class
if
ied
as
n
eg
ativ
e.
Valid
atio
n
ac
c
u
r
ac
y
a
n
d
tr
ain
i
n
g
ac
cu
r
ac
y
m
etr
ics
d
escr
ib
e
h
o
w
well
m
o
d
els
p
er
f
o
r
m
o
n
b
o
th
tr
ain
in
g
d
ata
a
n
d
u
n
s
ee
n
v
alid
atio
n
d
ata,
h
ig
h
lig
h
tin
g
th
e
ef
f
ec
tiv
en
ess
o
f
lear
n
in
g
an
d
th
e
ab
ilit
y
to
g
en
er
alize
.
T
h
e
AUC
-
R
OC
ev
alu
ates
th
e
m
o
d
el's
d
is
cr
im
in
atio
n
a
b
ilit
y
b
etwe
e
n
class
es
at
d
if
f
e
r
en
t
t
h
r
esh
o
ld
s
ettin
g
s
,
p
r
o
v
i
d
in
g
an
o
v
er
all
m
ea
s
u
r
e
o
f
p
er
f
o
r
m
an
ce
r
e
g
ar
d
less
o
f
cl
ass
d
is
tr
ib
u
tio
n
.
Mo
d
el
c
o
m
p
lex
ity
s
tu
d
ies
th
e
s
tr
u
ctu
r
al
c
o
m
p
lex
ity
o
f
ea
ch
m
o
d
el,
in
clu
d
in
g
th
e
n
u
m
b
er
o
f
p
ar
am
eter
s
an
d
co
m
p
u
tatio
n
al
r
eq
u
ir
em
en
ts
,
wh
ich
ca
n
af
f
ec
t
d
ep
lo
y
m
en
t
an
d
s
ca
lab
ilit
y
.
I
n
ter
p
r
etab
ilit
y
is
th
e
ex
ten
t
t
o
wh
ich
m
o
d
el
d
ec
i
s
io
n
s
ca
n
b
e
u
n
d
e
r
s
to
o
d
a
n
d
i
n
ter
p
r
eted
,
wh
ich
is
im
p
o
r
tan
t
f
o
r
co
n
f
id
en
ce
an
d
ap
p
licab
ilit
y
in
r
ea
l
-
wo
r
ld
s
ce
n
ar
io
s
,
esp
ec
ially
f
o
r
s
e
n
s
itiv
e
ap
p
licatio
n
s
.
C
o
n
f
u
s
io
n
Ma
tr
ix
is
u
s
ef
u
l
in
id
en
tify
in
g
s
p
ec
if
ic
s
tr
en
g
t
h
s
an
d
wea
k
n
ess
es
o
f
th
e
class
if
icatio
n
,
p
r
o
v
id
in
g
a
d
etailed
an
aly
s
is
o
f
m
o
d
el
p
r
e
d
ictio
n
s
ac
r
o
s
s
d
if
f
er
en
t c
ate
g
o
r
ies
[2
8
]
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
ex
p
e
r
im
en
t
r
esu
lts
f
o
r
ea
c
h
m
o
d
el,
s
tar
tin
g
with
ML
class
if
ier
s
,
f
o
llo
wed
b
y
DL
m
o
d
els.
T
h
e
s
tu
d
y
ex
p
lo
r
es
th
e
clas
s
if
icatio
n
o
f
Ar
ab
ic
tex
t
in
to
o
f
f
en
s
iv
e
an
d
n
o
n
-
o
f
f
en
s
iv
e
ca
teg
o
r
ies,
au
g
m
en
ted
with
e
m
o
jis
,
to
in
v
esti
g
ate
th
e
p
r
ec
is
e
r
o
le
o
f
t
h
ese
s
y
m
b
o
ls
in
co
m
m
u
n
icati
o
n
.
E
m
o
jis
th
at
a
r
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
.
3
,
J
u
n
e
20
25
:
3
3
3
2
-
3
3
4
5
3340
ty
p
ically
ass
o
ciate
d
with
p
o
s
itiv
e
em
o
tio
n
s
o
f
te
n
h
av
e
a
m
b
i
g
u
o
u
s
m
ea
n
in
g
s
in
Ar
ab
ic
tex
ts
an
d
m
ay
co
n
v
ey
n
eg
ativ
ity
o
r
s
ar
ca
s
m
.
