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.
9
, No
.
5
,
Octo
ber
201
9
, pp.
4441
~
44
45
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
9
i
5
.
pp4441
-
44
45
4441
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Su
gg
esti
n
g new wo
rds to
extract
keywords
from titl
e and ab
stract
Ha
deel
Qase
m Ghe
ni,
Ahm
ed
Moham
med H
ussein,
Wed K
adhi
m
Olei
w
i
Depa
rtment
o
f
C
om
pute
r,
Col
le
g
e
of
Sc
ie
n
ce for W
om
en,
Univer
sit
y
of
B
ab
y
lon
,
Ira
q
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Dec
22
, 201
8
Re
vised
A
pr
2
2
, 2
01
9
Accepte
d
Ma
y
6
, 2
01
9
W
hen
ta
lk
ing
ab
out
the
funda
m
e
nta
ls
of
writi
ng
r
ese
arc
h
pap
ers,
we
find
that
ke
y
words
ar
e
st
il
l
pr
ese
nt
in
m
ost
rese
ar
ch
pap
ers,
but
tha
t
doe
s
not
m
ea
n
tha
t
they
exi
st
in
all
of
the
m
,
we
c
an
find
pape
rs
th
at
do
not
contain
ke
y
words
.
Ke
ywords
are
those
words
or
phra
ses
tha
t
accurately
r
eflec
t
th
e
cont
en
t
of
the
rese
arc
h
pap
er.
Ke
y
words
are
a
n
exa
ct
abbr
evia
ti
on
of
wh
at
the
r
ese
arc
h
c
ar
rie
s
in
it
s
con
tent.
T
he
r
ight
k
e
y
words
m
a
y
i
ncr
ea
se
th
e
cha
nc
e
of
findi
n
g
the
art
i
cl
e
or
r
ese
arc
h
pap
er
an
d
cha
nce
s
of
reachi
ng
m
ore
peopl
e
who
should
re
ac
h
the
m
.
The
import
anc
e
of
ke
y
words
and
the
essenc
e
of
the
rese
arc
h
and
addr
ess
is
m
ai
nl
y
to
attra
ct
the
se
highly
sp
ec
i
al
i
ze
d
and
highly
infl
uen
tia
l
write
rs
in
the
ir
fie
lds
and
who
spec
ia
l
iz
e
in
re
adi
ng
what
holds
the
appr
o
pria
t
e
ch
aract
er
i
stic
s
but
they
d
o
not
re
ad
and
ca
nnot
rea
d
eve
r
y
th
ing.
In
t
his
pape
r
,
we
e
xtra
c
t
new
ke
ywords
b
y
s
ugge
sting
a
set
of
words
,
the
se
words
were
sugges
te
d
accordi
ng
to
the
m
an
y
m
ent
i
oned
in
the
rese
arc
h
es
with
m
ult
ipl
e
d
isci
pl
i
nes
in
th
e
fi
el
d
of
computer
.
In
our
s
y
stem,
we
ta
k
e
a
num
ber
of
words
(as
m
an
y
as
spe
ci
f
ied
in
th
e
progr
am)
tha
t
come
bef
ore
th
e
p
rop
osed
words
and
conside
r
it
as
new
ke
y
words
.
Thi
s
s
y
stem
prove
d
to
be
eff
ec
t
ive
in
find
in
g
ke
y
words
that
cor
respond
to
som
e
ext
ent
with
th
e
k
e
y
wor
ds de
vel
op
ed
b
y
the
aut
hor
in
h
is
rese
arc
h
.
Ke
yw
or
d
s
:
Inform
at
ion
r
et
rieval
Keyw
ord
e
xtra
ct
ion
Natu
ral
la
ngua
ge pr
ocessi
ng
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Had
eel
Qasem
G
he
ni
Dep
a
rtm
ent o
f C
om
pu
te
r,
Ba
bylon U
niv
e
rsity
,
Hill
a, Babil
, Ir
aq
.
Em
a
il
:
had
eel
qa
sem
84
@g
m
ail.co
m
1.
INTROD
U
CTION
Du
e
to
the
gre
at
dev
el
opm
ent
in
an
on
li
ne
publishi
ng
[
1]
and
a
vaila
bili
ty
of
el
ect
ro
nic
book
s
a
nd
web
sit
es
[2
]
th
e
retrieval
of
i
nfor
m
at
ion
can
achieve
d
quic
kly
and
ea
sil
y
us
in
g
sea
rch
e
ng
i
nes
[
3]
w
he
re
the
vast
am
ou
nt
of
inf
or
m
at
ion
is
acce
s
sible
t
o
people
t
hroug
h
the
In
te
rn
et
[
4]
.
