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
7
,
No
.
1
,
J
an
u
ar
y
20
2
5
,
p
p
.
2
5
0
~
2
5
6
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
7
.
i
1
.
pp
250
-
2
5
6
250
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
Att
en
tion ba
sed E
ng
lish to
Indo
-
Ar
y
a
n and
Dra
v
idia
n
la
ng
ua
g
e
trans
la
tion usin
g
spa
rsely
f
a
ctored
NM
T
Rit
esh
K
um
a
r
Dwiv
edi
1
,
P
a
rm
a
Na
nd
1
,
O
m
P
a
l
2
1
D
e
p
a
r
t
me
n
t
of
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
S
h
a
r
d
a
U
n
i
v
e
r
si
t
y
,
G
r
e
a
t
e
r
N
o
i
d
a
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
U
n
i
v
e
r
s
i
t
y
o
f
D
e
l
h
i
,
N
e
w
D
e
l
h
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
13
,
2
0
2
4
R
ev
is
ed
J
u
n
11
,
2
0
2
4
Acc
ep
ted
Sep
2
,
2
0
2
4
Ne
u
ra
l
m
a
c
h
in
e
tra
n
sla
ti
o
n
(NM
T)
is
a
so
p
h
ist
ica
ted
tec
h
n
iq
u
e
t
h
a
t
e
m
p
lo
y
s
a
larg
e
,
sin
g
u
lar
n
e
u
ra
l
n
e
two
r
k
to
lea
rn
a
n
d
e
x
e
c
u
te
a
u
t
o
m
a
ti
c
tran
sla
ti
o
n
tas
k
s.
Un
li
k
e
sta
ti
stica
l
m
a
c
h
i
n
e
t
ra
n
sla
ti
o
n
sy
ste
m
s,
NMT
h
a
n
d
les
th
e
e
n
ti
re
tran
sla
ti
o
n
p
ro
c
e
ss
in
a
n
e
nd
-
to
-
e
n
d
m
a
n
n
e
r,
re
m
o
v
i
n
g
t
h
e
n
e
e
d
fo
r
a
d
d
it
i
o
n
a
l
c
o
m
p
o
n
e
n
ts
.
T
h
is
a
p
p
ro
a
c
h
h
a
s
s
h
o
wn
si
g
n
ifi
c
a
n
t
p
ro
m
ise
in
tran
sla
ti
o
n
q
u
a
l
it
y
a
n
d
h
a
s
b
e
c
o
m
e
th
e
p
re
v
a
len
t
m
e
th
o
d
.
I
n
t
h
is
stu
d
y
,
we
a
p
p
ly
sp
a
rse
ly
fa
c
to
re
d
NMT
t
o
E
n
g
l
ish
a
n
d
se
v
e
ra
l
In
d
o
-
Ar
y
a
n
(Hin
d
i,
Be
n
g
a
li
)
a
n
d
Dra
v
id
ian
(
Tam
il
,
M
a
lay
a
lam
)
lan
g
u
a
g
e
p
a
irs.
S
p
e
c
ifi
c
a
ll
y
,
we
d
e
v
e
lo
p
th
e
m
a
c
h
i
n
e
tran
sla
ti
o
n
s
y
ste
m
u
sin
g
a
n
a
tt
e
n
t
io
n
-
b
a
se
d
m
e
c
h
a
n
ism
.
A
sig
n
ifi
c
a
n
t
p
ro
b
le
m
with
trad
it
io
n
a
l
tran
sfo
rm
e
r
s
is
th
e
h
u
g
e
m
e
m
o
ry
re
q
u
irem
e
n
t.
T
h
e
re
fo
re
,
a
sp
a
rse
ly
fa
c
to
re
d
NM
T
(
S
F
NM
T)
is
u
se
d
to
re
d
u
c
e
th
e
m
e
m
o
ry
re
q
u
irem
e
n
t
b
u
t
a
lso
im
p
ro
v
e
s
th
e
train
in
g
ti
m
e
,
th
e
re
b
y
,
re
d
u
c
i
n
g
t
h
e
c
o
m
p
u
ti
n
g
ti
m
e
.
In
t
h
is
p
a
p
e
r,
tak
e
i
n
sp
ira
ti
o
n
fro
m
Va
sw
a
n
i
t
ra
n
sfo
rm
e
r
a
n
d
m
o
d
if
y
it
to
g
e
t
t
h
e
b
e
st
re
su
lt
s.
T
h
e
sy
ste
m
’
s
p
e
rfo
rm
a
n
c
e
wa
s
e
v
a
lu
a
ted
u
sin
g
th
e
BLE
U
m
e
tri
c
.
T
h
e
p
r
o
p
o
s
e
d
m
o
d
e
l
in
d
trl
a
c
h
iev
e
s
a
BLUE
sc
o
re
o
f
3
2
.
1
3
(e
n
→
h
i)
,
2
9
.
3
1
(e
n
→
b
e
),
3
1
.
2
1
(e
n
→
ta),
2
1
.
1
2
(e
n
→
m
l)
a
n
d
3
2
.
6
7
(e
n
→
h
i
),
2
9
.
3
8
(e
n
→
b
e
),
3
1
.
7
5
(e
n
→
ta),
2
1
.
1
7
(e
n
→
m
l)
wit
h
o
u
t
b
a
c
k
t
ra
n
sla
ti
o
n
a
n
d
wit
h
b
a
c
k
tran
sl
a
ti
o
n
.
T
o
e
v
a
lu
a
te
th
e
p
e
rfo
rm
a
n
c
e
o
f
t
h
e
sy
ste
m
,
we
c
o
m
p
a
re
d
th
e
re
su
lt
s
with
th
o
se
o
f
e
x
isti
n
g
sy
ste
m
s.
Th
e
d
e
v
e
lo
p
e
d
sy
ste
m
d
e
m
o
n
stra
ted
a
m
a
rg
in
a
ll
y
h
i
g
h
e
r
BLE
U sc
o
re
th
a
n
b
o
th
A
n
g
laM
T
a
n
d
G
o
o
g
le
t
ra
n
sla
te.
K
ey
w
o
r
d
s
:
B
L
E
U
s
co
r
es
L
in
g
u
is
tic
d
r
o
p
o
u
t
Ma
ch
in
e
tr
an
s
latio
n
NM
T
SF
NM
T
T
r
an
s
f
o
r
m
e
r
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
:
R
ites
h
Ku
m
ar
Dwiv
ed
i
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
Sh
a
r
d
a
Un
iv
er
s
ity
Gr
ea
ter
No
id
a,
I
n
d
ia
E
m
ail: r
ites
h
.
d
wiv
ed
i@
n
ic.
in
1.
I
NT
RO
D
UCT
I
O
N
“
Ma
ch
in
e
t
r
an
s
latio
n
”
r
ef
er
s
to
th
e
au
to
m
ated
tr
an
s
latio
n
o
f
tex
t
b
etwe
en
lan
g
u
a
g
es
u
s
in
g
co
m
p
u
ter
s
.
I
n
th
e
f
ield
o
f
ar
tif
icial
in
tellig
en
ce
(
AI
)
,
m
ac
h
i
n
e
tr
an
s
latio
n
(
MT
)
is
co
n
s
id
er
ed
an
AI
-
c
o
m
p
lete
p
r
o
b
lem
,
m
ea
n
in
g
th
at
s
o
lv
in
g
MT
is
ak
in
to
ad
d
r
ess
in
g
th
e
co
r
e
ch
allen
g
e
o
f
AI
:
d
ev
el
o
p
in
g
a
s
y
s
tem
with
g
en
er
al
in
tellig
en
ce
.
W
ea
v
er
(
1
9
4
9
)
p
r
esen
ted
th
e
f
ir
s
t
wo
r
k
ab
le
m
eth
o
d
f
o
r
tr
an
s
latin
g
te
x
t
u
s
in
g
co
m
p
u
te
r
s
.
I
t
s
tim
u
lated
th
e
f
ield
o
f
MT
r
esear
ch
.
T
h
e
o
r
ig
i
n
al
m
ac
h
in
e
tr
an
s
latio
n
m
o
d
els
r
elied
s
o
lely
o
n
wo
r
d
-
f
o
r
-
wo
r
d
r
e
p
la
ce
m
en
ts
with
m
u
lti
lin
g
u
al
d
ictio
n
a
r
ies,
wh
ich
n
e
v
er
p
r
o
d
u
ce
d
s
atis
f
ac
to
r
y
tr
a
n
s
latio
n
s
.
T
h
e
p
o
o
r
ap
p
r
aisal
o
f
th
e
in
f
am
o
u
s
1
9
6
6
AL
PAC
r
ep
o
r
t1
h
in
d
er
ed
r
esear
ch
o
n
MT
.
T
h
e
ad
o
p
tio
n
o
f
s
tatis
tic
-
b
ased
m
eth
o
d
o
l
o
g
ies
m
ar
k
ed
a
s
ig
n
i
f
ican
t
r
esu
r
g
en
ce
o
f
r
esear
c
h
.
T
h
ese
m
eth
o
d
s
u
s
ed
a
s
en
ten
ce
-
alig
n
ed
p
a
r
allel
co
r
p
u
s
to
lear
n
th
e
b
ilin
g
u
al
d
ictio
n
ar
ies
(
o
r
tr
a
n
s
latio
n
m
o
d
els)
in
a
p
r
o
b
ab
ilis
tic
m
an
n
e
r
.
Statis
t
ica
l
m
ac
h
in
e
tr
an
s
latio
n
(
SMT
)
m
o
d
els is
th
e
ag
g
r
e
g
ate
n
am
e
f
o
r
th
ese
m
o
d
els.
Fo
r
o
v
er
a
d
ec
ad
e,
SM
T
was t
h
e
p
r
ev
ailin
g
s
tan
d
ar
d
.
Ph
r
ase
-
b
ased
m
o
d
el
s
[
1
]
,
wh
ile
lo
n
g
-
estab
lis
h
ed
an
d
u
tili
ze
d
in
b
o
th
co
m
m
er
c
ial
ap
p
licatio
n
s
an
d
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
A
tten
tio
n
b
a
s
ed
E
n
g
lis
h
to
I
n
d
o
-
A
r
ya
n
a
n
d
Dra
vid
ia
n
la
n
g
u
a
g
e
…
(
R
ites
h
K
u
ma
r
Dw
ived
i
)
251
m
ac
h
in
e
tr
an
s
latio
n
r
esear
c
h
,
h
av
e
s
ee
n
th
eir
tr
a
n
s
latio
n
q
u
ality
p
latea
u
o
v
er
ti
m
e.
T
h
e
lo
n
g
-
ter
m
d
ep
en
d
e
n
cies
in
a
s
en
ten
ce
ar
e
n
o
t
c
ap
tu
r
ed
b
y
p
h
r
ase
-
b
ased
SMT
m
o
d
els,
wh
ich
b
ase
th
eir
tr
an
s
latio
n
ju
d
g
m
en
ts
o
n
p
h
r
ases
.
T
h
e
ad
d
itio
n
o
f
n
u
m
er
o
u
s
co
m
p
o
n
e
n
ts
,
in
clu
d
in
g
lan
g
u
ag
e
m
o
d
els,
r
eo
r
d
e
r
in
g
m
o
d
els,
len
g
th
p
en
alties,
an
d
tr
an
s
latio
n
m
o
d
els,
h
as
m
a
d
e
t
h
e
en
tire
SMT
p
r
o
ce
s
s
m
o
r
e
co
m
p
lex
[
2
]
.
