I
nte
rna
t
io
na
l J
o
urna
l o
f
Rec
o
nfig
ura
ble a
nd
E
m
be
dd
e
d Sy
s
t
e
m
s
(
I
J
R
E
S)
Vo
l.
14
,
No
.
1
,
Ma
r
ch
20
25
,
p
p
.
2
4
3
~
2
5
3
I
SS
N:
2089
-
4864
,
DOI
:
1
0
.
1
1
5
9
1
/i
j
r
es
.
v
1
4
.
i
1
.
p
p
2
4
3
-
253
243
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
r
es.ia
esco
r
e.
co
m
Self
-
a
t
tent
io
n
enc
o
der
-
deco
der
w
it
h
m
o
d
el ada
ptati
o
n f
o
r
trans
literation
an
d t
ra
nsla
tion
tas
ks in regio
na
l lang
ua
g
e
Sh
a
ntha
la
Na
g
a
ra
j
a
,
K
ira
n
Y
.
Cha
nd
a
pp
a
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
G
l
o
b
a
l
A
c
a
d
e
my
o
f
T
e
c
h
n
o
l
o
g
y
,
B
a
n
g
a
l
o
r
e
,
I
n
d
i
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Oct
3
,
2
0
2
3
R
ev
i
s
ed
J
u
l 1
0
,
2
0
2
4
A
cc
ep
ted
A
u
g
1
2
,
2
0
2
4
T
h
e
re
c
e
n
t
a
d
v
a
n
c
e
m
e
n
ts
in
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
(N
L
P
)
h
a
v
e
h
ig
h
li
g
h
ted
t
h
e
sig
n
if
ica
n
c
e
o
f
in
teg
ra
ti
n
g
m
a
c
h
in
e
tran
slit
e
ra
ti
o
n
w
it
h
tran
sla
ti
o
n
f
o
r
e
n
h
a
n
c
e
d
la
n
g
u
a
g
e
se
rv
ic
e
s,
p
a
rti
c
u
larly
in
th
e
c
o
n
tex
t
o
f
re
g
io
n
a
l
lan
g
u
a
g
e
s.
T
h
is
p
a
p
e
r
in
tr
o
d
u
c
e
s
a
n
o
v
e
l
n
e
u
ra
l
n
e
tw
o
rk
a
rc
h
it
e
c
tu
re
th
a
t
lev
e
ra
g
e
s
a
se
lf
-
a
tt
e
n
ti
o
n
m
e
c
h
a
n
is
m
to
c
re
a
te
a
n
a
u
to
-
e
n
c
o
d
e
r
w
it
h
o
u
t
t
h
e
n
e
e
d
f
o
r
i
tera
ti
v
e
o
r
c
o
n
v
o
l
u
ti
o
n
a
l
p
ro
c
e
ss
e
s.
T
h
e
se
lf
-
a
tt
e
n
ti
o
n
m
e
c
h
a
n
is
m
o
p
e
ra
tes
o
n
p
ro
jec
ti
o
n
m
a
tri
c
e
s,
f
e
a
tu
re
m
a
tri
c
e
s,
a
n
d
targ
e
t
q
u
e
ries
,
u
ti
li
z
i
n
g
th
e
S
o
f
t
m
a
x
f
u
n
c
ti
o
n
f
o
r
o
p
ti
m
iza
t
io
n
.
T
h
e
in
tro
d
u
c
ti
o
n
o
f
th
e
se
l
f
-
a
tt
e
n
ti
o
n
e
n
c
o
d
e
r
-
d
e
c
o
d
e
r
w
it
h
m
o
d
e
l
a
d
a
p
tatio
n
(S
A
EDM
)
re
p
re
se
n
ts
a
b
re
a
k
th
ro
u
g
h
,
m
a
rk
in
g
a
su
b
sta
n
ti
a
l
e
n
h
a
n
c
e
m
e
n
t
in
tran
slit
e
ra
ti
o
n
a
n
d
tra
n
sla
ti
o
n
a
c
c
u
ra
c
y
o
v
e
r
p
re
v
io
u
s
m
e
th
o
d
o
l
o
g
ies
.
