I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
,
p
p
.
1
735
~
1
7
4
4
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
8
.
i
3
.
pp
1
7
3
5
-
1
7
4
4
1735
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
Text
Bug
g
er:
an e
x
ten
ded adv
ersa
r
ia
l t
ex
t
a
t
tack o
n
NLP
-
ba
sed
text c
la
ss
ificatio
n
mo
del
Sa
nja
ik
a
nth
E
.
Va
da
kk
et
hil
So
m
a
na
t
ha
n P
illa
i
1
,
Srini
v
a
s
A
.
Va
dd
a
di
2
,
Ro
hith V
a
lla
b
ha
neni
2
,
Sa
nto
s
h Re
dd
y
Ad
du
la
2
,
B
h
uv
a
nes
h Ana
ntha
n
3
1
S
c
h
o
o
l
o
f
E
l
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
U
n
i
v
e
r
si
t
y
o
f
N
o
r
t
h
D
a
k
o
t
a
,
G
r
a
n
d
F
o
r
k
s,
U
n
i
t
e
d
S
t
a
t
e
s
2
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
Te
c
h
n
o
l
o
g
y
,
U
n
i
v
e
r
si
t
y
o
f
t
h
e
C
u
m
b
e
r
l
a
n
d
s,
W
i
l
l
i
a
m
sb
u
r
g
,
U
n
i
t
e
d
S
t
a
t
e
s
3
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
i
c
a
l
a
n
d
El
e
c
t
r
o
n
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
P
S
N
C
o
l
l
e
g
e
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
Te
c
h
n
o
l
o
g
y
,
T
i
r
u
n
e
l
v
e
l
i
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Mar
19
,
2
0
2
4
R
ev
is
ed
No
v
19
,
202
4
Acc
ep
ted
No
v
24
,
2
0
2
4
Re
c
e
n
tl
y
,
a
d
v
e
rsa
rial
in
p
u
t
h
i
g
h
l
y
n
e
g
o
ti
a
tes
t
h
e
se
c
u
rit
y
c
o
n
c
e
rn
s
in
d
e
e
p
lea
rn
in
g
(DL)
tec
h
n
iq
u
e
s.
Th
e
m
a
in
m
o
ti
v
e
to
e
n
h
a
n
c
e
th
e
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
(NL
P
)
m
o
d
e
ls
is
t
o
l
e
a
rn
a
tt
a
c
k
s
a
n
d
se
c
u
re
a
g
a
in
st
a
d
v
e
rsa
rial
tex
t.
P
re
se
n
tl
y
,
th
e
a
n
tag
o
n
ist
ic
a
tt
a
c
k
tec
h
n
i
q
u
e
s
fa
c
e
so
m
e
issu
e
s
li
k
e
h
i
g
h
e
rro
r
a
n
d
tra
d
it
io
n
a
l
p
re
v
e
n
ti
o
n
a
p
p
ro
a
c
h
e
s
a
c
c
u
ra
tely
se
c
u
re
d
a
ta
a
g
a
in
st
h
a
rm
fu
l
a
tt
a
c
k
s.
He
n
c
e
,
so
m
e
a
tt
a
c
k
s
u
n
a
b
le
t
o
i
n
c
re
a
se
m
o
re
flaw
s
o
f
NLP
m
o
d
e
ls
th
e
re
b
y
in
tr
o
d
u
c
in
g
e
n
h
a
n
c
e
d
a
n
tag
o
n
isti
c
m
e
c
h
a
n
is
m
s
.
Th
e
p
ro
p
o
se
d
a
rti
c
le
i
n
tro
d
u
c
e
d
a
n
e
x
ten
d
e
d
te
x
t
a
d
v
e
rsa
rial
g
e
n
e
ra
ti
o
n
m
e
th
o
d
,
Tex
tBu
g
g
e
r.
I
n
it
iall
y
,
p
re
p
r
o
c
e
ss
in
g
ste
p
s
su
c
h
a
s
st
o
p
w
o
rd
(S
R)
re
m
o
v
a
l,
a
n
d
to
k
e
n
iza
ti
o
n
a
re
p
e
rfo
rm
e
d
t
o
re
m
o
v
e
n
o
ise
s
fro
m
th
e
tex
t
d
a
ta.
Th
e
n
,
v
a
rio
u
s
NLP
m
o
d
e
ls
li
k
e
B
i
-
d
irec
ti
o
n
a
l
e
n
c
o
d
e
r
re
p
re
se
n
tati
o
n
s
fro
m
tran
sfo
rm
e
rs
(BERT
),
ro
b
u
stl
y
o
p
ti
m
ize
d
BERT
(ROBERTa),
a
n
d
e
x
trem
e
lea
rn
in
g
m
a
c
h
i
n
e
n
e
u
ra
l
n
e
two
rk
(XLNe
t)
m
o
d
e
ls
a
re
a
n
a
ly
z
e
d
fo
r
o
u
t
p
u
tt
in
g
h
o
stil
e
tex
ts
.
Th
e
sim
u
latio
n
p
r
o
c
e
ss
is
c
a
rried
o
u
t
in
t
h
e
P
y
t
h
o
n
p
la
tfo
rm
a
n
d
a
p
u
b
li
c
ly
a
v
a
il
a
b
le
tex
t
c
las
sifica
ti
o
n
a
tt
a
c
k
d
a
tab
a
se
is
u
ti
li
z
e
d
f
o
r
t
h
e
trai
n
in
g
p
ro
c
e
ss
.
Va
rio
u
s
a
ss
e
ss
in
g
m
e
a
su
re
s
li
k
e
su
c
c
e
ss
ra
te,
ti
m
e
c
o
n
su
m
p
ti
o
n
,
p
o
si
ti
v
e
p
re
d
ictiv
e
v
a
lu
e
(P
P
V),
Ka
p
p
a
c
o
e
fficie
n
t
(KC),
a
n
d
F
-
m
e
a
su
re
a
re
a
n
a
ly
z
e
d
wit
h
d
iffere
n
t
Tex
tB
u
g
g
e
r
m
o
d
e
ls.
Th
e
o
v
e
ra
ll
su
c
c
e
ss
ra
te
a
c
h
iev
e
d
b
y
BERT
,
ROBERTa,
a
n
d
XLNe
t
is
a
b
o
u
t
9
8
.
6
%
,
9
9
.
7
%
,
a
n
d
9
6
.
8
%
re
sp
e
c
t
iv
e
ly
.
K
ey
w
o
r
d
s
:
Attack
d
etec
tio
n
B
E
R
T
Natu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
R
o
b
u
s
tly
o
p
tim
ized
B
E
R
T
T
ex
t a
d
v
er
s
ar
ies
XLNe
t
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
:
San
jaik
an
th
E
.
Vad
a
k
k
eth
il So
m
an
ath
an
Pil
lai
Sch
o
o
l o
f
E
lectr
ical
E
n
g
in
ee
r
in
g
an
d
C
o
m
p
u
ter
Scien
ce
,
Un
iv
er
s
ity
o
f
No
r
th
Dak
o
ta
Gr
an
d
Fo
r
k
s
,
ND
5
8
2
0
2
,
U
n
it
ed
States
E
m
ail: s.e
v
ad
ak
k
eth
il@
u
n
d
.
e
d
u
1.
I
NT
RO
D
UCT
I
O
N
I
n
to
d
a
y
’
s
s
ce
n
ar
io
,
th
e
u
s
e
o
f
th
e
d
ee
p
lea
r
n
in
g
(
DL
)
ap
p
r
o
ac
h
k
ee
p
s
o
n
in
cr
ea
s
in
g
r
esu
lts
in
th
e
in
tr
o
d
u
ctio
n
o
f
n
atu
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
in
g
(
NL
P
)
m
o
d
els.
I
t
is
n
o
ted
th
at
f
ascin
atin
g
r
esu
lts
ar
e
o
b
tain
ed
wh
ile
p
r
o
ce
s
s
in
g
th
e
NL
P
m
o
d
els
in
v
ar
io
u
s
f
ield
s
lik
e
que
s
tio
n
an
s
wer
in
g
,
s
en
tim
en
tal
an
aly
s
is
(
SA)
,
lan
g
u
ag
e
tr
an
s
latio
n
,
an
d
tex
t
m
an
ip
u
latio
n
.
Ast
u
d
illo
et
a
l.
[
1
]
,
it
is
n
o
ted
th
at
i
n
teg
r
atin
g
s
u
itab
le
p
er
tu
r
b
atio
n
s
ca
n
n
o
t
b
e
ea
s
ily
id
en
tifie
d
t
o
tex
t
d
ata
t
h
at
d
elib
er
ates
th
e
DL
m
o
d
els
to
p
r
o
d
u
ce
er
r
o
r
s
r
es
u
ltin
g
in
ad
v
er
s
ar
ial
attac
k
s
m
ain
ly
en
co
m
p
ass
ed
in
co
m
p
u
ter
v
is
io
n
ap
p
licatio
n
s
.
R
ec
en
tly
,
s
tu
d
ies
o
n
ad
v
er
s
ar
ial
attac
k
s
m
ad
e
o
u
ts
tan
d
in
g
in
tim
id
atio
n
in
N
L
P,
im
ag
e
p
r
o
ce
s
s
in
g
,
f
ac
e
id
en
tific
atio
n
,
an
d
in
tr
u
s
io
n
d
etec
tio
n
p
r
o
ce
s
s
es
[
2
]
,
[
3
]
.
I
t
is
an
aly
z
e
d
th
at
p
a
r
ticu
lar
NL
P
p
r
o
ce
s
s
es
lik
e
s
p
am
id
en
tific
atio
n
,
an
d
s
en
s
itiv
e
d
ata
d
etec
tio
n
ar
e
p
lay
in
g
an
in
teg
r
al
r
o
le
in
d
ata
p
r
o
ce
s
s
in
g
an
d
s
ec
u
r
ity
o
n
n
etwo
r
k
s
.
Hen
ce
,
it is
n
ec
ess
ar
y
to
en
h
an
ce
th
e
p
er
f
o
r
m
an
ce
o
f
NL
P m
o
d
els b
ased
o
n
DL
tech
n
iq
u
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
7
3
5
-
1
7
4
4
1736
Ho
wev
er
,
cr
ea
tin
g
a
d
v
er
s
ar
ial
in
p
u
ts
f
o
r
te
x
ts
is
h
ig
h
ly
ch
all
en
g
in
g
co
m
p
a
r
ed
t
o
cr
ea
tin
g
a
d
v
er
s
ar
ial
in
p
u
ts
in
im
a
g
es
[
4
]
,
[
5
]
.
T
h
e
tex
ts
ar
e
h
ig
h
l
y
r
a
n
d
o
m
,
co
n
q
u
e
r
in
g
th
e
p
er
s
is
ten
t
co
n
ce
p
t
o
f
an
im
ag
e.
Mo
r
eo
v
er
,
th
e
h
o
s
tile
tex
t
in
p
u
ts
ar
e
o
b
tain
e
d
v
ia
a
d
is
tu
r
b
in
g
ch
ar
ac
ter
-
lev
el
p
r
o
ce
s
s
th
at
ca
u
s
es
v
u
ln
er
ab
ilit
y
d
u
r
in
g
wo
r
d
co
r
r
ec
tio
n
an
d
r
ea
d
a
b
le
p
r
o
ce
s
s
es
[
6
]
,
[
7
]
.
T
h
is
p
r
o
ce
s
s
ca
n
cr
ea
te
h
ig
h
s
ec
u
r
ity
to
s
o
m
e
ex
ten
t
r
eg
ar
d
in
g
ch
ar
ac
ter
-
lev
el
attac
k
s
.
B
u
t
t
h
is
alter
atio
n
s
u
b
jects
to
in
cr
ea
s
ed
g
r
ad
ien
t
attac
k
s
th
at
ar
e
n
o
t
d
ir
ec
tly
im
p
lem
e
n
ted
o
n
th
e
tex
t.
I
n
ad
d
itio
n
to
th
i
s
,
in
teg
r
atin
g
s
u
b
-
wo
r
d
p
er
tu
r
b
atio
n
m
ay
ch
a
n
g
e
th
e
tex
t
in
to
o
u
t
-
of
-
v
o
ca
b
u
lar
y
(
OOV)
wo
r
d
s
.
T
h
e
tex
t
u
al
p
er
tu
r
b
atio
n
ca
n
cr
ea
te
a
n
e
n
h
an
ce
d
im
p
ac
t
o
n
s
em
an
tics
th
an
o
n
im
ag
es.
A
s
a
r
esu
lt,
it
is
d
if
f
icu
lt
to
en
h
an
ce
th
e
m
o
d
els
to
g
e
n
er
ate
ad
v
er
s
ar
ial
tex
tu
al
ex
am
p
les
[
8
]
,
[
9
]
.
T
o
o
v
er
c
o
m
e
th
e
co
n
s
o
f
ex
is
tin
g
m
eth
o
d
o
lo
g
ies,
th
is
ar
ticle
in
teg
r
ates
th
e
tex
tu
al
f
ea
tu
r
es
an
d
m
o
d
el
f
ea
tu
r
es to
d
ev
elo
p
a
m
u
ltip
le
attac
k
tech
n
i
q
u
e
n
a
m
ed
,
T
ex
tB
u
g
g
er
.
Mo
tiv
atio
n
:
n
o
wad
ay
s
,
th
e
DL
m
o
d
els
ar
e
b
ec
o
m
in
g
m
o
r
e
p
o
p
u
lar
in
class
if
y
in
g
ad
v
er
s
ar
ial
tex
t
b
ased
o
n
o
r
ig
in
al
tex
ts
.
Ho
w
ev
er
,
g
e
n
er
atin
g
a
d
v
er
s
ar
ial
d
ata
is
h
ig
h
ly
ch
allen
g
i
n
g
an
d
it
is
n
o
t
as
im
ag
e
ad
v
er
s
ar
ies.
T
o
o
v
er
c
o
m
e
t
h
is
is
s
u
e,
NL
P
-
ba
s
ed
DL
m
o
d
els
ar
e
in
t
r
o
d
u
ce
d
t
h
at
au
to
m
atica
lly
lear
n
m
ea
n
in
g
f
u
l
s
en
ten
ce
s
a
n
d
class
if
y
th
e
h
o
s
tile
tex
t
ef
f
ec
tiv
el
y
.