Owin
g
to
th
ese
ch
allen
g
es,
a
s
ig
n
i
f
ican
t
m
o
d
if
icatio
n
o
f
th
e
cu
r
r
en
t
s
y
s
tem
was
n
ec
ess
ar
y
.
Sev
er
al
o
f
SMT
s
y
s
tem
s
’
d
r
awb
ac
k
s
ar
e
a
d
d
r
ess
ed
b
y
N
MT
.
I
t
is
a
co
m
p
r
eh
en
s
iv
e
e
n
d
-
to
-
en
d
s
y
s
tem
th
at
u
s
es
lar
g
e
n
eu
r
al
n
etwo
r
k
t
o
s
im
u
late
th
e
wh
o
le
m
ac
h
in
e
tr
an
s
latio
n
p
r
o
ce
s
s
.
T
h
e
r
esear
ch
er
p
r
o
p
o
s
ed
a
s
y
n
ch
r
o
n
o
u
s
in
f
er
en
ce
m
eth
o
d
ca
p
ab
le
o
f
g
e
n
er
atin
g
tr
an
s
l
atio
n
s
in
m
u
ltip
le
lan
g
u
a
g
es
s
im
u
ltan
eo
u
s
ly
[
3
]
.
c
u
r
r
e
n
t
m
ac
h
in
e
lear
n
in
g
tr
an
s
latio
n
m
o
d
els,
asid
e
f
r
o
m
“
hi
-
en
”
tr
a
n
s
latio
n
,
ap
p
ea
r
to
tr
an
s
late
o
th
e
r
I
n
d
ian
lan
g
u
ag
es
(
B
en
g
ali,
T
am
il,
Pu
n
jab
i,
Ur
d
u
an
d
G
u
ajar
ati,
T
elu
g
u
,
Kan
n
ad
a,
Ma
la
y
alam
)
with
an
ac
cu
r
ac
y
o
f
o
n
ly
1
0
%,
Kh
an
et
a
l.
[
4
]
.
A
tr
an
s
latio
n
m
o
d
el
f
r
o
m
E
n
g
lis
h
to
T
am
il
with
a
B
L
E
U
s
co
r
e
o
f
8
.
3
3
h
as
b
ee
n
p
r
o
p
o
s
ed
b
y
C
h
o
u
d
h
ar
y
et
a
l
.
[
5
]
.
An
NM
T
s
y
s
tem
o
f
I
I
I
T
-
H
h
a
s
b
ee
n
p
r
o
p
o
s
ed
b
y
Vik
r
a
n
t
i
n
a
p
ap
er
f
o
r
th
e
ev
alu
atio
n
o
f
W
MT
1
9
.
T
h
ey
tr
an
s
lated
n
ews
f
r
o
m
G
u
jar
ati
to
E
n
g
lis
h
u
s
i
n
g
an
att
en
tio
n
m
o
d
el
f
o
r
th
eir
wo
r
k
,
an
d
th
ey
r
ec
eiv
ed
a
B
L
E
U
s
co
r
e
o
f
9
.
8
.
T
h
e
y
h
a
v
e
tack
led
th
e
lim
ited
r
eso
u
r
ce
d
ata
as
an
is
s
u
e.
Go
o
d
ef
f
o
r
ts
h
av
e
b
ee
n
m
a
d
e
in
tr
an
s
latin
g
E
n
g
lis
h
to
Pu
n
ja
b
i.
Kam
al
Dee
p
h
av
e
p
r
o
p
o
s
ed
a
m
o
d
el
t
h
at
claim
s
to
g
iv
e
a
B
E
L
U
s
co
r
e
o
f
3
8
.
3
0
f
o
r
Pu
n
jab
i
to
E
n
g
lis
h
a
n
d
3
6
.
9
6
f
o
r
E
n
g
lis
h
to
Pu
n
jab
i.
T
h
e
Pu
n
jab
i
lan
g
u
ag
e
d
o
es
n
o
t
h
av
e
as
lo
w
o
f
a
r
eso
u
r
ce
c
o
r
p
u
s
as
it
is
f
o
r
B
h
o
jp
u
r
i o
r
o
th
er
lan
g
u
ag
es
.
I
n
an
o
th
er
wo
r
k
b
y
Sin
g
h
et
a
l.
[
6
]
Stro
n
g
atten
tio
n
m
o
d
els
wer
e
em
p
lo
y
ed
,
a
n
d
t
h
e
r
esu
lts
s
h
o
wed
a
B
L
E
U
s
co
r
e
o
f
r
o
u
g
h
ly
2
4
.
4
8
f
o
r
b
o
th
ty
p
es
o
f
tr
an
s
latio
n
.
A
wo
r
k
b
y
Haq
u
e
a
n
d
Hasan
[
7
]
h
as
p
r
o
p
o
s
ed
a
m
o
d
el
claim
in
g
th
at
th
e
s
en
ten
ce
s
p
r
o
d
u
ce
d
in
tr
an
s
latin
g
E
n
g
lis
h
s
en
ten
ce
s
to
B
en
g
ali
b
y
th
eir
m
o
d
el
ar
e
m
o
r
e
s
em
an
tically
co
r
r
ec
t
th
an
th
e
s
en
t
en
ce
s
p
r
o
d
u
ce
d
b
y
Go
o
g
le
tr
an
s
lato
r
.
Ver
y
f
ew
r
e
s
ea
r
ch
es h
av
e
b
ee
n
d
o
n
e
in
th
e
E
n
g
lis
h
to
B
en
g
ali
lan
g
u
ag
e
,
an
d
m
an
y
o
f
th
em
ar
e
n
o
t
u
p
to
p
ar
[
7
]
.
T
h
is
r
e
s
ea
r
ch
p
r
o
p
o
s
es
s
o
m
e
s
p
ec
ial
tr
an
s
latin
g
r
u
les
b
ec
au
s
e
B
e
n
g
ali
is
a
v
er
y
h
ar
d
lan
g
u
ag
e
an
d
p
r
o
d
u
cin
g
r
i
g
h
t
s
en
ten
ce
s
is
th
e
tr
u
e
ch
alle
n
g
e,
e
v
en
th
o
u
g
h
th
ei
r
alg
o
r
i
th
m
h
as
p
r
o
d
u
ce
d
ac
cu
r
ac
y
r
ates
o
f
ab
o
v
e
9
7
%.
T
h
e
p
ap
er
b
y
Sip
r
a
[
8
]
f
in
d
in
g
s
o
n
th
e
s
u
b
ject
o
f
wo
r
d
b
o
r
r
o
win
g
f
r
o
m
E
n
g
lis
h
to
Ur
d
u
tr
a
n
s
latio
n
d
e
m
o
n
s
tr
at
e
th
at
t
h
er
e
a
r
e
t
h
r
ee
p
o
s
s
ib
le
ap
p
r
o
ac
h
es:
d
ir
ec
t
b
o
r
r
o
win
g
with
m
in
im
al
o
r
n
o
m
o
d
if
icatio
n
,
u
s
in
g
a
t
r
an
s
lato
r
to
tr
a
n
s
late
f
r
o
m
E
n
g
lis
h
to
Ur
d
u
,
an
d
co
m
b
i
n
in
g
Ur
d
u
an
d
E
n
g
lis
h
.
He
p
r
o
v
id
es
n
o
d
etailed
in
f
o
r
m
a
tio
n
o
n
an
y
NM
T
tech
n
iq
u
e
o
r
m
et
h
o
d
o
lo
g
y
r
elate
d
to
em
b
ed
d
in
g
v
ec
t
o
r
d
esig
n
in
g
in
h
is
wo
r
k
.
L
in
g
a
m
et
a
l.
[
9
]
p
r
o
p
o
s
ed
a
r
u
le
-
b
ased
m
eth
o
d
f
o
r
tr
an
s
latin
g
T
elu
g
u
in
to
E
n
g
lis
h
.
T
h
is
ap
p
r
o
ac
h
,
h
ig
h
lig
h
ted
d
u
r
in
g
a
c
o
m
p
r
e
h
en
s
iv
e
r
ev
i
ew
o
f
s
tu
d
ies
in
th
e
f
ield
,
d
em
o
n
s
tr
ated
9
2
%
tr
an
s
latio
n
ac
cu
r
ac
y
f
o
r
m
o
s
t
s
en
ten
ce
s
,
wh
i
le
o
th
er
s
en
ten
ce
s
ac
h
iev
ed
ar
o
u
n
d
5
0
%
ac
cu
r
ac
y
.
T
h
e
co
n
s
tr
u
ctio
n
o
f
a
m
o
r
e
ex
o
tic
m
o
d
el
th
at
is
b
ased
o
n
NM
T
p
r
in
cip
les
an
d
ca
n
g
e
n
er
ate
s
en
ten
ce
s
th
at
ar
e
m
o
r
e
s
em
an
tically
co
r
r
ec
t is wh
er
e
th
e
g
a
p
is
f
o
u
n
d
in
th
is
in
s
tan
ce
.
I
n
th
is
p
ap
er
,
a
n
ew
ap
p
r
o
ac
h
u
s
in
g
tr
an
s
f
o
r
m
er
-
b
ased
m
ac
h
in
e
tr
an
s
latio
n
is
s
h
o
wn
.
A
s
ig
n
if
ican
t
p
r
o
b
lem
with
tr
a
d
itio
n
al
tr
an
s
f
o
r
m
er
s
is
th
e
h
u
g
e
m
em
o
r
y
r
eq
u
ir
em
en
t.
T
h
er
ef
o
r
e,
a
s
p
ar
s
ely
f
ac
to
r
ed
n
e
u
r
al
m
ac
h
in
e
tr
an
s
latio
n
(
SF
NM
T
)
is
u
s
ed
th
at
r
ed
u
ce
s
th
e
m
em
o
r
y
r
e
q
u
ir
e
d
an
d
im
p
r
o
v
e
s
th
e
tr
ain
in
g
tim
e,
th
er
eb
y
,
r
e
d
u
cin
g
th
e
co
m
p
u
tin
g
tim
e.
I
n
th
is
p
ap
er
,
tak
e
in
s
p
ir
atio
n
f
r
o
m
Vaswan
i
tr
an
s
f
o
r
m
er
an
d
m
o
d
if
y
i
t
to
g
et
t
h
e
b
est
r
esu
lts
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
was
tr
ain
e
d
o
n
d
atasets
o
f
Sam
an
an
tar
.
B
L
UE
s
co
r
e
h
elp
s
in
ev
alu
atio
n
o
f
th
e
p
er
f
o
r
m
a
n
ce
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
in
d
tr
l
a
ch
iev
es
a
B
L
UE
s
co
r
e
o
f
3
2
.
1
3
(
e
n
→
h
i)
,
2
9
.
3
1
(
en
→
b
e)
,
3
1
.
2
1
(
en
→
ta)
,
2
1
.
1
2
(
en
→
m
l)
an
d
3
2
.
6
7
(
en
→
h
i)
,
2
9
.
3
8
(
e
n
→
b
e)
,
3
1
.
7
5
(
en
→
ta)
,
2
1
.
1
7
(
en
→
m
l)
with
o
u
t
b
ac
k
tr
an
s
latio
n
an
d
with
b
ac
k
tr
an
s
latio
n
.
