Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
11
,
No.
1
,
Febr
uar
y
2021
, pp.
753
~
76
2
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v11
i
1
.
pp753
-
762
753
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om
Differ
ent valuabl
e
t
ools
f
or
A
rabic
s
entim
en
t
a
nal
ysis
:
a
c
omp
arativ
e evaluati
on
Y
oussr
a
Z
ah
idi
1
, Y
acine
El
Youn
ou
ssi
2
, Y
as
sine
Al
-
Amr
an
i
3
1
,2
Inform
at
ion
S
y
stem
and
Softw
are
Engi
n
ee
ring
La
bora
tor
y
,
Abdelma
le
k
Essaa
d
i U
nive
rsit
y
,
Morocc
o
3
Te
chno
logi
es
d
e
l’Inf
orm
a
ti
on
e
t
Modél
isat
ion
d
es
S
y
st
èmes,
Ab
del
m
a
le
k
Essaa
d
i
Univer
si
t
y
,
Morocc
o
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
9
, 2
01
9
Re
vised
A
ug 11
,
2020
Accepte
d
Aug 26,
2020
Arabi
c
Natur
al
la
nguage
pro
c
essing
(AN
LP)
is
a
subfiel
d
of
art
ificial
int
ellige
n
ce
(AI
)
tha
t
tries
to
buil
d
v
ari
ous
a
ppli
c
at
ions
in
t
he
Arabi
c
la
nguag
e
li
ke
Arabi
c
senti
m
e
nt
ana
l
y
sis
(AS
A)
tha
t
is
the
oper
ation
of
cl
assif
y
ing
the
f
ee
l
ings
and
emo
ti
ons
expr
essed
for
def
ini
ng
the
at
ti
tud
e
of
the
write
r
(n
eu
tra
l
,
negative
o
r
positi
ve)
.
In
orde
r
to
work
on
AS
A,
rese
arc
h
ers
ca
n
use
var
ious
tool
s
in
the
ir
rese
ar
ch
proje
c
ts
without
expl
a
ini
ng
the
ca
use
beh
in
d
thi
s
use,
or
they
choose
a
set
of
li
bra
r
ie
s
a
cc
ording
to
the
ir
knowledg
e
about
a
spec
if
i
c
progra
m
m
ing
la
nguag
e
.
Be
ca
u
se
of
the
ir
li
bra
r
ie
s
'
abund
a
nce
in
the
AN
L
P
fie
ld
,
espe
ci
a
lly
in
AS
A,
we
ar
e
re
l
y
ing
on
JA
VA
and
Py
th
on
progra
m
m
ing
la
nguag
es
in
ou
r
rese
arc
h
work
.
Thi
s
pape
r
rel
i
es
on
m
aki
n
g
an
i
n
-
dept
h
c
om
par
at
ive
ev
aluati
on
o
f
diff
er
ent
v
al
uab
le
P
y
thon
and
Java
li
bra
rie
s
to
ded
uce
the
m
ost
useful
ones
in
Arab
ic
senti
m
ent
ana
l
y
sis
(AS
A).
Acc
ording
to
a
l
arg
e
var
ie
t
y
of
g
rea
t
and
infl
u
entia
l
works
in
the
dom
ai
n
of
AS
A,
we
dedu
ce
that
the
NL
TK,
Gensim
and
Te
xtBl
ob
li
bra
r
ie
s
are
th
e
m
ost
useful
for
P
y
thon
AS
A
ta
sk.
I
n
conne
c
ti
on
with
Java
AS
A
li
bra
rie
s,
w
e
conclude
that
W
eka
and
Cor
e
NLP
tool
s
are
th
e
m
ost
used,
and
th
e
y
h
ave gr
ea
t
result
s
in
t
his
rese
a
rch
dom
a
i
n.
Ke
yw
or
d
s
:
AN
L
P
ASA
A
SA
program
m
ing
lan
guage
s
Java
li
braries
Pyt
hon
li
brarie
s
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
B
Y
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
Youssra Za
hid
i
,
Inform
at
ion
System
an
d So
ftwar
e
Enginee
ri
ng
La
borato
ry,
Abdelm
al
ek
Essaadi
Un
i
ver
sit
y
, Tetua
n,
M
orocco.
Em
a
il
:
yous
sra
1994zahi
di@
gm
ai
l.co
m
1.
INTROD
U
CTION
Natu
ral
l
angu
age
p
ro
ces
sin
g
(NLP)
is
a
su
bfi
el
d
of
com
pu
te
r
sci
ence,
li
nguisti
cs,
arti
fici
al
intel
li
gen
ce
,
a
nd
in
form
ation
en
gin
ee
rin
g
i
nterested
by
t
he
interact
io
ns
betwee
n
hu
m
an
(n
at
ur
al
)
la
ngua
ges
and
c
om
pu
te
rs
,
in
pa
rtic
ular
how
to
pro
gr
a
m
co
m
pu
te
rs
t
o
treat
an
d
pr
ocess
a
m
assiv
e
qu
a
n
ti
ty
of
natu
ral
la
nguag
e
data.
Ar
a
bic
n
at
ura
l
la
ng
ua
ge
pr
oc
essing
A
N
LP
trie
s
to
bu
il
d
so
ftwa
re
el
igi
ble
to
treat
A
rab
ic
li
ng
uisti
c
data
autom
at
ic
ally
fo
r
a
sp
eci
fic
a
pp
li
cat
io
n.
T
he
Ar
a
bic
la
ng
ua
ge
is
rec
ogni
zed
as
the
4t
h
m
os
t
us
e
d
la
ngua
ge
of
the
In
te
rn
e
t.
It
is
the
fo
rm
al
la
ng
ua
ge
of
twenty
-
tw
o
c
ountries,
s
poke
n
by
m
or
e
than
f
our
hundre
d
m
illi
on
sp
e
ake
rs
.
It
is
a
Sem
it
ic
lan
gua
ge
that
is
char
act
eri
z
e
d
by
it
s
li
te
rar
y
abun
dan
ce
.
A
rab
ic
m
or
phology
is
rich,
com
plex,
an
d
highly
a
m
big
uous
.
F
or
this
reas
on,
it
po
ses
a
va
riet
y
of
prob
l
e
m
s
in
the
fiel
d
of
NL
P.
Nowa
days,
AN
L
P
has
obta
ined
sig
nific
a
nt
val
ue
.
A
la
r
ge
var
ie
ty
of
a
pp
li
cat
io
ns
have
bee
n
bu
il
t
li
ke:
sent
i
m
ent
analy
sis
[1
,
2]
,
m
achin
e
translat
io
n,
quest
io
n
ans
we
r
ing
,
nam
ed
ent
it
y
reco
gnit
ion,
et
c.
These
a
pp
li
cat
ion
s
m
us
t
adap
t
to
the
com
plic
at
ed
struct
ur
e
of
A
ra
bic
[
3]
.
This
S
em
itic
l
angua
ge
has
it
s
own
s
pecial
feat
ur
e
s
;
for
e
xam
ple,
it
has
no
c
ap
it
al
i
z
ation
;
the
A
rab
ic
al
ph
a
bet
co
n
ta
ins
29
c
onsona
nts
and
11
vowels.
Mo
re
over
,
the
Ar
a
bic
la
ng
ua
ge
is
w
ritt
en
from
righ
t
to
le
ft
,
and
i
ts
le
tt
ers
chan
ge
fo
rm
at
dep
en
ding
on their
place
i
n
the
wo
r
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
75
3
-
76
2
754
AN
L
P
-
relat
ed
pro
blem
s
[4
,
5]
can
be
s
um
m
arized
as
fo
ll
ows:
T
o
be
gin
with,
t
he
pro
blem
of
m
ul
ti
ple
vowell
at
ion
an
d
t
he
com
plexity
of
the
A
rab
ic
gr
a
ph
ic
wor
d
str
uc
ture
(a
n
Ar
a
bi
c
descr
ipti
ve
wor
d
can
co
rr
es
pond
to
a
w
ho
le
sentence
in
F
r
ench).
Be
side
s
,
the
w
ord
or
der
is
relat
ivel
y
fr
ee
in
an
Ar
a
bic
sentence
(v
e
r
b
+
obj
ect
+
subj
ect
(VOS
);
ve
rb
+
s
ubj
ect
+
obj
ect
(
VSO
);
ob
j
ect
+
verb
+
sub
j
ect
(
O
VS
)
).
Ar
a
bic
do
es
no
t
con
ta
in
capit
a
l l
et
te
rs
u
nlike
m
os
t Lat
in lang
ua
ges
. T
his m
akes
ANLP,
such as rec
ogniti
on of
entit
y
na
m
es,
ver
y
dif
ficult
.
A
no
t
her
disti
nctive
featu
re
of
Ar
a
bic
is
diacrit
ic
al
m
ark
s
(s
hort
vo
wels):
the
sa
m
e
wo
r
d
with
dif
fer
e
nt
diacrit
ic
s
can
ex
pr
ess
dif
fer
e
nt
m
eaning
s.
Diac
riti
cs
are
usual
ly
omi
t
te
d,
causin
g
am
big
uity
.
Also,
t
his
Sem
it
ic
lan
gua
ge
is
very
inflect
ion
al
an
d
der
i
vative
,
w
hich
m
akes
it
s
m
or
phologica
l
analy
sis
a
co
m
plex
ta
sk
.
It
is
der
ivat
ive
i
n
that
al
l
the
Ar
a
bic
w
ords
hav
e
a
t
hr
ee
-
or
four
-
char
act
e
r
r
oot
verb
,
a
nd
it
is
inflect
io
nal
bec
ause
eac
h
wor
d
c
on
sist
s
of
z
ero
or
m
or
e
affi
xes
(
pr
e
fix
,
in
fix
a
nd
su
f
fix
)
an
d
a
r
oo
t.
As
a
res
ult, the lack
of
Arabic res
ources,
su
c
h
as li
br
a
ries, that sup
port
the Arab
ic
lan
gu
a
ge
corp
or
a
m
akes
ANLP
re
sear
ch
m
or
e
chall
eng
i
ng.
F
or
t
his
reason,
A
rab
i
c
is
m
or
e
com
plex
a
nd
dif
fic
ult
to
process
in NL
P f
ie
ld
co
m
pared to
the
ot
her
fam
ou
s lan
gu
a
ges,
Applic
at
ion
s
of
A
NL
P
are
widely
sp
rea
d
beca
us
e
pe
op
le
com
m
un
ic
at
e
a
l
m
os
t
e
ver
yt
hi
ng
in
la
nguag
e
[
6]
.
Am
on
g
t
hese
a
pp
li
cat
io
ns
,
we
fo
c
us
on
A
ra
bi
c
senti
m
ent
analy
sis
in
ou
r
researc
h
st
ud
y.
ASA
or
op
i
nion
m
i
ning
ai
m
s
at
def
inin
g
the
at
ti
tud
e,
t
he
senti
m
ent
po
la
rity
(positi
vity
,
ne
utrali
ty
,
neg
at
i
vity
)
o
f
a
wr
it
er
or
a
not
he
r
s
ubj
ect
c
on
ce
r
ning
a
pa
rtic
ular
e
ven
t
[7
,
8]
.
A
la
r
ge
nu
m
ber
of
sen
tim
ents
are
bo
rn
e
i
n
po
sts
on
m
any
so
ci
al
m
edia
platfor
m
s
li
ke
(T
witt
e
r,
Face
bo
ok, Y
ouT
ub
e
,
I
ns
ta
gram
).
Sent
i
m
ent
a
l
An
al
ysi
s
is
perfo
rm
ed
us
i
ng
va
rio
us
m
achine
le
a
rn
i
ng
te
ch
niques
[9
-
11
,
sta
ti
sti
c
al
m
od
el
s
,
a
nd
N
LP
for
fe
at
ur
e
extracti
on
fro
m
extensiv
e d
a
ta
.
Sentim
ent
ana
ly
sis
has
var
iou
s
tre
ndin
g
app
li
cat
io
ns
in
m
any
fiel
ds
.
In
poli
ti
cs,
it
can
ai
d
in
inferrin
g
the
fr
ee
ori
entat
io
n
an
d
reacti
on
towa
r
ds
pol
it
ic
al
even
ts,
wh
ic
h
hel
ps
in
decisi
on
m
akin
g.
In
busines
s,
it
per
m
it
s
com
p
anies
to
aut
oma
ti
cal
ly
colle
ct
their
c
us
tom
ers'
op
i
nions
on
their
se
rv
ic
es
[12]
.
Sentim
ent an
al
ysi
s can be
done
at
seve
ral
le
ve
ls, doc
um
ent l
evel
[
13]
, s
e
ntence
-
le
vel
,
a
nd
su
bject
-
le
vel
.