So
m
e
o
f
th
e
co
m
m
o
n
l
y
u
s
ed
NL
P
s
ch
em
es
ar
e
Bi
-
d
ir
ec
tio
n
al
en
co
d
er
r
ep
r
es
en
tatio
n
s
f
r
o
m
tr
a
n
s
f
o
r
m
er
s
(
B
E
R
T
)
,
r
o
b
u
s
tly
o
p
tim
ized
B
E
R
T
(
R
OB
E
R
T
a
)
,
an
d
ex
tr
em
e
lear
n
i
n
g
m
ac
h
i
n
e
n
eu
r
al
n
etwo
r
k
(
XL
Net
)
m
o
d
els
th
at
u
s
e
co
n
tex
tu
al
em
b
e
d
d
in
g
p
r
o
p
e
r
ty
an
d
p
r
ev
en
t
lo
n
g
-
ter
m
d
e
p
en
d
e
n
c
y
p
r
o
b
lem
s
.
Mo
tiv
ated
b
y
th
i
s
,
th
e
d
ev
elo
p
ed
f
r
am
ewo
r
k
i
n
v
esti
g
ated
s
ev
er
al
NL
P
m
o
d
els
in
clas
s
if
y
in
g
ad
v
er
s
ar
ial
tex
ts
u
s
i
n
g
o
r
ig
in
a
l
tex
ts
.
T
h
e
k
ey
co
n
tr
ib
u
tio
n
s
o
f
th
e
d
ev
elo
p
ed
f
r
am
ewo
r
k
ar
e
d
escr
ib
e
d
as f
o
llo
ws:
−
T
o
in
tr
o
d
u
ce
a
n
ex
te
n
d
ed
tex
t
attac
k
NL
P
s
ch
em
e
to
an
aly
ze
its
p
er
f
o
r
m
an
ce
in
class
if
y
in
g
ad
v
er
s
ar
ial
o
u
tco
m
es
.
−
T
o
an
al
y
ze
v
a
r
io
u
s
n
atu
r
al
lan
g
u
ag
e
m
o
d
els
lik
e
B
E
R
T
,
R
OB
E
R
T
a,
an
d
XlNet
in
class
if
y
in
g
ad
v
er
s
ar
ial
tex
t b
ased
o
n
te
x
tu
al
o
u
t
p
u
t.
−
T
o
v
alid
ate
th
e
ex
is
tin
g
B
E
R
T
,
R
O
B
E
R
T
a,
an
d
XL
Net
-
b
ased
NL
P
m
o
d
els
b
y
ass
ess
in
g
d
if
f
er
en
t
p
er
f
o
r
m
an
ce
m
ea
s
u
r
es
lik
e
a
cc
u
r
ac
y
,
K
ap
p
a
co
ef
f
icien
t
(
KC
)
,
p
o
s
itiv
e
p
r
e
d
ictiv
e
v
alu
e
(
PP
V)
,
a
n
d
F
-
m
ea
s
u
r
e
m
etr
ics.
T
h
e
u
p
co
m
in
g
s
ec
tio
n
s
a
r
e
o
r
g
an
ized
as
f
o
llo
ws:
s
ec
tio
n
2
o
u
tlay
s
th
e
s
ec
tio
n
a
b
o
u
t
r
ela
ted
wo
r
k
,
s
ec
tio
n
3
d
elib
er
ates
o
v
er
th
e
s
u
g
g
ested
m
eth
o
d
o
lo
g
y
,
s
ec
tio
n
4
p
r
esen
ts
th
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
,
an
d
s
ec
tio
n
5
r
ep
r
e
s
en
ts
th
e
c
o
n
clu
s
io
n
o
f
th
e
p
r
o
p
o
s
ed
f
r
a
m
ewo
r
k
.
2.
L
I
T
E
R
AT
U
RE
SU
RVE
Y
Sey
y
ar
et
a
l.
[
1
0
]
d
ef
in
e
d
th
e
B
E
R
T
m
o
d
el
f
o
r
class
if
y
in
g
t
ex
t
attac
k
s
to
ass
is
t
v
ar
io
u
s
tex
t
-
r
elate
d
ap
p
licatio
n
s
.
I
n
th
is
s
tu
d
y
,
HT
T
P
r
eq
u
ests
wer
e
co
n
s
id
er
ed
to
d
etec
t
g
en
u
i
n
e
an
d
m
alicio
u
s
tex
ts
ef
f
ec
tiv
ely
.
Mo
r
eo
v
er
,
s
ix
f
u
lly
co
n
n
ec
te
d
(
FC
)
lay
er
s
o
f
m
u
ltil
ay
er
p
er
ce
p
tr
o
n
(
ML
P)
wer
e
u
tili
ze
d
to
class
if
y
th
e
ad
v
er
s
ar
ial
tex
ts
.
I
n
th
e
e
x
p
e
r
im
en
tal
p
ar
t,
ac
cu
r
ac
y
an
d
F
-
m
ea
s
u
r
e
wer
e
an
aly
ze
d
a
n
d
d
is
tin
g
u
is
h
ed
f
r
o
m
o
th
er
s
tu
d
ies.
Ho
we
v
er
,
th
e
lo
n
g
-
ter
m
d
ep
en
d
en
cy
p
r
o
b
lem
s
wer
e
u
n
s
o
lv
ed
f
o
r
lar
g
er
d
o
cu
m
en
ts
.
L
iu
et
a
l.
[
1
1
]
,
p
u
t
f
o
r
t
h
a
s
ec
u
r
e
tex
t
s
im
ilar
ity
p
r
o
to
co
l
f
o
r
m
alicio
u
s
tex
t
class
if
icat
io
n
attac
k
s
in
th
e
DL
m
o
d
el.
Her
e
,
th
e
ellip
tic
-
cu
r
v
e
c
r
y
p
to
g
r
ap
h
y
(
E
C
C
)
tech
n
iq
u
e
was
in
t
r
o
d
u
ce
d
to
en
h
an
ce
th
e
m
o
d
el
ex
ec
u
tio
n
e
f
f
icien
cy
.
T
h
e
n
,
th
e
m
alicio
u
s
b
eh
a
v
io
r
o
f
t
h
e
s
e
m
i
-
h
o
n
est
p
r
o
to
co
l
was
ex
am
i
n
ed
an
d
co
m
b
in
ed
with
ze
r
o
-
k
n
o
wled
g
e
-
p
r
o
o
f
a
n
d
cu
t
-
ch
o
o
s
e
s
ch
em
es.
I
n
th
e
ex
p
er
im
en
tal
p
ar
t,
ac
cu
r
ac
y
an
d
ex
ec
u
tio
n
tim
e
wer
e
an
aly
ze
d
a
n
d
d
is
tin
g
u
is
h
ed
f
r
o
m
o
th
er
s
tu
d
ies.
Ho
w
ev
er
,
th
is
m
eth
o
d
was
h
ig
h
ly
s
en
s
itiv
e
to
wo
r
d
len
g
th
an
d
in
cr
ea
s
ed
er
r
o
r
d
u
r
i
n
g
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
.
Z
h
an
g
et
a
l.
[
1
2
]
,
estab
lis
h
ed
th
e
DL
-
b
ased
ad
v
er
s
ar
ial
tex
t
class
if
icatio
n
tech
n
iq
u
e
u
s
in
g
a
v
ir
tu
al
tr
ain
in
g
p
r
o
ce
s
s
.
Fo
r
wo
r
d
em
b
ed
d
in
g
,
b
ag
-
of
-
wo
r
d
s
(
B
o
W
)
was
u
tili
ze
d
,
p
er
f
o
r
m
in
g
v
ec
to
r
izatio
n
o
v
er
ea
ch
d
atab
ase.
T
h
e
E
lec,
I
MD
B
,
an
d
R
o
tten
-
b
ased
th
ir
d
b
en
c
h
m
a
r
k
d
atasets
wer
e
u
s
ed
f
o
r
t
h
e
t
r
ain
in
g
p
r
o
ce
s
s
.
I
n
th
e
ex
p
er
im
en
tal
p
ar
t,
ac
c
u
r
ac
y
,
an
d
lo
s
s
wer
e
an
aly
ze
d
a
n
d
d
is
tin
g
u
is
h
ed
f
r
o
m
o
th
e
r
s
tu
d
ies.
Ho
wev
er
,
th
is
m
eth
o
d
ca
u
s
es h
ig
h
b
lack
-
b
o
x
is
s
u
es a
n
d
lack
s
its
in
ter
p
r
etab
ilit
y
o
v
er
u
n
s
tr
u
ctu
r
e
d
tex
t d
ata.
B
ajaj
an
d
Vis
h
wak
ar
m
a
[
1
3
]
,
a
h
o
s
tile
attac
k
p
r
o
to
co
l
f
o
r
o
u
tp
u
ttin
g
tex
t
v
u
ln
er
a
b
ilit
ies
o
v
er
DL
-
b
ased
s
en
tim
en
t c
lass
if
ier
s
.
V
ar
io
u
s
p
o
p
u
lar
NL
P
-
b
ased
DL
m
o
d
els lik
e
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN
)
an
d
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
alo
n
g
with
f
iv
e
d
if
f
er
en
t
tr
an
s
f
o
r
m
er
m
e
t
h
o
d
s
wer
e
u
tili
ze
d
.
Mo
r
eo
v
er
,
th
e
MR
an
d
I
MD
B
-
b
ased
tw
o
b
en
ch
m
a
r
k
d
atasets
wer
e
u
s
ed
f
o
r
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
I
n
th
e
ex
p
er
im
e
n
tal
p
ar
t,
ac
cu
r
ac
y
,
s
en
s
itiv
ity
,
an
d
r
u
n
-
tim
e
wer
e
a
n
aly
ze
d
an
d
d
is
tin
g
u
is
h
ed
f
r
o
m
o
th
er
s
tu
d
ie
s
.
Ho
wev
er
,
r
ec
en
t
NL
P m
o
d
els lik
e
R
OB
E
R
T
a
an
d
XL
Net
f
ailed
to
c
o
n
s
id
er
f
o
r
class
if
y
in
g
ad
v
er
s
ar
ial
tex
ts
.
B
ao
et
a
l.
[
1
4
]
,
in
tr
o
d
u
ce
d
a
s
co
r
e
lev
el
n
etwo
r
k
f
o
r
d
etec
tin
g
h
o
s
tile
tex
ts
ac
c
u
r
ately
.
Her
e,
th
e
class
-
awa
r
e
s
co
r
e
n
etwo
r
k
(
C
ASN)
m
o
d
el
was
em
p
h
asized
to
id
en
tify
th
e
tex
t
o
v
er
ad
v
er
s
ar
ial
tr
ain
in
g
.
Mo
r
eo
v
er
,
th
e
co
s
in
e
s
im
ilar
ity
was
p
er
f
o
r
m
ed
t
o
d
en
o
is
e
th
e
u
n
wan
te
d
tex
t
d
ata.
T
h
e
SS
T
-
1
,
SS
T
-
2
,
I
MD
B
,
an
d
AGNE
W
S
-
b
ased
f
o
u
r
b
e
n
ch
m
ar
k
d
atasets
wer
e
u
s
ed
f
o
r
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
I
n
th
e
ex
p
er
im
en
tal
p
ar
t,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
TextB
u
g
g
er:
a
n
ex
te
n
d
ed
a
d
ve
r
s
a
r
ia
l
text
a
tta
ck
on
…
(
S
a
n
j
a
ika
n
th
E
.
V
a
d
a
kk
eth
il S
o
m
a
n
a
th
a
n
P
illa
i
)
1737
a
r
e
a
u
n
d
e
r
t
h
e
r
e
c
ei
v
e
r
o
p
e
r
a
t
in
g
c
h
a
r
a
c
t
e
r
is
t
i
c
c
u
r
v
e
(
A
UR
OC
)
,
a
n
d
F
-
m
e
a
s
u
e
w
e
r
e
a
n
a
l
y
ze
d
a
n
d
d
i
s
ti
n
g
u
i
s
h
e
d
f
r
o
m
o
t
h
e
r
s
t
u
d
i
e
s
.
H
o
w
e
v
e
r
,
th
e
t
i
m
e
c
o
m
p
l
e
x
it
y
w
a
s
h
i
g
h
l
y
l
i
k
e
l
y
t
o
c
a
u
s
e
h
i
g
h
o
v
e
r
f
i
t
ti
n
g
i
s
s
u
e
s
.
Me
an
wh
ile,
d
ee
p
n
eu
r
al
n
et
wo
r
k
s
(
DNNs
)
-
b
ased
tex
t
c
lass
if
icatio
n
i
s
b
ec
o
m
i
n
g
in
cr
ea
s
in
g
ly
s
ig
n
if
ican
t
in
to
d
ay
’
s
in
f
o
r
m
atio
n
an
aly
s
is
an
d
co
m
p
r
eh
e
n
s
io
n
.
Fo
r
ex
am
p
le,
s
en
tim
en
t
an
aly
s
is
o
f
u
s
er
r
ev
iews
an
d
co
m
m
en
ts
is
a
k
ey
co
m
p
o
n
en
t
o
f
m
an
y
o
n
lin
e
r
ec
o
m
m
e
n
d
atio
n
s
y
s
tem
s
[
1
5
]
.
T
h
ese
k
in
d
s
o
f
alg
o
r
ith
m
s
wo
u
l
d
o
f
ten
d
iv
id
e
th
e
r
e
v
i
ews
an
d
c
o
m
m
en
ts
i
n
to
two
o
r
t
h
r
ee
g
r
o
u
p
s
,
th
en
r
an
k
t
h
e
m
o
v
ies
o
r
p
r
o
d
u
cts
b
ased
o
n
th
e
r
esu
lt
s
.
T
ex
t
class
if
icat
io
n
p
lay
s
a
cr
u
cial
r
o
le
in
im
p
r
o
v
in
g
t
h
e
s
af
ety
o
f
o
n
lin
e
d
is
cu
s
s
io
n
s
p
ac
es.