This
i
nfor
m
at
ion
is
a
vaila
ble
as
database
,
doc
um
ent,
or
m
ultim
edia
form
at
and
acce
ss
t
o
this
in
form
at
io
n
governe
d
by
the
a
vaila
bili
t
y
of
a
n
appr
opriat
e search
engine [5]. New
ly
, m
any
do
c
um
ents are
el
ect
ro
nical
ly
av
ai
la
ble an
d
yo
u
can easil
y cho
ose
any
do
c
um
ent
you
wa
nt
to
re
ad
or
to
know
the
relat
ion
s
hi
p
betwee
n
the
do
c
um
ents
by
extracti
ng
the
su
it
able
keyw
ords
[
6].
Keyw
ords
ar
e
w
ords
us
e
d
by
the
use
rs
of
sea
rch
en
gin
e
t
o
get
w
hat
they
wan
t
from
inf
or
m
at
ion
an
d
rese
arc
h,
s
o
they
are
te
rm
s
t
hat g
ive
n
to the w
ords
that i
ndic
at
e the co
nt
ent o
f
the sub
je
ct
[
7].
Keyw
ords
have
been
e
ver
y
w
her
e
in
our
daily
li
ves,
fr
om
searchi
ng
for
t
he
inf
orm
ation
we
nee
d
on
th
e
web
via
search
e
ngines
to
on
li
ne
ads
that
m
at
ch
the
con
te
nt
w
e
'
re
cur
ren
tl
y
browsi
ng
[8
]
.
Unfortu
natel
y,
m
any
do
c
um
ents
do
no
t
c
onta
in
ke
ywords
[
9].
Pe
op
le
do
no
t
ha
ve
e
nough
ti
m
e
to
read
the
e
ntire
re
searc
h;
the
be
st
for
them
is
a
br
ie
f
read
i
ng
of
it
l
ike
the
abstract
instea
d
of
the
entire
te
xt
[1
0].
Key
wor
ds
can
be
an
intensiv
e
su
m
m
ary
of
t
he
docum
ent,
dev
el
op
e
d
the
retrieval
of
inf
or
m
at
ion
,
or
beco
m
e
an
entry
for
the
docum
ent
set
[
11]
.
Keyw
ords
are
i
m
po
rtant
an
d
m
eaningfu
l
w
ords
in
the
doc
um
ent,
wh
ic
h
giv
e
an
acc
ur
a
te
ov
er
view
of
their
c
on
te
nt
and
ref
le
ct
th
e
auth
or'
s
intenti
on
to
wr
it
e
[
12]
.
Hen
ce
,
thes
e
wo
r
ds
are
al
l
fo
r
the
w
rite
r
of
the
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.
9
, N
o.
5
,
Oct
ober
201
9
:
4
4
4
1
-
4
4
4
5
4442
arti
cl
e
and
sho
uld
f
oc
us
on
the
sel
ect
ion
an
d
identific
at
io
n
to
increase
th
e
interest
of
vi
sit
or
s
to
rea
d
a
rtic
le
s
and
com
pete
with
oth
e
r
a
rtic
le
s
in
al
l
sea
r
ch
e
ng
i
nes.
I
n
the
e
xtracti
on
of
per
ti
ne
nt
i
nfor
m
at
ion
,
ke
ywords
hav
e
a
ve
ry
sign
i
ficant
r
ole
wh
e
re
the
sum
m
ary
of
the
entire
doc
ume
nt
co
ntent
ca
n
ext
racted
vi
a
a
few
keyw
ords
[
13]
.
It
m
ay
con
sist
of
on
e
or
t
wo
words
that
ha
ve
a
ge
ner
al
m
eanin
g
[
14]
,
th
eref
or
e
cal
le
d
sh
ort
k
ey
w
ords,
or
m
ay
con
sist
of
sh
ort
sentence
s
that
hav
e
an
aver
a
ge
le
ngth
and
be
m
or
e
sp
eci
al
iz
ed,
or
m
ay
be
long se
ntences
.
Keyw
ords
e
xtr
act
ion
is
to
e
xtract
sev
eral
prom
inent
wor
ds
f
r
om
a
par
ti
cular
te
xt
an
d
use
these
words
to
re
pr
e
sent
the
e
ntire
te
xt
[15]
,
it
is
an
a
uto
m
at
ic
pr
oces
s
to
def
i
ne
a
gro
up
of
te
rm
s,
wh
ic
h
re
present
the
inf
or
m
at
ion
in
a
doc
um
e
nt
that
hav
e
be
en
disc
us
se
d
[
16
]
.