In
t
his
resea
rc
h
pr
oj
ect
,
w
hic
h
re
l
ie
s
on
the
do
m
ai
n
of
A
S
A,
we
try
to
do
a
c
om
par
at
ive
evaluati
on
to
co
nclu
de
t
he
m
os
t
valuab
l
e
pro
gr
am
m
in
g
la
ng
uag
e
s,
wh
ic
h
a
re
a
bund
a
nt
at
th
e
le
vel
of
ASA
li
b
ra
ries.
We
com
par
e
these
li
br
a
ries
to
de
du
ce
the
m
os
t
po
we
rful
on
e
s.
Wh
e
n
w
e
ta
lk
abo
ut
th
e
Eng
li
sh
la
ng
uag
e
,
f
or
exam
ple,
there
are
var
i
ou
s
N
LP
too
ls
ad
va
ntage
ous
for
va
rio
us
NL
P
ta
sk
s
,
es
pecial
ly
i
n
sentim
ent
analy
sis
SA
.
Nev
e
rthel
ess
,
this
is
no
t
the
s
am
e
sit
uat
ion
f
or
t
he
A
ra
bic
la
ngua
ge.
Du
e
to
it
s
am
big
uity
,
synta
ct
ic
and
m
or
phologica
l
abun
dan
ce
a
nd
richness,
the
Ar
a
bic
la
ngua
ge
is
deem
ed
as
the
m
os
t
diffi
cult
la
ng
ua
ge
.
Ther
e
are
a
lim
it
ed
nu
m
ber
of
li
br
aries
that
sup
port
it
.
T
his
co
m
plex
natur
e
,
wit
h
the
la
c
k
of
it
s
res
ource
s
an
d
the
di
ver
sit
y
of
diale
ct
s
,
im
po
se
s
dif
ficult
ie
s
on
the
de
vel
op
m
ent
in
the
fiel
d
of
ASA
r
esearch
.
Ch
oos
ing
t
he
m
os
t
app
ropr
ia
te
gr
ou
p
of
li
br
aries
that
m
ee
ts
ou
r
s
pecific
need
s
is
ve
ry
diff
ic
ult
and
im
po
ses
a
n
in
-
dept
h
evaluati
on.
T
o
so
l
ve
t
his
m
ajor
pro
blem
,
we
rely
on
a
var
ie
ty
of
val
uab
le
aspects
in
this
com
par
at
ive
evaluati
on.
Th
is
com
par
at
ive
evaluati
on
is
crit
ic
al
,
in
th
at
it
wo
ul
d
e
na
ble
va
rio
us
r
esearche
rs
w
ho
are
interest
ed
in
usi
ng
ASA
in
t
heir
proj
ect
s
t
o
bui
ld
a
ppr
opriat
e
decisi
ons
about
avail
ab
le
li
br
aries
tha
t
m
eet
their r
e
quirem
e
nts a
nd n
ee
ds
a
ccur
at
el
y
.
The
r
est
of
th
e
p
a
p
er
is
des
cribe
d
li
ke
thi
s
:
t
he
seco
nd
sect
ion
em
ph
a
siz
es
A
SA
pr
ogram
m
ing
la
nguag
e
s
an
d
their
fam
ou
s
L
ibrar
ie
s
.
Sect
io
n
3
offer
s
our
in
-
de
pth
com
par
iso
n
betw
een
t
he
m
os
t
us
efu
l
Java
and
Pyt
hon
li
braries
f
or
ASA
.
The
resu
lt
s
a
re
de
bated
i
n
detai
l
in
s
ect
ion
4
,
an
d
th
is
work
is
finis
h
ed
with
final th
oughts i
n
s
ect
io
n
5
.
2.
APPLIE
D
P
R
OGRA
MMIN
G LA
NGUA
G
ES A
ND LIB
RARIES
V
ari
ou
s
pr
ogr
a
m
m
ing
la
ngua
ges
a
re
a
ppli
ed
i
n
Deep
L
earn
i
ng
an
d
AN
L
P
(like
C
++,
R,
Perl,
Pr
ol
og,
Lisp
...).
Th
ese,
how
ever,
a
re
known
by
their
s
carcit
y
of
ap
pro
pr
ia
te
gro
up
s
of
li
braries
us
e
d
i
n
m
od
ern
a
pp
li
c
at
ion
s
in
th
ese
do
m
ai
ns
.
N
owadays,
va
rio
us
Pyt
hon
an
d
Ja
va
li
br
a
ries
ha
ve
bee
n
bu
il
t
to
cat
er
to
the
re
quire
m
ents
and
nee
ds
in
c
urre
nt
Deep
Lear
ning
and
A
NLP
m
od
e
r
n
ta
sk
s
.
I
n
view
of
t
hese
too
ls'
abun
dan
ce
,
re
pu
ta
ti
on,
a
nd
high
pe
rfo
rm
a
nce,
Ja
va
a
nd
Pyt
hon
can
be
con
si
der
e
d
as
the
m
os
t
widely
us
e
d
pro
gr
am
m
ing
la
nguag
e
s
in
these
dom
ai
ns
.
That
is
wh
y
we
are
basin
g
on
Ja
va
a
nd
Pyt
hon
pro
gra
m
m
ing
la
nguag
e
s
in
t
his
in
-
de
pth
e
valuati
on,
as
t
hey
are
the
m
os
t
com
m
on
ly
us
ed
.
W
e
wi
ll
be
beg
i
nn
i
ng
by
choosi
ng
ap
pr
opriat
e
progra
m
m
ing
la
ng
ua
ges
to
co
nclu
de
wh
at
are
ide
ntify
their
m
os
t
po
te
nt
li
br
a
ries
in
the ASA
dom
a
in.
2
.
1.
A
r
ab
ic
se
nt
im
en
t an
alysi
s using P
yt
h
on
Pyt
hon
is
a
po
werfu
l
pro
gr
a
m
m
ing
la
ngua
ge
with
e
xcell
ent
f
unct
ion
al
it
y
and
feat
ur
e
f
or
proce
ssi
ng
natu
ral
la
ngua
ge,
it
s
sem
antic
s
an
d
sy
ntax
are
tra
nspare
nt,
an
d
it
has
e
xc
el
le
nt
string
-
ha
nd
li
ng
f
un
ct
i
on
al
it
y.
As
an
ob
j
ect
-
ori
ented
la
ng
ua
ge,
Pyt
hon per
m
it
s
m
et
ho
ds
a
nd
data to b
e e
ncapsulat
ed
a
nd r
e
us
ed
easi
ly
. A
s an
interp
reted
la
ngua
ge,
Pyt
ho
n
facil
it
at
es
interact
ive
ex
plorat
ion.
As
a
d
ynam
ic
la
nguag
e
,
Pyt
hon
al
lows
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Diff
erent val
uable to
ols
for
ar
ab
ic
sentime
nt
analysis:
a co
mpar
ative
eval
ua
ti
on
(
Y
oussr
a
Z
ahidi
)
755
var
ia
bles
to
be
ty
ped
dynam
icall
y
and
add
at
tribu
te
s
to
ob
j
e
ct
s
on
the
fly
,
facil
it
ating
rap
i
d
de
velo
pm
ent
[14]
.
Ma
ny
researc
he
rs
rec
omm
en
d
this
po
werful
pr
og
ram
m
ing
la
nguag
e
.
F
or
instance:
Stev
en
Bi
rd
a
nd
E
dwa
r
d
Lo
per
in
[
14
]
strongly
rec
om
m
end
the
use
of
Pyt
hon
in
NLP
pro
j
ect
s,
and
t
hro
ugh
t
heir
work
[
15
]
,
they
deduce t
hat thi
s progra
m
m
ing
langua
ge
is t
he
b
est
,
providi
ng a lar
ge vari
et
y of
be
nef
it
s.
In
his
pap
e
r
[16]
,
Niti
n
Ma
dn
a
ni
ch
ose
to
e
m
plo
y
Pyt
ho
n
beca
us
e
he
c
onfir
m
s
that
this
pro
gr
am
m
ing
la
ngua
ge
has
a
la
rg
e
var
ie
ty
of
be
ne
fits
over
the
oth
e
r
program
m
ing
la
ngua
ges,
s
uch
as
an
easy
-
to
-
us
e
ob
j
ect
-
or
ie
nted
pa
rad
i
gm
,
hig
h
read
a
bili
ty
,
strong
U
nico
de
support
,
easy
extensi
bili
ty
,
an
d
a pow
e
rful sta
nd
a
r
d
li
br
a
ry.
I
t i
s v
ery
e
ff
ic
ie
nt and
has
b
ee
n
a
pp
li
ed
in
com
plex
and
dif
f
ic
ult
NLP p
roj
ect
s.
The
T
heano
D
evelo
pm
ent
Tea
m
enco
ura
ge
s
the
us
e
of
Pyt
hon
th
rou
gh
t
hi
s
work
[17]
,
wh
ic
h
they
consi
der
a
fle
xib
le
pro
gr
am
m
ing
la
nguage
pr
ovidi
ng
a
strai
gh
tf
orwa
rd
m
ann
er
to
react
with
data
and
al
lowing
for
fast
pr
oto
ty
pi
ng.
M
or
e
ov
e
r,
pa
per
[
18
]
offer
s
a
c
riti
cal
assessm
ent
of
e
xisti
ng
Pyt
hon
infr
a
st
ru
ct
ur
e
f
or
NL
P
new
ta
sk
s.
T
hroug
h
t
heir
case
stu
dy
:
Au
t
om
at
ic
Asp
ect
ual
Cl
assi
ficat
ion
of
V
er
bs
i
n
an
U
ntag
ge
d
Corpus,
the
a
uthors
fou
nd
t
hat
Pyt
hon
’s
cor
e
li
braries
offer
perfect
cov
e
ra
ge
of
es
sentia
l
m
achine
le
ar
nin
g al
gorithm
s.
Pyt
hon
is
esp
eci
al
ly
m
o
re
a
ppr
opriat
e
for
var
i
ou
s
reas
ons:
fr
ee
a
nd
si
m
ple,
ob
j
ect
-
ori
ented
,
a
nd
com
patible
with
s
o
m
any
pla
tfor
m
s,
a
la
r
ge
num
ber
of
li
braries
f
or
Pyt
hon.
Neverthele
ss,
we
al
s
o
ha
ve
t
o
know
the
dow
ns
ides
of
c
hoosi
ng
it
ov
e
r
an
oth
e
r
pro
gr
am
m
ing
la
ngua
ge
:
Sp
eed
li
m
it
ation
s
,
W
ea
k
in
m
ob
il
e
com
pu
ti
ng
,
a
nd
browsers
.
2
.
1.1.
Py
thon l
ibrarie
s
It
is
fun
dam
ent
al
first
to
sho
w
the
m
os
t
us
ef
ul
Pyt
hon
li
bra
ries
that
ha
ve
been
pro
ve
n
in
the
dom
ai
n
of A
S
A:
NLT
K,
TextBl
ob, a
nd G
e
ns
im
.
a.
NLT
K:
is
a
le
adin
g
platfo
rm
for
NLP.
A
s
et
of
co
re
m
odules
(lib
rar
ie
s
and
pr
ogram
s)
offe
rs
basic
da
ta
ty
pes
that
are
ut
il
iz
ed
through
ou
t
the
to
ol.
N
LTK
is
a
pe
rf
e
ct
sta
rting
poin
t
fo
r
researc
he
r
s
and
st
ud
e
nts
in
the
do
m
ai
n
of
NLP
b
ecause
of
it
s
nu
m
erou
s
ben
e
fits.
Tha
t
is
wh
y
NLT
K
has
been
na
m
ed
"a
wo
nde
rful
too
l
f
or
te
ac
hi
ng
a
nd
w
orki
ng
in
c
om
pu
ta
ti
on
al
li
nguisti
c
s
us
in
g
Pyt
ho
n"
and
t
he
"m
oth
er"
of
al
l
N
LP
li
br
aries.
T
he
sign
ific
a
nt
ad
va
ntage
of
us
in
g
NL
TK
is
th
a
t
it
is
entirel
y
sel
f
-
c
on
ta
ine
d.
No
t
on
ly
does
it
pro
vid
e
s
uitabl
e
functi
ons
t
ha
t
can
be
use
d
as
buil
ding
bl
ock
s
f
or
c
omm
on
NLP
ta
s
ks.
T
his
gr
oup
of
app
li
cat
io
ns
an
d
li
br
aries
f
rom
the
Un
iver
sit
y
of
Pennsyl
va
nia
has
ear
ne
d
co
ns
ide
ra
ble
tract
ion
in
Pyt
hon
-
base
d
S
A
syst
e
m
s since its con
cept
io
n
i
n 2
001.
b.
TextBl
ob
:
it
is
a
python
li
br
a
r
y
fo
r
processin
g
te
xtu
al
data
;
it
pr
ovides
a
si
m
ple
AP
I
to
acce
ss
it
s
m
et
ho
ds
and
do
basic
NLP
ta
s
ks
su
c
h
as
s
entim
ent
a
naly
sis,
par
t
-
of
-
spe
ech
ta
gg
ing
,
c
la
ssific
at
ion
,
t
ra
ns
la
ti
on.