A
co
m
p
ar
is
o
n
with
o
th
er
m
o
d
els
s
u
ch
as
An
g
laM
T
an
d
Go
o
g
le
t
r
an
s
late
was
d
o
n
e
an
d
it
was
o
b
s
er
v
ed
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
p
er
f
o
r
m
s
s
lig
h
tly
(
as
p
er
B
L
E
U
s
co
r
e)
th
an
th
e
o
th
er
m
o
d
els.
T
h
e
s
tr
u
ctu
r
e
o
f
t
h
e
p
ap
er
is
as
f
o
llo
ws:
s
ec
tio
n
2
d
escr
ib
es
th
e
m
eth
o
d
o
lo
g
y
,
b
e
g
in
n
in
g
with
an
ex
p
lan
atio
n
o
f
h
o
w
SF
NM
T
is
em
p
lo
y
ed
,
f
o
llo
wed
b
y
d
etails
o
n
th
e
m
o
d
if
ie
d
tr
a
n
s
f
o
r
m
e
r
.
Nex
t,
t
h
e
d
ataset
d
etails
ar
e
p
r
esen
ted
.
I
n
s
ec
tio
n
3
o
u
tlin
es
th
e
ex
p
er
im
en
tal
s
etu
p
an
d
r
e
s
u
lts
.
Fin
ally
,
s
ec
t
io
n
4
d
is
cu
s
s
es
th
e
co
n
clu
s
io
n
an
d
s
u
g
g
ests
f
u
tu
r
e
d
ir
ec
tio
n
s
.
2.
M
E
T
H
O
D
2
.
1
.
Sp
a
rsely
f
a
c
t
o
re
d NMT
I
n
o
u
r
s
u
g
g
ested
m
eth
o
d
,
u
n
li
k
e
o
th
er
s
y
s
tem
s
th
at
o
n
l
y
r
ec
eiv
e
p
lain
te
x
t
f
o
r
tr
an
s
latio
n
,
th
is
m
o
d
el
r
ec
eiv
es
tex
t
an
n
o
tated
b
y
a
lin
g
u
is
tic
an
n
o
tatio
n
s
y
s
tem
f
o
r
s
o
u
r
ce
-
s
id
e
tr
an
s
latio
n
[
1
0
]
-
[
1
2
]
.
R
aw
tex
t
is
u
s
ed
in
th
e
tar
g
et
s
id
e
f
o
r
tr
ain
in
g
.
T
h
e
m
o
r
p
h
o
lo
g
ical
f
e
atu
r
es
co
m
p
o
n
en
t
h
as
a
wid
e
r
an
g
e
o
f
p
o
s
s
ib
le
v
alu
es
s
in
ce
ea
ch
wo
r
d
m
ig
h
t
co
n
tain
a
m
ix
tu
r
e
o
f
th
ese
f
e
atu
r
e
v
alu
es.
I
ts
in
d
iv
id
u
al
f
ea
tu
r
e
v
alu
es
m
ay
b
e
r
ar
e,
lead
in
g
to
in
f
r
eq
u
e
n
t
u
p
d
ates
to
th
e
em
b
ed
d
e
d
v
ec
to
r
s
f
o
r
m
o
r
p
h
o
l
o
g
ical
ch
ar
ac
te
r
is
tics
d
u
r
in
g
tr
ai
n
in
g
.
W
h
en
s
p
litt
in
g
th
e
d
ata
f
o
r
tr
ain
in
g
,
th
e
m
o
d
el
s
h
o
ws
th
e
f
r
eq
u
e
n
cy
o
f
co
m
b
in
atio
n
s
o
f
m
o
r
p
h
o
lo
g
ical
f
ea
tu
r
es
co
m
p
ar
e
d
to
th
e
f
r
e
q
u
en
cy
o
f
ea
ch
in
d
i
v
id
u
al
m
o
r
p
h
o
lo
g
ical
f
ea
tu
r
e
.
I
t
is
ev
id
en
t
th
at
th
er
e
ar
e
o
r
d
er
s
o
f
m
ag
n
it
u
d
e
m
o
r
e
p
o
s
s
ib
le
co
m
b
in
atio
n
s
th
an
th
er
e
ar
e
in
d
iv
id
u
al
f
ea
tu
r
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
7
,
No
.
1
,
J
an
u
ar
y
20
2
5
:
250
-
2
5
6
252
I
t
is
s
u
g
g
ested
lab
ellin
g
ea
ch
wo
r
d
ac
co
r
d
i
n
g
to
th
e
m
o
r
p
h
o
lo
g
ical
f
ea
tu
r
e
s
p
ac
e
r
ath
er
th
an
tr
ea
tin
g
ea
ch
co
m
b
in
atio
n
as
a
s
ep
ar
at
e
f
ac
to
r
v
alu
e.
T
o
d
o
th
is
,
th
e
m
o
d
el
m
ain
tain
s
a
n
em
b
ed
d
in
g
tab
le
with
a
v
alu
e
f
o
r
ea
ch
m
o
r
p
h
o
lo
g
ical
f
ea
tu
r
e
f
o
r
ea
ch
en
tr
y
.
I
n
a
d
d
itio
n
t
o
th
e
p
r
ev
i
o
u
s
ly
m
en
tio
n
e
d
m
o
r
p
h
o
lo
g
ical
f
ea
tu
r
e
v
o
ca
b
u
lar
y
,
a
lem
m
a
-
v
o
ca
b
u
l
ar
y
is
also
m
ain
tain
ed
.
Sp
ec
if
ically
,
wh
ile
en
co
d
in
g
tex
t
f
o
r
s
p
ar
s
ely
f
ac
to
r
ed
NM
T
,
it
is
v
er
if
ied
th
at
e
v
e
r
y
wo
r
d
is
lem
m
atiza
b
le
with
in
clu
s
io
n
in
th
e
v
o
ca
b
u
lar
y
(
lem
m
a)
.
I
f
s
o
,
th
e
m
o
d
el
en
co
d
es
th
e
w
o
r
d
b
y
ad
d
in
g
th
e
lem
m
a
’
s
em
b
e
d
d
ed
v
ec
to
r
alo
n
g
with
th
e
e
m
b
ed
d
e
d
v
ec
to
r
s
o
f
ea
ch
m
o
r
p
h
o
lo
g
ical
c
h
ar
ac
ter
i
s
tic
it
p
o
s
s
ess
ed
.
I
n
th
e
ev
en
t
th
at
a
wo
r
d
ca
n
n
o
t
b
e
lem
m
atize
d
o
r
is
n
o
t
f
o
u
n
d
in
th
e
v
o
ca
b
,
B
PE
is
u
s
ed
f
o
r
t
o
k
en
izatio
n
,
an
em
b
ed
d
i
n
g
ta
b
le
is
also
m
ain
tain
ed
.
As
a
r
e
s
u
lt,
o
u
r
to
k
en
s
ca
n
b
e
s
u
b
wo
r
d
s
o
r
(
lem
m
a
+
m
o
r
p
h
o
lo
g
ical
ch
a
r
ac
ter
is
tics
)
.
T
h
e
tex
t
is
f
ed
in
to
a
ty
p
ical
t
r
an
s
f
o
r
m
er
m
o
d
e
l
af
ter
it h
as b
ee
n
e
n
co
d
e
d
as a
s
er
ies o
f
em
b
ed
d
e
d
v
ec
to
r
s
.
B
u
ild
in
g
u
p
o
n
th
e
p
r
e
v
io
u
s
ly
p
u
b
lis
h
ed
b
ase
v
e
r
s
io
n
,
it
is
s
u
g
g
ested
a
n
a
d
d
itio
n
al
ex
ten
s
i
o
n
:
a
n
ew
h
y
p
er
p
ar
am
eter
ca
lled
‘
lin
g
u
i
s
tic
d
r
o
p
o
u
t
’
(
L
D)
i
s
in
tr
o
d
u
ce
d
,
r
e
p
r
esen
tin
g
th
e
p
r
o
b
ab
il
ity
o
f
to
k
en
izin
g
a
wo
r
d
u
s
in
g
a
s
u
b
wo
r
d
in
s
te
ad
o
f
t
h
e
(
lem
m
a
+
m
o
r
p
h
o
lo
g
ical
ch
ar
ac
ter
is
tics
)
r
ep
r
es
en
tatio
n
.
B
o
th
th
e
lem
m
atize
d
r
ep
r
esen
tatio
n
(
if
av
ailab
le)
an
d
th
e
s
u
b
wo
r
d
r
e
p
r
esen
tatio
n
ar
e
cr
ea
ted
d
u
r
in
g
d
ata
p
r
ep
ar
atio
n
.
A
B
er
n
o
u
lli
d
is
tr
ib
u
tio
n
ass
o
c
iated
with
th
e
L
D
p
r
o
b
a
b
ilit
y
i
s
u
s
ed
to
d
eter
m
in
e
th
e
ty
p
e
o
f
r
ep
r
esen
tatio
n
f
o
r
ea
ch
wo
r
d
d
u
r
in
g
b
atc
h
cr
ea
tio
n
.
T
h
e
g
o
al
o
f
L
D
is
to
t
r
ain
th
e
m
o
d
el
to
o
p
e
r
ate
in
s
ce
n
ar
io
s
with
o
u
t
lin
g
u
is
tic
d
ata,
s
u
ch
as
wh
en
a
wo
r
d
is
n
o
t
in
its
lex
ico
n
.
T
r
ain
in
g
with
L
D
r
esu
lts
in
m
o
r
e
f
r
eq
u
e
n
t
u
p
d
ates
to
th
e
s
u
b
wo
r
d
to
k
en
em
b
ed
d
in
g
s
,
lead
in
g
to
m
o
r
e
r
o
b
u
s
t
s
y
s
tem
s
,
esp
ec
ially
wh
en
h
an
d
lin
g
o
u
t
-
of
-
d
o
m
ain
m
ater
ial.
2
.
2
.
T
ra
ns
f
o
rm
er
m
o
del
T
r
an
s
f
o
r
m
e
r
s
h
av
e
g
ain
e
d
s
ig
n
if
ican
t
im
p
o
r
ta
n
ce
in
tr
an
s
lat
io
n
s
o
r
d
ee
p
e
r
n
eu
r
al
n
etwo
r
k
s
s
h
o
ws
a
g
o
o
d
p
er
f
o
r
m
a
n
ce
[
1
3
]
-
[
1
5
]
i
n
s
p
ir
atio
n
is
tak
en
f
r
o
m
[
1
6
]
an
d
m
a
k
e
m
o
d
if
icatio
n
s
t
o
th
e
t
r
an
s
f
o
r
m
e
r
.
T
h
e
en
co
d
er
r
ec
eiv
es
th
e
f
o
llo
win
g
s
eq
u
e
n
ce
=
(
1
,
2
,
3
,
…
,
)
an
d
m
a
p
s
it
to
=
(
1
,
2
,
3
,
…
,
)
.
Giv
en
z,
o
u
tp
u
t
is
g
en
er
ated
with
th
e
h
elp
o
f
t
h
e
d
ec
o
d
er
,
th
e
o
u
tp
u
t
is
=
(
1
,
2
,
3
,
…
,
)
.
B
ein
g
au
to
-
r
eg
r
ess
iv
e
at
ea
ch
s
tep
[
1
7
]
,
u
s
in
g
th
e
s
y
m
b
o
ls
th
at
wer
e
g
en
er
ated
ea
r
lier
as
ex
t
r
a
in
p
u
t
f
o
r
th
e
f
o
llo
win
g
g
en
er
atio
n
.