E
xtracti
ng
of
ke
ywo
r
ds
is
an
essenti
al
ta
sk
an
d
it
is
an
i
m
po
rtant
researc
h
tr
end
in
te
xt
m
i
ning,
inf
orm
ati
on
ret
rieval,
a
nd
na
tural
la
ng
uag
e
processin
g
[17],
it
le
t
us
to
represent
te
xt
doc
um
ent
in
an
in
te
ns
ive
way
[18].
Keyw
ords
us
ua
ll
y
extracte
d
by
e
xtracti
ng
th
e
releva
nt
highe
r
fr
e
quency
w
ords
f
ro
m
othe
rs,
wit
h
em
p
hasis
on
im
po
rtant
w
ords
[
19]
.
The
al
go
rithm
of
keyw
ord
E
x
tra
ct
ion
has gr
eat
capab
il
it
ie
s in sum
m
ariz
ing
t
he
e
ntire
do
c
um
ent [
20]
.
Extracti
on
of
keyw
ords
is
the
basis
of
inf
orm
ation
retriev
al
pr
ocess
a
nd
nu
m
ero
us
of
te
chn
iq
ue
s
hav
e
been
pro
po
s
ed
to
ad
dre
ss
this
pr
oble
m
[2
1],
[10]
su
ggest
a
form
that
extracts
ke
ywo
rd
s
f
r
om
the
ti
tle
and
a
bs
tract
by
con
str
uctin
g
a
li
st
of
w
ords
arr
a
ng
e
d
in
de
scen
ding
orde
r
de
pe
nd
i
ng
on
the
nu
m
ber
of
t
heir
app
ea
ra
nce
in
abstract
a
nd
ti
t
le
.
[22]
Em
bodim
ents
can
in
cl
ud
e
w
ord
a
na
ly
sis
in
an
in
div
id
ual
do
c
um
ent
by
stop
w
ords,
del
i
m
i
t
ers,
or
bot
h
to
i
den
ti
fy
ca
nd
i
date
key
word
s
.
T
hen,
f
or
each
w
ord
,
w
ord
sc
ores
are
c
ount
in
the
cand
i
date
keyw
ords
base
d
on
the
fr
e
qu
ency
functi
on.
[23]
Test
ed
a
set
of
central
m
et
rics
on
word
s
a
nd
netw
orks
c
ompil
e
the
noun
s
entences
an
d
a
naly
sis
of
the
a
ch
ie
vem
ent
on
fou
r
sta
nda
rd
databases
.
[24]
Buil
d
appr
oach
with
the
scal
e
of
a
ne
w
net
work
-
node
sel
ect
ivit
y,
dr
ive
n
by
the
researc
h
of
a
c
entrali
zed
a
ppr
oach
base
d
on
t
he
gr
a
ph.
N
od
e
sel
ect
ivit
y
is
def
i
ne
as
t
he
distrib
utio
n
of
ave
ra
ge
weig
ht
on
si
ngle
no
d
e
connecti
ons;
they
extract
no
des
(
keyw
ord
cand
i
dates)
ba
sed
on
t
he
value
of
sel
ect
ivi
ty
.
In
a
dd
it
io
n,
they
exp
a
nd
the
e
xtracted
nodes
i
nto
word
gro
ups
with
t
he
hi
gh
e
st
in
/
ou
t
sel
ect
ive
value
s.
[
25
]
S
ugges
ti
ng
a
n
al
gorithm
dep
end
i
ng
on
a
n
en
d
-
to
-
e
nd
neur
a
l
keyword
s
e
xtracti
on
by
us
in
g
the
netw
ork
of
Siam
ese
LSTM,
rem
ov
ing t
he n
eed to en
gine
er
m
anu
al
f
eat
ur
es.
2.
PROP
OSE
D
METHO
D
In g
e
ner
al
,
the
basic idea
is to
prop
os
e a
set
of wor
ds
e
xpec
te
d
to
be
m
ention o
r
al
ways c
om
e in the
ti
tl
e and
a
bs
tra
ct
, s
eve
ral
area
s of c
om
pu
te
r a
pp
li
cat
io
ns
w
ere c
on
si
der an
d
ide
ntify
wor
ds
t
hat are
e
xtr
e
m
el
y
us
e
d
in
r
esea
rc
h,
t
his
word ar
e cla
rified
in
T
able 1.
Table
1.