.
.
The
se
ntim
ent
functi
on
of
T
e
xt
B
lob
retu
rn
s
tw
o
pr
operti
es,
s
ubj
ect
ivit
y,
and
po
l
arit
y.
Su
bject
iv
e
sentences
us
ua
ll
y
ref
er
to
per
sonal
opinio
n,
j
ud
gm
ent,
or
e
m
otion
,
w
he
reas
obj
ect
ive
ref
e
rs
to
factu
al
inf
or
m
at
ion
.
Su
bject
ivit
y
is
al
so
a
float
w
hich
li
es
in
the
range
of
[0,
1].
Po
la
rit
y
is
a
float
that
li
e
s
in
the r
a
nge
of
[
-
1,1] w
her
e
1 m
eans a
posit
ive
sta
tem
ent
and
-
1
m
eans a nega
ti
ve
sta
tem
ent.
c.
Gen
sim
:
i
t
is
an
ope
n
-
s
our
ce
li
br
ary
f
or
un
s
up
e
r
vised
su
bject
m
od
el
ing
a
nd
N
LP
,
us
ing
m
od
e
r
n
sta
ti
sti
cal
m
ac
hin
e
le
a
rn
i
ng
.
It
is
co
ns
i
dered
as
a
r
obus
t
vecto
r
s
pace
m
od
el
ing
to
ol
i
m
ple
m
ented
i
n
Pyt
hon.
Co
ntr
ary
to
NL
TK
,
Ge
ns
im
is
the
best
way
t
o
proce
ss
m
assive
dataset
s.
G
ensim
li
br
ary
was
pr
im
aril
y
bu
il
t
fo
r
doc
um
e
nt
si
m
il
ari
ty
est
i
m
ation
,
an
d
this
treatm
ent
is
the
m
os
t
devel
oped
in
the
pa
cka
ge.
It
sup
ports
th
ree
m
ai
n
NLP
m
od
er
n
ta
s
ks
:
retr
ie
ve
sem
antic
al
ly
si
m
il
ar
do
c
um
ents,
scal
ab
le
sta
ti
sti
cal
se
m
antic
s,
an
d
an
al
yz
e
plain
-
te
xt
do
c
u
m
ents
for
sem
antic
structu
re
[
18
]
.
Gen
sim
include
s
stream
ed
par
al
le
li
zed
i
m
ple
m
entat
ion
s
of
m
any
al
gorithm
s
li
ke
fa
stTe
xt
,
word2
vec
,
an
d
doc2
vec
that
are
us
e
d
a
lot
in
t
he
fiel
d
of
A
r
abic
sentim
ent
analy
sis
.
Its
hig
hly
and
native
optim
iz
ed
i
m
ple
m
entat
ion
of
Goo
gle'
s
wo
r
d2vec
m
achine
le
arn
in
g
m
od
el
s
m
akes
it
a
s
t
ron
g
co
nten
der
fo
r
incl
us
io
n
in
a
SA
pro
j
ec
t,
ei
ther
as a
c
or
e
f
ram
ework o
r a
s a libra
ry r
e
s
ource.
In
Ta
ble
1
,
w
e
try
to
hig
hligh
t
m
any
adv
antages
a
nd
disad
van
ta
ges
of
the
m
os
t
us
ed
Pyt
hon
li
br
a
ries
in
A
ra
bic
se
ntim
e
nt an
al
ysi
s
.
2.2.
Ar
ab
ic
se
nt
im
en
t an
alysi
s using
Ja
va
Be
cause
of
it
s
best
featur
e
s,
Jav
a
is
a
powe
rful
progr
a
m
m
ing
la
ng
ua
ge
f
or
pe
rfo
r
m
ing
NL
P.
The
Java
a
pp
li
cat
ion
,
li
ke
j
ust
in
tim
e,
pr
oc
esses
a
la
rg
e
qu
antit
y
of
data
as
rap
id
ly
as
po
s
sible.
T
he
m
ul
ti
-
threa
ding
cha
r
act
erist
ic
of
Java
is
ver
y
sig
nificant
f
or
t
he
heav
il
y
load
ed
ap
plica
ti
on
.
This
ap
plica
tio
n
i
s
us
ef
ul in
NLP i
n
that t
he
ta
s
k
i
s d
i
vid
e
d
int
o
s
ever
al
t
hr
ea
ds
,
thu
s
r
e
du
ci
ng the tim
e.
NLP
st
or
e
d
a
wide
var
ie
ty
of
li
nguisti
c
file
s.
Java
has
a
n
excell
ent
abili
ty
to
store
data
without
any
changin
g
a
sin
gle
cod
e
.
The
Java
data
base
connecti
vity
AP
I
ser
ves
as
a
br
i
dg
e
bet
wee
n
Java
ap
pli
cat
ion
an
d
the
databa
se.
T
he
li
nguisti
c
knowle
dge u
pda
te
d
with
ou
t
c
ha
ng
i
ng
t
he
sin
gl
e
li
ne
of
Ja
va
cod
e
,
an
d
it
sto
red
in
the
da
ta
base.
I
n
[19]
,
t
he
a
uth
ors
stron
gly
r
ecom
m
end
e
d
t
he
us
e
of
Ja
va
pro
gr
am
m
ing
la
ngua
ge
.
Be
s
ides,
the au
t
hors o
f
[
20
]
fou
nd that
Java is t
he
be
st
and
the
m
os
t u
sef
u
l
pro
gr
am
m
ing
lan
guage
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
75
3
-
76
2
756
The
f
ollow
i
ng
sect
ion
p
r
ese
nts
sever
al
be
nefi
t
s
of
Java:
it
i
s
s
i
m
ple,
s
ecure
,
i
nterprete
d,
d
ist
rib
uted
,
o
bject
-
or
ie
nted
,
p
la
tfo
rm
-
inde
pende
nt
,
an
d
m
ulti
-
thread
e
d.
Ac
co
rd
i
ng
to
s
un
m
ic
ro
syst
em
s,
Java
has
the
fo
ll
owin
g
essenti
al
stren
gth
s
:
sec
ur
it
y
,
portabil
it
y,
ease
of
use
,
r
obust
ness
,
a
nd
distr
ibu
te
d
process
acro
s
s
the
W
eb
.
The
r
e
is,
howev
e
r,
the
scop
e
for
Java
i
m
pr
ove
m
ent
as
it
cont
inu
es
to
ha
ve
so
m
e
disadv
a
ntages:
Java
can
be
s
een
as
sig
nificantl
y
slow
er
a
n
d
m
or
e
m
e
mo
ry
-
i
ntensi
ve
than
nativel
y
com
piled
la
ng
ua
ges
,
the
sin
gle
par
a
dig
m
la
ng
ua
ge
,
Lo
ok
an
d
fee
l.
The
def
a
ult
feel
an
d
lo
ok
of
GUI
a
pp
li
c
at
ion
s
wr
it
te
n
in
Jav
a
us
in
g
t
he
S
wing to
ol a
re
ver
y
diff
e
re
nt fro
m
n
at
ive a
pp
li
cat
ion
s
.
2
.
2.1.
Java li
b
raries
In th
is
sect
io
n
,
we wil
l
show
the m
os
t
powe
r
fu
l
Ja
va
li
brar
y
f
or
A
S
A
:
Weka
,
Co
reNLP
,
a
nd G
at
e.
a.
The
Stan
ford
Core
NLP
:
it
offe
rs
a
set
of
hu
m
an
la
ng
ua
ge
te
c
hnology
too
ls
.
It
is
a
Java
an
nota
ti
on
pip
el
ine
fr
am
ework
that
pro
vid
es
la
ngua
ge
processi
ng
ta
sk
s
a
nd
offe
rs
m
os
t
of
the
c
omm
on
essent
ia
l
NLP
ste
ps,
f
rom
tok
enizat
ion
throu
gh
to
c
o
-
re
fer
e
nce
res
ol
ution
[
21]
.
Stanf
ord
Co
reNLP'
s
pu
r
p
ose
is
to
m
ake
it
si
m
ple
to
ap
ply
a
bunch
of
li
ng
uisti
c
analy
sis
tool
s
to
a
te
xt.
This
li
br
a
ry
is
buil
t
to
be
hi
ghly
flexible
an
d
e
xtensi
ble.
T
he
m
os
t
su
pp
or
t
ed
la
ngua
ge
i
s
the
En
glish
la
nguag
e
,
but
oth
e
r
la
ng
uages,
li
ke
Ar
a
bic,
G
erm
an,
Chines
e,
Sp
a
nish,
an
d
Fr
e
nc
h,
ar
e
al
so
avail
abl
e.
Its
featu
res,
relat
ive
ease
of
i
m
ple
m
entat
io
n,
de
dicat
ed
S
A
to
ols,
an
d
e
xc
el
le
nt
com
m
u
nity
support
m
ake
C
or
e
NLP
a
seve
re
c
on
te
nd
e
r
to
pro
duct
ion,
even
i
f
it
s
Jav
a
-
base
d
a
rch
it
e
ct
ur
e
c
ou
l
d
ent
ai
l
a
li
t
tl
e
extra
eng
i
neer
i
ng
an
d
over
hea
d,
in
certai
n
ci
rc
umst
ances.
T
he
S
ta
nford
NLP
l
ibrar
y
ca
n
be
us
e
d
us
i
ng
Py
thon
becau
se
there
a
re
se
veral
pack
a
ges
and i
nterf
ace
s fo
r u
sing Sta
nfo
r
d C
or
e
NLP i
n P
y
tho
n
(i
nd
e
pe
ndent
of
NLT
K).
b.
Wek
a:
it
is
op
e
n
-
s
ource
s
of
tw
are
avail
able
unde
r
the
G
NU
gen
e
ral
public
li
cense
.
It
is
an
acce
ssible
su
it
e
of
m
achine
le
arn
i
ng
s
of
twa
r
e
w
ritt
en
in
Ja
va,
de
velo
pe
d
at
the
Un
i
ver
s
it
y
of
Waikat
o,
Ne
w
Zeal
an
d.
The
We
ka
wor
kb
e
nc
h
in
cl
ude
s
a
group
of
al
gorithm
s
and
vi
su
al
iz
at
ion
to
ols
f
or
pr
e
dicti
ve
m
od
el
in
g
a
nd
data
an
al
ysi
s,
with
gr
a
phic
al
us
e
r
inte
rf
ace
s
f
or
easy
acc
ess
to
t
his
f
un
ct
ion
al
it
y.
W
E
KA
was
use
d
to
perform
sentim
ent
cl
assifi
cat
ion
to
s
olv
e
pro
blem
s
in
var
i
ou
s
fiel
ds.
It
ha
s
bee
n
us
e
d
f
or
S
A
purpose
s
by
a large
v
a
riet
y of resea
rch
e
s a
nd p
a
pe
rs.
c.
G
at
e
:
it
is
a
n
open
-
s
ource
and
Cr
os
s
-
pl
at
fo
rm
Java
s
of
t
war
e
to
olk
i
t
capab
le
of
r
esolvi
ng
al
l
te
xt
processi
ng
prob
le
m
s.
It
co
ntains
a
near
l
y
-
ne
w
inf
orm
at
ion
ext
racti
on
syst
em
"A
NNIE,"
w
hich
is
a
gro
up
of
m
od
ules
co
nta
ining
a
nam
e
d
e
ntit
ie
s
tran
sd
uc
er,
a
par
t
-
of
-
s
peec
h
ta
gger
,
a
gazett
eer
,
a tok
e
nizer,
a c
o
-
ref
e
ren
ce ta
gger
, and
a sent
ence sp
li
tt
er.
T
his librar
y s
up
ports v
a
rio
us
l
angua
g
es:
Ara
bic,
En
glish,
Fr
e
nc
h,
Ge
rm
an,
Chinese
,
Ital
ia
n,
Bulga
rian
,
R
om
anian,
Hind
i,
Ce
buan
o,
R
om
anian,
Dan
i
sh
,
and
R
us
sia
n.
T
her
e
a
re
so
m
e
valua
ble
Gate plugins
that
a
r
e
ver
y
us
e
f
ul
in
A
rab
ic
se
ntim
ent
analy
sis
,
su
c
h
as
SE
AS
a
nd S
AGA.
Table
1.
C
om
par
iso
n
of the m
os
t
us
ed
Pyt
ho
n
li
braries
i
n A
S
A
Librar
y
Ad
v
an
tag
es
D
isad
v
an
tag
es
NLT
K
-
Su
p
p
o
rt
th
e
m
o
s
t
sig
n
if
ican
t
n
u
m
b
er
o
f
lan
g
u
ag
es
co
m
p
a
red to
oth
er
lib
rar
ies.
-
The
M
o
st
W
ell
-
Kn
o
wn
and
f
u
ll NL
P library.