Fig
u
r
e
1
o
f
th
e
tr
a
n
s
f
o
r
m
er
,
r
esp
ec
tiv
ely
,
d
etails
th
e
tr
an
s
f
o
r
m
er
ar
ch
itectu
r
e
with
f
u
lly
c
o
n
n
ec
ted
lay
er
s
an
d
la
y
er
ed
s
elf
-
atten
tio
n
,
wh
ic
h
ad
h
er
es
to
t
h
is
o
v
er
all
d
esig
n
.
I
n
th
e
f
o
llo
win
g
s
ec
tio
n
s
,
ch
an
g
es
m
ad
e
to
th
e
[
1
6
]
tr
a
n
s
f
o
r
m
e
r
ar
e
d
escr
ib
ed
.
T
h
ese
ch
a
n
g
es
wer
e
f
in
alize
d
as
th
e
y
p
r
o
v
ed
th
e
b
est
ef
f
icien
c
y
o
f
th
e
m
o
d
el.
2
.
2
.
1
.
Sta
c
k
s
-
enco
der
a
nd
deco
der
−
E
n
co
d
er
:
t
o
e
n
h
an
ce
ca
p
ac
ity
,
a
s
tack
o
f
N=
8
lay
er
s
is
em
p
lo
y
ed
f
o
r
th
e
s
elf
-
atten
tio
n
m
ec
h
an
is
m
an
d
p
o
s
itio
n
al
en
co
d
in
g
.
Mo
r
e
o
v
e
r
,
a
ch
an
g
e
is
m
ad
e
to
th
e
o
u
tp
u
t
d
im
en
s
io
n
f
r
o
m
5
1
2
t
o
2
5
6
to
en
s
u
r
e
f
aster
tr
ain
in
g
o
f
t
h
e
m
o
d
el.
T
h
e
r
e
ar
e
n
o
o
th
e
r
ch
a
n
g
es in
th
e
ar
ch
itectu
r
e.
−
Dec
o
d
er
:
h
er
e
,
a
s
tack
o
f
N=
8
id
en
tical
lay
er
s
is
u
s
ed
.
T
h
er
e
ar
e
n
o
o
th
er
c
h
an
g
es in
th
e
ar
c
h
itectu
r
e
.
2
.
2
.
2
.
At
t
ent
io
n
t
hro
ug
h sca
l
ed
do
t
-
pro
du
ct
T
h
e
dot
-
p
r
o
d
u
ct
is
g
iv
en
b
y
t
h
e
(
1
)
:
(
,
,
)
=
(
√
)
(
1
)
2
.
2
.
3
.
O
ptim
izer
Ad
am
o
p
tim
izer
is
u
s
ed
as
it
is
o
r
ig
in
ally
in
[
1
6
]
,
h
o
wev
e
r
,
th
e
b
est
r
esu
lts
ar
e
with
1
=
0
.
92
an
d
2
=
0
.
96
.
2
.
3
.
Da
t
a
s
et
Sam
an
an
tar
d
ataset,
d
ev
el
o
p
e
d
b
y
AI
4
B
h
ar
at,
is
a
m
u
ltil
in
g
u
al
p
a
r
allel
co
r
p
u
s
d
esig
n
ed
to
en
h
a
n
ce
m
ac
h
in
e
tr
an
s
latio
n
s
y
s
tem
s
f
o
r
I
n
d
ian
lan
g
u
ag
es.
I
t
co
n
s
is
ts
o
f
h
ig
h
-
q
u
ality
,
alig
n
ed
tr
an
s
latio
n
s
in
m
u
ltip
le
lan
g
u
ag
e
p
air
s
,
in
clu
d
i
n
g
m
aj
o
r
I
n
d
ian
lan
g
u
ag
es.
T
h
e
d
a
t
aset
was
cr
ea
ted
to
ad
d
r
ess
th
e
n
ee
d
f
o
r
r
o
b
u
s
t
tr
ain
in
g
d
ata
i
n
lo
w
-
r
eso
u
r
ce
lan
g
u
ag
e
s
ettin
g
s
an
d
aim
s
to
s
u
p
p
o
r
t
t
h
e
d
ev
el
o
p
m
en
t
o
f
ad
v
an
ce
d
m
ac
h
in
e
tr
an
s
latio
n
m
o
d
els
b
y
p
r
o
v
id
i
n
g
ex
ten
s
iv
e
an
d
d
iv
er
s
e
lin
g
u
is
tic
d
ata.
T
h
e
Sam
an
an
tar
d
at
aset
is
a
s
ig
n
if
ican
t
r
eso
u
r
ce
f
o
r
im
p
r
o
v
in
g
t
r
an
s
latio
n
ac
cu
r
ac
y
an
d
p
e
r
f
o
r
m
an
ce
in
I
n
d
ian
la
n
g
u
a
g
es,
f
ac
ilit
atin
g
r
esear
ch
an
d
ap
p
licatio
n
s
in
n
at
u
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
.
I
n
T
ab
le
1
,
f
o
u
r
la
n
g
u
a
g
e
p
air
s
a
r
e
s
h
o
wn
with
th
e
n
o
.
o
f
s
en
ten
ce
s
f
o
r
ea
ch
la
n
g
u
a
g
e
p
air
.
Par
a
llel
co
r
p
o
r
a
s
h
ar
ed
b
y
[
1
8
]
,
[
1
9
]
is
lev
er
ag
ed
a
n
d
d
at
a
au
g
m
en
ted
to
e
n
h
an
ce
t
h
e
tr
an
s
latio
n
q
u
ality
f
o
r
lo
w
r
eso
u
r
ce
d
lan
g
u
ag
es.
T
o
k
en
izin
g
s
en
ten
ce
s
a
cr
o
s
s
all
lan
g
u
a
g
es
is
d
o
n
e
u
s
i
n
g
tr
ain
e
d
m
o
d
els,
an
d
th
e
n
f
ilter
s
en
ten
ce
s
b
as
ed
o
n
th
e
r
atio
o
f
s
o
u
r
ce
to
tar
g
et
to
k
en
c
o
u
n
t
[
2
0
]
.
R
esear
ch
er
s
ex
am
in
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
A
tten
tio
n
b
a
s
ed
E
n
g
lis
h
to
I
n
d
o
-
A
r
ya
n
a
n
d
Dra
vid
ia
n
la
n
g
u
a
g
e
…
(
R
ites
h
K
u
ma
r
Dw
ived
i
)
253
tr
an
s
latio
n
f
r
o
m
v
er
y
lo
w
-
r
eso
u
r
ce
lan
g
u
ag
es,
s
u
ch
as
Dzo
n
g
k
h
a
to
E
n
g
lis
h
,
u
s
in
g
v
ar
io
u
s
NM
T
m
o
d
els.
T
h
e
r
esu
lts
in
d
icate
th
at
th
e
B
L
E
U
s
co
r
e
o
f
Seq
2
Seq
m
o
d
els
f
lu
ctu
ates,
wh
ile
th
e
B
L
E
U
s
co
r
e
o
f
th
e
tr
an
s
f
o
r
m
er
m
o
d
el
s
h
o
ws a
s
tead
y
in
cr
ea
s
e
[
2
1
]
.
C
o
m
p
ar
in
g
s
o
m
e
s
y
s
te
m
’
s
ac
cu
r
ac
y
to
Go
o
g
le
t
r
an
s
late
’
s
3
8
.
6
7
% f
o
r
th
e
i
d
e
n
t
i
c
a
l
t
e
s
t
s
a
m
p
l
e
s
,
t
h
e
r
e
s
e
a
r
c
h
e
r
s
’
r
e
s
u
l
t
s
s
h
o
w
t
h
a
t
t
h
e
i
r
a
p
p
r
o
a
c
h
i
s
m
u
c
h
m
o
r
e
e
f
f
e
c
t
i
v
e
a
t
7
9
.
3
3
%
[
2
2
]
.
B
esei
s
o
et
a
l.
[
2
3
]
in
tr
o
d
u
ce
an
in
n
o
v
ativ
e
lin
g
u
is
tic
-
b
ased
ev
alu
atio
n
m
eth
o
d
f
o
r
E
n
g
lis
h
-
tr
an
s
lated
Ar
a
b
ic
s
en
ten
ce
s
,
wh
ich
s
ig
n
if
ica
n
tly
o
u
t
p
er
f
o
r
m
s
tr
ad
itio
n
al
M
T
ev
alu
a
tio
n
m
eth
o
d
s
lik
e
B
L
E
U.
As
a
p
r
e
-
p
r
o
ce
s
s
in
g
s
tep
f
o
r
m
ac
h
in
e
tr
an
s
latio
n
,
Ny
ein
an
d
So
e
[
2
4
]
em
p
lo
y
ed
a
m
o
d
el
t
o
r
eo
r
d
er
E
n
g
lis
h
p
h
r
ases
to
m
atch
th
e
wo
r
d
o
r
d
er
o
f
My
an
m
ar
,
r
esu
ltin
g
in
q
u
ality
im
p
r
o
v
e
m
en
ts
o
n
p
ar
with
th
e
b
aselin
e
r
u
le
-
b
ase
d
r
eo
r
d
er
i
n
g
tech
n
iq
u
e
.
Ay
u
et
a
l.
[
2
5
]
u
s
ed
in
-
d
e
p
th
lin
g
u
i
s
tic
k
n
o
wled
g
e
way
s
to
ex
p
lain
wh
y
a
s
p
ec
if
ic
s
eq
u
en
ce
k
n
o
wled
g
e/v
alu
e
p
ar
am
eter
in
th
e
tr
a
n
s
latio
n
p
r
o
ce
s
s
p
er
f
o
r
m
s
b
etter
th
a
n
th
e
o
t
h
er
m
eth
o
d
.
T
ab
le
1
.
Sam
an
a
n
tar
d
ataset
d
escr
ip
tio
n
f
o
r
Hin
d
i
,
Or
iy
a,
Pu
n
jab
i
,
an
d
T
am
il
La
n
g
u
a
g
e
p
a
i
r
P
a
r
a
l
l
e
l
C
o
r
p
u
s (N
o
.
o
f
s
e
n
t
e
n
c
e
s)
En
g
l
i
sh
-
H
i
n
d
i
1
0
1
2
5
7
0
6
En
g
l
i
sh
-
B
e
n
g
a
l
i
8
6
0
4
5
8
0
En
g
l
i
sh
-
M
a
l
a
y
a
l
a
m
5
9
2
4
4
2
6
En
g
l
i
sh
-
T
a
mi
l
5
2
6
4
8
6
7
3.
E
XP
E
R
I
M
E
N
T
AND
R
E
SU
L
T
S
T
h
e
m
o
d
el
is
tr
ain
ed
o
n
o
n
e
m
ac
h
in
e
with
8
NVI
DI
A
A
1
0
0
GPUs
.
I
n
T
a
b
le
2
d
escr
ib
es
th
e
in
f
r
astru
ctu
r
e
d
etails.
T
o
ev
al
u
ate
th
e
im
p
o
r
tan
ce
o
f
d
if
f
er
e
n
t
co
m
p
o
n
e
n
ts
o
f
th
e
t
r
an
s
f
o
r
m
er
,
ch
an
g
es
to
th
e
h
y
p
er
-
p
ar
am
eter
s
o
f
th
e
b
ase
m
o
d
el
wer
e
m
a
d
e
in
d
if
f
er
en
t
way
s
,
m
ea
s
u
r
i
n
g
th
e
c
h
an
g
e
in
p
er
f
o
r
m
a
n
ce
o
n
th
e
tr
an
s
latio
n
s
o
n
th
e
d
ev
elo
p
m
en
t set
.