Fo
r
ex
a
m
p
le,
it
ca
n
b
e
u
s
ed
t
o
au
t
o
m
atica
lly
id
en
tify
o
n
lin
e
to
x
ic
co
n
ten
t
[
1
6
]
,
wh
ich
in
clu
d
es
in
s
u
lts
,
s
ar
ca
s
m
,
ab
u
s
e,
h
ar
ass
m
en
t,
a
n
d
ir
o
n
y
.
Nu
m
er
o
u
s
r
esear
ch
wo
r
k
s
h
av
e
ex
am
i
n
ed
t
h
e
s
ec
u
r
ity
o
f
ex
is
tin
g
m
ac
h
in
e
lear
n
in
g
m
o
d
els
an
d
h
av
e
p
u
t
f
o
r
th
s
ev
er
al
attac
k
t
ec
h
n
iq
u
es,
s
u
ch
as
ex
p
lo
r
ato
r
y
an
d
ca
u
s
al
attac
k
s
[
1
7
]
–
[
1
9
]
.
E
x
p
lo
r
at
o
r
y
attac
k
s
cr
ea
te
h
o
s
tile
test
in
g
ca
s
es
(
ad
v
er
s
ar
ial
ex
am
p
les)
in
o
r
d
er
to
el
u
d
e
a
p
ar
ticu
lar
class
if
ier
,
wh
ile
ca
u
s
ativ
e
attac
k
s
tr
y
to
m
o
d
if
y
th
e
tr
ain
in
g
d
ata
i
n
o
r
d
er
to
tr
ick
th
e
class
if
ier
its
elf
.
Nu
m
er
o
u
s
m
eth
o
d
s
h
av
e
b
ee
n
p
u
t
f
o
r
t
h
to
p
r
o
d
u
ce
r
o
b
u
s
t
class
if
ier
s
in
o
r
d
er
to
f
e
n
d
o
f
f
th
ese
attac
k
s
[
2
0
]
,
[
2
1
]
.
Ad
v
er
s
ar
ial
ass
au
lts
h
av
e
d
em
o
n
s
tr
ated
a
h
ig
h
attac
k
s
u
cc
ess
r
ate
in
im
ag
e
class
if
icatio
n
task
s
r
ec
en
tly
[
2
2
]
,
wh
ich
h
as
p
u
t
m
a
n
y
in
tellig
en
t
d
ev
ices
s
u
ch
as
s
elf
-
d
r
iv
in
g
ca
r
s
in
g
r
av
e
d
a
n
g
er
[
2
3
]
,
[
2
4
]
.
R
esear
ch
g
ap
s
in
T
ex
tB
u
g
g
er
:
an
ex
te
n
d
ed
ad
v
e
r
s
ar
ial
tex
t
atta
ck
o
n
NL
P
-
b
ased
tex
t
cla
s
s
if
icatio
n
m
o
d
el
s
p
r
esen
t
s
ev
er
al
o
p
p
o
r
t
u
n
ities
f
o
r
ex
p
lo
r
atio
n
.
On
e
k
ey
ar
ea
is
th
e
r
o
b
u
s
tn
ess
o
f
m
o
d
els
ag
ain
s
t
m
o
r
e
s
o
p
h
is
ticated
ad
v
er
s
ar
ial
attac
k
s
.
T
ex
tB
u
g
g
er
h
as
d
em
o
n
s
tr
ated
v
u
ln
er
ab
ilit
ies
in
tex
t
cla
s
s
if
icatio
n
m
o
d
els,
b
u
t
f
u
r
th
er
r
esear
c
h
is
n
ee
d
e
d
to
ex
p
lo
r
e
m
o
r
e
co
m
p
lex
an
d
co
n
tex
t
-
awa
r
e
p
e
r
tu
r
b
ati
o
n
s
.
Su
ch
ad
v
an
ce
d
attac
k
s
co
u
ld
ex
p
lo
it
d
ee
p
er
lin
g
u
is
tic
f
ea
t
u
r
es,
r
eq
u
ir
in
g
m
o
d
els
to
b
e
eq
u
ip
p
ed
with
s
tr
o
n
g
er
d
e
f
en
s
es
ca
p
ab
le
o
f
r
ec
o
g
n
izin
g
s
u
b
tle
ch
an
g
es in
ad
v
er
s
ar
ial
in
p
u
ts
.
An
o
th
er
s
ig
n
if
ican
t
r
esear
ch
g
ap
lies
in
d
ev
elo
p
in
g
d
ef
e
n
s
e
m
ec
h
an
is
m
s
s
p
ec
if
ically
t
ailo
r
ed
to
tex
tu
al
d
ata.
W
h
ile
T
ex
tB
u
g
g
er
ex
p
o
s
es
wea
k
n
ess
es
in
ex
i
s
tin
g
NL
P
m
o
d
els,
th
e
s
tu
d
y
o
f
ef
f
ec
tiv
e
d
ef
en
s
e
s
tr
ateg
ies
r
em
ain
s
u
n
d
er
d
ev
elo
p
ed
.
T
ec
h
n
iq
u
es
lik
e
ad
v
er
s
ar
ial
tr
ain
in
g
,
n
o
is
e
-
in
je
ctio
n
,
an
d
ce
r
tifie
d
r
o
b
u
s
tn
ess
h
av
e
b
ee
n
e
x
p
lo
r
e
d
in
v
is
io
n
m
o
d
els
b
u
t
n
ee
d
f
u
r
th
er
r
ef
in
e
m
en
t
a
n
d
test
in
g
in
th
e
NL
P
d
o
m
ain
,
p
ar
ticu
lar
ly
in
h
an
d
lin
g
d
iv
e
r
s
e
tex
t stru
ctu
r
es a
n
d
m
ea
n
in
g
s
.
T
h
e
cr
o
s
s
-
lin
g
u
al
an
d
m
u
lti
-
task
v
u
ln
er
a
b
ilit
ies
o
f
NL
P
m
o
d
els
u
n
d
er
ad
v
er
s
ar
ial
attac
k
s
also
war
r
an
t
f
u
r
th
er
i
n
v
esti
g
atio
n
.
T
ex
tB
u
g
g
er
p
r
im
ar
ily
f
o
c
u
s
es
o
n
E
n
g
lis
h
tex
t,
l
ea
v
in
g
o
p
en
q
u
esti
o
n
s
ab
o
u
t
h
o
w
ad
v
er
s
ar
ial
attac
k
s
im
p
a
ct
m
o
d
els
th
at
o
p
er
ate
in
m
u
ltip
le
lan
g
u
ag
es
o
r
p
e
r
f
o
r
m
v
ar
io
u
s
task
s
lik
e
s
en
tim
en
t
an
aly
s
is
an
d
n
am
ed
en
tity
r
ec
o
g
n
itio
n
.
R
esear
ch
in
th
is
ar
ea
ca
n
p
r
o
v
i
d
e
in
s
ig
h
ts
in
to
th
e
g
en
er
aliza
tio
n
a
n
d
tr
an
s
f
e
r
ab
il
ity
o
f
ad
v
er
s
ar
ial
v
u
ln
e
r
ab
ilit
ies ac
r
o
s
s
lin
g
u
is
tic
b
o
u
n
d
a
r
ies.
An
o
th
er
g
ap
r
elate
s
to
t
h
e
tr
an
s
f
er
ab
ilit
y
o
f
ad
v
e
r
s
ar
ial
ex
am
p
les.
W
h
ile
T
ex
tB
u
g
g
er
s
h
o
wca
s
es
v
u
ln
er
ab
ilit
ies
in
s
p
ec
if
ic
m
o
d
els,
it
r
em
ain
s
u
n
clea
r
h
o
w
tr
an
s
f
er
ab
le
th
ese
ad
v
er
s
ar
ial
attac
k
s
ar
e
ac
r
o
s
s
d
if
f
er
en
t
ar
ch
itectu
r
es,
p
ar
ticu
lar
ly
in
m
o
d
er
n
tr
a
n
s
f
o
r
m
e
r
-
b
ased
m
o
d
els
lik
e
B
E
R
T
an
d
GPT.
E
x
p
lo
r
i
n
g
t
h
e
cr
o
s
s
-
m
o
d
el
tr
an
s
f
er
a
b
ilit
y
o
f
ad
v
er
s
ar
ial
attac
k
s
ca
n
h
elp
u
n
d
er
s
tan
d
h
o
w
to
b
u
ild
m
o
r
e
r
o
b
u
s
t
ar
ch
itectu
r
es
th
at
ca
n
d
ef
e
n
d
ag
ai
n
s
t a
wid
er
ar
r
ay
o
f
th
r
ea
ts
.
Fu
r
th
er
m
o
r
e
,
h
u
m
an
p
er
ce
p
ti
b
ilit
y
an
d
s
em
an
tic
p
r
eser
v
ati
o
n
is
an
o
th
e
r
im
p
o
r
tan
t
ar
ea
f
o
r
f
u
tu
r
e
r
esear
ch
.
Alth
o
u
g
h
T
ex
tB
u
g
g
er
aim
s
to
cr
ea
te
ad
v
er
s
ar
ial
ex
am
p
les
th
at
r
em
ain
im
p
er
ce
p
tib
le
to
h
u
m
an
s
,
th
e
ex
ten
t
to
wh
ich
t
h
ese
attac
k
s
p
r
eser
v
e
th
e
o
r
ig
in
al
m
ea
n
in
g
an
d
c
o
h
er
en
ce
o
f
th
e
te
x
t
r
eq
u
ir
es
f
u
r
th
er
ev
alu
atio
n
.
Stu
d
ies
ar
e
n
ee
d
ed
to
ass
ess
th
e
b
alan
ce
b
e
twee
n
attac
k
s
u
cc
ess
an
d
th
e
p
r
eser
v
atio
n
o
f
s
em
an
tics
,
esp
ec
ially
f
o
r
m
o
r
e
co
m
p
lex
NL
P task
s
wh
er
e
m
a
in
tain
in
g
m
ea
n
i
n
g
is
cr
u
cial.
T
h
e
r
ea
l
-
wo
r
ld
a
p
p
licab
ilit
y
o
f
T
ex
tB
u
g
g
er
-
s
ty
le
attac
k
s
a
ls
o
r
eq
u
ir
es
f
u
r
th
er
r
esear
ch
.
E
v
alu
atin
g
h
o
w
ad
v
er
s
ar
ial
tex
t
m
an
ip
u
latio
n
s
im
p
ac
t
r
ea
l
-
tim
e
ap
p
licatio
n
s
,
s
u
ch
as
s
p
am
d
etec
tio
n
,
f
ak
e
n
ews
m
o
d
er
atio
n
,
an
d
c
o
n
ten
t
f
ilt
er
in
g
s
y
s
tem
s
,
i
s
ess
en
tial.
Un
d
er
s
tan
d
in
g
t
h
e
b
eh
a
v
io
r
o
f
th
ese
attac
k
s
in
p
r
ac
tical
s
ettin
g
s
,
esp
ec
ially
t
h
o
s
e
with
h
u
m
an
-
in
-
t
h
e
-
lo
o
p
s
y
s
tem
s
o
r
m
u
ltip
le
lay
er
s
o
f
f
ilter
in
g
,
ca
n
s
h
ed
lig
h
t o
n
p
o
ten
tial d
ef
e
n
s
e
m
ec
h
an
is
m
s
an
d
s
y
s
tem
v
u
ln
e
r
ab
i
liti
es.
Mo
r
eo
v
er
,
ad
v
e
r
s
ar
ia
l
attac
k
s
o
n
tr
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
r
em
ain
r
elativ
ely
u
n
ex
p
l
o
r
ed
.
T
ex
tB
u
g
g
er
’
s
an
aly
s
is
f
o
cu
s
es
p
r
im
ar
ily
o
n
tr
ad
itio
n
al
N
L
P
m
o
d
els,
b
u
t
with
th
e
in
cr
ea
s
in
g
ad
o
p
tio
n
o
f
tr
an
s
f
o
r
m
er
s
,
th
e
r
e
is
a
p
r
ess
in
g
n
ee
d
to
u
n
d
er
s
tan
d
h
o
w
r
esil
ien
t
th
ese
n
ewe
r
m
o
d
els
ar
e
to
a
d
v
er
s
ar
ial
attac
k
s
.
R
esear
ch
in
to
ex
ten
d
in
g
T
ex
tB
u
g
g
er
’
s
m
eth
o
d
o
lo
g
y
to
tr
an
s
f
o
r
m
er
-
b
ased
ar
ch
it
ec
tu
r
es
lik
e
B
E
R
T
,
GPT,
an
d
T
5
will p
r
o
v
id
e
i
n
s
ig
h
ts
in
to
th
e
r
o
b
u
s
tn
ess
o
f
s
tate
-
of
-
th
e
-
ar
t m
o
d
els.
L
astl
y
,
th
er
e
is
a
g
ap
i
n
u
n
d
e
r
s
tan
d
in
g
attac
k
g
e
n
er
aliza
tio
n
ac
r
o
s
s
v
ar
io
u
s
NL
P
task
s
b
ey
o
n
d
tex
t
class
if
icatio
n
.
T
h
e
v
er
s
atility
o
f
ad
v
er
s
ar
ial
attac
k
s
,
s
u
ch
as
T
ex
tB
u
g
g
er
,
in
o
th
er
d
o
m
ain
s
lik
e
m
ac
h
in
e
tr
an
s
latio
n
,
tex
t
s
u
m
m
ar
izati
o
n
,
o
r
q
u
esti
o
n
-
an
s
wer
in
g
s
y
s
tem
s
r
em
ain
s
lar
g
ely
u
n
ex
p
lo
r
ed
.
I
n
v
esti
g
at
in
g
h
o
w
th
ese
attac
k
s
g
en
er
alize
t
o
m
o
r
e
c
o
m
p
lex
an
d
d
iv
er
s
e
NL
P
task
s
will
h
elp
id
en
tify
m
o
r
e
co
m
p
r
eh
e
n
s
iv
e
d
ef
en
s
e
s
tr
ateg
ies.