T
h
is
in
n
o
v
a
ti
v
e
a
p
p
ro
a
c
h
e
m
p
lo
y
s
b
o
t
h
stu
d
e
n
t
a
n
d
tea
c
h
e
r
m
o
d
e
ls,
w
it
h
t
h
e
stu
d
e
n
t
m
o
d
e
l'
s
lo
ss
c
a
lcu
late
d
th
ro
u
g
h
th
e
p
r
o
b
a
b
il
it
ies
a
n
d
p
re
d
ictio
n
lab
e
ls
v
ia
th
e
n
e
g
a
ti
v
e
lo
g
e
n
tro
p
y
f
u
n
c
ti
o
n
.
T
h
e
p
ro
p
o
se
d
a
rc
h
i
tec
tu
re
is
d
isti
n
c
ti
v
e
ly
d
e
sig
n
e
d
a
t
th
e
c
h
a
ra
c
ter
lev
e
l,
in
c
o
rp
o
ra
ti
n
g
a
w
o
rd
-
t
o
-
w
o
rd
e
m
b
e
d
d
in
g
f
ra
m
e
w
o
rk
,
a
b
e
a
m
se
a
rc
h
a
lg
o
rit
h
m
f
o
r
se
n
ten
c
e
g
e
n
e
ra
ti
o
n
,
a
n
d
a
b
in
a
ry
c
las
sif
ier
w
it
h
in
th
e
e
n
c
o
d
e
r
-
d
e
c
o
d
e
r
stru
c
t
u
re
to
e
n
su
re
th
e
u
n
i
q
u
e
n
e
ss
o
f
th
e
c
o
n
ten
t.
T
h
e
e
ff
e
c
ti
v
e
n
e
ss
o
f
th
e
p
ro
p
o
se
d
m
o
d
e
l
is
v
a
li
d
a
ted
th
ro
u
g
h
c
o
m
p
re
h
e
n
si
v
e
e
v
a
lu
a
ti
o
n
s
u
sin
g
tra
n
slit
e
r
a
ti
o
n
a
n
d
tran
sla
ti
o
n
d
a
tas
e
ts
i
n
Ka
n
n
a
d
a
a
n
d
Hin
d
i
la
n
g
u
a
g
e
s,
d
e
m
o
n
stra
ti
n
g
it
s
su
p
e
rio
r
p
e
rf
o
rm
a
n
c
e
c
o
m
p
a
re
d
t
o
e
x
isti
n
g
m
o
d
e
ls.
K
ey
w
o
r
d
s
:
Au
to
-
e
n
co
d
er
Natu
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
Neu
r
al
n
et
w
o
r
k
Self
-
atte
n
tio
n
en
co
d
er
-
d
ec
o
d
er
w
it
h
m
o
d
el
ad
ap
tatio
n
Self
-
atte
n
tio
n
m
ec
h
a
n
is
m
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
:
Sh
a
n
t
h
ala
Nag
ar
aj
a
Dep
ar
t
m
en
t o
f
I
n
f
o
r
m
atio
n
Sc
ien
ce
an
d
E
n
g
i
n
ee
r
in
g
,
Glo
b
al
A
ca
d
e
m
y
o
f
T
ec
h
n
o
lo
g
y
R
aj
ar
aj
esh
w
ar
in
a
g
ar
,
(
o
f
f
M
y
s
o
r
e
R
o
ad
)
,
I
d
ea
l H
o
m
es T
o
w
n
s
h
ip
,
B
an
g
alo
r
e
-
5
6
0
0
9
8
,
Ka
r
n
atak
a,
I
n
d
ia
E
m
ail: s
h
a
n
t
h
ala_
1
2
@
r
ed
if
f
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
I
n
to
d
ay
's
m
o
d
er
n
w
o
r
ld
,
ef
f
ec
tiv
e
cr
o
s
s
-
lan
g
u
ag
e
co
m
m
u
n
ica
tio
n
an
d
i
n
f
o
r
m
atio
n
ac
ce
s
s
h
a
v
e
b
ec
o
m
e
i
n
d
is
p
en
s
ab
le.