T
ab
le
3
s
h
o
ws
B
L
E
U
s
co
r
es
an
d
p
r
o
v
id
es
d
etails
o
f
h
y
p
e
r
p
ar
am
eter
s
tak
en
d
u
r
in
g
m
o
d
el
tr
ain
in
g
.
Fo
llo
win
g
Fig
u
r
e
1
s
h
o
ws
th
e
co
m
p
ar
is
o
n
b
etwe
en
p
r
o
p
o
s
ed
m
o
d
el
an
d
o
th
er
s
tan
d
ar
d
m
o
d
els.
I
t
is
s
h
o
win
g
b
etter
B
L
UE
s
co
r
e
f
o
r
all
f
o
u
r
lan
g
u
ag
e
p
a
ir
s
.
Pro
p
o
s
ed
m
o
d
el
u
s
in
g
SF
NM
T
.
I
t
n
o
t
o
n
ly
r
ed
u
ce
s
th
e
m
em
o
r
y
r
eq
u
ir
e
d
b
u
t a
ls
o
im
p
r
o
v
es th
e
tr
ain
in
g
tim
e,
th
er
eb
y
,
r
ed
u
cin
g
th
e
co
m
p
u
tin
g
tim
e.
T
ab
le
4
co
m
p
ar
es
th
e
p
er
f
o
r
m
an
ce
o
f
f
iv
e
m
ac
h
in
e
tr
a
n
s
latio
n
m
o
d
els
f
o
r
tr
a
n
s
latin
g
E
n
g
l
is
h
to
f
o
u
r
lan
g
u
ag
es.
T
h
e
r
esu
lts
ar
e
m
ea
s
u
r
ed
u
s
in
g
two
m
etr
ics:
B
L
E
U
s
co
r
e
(
in
d
icatin
g
o
v
er
all
s
im
ilar
ity
to
r
ef
er
en
ce
t
r
an
s
latio
n
s
)
an
d
R
I
B
E
S
s
co
r
e
(
in
d
icati
n
g
f
lu
en
c
y
b
ased
o
n
h
u
m
a
n
ju
d
g
m
en
t)
.
E
n
g
lis
h
to
Hin
d
i
(
en
-
h
i)
:
Go
o
g
le
t
r
a
n
s
late
lead
s
with
a
B
L
E
U
s
co
r
e
o
f
3
2
.
3
3
,
s
u
g
g
esti
n
g
its
tr
an
s
latio
n
s
ar
e
m
o
s
t
s
im
ilar
to
h
u
m
an
r
ef
er
e
n
ce
s
f
o
r
th
is
lan
g
u
ag
e
p
air
.
Ho
we
v
er
,
I
n
d
tr
l
a
ch
iev
es
a
s
tr
o
n
g
s
co
r
e
o
f
3
2
.
1
3
,
in
d
icatin
g
v
er
y
co
m
p
etitiv
e
p
er
f
o
r
m
an
ce
.
I
n
te
r
esti
n
g
ly
,
ad
d
in
g
b
ac
k
tr
a
n
s
latio
n
to
I
n
d
t
r
l
im
p
r
o
v
es
its
B
L
E
U
s
co
r
e
s
lig
h
tly
to
3
2
.
6
7
.
I
n
ter
m
s
o
f
R
I
B
E
S sco
r
e,
I
n
d
tr
l (
with
o
r
with
o
u
t b
ac
k
tr
an
s
latio
n
)
o
u
tp
e
r
f
o
r
m
s
all
o
th
er
m
o
d
els (
0
.
7
8
6
8
an
d
0
.
7
8
7
2
,
r
esp
ec
tiv
ely
)
,
s
u
g
g
esti
n
g
its
tr
an
s
latio
n
s
ar
e
ju
d
g
ed
as
m
o
r
e
f
lu
en
t
b
y
h
u
m
a
n
s
f
o
r
E
n
g
lis
h
-
Hin
d
i.
E
n
g
lis
h
to
B
en
g
ali
(
en
-
b
e)
:
I
n
d
tr
l
tak
es
th
e
lead
with
a
B
L
E
U
s
co
r
e
o
f
2
9
.
3
1
,
f
o
llo
wed
b
y
Go
o
g
le
t
r
an
s
late
at
2
8
.
2
6
.
B
ac
k
tr
an
s
latio
n
d
o
esn
’
t
s
ee
m
to
s
ig
n
if
ican
tly
im
p
ac
t
I
n
d
tr
l
’
s
p
er
f
o
r
m
an
ce
f
o
r
B
en
g
ali
(
B
L
E
U
s
co
r
e
r
em
ain
s
at
2
9
.
3
8
)
.
W
h
en
it
co
m
es
to
R
I
B
E
S
s
co
r
es,
I
n
d
tr
l
ag
ain
d
o
m
in
ates
(
0
.
7
6
3
5
)
f
o
llo
wed
b
y
B
in
g
t
r
an
s
latio
n
(
0
.
7
8
8
2
)
.
T
h
is
in
d
icate
s
th
at
B
in
g
t
r
an
s
latio
n
ac
h
iev
es
a
h
ig
h
er
lev
el
o
f
f
lu
en
cy
f
o
r
B
en
g
ali
d
esp
it
e
h
av
in
g
a
lo
wer
B
L
E
U
s
co
r
e.
E
n
g
lis
h
to
T
am
il
(
e
n
-
ta)
:
Go
o
g
le
t
r
a
n
s
late
ac
h
iev
es
a
B
L
E
U
s
co
r
e
o
f
3
0
.
9
8
.
I
n
d
t
r
l
f
o
llo
ws
clo
s
ely
b
eh
in
d
at
3
1
.
2
1
,
a
n
d
b
ac
k
tr
a
n
s
latio
n
o
f
f
er
s
a
m
in
o
r
im
p
r
o
v
em
en
t
(
3
1
.
7
5
)
.
I
n
T
ab
le
5
,
a
co
m
p
ar
is
o
n
o
f
p
e
r
f
o
r
m
a
n
ce
o
f
in
d
t
r
l
was
d
o
n
e
with
o
th
er
s
tan
d
ar
d
m
o
d
el
s
in
clu
d
in
g
Go
o
g
le
tr
an
s
latio
n
,
B
in
g
tr
an
s
latio
n
,
AI
4
B
h
ar
at
I
n
d
icT
r
an
s
tr
an
s
latio
n
.
T
ab
le
2
.
E
x
p
er
im
en
tal
s
etu
p
d
etails
R
e
s
o
u
r
c
e
t
y
p
e
D
e
t
a
i
l
s
P
r
o
c
e
ss
o
r
I
n
t
e
l
X
e
o
n
R
a
n
d
o
m
a
c
c
e
ss
mem
o
r
y
5
1
2
G
B
G
r
a
p
h
i
c
s
p
r
o
c
e
ssi
n
g
u
n
i
t
8
x
N
V
I
D
I
A
A
1
0
0
G
P
U
s
La
n
g
u
a
g
e
P
y
t
h
o
n
T
ab
le
3
.
B
L
UE
s
co
r
es with
d
if
f
er
en
t c
o
n
f
ig
u
r
atio
n
s
o
f
t
h
e
p
r
o
p
o
s
ed
tr
an
s
f
o
r
m
er
N
d
m
o
d
e
l
d
ff
h
d
k
d
v
P
d
r
o
p
B
LU
E
(
e
n
-
h
i
)
B
LU
E
(
e
n
-
b
e
)
B
LU
E
(
e
n
-
t
a
)
B
LU
E
(
e
n
-
m
l
)
F
i
n
a
l
8
2
5
6
2
0
4
8
8
64
64
0
.
1
3
2
.
1
3
2
9
.
3
1
3
1
.
2
1
2
1
.
1
2
4
1
2
8
1
2
8
2
6
.
6
5
2
6
.
6
3
2
8
.
7
5
1
8
.
8
16
32
32
2
6
.
8
5
2
8
.
2
9
2
7
.
7
8
1
9
.
5
3
6
2
7
.
7
2
2
8
.
3
4
2
7
.
0
5
1
9
.
4
6
4
2
6
.
2
2
2
8
.
1
9
2
8
.
2
5
1
9
.
2
8
2
2
7
.
7
1
2
7
.
6
2
6
.
6
1
1
9
.
7
7
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
7
,
No
.
1
,
J
an
u
ar
y
20
2
5
:
250
-
2
5
6
254
F
ig
u
r
e
1
.
B
L
UE
s
co
r
e
o
f
p
r
o
p
o
s
ed
m
o
d
el
a
n
d
o
t
h
er
s
tan
d
ar
d
m
o
d
els
T
ab
le
4
.
Per
f
o
r
m
an
ce
co
m
p
a
r
is
o
n
o
f
in
d
tr
l
with
o
th
e
r
s
tan
d
a
r
d
m
o
d
els
M
o
d
e
l
B
LU
E
(
e
n
-
h
j
)
B
LU
E
(
e
n
-
b
e
)
B
LU
E
(
e
n
-
t
a
)
B
LU
E
(
e
n
-
m
l
)
G
o
o
g
l
e
t
r
a
n
s
l
a
t
i
o
n
3
2
.
3
3
2
8
.
2
6
3
0
.
9
8
1
9
.
4
8
B
i
n
g
t
r
a
n
s
l
a
t
i
o
n
3
1
.
6
5
2
0
.
8
7
7
2
9
.
8
5
2
0
.
3
7
A
I
4
B
h
a
r
a
t
I
n
d
i
c
Tr
a
n
s
t
r
a
n
sl
a
t
i
o
n
3
1
.
3
6
1
7
.
5
4
9
2
8
.
2
7
1
9
.
1
8
I
n
d
t
r
l
3
2
.
1
3
2
9
.
3
1
3
1
.
2
1
2
1
.
1
2
I
n
d
t
r
l
+
B
a
c
k
t
r
a
n
sl
a
t
i
o
n
3
2
.
6
7
2
9
.
3
8
3
1
.
7
5
2
1
.
1
7
T
ab
le
5
.
Per
f
o
r
m
an
ce
co
m
p
a
r
is
o
n
o
f
in
d
tr
l
with
o
th
e
r
s
tan
d
a
r
d
m
o
d
els
M
o
d
e
l
R
I
B
ES
(
e
n
-
h
j
)
R
I
B
ES
(
e
n
-
b
e
)
R
I
B
ES
(
e
n
-
t
a
)
R
I
B
ES
(
e
n
-
m
l
)
G
o
o
g
l
e
t
r
a
n
s
l
a
t
i
o
n
0
.
6
6
9
3
0
.
6
1
7
5
0
.
7
1
1
8
0
.
6
3
3
4
B
i
n
g
t
r
a
n
s
l
a
t
i
o
n
0
.
6
6
5
2
0
.
7
8
8
2
0
.
6
8
8
1
0
.
7
3
0
7
A
I
4
B
h
a
r
a
t
I
n
d
i
c
Tr
a
n
s
t
r
a
n
sl
a
t
i
o
n
0
.
6
5
7
6
0
.
6
2
2
1
0
.
6
3
5
8
0
.
7
8
0
3
I
n
d
t
r
l
0
.
7
8
6
8
0
.
7
6
3
5
0
.
7
4
7
2
0
.