B
y
ad
d
r
ess
in
g
th
ese
r
esear
ch
g
ap
s
,
ad
v
an
ce
m
en
ts
ca
n
b
e
m
ad
e
in
b
u
ild
in
g
m
o
r
e
s
ec
u
r
e
an
d
r
esil
ien
t
NL
P
m
o
d
els,
wh
ich
ar
e
cr
u
cial
f
o
r
t
h
e
r
eliab
le
d
ep
lo
y
m
e
n
t
o
f
AI
-
d
r
iv
en
s
y
s
tem
s
in
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
7
3
5
-
1
7
4
4
1738
Pro
b
lem
s
tatem
en
t
:
Fro
m
th
e
d
ee
p
an
aly
s
is
o
f
th
e
c
o
n
v
en
tio
n
al
s
tu
d
ies,
it
is
n
o
te
d
th
at
th
e
v
u
ln
er
ab
ilit
y
o
f
th
ese
m
o
d
els
h
as
f
ailed
to
m
itig
ate
in
ter
m
s
o
f
h
ar
m
f
u
l
h
o
s
tile
attac
k
s
.
T
h
e
m
in
o
r
ch
an
g
es
in
in
p
u
t
tex
ts
ca
n
lead
t
o
in
ac
c
u
r
ate
class
if
icatio
n
s
.
T
h
ese
a
d
v
er
s
ar
ial
attac
k
s
wea
k
en
th
e
m
o
d
el
a
n
d
ca
u
s
e
s
er
io
u
s
co
n
s
eq
u
e
n
ce
s
lik
e
th
e
tr
an
s
m
is
s
io
n
o
f
m
an
ip
u
lated
d
ata
o
r
v
u
ln
er
a
b
ilit
y
to
au
t
o
m
ated
tech
n
iq
u
es.
R
ec
en
tly
,
s
ev
er
al
c
h
allen
g
es
h
av
e
b
ee
n
f
ac
ed
to
id
en
tify
ef
f
ec
tiv
e
tech
n
i
q
u
es
th
at
ca
n
ac
cu
r
ately
s
ec
u
r
e
tex
t
attac
k
s
an
d
en
h
an
c
e
th
e
r
elia
b
ilit
y
o
f
th
e
tex
t
class
if
icatio
n
p
r
o
ce
s
s
.
No
wad
ay
s
,
NL
P
m
o
d
els
ar
e
p
lay
in
g
a
n
in
teg
r
al
r
o
le
in
s
ev
e
r
al
tex
t
-
r
el
ated
ap
p
licatio
n
s
t
h
at
m
ain
tai
n
th
eir
p
o
p
u
lar
ity
e
v
en
th
o
u
g
h
lar
g
er
s
am
p
les
ar
e
p
r
o
ce
s
s
ed
.
Hen
ce
,
th
is
ar
ticle
in
v
esti
g
ated
v
ar
io
u
s
NL
P
m
o
d
els
in
tex
t
class
if
icatio
n
attac
k
s
o
v
er
o
r
i
g
in
al
ex
am
p
les.
3.
P
RO
P
O
SE
D
M
E
T
H
O
D
T
h
e
p
r
o
p
o
s
ed
ar
ticle
in
tr
o
d
u
ce
d
an
ex
ten
d
e
d
tex
t
ad
v
e
r
s
ar
ial
g
en
er
atio
n
m
et
h
o
d
,
T
e
x
tB
u
g
g
er
.
I
n
itially
,
p
r
e
p
r
o
ce
s
s
in
g
s
tep
s
s
u
ch
as
k
ey
wo
r
d
s
elec
tio
n
(
KS)
ar
e
p
er
f
o
r
m
e
d
to
r
em
o
v
e
n
o
is
es
f
r
o
m
th
e
tex
t
d
ata.
T
h
en
,
v
a
r
io
u
s
NL
P
m
o
d
els
lik
e
B
E
R
T
,
R
O
B
E
R
T
a,
an
d
XL
Net
m
o
d
els
ar
e
an
aly
ze
d
f
o
r
o
u
tp
u
ttin
g
h
o
s
tile te
x
ts
.
Fig
u
r
e
1
in
d
icate
s
th
e
wo
r
k
f
lo
w
o
f
th
e
d
ev
elo
p
ed
f
r
am
ewo
r
k
.
Fig
u
r
e
1
.
W
o
r
k
f
lo
w
o
f
th
e
d
e
v
elo
p
ed
f
r
am
ewo
r
k
3
.
1
.
P
re
pro
ce
s
s
ing
s
t
a
g
e
I
n
itially
,
th
e
r
aw
tex
t
d
ata
co
llected
f
r
o
m
p
u
b
lic
s
o
u
r
ce
s
ar
e
p
r
ep
r
o
ce
s
s
ed
b
y
p
e
r
f
o
r
m
in
g
th
e
T
o
k
en
izatio
n
p
r
o
ce
s
s
.
T
h
e
d
et
ailed
an
aly
s
is
o
f
ea
ch
s
tag
e
is
d
ep
icted
b
elo
w.
3.
1
.
1
.
K
ey
wo
rd
s
elec
t
io
n
I
t
is
th
e
p
r
o
ce
s
s
o
f
s
ep
ar
atin
g
th
e
tex
tu
al
d
ata
in
to
m
i
n
u
te
u
n
its
(
k
ey
wo
r
d
s
)
th
at
ca
n
ea
s
ily
r
ec
o
g
n
ize
th
e
tex
t a
ttack
s
ac
cu
r
ately
.
An
ex
am
p
le
o
f
th
e
KS
p
r
o
ce
s
s
is
co
n
q
u
e
r
ed
b
el
o
w
in
T
ab
le
1
.
T
ab
le
1
.
A
f
ew
e
x
am
p
les
o
f
th
e
KS p
r
o
ce
s
s
I
n
p
u
t
K
S
p
r
o
c
e
ss
O
n
a
mi
ss
i
o
n
t
o
f
i
n
d
s
o
me
z
e
b
r
a
c
a
k
e
s
O
n
a
mi
ss
i
o
n
,
mi
s
si
o
n
t
o
f
i
n
d
,
t
o
f
i
n
d
,
so
m
e
z
e
b
r
a
,
z
e
b
r
a
c
a
k
e
s
Th
i
s
b
i
t
c
h
h
a
d
h
o
r
ser
a
d
i
s
h
p
o
n
y
t
a
i
l
t
o
d
a
y
s
h
e
.
d
y
e
d
h
e
r
b
a
l
d
h
e
a
d
a
s
s
h
a
i
r
r
e
d
a
n
d
p
u
t
t
h
a
t
b
i
t
c
h
i
n
a
p
o
n
y
t
a
i
l
s
mh
Th
i
s
b
i
t
c
h
h
a
d
,
h
a
d
h
o
r
sera
d
i
s
h
p
o
n
y
t
a
i
l
,
p
o
n
y
t
a
i
l
s
h
e
.
d
y
e
d
h
e
r
,
b
a
l
d
h
e
a
d
a
s
s
h
a
i
r
r
e
d
,
a
n
d
p
u
t
t
h
a
t
,
b
i
t
c
h
i
n
a
p
o
n
y
t
a
i
l
sm
h
3
.
2
.
T
ex
t
cl
a
s
s
if
ica
t
io
n
a
t
t
a
c
k
s
o
n diff
er
ent
NL
P
m
o
dels
T
h
e
s
elec
ted
k
e
y
wo
r
d
s
ar
e
t
h
en
f
e
d
in
t
o
th
e
d
if
f
er
en
t
NL
P
m
o
d
els
lik
e
B
E
R
T
,
R
OB
E
R
T
a,
an
d
XL
Net
m
o
d
els
to
an
aly
ze
th
eir
p
er
f
o
r
m
an
ce
o
n
th
e
tex
t
class
if
icatio
n
attac
k
p
r
o
ce
s
s
.
T
h
e
d
etailed
an
aly
s
is
o
f
th
e
d
if
f
er
e
n
t N
L
P m
o
d
els is
d
ep
icted
b
elo
w.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
TextB
u
g
g
er:
a
n
ex
te
n
d
ed
a
d
ve
r
s
a
r
ia
l
text
a
tta
ck
on
…
(
S
a
n
j
a
ika
n
th
E
.
V
a
d
a
kk
eth
il S
o
m
a
n
a
th
a
n
P
illa
i
)
1739
3
.
2
.
1
.
B
E
RT
-
ba
s
ed
NL
P
m
o
del
I
n
th
e
B
E
R
T
tech
n
iq
u
e,
t
h
e
s
y
n
o
n
y
m
s
o
f
th
e
wo
r
d
i
n
a
g
i
v
en
s
en
ten
ce
ar
e
m
a
n
ip
u
lated
b
ased
o
n
o
th
er
wo
r
d
s
ad
jace
n
t
to
it.
T
h
e
B
E
R
T
m
o
d
el
p
r
o
v
id
es
all
th
e
in
p
u
t
in
a
s
in
g
le
d
u
r
atio
n
to
s
o
lv
e
th
e
lo
n
g
-
ter
m
d
ep
en
d
e
n
cies
b
etwe
en
wo
r
d
s
an
d
it
is
o
f
two
ty
p
es:
B
E
R
T
b
ase
an
d
B
E
R
T
lar
g
e
m
o
d
el.
I
n
th
e
B
E
R
T
b
ase
tech
n
iq
u
e,
a
to
tal
o
f
twelv
e
t
r
an
s
f
o
r
m
er
en
c
o
d
er
s
ar
e
p
r
es
en
t
f
o
r
t
h
e
tr
ain
i
n
g
p
r
o
ce
s
s
.
I
n
th
e
B
E
R
T
lar
g
e
tech
n
iq
u
e,
a
to
tal
o
f
twen
ty
-
f
o
u
r
tr
an
s
f
o
r
m
er
en
c
o
d
er
s
ar
e
p
r
esen
t.
Her
e,
t
h
e
tu
n
in
g
p
r
o
c
ess
i
s
v
er
y
ea
s
y
an
d
p
r
o
v
id
es
o
u
ts
tan
d
in
g
class
if
icatio
n
p
e
r
f
o
r
m
an
ce
.
T
h
e
f
o
llo
win
g
s
tep
s
ar
e
p
er
f
o
r
m
e
d
in
t
h
e
B
E
R
T
-
b
ased
NL
P
m
o
d
el
f
o
r
th
e
tex
t c
lass
if
icatio
n
attac
k
p
r
o
ce
s
s
:
−
Sep
ar
ate
th
e
co
llected
tex
t
d
ata
b
ased
o
n
t
r
ain
in
g
a
n
d
test
in
g
s
et
s
u
s
in
g
th
e
tr
ain
-
test
s
p
lit p
r
o
ce
s
s
.
−
T
r
an
s
f
o
r
m
t
h
e
tr
ain
in
g
s
et
b
ased
o
n
c
o
r
r
esp
o
n
d
in
g
Py
th
o
n
te
n
s
o
r
s
f
o
r
th
e
NL
P tec
h
n
iq
u
e.
−
Dete
r
m
in
e
th
e
b
atch
s
ize
to
g
e
n
er
ate
ten
s
o
r
s
r
ep
ea
ted
l
y
to
en
h
an
ce
th
e
B
E
R
T
tech
n
iq
u
e
.
−
T
r
ain
th
e
B
E
R
T
u
s
in
g
th
e
n
et
wo
r
k
p
ar
am
eter
s
an
d
a
n
aly
ze
th
e
s
u
cc
ess
r
ate
p
er
f
o
r
m
an
ce
.
T
h
e
o
u
tco
m
e
o
f
B
E
R
T
-
NL
P m
o
d
el
is
d
ep
icted
in
T
ab
le
2
.
Fig
u
r
e
2
in
d
icate
s
th
e
ar
ch
itectu
r
e
o
f
B
E
R
T
m
o
d
el.
T
ab
le
2
.
Ad
v
er
s
ar
ial
tex
t o
u
tc
o
m
e
f
r
o
m
th
e
B
E
R
T
-
NL
P m
o
d
el
O
r
i
g
i
n
a
l
i
n
p
u
t
A
d
v
e
r
sari
a
l
o
u
t
c
o
m
e
M
o
d
e
r
n
d
a
y
s
i
n
g
e
r
s t
a
l
k
a
b
o
u
t
t
h
e
s
a
me
s
h
i
t
r
a
p
p
e
r
s
t
a
l
k
a
b
o
u
t
l
o
l
.
.
.
.
h
o
e
s
M
o
d
e
r
n
d
a
y
s
i
n
g
e
r
s t
a
l
k
a
b
o
u
t
t
h
e
s
a
me
si
t
r
a
p
p
e
r
s
t
a
l
k
a
b
o
u
t
l
o
l
.
.
.
.
h
K
e
s
Y
o
u
h
a
d
t
o
t
h
r
o
w
i
n
t
h
e
f
a
g
g
o
t
w
o
r
d
s
mh
sm
h
s
mh
Y
o
u
h
a
d
t
o
t
o
ss
i
n
t
t
e
f
a
o
g
g
t
w
o
d
s
n
h
sn
h
sn
h
Fig
u
r
e
2
.
Ar
c
h
itectu
r
e
o
f
B
E
R
T
m
o
d
el
3
.
2
.
2
.
RO
B
E
RT
a
-
ba
s
e
d NLP
m
o
del
T
h
e
R
OB
E
R
T
a
tech
n
iq
u
e
is
th
e
im
p
r
o
v
ed
v
er
s
io
n
o
f
t
h
e
B
E
R
T
s
ch
em
e
an
d
aid
s
in
s
o
lv
in
g
lo
n
g
-
ter
m
d
ep
e
n
d
en
c
y
p
r
o
b
lem
s
d
u
r
in
g
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
As
lik
e
B
E
R
T
m
o
d
el,
th
e
R
OB
E
R
T
a
m
o
d
el
also
u
s
es
th
e
tr
an
s
f
o
r
m
e
r
s
th
ta
co
n
s
is
t
o
f
th
r
ee
elem
en
ts
:
h
ea
d
s
,
tr
an
s
f
o
r
m
er
s
,
an
d
to
k
e
n
izer
.
T
h
e
tr
a
n
s
f
o
r
m
er
s
co
n
v
er
t
th
e
s
p
ar
s
e
d
ata
in
to
co
n
tex
t
u
al
em
b
ed
d
in
g
’
s
f
o
r
d
ep
th
-
le
v
el
tr
ain
in
g
.