T
he
s
uggeste
d keyw
ords
Su
g
g
ested
Key
wo
r
d
s
Netwo
rks
p
rocess
in
g
ex
traction
d
etectio
n
ex
p
an
sio
n
m
e
th
o
d
to
o
l
Seg
m
en
tatio
n
way
m
o
d
el
syste
m
step
s
Ap
p
roach
lo
calization
Sch
e
m
e
Techn
iq
u
e
d
iag
ra
m
Flo
wch
art
Fra
m
e
wo
rk
Co
m
p
r
ess
io
n
Cry
p
to
g
raph
y
Proces
so
r
stitch
in
g
tex
ts
Circu
it
steg
an
o
g
raph
y
Secu
rity
Structu
re
Paradig
m
f
eatu
res
Alg
o
rith
m
d
iscri
m
in
atio
n
p
rivacy
ex
p
o
su
re
ch
ain
s
W
ire
less
en
v
iron
m
en
t
Co
m
m
u
n
icatio
n
s
tech
n
o
lo
g
y
ap
p
licatio
n
s
p
atterns
an
aly
sis
The pr
opos
e
d
s
yst
e
m
co
ns
ist
s o
f
se
ve
ral step
s: (1) T
ok
e
niza
ti
on
, (2
)
St
op
word rem
ov
in
g, (
3)
Searc
h
about
the
s
ugge
ste
d
w
ords
in
the
ti
tl
e
and
abstract
an
d
ta
ke
(
N)
wor
ds
that
com
es
bef
or
e
it
,
(
4)
Fin
d
the
si
m
il
arity
between
the
extract
ed
keyw
ords
a
nd
keyw
ords
m
entioned
by
the
auth
or
in
t
he
resea
rch.
Figure
1
il
lustrate
s the
m
echan
ism
o
f work
of t
he pr
opos
e
d
syst
em
.
The
fi
rst
ste
p
is
to
div
i
de
the
par
a
gr
a
phs
of
ti
tl
e
and
abstr
act
into
sin
gle
words
by
the
t
ok
e
nizat
io
n
process
,
then
,
delet
e
the
stop
word
s
i
n
the
s
econd
ste
p
a
nd
conver
t
al
l
w
ords
to
lo
we
r
case
to
m
ini
m
i
ze
the
com
par
ison
ti
m
e.
Af
te
r
t
hat, w
e
will
searc
h
in
the
ti
tl
e
and abstract par
a
graphs
ab
out
t
he
pro
po
se
d
wor
ds,
an
d
if
they
f
ound,
the
wor
ds
that
com
e
bef
ore
th
e
m
will
be
ta
ke
n
in
am
ou
nt
of
(
N)
s
pecified
in
the
pr
ogra
m
.
(N)
Re
pr
ese
nts
the
nu
m
ber
of
w
ords
will
be
tru
ncate
from
the
sentence
t
hat
com
es
bef
or
e
the
wor
ds
that
su
ggest
e
d.
Th
en
analy
zi
ng
t
he
res
ults
and
find
the
sim
i
la
rity
between
the
extracte
d
keywords
an
d
the
keyw
ords
dev
e
lop
e
d
by
t
he
a
uthor
.
The
pro
po
s
ed
syst
em
s
how
n
in
Fi
gur
e
1.
Fi
gure
2
e
xp
la
in
our
al
gorithm
that dem
on
stra
te
the wo
rk of t
he pr
opos
e
d sy
stem
.
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
Suggest
in
g ne
w w
or
ds t
o ex
tract ke
yw
ords
from
ti
tl
e an
d ab
str
act
(
H
ade
el
Q
ase
m
Ghe
ni
)
4443
Figure
1. The
pro
po
se
d
syst
e
m
Figure
2. The
a
lgorit
hm
3.
RESU
LT
S
AND A
N
ALYSIS
Database
buil
t
to
st
or
e
resea
rch
es
that
will
te
ste
d
by
our
syst
em
.
This
database
co
ns
i
sts
of
th
re
e
fiel
ds
, n
am
ely:
the
first
fiel
d
t
o
sto
re
the
ti
tl
e
of
re
searc
h,
th
e
second
fiel
d
to
sto
re
the
abst
ract
of
r
esearc
h
a
nd
the
third
fiel
d
is
to
store
the
ke
ywo
rd
s
wr
it
te
n
by
the
aut
hor
of
t
he
resea
rc
h.
A
bout
100
s
ci
entifi
c
resear
ches
wer
e
st
or
e
d
in
the
databas
e
for
te
sti
ng
;
these
resea
rc
hes
ra
ndom
ly
sel
ect
ed
fro
m
pu
blica
ti
ons.