7
2
5
3
I
n
d
t
r
l
+
B
a
c
k
t
r
a
n
sl
a
t
i
o
n
0
.
7
8
7
2
0
.
7
9
2
1
0
.
7
4
9
1
0
.
7
3
1
0
Ho
wev
er
,
in
T
ab
le
5
,
th
e
R
I
B
E
S
s
co
r
e
p
r
o
v
id
es
m
o
r
e
in
s
ig
h
ts
.
I
n
d
tr
l
(
with
a
n
d
with
o
u
t
b
ac
k
tr
an
s
latio
n
)
ac
h
iev
es
th
e
h
ig
h
est
s
co
r
es
(
0
.
7
4
7
2
a
n
d
0
.
7
4
9
1
)
,
s
u
g
g
esti
n
g
its
tr
an
s
latio
n
s
ar
e
ju
d
g
ed
as
m
o
r
e
f
l
u
en
t
f
o
r
E
n
g
lis
h
-
T
am
il
.
E
n
g
lis
h
t
o
Ma
lay
alam
(
en
-
m
l)
:
t
h
is
is
th
e
m
o
s
t
c
h
allen
g
in
g
tr
an
s
latio
n
f
o
r
all
m
o
d
els,
r
ef
l
ec
ted
in
th
e
lo
wer
B
L
E
U
s
co
r
es.
Go
o
g
le
t
r
an
s
late
lead
s
with
1
9
.
4
8
,
f
o
llo
we
d
b
y
I
n
d
t
r
l
at
2
1
.
1
2
.
B
ac
k
tr
an
s
latio
n
o
f
f
e
r
s
m
in
i
m
al
im
p
r
o
v
em
en
t
f
o
r
i
n
d
tr
l
(
2
1
.
1
7
)
.
I
n
ter
esti
n
g
ly
,
AI
4
B
h
ar
at
I
n
d
icT
r
an
s
T
r
an
s
latio
n
ac
h
iev
es
th
e
h
i
g
h
e
s
t
R
I
B
E
S
s
co
r
e
(
0
.
7
8
0
3
)
d
esp
ite
h
av
in
g
a
l
o
wer
B
L
E
U
s
co
r
e
.
T
h
is
s
u
g
g
ests
its
tr
an
s
latio
n
s
is
ju
d
g
e
d
as
m
o
r
e
f
lu
en
t
d
esp
ite
b
ein
g
less
s
im
ilar
to
r
ef
er
e
n
ce
tr
a
n
s
latio
n
s
in
ter
m
s
o
f
c
o
n
ten
t.
Ov
er
all,
I
n
d
tr
l
with
b
ac
k
tr
an
s
latio
n
em
er
g
es
as
a
s
tr
o
n
g
co
n
ten
d
er
,
p
a
r
ticu
lar
ly
f
o
r
ac
h
ie
v
in
g
h
u
m
an
-
q
u
ality
f
lu
en
cy
as
m
ea
s
u
r
e
d
b
y
R
I
B
E
S.
Ho
wev
er
,
Go
o
g
le
t
r
an
s
late
r
em
ain
s
co
m
p
etitiv
e
f
o
r
E
n
g
li
s
h
-
Hin
d
i
an
d
T
am
il
b
ased
o
n
B
L
E
U
s
co
r
e.
Fo
r
Ma
lay
alam
tr
an
s
latio
n
,
AI
4
B
h
ar
a
t
I
n
d
icT
r
a
n
s
s
ee
m
s
to
p
r
io
r
itiz
e
f
lu
en
c
y
b
ased
o
n
R
I
B
E
S sco
r
e.
4.
CO
NCLU
SI
O
N
An
attem
p
t
to
tr
a
n
s
late
E
n
g
lis
h
lan
g
u
a
g
e
in
to
Hin
d
i,
B
en
g
al
i,
T
am
il,
an
d
Ma
lay
alam
was
m
ad
e
with
an
in
n
o
v
ativ
e
ap
p
r
o
ac
h
o
f
u
s
in
g
SF
NM
T
an
d
a
m
o
d
if
ie
d
tr
an
s
f
o
r
m
er
m
o
d
el
.
Mo
r
eo
v
er
,
to
im
p
r
o
v
e
th
e
d
ataset
an
d
tr
an
s
latio
n
,
a
m
e
th
o
d
o
f
b
ac
k
tr
an
s
latio
n
was
ap
p
lied
.
C
h
an
g
es
to
th
e
tr
a
n
s
f
o
r
m
er
i
n
clu
d
e
t
h
e
n
u
m
b
er
o
f
lay
er
s
i
n
en
c
o
d
er
-
d
ec
o
d
er
,
ch
a
n
g
es
in
th
e
o
u
tp
u
t
d
im
en
s
io
n
,
an
d
ch
a
n
g
es
in
th
e
h
y
p
er
p
a
r
am
eter
s
o
f
th
e
Ad
am
o
p
tim
izer
.
T
h
e
m
o
d
el
n
am
ed
in
d
tr
l
was
co
m
p
ar
ed
with
o
th
er
s
tan
d
ar
d
m
o
d
els
an
d
p
er
f
o
r
m
ed
s
lig
h
tly
b
e
tter
an
d
b
est
with
b
ac
k
tr
an
s
latio
n
.
W
h
ile
i
n
d
tr
l
wi
th
b
ac
k
t
r
an
s
latio
n
e
m
er
g
es
as
a
s
tr
o
n
g
all
-
ar
o
u
n
d
p
er
f
o
r
m
er
,
ac
h
iev
in
g
g
o
o
d
s
co
r
es
in
b
o
th
s
im
ilar
ity
(
B
L
E
U
)
an
d
f
lu
e
n
cy
(
R
I
B
E
S),
o
th
er
m
o
d
els
h
av
e
th
eir
s
tr
en
g
th
s
.
Go
o
g
le
t
r
an
s
late
ex
ce
ls
in
co
n
ten
t
s
im
il
ar
ity
f
o
r
E
n
g
lis
h
to
Hin
d
i
an
d
T
am
il
(
h
ig
h
B
L
E
U
s
co
r
es).
Ho
wev
er
,
B
L
E
U
an
d
R
I
B
E
S
s
co
r
es
co
n
f
lict
ea
ch
o
th
er
.
Fo
r
in
s
tan
ce
,
a
m
o
d
el
with
h
ig
h
B
L
E
U
m
ig
h
t
n
o
t
b
e
th
e
m
o
s
t
f
lu
en
t
b
ased
o
n
R
I
B
E
S
s
co
r
e.
E
n
g
lis
h
to
Ma
la
y
alam
tr
an
s
latio
n
s
ee
m
s
m
o
s
t
ch
allen
g
in
g
,
wi
th
AI
4
B
h
ar
at
I
n
d
icT
r
an
s
p
r
io
r
iti
zin
g
f
l
u
en
cy
(
h
i
g
h
R
I
B
E
S)
e
v
en
if
co
n
ten
t
s
im
ilar
ity
is
lo
wer
(
lo
wer
B
L
E
U)
.
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
A
tten
tio
n
b
a
s
ed
E
n
g
lis
h
to
I
n
d
o
-
A
r
ya
n
a
n
d
Dra
vid
ia
n
la
n
g
u
a
g
e
…
(
R
ites
h
K
u
ma
r
Dw
ived
i
)
255
Fin
ally
,
b
ac
k
tr
a
n
s
latio
n
o
f
f
er
s
s
o
m
e
im
p
r
o
v
e
m
en
t
f
o
r
in
d
tr
l
b
u
t
th
e
im
p
ac
t
is
m
o
d
es.
A
s
ig
n
if
ican
t
p
r
o
b
lem
with
tr
ad
itio
n
al
tr
an
s
f
o
r
m
er
s
is
th
e
h
u
g
e
m
em
o
r
y
r
eq
u
ir
e
m
e
n
t.
T
h
er
ef
o
r
e,
a
SF
NM
T
was
u
s
ed
th
at
n
o
t
o
n
ly
r
ed
u
ce
d
th
e
m
em
o
r
y
r
eq
u
ir
ed
b
u
t
also
im
p
r
o
v
e
d
th
e
tr
ai
n
in
g
tim
e,
th
er
e
b
y
,
r
e
d
u
cin
g
th
e
co
m
p
u
tin
g
tim
e
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
was
tr
ain
ed
o
n
d
atasets
:
B
ac
k
tr
an
s
lated
-
Hin
d
i,
I
I
T
B
-
hi
-
e
n
,
W
AT
-
I
L
MPC
,
I
L
C
I
.
T
h
e
m
o
d
el
ac
h
iev
es
an
av
e
r
ag
e
s
p
ee
d
o
f
1
1
.
3
7
m
illi
s
ec
o
n
d
s
p
er
s
en
ten
ce
,
wh
ich
o
t
h
er
wis
e
was
1
4
.
5
6
m
illi
s
ec
o
n
d
s
with
o
u
t
th
e
SF
NM
T
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
i
n
d
tr
l
ac
h
iev
es
a
B
L
UE
s
co
r
e
o
f
3
2
.
1
3
(
e
n
→
h
i)
,
2
9
.
3
1
(
en
→
b
e)
,
3
1
.
2
1
(
en
→
ta)
,
2
1
.
1
2
(
en
→
m
l)
an
d
3
2
.
6
7
(
en
→
h
i)
,
2
9
.
3
8
(
en
→
b
e)
,
3
1
.
7
5
(
e
n
→
ta)
,
2
1
.
1
7
(
en
→
m
l)
with
o
u
t
b
ac
k
tr
a
n
s
latio
n
an
d
with
b
ac
k
tr
a
n
s
latio
n
.
A
lim
ited
d
ataset
is
a
p
r
im
e
s
o
u
r
ce
o
f
lo
w
B
L
UE
s
co
r
es;
h
o
wev
er
,
d
if
f
er
en
t
c
o
n
f
ig
u
r
atio
n
s
o
f
t
h
e
tr
a
n
s
f
o
r
m
er
m
o
d
el
ca
n
p
r
o
v
id
e
b
etter
r
esu
lts
.
I
t
i
s
r
ec
o
m
m
en
d
ed
to
e
m
p
lo
y
tech
n
iq
u
es
s
u
ch
as
g
r
id
s
ea
r
c
h
o
r
r
an
d
o
m
s
ea
r
ch
to
s
y
s
tem
atica
lly
ex
p
lo
r
e
d
if
f
er
e
n
t
co
m
b
in
atio
n
s
o
f
h
y
p
er
p
ar
a
m
eter
s
.
Oth
er
lim
itin
g
f
ac
to
r
s
in
clu
d
e
lim
ited
ca
p
ac
ity
o
f
t
h
e
ex
p
er
im
en
tal
s
etu
p
,
lead
in
g
to
s
tag
n
atio
n
in
ex
p
l
o
r
in
g
th
e
m
o
d
el
’
s
co
m
p
lete
p
o
ten
tial.
SF
NM
T
m
ay
b
e
u
s
ed
f
o
r
all
o
th
er
22
o
f
f
icial
I
n
d
ia
n
r
eg
i
o
n
al
lan
g
u
a
g
es
.
RE
F
E
R
E
NC
E
S
[
1
]
P
.
K
o
e
h
n
,
S
t
a
t
i
s
t
i
c
a
l
m
a
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
,
v
o
l
.
9
7
8
0
5
2
1
8
7
4
1
5
1
.
C
a
m
b
r
i
d
g
e
U
n
i
v
e
r
si
t
y
P
r
e
ss,
2
0
0
9
.