T
h
e
h
ea
d
co
v
er
s
th
e
tr
an
s
f
o
r
m
e
r
th
at
ass
is
ts
th
e
co
n
tex
tu
al
em
b
e
d
d
i
n
g
f
o
r
u
p
c
o
m
i
n
g
tr
ain
i
n
g
p
r
o
c
ess
.
T
h
e
to
k
en
izer
ass
is
ts
in
alter
in
g
o
r
ig
in
al
tex
t
in
to
in
d
ex
s
p
ar
s
e
en
co
d
i
n
g
s
.
T
h
e
R
OB
E
R
T
a
u
s
es
th
e
b
y
te
-
p
air
ch
ar
ac
ter
-
lev
el
en
co
d
in
g
s
ca
p
ab
le
o
f
tr
ain
in
g
lar
g
er
tex
t
d
ata
o
v
e
r
5
0
,
0
0
0
s
u
b
s
et
u
n
its
.
Ap
ar
t
f
r
o
m
th
is
,
t
h
e
R
OB
E
R
T
a
m
o
d
el
f
in
e
-
tu
n
es
m
o
r
e
ef
f
ec
tiv
ely
co
m
p
ar
ed
to
B
E
R
T
m
o
d
els.
T
h
e
f
o
llo
win
g
s
tep
s
ar
e
p
er
f
o
r
m
ed
i
n
th
e
R
OB
E
R
T
a
-
b
ased
NL
P
m
o
d
el
f
o
r
th
e
tex
t c
lass
if
icatio
n
attac
k
p
r
o
ce
s
s
:
−
I
n
itially
,
th
e
ac
tu
al
tex
t
d
ata
is
to
k
en
ized
in
to
s
u
b
-
wo
r
d
s
s
o
th
e
wo
r
d
em
b
ed
d
in
g
ar
e
en
co
d
ed
ea
s
ily
.
A
s
p
ec
ialized
to
k
en
s
u
ch
as
<
s
>
an
d
</s>
to
r
ep
r
esen
t
t
h
e
s
tar
tin
g
an
d
en
d
in
g
wo
r
d
s
eq
u
en
ce
.
Mo
r
e
o
v
er
,
<p
ad
>
to
k
e
n
ass
is
tin
g
tex
t p
ad
d
in
g
to
i
n
cr
ea
s
e
th
e
len
g
t
h
o
f
wo
r
d
v
ec
to
r
.
−
Fo
r
tex
t
lear
n
in
g
,
th
e
wo
r
d
s
ar
e
co
n
v
er
ted
in
to
u
s
ef
u
l
n
u
m
er
ical
in
ter
p
r
e
tatio
n
.
T
h
e
to
k
en
i
ze
r
en
c
o
d
es
th
e
ac
tu
al
tex
t
in
to
an
atten
tio
n
m
ask
(
d
elib
er
ates
th
e
p
r
esen
ts
a
n
d
ab
s
en
ce
o
f
to
k
en
s
f
o
r
th
e
t
r
ain
in
g
p
r
o
ce
s
s
)
an
d
tex
t I
Ds (
co
n
tain
s
to
k
en
in
d
ex
a
n
d
to
k
en
n
u
m
er
ical
in
te
r
p
r
etatio
n
)
.
−
T
h
e
tex
t
I
Ds
an
d
atten
tio
n
m
ask
s
ar
e
th
en
f
ed
in
to
th
e
R
OB
E
R
T
a
s
ch
em
e
th
at
co
n
s
is
t
o
f
1
2
b
ase
lay
er
s
,
m
o
r
e
th
an
1
2
0
m
illi
o
n
p
a
r
a
m
eter
s
an
d
7
6
8
h
i
d
d
en
v
ec
to
r
s
th
at
cr
ea
tes
u
s
ef
u
l
wo
r
d
em
b
ed
d
in
g
as
th
e
f
ea
tu
r
e
en
g
in
ee
r
in
g
.
T
h
e
o
u
tco
m
e
o
f
R
OB
E
R
T
a
-
NL
P m
o
d
el
is
d
ep
icted
in
T
ab
le
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
7
3
5
-
1
7
4
4
1740
T
ab
le
3
.
Ad
v
er
s
ar
ial
tex
t o
u
tc
o
m
e
f
r
o
m
th
e
R
OB
E
R
T
-
NL
P
m
o
d
el
O
r
i
g
i
n
a
l
i
n
p
u
t
A
d
v
e
r
sari
a
l
o
u
t
c
o
m
e
M
o
d
e
r
n
d
a
y
s
i
n
g
e
r
s t
a
l
k
a
b
o
u
t
t
h
e
s
a
me
s
h
i
t
r
a
p
p
e
r
s
t
a
l
k
a
b
o
u
t
l
o
l
.
.
.
.
h
o
e
s
M
o
d
e
r
n
d
a
y
s
i
n
g
e
r
s t
a
l
k
a
b
o
u
t
t
h
e
s
a
me
si
h
t
r
a
p
p
e
r
s
t
a
l
k
a
b
o
u
t
l
o
l
.
.
.
.
h
o
o
k
e
r
s
W
t
f
w
a
s
d
r
a
k
e
a
s
k
i
n
g
u
s t
o
p
u
l
l
o
v
e
r
s
o
h
e
c
a
n
g
e
t
my
a
u
t
o
g
r
a
p
h
b
i
t
c
h
W
t
f
w
a
s
d
r
a
k
e
a
s
k
i
n
g
u
s t
o
p
u
l
l
o
v
e
r
s
o
h
e
c
a
n
g
e
t
m
y
a
u
t
o
g
r
a
p
h
b
i
t
h
c
3
.
2
.
3
.
XL
Net
-
ba
s
ed
NL
P
m
o
del
T
h
e
XL
Net
u
tili
ze
s
th
e
p
r
o
p
er
ty
o
f
p
e
r
m
u
tatio
n
lan
g
u
ag
e
m
o
d
el
(
PLM
)
t
o
in
teg
r
ate
t
h
e
p
r
o
s
o
f
au
to
r
eg
r
ess
iv
e
(
AR
)
,
an
d
a
u
t
o
en
co
d
er
(
AE
)
.
T
h
e
AR
m
o
d
el
ac
ts
as
a
d
ec
o
d
er
o
f
tr
an
s
f
o
r
m
er
an
d
p
r
o
ce
s
s
th
e
p
r
esen
t
d
ata
to
class
if
y
th
e
co
r
r
esp
o
n
d
i
n
g
o
u
tc
o
m
e.
I
n
th
e
AE
tech
n
iq
u
e,
B
E
R
T
m
o
d
el
is
u
tili
ze
d
wh
er
e
th
e
p
ar
ticu
lar
wo
r
d
s
o
f
t
h
e
in
p
u
t
tex
t
ar
e
m
ask
ed
an
d
th
e
o
u
tco
m
e
is
r
etain
ed
.
T
h
e
to
k
en
s
ar
e
ar
r
an
g
ed
d
y
n
am
ically
i
n
PLM
in
a
s
en
ten
ce
f
o
r
m
at
an
d
u
tili
ze
AE
to
d
etec
t
f
i
n
al
f
ew
to
k
en
s
.
W
h
ile
d
etec
tin
g
th
e
to
k
en
,
d
u
al
to
k
e
n
in
f
o
r
m
atio
n
is
u
tili
ze
d
an
d
u
n
d
er
s
tan
d
th
e
d
ep
en
d
e
n
cy
am
o
n
g
th
e
to
k
e
n
s
.
Mo
r
eo
v
er
,
XL
Net
im
p
lem
en
t
s
th
e
r
ec
u
r
s
iv
e
m
ec
h
an
is
m
a
n
d
au
th
o
r
ized
p
o
s
itio
n
e
n
co
d
in
g
in
th
e
tr
an
s
f
o
r
m
er
.
XL
Net
s
to
r
e
th
e
h
id
d
e
n
u
n
it
s
eq
u
en
ce
d
u
r
i
n
g
ev
er
y
p
e
r
m
u
tatio
n
an
d
th
e
a
u
th
o
r
ized
p
o
s
itio
n
en
co
d
in
g
is
b
ala
n
ce
d
b
etwe
e
n
v
a
r
io
u
s
p
er
m
u
tatio
n
s
.
Du
e
to
t
h
e
u
s
e
o
f
tr
an
s
f
o
r
m
er
s
,
it
ca
n
e
n
h
an
ce
th
e
ex
tr
ac
ted
f
ea
tu
r
es
b
y
u
tili
zin
g
th
e
p
r
o
s
o
f
NL
P
o
v
er
lar
g
e
r
t
ex
ts
.
B
ec
au
s
e
o
f
th
e
af
o
r
em
e
n
tio
n
ed
p
r
o
p
er
ty
o
f
XL
Net,
it c
an
co
m
p
letel
y
in
d
i
ca
te
ev
er
y
to
k
en
b
ased
o
n
s
em
an
tic
r
ep
r
esen
tatio
n
in
T
ab
le
4
.
T
ab
le
4
.
Ad
v
er
s
ar
ial
tex
t o
u
tc
o
m
e
f
r
o
m
th
e
XL
Net
-
NL
P m
o
d
el
O
r
i
g
i
n
a
l
i
n
p
u
t
A
d
v
e
r
sari
a
l
o
u
t
c
o
m
e
M
o
d
e
r
n
d
a
y
s
i
n
g
e
r
s t
a
l
k
a
b
o
u
t
t
h
e
s
a
me
sh
i
t
r
a
p
p
e
r
s
t
a
l
k
a
b
o
u
t
l
o
l
.
.
.
.
h
o
e
s
M
o
d
e
r
n
d
a
y
s
i
n
g
e
r
s t
a
l
k
a
b
o
u
t
t
h
e
s
a
me
p
o
o
p
r
a
p
p
e
r
s
t
a
l
k
a
b
o
u
t
l
o
l
.
.
.
.
d
u
c
k
l
i
n
g
s
W
t
f
w
a
s
d
r
a
k
e
a
s
k
i
n
g
u
s t
o
p
u
l
l
o
v
e
r
s
o
h
e
c
a
n
g
e
t
my
a
u
t
o
g
r
a
p
h
b
i
t
c
h
W
t
f
w
a
s
d
r
a
k
e
a
s
k
i
n
g
u
s t
o
p
u
l
l
o
v
e
r
s
o
h
e
c
a
n
e
g
t
m
y
a
u
t
o
g
r
a
p
h
b
i
c
t
h
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
NS
T
h
e
d
ev
elo
p
e
d
m
eth
o
d
is
p
r
o
ce
s
s
ed
an
d
an
aly
ze
d
v
ia
th
e
Py
th
o
n
p
latf
o
r
m
.
Fo
r
th
e
s
im
u
latio
n
p
r
o
ce
s
s
,
a
tex
t
class
if
icatio
n
a
ttack
b
en
ch
m
ar
k
d
atab
ase
(
T
C
AB
)
[
2
5
]
is
u
tili
ze
d
wh
ich
c
o
n
s
is
ts
o
f
d
if
f
er
e
n
t
ad
v
er
s
ar
ial
attac
k
s
o
n
tr
a
d
itio
n
al
tex
t
class
if
icatio
n
m
o
d
els
t
r
ain
ed
o
n
v
ar
io
u
s
s
en
tim
en
ts
a
n
d
a
b
u
s
iv
e
d
o
m
ain
co
n
ten
ts
.
I
n
th
e
tr
ain
in
g
p
ar
t,
5
5
2
,
3
6
4
s
am
p
les
ar
e
co
n
s
id
er
ed
clea
n
,
an
d
th
e
r
em
ai
n
in
g
a
s
u
n
p
er
tu
r
b
e
d
d
ata.
Fo
r
th
e
test
in
g
p
r
o
ce
s
s
,
1
7
8
,
6
0
7
s
am
p
les
ar
e
co
n
s
id
er
e
d
clea
n
,
an
d
th
e
r
em
ai
n
in
g
as
u
n
p
er
tu
r
b
ed
tex
ts
.
Var
io
u
s
p
er
f
o
r
m
an
ce
an
aly
s
e
s
lik
e
ac
cu
r
ac
y
,
KC
,
F
-
m
ea
s
u
r
e,
an
d
PP
V
ar
e
co
m
p
u
ted
an
d
co
m
p
a
r
ed
with
d
if
f
er
en
t N
L
P m
o
d
els.
4
.
1
.
Ass
ess
m
ent
m
et
rics
_
=
+
+
+
+
(
1
)
1
−
=
2
×
(
×
+
)
(
2
)
=
2
×
(
×
−
×
)
(
+
)
(
+
)
+
(
+
)
(
+
)
(
3
)
(
%
)
=
+
(
4)
Her
e,
,
,
,
in
d
icate
s
th
e
tr
u
e
n
eg
ativ
e
(
T
N)
,
tr
u
e
p
o
s
itiv
e
(
T
P),
f
alse
n
eg
ativ
e
(
FN)
,
an
d
f
alse
p
o
s
itiv
e
(
FP
)
r
esp
ec
tiv
ely
.
4
.
2
.
Co
m
pa
ra
t
iv
e
a
na
ly
s
is
o
f
dev
elo
ped m
et
ho
d o
v
er
co
n
v
ent
io
na
l t
ec
hn
iqu
es
I
n
th
is
s
ec
tio
n
,
th
e
o
u
tco
m
es
ac
h
iev
ed
b
y
v
ar
io
u
s
NL
P
m
o
d
els
in
p
r
o
d
u
cin
g
h
o
s
tile
tex
ts
ar
e
an
aly
ze
d
b
y
ass
ess
in
g
s
u
cc
ess
r
ate,
F
-
m
ea
s
u
r
e,
KC
,
tim
e
co
n
s
u
m
p
tio
n
,
a
n
d
PP
V
m
etr
ics.
T
h
e
d
etailed
an
aly
s
is
o
f
th
e
o
b
tain
ed
o
u
tco
m
es
is
co
n
q
u
e
r
ed
b
elo
w.
Fig
u
r
e
3
d
ep
icts
th
e
o
v
er
all
ac
c
u
r
a
cy
an
d
lo
s
s
an
aly
s
is
o
f
th
e
NL
P
m
o
d
els.