T
h
e
r
a
p
id
ex
p
an
s
io
n
o
f
th
e
i
n
ter
n
et
h
as
led
to
a
v
ast
a
m
o
u
n
t
o
f
d
ig
ital
in
f
o
r
m
at
io
n
b
ein
g
g
e
n
er
ated
in
v
ar
io
u
s
la
n
g
u
a
g
es,
in
cl
u
d
i
n
g
Hin
d
i
a
n
d
E
n
g
l
is
h
[
1
]
.
C
r
o
s
s
-
lan
g
u
ag
e
i
n
f
o
r
m
atio
n
r
etr
ie
v
al
(
C
L
I
R
)
en
ab
les
u
s
er
s
to
f
i
n
d
r
elev
an
t
d
ata
in
d
i
f
f
er
en
t
lan
g
u
ag
es,
m
ak
i
n
g
it
p
ar
tic
u
lar
l
y
cr
u
cial
f
o
r
lan
g
u
a
g
es
lik
e
Hin
d
i
an
d
E
n
g
lis
h
,
w
h
i
ch
h
av
e
d
is
ti
n
ct
lin
g
u
is
t
ic
f
r
a
m
e
w
o
r
k
s
an
d
w
r
iti
n
g
s
y
s
te
m
s
[
2
]
.
T
h
is
s
tu
d
y
in
v
e
s
t
i
g
ates
t
h
e
p
er
f
o
r
m
a
n
ce
a
n
d
ef
f
ec
ti
v
e
n
es
s
o
f
C
L
I
R
tec
h
n
iq
u
e
s
,
s
p
ec
i
f
icall
y
f
o
cu
s
i
n
g
o
n
Hi
n
d
i
-
to
-
E
n
g
lis
h
tr
an
s
liter
at
io
n
a
n
d
tr
an
s
latio
n
[
3
]
–
[
5
]
.
Hin
d
i
h
o
ld
s
s
i
g
n
i
f
ic
an
t
i
m
p
o
r
tan
ce
as
it
i
s
s
p
o
k
e
n
b
y
o
v
er
4
0
%
o
f
I
n
d
ian
s
a
n
d
ac
ts
as
a
lin
g
u
a
f
r
an
ca
,
b
r
in
g
i
n
g
to
g
eth
er
p
eo
p
le
f
r
o
m
d
iv
er
s
e
li
n
g
u
i
s
tic
an
d
g
eo
g
r
ap
h
ical
b
ac
k
g
r
o
u
n
d
s
[
6
]
.
B
esid
es
E
n
g
lis
h
,
Hin
d
i
is
o
n
e
o
f
th
e
2
2
o
f
f
icia
ll
y
r
ec
o
g
n
ized
lan
g
u
a
g
es
in
I
n
d
ia
[
7
]
,
[
8
]
.
Du
e
to
its
o
f
f
icial
s
tat
u
s
,
a
co
n
s
id
er
ab
le
a
m
o
u
n
t
o
f
o
f
f
icia
l
p
ap
er
w
o
r
k
,
le
g
al
d
o
cu
m
e
n
ts
,
an
d
co
m
m
u
n
icatio
n
ar
e
p
r
o
d
u
ce
d
in
Hin
d
i,
e
m
p
h
a
s
izin
g
t
h
e
n
ee
d
f
o
r
ef
f
icie
n
t a
n
d
ac
cu
r
ate
C
L
I
R
s
y
s
te
m
s
to
e
n
h
an
ce
ac
ce
s
s
ib
ilit
y
to
s
u
ch
cr
itical
i
n
f
o
r
m
atio
n
[
9
]
,
[
1
0
]
.