-
Many
th
i
rd
-
p
art
y
e
x
ten
sio
n
s.
-
Fast sen
ten
ce tok
en
izatio
n
.
-
Plen
ty
of
app
roach
es to
eac
h
N
LP
tas
k
-
Qu
ite slo
w.
-
It
is c
o
m
p
li
cated t
o
lear
n
and
us
e.
-
Proces
ses
strin
g
s
wh
ich
are
n
o
t
v
e
ry
ty
p
ical
f
o
r
o
b
ject
-
o
riented
lang
u
ag
e Py
th
o
n
.
-
In
sen
ten
ce
to
k
e
n
izatio
n
,
NLT
K
o
n
ly
sp
lits
tex
t
b
y
sen
ten
ces, witho
u
t analyzing
the se
m
a
n
tic structu
re.
Gen
si
m
-
Prov
id
es
tf
-
id
f
v
ec
to
rization
,
wo
rd2
v
ec,
d
o
cu
m
en
t2
v
ec,
laten
t se
m
an
tic
ana
ly
sis
,
lat
en
t Dir
i
ch
let allocatio
n
.
-
W
o
rks
with la
rge
d
atasets
and
pro
cess
es d
ata strea
m
s.
-
Su
p
p
o
rts deep
lear
n
in
g
.
-
Do
es
no
t
h
av
e
en
o
u
g
h
to
o
ls
to
p
ro
v
id
e
full
NLP
p
ip
elin
e,
so
s
h
o
u
l
d
b
e
u
sed
with
so
m
e
o
th
er
lib
rar
y
(Spacy
or N
LT
K
)
-
Desig
n
ed
pri
m
aril
y
f
o
r
u
n
su
p
ervis
ed
text
m
o
d
elin
g
TextB
lo
b
-
Of
f
ers
lan
g
u
ag
e
trans
latio
n
an
d
d
etectio
n
wh
ich
is
p
o
were
d
by
Go
o
g
le T
r
an
slate
-
Si
m
p
le
to
ap
p
ly
an
d
intu
itiv
e interf
ace
to
NL
TK
lib
r
ar
y
-
S
lo
w
-
N
o
integ
rated wor
d
vecto
rs
-
N
o
neu
ral
n
etwo
rk
m
o
d
els
3.
COMP
ARAT
IVE ST
UDY OF
AS
A
LIB
RARIES
In
our
in
-
de
pth
com
par
at
ive
stud
y,
we
try
to
c
ho
os
e
the
m
os
t
valua
ble
gr
oup
of
li
br
a
rie
s
that
m
eet
s
our nee
ds
rely
ing
on a
v
a
riet
y o
f valua
ble as
pects
a
nd level
s:
3.1.
C
ompar
ati
ve
s
tu
d
y
of
t
he m
os
t po
ten
t
A
SA
li
brarie
s
based
on the
li
tera
t
ure
In
Tabl
e
2
,
we
try
to
highli
ght
num
ero
us
c
har
act
erist
ic
s
of
A
ra
bic
s
enti
m
ent
a
naly
sis
li
br
aries
a
nd
fam
ou
s
w
orks
base
d
on
t
he
li
te
ratur
e
.
Acc
or
ding
to
t
he
li
te
ratur
e
,
we
c
oncl
ud
e
d
that
NL
TK,
We
ka,
Ge
ns
im
,
T
extBl
ob,
a
nd
Stanf
ord
C
or
e
NLP
li
braries
are
be
ne
fici
a
l
com
par
ed
to
ot
her
fam
ou
s
Li
br
a
ries
in
t
he
f
ie
ld
of
ASA
an
d
we
f
ound
m
any
Ar
ti
cl
es
wh
ic
h
ad
op
te
d
the
us
e
of
NLT
K
,
Wek
a
,
TextBl
ob
a
nd
Gen
sim
l
ibrar
i
es
in
their
works m
or
e tha
n St
anfo
r
d
C
or
e
NLP
an
d Gate
li
br
a
r
ie
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Diff
erent val
uable to
ols
for
ar
ab
ic
sentime
nt
analysis:
a co
mpar
ative
eval
ua
ti
on
(
Y
oussr
a
Z
ahidi
)
757
Table
2
.
C
om
par
iso
n of t
he
m
os
t
po
te
nt
A
S
A
li
braries
Librar
y
Licens
e
P
latf
o
r
m
Hig
h
lig
h
ts
W
o
rks
in AS
A
NLT
K
Ap
ache 2
.0
Cro
ss
-
p
latf
o
r
m
Massiv
e
n
u
m
b
ers
o
f
lang
u
a
g
es an
d
too
ls
su
p
p
o
rted; well
-
d
ev
elo
p
ed
co
m
m
u
n
it
y
an
d
d
o
cu
m
en
tatio
n
[
2
2
]
[23
]
[24
]
[25
]
[
2
6
]
[
2
7
]
[28
]
[29
]
[30
]
[
3
1
]
[
3
2
]
Gen
si
m
LGPL
W
in
d
o
ws
,
Linu
x
,
Mac
OS X
an
d
sh
o
u
ld
wo
rk
s
o
n
any
o
th
er
p
latf
o
r
m
th
at su
p
p
o
rts Py
th
o
n
2
.6+ an
d
Nu
m
Py
S
p
eedy
, s
cal
ab
le
,
s
tron
g
nativ
e ca
p
ab
ilities;
co
m
m
e
rcial
sp
in
o
ff
s av
ailab
le
[
3
3
]
[34
]
[35
]
[36
]
[
3
7
]
[
2
3
]
[
3
2
]
TextB
lo
b
MI
T
Licens
e
Cro
ss
-
p
latf
o
r
m
This
librar
y
stan
d
s o
n
the g
ian
t sh
o
u
l
d
ers
of
NLT
K
an
d
p
atte
rn
an
d
plays nicel
y
with
b
o
th
.
th
erefo
re,
m
a
k
in
g
it
easy
f
o
r
b
e
g
in
n
ers by
p
rov
id
in
g
an in
tu
it
iv
e interf
ace
to
N
L
TK
[
3
8
]
[39
]
[40
]
[41
]
[
4
2
]
[
4
3
]
[
4
4
]
Stan
f
o
rd
Co
reNL
P
GNU G
PL
Cro
ss
-
p
latf
o
r
m
P
latf
o
r
m
-
ag
n
o
stic;
m
u
lti
-
lan
g
u
ag
e su
p
p
o
rt;
a live d
e
m
o
avail
a
b
le
[
2
5
]
[
4
5
]
[
3
5
]
[
4
6
]
W
ek
a
GNU
GPL
IA
-
32
,
x86
-
64
;
Jav
a
SE
Po
rtability,
sin
ce it
is
co
m
p
letel
y
i
m
p
le
m
en
ted
in
Jav
a
an
d
thu
s
wo
rk
s
o
n
al
m
o
st
an
y
n
ew
co
m
p
u
tin
g
platf
o
rm
.
A
la
rg
e
g
rou
p
o
f
m
o
d
elin
g
tech
n
iq
u
es
an
d
d
at
a prepro
cess
in
g
.
Si
m
p
le
o
f
us
e du
e to its g
raph
ical us
er
in
terfaces
.
[
4
7
]
[48
]
[49
]
[50
]
[
5
1
]
[
5
2
]
[53
]
[54
]
[55
]
[
5
6
]
[
5
7
]
Gate
LGPL
Cro
ss
-
p
latf
o
r
m
G
ate
h
as p
lu
g
in
s for
m
achi
n
e l
earnin
g
with
W
ek
a,
M
A
XEN
T
,
SVM
lig
h
t
,
R
ASP,
and
f
ast
LibSVM
in
teg
ration
,
a percept
i
on
i
m
p
l
e
m
en
tatio
n
f
o
r
m
an
ag
in
g
on
to
lo
g
ies lik
e
W
o
rdNet,
plu
g
in
s
f
o
r
q
u
ery
in
g
searc
h
eng
in
es
lik
e Yaho
o
o
r
Go
o
g
le,
an
d
plu
g
in
s fo
r
Po
S
tag
g
in
g
with Brill or T
re
eTagg
er.
[
5
8
]
[
5
9
]
[
6
0
]
3.2.
C
ompar
ati
ve
s
tu
d
y
of
open s
oftware
li
braries b
as
ed
o
n
t
he
c
omm
unity on
GitH
ub
r
esul
ts
GitHub
pr
ov
i
de
s
plans
f
or
both
fr
ee
acc
ount
s
and
pr
ivate
r
eposi
tories,
whic
h
are
com
m
o
n
ly
app
li
e
d
to
host
ope
n
-
s
ource
project
s.
It
is
the
bi
gg
e
st
ho
st
of
sour
ce
code
in
the
glob.
T
he
nu
m
ber
s
i
n
the
GitHub
sit
e
are
per
m
anen
tl
y
var
ia
ble.
T
ha
t
is
wh
y
we
will
design
at
e
the
visit
at
ion
date
of
th
e
se
pieces
of
i
nform
at
ion
(10/0
6/2
020).
Table
3
sho
w
s
the
GitHub
r
e
su
lt
s.
T
hro
ugh
the
resu
lt
s,
we
deduce
that
Ge
ns
im
and
NLT
K
are
the m
os
t app
li
e
d,
purs
ue
d by
Core
NLP, Te
xt
Bl
ob
,
We
ka,
a
nd last
ly
,
G
AT
E
.
Table
3
.
GitH
ub
r
esults
Librar
y
NLT
K
Gen
si
m
TextB
lo
b
Co
reNL
P
W
ek
a
Gate
Lang
u
ag
e
Py
th
o
n
Py
th
o
n
Py
th
o
n
Jav
a
Jav
a
Jav
a
Stars
8
,97
5
1
0
,86
7
7
,08
4
7
,
254
302
105
Fo
rks
2
,34
6
3
,80
0
942
2,
400
240
139
Co
n
tribu
to
rs
291
336
22
10
0
1
44
Co
m
m
its
1
3
,88
8
3
,92
8
537
1
6
,
02
2
9
,61
2
3
.40
3
3.3.
C
ompar
ati
ve
s
tu
d
y
of
t
he open s
oftw
are librarie
s
b
as
ed
on m
ulti
ple cri
teria
In
Ta
ble
4
(se
e
app
e
nd
i
x)
,
w
e
try
to
sh
o
w
var
i
ou
s
c
rite
ria
of
NL
P
To
o
ls
li
ke
the
Do
cu
m
entat
ion
,
Characte
risti
cs
,
al
so
,
the
s
upported
treat
m
e
nts
of
e
ach
N
LP
li
br
a
r
y
,
i.e
.
,
N
LTK
,
Gen
i
sm
,
TextBl
ob
Pyt
hon
Libra
ries
and
Core
NLP,
W
e
ka
,
G
ATE
Ja
va
Librar
ie
s
.
W
e
will
m
ake
a
com
par
ison
be
tween
these
li
braries
to
reach
a c
on
cl
usi
on on t
he
m
os
t p
o
te
nt
li
brari
es that
m
eet
s
ou
r
n
ee
ds
v
e
ry
well
.
4.
RESU
LT
S
AND DI
SCUS
S
ION
In
this
c
om
par
at
ive
stud
y
,
we
try
to
adopt
two
ma
jor
m
at
t
ers
.
T
he
first
is
that
nu
m
ero
us
researc
her
s
are
co
nfo
unde
d
ab
out
wh
at
pro
gr
am
m
ing
la
nguag
e
t
hey
hav
e
to
a
pply
fo
r
var
i
ou
s
AN
L
P
m
od
er
n
ta
sk
s
,
especial
ly
fo
r
the
Ar
a
bic
sent
i
m
ent
analy
sis
fiel
d
.
The
sec
ond
issue
is
th
at
there
are
a
l
arg
e
var
ie
ty
of
NLP
li
br
aries,
w
hic
h
is
wh
y
m
any
researc
her
s
f
ind
it
ver
y
hard
to
sel
ect
a
s
uitable
set
of
l
ibrar
ie
s
i
n
thei
r
A
SA
researc
h
pro
j
e
ct
s
an
d
wh
ic
h
on
e
s
m
eet
their
needs
best.
F
or
this
reas
on,
they
us
e
A
N
LP
li
braries
for
thei
r
AS
A
re
searc
h pro
j
ect
s, bu
t
w
it
ho
ut
justi
fyi
ng
thei
r op
ti
on.
Both m
at
te
rs
are
deb
at
e
d
in
m
or
e d
et
ai
l below
:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
75
3
-
76
2
758
4.1. Sele
ctin
g th
e s
uitabl
e p
rog
r
ammin
g
l
angu
ag
e
Am
on
g
v
a
rio
us
program
m
ing
la
ngua
ges
(
su
ch
as
C+
+
,
R,
Perl
,
Prolo
g,
Lisp..
.
),
w
e
sel
ect
ed
Java
and
Pyt
hon
pr
ogram
m
ing
la
nguag
e
s
in
our
st
ud
y.