T
h
e
NL
P
m
o
d
els
lik
e
B
E
R
T
,
R
OB
E
R
T
a,
an
d
XL
Net
m
o
d
els
ar
e
tr
ain
ed
an
d
test
ed
f
o
r
class
if
y
in
g
h
o
s
tile
at
tack
s
o
n
in
p
u
t
tex
ts
.
Fr
o
m
th
e
g
r
ap
h
ic
al
in
ter
p
r
etatio
n
,
it
is
n
o
ted
t
h
at
th
e
NL
P
m
o
d
els
o
u
tp
er
f
o
r
m
well
b
y
m
in
im
izin
g
lo
s
s
es
d
u
r
in
g
th
e
tr
ain
in
g
,
an
d
test
in
g
p
r
o
ce
s
s
.
T
ab
le
5
tab
u
lates
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
TextB
u
g
g
er:
a
n
ex
te
n
d
ed
a
d
ve
r
s
a
r
ia
l
text
a
tta
ck
on
…
(
S
a
n
j
a
ika
n
th
E
.
V
a
d
a
kk
eth
il S
o
m
a
n
a
th
a
n
P
illa
i
)
1741
co
m
p
ar
ativ
e
an
aly
s
is
o
f
d
if
f
er
en
t
NL
P
m
o
d
els.
W
h
ile
an
aly
zin
g
th
e
p
er
f
o
r
m
an
ce
o
f
d
i
f
f
e
r
en
t
NL
P
m
o
d
els,
R
OB
E
R
T
a
m
o
d
el
o
u
tp
er
f
o
r
m
s
b
etter
in
ter
m
s
o
f
s
u
cc
ess
r
ate,
an
d
tim
e
c
o
n
s
u
m
p
tio
n
.
Fig
u
r
e
3
.
Ov
e
r
all
ac
cu
r
ac
y
an
d
lo
s
s
an
aly
s
is
T
ab
le
5
.
C
o
m
p
a
r
ativ
e
an
aly
s
is
o
f
d
if
f
er
en
t N
L
P m
o
d
els
M
e
t
h
o
d
s
S
u
c
c
e
ss ra
t
e
(
%)
F
-
mea
su
r
e
(
%)
K
C
(
%)
P
P
V
(
%)
Ti
me
c
o
n
s
u
m
p
t
i
o
n
(
s)
R
O
B
E
R
Ta
9
9
.
7
9
9
.
6
5
9
8
.
9
0
9
9
.
6
8
1
0
6
.
2
8
B
ER
T
9
8
.
6
9
8
.
5
9
9
7
.
8
7
9
8
.
6
2
1
8
4
.
0
8
X
LN
e
t
9
6
.
8
9
6
.
6
6
9
5
.
4
5
9
5
.
9
2
7
6
9
1
.
0
1
8
4
.
3
.
P
r
a
ct
ica
l i
m
pa
ct
s
o
f
T
e
x
t
B
ug
g
er
An
ex
ten
d
e
d
ad
v
er
s
ar
ial
tex
t
attac
k
o
n
NL
P
-
b
ased
tex
t
class
if
icatio
n
m
o
d
el
s
ar
e
s
u
b
s
t
an
tial
an
d
m
u
ltifa
ce
ted
.
Firstl
y
,
T
ex
tB
u
g
g
er
h
i
g
h
lig
h
ts
th
e
v
u
ln
er
a
b
ilit
ies
o
f
NL
P
m
o
d
els
to
ad
v
er
s
ar
ial
attac
k
s
,
s
ig
n
if
ican
tly
r
aisi
n
g
awa
r
en
es
s
ab
o
u
t
th
e
n
ee
d
f
o
r
en
h
an
ce
d
s
ec
u
r
ity
m
ea
s
u
r
es.
T
h
is
n
e
wf
o
u
n
d
awa
r
en
ess
d
r
iv
es
r
esear
ch
e
r
s
an
d
p
r
ac
tit
io
n
er
s
to
ad
d
r
ess
th
ese
wea
k
n
ess
es
an
d
d
ev
el
o
p
m
o
r
e
r
o
b
u
s
t
m
o
d
els
th
at
ca
n
r
esis
t
s
u
ch
m
an
ip
u
lativ
e
in
p
u
ts
.
T
ex
tB
u
g
g
er
also
p
r
esen
ts
ch
allen
g
es
r
elate
d
to
m
ai
n
tain
in
g
s
em
an
ti
c
in
teg
r
ity
in
t
h
e
f
ac
e
o
f
a
d
v
er
s
ar
ial
ex
am
p
les.
T
h
e
attac
k
’
s
a
b
ilit
y
to
g
e
n
er
ate
tex
t
m
o
d
if
ic
atio
n
s
th
at
r
em
ain
s
em
an
tically
s
im
ilar
to
th
e
o
r
i
g
in
al
co
n
ten
t
h
ig
h
lig
h
ts
th
e
n
e
ed
f
o
r
m
eth
o
d
s
th
at
ca
n
d
etec
t
an
d
m
itig
ate
s
u
c
h
s
u
b
tle
m
an
ip
u
latio
n
s
with
o
u
t
co
m
p
r
o
m
is
in
g
th
e
m
o
d
el
’
s
p
er
f
o
r
m
an
ce
o
r
u
n
d
e
r
s
tan
d
in
g
.
Fin
ally
,
th
e
ex
p
lo
r
atio
n
o
f
a
d
v
er
s
ar
ial
tex
t
attac
k
s
b
y
T
ex
tB
u
g
g
er
m
ay
d
r
iv
e
cr
o
s
s
-
d
is
cip
lin
ar
y
r
esear
ch
ef
f
o
r
ts
.
B
y
in
te
g
r
ati
n
g
in
s
ig
h
ts
f
r
o
m
NL
P,
cy
b
er
s
ec
u
r
ity
,
an
d
ar
tific
ial
in
telli
g
en
ce
,
it
f
o
s
ter
s
a
co
m
p
r
eh
e
n
s
iv
e
ap
p
r
o
ac
h
to
d
ev
elo
p
in
g
s
o
lu
tio
n
s
th
at
e
n
h
an
ce
th
e
o
v
er
all
s
ec
u
r
ity
f
r
am
ewo
r
k
f
o
r
tex
t
class
if
icatio
n
s
y
s
tem
s
.
T
h
is
i
n
ter
d
is
cip
lin
ar
y
c
o
llab
o
r
atio
n
ca
n
lead
to
m
o
r
e
ef
f
ec
tiv
e
a
n
d
r
esil
ien
t
s
ec
u
r
ity
m
ea
s
u
r
es
in
NL
P
ap
p
licatio
n
s
.
Ov
er
a
ll,
T
ex
tB
u
g
g
er
’
s
p
r
ac
ti
ca
l
im
p
ac
ts
ar
e
s
ig
n
if
ican
t,
le
ad
in
g
to
im
p
r
o
v
ed
m
o
d
el
s
ec
u
r
ity
,
r
ef
in
ed
ev
al
u
atio
n
m
etr
ics,
b
etter
m
o
d
el
d
esig
n
,
an
d
en
h
an
ce
d
d
ef
e
n
s
es
in
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
.
I
ts
f
in
d
in
g
s
d
r
iv
e
ad
v
an
ce
m
en
ts
in
cr
ea
tin
g
r
o
b
u
s
t
an
d
s
ec
u
r
e
NL
P
s
y
s
te
m
s
,
ad
d
r
ess
in
g
k
ey
ch
allen
g
es a
n
d
f
o
s
ter
in
g
cr
o
s
s
-
d
is
cip
lin
ar
y
r
esear
ch
.
5.
CO
NCLU
SI
O
N
T
h
e
d
ev
el
o
p
ed
m
eth
o
d
in
v
est
ig
ated
v
ar
io
u
s
ex
is
tin
g
NL
P
m
o
d
els
to
class
if
y
ad
v
er
s
ar
ial
tex
ts
o
n
o
r
ig
in
al
e
x
am
p
les.
C
o
m
m
o
n
tex
tB
u
g
g
er
s
lik
e
B
E
R
T
,
R
OB
E
R
T
a,
an
d
XL
Net
m
o
d
e
l
s
ar
e
an
aly
ze
d
b
y
in
p
u
ttin
g
ac
t
u
al
tex
ts
f
o
r
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
T
h
e
ex
ten
s
iv
e
s
im
u
latio
n
is
ca
r
r
ied
o
u
t
i
n
th
e
p
u
b
licly
av
ailab
le
T
C
AB
d
ataset
to
an
aly
ze
th
e
s
e
m
o
d
els.
T
h
e
o
u
tco
m
es
o
f
th
is
s
im
u
latio
n
p
r
o
v
ed
th
at
th
e
R
O
B
E
R
T
a
-
b
ased
tex
tb
u
g
g
e
r
m
o
d
el
is
h
ig
h
ly
ef
f
ec
tiv
e
an
d
f
ast.
T
o
p
r
o
v
e
th
e
r
o
b
u
s
tn
ess
o
f
d
ev
elo
p
ed
s
ch
em
e,
o
th
er
ex
is
tin
g
ap
p
r
o
ac
h
es
ar
e
also
ex
p
er
im
e
n
ted
with
in
ter
m
s
o
f
s
u
cc
ess
r
ate,
tim
e
co
n
s
u
m
p
tio
n
,
PP
V,
F
-
m
ea
s
u
r
e,
an
d
KC
.
T
h
e
s
im
u
latio
n
p
r
o
ce
s
s
is
ca
r
r
ied
o
u
t
in
th
e
Py
th
o
n
p
latf
o
r
m
an
d
t
h
e
o
v
er
all
s
u
cc
ess
r
ate
ac
h
iev
ed
b
y
B
E
R
T
,
R
OB
E
R
T
a,
an
d
XL
Net
is
ab
o
u
t
9
8
.
6
%,
9
9
.
7
%,
an
d
9
6
.
8
%
r
esp
ec
tiv
ely
.
Ho
wev
er
,
th
e
d
ev
elo
p
ed
s
ch
em
e
f
ailed
to
co
n
s
id
er
o
t
h
er
DL
m
o
d
els
lik
e
B
i
-
L
STM
,
C
NN,
an
d
L
STM
m
o
d
els
to
class
i
f
y
ad
v
er
s
ar
ial
tex
ts
b
ased
o
n
in
p
u
t
tex
ts
.
I
n
f
u
tu
r
e
s
tu
d
ies,
o
th
er
DL
m
o
d
els
ar
e
also
co
n
s
id
er
ed
an
d
th
eir
p
er
f
o
r
m
a
n
ce
will
b
e
an
aly
ze
d
b
y
in
p
u
ttin
g
v
a
r
io
u
s
tex
t
ex
am
p
les.
An
ex
ten
d
ed
ad
v
er
s
ar
ial
tex
t
attac
k
o
n
NL
P
-
b
ased
tex
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
7
3
5
-
1
7
4
4
1742
class
if
icatio
n
m
o
d
el
s
s
h
o
u
ld
f
o
cu
s
o
n
s
ev
er
al
k
e
y
ar
ea
s
.
Ad
v
an
cin
g
a
ttack
s
tr
ateg
ies
to
ex
p
lo
it
d
ee
p
er
NL
P
m
o
d
el
f
ea
tu
r
es,
d
ev
elo
p
in
g
m
o
r
e
ef
f
ec
tiv
e
an
d
tailo
r
ed
d
e
f
e
n
s
e
m
ec
h
an
is
m
s
,
an
d
u
n
d
e
r
s
tan
d
in
g
t
h
e
im
p
ac
t
o
f
ad
v
er
s
ar
ial
attac
k
s
o
n
cr
o
s
s
-
l
in
g
u
al
an
d
m
u
lti
-
task
m
o
d
els
ar
e
cr
u
cial.
R
esear
ch
s
h
o
u
l
d
also
ex
p
l
o
r
e
th
e
tr
an
s
f
er
ab
ilit
y
o
f
ad
v
er
s
ar
ial
ex
am
p
les
ac
r
o
s
s
d
if
f
er
en
t
ar
c
h
itectu
r
es,
ex
am
in
e
h
o
w
th
ese
attac
k
s
af
f
ec
t
tex
t
r
ea
d
ab
ilit
y
an
d
s
em
an
tic
in
teg
r
ity
,
an
d
ass
ess
th
eir
r
ea
l
-
wo
r
ld
ap
p
licab
ilit
y
in
s
y
s
tem
s
lik
e
au
to
m
ated
co
n
ten
t
m
o
d
er
atio
n
an
d
s
en
tim
e
n
t a
n
a
ly
s
is
.
AC
K
NO
WL
E
DG
E
M
E
NT
S
T
h
e
au
th
o
r
with
a
d
ee
p
s
en
s
e
o
f
g
r
atitu
d
e
wo
u
l
d
th
an
k
th
e
s
u
p
er
v
is
o
r
f
o
r
h
is
g
u
i
d
an
ce
a
n
d
co
n
s
tan
t
s
u
p
p
o
r
t r
e
n
d
er
e
d
d
u
r
in
g
t
h
is
r
esear
ch
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
No
f
u
n
d
in
g
in
v
o
lv
e
d
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
San
jaik
an
th
E
.
Vad
a
k
k
eth
il
So
m
an
ath
an
Pil
lai
✓
✓
✓
✓
✓
✓
✓
Srin
iv
as A.
Vad
d
ad
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
R
o
h
ith
Vallab
h
an
en
i
✓
✓
✓
✓
✓
✓
✓
✓
San
to
s
h
R
ed
d
y
Ad
d
u
la
✓
✓
✓
✓
✓
✓
✓
✓
B
h
u
v
an
esh
An
an
th
a
n
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
t
h
at
s
u
p
p
o
r
t
th
e
f
i
n
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
a
ilab
le
f
r
o
m
th
e
c
o
r
r
esp
o
n
d
in
g
au
th
o
r
,
[
S.E
.
V.
S.P]
,
u
p
o
n
r
ea
s
o
n
a
b
le
r
eq
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
E.
B
.
-
A
s
t
u
d
i
l
l
o
,
W
.
F
u
e
r
t
e
s,
S
.
S
.
-
G
o
r
d
o
n
,
D
.
N
.
-
A
g
u
r
t
o
,
a
n
d
G
.
R
.