T
h
ese
co
m
p
lex
ities
p
o
s
e
s
ig
n
if
ican
t
ch
allen
g
es
t
o
NL
P
m
o
d
els,
wh
ich
ar
e
an
aly
ze
d
ac
r
o
s
s
a
v
ar
iety
o
f
co
m
p
u
tatio
n
al
f
r
am
ewo
r
k
s
,
i
n
clu
d
i
n
g
ML
class
if
ier
s
an
d
DL
m
o
d
els.
3
.
1
.
M
L
ex
periment
s
T
r
ad
itio
n
al
ML
class
if
ier
s
’
r
e
s
u
lts
ar
e
s
h
o
wn
in
T
ab
le
4
wit
h
em
o
jis
an
d
in
T
ab
le
5
with
o
u
t
em
o
jis
.
L
R
an
d
NB
s
h
o
w
lim
itatio
n
s
in
d
ea
lin
g
with
t
h
e
f
in
e
d
etails
o
f
em
o
jis
-
en
h
a
n
ce
d
tex
t
.
T
h
e
i
n
tr
icate
im
p
licatio
n
s
o
f
em
o
jis
in
s
en
tim
en
t
an
aly
s
i
s
ar
e
d
if
f
icu
l
t
f
o
r
th
eir
u
n
d
er
l
y
in
g
alg
o
r
ith
m
s
to
h
an
d
le
b
ec
au
s
e
th
ey
ar
e
m
ad
e
to
ev
alu
ate
s
tr
aig
h
tf
o
r
war
d
s
ig
n
a
ls
.
T
h
is
lim
itatio
n
is
ev
i
d
en
t
i
n
th
e
p
er
f
o
r
m
a
n
ce
d
is
cr
ep
an
c
y
o
b
s
er
v
ed
b
etwe
en
tex
ts
p
r
o
ce
s
s
ed
with
an
d
with
o
u
t
em
o
jis
,
h
ig
h
lig
h
tin
g
th
e
n
ee
d
f
o
r
m
o
d
els
with
en
h
an
ce
d
co
n
tex
tu
al
awa
r
en
ess
.
On
th
e
o
th
er
h
an
d
,
XGBo
o
s
t
,
R
F,
an
d
L
ig
h
tGB
M
s
h
o
wed
s
ig
n
if
ican
t
ac
cu
r
ac
y
an
d
r
e
ca
ll
in
n
o
n
-
em
o
jis
s
ce
n
ar
io
s
.
I
n
p
ar
ticu
lar
,
XGB
s
h
o
wed
h
ig
h
ac
cu
r
ac
y
,
in
d
icati
n
g
its
ef
f
ec
tiv
en
ess
in
s
ce
n
ar
io
s
with
well
-
d
ef
in
ed
f
ea
tu
r
e
s
ets a
s
s
h
o
wn
in
Fig
u
r
e
7
f
o
r
th
e
XGB co
n
f
u
s
io
n
m
a
tr
ix
an
d
Fig
u
r
e
8
f
o
r
XGB AUC
-
R
O
C
.
T
ab
le
4
.
ML
e
x
p
er
im
e
n
t r
esu
lts
with
em
o
jis
ac
r
o
n
y
m
s
M
o
d
e
l
n
a
m
e
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
A
c
c
u
r
a
c
y
Tr
a
i
n
i
n
g
a
c
c
u
r
a
c
y
AUC
-
R
O
C
M
o
d
e
l
c
o
m
p
l
e
x
i
t
y
I
n
t
e
r
p
r
e
t
a
b
i
l
i
t
y
LR
0
.
9
5
2
0
4
0
.
9
6
6
5
3
1
0
.
9
5
9
2
3
5
0
.
9
5
8
8
6
2
0
.
9
8
1
7
1
6
0
.
9
8
9
5
6
6
Lo
w
H
i
g
h
S
V
C
0
.
9
6
0
6
4
5
0
.
9
6
5
5
1
7
0
.
9
6
3
0
7
5
0
.
9
6
2
9
2
5
0
.
9
9
7
0
7
0
.
9
8
9
2
5
3
M
e
d
i
u
m
M
e
d
i
u
m
NB
0
.
8
8
2
4
6
0
0
.
9
7
4
6
4
5
0
.
9
2
6
2
6
5
0
.
9
2
2
2
9
0
.
9
7
2
7
0
1
0
.
9
8
4
4
5
4
Lo
w
H
i
g
h
RF
0
.
9
5
3
8
6
1
0
.