Wh
e
n
searchi
ng
f
or
a
nd
fin
ding
t
he
words
we
wa
nt
,
we
will
determ
ine
the
nu
m
ber
of
w
ords
t
o
be
tr
un
cat
e
be
fore
the pr
opos
e
d w
ords, an
d
t
his i
s what (
N) e
xpresse
d.
Thr
ee
values
wer
e
te
ste
d
f
or
(N)
an
d
thes
e
values
are
w
hen
(
N
)
is
on
e
,
two
an
d
thre
e.
W
he
n
the
value
of
(
N)
is
on
e,
it
m
eans
tru
ncati
ng
on
e
word
befor
e
th
e
pr
op
os
e
d
word
that
f
ound
in
the
ti
tl
e
or
a
bs
tract
,
and
w
hen
t
he
value
of
(N)
is
two,
it
m
eans
tru
ncati
ng
t
wo
w
ords,
a
nd
therefor
e
w
he
n
the
value
of
(N)
is
three,
it
m
eans
trun
cat
i
ng
t
hree
words.
Wh
e
n
tru
ncates
the
words
by
(
N),
these
w
ords
w
il
l
be
a
sentenc
e
with
the
pro
po
se
d
word
acc
ordin
g
to
the
seq
ue
nce
in
the
pa
ragrap
h
an
d
thu
s
we
will
ha
ve
a
short
se
ntence
represe
nting t
he
n
e
w keyw
ord
.
The
res
ults
showe
d
that
ther
e
is
78
%
m
atch
in
g
betwee
n
the
extracte
d
keyword
s
an
d
keyw
ords
dev
el
op
e
d
by
the
auth
or
in
hi
s
researc
h
wh
e
n
ch
oo
si
ng
the
value
of
(
N)
e
qu
al
to
one,
w
her
e
the
w
ord
ta
ke
n
by (N)
and t
he pr
opos
e
d w
ord
are fo
rm
ed
a s
hort se
ntence
of tw
o wor
ds
.
Wh
il
e
wh
e
n
se
le
ct
ing
the
val
ue
of
(
N)
eq
ua
l
to
tw
o,
t
her
e
is
54%
m
at
ching
betwee
n
th
e
extracte
d
keyw
ords
a
nd
keyw
ords
dev
e
lop
e
d
by
the
a
uthor
,
w
her
e
t
he
w
ords
ta
ke
n
by
(
N)
a
nd
the
pro
pose
d
w
ord
ar
e
form
ed
a
sho
rt
sente
nce
of
th
ree
words.
Fin
al
ly
,
wh
e
n
sel
e
ct
ing
t
he
value
of
(N)
e
qu
al
to
th
ree,
the
re
i
s
23
%
m
at
ching
bet
w
een
the
e
xtract
ed
key
wor
ds
a
nd
keyw
ords
dev
el
op
e
d
by
th
e
aut
hor,
where
the
wor
ds
ta
ken
by
(
N)
a
nd
the
pro
posed
wor
d
a
re
form
ed
a
sente
nce
of
f
our
w
ords
.
Ta
bl
e
2
sho
ws
t
he
m
at
ching
rati
o
of
(
N
)
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.
9
, N
o.
5
,
Oct
ober
201
9
:
4
4
4
1
-
4
4
4
5
4444
values
.
T
he
res
ults
obta
ine
d
i
n
Ta
ble
3
s
how
that
t
her
e
is
a
s
ig
nificant
c
orrelat
ion
bet
w
een
the
ke
ywo
rd
s
that
wer
e
draw
n wi
th the key
w
ord
s d
e
velo
pe
d by the a
uthor
in h
is researc
h.
Table
2.
Mat
c
hi
ng
rati
os
Extracted K
ey
wo
r
d
s W
h
ere
Au
th
o
r
Key
wo
rds
N=1
N=2
N=3
78%
54%
23%
Table
3.
Mat
c
hi
ng
res
ults
Total
Ke
y
wo
rds
Total Mat
ch
es
Partial
Matches
Failu
res
552
53
331
168
Ratio
9
.6 %
5
9
.9 %
3
0
.4 %
4.
CONCL
US
I
O
N
In
ge
ner
al
,
as
sho
wn
in
T
ables
1
-
3,
10
0
researc
h
pa
per
s
c
onta
ini
ng
keyw
ords
wer
e
te
ste
d.
The
res
ults
showe
d
that
the
re
we
re
82
st
ud
ie
s
fou
nd
a
m
at
ch
betwe
en
it
s
keyw
ords
an
d
the
pr
opos
e
d
keyw
ords
a
nd
that
there
we
re
18
stu
dies
tha
t
fail
ed
to
m
atch
.