T
his
ch
oi
ce
is
j
us
ti
fied
by
their
broad
popula
rity
us
ef
ul
ness
an
d
im
po
rtanc
e
fo
r
c
urren
t
A
NLP
ta
sk
s
,
esp
eci
al
ly
fo
r
the
Ar
a
bic
Sentim
ent
An
al
ysi
s
dom
ai
n
.
Be
sides
,
these
two p
rogr
am
m
ing
la
ngua
ges have
a lar
ge va
riet
y of
powe
rful li
br
a
ries in
the
ASA
fiel
d
.
4.2.
C
hoosin
g an
adequ
at
e
l
ibrary
f
or Ar
ab
ic
se
nt
im
en
t ana
l
ys
is p
ro
j
ect
Tha
nk
s
to
t
he
div
e
rsity
of
av
ai
la
ble
NLP
li
br
a
ries,
m
os
t
r
esearche
rs
us
e
va
rio
us
li
brar
ie
s
in
thei
r
researc
h
proje
ct
s
without
e
xp
la
ini
ng
the
cause
beh
i
nd
this
us
e
.
We
aim
ed
to
rely
o
n
our
rev
ie
w
of
the
li
te
ratur
e
on
t
he
m
os
t
po
te
nt
a
nd
use
f
ul
A
S
A
li
br
a
ries
,
na
m
ely:
NLTK
,
Ge
nism
,
T
extBl
ob,
C
oreNLP
,
Wek
a
,
a
nd
Gate
.
T
he
c
ho
ic
e
of
the
li
b
rar
y
r
el
ie
s
on
the
spe
ci
fic
pro
blem
you
are
de
al
in
g
with
i
n
Se
ntim
ent
An
al
ysi
s.
W
e
can
us
e
each
of
them
in
var
io
us
scena
rios.
we
trie
d
to
give
you
a
gen
era
l
su
m
m
arize
of
the
m
,
and w
e
ho
pe
it
can
help
y
ou m
ake th
e
r
i
gh
t
opti
on for y
our
pro
blem
:
a.
NLT
K:
is
v
e
ry
us
ef
ul.
If
y
ou
know
to
pro
gram
in
Pyt
ho
n,
then
NLT
K
is
a
sm
art
cho
ic
e
as
it
con
ta
in
s
the
f
unct
ion
al
i
sts
of
Sta
nford
Core
NL
P
a
nd
W
e
ka
T
ools.
Othe
r
tha
n
t
his
,
yo
u
ca
n
be
ne
fit
from
le
xical
resou
rces
with
ease,
su
c
h
as
Wor
dN
et
,
of
te
n
ind
is
pe
ns
a
ble
in
the
do
m
ain
of
ASA.
S
uc
h
as
Core
NL
P
,
NLT
K
pro
vide
s
var
io
us
w
r
app
e
rs
f
or
m
a
ny
program
m
i
ng
la
ng
uag
es
and
c
om
es
wi
th
a
var
ie
ty
of
resou
rces.
b.
Gen
sim
:
i
ts
hig
hly
an
d
native
opti
m
iz
ed
i
m
ple
m
entat
io
n
of
Goo
gle'
s
wor
d2vec
m
achine
le
ar
ni
ng
m
od
el
s
m
akes
it
a
strong
ca
nd
i
date
f
or
in
cl
us
io
n
in
a
s
entim
ent
analy
sis
project
,
ei
ther
as
a
li
brar
y
resou
rce
or
as
a
cor
e
f
ram
ewo
r
k.
C
on
t
rar
y
t
o
N
LTK
,
Gen
i
sm
is
a
gr
eat
opti
on
f
or
proce
ssing
m
assive
dataset
s. At t
he
sam
e tim
e, it do
es
not
acce
p
t
a sig
nificant
num
ber
of
c
urre
nt
N
LP ta
s
ks
s
uch as
NLT
K.
c.
TextBl
ob
:
it
is
re
l
ie
d
on
N
LT
K
an
d
Patt
er
n.
It
has
a
n
excell
ent
API
f
or
al
l
the
com
m
on
N
LP
treat
m
ents
.
It
is
a
m
or
e
pr
act
ic
al
li
br
ary
fo
c
us
e
d
on
ev
eryday
us
a
ge.
It
is
per
fect
for
init
ia
l
pr
oto
t
yping
in
al
m
os
t
ever
y
NL
P
pro
j
ect
.
U
nfor
t
unat
el
y,
i
t
inh
erit
s
the
low
pe
rfor
m
ance
from
NLT
K,
an
d
th
eref
or
e
it
is
not
su
it
able
f
or
la
rg
e
scal
e
pr
oduction
us
a
ge.
Ma
ny
researc
he
rs
co
ns
ide
re
d
TextBl
ob
Li
brary
as
one
of
Pyt
hon'
s li
br
ari
es to e
xec
ute S
entim
ent Analy
sis.
d.
Stanf
ordC
or
e
N
LP:
it
is
help
f
ul
if
yo
u
need
pa
rt
of
s
peec
h
c
at
egories,
c
o
-
r
efere
nce,
or
na
m
ed
entit
ie
s
in
te
xt.
Thes
e
ha
ve
bee
n
em
ploy
ed
as
pote
nti
al
featur
es
by
the
sentim
ent
analy
sis
resear
ch
com
m
un
it
y.
The
Stan
ford
Core
NLP
is
one
of
th
e
m
os
t
po
te
nt
li
br
a
ries
am
on
g
a
la
r
ge
va
riet
y
of
gr
e
at
NLP
li
br
a
rie
s
because
it
is
easi
ly
co
m
pr
eh
ensible.
C
om
par
ed
t
o
ot
her
l
ibrar
ie
s
,
Co
re
NLP
is
easy
t
o
set
up
a
nd
r
un
since
us
ers
do
no
t
nee
d
t
o
under
sta
nd
c
om
plex
instal
la
ti
ons
an
d
proce
dur
es,
a
nd
it
s
use
r
s
only
re
quire
t
o
hav
e
a lit
tl
e b
a
ckgr
ound
of
pi
eces o
f
in
f
or
m
at
ion
a
bout
Ja
va
b
e
fore they
c
an get st
arte
d.
e.
Wek
a
:
i
t
is
use
fu
l
if
we
al
re
ady
h
o
l
d
data
with
each
data
po
i
nt
h
o
l
ding
a
featur
e
vector,
then
we
ca
n
em
pl
oy
t
h
is
to
o
l
f
or
cl
us
te
ri
ng
our
data.
Hel
pful
if
we
al
s
o
h
o
l
d
t
he
go
l
d
pr
e
dicte
d
outp
uts
f
or
our
data,
we
ca
n bu
il
d cl
assifi
ers.
Sim
pl
e
to
em
pl
oy
G
UI
acce
ssible
a
nd h
i
gh
ly
c
onfi
gura
ble.
f.
G
at
e
:
i
t
is
ad
van
ta
geous
if
we
wa
nt
to
cr
eat
e
a
pip
el
ine.
Dev
el
opers
con
tri
bu
te
la
ngua
ge
analy
s
is
m
od
ules
for
va
rio
us
la
ngua
ge
s
that
are
a
va
il
able
to
be
use
d
pl
ugge
d
int
o
yo
ur
pip
el
in
e.
Help
f
ul
if
you
hav
e
a
ne
w
ap
proac
h,
yo
u
ca
n
w
rite
a
cust
om
iz
ed
m
od
ule
in
JA
VA
a
nd
plug
it
into
the
pip
el
ine
,
an
d
a com
plete
syst
e
m
w
il
l be obt
ai
nab
le
.
As
a
c
on
cl
us
io
n
of
this
par
t,
each
li
br
a
ry
ha
s
it
s
adv
a
ntage
s
to
A
NLP
Tas
ks
,
a
nd
eac
h
one
wa
s
buil
t
to m
eet
the r
es
earche
r
'
s purposes.
Our
in
fe
re
nce
raises tw
o m
ai
n
par
ts
:
a.
The
first
on
e
is
to
do
with
A
NLP
pr
ogram
m
ing
la
nguag
e
s
Pyt
ho
n
a
nd
J
ava,
w
hich
ar
e
ver
y
popula
r
in
the
A
NLP
dom
ai
n
.
Howe
ve
r,
we
reco
m
m
end
P
yt
hon
be
cause
it
is
le
ss
com
pl
ic
at
ed
t
han
Ja
va,
it
ha
s
po
we
rful
and
va
luable
A
N
LP
li
br
aries
c
om
par
ed
to
J
ava
,
an
d
T
hroug
h
our
co
m
par
at
ive
stud
y,
we
c
on
cl
ud
e
that
the
m
os
t
va
luable,
r
obus
t
an
d
us
e
d
ASA
Li
br
a
ries
(
N
LTK,
Ge
ns
im
,
an
d
Te
xtBl
ob
)
belo
ng
to
Pyt
hon
pro
gram
m
i
ng
la
ngua
ge.
F
or
t
his
reas
on,
we
will
adopt
P
yt
hon
in
orde
r
to
acc
om
plish
our
A
S
A
resea
rch p
roject easi
ly
an
d pe
rf
ect
l
y.
b.
The
s
ec
ond
point
r
el
at
es
to
A
NLP
li
brar
ie
s
,
w
hich
a
r
e
al
l
ver
y
use
fu
l.
H
ow
e
ve
r
,
acco
r
ding
to
the
li
te
ratur
e
a
nd
la
rge
va
riet
y
of
powe
rful
and
sig
nifican
t
w
orks
in
the
do
m
ai
n
of
A
S
A
,
we
c
oncl
ud
e
that
the
NLT
K
,
Gensi
m
,
and
TextBl
ob
li
bra
r
ie
s
are
the
m
os
t
us
e
d
f
or
P
yt
ho
n
ASA
ta
sk
beca
us
e
they
hav
e
num
ero
us
ad
van
ta
ges
c
om
par
ed
to
o
t
he
r
A
NL
P
li
br
a
ri
es
.
A
s
f
or
the
Java
A
SA
to
ol
s
,
we
fin
d
t
hat
W
eka
an
d
C
oreNLP
to
ols
are
the m
os
t
us
e
d and
fam
ou
s,
an
d
they
ha
ve gre
at
r
esults i
n
thi
s f
ie
ld
.
5.
C
O
NC
L
US
I
O
N
Be
cause
of
th
ei
r
popula
rity
and
la
r
ge
ab
unda
nce
in
li
braries
for
the
AN
L
P
dom
ai
n,
we
se
le
ct
ed
Java
a
nd
Pyt
hon
pr
ogram
m
i
ng
la
ngua
ges
in
our
c
om
par
a
ti
ve
stu
dy.
In
this
wo
rk
,
we
de
scribe
d
a
va
riet
y
of
AN
L
P
to
ols
w
hich
a
re
co
ns
i
der
e
d
as
the
m
os
t
po
we
rful
an
d
us
ed
.
H
oweve
r,
ther
e
a
re
oth
er
to
ols
in
ot
her
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Diff
erent val
uable to
ols
for
ar
ab
ic
sentime
nt
analysis:
a co
mpar
ative
eval
ua
ti
on
(
Y
oussr
a
Z
ahidi
)
759
pro
gr
am
m
ing
la
nguag
e
s
that
al
so
co
uld
be
ver
y
hel
pful
and
us
e
fu
l
.
Be
sides
,
we
ha
ve
trie
d
to
e
valuate
the
va
rio
us
li
brar
ie
s
usi
ng se
ve
ral aspects
and m
ulti
ple crite
ria.
It
can
be
de
duc
ed
that
each
pr
ogram
m
ing
la
ngua
ge
ha
s
it
s
ben
efit
s
an
d
ad
va
ntages
,
each
li
br
ary
al
so
has
it
s
char
act
erist
ic
s
fo
r
A
NLP
ne
w
ta
sks
,
and
eac
h
one
was
buil
t
t
o
m
ee
t
the
research
e
r
'
s
pu
r
po
s
es
.
Ther
e
f
or
e,
it
is
tou
gh
to
sel
ec
t
the
best
Ar
ab
ic
NLP
li
br
ari
es
becau
se
the
re
is
no
t
only
on
e
si
ng
le
asp
ect
or
crit
erion
t
o do
this.
The se
le
ct
ion
of
t
he
m
os
t
su
it
able li
br
a
ries d
e
pends on
the r
esea
rch
proj
ect
a
nd
w
hic
h
pa
r
t
of
the
ANLP
fi
el
d
is
con
ce
rn
e
d.
F
or
this
rea
s
on,
we
reli
ed
on
our
w
ork
,
w
hich
de
al
s
with
the
do
m
ai
n
of
ASA
in or
der
t
o
sel
e
ct
it
s
m
os
t po
te
nt
an
d use
fu
l l
i
br
a
ries
with ea
se.
APPE
ND
I
X
Table
4
.