9
6
4
5
0
3
0
.
9
5
9
1
5
2
0
.
9
5
8
8
6
2
0
.
9
9
9
4
9
2
0
.
9
9
0
2
7
1
H
i
g
h
M
e
d
i
u
m
DT
0
.
9
5
2
7
6
3
0
.
9
6
1
4
6
0
0
.
9
5
7
0
9
2
0
.
9
5
6
8
3
0
0
.
9
9
9
4
9
2
0
.
9
5
7
3
1
2
H
i
g
h
H
i
g
h
XGB
0
.
9
6
8
3
9
9
0
.
9
6
3
4
8
8
0
.
9
6
5
9
3
7
0
.
9
6
5
9
7
2
0
.
9
8
1
4
6
2
0
.
9
9
1
8
7
4
H
i
g
h
Lo
w
t
o
M
e
d
i
u
m
Li
g
h
t
G
B
M
0
.
9
5
9
5
5
5
0
.
9
6
2
4
7
4
0
.
9
6
1
0
1
2
0
.
9
5
7
1
0
.
9
6
0
8
9
3
0
.
9
9
1
3
8
3
H
i
g
h
Lo
w
t
o
M
e
d
i
u
m
T
ab
le
5
.
ML
e
x
p
er
im
e
n
t r
esu
lts
with
o
u
t e
m
o
jis
ac
r
o
n
y
m
s
M
o
d
e
l
N
a
me
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
A
c
c
u
r
a
c
y
Tr
a
i
n
i
n
g
A
c
c
u
r
a
c
y
AUC
-
R
O
C
M
o
d
e
l
C
o
m
p
l
e
x
i
t
y
I
n
t
e
r
p
r
e
t
a
b
i
l
i
t
y
LR
0
.
9
5
2
0
4
8
0
.
9
6
6
5
3
0
.
9
5
9
2
3
0
.
9
5
8
8
6
2
0
.
9
8
1
7
1
7
0
.
9
8
9
5
6
6
Lo
w
H
i
g
h
S
V
C
0
.
9
6
0
6
4
6
0
.
9
6
5
5
1
7
0
.
9
6
3
0
7
0
.
9
6
2
9
2
5
0
.
9
9
7
0
8
0
0
.
9
8
9
2
5
3
M
e
d
i
u
m
NB
0
.
8
9
2
8
5
7
0
.
9
7
4
2
4
0
0
.
9
3
1
7
7
0
.
9
2
8
0
9
6
0
.
9
7
9
4
4
9
0
.
9
8
7
9
0
8
Lo
w
RF
0
.
9
6
9
4
9
6
0
.
9
6
5
6
5
4
0
.
9
6
7
5
7
0
.
9
6
7
3
7
7
0
.
9
9
8
7
1
6
0
.
9
8
9
7
1
0
H
i
g
h
M
e
d
i
u
m
DT
0
.
9
6
4
2
1
5
0
.
9
6
1
0
3
0
0
.
9
6
2
6
2
0
.
9
6
2
3
8
3
0
.
9
9
8
8
5
8
0
.
9
6
3
3
3
4
H
i
g
h
H
i
g
h
XGB
0
.
9
7
0
8
2
2
0
.
9
6
6
9
7
5
0
.
9
6
8
8
9
0
.
9
6
8
7
0
8
0
.
9
7
9
8
7
7
0
.
9
9
3
6
0
4
H
i
g
h
Lo
w
t
o
M
e
d
i
u
m
Li
g
h
t
G
B
M
0
.
9
6
6
1
5
8
0
.
9
6
1
6
9
1
0
.
9
6
3
9
1
0
.
9
6
3
7
1
5
0
.
9
8
5
4
4
3
0
.
9
9
3
2
1
4
H
i
g
h
Lo
w
t
o
M
e
d
i
u
m
Fig
u
r
e
7
.
XGB co
n
f
u
s
io
n
m
atr
ix
Fig
u
r
e
8
.
AUC
-
R
OC
f
o
r
XGB
3
.
2
.
DL
ex
periment
s
r
e
s
ult
s
T
h
is
s
ec
tio
n
d
is
cu
s
s
es
th
e
r
e
s
u
lts
f
o
r
th
e
DL
m
o
d
el.
T
h
is
s
tu
d
y
co
n
s
id
er
s
th
r
ee
s
tate
-
of
-
th
e
-
a
r
t
ar
ch
itectu
r
es:
L
STM
,
B
iLST
M,
an
d
B
iGR
U.