Giv
en
its
w
id
e
s
p
r
ea
d
u
s
ag
e,
o
f
f
icial
s
tatu
s
,
an
d
th
e
g
r
o
w
in
g
d
ig
i
tal
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2089
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
1
,
Ma
r
c
h
2025
:
2
4
3
-
253
244
co
n
ten
t,
H
in
d
i
p
la
y
s
a
cr
u
cial
r
o
le
in
C
L
I
R
i
n
I
n
d
ia,
f
ac
ilit
atin
g
i
n
ter
cu
lt
u
r
al
co
m
m
u
n
ica
tio
n
an
d
p
r
o
v
id
i
n
g
ac
ce
s
s
to
o
f
f
i
cial
i
n
f
o
r
m
atio
n
[
1
1
]
,
[
1
2
]
.
Fo
r
Hin
d
i
-
to
-
E
n
g
lis
h
C
L
I
R
,
tr
an
s
liter
atio
n
is
p
ar
ticu
lar
l
y
u
s
e
f
u
l,
as
it
co
n
v
er
ts
tex
t
b
et
w
ee
n
th
e
Dev
a
n
a
g
ar
i
s
cr
ip
t
(
u
s
ed
in
Hin
d
i)
an
d
th
e
L
ati
n
s
cr
ip
t
(
u
s
ed
in
E
n
g
li
s
h
)
w
h
ile
p
r
eser
v
in
g
p
h
o
n
et
ic
s
i
m
ilar
it
y
[
1
3
]
,
[
1
4
]
.
Ho
w
ev
er
,
tr
an
s
liter
atio
n
m
a
y
n
o
t
al
w
a
y
s
ca
p
tu
r
e
th
e
e
x
ac
t
s
e
m
a
n
tic
m
ea
n
in
g
o
f
t
h
e
te
x
t.
On
th
e
o
th
er
h
a
n
d
,
tr
an
s
latio
n
in
v
o
lv
e
s
co
n
v
e
y
i
n
g
t
h
e
m
e
an
in
g
o
f
a
tex
t
f
r
o
m
o
n
e
la
n
g
u
a
g
e
to
an
o
th
er
,
m
ak
in
g
it
m
o
r
e
s
u
itab
le
f
o
r
t
r
an
s
f
er
r
in
g
s
e
m
an
t
ic
in
f
o
r
m
at
io
n
b
et
w
ee
n
lan
g
u
ag
e
s
lik
e
Hin
d
i
an
d
E
n
g
li
s
h
.
A
lt
h
o
u
g
h
tr
an
s
latio
n
-
b
ased
te
ch
n
iq
u
es
h
av
e
s
h
o
w
n
ad
v
an
t
ag
es
o
v
er
tr
an
s
liter
atio
n
al
g
o
r
ith
m
s
i
n
p
r
ev
io
u
s
C
L
I
R
r
esear
ch
,
it
r
e
m
ai
n
s
u
n
ce
r
t
ain
if
t
h
ese
f
i
n
d
i
n
g
s
ap
p
ly
to
a
w
id
e
r
an
g
e
o
f
lan
g
u
ag
e
p
air
s
an
d
d
o
m
ain
s
,
esp
ec
iall
y
f
o
r
Hin
d
i
-
to
-
E
n
g
lis
h
C
L
I
R
[
1
5
]
.
R
esear
ch
er
s
an
d
li
n
g
u
is
ts
h
av
e
ex
p
lo
r
ed
v
ar
io
u
s
ch
alle
n
g
es
r
elate
d
to
tr
an
s
liter
atio
n
an
d
tr
an
s
latio
n
m
o
d
el
s
.
T
r
an
s
latio
n
s
tu
d
ie
s
h
a
v
e
ex
a
m
i
n
ed
asp
ec
ts
s
u
c
h
as
c
u
lt
u
r
al
b
ac
k
g
r
o
u
n
d
,
tr
an
s
latio
n
ac
cu
r
ac
y
,
an
d
t
h
e
i
m
p
ac
t
o
f
tr
an
s
latio
n
o
n
li
ter
ar
y
w
o
r
k
s
.