-
G
a
l
á
n
,
“
A
p
h
i
s
h
i
n
g
-
a
t
t
a
c
k
-
d
e
t
e
c
t
i
o
n
mo
d
e
l
u
s
i
n
g
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ssi
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
(
S
w
i
t
zerl
a
n
d
)
,
v
o
l
.
1
3
,
n
o
.
9
,
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
3
0
9
5
2
7
5
.
[
2
]
B
.
H
e
,
M
.
A
h
a
ma
d
,
a
n
d
S
.
K
u
mar,
“
P
e
t
g
e
n
:
p
e
r
so
n
a
l
i
z
e
d
t
e
x
t
g
e
n
e
r
a
t
i
o
n
a
t
t
a
c
k
o
n
d
e
e
p
s
e
q
u
e
n
c
e
e
m
b
e
d
d
i
n
g
-
b
a
s
e
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
mo
d
e
l
s
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
7
t
h
AC
M
S
I
G
K
D
D
C
o
n
f
e
re
n
c
e
o
n
K
n
o
w
l
e
d
g
e
D
i
sc
o
v
e
ry
&
D
a
t
a
M
i
n
i
n
g
,
A
u
g
.
2
0
2
1
,
p
p
.
5
7
5
–
5
8
4
,
d
o
i
:
1
0
.
1
1
4
5
/
3
4
4
7
5
4
8
.
3
4
6
7
3
9
0
.
[
3
]
I
.
A
l
smad
i
e
t
a
l
.
,
“
A
d
v
e
r
s
a
r
i
a
l
a
t
t
a
c
k
s
a
n
d
d
e
f
e
n
se
s
f
o
r
s
o
c
i
a
l
n
e
t
w
o
r
k
t
e
x
t
p
r
o
c
e
ssi
n
g
a
p
p
l
i
c
a
t
i
o
n
s:
t
e
c
h
n
i
q
u
e
s,
c
h
a
l
l
e
n
g
e
s
a
n
d
f
u
t
u
r
e
r
e
s
e
a
r
c
h
d
i
r
e
c
t
i
o
n
s,”
Arx
i
v
,
2
0
2
1
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
2
1
1
0
.
1
3
9
8
0
.
[
4
]
F
.
M
a
r
u
l
l
i
,
L
.
V
e
r
d
e
,
a
n
d
L
.
C
a
mp
a
n
i
l
e
,
“
Ex
p
l
o
r
i
n
g
d
a
t
a
a
n
d
m
o
d
e
l
p
o
i
so
n
i
n
g
a
t
t
a
c
k
s
t
o
d
e
e
p
l
e
a
r
n
i
n
g
-
b
a
se
d
N
LP
s
y
st
e
ms,
”
Pro
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
1
9
2
,
p
p
.
3
5
7
0
–
3
5
7
9
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.
2
0
2
1
.
0
9
.
1
3
0
.
[
5
]
S
.
U
p
l
e
n
c
h
w
a
r
,
V
.
S
a
w
a
n
t
,
P
.
S
u
r
v
e
,
S
.
D
e
sh
p
a
n
d
e
,
a
n
d
S
.
K
e
l
k
a
r
,
“
P
h
i
s
h
i
n
g
a
t
t
a
c
k
d
e
t
e
c
t
i
o
n
o
n
t
e
x
t
m
e
ssa
g
e
s
u
s
i
n
g
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s,”
i
n
2
0
2
2
I
EEE
P
u
n
e
S
e
c
t
i
o
n
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
(
P
u
n
e
C
o
n
)
,
D
e
c
.
2
0
2
2
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
1
0
9
/
P
u
n
e
C
o
n
5
5
4
1
3
.
2
0
2
2
.
1
0
0
1
4
8
7
6
.
[
6
]
Z
.
Zh
o
u
,
H
.
G
u
a
n
,
M
.
B
h
a
t
,
a
n
d
J
.
H
s
u
,
“
F
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
v
i
a
N
L
P
i
s
v
u
l
n
e
r
a
b
l
e
t
o
a
d
v
e
r
s
a
r
i
a
l
a
t
t
a
c
k
s
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
1
1
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
ren
c
e
o
n
Ag
e
n
t
s
a
n
d
A
rt
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
2
0
1
9
,
v
o
l
.
2
,
p
p
.
7
9
4
–
8
0
0
,
d
o
i
:
1
0
.
5
2
2
0
/
0
0
0
7
5
6
6
3
0
7
9
4
0
8
0
0
.
[
7
]
H
.
A
l
i
e
t
a
l
.
,
“
A
l
l
y
o
u
r
f
a
k
e
d
e
t
e
c
t
o
r
a
r
e
b
e
l
o
n
g
t
o
u
s:
e
v
a
l
u
a
t
i
n
g
a
d
v
e
r
sar
i
a
l
r
o
b
u
st
n
e
ss
o
f
f
a
ke
-
n
e
w
s
d
e
t
e
c
t
o
r
s
u
n
d
e
r
b
l
a
c
k
-
b
o
x
set
t
i
n
g
s
,
”
I
E
EE
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
8
1
6
7
8
–
8
1
6
9
2
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
0
8
5
8
7
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
TextB
u
g
g
er:
a
n
ex
te
n
d
ed
a
d
ve
r
s
a
r
ia
l
text
a
tta
ck
on
…
(
S
a
n
j
a
ika
n
th
E
.
V
a
d
a
kk
eth
il S
o
m
a
n
a
th
a
n
P
illa
i
)
1743
[
8
]
X
.
Li
,
L
.
C
h
e
n
,
a
n
d
D
.
W
u
,
“
T
u
r
n
i
n
g
a
t
t
a
c
k
s
i
n
t
o
p
r
o
t
e
c
t
i
o
n
:
s
o
c
i
a
l
m
e
d
i
a
p
r
i
v
a
c
y
p
r
o
t
e
c
t
i
o
n
u
s
i
n
g
a
d
v
e
r
s
a
r
i
a
l
a
t
t
a
c
k
s
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
2
1
S
I
AM
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
D
a
t
a
Mi
n
i
n
g
(
S
D
M)
,
P
h
i
l
a
d
e
l
p
h
i
a
,
P
A
:
S
o
c
i
e
t
y
f
o
r
I
n
d
u
s
t
r
i
a
l
a
n
d
A
p
p
l
i
e
d
M
a
t
h
e
m
a
t
i
c
s
,
2
0
2
1
,
p
p
.
2
0
8
–
2
1
6
.
[
9
]
J.
R
.
A
s
l
,
M
.
H
.
R
a
f
i
e
i
,
M
.
A
l
o
h
a
l
y
,
a
n
d
D
.
Ta
k
a
b
i
,
“
A
sem
a
n
t
i
c
,
s
y
n
t
a
c
t
i
c
,
a
n
d
c
o
n
t
e
x
t
-
a
w
a
r
e
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
a
d
v
e
r
s
a
r
i
a
l
e
x
a
mp
l
e
g
e
n
e
r
a
t
o
r
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
D
e
p
e
n
d
a
b
l
e
a
n
d
S
e
c
u
re
C
o
m
p
u
t
i
n
g
,
v
o
l
.
2
1
,
n
o
.
5
,
p
p
.
4
7
5
4
–
4
7
6
9
,
S
e
p
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
TD
S
C
.
2
0
2
4
.
3
3
5
9
8
1
7
.
[
1
0
]
Y
.
E.
S
e
y
y
a
r
,
A
.
G
.
Y
a
v
u
z
,
a
n
d
H
.
M
.
U
n
v
e
r
,
“
A
n
a
t
t
a
c
k
d
e
t
e
c
t
i
o
n
f
r
a
mew
o
r
k
b
a
s
e
d
o
n
B
E
R
T
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
6
8
6
3
3
–
6
8
6
4
4
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
1
8
5
7
4
8
.
[
1
1
]
X
.
Li
u
e
t
a
l
.
,
“
S
e
c
u
r
e
c
o
mp
u
t
a
t
i
o
n
p
r
o
t
o
c
o
l
o
f
t
e
x
t
si
mi
l
a
r
i
t
y
a
g
a
i
n
s
t
m
a
l
i
c
i
o
u
s
a
t
t
a
c
k
s
f
o
r
t
e
x
t
c
l
a
ssi
f
i
c
a
t
i
o
n
i
n
d
e
e
p
-
l
e
a
r
n
i
n
g
t
e
c
h
n
o
l
o
g
y
,
”
E
l
e
c
t
r
o
n
i
c
s
,
v
o
l
.
1
2
,
n
o
.
1
6
,
p
.
3
4
9
1
,
A
u
g
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
1
2
1
6
3
4
9
1
.
[
1
2
]
W
.
Z
h
a
n
g
,
Q
.
C
h
e
n
,
a
n
d
Y
.
C
h
e
n
,
“
D
e
e
p
l
e
a
r
n
i
n
g
b
a
se
d
r
o
b
u
s
t
t
e
x
t
c
l
a
ssi
f
i
c
a
t
i
o
n
m
e
t
h
o
d
v
i
a
v
i
r
t
u
a
l
a
d
v
e
r
sari
a
l
t
r
a
i
n
i
n
g
,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
8
,
p
p
.
6
1
1
7
4
–
6
1
1
8
2
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
2
9
8
1
6
1
6
.
[
1
3
]
A
.
B
a
j
a
j
a
n
d
D
.
K
.
V
i
s
h
w
a
k
a
r
ma
,
“
H
O
M
O
C
H
A
R
:
a
n
o
v
e
l
a
d
v
e
r
s
a
r
i
a
l
a
t
t
a
c
k
f
r
a
mew
o
r
k
f
o
r
e
x
p
o
si
n
g
t
h
e
v
u
l
n
e
r
a
b
i
l
i
t
y
o
f
t
e
x
t
b
a
s
e
d
n
e
u
r
a
l
se
n
t
i
m
e
n
t
c
l
a
ssi
f
i
e
r
s,
”
E
n
g
i
n
e
e
r
i
n
g
A
p
p
l
i
c
a
t
i
o
n
s
o
f
A
rt
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
2
6
,
p
.
1
0
6
8
1
5
,
N
o
v
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
a
p
p
a
i
.
2
0
2
3
.
1
0
6
8
1
5
.
[
1
4
]
R
.
B
a
o
,
R
.
Zh
e
n
g
,
L
.
D
i
n
g
,
Q
.
Z
h
a
n
g
,
a
n
d
D
.
Ta
o
,
“
C
A
S
N
:
c
l
a
ss
-
a
w
a
r
e
sc
o
r
e
n
e
t
w
o
r
k
f
o
r
t
e
x
t
u
a
l
a
d
v
e
r
s
a
r
i
a
l
d
e
t
e
c
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
6
1
s
t
An
n
u
a
l
Me
e
t
i
n
g
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
s
t
i
c
s
(
V
o
l
u
m
e
1
:
L
o
n
g
Pa
p
e
rs)
,
2
0
2
3
,
v
o
l
.
1
,
p
p
.
6
7
1
–
6
8
7
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
2
0
2
3
.
a
c
l
-
l
o
n
g
.
4
0
.
[
1
5
]
W
.
M
e
d
h
a
t
,
A
.
H
a
ssa
n
,
a
n
d
H
.
K
o
r
a
sh
y
,
“
S
e
n
t
i
m
e
n
t
a
n
a
l
y
s
i
s
a
l
g
o
r
i
t
h
ms
a
n
d
a
p
p
l
i
c
a
t
i
o
n
s
:
a
su
r
v
e
y
,
”
A
i
n
S
h
a
m
s
En
g
i
n
e
e
ri
n
g
J
o
u
rn
a
l
,
v
o
l
.
5
,
n
o
.
4
,
p
p
.
1
0
9
3
–
1
1
1
3
,
D
e
c
.
2
0
1
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
se
j
.
2
0
1
4
.
0
4
.
0
1
1
.
[
1
6
]
C
.
N
o
b
a
t
a
,
J
.
T
e
t
r
e
a
u
l
t
,
A
.
T
h
o
m
a
s,
Y
.
M
e
h
d
a
d
,
a
n
d
Y
.
C
h
a
n
g
,
“
A
b
u
s
i
v
e
l
a
n
g
u
a
g
e
d
e
t
e
c
t
i
o
n
i
n
o
n
l
i
n
e
u
s
e
r
c
o
n
t
e
n
t
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
W
o
rl
d
W
i
d
e
W
e
b
,
A
p
r
.
2
0
1
6
,
p
p
.
1
4
5
–
1
5
3
,
d
o
i
:
1
0
.
1
1
4
5
/
2
8
7
2
4
2
7
.
2
8
8
3
0
6
2
.
[
1
7
]
M
.
B
a
r
r
e
n
o
,
B
.
N
e
l
s
o
n
,
A
.
D
.
Jo
s
e
p
h
,
a
n
d
J
.
D
.
T
y
g
a
r
,
“
T
h
e
s
e
c
u
r
i
t
y
o
f
mac
h
i
n
e
l
e
a
r
n
i
n
g
,
”
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
v
o
l
.
8
1
,
n
o
.
2
,
p
p
.
1
2
1
–
1
4
8
,
N
o
v
.
2
0
1
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
9
9
4
-
010
-
5
1
8
8
-
5.
[
1
8
]
M
.
B
a
r
r
e
n
o
,
B
.
N
e
l
s
o
n
,
R
.
S
e
a
r
s,
A
.
D
.
Jo
se
p
h
,
a
n
d
J.
D
.
T
y
g
a
r
,
“
C
a
n
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
b
e
s
e
c
u
r
e
?
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
0
6
AC
M
S
y
m
p
o
s
i
u
m
o
n
I
n
f
o
rm
a
t
i
o
n
,
c
o
m
p
u
t
e
r
a
n
d
c
o
m
m
u
n
i
c
a
t
i
o
n
s
sec
u
r
i
t
y
,
M
a
r
.
2
0
0
6
,
v
o
l
.
2
0
0
6
,
p
p
.
1
6
–
2
5
,
d
o
i
:
1
0
.
1
1
4
5
/
1
1
2
8
8
1
7
.
1
1
2
8
8
2
4
.
[
1
9
]
L.
H
u
a
n
g
,
A
.
D
.
J
o
se
p
h
,
B
.
N
e
l
so
n
,
B
.
I
.
P
.