[
2
]
F
.
J.
O
c
h
a
n
d
H
.
N
e
y
,
“
D
i
s
c
r
i
m
i
n
a
t
i
v
e
t
r
a
i
n
i
n
g
a
n
d
ma
x
i
m
u
m
e
n
t
r
o
p
y
mo
d
e
l
s
f
o
r
s
t
a
t
i
s
t
i
c
a
l
m
a
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
A
n
n
u
a
l
Me
e
t
i
n
g
o
f
t
h
e
Asso
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
st
i
c
s
,
2
0
0
2
,
v
o
l
.
2
0
0
2
-
J
u
l
y
,
p
p
.
2
9
5
–
3
0
2
,
d
o
i
:
1
0
.
3
1
1
5
/
1
0
7
3
0
8
3
.
1
0
7
3
1
3
3
.
[
3
]
Q
.
W
a
n
g
,
J.
Z
h
a
n
g
,
a
n
d
C
.
Z
o
n
g
,
“
S
y
n
c
h
r
o
n
o
u
s
I
n
f
e
r
e
n
c
e
f
o
r
M
u
l
t
i
l
i
n
g
u
a
l
N
e
u
r
a
l
M
a
c
h
i
n
e
Tr
a
n
s
l
a
t
i
o
n
,
”
I
E
EE/
AC
M
T
ra
n
s
a
c
t
i
o
n
s
o
n
A
u
d
i
o
S
p
e
e
c
h
a
n
d
L
a
n
g
u
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
3
0
,
p
p
.
1
8
2
7
–
1
8
3
9
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
TA
S
LP.
2
0
2
2
.
3
1
7
8
2
4
1
.
[
4
]
N
.
J.
K
h
a
n
,
W
.
A
n
w
a
r
,
a
n
d
N
.
D
u
r
r
a
n
i
,
“
M
a
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
a
p
p
r
o
a
c
h
e
s
a
n
d
s
u
r
v
e
y
f
o
r
i
n
d
i
a
n
l
a
n
g
u
a
g
e
s,
”
a
rXi
v
,
J
a
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
1
.
0
4
2
9
0
.
[
5
]
H
.
C
h
o
u
d
h
a
r
y
,
A
.
K
.
P
a
t
h
a
k
,
R
.
R
.
S
a
h
a
,
a
n
d
P
.
K
u
m
a
r
a
g
u
r
u
,
“
N
e
u
r
a
l
m
a
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
f
o
r
En
g
l
i
s
h
-
Ta
m
i
l
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
T
h
i
r
d
C
o
n
f
e
re
n
c
e
o
n
M
a
c
h
i
n
e
T
ra
n
sl
a
t
i
o
n
:
S
h
a
r
e
d
T
a
s
k
P
a
p
e
rs
,
2
0
1
9
,
p
p
.
7
7
0
–
7
7
5
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
w
1
8
-
6
4
5
9
.
[
6
]
S
.
S
i
n
g
h
,
M
.
A
n
a
n
d
K
u
mar,
a
n
d
K
.
P
.
S
o
ma
n
,
“
A
t
t
e
n
t
i
o
n
b
a
se
d
En
g
l
i
s
h
t
o
P
u
n
j
a
b
i
n
e
u
r
a
l
m
a
c
h
i
n
e
t
r
a
n
s
l
a
t
i
o
n
,
”
J
o
u
rn
a
l
o
f
I
n
t
e
l
l
i
g
e
n
t
a
n
d
F
u
zzy
S
y
st
e
m
s
,
v
o
l
.
3
4
,
n
o
.
3
,
p
p
.
1
5
5
1
–
1
5
5
9
,
M
a
r
.
2
0
1
8
,
d
o
i
:
1
0
.
3
2
3
3
/
JI
F
S
-
1
6
9
4
5
0
.
[
7
]
M
.
H
a
q
u
e
a
n
d
M
.
H
a
sa
n
,
“
E
n
g
l
i
s
h
t
o
B
e
n
g
a
l
i
ma
c
h
i
n
e
t
r
a
n
s
l
a
t
i
o
n
:
a
n
a
n
a
l
y
si
s
o
f
sem
a
n
t
i
c
a
l
l
y
a
p
p
r
o
p
r
i
a
t
e
v
e
r
b
s,”
i
n
2
0
1
8
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
n
n
o
v
a
t
i
o
n
s
i
n
S
c
i
e
n
c
e
,
E
n
g
i
n
e
e
ri
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
I
C
I
S
ET
2
0
1
8
,
O
c
t
.
2
0
1
8
,
p
p
.
2
1
7
–
2
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
S
ET.
2
0
1
8
.
8
7
4
5
6
2
6
.
[
8
]
M
.
A
.
S
i
p
r
a
,
“
A
l
i
n
g
u
i
s
t
i
c
s
t
u
d
y
o
f
b
o
r
r
o
w
i
n
g
s
f
r
o
m
E
n
g
l
i
sh
t
o
u
r
d
u
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Ap
p
l
i
e
d
L
i
n
g
u
i
st
i
c
s
a
n
d
En
g
l
i
s
h
L
i
t
e
r
a
t
u
r
e
,
v
o
l
.
2
,
n
o
.
1
,
p
p
.
2
0
3
–
2
0
7
,
Jan
.
2
0
1
3
,
d
o
i
:
1
0
.
7
5
7
5
/
i
j
a
l
e
l
.
v
.
2
n
.
1
p
.
2
0
3
.
[
9
]
K
.
Li
n
g
a
m,
E.
R
a
ma
l
a
k
sh
m
i
,
a
n
d
S
.
I
n
t
u
r
i
,
“
En
g
l
i
s
h
t
o
Te
l
u
g
u
r
u
l
e
b
a
sed
mac
h
i
n
e
t
r
a
n
s
l
a
t
i
o
n
sy
s
t
e
m
:
a
h
y
b
r
i
d
a
p
p
r
o
a
c
h
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
C
o
m
p
u
t
e
r
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
0
1
,
n
o
.
2
,
p
p
.
1
9
–
2
4
,
2
0
1
4
,
d
o
i
:
1
0
.
5
1
2
0
/
1
7
6
5
9
-
8
4
7
4
.
[
1
0
]
D
.
B
a
h
d
a
n
a
u
,
K
.
H
.
C
h
o
,
a
n
d
Y
.
B
e
n
g
i
o
,
“
N
e
u
r
a
l
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
b
y
j
o
i
n
t
l
y
l
e
a
r
n
i
n
g
t
o
a
l
i
g
n
a
n
d
t
r
a
n
sl
a
t
e
,
”
2
0
1
5
.
[
1
1
]
J.
A
r
men
g
o
l
-
Es
t
a
p
é
,
M
.
R
.
C
o
st
a
-
j
u
ssà,
a
n
d
C
.
Es
c
o
l
a
n
o
,
“
E
n
r
i
c
h
i
n
g
t
h
e
t
r
a
n
sf
o
r
mer
w
i
t
h
l
i
n
g
u
i
s
t
i
c
f
a
c
t
o
r
s
f
o
r
l
o
w
-
r
e
so
u
r
c
e
mac
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
C
o
n
f
e
re
n
c
e
R
e
c
e
n
t
Ad
v
a
n
c
e
s
i
n
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
Pro
c
e
ss
i
n
g
-
D
e
e
p
L
e
a
r
n
i
n
g
f
o
r
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
Pro
c
e
ss
i
n
g
Me
t
h
o
d
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
2
0
2
1
,
p
p
.
7
3
–
7
8
,
d
o
i
:
1
0
.
2
6
6
1
5
/
9
7
8
-
9
5
4
-
4
5
2
-
072
-
4
_
0
0
9
.
[
1
2
]
M
.
G
a
r
c
í
a
-
M
a
r
t
í
n
e
z
,
W
.
A
r
a
n
s
a
,
F
.
B
o
u
g
a
r
e
s,
a
n
d
L.
B
a
r
r
a
u
l
t
,
“
A
d
d
r
e
ssi
n
g
d
a
t
a
s
p
a
r
si
t
y
f
o
r
n
e
u
r
a
l
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
b
e
t
w
e
e
n
mo
r
p
h
o
l
o
g
i
c
a
l
l
y
r
i
c
h
l
a
n
g
u
a
g
e
s,”
M
a
c
h
i
n
e
T
r
a
n
s
l
a
t
i
o
n
,
v
o
l
.
3
4
,
n
o
.
1
,
p
p
.
1
–
2
0
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
5
9
0
-
0
1
9
-
0
9
2
4
2
-
9.
[
1
3
]
I
.
S
u
t
sk
e
v
e
r
,
O
.
V
i
n
y
a
l
s
,
a
n
d
Q
.
V
.
L
e
,
“
S
e
q
u
e
n
c
e
t
o
se
q
u
e
n
c
e
l
e
a
r
n
i
n
g
w
i
t
h
n
e
u
r
a
l
n
e
t
w
o
r
k
s
,
”
A
d
v
a
n
c
e
s
i
n
N
e
u
ra
l
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ssi
n
g
S
y
s
t
e
m
s
,
v
o
l
.
4
,
n
o
.
J
a
n
u
a
r
y
,
p
p
.
3
1
0
4
–
3
1
1
2
,
2
0
1
4
.
[
1
4
]
S
.
A
.
M
o
h
a
m
e
d
,
M
.
A
.
A
b
d
o
u
,
a
n
d
A
.
A
.
E
l
sa
y
e
d
,
“
R
e
s
i
d
u
a
l
I
n
f
o
r
mat
i
o
n
F
l
o
w
f
o
r
N
e
u
r
a
l
M
a
c
h
i
n
e
Tr
a
n
sl
a
t
i
o
n
,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
1
1
8
3
1
3
–
1
1
8
3
2
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
2
2
0
6
9
1
.
[
1
5
]
K
.
C
h
o
e
t
a
l
.
,
“
Le
a
r
n
i
n
g
p
h
r
a
s
e
r
e
p
r
e
s
e
n
t
a
t
i
o
n
s
u
s
i
n
g
R
N
N
e
n
c
o
d
e
r
–
d
e
c
o
d
e
r
f
o
r
st
a
t
i
st
i
c
a
l
ma
c
h
i
n
e
t
r
a
n
s
l
a
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
1
4
C
o
n
f
e
re
n
c
e
o
n
Em
p
i
ri
c
a
l
Me
t
h
o
d
s
i
n
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
Pr
o
c
e
ssi
n
g
(
E
MN
L
P)
,
2
0
1
4
,
p
p
.
1
7
2
4
–
1
7
3
4
,
d
o
i
:
1
0
.
3
1
1
5
/
v
1
/
D
1
4
-
1
1
7
9
.
[
1
6
]
A
.
V
a
sw
a
n
i
e
t
a
l
.
,
“
A
t
t
e
n
t
i
o
n
i
s
a
l
l
y
o
u
n
e
e
d
,
”
A
d
v
a
n
c
e
s
i
n
N
e
u
r
a
l
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ss
i
n
g
S
y
st
e
m
s
,
v
o
l
.
2
0
1
7
-
D
e
c
e
mb
e
r
,
n
o
.
N
i
p
s,
p
p
.
5
9
9
9
–
6
0
0
9
,
2
0
1
7
.
[
1
7
]
A
.
G
r
a
v
e
s
,
“
G
e
n
e
r
a
t
i
n
g
s
e
q
u
e
n
c
e
s
w
i
t
h
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
a
r
x
i
v
,
2
0
1
3
,
[
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
3
0
8
.