N
LP
Libra
ries
'
Com
par
is
on
Variable
Ch
arac
teristics
Su
p
p
o
rted trea
t
m
e
n
ts
NLT
K
(Py
th
o
n
)
-
It
is th
e
"orig
in
"
o
f
all
N
LP
lib
ra
ries.
-
Ver
y
go
o
d
f
o
r
th
e de
-
f
acto
stan
d
ard for
v
ariou
s NLP
t
as
k
s
an
d
edu
catio
n
al pur
p
o
ses
.
-
It
o
f
f
ers an exten
sib
le,
si
m
p
le,
un
if
o
rm
f
ra
m
ewo
rk f
o
r
p
rojects
,
class
de
m
o
n
stratio
n
s, and
assig
n
m
en
ts.
It
is
we
ll
d
o
cu
m
en
ted
,
si
m
p
l
e to learn,
and
eas
y
to ap
p
ly
.
Accessin
g
corp
o
ra,
string
pro
cess
in
g
,
co
llectio
n
d
isco
v
ery
,
to
k
en
izatio
n
,
ste
m
m
in
g
,
P
OS
tag
g
in
g
,
ch
u
n
k
in
g
,
na
m
ed
entities
id
en
tif
icatio
n
,
se
m
an
tic inte
rpreta
tio
n
,
class
if
icatio
n
,
p
rob
ab
ility
esti
m
at
io
n
,
ev
alu
atio
n
m
e
trics,
trans
latio
n
,
d
ep
en
d
en
cy
parsin
g
,
au
to
m
a
tic
su
m
m
a
rization
,
se
n
ti
m
en
t analysis
,
l
an
g
u
ag
e
m
o
d
elin
g
,
t
witter
pro
cess
in
g
,
lo
g
ical se
m
an
tics.
Gen
si
m
(Py
th
o
n
)
-
A
su
b
ject
m
o
d
elin
g
to
o
lk
it i
m
p
le
m
en
ted
in
Py
th
o
n
.
-
It
ap
p
lies
SciP
y
,
op
tio
n
ally
Cytho
n
,
an
d
Nu
m
Py
f
o
r
p
erfo
r
m
an
ce.
-
Scalab
ility
-
Ver
y
ef
f
icien
t i
m
p
l
e
m
en
tatio
n
s
-
Co
n
v
erter
s &
I/O
f
o
r
m
ats
-
Fast:
Ro
b
u
st
-
Si
m
ila
rity q
u
eries
-
Gen
si
m
w
as o
rigin
ally
des
ig
n
ed
f
o
r
esti
m
atin
g
do
cu
m
e
n
t si
m
ilarit
y
,
and
t
h
is
f
eatu
re
is th
e
m
o
st
so
p
h
isticated
of
th
e
p
ackag
e.
-
Evo
lu
tiv
e statistical se
m
an
tics;
-
An
aly
ze
plain
-
tex
t
d
o
cu
m
en
ts f
o
r
se
m
an
tic
stru
ctu
re;
-
Reco
v
er
se
m
an
tic
a
lly
si
m
ilar
do
cu
m
e
n
ts;
TextB
lo
b
(Py
th
o
n
)
Exa
m
p
les o
f
NL
P
TextB
lo
b
Quick
start us
e ca
ses
:
-
Sen
ti
m
en
t Anal
y
si
s
-
Sp
ellin
g
Co
rr
ectio
n
-
Tr
an
slatio
n
and
L
a
n
g
u
ag
e
Detectio
n
Tok
en
izatio
n
,
NE
R, POS
,
class
if
icat
io
n
,
sen
ti
m
en
t
an
aly
sis
,
parsin
g
,
sp
ellch
ec
k
,
lan
g
u
ag
e detectio
n
,
an
d
tr
an
slatio
n
Co
reNL
P
(Jav
a)
-
An
integ
rated NL
P
too
l with a wide
r
an
g
e of
gra
m
m
a
r
an
aly
sis
too
ls;
-
A r
o
b
u
st an
n
o
tato
r
f
o
r
arbitrar
y
t
ex
ts,
widely
app
lied
in
p
rod
u
ctio
n
;
-
A r
eg
u
larl
y
up
d
ate
d
pack
ag
e,
with
th
e hig
h
est q
u
ality
te
x
t
an
aly
tics su
p
p
o
rt
f
o
r
sev
eral
m
ajo
r
(
h
u
m
an
)
lan
g
u
ag
es;
-
Av
ailab
le API
s
f
o
r
m
o
st sig
n
if
ican
t new p
rog
ra
m
m
i
n
g
lan
g
u
ag
es;
-
Ab
ility
to
run
as a
n
easy
web
service
.
Sen
ti
m
en
t anal
y
sis
,
in
f
o
r
m
atio
n
extra
ctio
n
,
n
a
m
ed
entity
r
ecog
n
itio
n
,
p
art
-
of
-
sp
eech
tag
g
in
g
,
co
-
referen
ce r
eso
lu
tio
n
sy
ste
m
,
p
arsin
g
,
b
o
o
tstrap
p
ed
pattern
lear
n
in
g
W
ek
a
(Jav
a)
-
Po
rtable and
si
m
p
l
e to ap
p
ly
.
-
Ad
ap
ted
to
m
ak
e n
ew wa
y
s to
m
achi
n
e lear
n
in
g
des
ig
n
s
-
Latest t
rend
s in
ar
t
if
icial intellig
en
ce
-
Free
o
n
lin
e cou
rse
s av
a
ilab
le
-
Extre
m
e
ly
r
eso
u
rc
ef
u
l bo
o
k
s an
d
pub
licatio
n
s av
ailab
le
-
Hig
h
ly
edu
cated, s
k
illed
and
co
m
m
itt
ed
pro
f
ess
o
rs
Sen
ti
m
en
t Anal
y
si
s d
ata prepro
cess
in
g
,
clu
sterin
g
,
regressi
o
n
,
class
if
icatio
n
,
v
isualization
,
an
d
f
eatu
re
selectio
n
.
Gate
(Jav
a)
-
SEAS
(Gate
plu
g
in
):
is a set
of
pro
cess
in
g
and
linguis
ti
c
reso
u
rces,
w
ritten i
n
Jav
a,
d
ev
elo
p
ed
to run
s
en
ti
m
en
t an
d
e
m
o
tio
n
analysis o
v
er
tex
t us
in
g
th
e
GATE
p
latf
o
r
m
.
Becau
se o
f
the n
atu
re
o
f
GAT
E
,
th
e t
ex
t f
o
r
m
at sh
o
u
ld
be
p
lain
or XM
L.
The
sen
ti
m
en
t anal
y
sis
m
o
d
u
les are
exec
u
ted
in
e
m
b
ed
d
ed
ins
id
e SE
A
S.
-
SAGA
(Senti
m
en
t
an
d
E
m
o
tio
n
Anal
y
sis
integ
rated in
to
GATE
)
is
a set of
p
rocess
in
g
and
lin
g
u
istic reso
u
rces,
written in
Jav
a,
dev
elo
p
ed
to run
sen
t
i
m
en
t and
e
m
o
tio
n
an
aly
sis
ov
er
tex
t
u
sin
g
th
e
GATE
p
l
atf
o
r
m
.
SA
GA is
d
istr
ib
u
ted
as a
G
ATE
plu
g
in
.
Sen
ti
m
en
t anal
y
sis
,
in
f
o
r
m
atio
n
extra
ctio
n
,
p
art
-
of
-
sp
eech tag
g
in
g
,
s
en
ten
ce seg
m
en
ta
tio
n
,
n
a
m
ed
entity
r
ecog
n
itio
n
,
to
k
en
izatio
n
,
co
-
referen
ce
tag
g
in
g
,
REFERE
NCE
S
[1]
N.
S.
Redd
y
,
B.
Praba
devi,
and
B.
Dee
p
a,
"H
ear
t
rate
en
ca
psula
t
ion
and
respons
e
tool
using
sen
t
iment
anal
y
s
is,"
Inte
rnational
Jo
urnal
of El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
(
IJE
CE)
,
vol
.
9
,
no
.
4,
pp.
2585
–
259
2,
Aug.
2019.
[2]
K.
V.
Ghag
and
K.
Shah,
"Conce
ptua
l
senti
m
ent
ana
l
y
sis m
odel
,
"
Inte
rnational
Jo
urnal
of
El
ec
tri
c
al
and
Computer
Engi
ne
ering
(
IJ
ECE
)
,
vol
.
8
,
no
.
4,
pp.
235
8
–
236
6,
Aug.
2018.
[3]
A.
Fargha
l
y
and
K.
Shaal
an,
"A
rab
ic
na
tura
l
l
an
guage
proc
essin
g:
Chal
le
ng
es
and
soluti
ons,"
ACM
Tr
an
s.
Asian
Lang.
Inf. P
roc
e
ss
.
,
vol
.
8
,
no
.
4
,
pp.
1
–
22
,
De
c. 2
009.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
75
3
-
76
2
760
[4]
H.
Abdelna
ss
er
et
al
.
,
"A
l
-
Ba
y
a
n:
An
Arabi
c
Q
uesti
on
Ans
weri
ng
S
y
stem
for
th
e
Hol
y
Quran
,
"
i
n
Proceedi
ngs
o
f
the
EMNLP
201
4
Workshop on Arabic
Na
tural L
anguage
Proces
sing (
ANLP
)
,
pp.
57
–
64
,
2014
.
[5]
W
.
Baka
ri,
P.
Bel
lot,
and
M.
Neji
,
"A
QA
-
W
ebC
orp:
W
eb
-
base
d
Fact
ua
l
Questions
for
Arabi
c,
"
Proce
dia
Comput
.
Sci
.
,
vo
l. 9
6,
pp.
275
–
284,
2016
.
[6]
Y.
Za
hidi,
Y.
El
Younous
si,
and
C.
Azroum
ahl
i,
"Com
par
at
ive
St
ud
y
of
the
Mos
t
Us
efu
l
Arabi
c
-
supporting
Natur
al
La
nguag
e
Proc
e
ss
ing
and
Dee
p
Learni
ng
Li
br
ari
es,
"
in
2019
Inte
rnat
ional
Confe
renc
e
on
Optimizati
on
an
d
Appl
ic
a
ti
ons,
IC
OA
2019
,
2019.
[7]
H.
G.
Hass
an,
H.
M.
Abo
B
akr
,
and
I.
E.
Zieda
n
,
"A
fra
m
ework
for
Arabi
c
conce
pt
-
le
v
el
sen
ti
m
e
nt
anal
y
s
is
using
Senti
cNe
t,
"
In
ter
nati
onal
Journ
al
of
Elec
tri
cal
and
Computer
E
ngine
ering
(
IJE
CE)
,
vol.
8,
no
.
5,
pp.
4015
–
402
2
,
Oct.
2018
.
[8]
A.
Alrum
ai
h,
A
.
Al
-
Sabbagh,
R.
Alsaba
h,
H.
Kha
rrufa
,
and
J.
B
aldw
in,
"S
ent
imen
t
anal
y
s
is
of
co
m
m
ent
s
in
social
m
edi
a,
"
Inte
rnat
ional
Journal
of
El
ectric
al
and
Computer
Engi
n
ee
ring
(
IJE
CE)
,
vol.
10,
no
.
6,
pp.
5917
-
5922,
Dec
.
2020.
[9]
Y.
Al
-
Am
ran
i,
M.
La
z
aa
r
,
and
K.
E.
El
k
adi
r
i,
"
Senti
m
ent
anal
ysis
using
superv
ised
cl
assifi
catio
n
al
gorit
hm
s,"
i
n
ACM
Int
ernati
o
nal
Conf
ere
nce
Proce
ed
ing
S
eri
es
,
vol
.
Par
t
F12
94
,
2017
.
[10]
Y.
Al
-
Am
ran
i,
M.
La
zaar,
K.
E
ddine
,
and
E
.
L.
Kadir
i,
"S
ent
ime
nt
ana
l
y
sis
using
h
y
brid
m
et
hod
of
support
vec
tor
m
ac
hine
and
d
ecision
tr
ee
,
"
J
.
Th
eor.
App
l. Inf.
T
ec
hnol
.
,
vol
.
9
5
,
no.
7
,
pp
.
1886
-
1895
2018.
[11]
Y.
Al
Am
ran
i,
M.
La
z
aa
r
,
and
K.
E.
E
l
Kadir
p
,
"Random
fore
st
and
support
vector
m
ac
hine
b
ase
d
h
y
brid
appr
oa
c
h
to
sent
iment anal
y
sis
,
"
in
Proce
d
ia
Computer
S
cienc
e
,
vo
l. 127, p
p.
511
–
520
,
201
8
.
[12]
Y.
Al
Am
ran
i,
M.
La
z
aa
r
,
and
K.
E.