T
h
e
co
n
s
tr
u
ctio
n
o
f
th
e
m
o
d
els
is
th
e
s
am
e
i
n
b
o
th
e
x
p
er
im
e
n
ts
:
u
s
in
g
em
o
jis
an
d
with
o
u
t e
m
o
jis
.
T
ab
le
6
p
r
o
v
id
es th
e
m
o
d
e
l in
d
etail.
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
A
r
a
b
ic
o
ffen
s
ive
text
cla
s
s
ific
a
tio
n
u
s
in
g
em
o
jis
:
I
n
clu
d
i
n
g
emo
ji d
a
ta
in
A
r
a
b
ic
…
(
A
ma
l
A
lb
a
la
w
i
)
3341
E
ac
h
m
o
d
el
was
ev
al
u
ated
b
as
ed
o
n
t
h
e
n
u
m
b
er
o
f
lay
er
s
,
n
e
u
r
o
n
s
,
b
atc
h
s
ize,
d
r
o
p
o
u
t,
lear
n
in
g
r
ate,
p
r
ec
is
io
n
,
r
ec
all,
F1
-
S
co
r
e
,
AUC
-
R
O
C
,
m
o
d
el
co
m
p
lex
it
y
,
an
d
in
ter
p
r
etab
ilit
y
.
T
h
e
m
o
d
els’
p
er
f
o
r
m
an
ce
r
esu
lts
with
em
o
ji
ac
r
o
n
y
m
s
ar
e
p
r
o
v
id
e
d
in
T
ab
le
7
,
an
d
th
e
m
o
d
els’
r
esu
lts
with
o
u
t
e
m
o
ji
ac
r
o
n
y
m
s
a
r
e
p
r
o
v
id
e
d
in
T
ab
le
8
.
T
h
ese
a
r
c
h
itectu
r
es
ex
ce
l
at
ca
p
t
u
r
in
g
lo
n
g
-
ter
m
d
ep
e
n
d
en
cies
an
d
s
h
o
w
h
ig
h
s
en
s
itiv
ity
to
co
n
tex
tu
al
a
n
d
em
o
tio
n
al
c
u
es p
r
o
v
id
e
d
b
y
em
o
jis
.
As ev
id
en
ce
d
b
y
th
eir
co
n
tin
u
ally
b
etter
p
er
f
o
r
m
an
ce
in
tex
ts
with
a
h
ig
h
e
m
o
ji
co
n
ten
t,
th
ese
m
o
d
els
f
r
eq
u
en
t
ly
s
u
cc
ess
f
u
lly
id
en
tify
t
h
e
n
u
an
ce
d
em
o
tio
n
al
u
n
d
er
to
n
es
p
r
es
en
t
in
e
m
o
jis
.
T
h
is
u
n
d
er
lin
es
th
e
i
m
p
o
r
ta
n
t
r
o
le
o
f
em
o
jis
in
co
n
v
ey
i
n
g
co
m
p
lex
em
o
tio
n
al
s
tates,
esp
ec
ially
in
a
lin
g
u
is
ti
ca
lly
r
ich
lan
g
u
ag
e
lik
e
Ar
ab
ic
.
B
iLST
M
d
em
o
n
s
tr
ated
th
e
h
i
g
h
est
lev
els
o
f
ac
cu
r
ac
y
an
d
e
x
ce
llen
t
o
v
e
r
all
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
all
m
etr
ics
in
DL
m
o
d
els.
Fig
u
r
e
9
s
h
o
ws
th
e
t
r
ain
in
g
an
d
v
al
id
atio
n
ac
cu
r
ac
y
,
in
a
d
d
itio
n
t
o
th
e
tr
ain
i
n
g
a
n
d
v
alid
atio
n
l
o
s
s
f
o
r
B
iLST
M
with
em
o
jis
.
Als
o
,
Fig
u
r
e
1
0
s
h
o
ws
th
e
tr
ain
in
g
a
n
d
v
alid
atio
n
ac
cu
r
ac
y
,
in
ad
d
itio
n
to
th
e
tr
ain
in
g
an
d
v
alid
atio
n
l
o
s
s
f
o
r
B
iLST
M
b
u
t w
ith
o
u
t e
m
o
jis
.
T
ab
le
6
.