M
or
e
sp
eci
fical
ly
,
552
ke
ywords
we
re
t
est
ed
and
fou
nd
53
total
m
at
ches,
331
pa
rtia
l
m
a
tc
hes
an
d
16
8
m
at
ch
fail
ur
es.
As
sta
ti
sti
cal
nu
m
ber
s,
f
or
100
abstract
an
d
100
ti
tl
e
there
are
82%
m
a
tc
h
and
the
re
are
18%
fail
ure.
F
or
552
keyw
ord
s
there
are
9.6
%
total
m
at
ch,
wh
ic
h
is
m
ean
53
st
udy.
The
re
are
59.9%
pa
rtia
l
m
at
ch,
wh
ic
h
is
m
ean
331
st
ud
y.
T
he
re
are
30.4%
fail
ur
e m
at
ch,
wh
ic
h
is m
ean
168
st
udy.
REFERE
NCE
S
[1]
C.
Jiang,
et
al
.
,
"Im
pli
ci
t
sem
antics
base
d
m
et
ad
at
a
ext
ra
ct
ion
a
nd
m
at
chi
ng
of
schola
rl
y
do
cuments,
"
Journ
al
o
f
Database
Mana
geme
nt
(
JDM,
)
v
ol.
29
,
no
.
2
,
pp
.
1
-
22,
2018
.
[2]
E.
S.
Al
-
Sham
ery
and
H
.
Q.
Ghe
ni,
"P
la
g
ia
r
ism
det
e
ct
ion
using
sem
ant
ic
anal
y
s
i
s
,
"
Indian
Journal
of
Scienc
e
an
d
Technol
ogy
,
vo
l. 9, no.
1,
2016.
[3]
S.
Niwatt
ana
ku
l
,
et
al
,
"U
sing
of
Jac
ca
rd
co
eff
ic
i
ent
for
ke
y
w
ords
sim
il
ari
t
y
,
"
Proce
edi
ngs
of
the
Inte
rnation
al
Mult
iConf
ere
nc
e
of Engi
ne
ers an
d
Computer
Sc
ientists
,
vo
l. 1
-
6,
2
013.
[4]
A
M.
Hu
ss
ei
n,
"P
ara
phra
sing:
Soluti
o
n
to
the
Problem
of
Plagi
ari
sm
,
"
Asian
Journal
of
Infor
mation
Technology
,
vol.
15
-
16
,
pp
.
2
743
-
2746,
2016
.
[5]
M.
Habibi,
A.
Popescu
-
Belis,
"
Ke
y
word
ext
ra
ction
and
cl
ust
eri
n
g
for
document
r
ec
om
m
enda
ti
on
in
conv
ersa
ti
on
s,
"
IEE
E
/A
CM
Tr
an
sacti
ons on
aud
i
o,
spee
ch, and la
nguage
proce
ss
i
ng
,
vol
.
23
,
no
.
4
,
pp
.
746
-
759
,
2
015.
[6]
Y.
Matsuo
and
M.
Ishizuka,
"K
e
y
word
ext
r
acti
on
from
a
singl
e
document
usi
ng
word
co
-
oc
c
ur
ren
ce
sta
ti
sti
c
al
informati
on,
"
Int
ernati
onal
Journ
al
on
A
rtif
ic
ia
l
I
nte
lligen
ce Tools
,
vol.
13
,
no.
01
,
pp
.
157
-
169
,
2
004.
[7]
G.
K.
Palshika
r,
"K
e
y
word
ext
ra
ct
ion
from
a
single
doc
um
ent
using
ce
ntr
al
i
t
y
m
ea
sures,
"
Inte
rn
ati
onal
Confe
re
nc
e
on
Pattern
R
ec
o
gnit
ion
an
d
Mac
hine
In
te
l
li
gen
ce
,
Springer
,
B
erli
n,
Heid
el
b
erg
,
2
007
.
[8]
F.
Rouss
ea
u
an
d
V.
Michalis,
"
Main
cor
e
reten
ti
on
on
gra
ph
-
of
-
words
for
single
-
document
k
e
yword
ext
ra
ct
ion
,
"
European
Conf
e
renc
e
on
Inform
ati
on
Re
tri
ev
al
.
Springer,
Ch
am,
2015.
[9]
C.
Zh
ang,
"A
utomati
c
ke
y
wor
d
ext
r
ac
t
ion
fr
om
documents
using
cond
it
io
nal
r
andom
fields"
,
Journal
o
f
Computati
onal I
nformation
Syst
e
ms
,
vol. 4
-
3,
pp.