E
l
Kadir
i
,
"A
novel
h
y
bri
d
cl
assifi
ca
t
ion
appr
oac
h
for
se
nti
m
ent
an
aly
s
is
of
te
xt
do
cume
nt,
"
Int
ernati
on
al
Journal
of
El
e
ct
rica
l
and
Computer
Enginee
ring
(
IJE
CE
)
,
vol.
8,
no
.
6,
pp.
4554
–
4567
,
2018.
[13]
Y.
Al
Am
ran
i,
M.
Laza
a
r,
and
K.
E.
El
Kadir
i
,
"Rec
over
y
of
t
he
opini
ons
thro
ugh
the
spe
ci
f
ic
i
ti
es
of
do
cumen
ts
te
xt
,
"
in
2019
In
te
rnational
Conf
ere
nce
on
Wireless
Technol
ogies
,
Embe
dded
an
d
Inte
l
li
gen
t
Syst
ems,
WITS
2019
,
2019.
[14]
K.
Ewa
n
and
E.
Lope
r,
Natural
Language
Proc
essing
wit
h
Python
,
1st.
ed
.
O
'
Re
il
l
y
M
edi
a
,
In
c.,
2
009.
[15]
S.
Bird
and
E.
L
oper
,
"N
LT
K:
T
he
Natur
a
l
L
ang
uage
Too
lki
t
,
"
i
n
Proce
ed
ings
of
the
ACL
Inte
ra
ct
i
ve
Post
er
and
Demons
trati
on
S
essions
,
2004,
p
p.
214
–
217
.
[16]
N.
Madna
ni
,
"G
et
ti
ng
star
te
d
o
n
nat
ur
al
la
ngu
ag
e
proc
essing
wit
h
P
y
thon
,
"
Cros
sr
oads
,
vol.
13
,
no.
4
,
pp
.
5
–
5
,
2007.
[17]
The
The
ano
De
vel
opm
ent
Te
a
m
et
al
.
,
"
The
a
no:
A
P
y
thon
f
ramework
for
f
ast
computa
ti
on
of
m
at
hemat
i
c
al
expr
essions,"
Co
RR
,
Ma
y
2016
.
[18]
A.
Drozd,
A.
Gladkova
,
and
S.
Matsuoka,
"P
y
th
on,
per
form
ance
,
and
nat
ur
al
l
an
guage
proc
essing,
"
in
Proceedi
n
gs
of
P
yHP
C
2015
:
5th
Workshop
on
Python
for
H
igh
-
Pe
rform
anc
e
and
S
ci
en
ti
f
ic
Computing
-
He
ld
in
conj
un
ct
io
n
wit
h
SC
2015:
T
he
Int
ernati
onal
Confe
renc
e
for
High
Pe
rform
an
ce
Comput
ing,
N
et
working
,
Stora
ge
and
Analysis
,
pp.
1
–
10
,
2015
.
[19]
T.
R
ea
d
,
E
.
Bhr
ce
na
,
and
P.
Fa
ber
,
"J
av
a
and
its
role
in
Natur
al
La
ngu
age
Pro
c
essing
and
Ma
c
hine
Tra
nsla
ti
on
,
"
Proc.
Ma
ch. Tra
nsl.
Summ
it
V1
,
pp.
224
–
231
,
19
97.
[20]
A.
Pinto,
H
.
G.
Olive
ir
a,
and
A.
O.
Alves,
"Com
par
ing
the
p
er
form
anc
e
of
d
iff
ere
nt
NLP
tool
k
it
s
in
form
al
an
d
socia
l
m
edi
a
te
x
t
,
" i
n
OpenA
cc
ess
Serie
s
in
In
formatic
s
,
vol
.
51
,
pp
.
31
–
316
,
2016
.
[21]
C.
D.
M
anni
ng,
M.
Surdea
nu
,
J.
Baue
r,
J.
Fink
el,
S.
J.
B
et
h
ard
,
a
nd
D.
Mc
cl
osk
y
,
"The
St
anf
ord
CoreNL
P
Natur
al
La
nguag
e
Proc
e
ss
ing
Tool
kit,"
in
Proc
ee
dings
of
52nd
Annual
Me
e
ti
ng
of
the
Association
for
Computati
onal
Linguisti
cs:
S
yst
em
Demons
trati
ons
,
pp.
55
–
60
,
2014
.
[22]
A.
A.
Alia
ne
,
H.
Alia
n
e,
M.
Z
iane,
and
N.
Bens
aou,
"A
gene
t
ic
al
gorit
hm
featu
re
select
ion
b
ase
d
appr
oa
ch
for
Arabi
c
Sent
iment
Cla
ss
ifi
c
ation,
"
in
Proceedi
ngs
of
IEEE/A
CS
In
t
ernati
onal
Conf
e
renc
e
on
Compu
te
r
Syste
ms
and
Appl
ic
a
ti
ons,
AI
CCSA
,
2017
.
[23]
G.
Bada
ro
et
al
.
,
"EMA
at
Se
m
Eva
l
-
2018
Tas
k
1:
Emotion
Mining
for
Ara
bic
,
"
in
Procee
dings
of
The
12
th
Inte
rnational
W
orkshop on
Sem
anti
c
Ev
a
luat
ion
,
pp
.
236
–
244
,
2
018
.
[24]
M.
Alfonse
and
E.
-
S.
M
.
E
l
-
Hor
baty
,
"O
pini
on
Mining
for
Arab
ic
Di
al
e
ct
s
on
T
witt
er
,
"
E
g
ypt.
C
omput.
Sc
i.
J
.
,
v
ol.
42,
no
.
4
,
pp
.
52
–
61,
2018
.
[25]
S.
Rosentha
l
,
N.
Farra
,
and
P.
N
akov,
"S
emEva
l
-
2017
Ta
sk
4:
Senti
m
ent
Ana
l
y
s
is
in
Twit
t
er,
"
i
n
Proce
ed
ings
of
the
11
th
Int
ernat
ional
Workshop
on
Semantic Eva
luat
ion
(
SemE
va
l
-
2017)
,
pp.
502
–
518
, 2
017
.
[26]
L.
Al
-
Hor
ai
bi
,
M.
Badruddi
n
K
han,
and L. Al
-
Horai
bi
Muham
ma
d
Badrudd
in
K
han,
"S
en
ti
m
ent
Anal
y
sis
of
Ara
bi
c
Twe
et
s Us
ing
Se
m
ant
ic
R
esourc
e
s,"
Int. J.
Compu
t.
In
f. Sc
i
.
,
vol
.
1
3,
no
.
1
,
2017
.
[27]
N.
El
-
Nagga
r
,
Y.
El
-
Sonbat
y
,
an
d
M.
Abou
El
-
Nasr,
"
Senti
m
ent
ana
l
y
sis
of
m
od
ern
standa
rd
Ara
bic
and
Eg
y
ptia
n
dia
l
ec
t
al Arabi
c
twee
ts,"
in
Proc
ee
dings o
f
Comp
uti
ng
Conf
ere
nc
e
2017
,
vol
.
201
8
-
Janua
,
pp.
880
–
887
,
2018
.
[28]
A.
Mourad
and
K.
Darwish,
"
Subjec
ti
v
ity
and
Senti
m
ent
Anal
y
sis
of
Moder
n
Standa
rd
Arabi
c
and
Arab
i
c
Microbl
ogs,"
in
Proceedi
ngs
of
the
4th
Worksh
op
on
Computat
ional
Approa
ch
es
to
Sub
je
c
ti
v
ity,
S
ent
im
ent
an
d
Soci
al
Me
dia
An
aly
sis
,
pp
.
55
–
64
,
2013
.
[29]
D.
G
amal,
M.
Alfonse,
E.
-
S.
M.
El
-
Horbat
y
,
and
A.
-
B.
M.
Sale
m
,
"Twit
ter
Benc
hm
ark
Data
se
t
for
Arabi
c
Senti
m
ent
Ana
l
ysis,"
Mod.
Educ.
Comput.
S
ci.
,
v
ol.
1
,
pp
.
33
–
38
,
2019.
[30]
H.
Mulki,
H.
Ha
ddad,
M.
Grid
ach,
and
I
.
Baba
oğ
lu,
"Tw
-
StAR
at
Sem
Eva
l
-
2017
Ta
sk
4:
Senti
m
e
nt
Cla
ss
ifi
c
ation
of
Arabi
c
Twe
ets
,
"
in
Proceedi
n
gs
of
the
11th
I
nte
rnational
Wo
rkshop
on
Semanti
c
Ev
alua
ti
on
(
SemE
val
-
2017)
,
pp.
664
–
669
,
20
17
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Diff
erent val
uable to
ols
for
ar
ab
ic
sentime
nt
analysis:
a co
mpar
ative
eval
ua
ti
on
(
Y
oussr
a
Z
ahidi
)
761
[31]
D.
Sulei
m
an
an
d
A.
Aw
aj
an,
"
Com
par
at
ive
St
ud
y
of
W
ord
E
m
beddi
ngs
Mod
el
s
and
Their
Us
age
in
Arabi
c
L
angua
g
e
Appl
i
ca
t
ions,"
in
ACI
T 2018
-
19th
Int
ernati
onal
Arab
Confe
renc
e
on
I
nformation
Tech
nology
,
2019.
[32]
A.
A.
Altowa
y
a
n
and
L.
Ta
o,
"
W
ord
embeddings
for
Arabi
c
senti
m
ent
anal
y
s
is,
"
Proc.
-
2016
I
EE
E
Int
.
Conf.
Bi
g
Data,
B
ig
Data
2016
,
pp
.
3820
–
3825,
2016
.
[33]
A.
Barhoumi,
Y.
Estè
ve,
C
.
Aloulou,
and
L
.
H.
Bel
guit
h
,
"D
ocument
embeddings
for
Arabi
c
Sentim
ent
Anal
y
sis
,
"
in
Proc
ee
dings
o
f
th
e
F
irst Confe
renc
e
on
Langua
ge
Proc
essing
a
nd
Knowle
dg
e Manageme
nt
,
20
17,
vol
.
1988
.
[34]
A.
Aziz
Altowa
y
an
and
A.
El
n
a
gar
,
"Im
proving
Arabi
c
senti
m
en
t
ana
l
y
sis
with
senti
m
ent
-
spe
ci
fi
c
embeddings,
"
in
Proce
ed
ings
-
2
017
IEEE
In
te
rn
ati
onal
Con
fe
ren
ce
on
B
ig
Data
,
Bi
g
Data
2017
,
vol.
2018
,
pp
.
43
14
–
4320
,
2018
.
[35]
A.
M.
Ala
y
ba
,
V.
Pala
de
,
M.
Engl
and
,
and
R
.
Iqba
l
,
"Im
pr
oving
Senti
m
ent
Anal
y
sis
in
Ara
bic
Us
ing
W
ord
Repre
sent
at
ion
,
"
in
2nd
IEE
E
I
nte
rnational
Wo
rkshop
on
Arabic
and
Deriv
ed
Script
Analysis
and
Re
cognition
,
ASA
R
2018
,
pp.
13
–
18
,
2018
.
[36]
S.
Medha
ffa
r,
F.
Bougare
s,
Y.
Estè
ve
,
and
L.
Hadri
ch
-
B
elguith,
"S
ent
im
e
nt
Anal
y
sis
of
Tuni
sian
Dialec
ts:
Li
nguisti
c
Ress
ourc
es
and
Exp
eri
m
ent
s,"
in
P
roce
edi
ngs
of
t
he
Thir
d
Arabi
c
Natural
Lang
uage
Proce
ss
ing
Workshop
,
pp.
5
5
–
61
,
2017
.
[37]
A.
El
-
Ki
la
n
y
,
A
.
Azz
am,
and
S.
R.
El
-
Be
ltag
y
,
"U
sing
dee
p
n
e
ura
l
n
et
works
fo
r
ext
r
ac
t
i
ng
sen
t
iment
t
arg
et
s
in
ara
bi
c
tw
ee
ts,"
i
n
Studies i
n
Computati
onal
Int
el
l
i
genc
e
,
vo
l. 740,
Springer
Verl
ag, pp. 3
–
15
,
2018
.
[38]
K.
H.
Manguri
,
R.
N.
Ramadh
an,
and
P.
R
.
Moham
m
ed
A
m
in,
"Twit
t
er
S
ent
iment
An
alys
is
on
W
orldwide
COV
ID
-
19
Outbre
aks,
"
Kurdis
ta
n
J. A
pp
l. R
es
.
,
p
p.
54
–
65
,
Ma
y
2
020.
[39]
R.
Yada
v
and
N.
Sharm
a,
"A
n
aly
s
is
of
Br
and
Value
Pred
ic
t
io
n
base
d
on
Soci
al
Medi
a
Da
ta,"
Int.
Re
s
.
J
.
En
g.
Technol
.
,
vo
l. 06
,
no
.
08
,
pp
.
127
4
–
1278,
2019
.
[40]
Z.
Nass
r,
N.
Sae
l,
and
F.