0
8
5
0
.
[
1
8
]
S
.
S
i
r
i
p
r
a
g
a
d
a
,
J.
P
h
i
l
i
p
,
V
.
P
.
N
a
m
b
o
o
d
i
r
i
,
a
n
d
C
.
V
.
Jaw
a
h
a
r
,
“
A
mu
l
t
i
l
i
n
g
u
a
l
p
a
r
a
l
l
e
l
c
o
r
p
o
r
a
c
o
l
l
e
c
t
i
o
n
e
f
f
o
r
t
f
o
r
I
n
d
i
a
n
l
a
n
g
u
a
g
e
s
,
”
L
REC
2
0
2
0
-
1
2
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
L
a
n
g
u
a
g
e
R
e
so
u
rc
e
s
a
n
d
Ev
a
l
u
a
t
i
o
n
,
C
o
n
f
e
r
e
n
c
e
Pro
c
e
e
d
i
n
g
s
,
p
p
.
3
7
4
3
–
3
7
5
1
,
2
0
2
0
.
[
1
9
]
N
.
C
h
o
u
d
h
a
r
y
a
n
d
G
.
N
.
J
h
a
,
“
C
r
e
a
t
i
n
g
mu
l
t
i
l
i
n
g
u
a
l
p
a
r
a
l
l
e
l
c
o
r
p
o
r
a
i
n
I
n
d
i
a
n
l
a
n
g
u
a
g
e
s,
”
i
n
L
e
c
t
u
re
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
i
n
c
l
u
d
i
n
g
s
u
b
s
e
ri
e
s
L
e
c
t
u
re
N
o
t
e
s
i
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
L
e
c
t
u
r
e
N
o
t
e
s
i
n
Bi
o
i
n
f
o
rm
a
t
i
c
s)
,
v
o
l
.
8
3
8
7
LN
A
I
,
2
0
1
4
,
p
p
.
5
2
7
–
5
3
7
.
[
2
0
]
T.
K
u
d
o
a
n
d
J.
R
i
c
h
a
r
d
s
o
n
,
“
S
e
n
t
e
n
c
e
P
i
e
c
e
:
a
si
m
p
l
e
a
n
d
l
a
n
g
u
a
g
e
i
n
d
e
p
e
n
d
e
n
t
s
u
b
w
o
r
d
t
o
k
e
n
i
z
e
r
a
n
d
d
e
t
o
k
e
n
i
z
e
r
f
o
r
n
e
u
r
a
l
t
e
x
t
p
r
o
c
e
ss
i
n
g
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
1
8
C
o
n
f
e
re
n
c
e
o
n
Em
p
i
r
i
c
a
l
Me
t
h
o
d
s
i
n
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
Pr
o
c
e
ssi
n
g
:
S
y
s
t
e
m
D
e
m
o
n
s
t
r
a
t
i
o
n
s
,
2
0
1
8
,
p
p
.
6
6
–
7
1
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
D
1
8
-
2
0
1
2
.
[
2
1
]
K
.
W
a
n
g
c
h
u
k
,
S
.
C
.
N
a
v
a
n
e
e
t
h
a
k
r
i
s
h
n
a
n
,
Y
.
J
a
mt
s
h
o
,
a
n
d
Y
.
W
a
n
g
c
h
u
k
,
“
D
z
o
n
g
k
h
a
t
o
En
g
l
i
s
h
t
r
a
n
sl
a
t
i
o
n
u
s
i
n
g
t
h
e
n
e
u
r
a
l
mac
h
i
n
e
t
r
a
n
s
l
a
t
i
o
n
a
p
p
r
o
a
c
h
,
”
I
n
d
o
n
e
s
i
a
n
J
o
u
rn
a
l
o
f
E
l
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
ri
n
g
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
I
J
EE
C
S
)
,
v
o
l
.
3
1
,
n
o
.
2
,
p
p
.
8
8
5
–
8
9
2
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
e
c
s.v
3
1
.
i
2
.
p
p
8
8
5
-
8
9
2
.
[
2
2
]
D
.
A
.
H
a
me
e
d
,
T.
A
.
F
a
i
sa
l
,
A
.
K
.
A
b
b
a
s,
H
.
A
.
A
l
i
,
a
n
d
G
.
T
.
H
a
s
a
n
,
“
D
I
A
-
En
g
l
i
s
h
-
A
r
a
b
i
c
n
e
u
r
a
l
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
d
o
m
a
i
n
:
su
l
f
u
r
i
n
d
u
st
r
y
,
”
I
n
d
o
n
e
s
i
a
n
J
o
u
rn
a
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
(
I
J
EE
C
S
)
,
v
o
l
.
2
7
,
n
o
.
3
,
p
p
.
1
6
1
9
–
1
6
2
4
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
e
c
s.
v
2
7
.
i
3
.
p
p
1
6
1
9
-
1
6
2
4
.
[
2
3
]
M
.
B
e
se
i
s
o
,
S
.
Tr
i
p
a
t
h
i
,
B
.
A
l
-
S
h
b
o
u
l
,
a
n
d
R
.
A
l
j
a
d
i
d
,
“
S
e
ma
n
t
i
c
s
b
a
sed
e
n
g
l
i
sh
-
a
r
a
b
i
c
m
a
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
e
v
a
l
u
a
t
i
o
n
,
”
I
n
d
o
n
e
si
a
n
J
o
u
r
n
a
l
o
f
E
l
e
c
t
r
i
c
a
l
En
g
i
n
e
e
ri
n
g
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
I
J
EEC
S
)
,
v
o
l
.
2
7
,
n
o
.
1
,
p
p
.
1
8
9
–
1
9
7
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
e
c
s
.
v
2
7
.
i
1
.
p
p
1
8
9
-
1
9
7
.
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
7
,
No
.
1
,
J
an
u
ar
y
20
2
5
:
250
-
2
5
6
256
[
2
4
]
M
.
K
.
N
y
e
i
n
a
n
d
K
.
M
.
S
o
e
,
“
S
o
u
r
c
e
si
d
e
p
r
e
-
o
r
d
e
r
i
n
g
u
s
i
n
g
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
E
n
g
l
i
sh
-
M
y
a
n
mar
mac
h
i
n
e
t
r
a
n
s
l
a
t
i
o
n
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
E
l
e
c
t
ri
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
(
I
J
E
EC
S
)
,
v
o
l
.
1
1
,
n
o
.
5
,
p
p
.
4
5
1
3
–
4
5
2
1
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
1
i
5
.
p
p
4
5
1
3
-
4
5
2
1
.
[
2
5
]
M
.
A
.
A
y
u
,
T
.
M
a
n
t
o
r
o
,
a
n
d
J
.
A
si
a
n
,
“
Q
u
a
l
i
t
y
t
r
a
n
s
l
a
t
i
o
n
e
n
h
a
n
c
e
me
n
t
u
si
n
g
s
e
q
u
e
n
c
e
k
n
o
w
l
e
d
g
e
a
n
d
p
r
u
n
i
n
g
i
n
s
t
a
t
i
st
i
c
a
l
mac
h
i
n
e
t
r
a
n
s
l
a
t
i
o
n
,
”
T
e
l
k
o
m
n
i
k
a
(
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
C
o
m
p
u
t
i
n
g
El
e
c
t
r
o
n
i
c
s
a
n
d
C
o
n
t
ro
l
)
,
v
o
l
.
1
6
,
n
o
.
2
,
p
p
.
7
1
8
–
7
2
7
,
2
0
1
8
,
d
o
i
:
1
0
.
1
2
9
2
8
/
T
EL
K
O
M
N
I
K
A
.
v
1
6
i
2
.
8
6
8
7
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Rite
sh
K
u
m
a
r
D
wiv
e
d
i
he
jo
in
t
d
irec
to
r
(IT
)
i
n
G
o
v
t.
o
f
I
n
d
ia.
He
p
u
rsin
g
Ph
.
D
.
fro
m
De
p
a
rtme
n
t
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
in
e
e
ri
n
g
,
S
h
a
rd
a
Un
i
v
e
rsity
,
G
re
a
ter
No
id
a
,
I
n
d
ia.
He
h
a
s
m
o
re
t
h
a
n
1
8
y
e
a
rs
o
f
e
x
p
e
rien
c
e
i
n
so
f
twa
re
a
p
p
li
c
a
ti
o
n
d
e
v
e
lo
p
m
e
n
t.
He
h
a
s
k
e
e
n
in
tere
st
in
d
e
v
e
l
o
p
i
n
g
AI
a
n
d
M
L
a
p
p
li
c
a
ti
o
n
s.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
rit
e
sh
.
d
wi
v
e
d
i@n
ic.i
n
.
Pa
r
m
a
Na
n
d
p
ro
fe
ss
o
r
(Dr.)
P
a
rm
a
Na
n
d
is
P
ro
Vic
e
Ch
a
n
c
e
ll
o
r
a
n
d
P
ro
fe
ss
o
r
De
p
t.
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
,
S
h
a
r
d
a
Un
iv
e
rsity
G
re
a
ter
No
id
a
.
He
h
a
s
o
v
e
r
2
9
y
e
a
rs
o
f
tea
c
h
in
g
,
i
n
d
u
stry
a
n
d
re
se
a
rc
h
e
x
p
e
rien
c
e
.
He
h
a
s
e
x
p
e
rti
se
in
W
irele
ss
a
n
d
S
e
n
so
r
Ne
two
r
k
,
Cr
y
p
t
o
g
ra
p
h
y
,
Alg
o
rit
h
m
a
n
d
Co
m
p
u
ter
G
ra
p
h
i
c
s.
He
h
a
s
e
a
rn
e
d
h
is
P
h
.
D
.
fro
m
IIT
Ro
o
rk
e
e
,
M
.
Tec
h
.
a
n
d
B.
Tec
h
.
i
n
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
E
n
g
i
n
e
e
rin
g
fro
m
IIT
De
lh
i.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
p
a
rm
a
.
n
a
n
d
@s
h
a
rd
a
.
a
c
.
in
.
O
m
Pa
l
re
c
e
iv
e
d
B.
E.
i
n
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
E
n
g
i
n
e
e
rin
g
fro
m
Dr.
B
.
R.
Am
b
e
d
k
a
r
Un
i
v
e
rsity
,
Ag
ra
,
M
BA(O
&
M
)
fro
m
IG
NO
U,
M
S
(Re
se
a
rc
h
)
in
f
ield
o
f
Cry
p
t
o
g
ra
p
h
y
fr
o
m
IIT
Bo
m
b
a
y
a
n
d
P
h
.
D.
in
fiel
d
o
f
Cy
b
e
r
S
e
c
u
rit
y
fro
m
Ja
m
ia
M
il
li
a
Isla
m
ia,
Ne
w
De
lh
i.
P
re
se
n
tl
y
h
e
is
P
u
rsu
i
n
g
“
P
G
Dip
lo
m
a
p
ro
g
ra
m
(o
n
li
n
e
m
o
d
e
)
i
n
C
y
b
e
r
Law
,
Cy
b
e
r
Crime
In
v
e
sti
g
a
ti
o
n
a
n
d
Di
g
it
a
l
F
o
re
n
sic
s”
fro
m
Na
ti
o
n
a
l
Law
In
stit
u
te
Un
iv
e
rsity
,
Bh
o
p
a
l.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
o
p
a
l@c
s.d
u
.
a
c
.
i
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.