On
t
h
e
o
th
er
h
a
n
d
,
m
o
s
t
o
f
th
e
f
o
cu
s
o
n
tr
an
s
lit
er
atio
n
s
tu
d
ie
s
h
a
s
ce
n
ter
ed
ar
o
u
n
d
d
ev
elo
p
in
g
t
o
o
l
s
an
d
alg
o
r
ith
m
s
f
o
r
tex
t
tr
an
s
latio
n
ac
r
o
s
s
s
cr
ip
ts
.
Ho
w
e
v
er
,
th
e
r
elati
v
e
ef
f
ec
ts
o
f
tr
an
s
liter
atio
n
a
n
d
tr
an
s
latio
n
o
n
cr
o
s
s
-
li
n
g
u
i
s
tic
co
m
m
u
n
icatio
n
i
n
th
e
co
n
tex
t
o
f
r
eg
io
n
al
lan
g
u
a
g
es
li
k
e
Kan
n
ad
a
,
Hin
d
i
an
d
th
eir
in
ter
ac
tio
n
w
it
h
E
n
g
lis
h
h
a
v
e
n
o
t
b
e
en
ex
te
n
s
i
v
el
y
i
n
v
esti
g
ated
[
1
6
]
.
R
ec
en
t
ad
v
an
ce
m
e
n
ts
in
d
ee
p
lear
n
in
g
a
n
d
n
at
u
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
(
N
L
P
)
h
a
v
e
s
i
g
n
if
ica
n
tl
y
i
m
p
r
o
v
ed
m
ac
h
in
e
tr
a
n
s
latio
n
an
d
tr
an
s
liter
atio
n
f
o
r
r
eg
io
n
al
lan
g
u
ag
es
li
k
e
Kan
n
ad
a.
T
h
ese
d
ev
elo
p
m
e
n
ts
h
av
e
f
o
cu
s
ed
o
n
o
v
er
co
m
i
n
g
r
eso
u
r
ce
co
n
s
tr
ai
n
ts
a
n
d
lan
g
u
ag
e
co
m
p
lex
i
t
y
w
h
i
le
en
h
a
n
cin
g
t
h
e
ac
cu
r
ac
y
,
f
l
u
en
c
y
,
an
d
ef
f
ec
t
iv
e
n
es
s
o
f
lan
g
u
ag
e
m
o
d
els.
Neu
r
al
m
ac
h
i
n
e
tr
an
s
lat
io
n
(
NM
T
)
m
o
d
els
u
tili
zi
n
g
d
ee
p
lear
n
in
g
tech
n
iq
u
es
h
a
v
e
p
r
o
v
en
to
b
e
m
o
r
e
p
r
ec
is
e
an
d
r
eliab
le
th
an
tr
ad
itio
n
al
s
tati
s
ti
ca
l
an
d
r
u
le
-
b
ased
ap
p
r
o
ac
h
es
[
1
7
]
,
[
1
8
]
.
Fu
r
th
er
m
o
r
e,
lar
g
e
-
s
ca
le
p
r
e
-
tr
ain
ed
lan
g
u
a
g
e
m
o
d
el
s
li
k
e
b
i
d
ir
ec
tio
n
al
en
co
d
er
r
ep
r
esen
tatio
n
s
f
r
o
m
tr
an
s
f
o
r
m
er
s
(
B
E
R
T
)
an
d
g
e
n
er
ativ
e
p
r
e
-
tr
ain
ed
tr
an
s
f
o
r
m
er
s
(
GP
T
)
,
alo
n
g
w
it
h
t
h
eir
m
u
ltil
i
n
g
u
a
l
v
er
s
io
n
s
,
h
a
v
e
d
em
o
n
s
tr
ated
r
e
m
ar
k
ab
le
p
er
f
o
r
m
an
ce
in
v
ar
io
u
s
N
L
P
tas
k
s
,
i
n
clu
d
i
n
g
tr
an
s
latio
n
an
d
tr
an
s
li
ter
atio
n
w
h
en
tr
ai
n
e
d
o
n
Hin
d
i a
n
d
Kan
n
ad
a
d
ata.