Ben
abb
ou,
"A
compara
t
ive
stud
y
of
sent
iment
ana
l
y
sis a
pproa
che
s,"
in
A
CM
Inte
rnationa
l
Confe
renc
e
Pro
c
ee
ding
Serie
s
,
2
019.
[41]
K.
Dashtipour,
A.
Hus
sain,
Q.
Zhou,
A.
Gelbu
kh,
A.
Y.
A.
Ha
wala
h,
and
E
.
Cambria,
"P
erSe
n
t:
A
fre
ely
a
va
ilabl
e
per
sian
senti
m
e
nt
le
xi
con,
"
in
L
ec
ture
Not
es
in
Computer
Sci
en
ce
(
inc
ludi
ng
su
bseries
Lect
ure
Note
s
in
Arti
f
ici
al
Inte
lligen
ce and L
ec
ture
No
te
s in
Bi
o
inf
orm
ati
cs)
,
vol. 10023 LNA
I,
pp.
310
–
320
,
2016
.
[42]
V.
Agarwal
,
P.
Aher,
and
V.
Sa
want,
"A
uto
m
ated
As
pec
t
Extra
ct
ion
and
As
pe
c
t
Orien
te
d
Sen
tim
ent
Anal
y
sis
o
n
Hotel
Review
Data
sets,
"
in
Pr
oce
ed
ings
-
2018
4th
Inte
rnatio
nal
Confe
renc
e
on
Computing,
Comm
unic
ati
o
n
Control
and
Autom
ati
on,
ICCUB
EA
2018
,
2018
.
[43]
S.
M.
Alza
nin
and
A.
M.
Azm
i,
"Rum
or
d
et
ec
t
ion
in
Arabi
c
twee
ts
using
sem
i
-
supervise
d
and
unsupervise
d
expe
c
ta
t
ion
–
m
axi
m
iz
at
ion
,
"
Kn
owle
dge
-
Based Syst.
,
vol
.
185
,
p
.
104945
,
De
c. 2
019.
[44]
N.
Albadi,
M.
Kurdi,
and
S.
Mishra,
"H
at
efu
l
people
or
h
ateful
bots
?
D
et
e
ct
ion
and
ch
aract
er
iz
a
ti
on
of
b
ots
sprea
ding
re
li
gi
ous
hat
red
in
A
rab
ic
soc
ia
l
m
edi
a,
"
Proc.
ACM
Hum
an
-
Com
pute
r
Inte
rac
t.
,
vol.
3,
no.
CS
C
W
,
pp.
1
–
25
,
Nov.
2
019.
[45]
R.
Baly
,
H.
Haj
j
,
N.
Haba
sh,
K.
B.
Shaban,
and
W
.
El
-
Hajj,
"A
senti
m
ent
treeba
nk
and
m
orphologi
call
y
enr
i
ched
rec
ursive
dee
p
m
odel
s
for
eff
e
ct
iv
e
senti
m
en
t
ana
l
y
sis
in
Arab
ic
,
"
ACM
Tr
ans.
Asian
Low
-
R
e
sour
ce
Lang.
In
f.
Proce
ss
.
,
vol. 16
,
no
.
4
,
p
.
23
(21
page
s)
,
Jul.
201
7
.
[46]
R.
Eskand
er
an
d
O.
Rambow
,
"S
LSA:
A
Senti
m
ent
Le
xi
con
f
or
Standa
rd
Ar
a
bic
,
"
in
Proc
ee
d
ings
of
th
e
201
5
Confe
renc
e
on
E
mpirical
M
et
hod
s in
Natural
Language
Proc
essing
,
pp
.
2545
–
255
0
,
2015
.
[47]
A.
Mahm
oud
a
nd
T.
E
lgha
z
aly,
"U
sing
twi
tt
er
to
m
onit
or
p
ol
it
ical
senti
m
ent
for
Arabi
c
slan
g,
"
in
Studies
i
n
Computati
onal I
nte
lligen
ce
,
vol
.
740,
Springer
V
erl
ag
,
pp
.
53
–
66
,
2018
.
[48]
M.
A.
Sghaie
r
and
M.
Zri
gui,
"S
ent
iment
ana
l
y
sis
for
Arabi
c
e
-
comm
erc
e
website
s,"
in
Pro
ce
ed
ings
-
2016
Inte
rnational
Co
nfe
renc
e
on
Eng
ine
ering
and
MI
S,
ICEMI
S
2016
,
2016.
[49]
N.
F.
Bin
Hath
lian
and
A.
M.
H
afe
zs,
"S
ent
imen
t
-
Subjective
an
aly
s
is
fra
m
ework
for
Arabi
c
soc
ia
l
m
edia
posts,"
in
2016
4
th
Saud
i
Int
ernati
onal
C
onfe
renc
e
on
Inf
orm
ati
on
Techn
ology
(
Bi
g
Data
Anal
ysis)
,
KACS
TIT
2016
,
2016.
[50]
R.
Bouchlaghem
,
A.
El
khe
li
fi
,
a
nd
R.
Fai
z,
"S
e
nti
m
ent
an
aly
sis
in
ar
abi
c
twitte
r
posts
using
supervise
d
m
et
hod
s
with
combined
fea
tur
es,
"
in
Lec
ture
Note
s
in
C
omputer
Sci
en
ce
(
inc
ludi
ng
subs
erie
s
Lect
ure
N
ote
s
in
Artificial
Inte
lligen
ce
and
Lect
ure
No
te
s in
Bi
o
inf
orm
ati
cs)
,
vol. 9624 LNCS
,
pp.
320
–
334
,
2018
.
[51]
S.
Al
-
Saqqa,
N.
Obeid,
and
A.
Aw
aj
an,
"S
ent
i
m
ent
Anal
y
sis
for
Arabi
c
Te
xt
using
Ensembl
e
Le
arn
ing,
"
i
n
Proce
ed
ings
of
IEE
E/ACS
Inte
rn
ati
onal
Confe
r
e
nce
on
Compute
r
Syste
ms
and
A
ppli
cations,
AIC
CSA
,
vol.
2018
-
Novem
ber
,
2019
.
[52]
J.
O.
Atoum
an
d
M.
Noum
an,
"
Senti
m
ent
anal
ysis
of
Arabi
c
Jo
rda
nia
n
di
al
e
ct
t
wee
ts,"
Int.
J.
A
dv.
Comput.
Sc
i
.
Appl
.
,
vo
l. 10, n
o.
2
,
pp
.
256
–
26
2,
2019
.
[53]
B.
Haida
r
,
M.
Chamoun,
and
A.
Serhrouc
hni,
"M
ul
ti
l
ing
ual
c
y
ber
bu
lly
i
ng
det
ection
sy
stem:
Det
ec
t
in
g
c
y
b
erb
ul
l
y
ing
in
Arabi
c
content
,
"
in
2017
1st
C
ybe
r
Se
curit
y
in
Net
working
Co
nfe
renc
e,
CSNet
2017
,
vo
l.
2017
-
Janua
r
y
,
pp.
1
–
8
,
2017
.
[54]
M.
Abdulla
h,
M
.
Hadz
ik
adi
c
y
,
a
nd
S.
Shaikhz
,
"S
EDAT:
Senti
m
ent
and
Emotion
Dete
ction
in
Arabi
c
T
ext
Us
ing
CNN
-
LSTM
Dee
p
Learni
ng
,
"
in
Proce
ed
ings
-
17th
IEEE
Int
ernati
onal
Conf
ere
nce
on
Mac
hin
e
Learning
and
Appl
ic
a
ti
ons,
IC
MLA
2018
,
pp.
8
35
–
840
,
2019
.
[55]
S.
Alhum
oud,
T.
Albuhai
r
i,
a
nd
M.
Altuwai
j
ri,
"A
rab
ic
sent
iment
a
naly
sis
using
W
EKA
a
h
y
brid
learni
n
g
appr
oac
h
,
"
in
I
C3K
2015
-
Pr
oce
ed
ings
of
th
e
7th
In
te
rnatio
nal
Joi
n
t
Conf
e
renc
e
on
Know
l
edge
Disco
ve
ry
,
Knowle
dge
Enginee
ring a
nd
Kno
wle
dge
Manage
ment
,
vo
l. 1, pp.
402
–
408
,
2015
.
[56]
T.
El
gh
azal
y
,
A.
Mahm
oud,
and
H.
A.
Hefn
y
,
"P
oli
ti
c
al
senti
m
ent
an
aly
s
is
using
twit
te
r
dat
a,
"
in
AC
M
Inte
rnational
Co
nfe
renc
e Proce
e
ding
Seri
es
,
vo
l.
22
-
23
-
Marc
h
,
pp
.
1
–
5
,
2016
.
[57]
S.
R.
El
-
B
el
t
ag
y,
T.
Khalil,
A.
Hala
b
y
,
and
M.
Ham
m
ad,
"Combini
ng
le
x
ical
f
e
at
ure
s
and
a
sup
erv
ised
l
ea
rnin
g
appr
oac
h
for
arabic
senti
m
ent
an
aly
s
is,"
in
Lect
u
re
Note
s
in
Computer
Sci
en
ce
(
inc
ludi
ng
subs
eri
es
Lect
ure
Note
s
in
Artifi
ci
al
Int
ellige
n
ce
and
Lec
t
ure
Note
s
in Bi
oi
nformatic
s)
,
vol. 9624
LNCS,
pp.
307
–
319
,
2018
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
75
3
-
76
2
762
[58]
A.
Am
ja
d
and
U.
Qam
ar,
"U
AMSA:
Unifie
d
app
roa
ch
for
m
ult
ili
ngual
senti
m
ent
ana
l
y
sis
using
GA
TE
,
"
in
ACM
Inte
rnational
Co
nfe
renc
e Proce
e
ding
Seri
es
,
pp
.
1
–
5
,
2019
.
[59]
M.
A.
Jerbi,
H.
Achour,
and
E
.
Souis
si,
"S
ent
iment
Anal
y
sis
of
Code
-
Sw
it
che
d
Tuni
sian
Dia
lect
:
Expl
oring
RN
N
-
Based
T
ec
hni
qu
es,
" i
n
Comm
unic
ati
ons
in
Comp
ute
r and
In
formation
S
ci
en
ce
,
v
ol.
1108
,
pp
.
122
–
131
,
2019
.
[60]
F.
Am
iri
,
S.
Sce
rri,
and
M.
H.
Khodashahi
,
"
L
exi
con
-
ba
sed
se
nti
m
ent
ana
l
y
sis
for
Persian
t
ext,"
in
Inte
rnat
ion
al
Confe
renc
e
R
ece
nt
Ad
vanc
es
in
N
atural
Language
Proce
ss
ing, RA
NL
P
,
vol
.
2015
-
Jan.
,
pp.
9
–
16
,
2
015
.
BIOGR
AP
HI
ES OF
A
UTH
ORS
You
ss
ra
Z
ahid
i
is
a
Ph
.
D.
st
udent
in
Com
pute
r
Sci
ence,
In
form
at
ion
S
y
st
e
m
,
and
Softwar
e
Engi
ne
eri
ng
L
a
bora
tor
y
,
Abde
l
m
al
ek
Essaa
di
Univer
sit
y
,
Te
tu
an,
Morocc
o
.
Sh
e
is
a
Com
pute
r
Scie
nc
es
engi
ne
er,
gra
du
ated
in
201
7
from
the
Nati
ona
l
School
of
Applie
d
Science
s,
Abdelmal
e
k
Essaa
di
Univ
ersi
t
y
.
Yacine
El
You
n
ous
si
is
a
Ph
.
D.
doct
or
and
prof
e
ss
or
of
computer
scie
nc
es
at
th
e
Nati
ona
l
School
of
Applie
d
Sc
i
enc
es
of
Te
tu
a
n,
Inform
at
ion
S
y
stem
and
So
ftwa
re
Engi
ne
er
ing
La
bor
at
or
y
,
Abdelmale
k
Ess
aa
di
Univ
ersity
,
Te
tu
an,
Moroc
co
.
He
is
a
supe
rvisor
of
m
an
y
The
s
is
,
and
h
e
i
s
par
t
of
m
an
y
bo
a
rds of
interna
ti
o
nal
journa
ls
and int
ern
at
ion
al
con
fer
ences.
Yas
sin
e
Al
-
Am
rani
is
a
Ph
.
D.
doct
or
and
profe
s
sor
of
Com
pute
r
Scie
nce
s
a
t
the
Multi
disci
p
li
n
a
r
y
Facul
t
y
of
L
ar
ac
he
,
Abdem
al
e
k
Essaa
di
Univ
ersity
,
Moroc
co
.
He
is
a
Com
pute
r
Scie
nc
e
s
E
ngineer
,
and
h
e
is pa
r
t
of
m
an
y
boar
ds of
I
n
te
rn
at
ion
al
J
ourna
ls
and
I
nt
ern
a
ti
ona
l
C
onfe
ren
ce
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.