Ar
c
h
itectu
r
al
d
etails o
f
DL
m
o
d
els u
s
ed
ac
r
o
s
s
b
o
t
h
ex
p
er
im
e
n
ts
M
o
d
e
l
n
a
m
e
N
u
mb
e
r
o
f
l
a
y
e
r
s
N
e
u
r
o
n
s
B
a
t
c
h
si
z
e
D
r
o
p
o
u
t
Le
a
r
n
i
n
g
r
a
t
e
Ep
o
c
h
O
p
t
i
mi
z
e
r
LSTM
2
1
2
8
32
0
.
5
0
.
0
0
0
1
10
A
d
a
m
B
i
LST
M
2
1
2
8
32
S
p
a
t
i
a
l
:
0
.
3
,
B
i
LST
M
:
0
.
2
5
,
A
d
d
i
t
i
o
n
a
l
:
0
.
5
0
.
0
0
0
5
10
A
d
a
m
B
i
G
R
U
2
64
32
0
.
3
D
e
f
a
u
l
t
o
f
A
d
a
m
10
A
d
a
m
T
ab
le
7
.
DL
e
x
p
er
im
en
t
r
esu
lts
with
em
o
jis
ac
r
o
n
y
m
s
M
o
d
e
l
N
a
me
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
Tr
a
i
n
i
n
g
a
c
c
u
r
a
c
y
AUC
-
R
O
C
M
o
d
e
l
c
o
m
p
l
e
x
i
t
y
I
n
t
e
r
p
r
e
t
a
b
i
l
i
t
y
LSTM
0
.
9
6
3
1
0
.
9
5
6
8
0
.
9
5
9
9
0
.
9
5
8
7
0
.
9
9
3
5
0
.
9
5
8
7
H
i
g
h
Lo
w
B
i
LST
M
0
.
9
7
3
1
0
.
9
5
0
2
0
.
9
6
1
5
0
.
9
6
0
7
0
.
9
9
1
1
0
.
9
6
1
1
H
i
g
h
Lo
w
B
i
G
R
U
0
.
9
4
6
2
0
.
9
4
1
8
0
.
9
4
4
0
0
.
9
4
3
4
0
.
9
8
0
3
0
.
9
4
3
4
H
i
g
h
Lo
w
T
ab
le
8
.
DL
e
x
p
er
im
en
t
r
esu
lts
with
o
u
t e
m
o
jis
ac
r
o
n
y
m
s
M
o
d
e
l
N
a
me
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
Tr
a
i
n
i
n
g
a
c
c
u
r
a
c
y
AUC
-
R
O
C
M
o
d
e
l
c
o
m
p
l
e
x
i
t
y
I
n
t
e
r
p
r
e
t
a
b
i
l
i
t
y
LSTM
0
.
9
7
0
3
0
.
9
4
8
6
0
.
9
5
9
3
0
.
9
5
9
3
0
.
9
9
1
3
0
.
9
5
9
5
0
H
i
g
h
Lo
w
B
i
LST
M
0
.
9
5
9
3
0
.
9
6
4
4
0
.
9
6
1
8
0
.
9
6
1
3
0
.
9
9
7
1
0
.
9
6
1
3
5
H
i
g
h
Lo
w
B
i
G
R
U
0
.
9
3
6
1
0
.
9
5
1
1
0
.
9
4
3
5
0
.
9
4
2
4
0
.
9
8
9
8
0
.
9
4
2
3
4
H
i
g
h
Lo
w
Fig
u
r
e
9
.
T
est
ac
cu
r
ac
y
v
s
v
ali
d
atio
n
ac
cu
r
ac
y
an
d
test
lo
s
s
v
s
v
alid
atio
n
lo
s
s
f
o
r
B
iLST
M
m
o
d
el
with
o
u
t
em
o
jis
3
.
3
.
Dis
cu
s
s
io
n
ML
tr
ee
-
b
ased
class
if
ier
s
,
esp
ec
ially
XGBo
o
s
t
(
XG
B
)
,
p
er
f
o
r
m
ed
well,
p
ar
ticu
lar
ly
in
th
e
ab
s
en
ce
o
f
em
o
jis
.
XGB
o
u
tp
er
f
o
r
m
e
d
b
o
th
DL
a
n
d
tr
a
n
s
f
o
r
m
e
r
m
o
d
els
in
ter
m
s
o
f
ac
c
u
r
ac
y
in
th
is
d
ataset.
T
h
is
Evaluation Warning : The document was created with Spire.PDF for Python.