I
n
co
n
cl
u
s
io
n
,
ef
f
ec
ti
v
e
cr
o
s
s
-
lan
g
u
a
g
e
co
m
m
u
n
icatio
n
an
d
ac
ce
s
s
to
i
n
f
o
r
m
atio
n
ar
e
ess
e
n
tial i
n
t
h
e
d
ig
ital
er
a.
T
h
e
s
tu
d
y
ai
m
s
to
in
v
est
ig
ate
th
e
p
er
f
o
r
m
an
ce
an
d
e
f
f
ec
tiv
e
n
es
s
o
f
Hin
d
i
-
to
-
E
n
g
lis
h
tr
an
s
liter
at
io
n
an
d
tr
a
n
s
lat
io
n
tech
n
iq
u
es
i
n
C
L
I
R
.
C
o
n
s
id
er
in
g
t
h
e
s
ig
n
i
f
ica
n
ce
o
f
Hi
n
d
i
in
I
n
d
ia,
ac
cu
r
ate
C
L
I
R
s
y
s
te
m
s
ar
e
cr
u
cial
f
o
r
f
o
s
ter
i
n
g
in
ter
cu
lt
u
r
al
co
m
m
u
n
icatio
n
a
n
d
b
r
ea
k
in
g
d
o
w
n
lan
g
u
a
g
e
b
ar
r
ier
s
in
th
e
co
u
n
tr
y
.
R
ec
e
n
t
ad
v
an
ce
m
en
ts
i
n
d
ee
p
lear
n
in
g
an
d
NL
P
h
av
e
s
h
o
w
n
p
r
o
m
i
s
i
n
g
r
esu
lt
s
in
en
h
a
n
cin
g
tr
an
s
latio
n
an
d
tr
an
s
liter
atio
n
m
o
d
el
s
,
p
r
o
v
id
in
g
b
etter
ac
ce
s
s
to
in
f
o
r
m
atio
n
ac
r
o
s
s
d
i
f
f
er
en
t la
n
g
u
ag
e
s
[
1
9
]
.
I
n
th
is
n
o
v
el
ap
p
r
o
ac
h
,
t
w
o
p
r
e
-
tr
ain
ed
B
E
R
T
m
o
d
els
f
r
o
m
d
if
f
er
e
n
t
d
o
m
ai
n
s
ar
e
s
ea
m
les
s
l
y
in
te
g
r
ated
in
to
a
s
eq
u
en
ce
-
to
-
s
eq
u
en
ce
m
o
d
el
u
s
in
g
ad
ap
t
er
m
o
d
u
les
[
2
0
]
.
T
h
ese
ad
ap
ter
s
ar
e
in
tr
o
d
u
ce
d
b
et
w
ee
n
B
E
R
T
lay
er
s
an
d
f
in
e
-
t
u
n
ed
,
w
h
ile
late
n
t
v
ar
iab
les
d
u
r
in
g
f
i
n
e
-
t
u
n
i
n
g
d
eter
m
i
n
e
w
h
ic
h
la
y
er
s
u
t
ilize
th
e
ad
ap
ter
s
.
T
h
is
in
telli
g
en
t
ad
ap
tatio
n
s
ig
n
if
ican
tl
y
e
n
h
an
ce
s
p
ar
a
m
eter
ef
f
icie
n
c
y
a
n
d
d
ec
o
d
in
g
s
p
ee
d
.