1169
-
1180,
200
8.
[10]
Y.
HaCohe
n
-
Kerne
r,
"A
utomati
c
ext
racti
on
of
ke
y
words
from
abstra
ct
s
,
"
Int
ernat
ional
Confe
ren
c
e
on
Knowle
dge
-
Based
an
d
I
ntell
ige
nt
Informatio
n
and
Eng
ineeri
ng
Syste
ms
,
Spri
nger
,
B
erlin,
Hei
del
ber
g
,
2003
.
[11]
A.
Hulth,
"Im
prove
d
aut
om
atic
ke
y
word
ext
ra
ction
give
n
m
ore
li
nguisti
c
knowledge
,
"
Proceed
ings
of
the
200
3
conf
ere
n
ce
on
E
mpirical
me
thod
s in
natural
lang
uage
proce
ss
ing
,
Association
for
Computati
onal
L
ingui
stic
s
,
2003
.
[12]
L.
Yang
,
et
a
l
.
,
"A
new
net
work
m
od
el
for
ext
ra
ct
ing
te
x
t
ke
y
wo
rds,
"
Sci
en
tometr
ic
s
,
vol
.
116
,
n
o
.
1,
pp
.
339
-
361
,
2018.
[13]
Sh.
Lut
hr
a,
et
al
,
"A
Statis
ti
c
al
Approac
h
of
K
e
y
word
Ext
ra
ct
io
n
for
Eff
ic
i
ent
R
et
ri
eva
l
,
"
Int
ernati
onal
Journal
of
Computer
Appli
cat
ions
,
vol
.
168
-
7,
2017
.
[14]
R.
C.
Anju,
et
al
.
,
"K
e
y
phra
se
an
d
Rel
ation
Ext
r
a
ct
ion
from
Scie
n
ti
fic
Publ
icati
ons
,
"
Adv
anc
es
in
Mac
hine
Learni
ng
and
Data
Sc
ie
n
c
e
,
Spring
er
,
Sing
apor
e, pp. 113
-
1
20,
2018
.
[15]
X.
W
an,
et
al
.
,
"Towards
an
i
te
ra
ti
ve
re
inforcem
ent
appr
oac
h
for
si
m
ult
ane
o
us
document
su
m
m
a
riz
at
ion
an
d
ke
y
word
ext
r
acti
on,
"
Proc
ee
ding
s of
th
e
45
th
ann
ual
mee
ti
ng
of th
e
ass
oci
at
ion
of
computat
ional l
i
nguisti
cs
,
2007
.
[16]
A.
Ruhul,
and
M.
Chakra
bort
y
,
"A
lgori
thm
for
Benga
li
Ke
y
wo
rd
Ext
racti
on
,
"
2018
Inte
rnatio
nal
Confe
renc
e
on
Bangl
a
Sp
ee
c
h
a
nd
Language
Pr
oce
ss
ing
(
ICBSLP)
.
IEEE
,
2018
.
[17]
A.
Ti
xie
r
,
et
al
,
"A
gra
ph
dege
ner
acy
-
b
ase
d
app
roa
ch
to
ke
y
wor
d
ext
racti
on
,
"
Pr
oce
ed
ings
of
the
2
016
Confe
renc
e
on
Empiric
a
l
M
e
thods i
n
Na
tural L
anguage
Proces
sing
,
2016.
[18]
A.
Onan,
"Ense
m
ble
of
ke
y
wor
d
ext
ra
ct
ion
m
e
thods
and
class
ifi
ers
in
t
ext
c
las
sific
at
ion
,
"
E
xp
ert
Syste
ms
wit
h
Appl
ic
a
ti
ons
,
vo
l.
57
,
pp
.
232
-
24
7,
2016
.
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
Suggest
in
g ne
w w
or
ds t
o ex
tract ke
yw
ords
from
ti
tl
e an
d ab
str
act
(
H
ade
el
Q
ase
m
Ghe
ni
)
4445
[19]
R.
Naidu
,
et
a
l
.
,
"
Te
xt
sum
m
ari
z
at
ion
with
aut
om
at
i
c
ke
y
word
ex
t
rac
t
ion
in
t
el
ugu
e
-
newspap
ers,
"
Smar
t
Computing
and
Informatic
s,
Springer
,
Singa
pore
,
pp
.
55
5
-
56
4,
2018
.
[20]
X.
W
u,
e
t
al
.
,
"
A
visual
at
t
ent
i
on
-
base
d
ke
y
wo
rd
ext
r
ac
t
ion
for
document
class
ifi
c
at
ion
,
"
Mul
tim
edi
a
Tools
an
d
Appl
ic
a
ti
ons
,
vo
l.