R
u
b
i
n
s
t
e
i
n
,
a
n
d
J
.
D
.
T
y
g
a
r
,
“
A
d
v
e
r
sar
i
a
l
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
4
t
h
A
C
M
w
o
rks
h
o
p
o
n
S
e
c
u
ri
t
y
a
n
d
a
rt
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
,
O
c
t
.
2
0
1
1
,
p
p
.
4
3
–
5
8
,
d
o
i
:
1
0
.
1
1
4
5
/
2
0
4
6
6
8
4
.
2
0
4
6
6
9
2
.
[
2
0
]
B
.
B
i
g
g
i
o
,
G
.
F
u
mer
a
,
a
n
d
F
.
R
o
l
i
,
“
D
e
si
g
n
o
f
r
o
b
u
s
t
c
l
a
ss
i
f
i
e
r
s
f
o
r
a
d
v
e
r
sar
i
a
l
e
n
v
i
r
o
n
m
e
n
t
s,”
i
n
2
0
1
1
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
y
st
e
m
s,
Ma
n
,
a
n
d
C
y
b
e
r
n
e
t
i
c
s
,
O
c
t
.
2
0
1
1
,
p
p
.
9
7
7
–
9
8
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
M
C
.
2
0
1
1
.
6
0
8
3
7
9
6
.
[
2
1
]
D
.
S
c
u
l
l
e
y
,
G
.
M
.
W
a
c
h
ma
n
,
a
n
d
C
.
E.
B
r
o
d
l
e
y
,
“
S
p
a
m
f
i
l
t
e
r
i
n
g
u
si
n
g
i
n
e
x
a
c
t
st
r
i
n
g
ma
t
c
h
i
n
g
i
n
e
x
p
l
i
c
i
t
f
e
a
t
u
r
e
s
p
a
c
e
w
i
t
h
o
n
-
l
i
n
e
l
i
n
e
a
r
c
l
a
ss
i
f
i
e
r
s
,
”
N
I
S
T
S
p
e
c
i
a
l
P
u
b
l
i
c
a
t
i
o
n
,
2
0
0
6
.
[
2
2
]
N
.
C
a
r
l
i
n
i
a
n
d
D
.
W
a
g
n
e
r
,
“
To
w
a
r
d
s
e
v
a
l
u
a
t
i
n
g
t
h
e
r
o
b
u
s
t
n
e
ss
o
f
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
i
n
2
0
1
7
I
EEE
S
y
m
p
o
si
u
m
o
n
S
e
c
u
ri
t
y
a
n
d
Pri
v
a
c
y
(
S
P)
,
M
a
y
2
0
1
7
,
p
p
.
3
9
–
5
7
,
d
o
i
:
1
0
.
1
1
0
9
/
S
P
.
2
0
1
7
.
4
9
.
[
2
3
]
I
.
Ev
t
i
m
o
v
e
t
a
l
.
,
“
R
o
b
u
s
t
p
h
y
si
c
a
l
-
w
o
r
l
d
a
t
t
a
c
k
s
o
n
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
mo
d
e
l
s
,
”
C
o
m
p
u
t
e
r
V
i
si
o
n
a
n
d
P
a
t
t
e
rn
R
e
c
o
g
n
i
t
i
o
n
,
2
0
1
7
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s
/
1
7
0
7
.
0
8
9
4
5
.
[
2
4
]
J.
G
a
o
,
J.
La
n
c
h
a
n
t
i
n
,
M
.
L
o
u
S
o
f
f
a
,
a
n
d
Y
.
Q
i
,
“
B
l
a
c
k
-
b
o
x
g
e
n
e
r
a
t
i
o
n
o
f
a
d
v
e
r
s
a
r
i
a
l
t
e
x
t
se
q
u
e
n
c
e
s
t
o
e
v
a
d
e
d
e
e
p
l
e
a
r
n
i
n
g
c
l
a
ssi
f
i
e
r
s,
”
i
n
2
0
1
8
I
E
EE
S
e
c
u
ri
t
y
a
n
d
Pr
i
v
a
c
y
W
o
rks
h
o
p
s
(
S
PW
)
,
M
a
y
2
0
1
8
,
p
p
.
5
0
–
5
6
,
d
o
i
:
1
0
.
1
1
0
9
/
S
P
W
.
2
0
1
8
.
0
0
0
1
6
.
[
2
5
]
A
.
K
a
l
y
a
n
i
e
t
a
l
.
,
“
TC
A
B
:
t
e
x
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
a
t
t
a
c
k
b
e
n
c
h
mar
k
d
a
t
a
s
e
t
,
”
Z
e
n
o
d
o
,
2
0
2
2
.
h
t
t
p
s:
/
/
z
e
n
o
d
o
.
o
r
g
/
r
e
c
o
r
d
s/
7
2
2
6
5
1
9
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
S
a
n
j
a
ik
a
n
th
E
.
Va
d
a
k
k
e
th
il
S
o
m
a
n
a
th
a
n
Pi
ll
a
i
(S
e
n
i
o
r
M
e
m
b
e
r,
IEE
E)
h
o
ld
s
a
n
M
S
in
so
ftwa
re
e
n
g
i
n
e
e
rin
g
f
ro
m
Th
e
Un
i
v
e
rsity
o
f
Tex
a
s
a
t
Au
stin
,
Tex
a
s,
USA
,
a
n
d
a
BE
fro
m
t
h
e
Un
i
v
e
rsity
o
f
Ca
li
c
u
t,
Ke
ra
la,
In
d
ia.
Cu
rre
n
tl
y
p
u
r
su
in
g
a
P
h
.
D
.
i
n
c
o
m
p
u
ter
sc
ien
c
e
a
t
th
e
Un
i
v
e
rsity
o
f
N
o
r
th
Da
k
o
ta,
G
ra
n
d
F
o
rk
s,
No
rt
h
Da
k
o
ta,
USA,
h
is
re
se
a
rc
h
sp
a
n
s
d
iv
e
rse
a
re
a
s
su
c
h
a
s
m
o
b
il
e
n
e
tw
o
rk
s,
n
e
tw
o
rk
se
c
u
rit
y
,
p
ri
v
a
c
y
,
lo
c
a
ti
o
n
-
b
a
se
d
se
rv
ice
s,
a
n
d
m
isin
fo
rm
a
ti
o
n
d
e
t
e
c
ti
o
n
.
He
is
a
p
ro
u
d
m
e
m
b
e
r
o
f
S
ig
m
a
Xi,
Th
e
S
c
ien
ti
fic
Re
se
a
rc
h
Ho
n
o
r
S
o
c
iety
,
u
n
d
e
rli
n
in
g
h
is
c
o
m
m
it
m
e
n
t
t
o
a
d
v
a
n
c
i
n
g
sc
ien
ti
f
ic
k
n
o
wle
d
g
e
a
n
d
re
se
a
rc
h
e
x
c
e
ll
e
n
c
e
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
s.e
v
a
d
a
k
k
e
th
il
@
u
n
d
.
e
d
u
.
S
r
in
iv
a
s
A.
Va
d
d
a
d
i
is
a
d
y
n
a
m
ic
a
n
d
f
o
rwa
rd
-
t
h
in
k
i
n
g
p
r
o
fe
ss
io
n
a
l
in
t
h
e
field
o
f
Clo
u
d
a
n
d
De
v
S
e
c
Op
s.
Wi
t
h
a
so
li
d
e
d
u
c
a
ti
o
n
a
l
fo
u
n
d
a
ti
o
n
in
c
o
m
p
u
ter
sc
ien
c
e
,
he
e
m
b
a
rk
e
d
o
n
a
jo
u
rn
e
y
o
f
c
o
n
t
in
u
o
u
s
lea
rn
in
g
a
n
d
p
r
o
fe
ss
io
n
a
l
g
r
o
wth
.
T
h
e
ir
re
len
t
les
s
p
u
rsu
i
t
o
f
k
n
o
wle
d
g
e
a
n
d
c
o
m
m
it
m
e
n
t
to
sta
y
in
g
a
t
th
e
fo
re
fro
n
t
o
f
in
d
u
stry
a
d
v
a
n
c
e
m
e
n
ts
h
a
s
e
a
rn
e
d
th
e
m
re
c
o
g
n
it
i
o
n
a
s
a
th
o
u
g
h
t
lea
d
e
r
in
th
e
Cl
o
u
d
a
n
d
De
v
S
e
c
Op
s
sp
a
c
e
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
v
sa
d
9
3
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
7
3
5
-
1
7
4
4
1744
Dr
.
Ro
h
ith
Va
ll
a
b
h
a
n
e
n
i
is
a
d
e
d
ica
ted
wo
rk
e
r
with
a
str
o
n
g
wo
r
k
e
th
ic
i
n
lea
d
in
g
tea
m
s
to
s
o
lv
e
o
rg
a
n
iza
ti
o
n
a
l
iss
u
e
s.
He
is
c
a
p
a
b
le
o
f
lea
rn
i
n
g
a
ll
a
sp
e
c
ts
o
f
in
fo
rm
a
ti
o
n
with
i
n
a
c
o
m
p
a
n
y
a
n
d
u
sin
g
th
e
tec
h
n
ica
l
k
n
o
wle
d
g
e
a
n
d
b
u
si
n
e
ss
b
a
c
k
g
ro
u
n
d
to
e
ffe
c
ti
v
e
ly
a
n
a
ly
z
e
se
c
u
rit
y
m
e
a
su
re
s
to
d
e
term
in
e
th
e
ir
e
ffe
c
ti
v
e
n
e
ss
in
o
rd
e
r
t
o
stre
n
g
th
e
n
th
e
o
v
e
ra
ll
se
c
u
rit
y
p
o
stu
re
.
He
h
a
s
g
re
a
t
wo
rk
e
t
h
ic
a
n
d
o
u
tstan
d
in
g
tea
m
lea
d
e
rsh
ip
sk
i
ll
s
a
n
d
se
e
k
to
a
c
c
o
m
p
li
sh
o
rg
a
n
iza
ti
o
n
a
l
g
o
a
ls,
wh
i
le
g
ro
wi
n
g
i
n
k
n
o
wle
d
g
e
a
n
d
e
x
p
e
rien
c
e
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
r
o
h
it
.
v
a
ll
a
b
h
a
n
e
n
i
.
2
2
2
2
@g
m
a
il
.
c
o
m
.
S
a
n
t
o
sh
Re
d
d
y
Add
u
l
a
a
se
n
io
r
m
e
m
b
e
r
o
f
IEE
E
,
is
a
re
se
a
rc
h
sc
h
o
lar
a
t
th
e
Un
iv
e
rsity
o
f
th
e
Cu
m
b
e
rlan
d
s.
H
is ed
u
c
a
ti
o
n
a
l
q
u
a
li
fica
ti
o
n
s in
c
lu
d
e
a
P
h
.
D.
a
n
d
a
M
a
ste
r
o
f
S
c
ien
c
e
in
in
fo
rm
a
ti
o
n
tec
h
n
o
lo
g
y
.
Wi
th
e
x
ten
si
v
e
e
x
p
e
rien
c
e
i
n
th
e
IT
in
d
u
str
y
,
h
e
h
a
s
d
e
m
o
n
stra
ted
e
x
p
e
rti
se
a
c
ro
ss
m
u
lt
ip
le
d
o
m
a
in
s.
He
is
a
n
in
n
o
v
a
to
r
wh
o
h
a
s
m
a
d
e
sig
n
ifi
c
a
n
t
c
o
n
tri
b
u
ti
o
n
s
t
o
a
c
a
d
e
m
ic
re
se
a
rc
h
th
ro
u
g
h
h
is
a
rti
c
les
a
s
a
n
a
u
th
o
r
a
n
d
c
o
-
a
u
th
o
r
.
Ad
d
it
i
o
n
a
ll
y
,
h
e
se
rv
e
s
a
s
a
re
v
iew
e
r
fo
r
e
ste
e
m
e
d
jo
u
rn
a
ls,
d
e
m
o
n
stra
ti
n
g
h
is
c
o
m
m
it
m
e
n
t
to
a
d
v
a
n
c
in
g
k
n
o
wle
d
g
e
a
n
d
u
p
h
o
l
d
in
g
h
ig
h
sta
n
d
a
rd
s
i
n
sc
h
o
larl
y
p
u
b
li
c
a
ti
o
n
s
with
in
h
is
field
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
sa
n
to
sh
a
d
d
u
lait@
g
m
a
il
.
c
o
m
.
Dr
.
Bh
u
v
a
n
e
sh
Ana
n
th
a
n
re
c
e
iv
e
d
th
e
B.
E
.
d
e
g
re
e
i
n
e
lec
tri
c
a
l
a
n
d
e
lec
tro
n
ics
e
n
g
in
e
e
rin
g
fro
m
An
n
a
Un
iv
e
rs
it
y
in
2
0
1
2
,
M
.
Tec
h
.
in
p
o
we
r
sy
ste
m
e
n
g
in
e
e
rin
g
fr
o
m
Ka
las
a
li
n
g
a
m
Un
iv
e
rsit
y
i
n
2
0
1
4
a
n
d
P
h
.
D.
d
e
g
re
e
fro
m
F
a
c
u
lt
y
o
f
El
e
c
tri
c
a
l
E
n
g
i
n
e
e
rin
g
o
f
An
n
a
Un
iv
e
rsity
i
n
2
0
1
9
.
He
h
a
s
p
u
b
li
s
h
e
d
m
o
re
th
a
n
1
0
0
p
a
p
e
rs
in
re
p
u
ted
in
tern
a
ti
o
n
a
l
jo
u
r
n
a
ls,
7
5
p
a
p
e
rs
i
n
i
n
tern
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
s
a
n
d
2
0
b
o
o
k
s
.
H
e
is
a
li
fe
ti
m
e
m
e
m
b
e
r
o
f
In
tern
a
ti
o
n
a
l
S
o
c
iety
f
o
r
Re
se
a
rc
h
a
n
d
De
v
e
lo
p
m
e
n
t,
In
tern
a
ti
o
n
a
l
As
so
c
iatio
n
o
f
E
n
g
i
n
e
e
rs.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
b
h
u
v
a
n
e
sh
.
a
n
a
n
th
a
n
@
g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.