77
,
no
.
19
,
pp
.
25355
-
25367,
2
018.
[21]
Y.
Chen,
et
al
.
,
"S
ingl
e
document
ke
y
word
ext
r
ac
t
ion
via
quantif
y
ing
highe
r
-
or
der
struct
ura
l
fe
at
u
res
of
word
co
-
occ
urre
n
ce gra
p
h,
"
Computer
Sp
ee
ch
&
Languag
e
,
2019
.
[22]
S.
Ro
se,
e
t
al
,
"Rapi
d
aut
om
ati
c
ke
y
word
ex
tr
ac
t
ion
for
infor
m
at
ion
ret
r
ie
v
al
and
an
aly
s
is
,
"
U.S.
Pate
nt
No.
8,
131,
735
.
6
Ma
r
2012.
[23]
Sh.
La
hir
i,
"K
e
yword
and
ke
y
ph
rase
ext
r
action
u
sing
ce
ntr
al
i
t
y
m
ea
sures
on
collocation
ne
tworks
,
"
arXiv
pre
pr
int
arXiv
:
1401
.
657
1,
2014
.
[24]
S.
Beliga
,
"Tow
ard
sel
ectivit
y
b
ase
d
ke
y
word
ex
tra
c
ti
on
for
Croa
ti
an
news
,
"
arXi
v
pre
prin
t arXiv:
1407.
4723
,
201
4.
[25]
J.
Vill
m
ow,
et
a
l
.
,
"A
utoma
ti
c
k
e
y
phra
se
ex
tra
c
ti
on
using
r
e
cur
ren
t
neur
a
l
n
et
works
,
"
Inte
rn
ati
onal
Con
fe
re
nce
on
Mac
h
ine Learni
ng
and
Dat
a
Mi
ning
in
Pattern
Re
cogn
it
ion
,
Spr
inge
r
,
Ch
am,
20
18.
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
Hade
el
Qasem
Gheni
obta
in
ed
a
Bac
he
lor
'
s
Degre
e
in
Com
puter
Scie
nce
from
t
he
Univer
sit
y
of
Bab
y
lon
-
Facult
y
of
Sci
ence
f
or
W
om
en'
s
-
Com
pute
r
Depa
rtment
in
2006,
and
the
n
got
a
Master
'
s
Degre
e
in
Artificial
Inte
lligence
fr
om
the
Univer
sit
y
of
Bab
y
l
o
n
-
Facul
t
y
of
Inform
at
ion
T
echno
log
y
-
Soft
ware
Depa
r
tme
nt
in
2016
.
D
o
as
assistence
le
c
ture
r
at
the
Univer
sit
y
of
B
ab
y
lon/
Col
le
g
e
of
Scie
nc
e
for
W
om
en/
Com
put
er
Depa
r
tment
Since
2006
until
now.
Ahm
ed
Moha
mmed
Hussei
n,
Bac
he
lor
of
Comput
er
Scie
n
ce
in
2004
from
the
Univer
sit
y
o
f
Bab
y
lon/
Facu
lty
of
Sc
ie
nc
e,
M
aste
r
o
f
I
T/
univ
ersity
of
T
ena
g
a/
Mal
a
y
s
ia
in
2
012
,
and
do
as
le
c
ture
r
at
the
Univer
sit
y
of
B
ab
y
lon/
Col
le
g
e
of
Scie
n
ce
for
W
om
en/
Com
put
er
Depa
r
tment
Since
2006
till
n
ow.
W
ed
Kadhim
Olei
wi
got
a
B
ac
he
lor
Degr
ee
of
Com
pute
r
Scie
nc
e
from
t
he
Univer
si
t
y
o
f
Bab
y
lon/
Col
le
g
e
of
Scie
n
ce/Dep
art
m
ent
of
Com
pute
r
in
2005,
t
hen
e
arn
ed
a
M
aste
r
'
s
Degre
e
from
the
Univer
sit
y
of
Bab
y
l
on/
Coll
ege
of
Sc
ie
n
ce
/D
ep
art
m
ent
o
f
Com
pute
r
in
2012
in
the
fi
el
d
of
Artificial
Int
e
ll
ige
n
ce,
and
do
as
lectur
er
at
th
e
Univer
sit
y
of
B
ab
y
lon/
Col
le
g
e
of
Scie
n
ce
for
W
om
en/
Com
put
er
Dep
art
m
ent
S
inc
e
2007
ti
l
l
no
w.
Evaluation Warning : The document was created with Spire.PDF for Python.