T
esti
n
g
t
h
e
p
r
o
p
o
s
ed
f
r
am
e
w
o
r
k
ag
ai
n
s
t
v
ar
io
u
s
NM
T
ch
allen
g
e
s
d
e
m
o
n
s
tr
ates it
s
ef
f
ec
t
iv
en
es
s
.
T
h
e
p
ap
er
in
tr
o
d
u
ce
s
th
e
iter
a
tiv
e
a
n
d
len
g
th
-
ad
j
u
s
tab
le
n
o
n
-
au
to
r
eg
r
ess
i
v
e
d
ec
o
d
er
(
I
L
AND)
[
2
1
]
,
a
u
n
iq
u
e
p
ar
ad
ig
m
f
o
r
m
ac
h
in
e
tr
an
s
latio
n
.
I
L
AND
e
m
p
lo
y
s
a
len
g
t
h
-
ad
j
u
s
tab
le
n
o
n
-
au
to
r
eg
r
ess
i
v
e
d
ec
o
d
er
th
at
u
s
es
a
h
id
d
en
la
n
g
u
a
g
e
m
o
d
el
to
p
r
ev
en
t
lo
w
-
co
n
f
i
d
en
ce
to
k
en
cr
ea
tio
n
s
w
h
ile
m
a
x
i
m
izi
n
g
tar
g
et
s
en
te
n
ce
len
g
t
h
.
C
o
m
p
r
is
i
n
g
th
r
ee
s
u
b
-
m
o
d
u
le
s
-
to
k
e
n
m
ask
er
,
len
g
t
h
m
o
d
u
lato
r
,
an
d
to
k
en
g
en
er
ato
r
-
I
L
A
ND
co
llab
o
r
ativ
el
y
ac
h
ie
v
es
its
o
b
j
ec
tiv
es.
T
h
e
to
k
en
m
a
s
k
er
an
d
to
k
en
g
e
n
er
ato
r
h
an
d
le
th
e
m
as
k
ed
lan
g
u
a
g
e
m
o
d
el,
w
h
ile
t
h
e
le
n
g
th
m
o
d
u
lato
r
o
p
ti
m
izes
s
en
te
n
ce
le
n
g
t
h
.
T
h
e
s
eq
u
e
n
ce
-
to
-
s
eq
u
en
ce
tr
ai
n
i
n
g
o
f
th
e
tr
an
s
latio
n
m
o
d
el
i
s
ef
f
ec
t
iv
el
y
d
e
m
o
n
s
tr
ated
.
T
h
e
co
n
c
u
r
r
en
t
tr
ain
in
g
o
f
t
h
e
len
g
t
h
m
o
d
u
lato
r
an
d
to
k
e
n
g
en
er
ato
r
,
w
h
ic
h
s
h
ar
e
s
i
m
ilar
s
tr
u
ctu
r
e
s
,
co
n
tr
ib
u
tes
to
th
e
m
o
d
el
's
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
co
m
p
ar
ed
to
o
th
er
n
o
n
-
au
to
r
eg
r
ess
i
v
e
d
ec
o
d
er
s
,
p
r
o
v
id
in
g
e
m
p
ir
ical
v
alid
atio
n
.
An
o
th
er
p
r
o
p
o
s
ed
tech
n
iq
u
e
[
2
2
]
in
tr
o
d
u
ce
s
k
n
o
w
led
g
e
-
a
w
ar
e
NM
T
,
in
co
r
p
o
r
atin
g
ad
d
itio
n
al
lan
g
u
a
g
e
p
r
o
p
er
ties
at
t
h
e
wo
r
d
lev
el
u
s
i
n
g
r
ec
u
r
r
en
t
n
e
u
r
al
n
et
w
o
r
k
s
(
R
NN)
.
T
h
e
s
e
n
ten
ce
le
v
el
u
s
e
s
a
n
R
NN
e
n
co
d
er
to
en
co
d
e
th
ese
w
o
r
d
-
le
v
el
f
ea
t
u
r
e
u
n
it
s
.
A
d
d
itio
n
all
y
,
a
k
n
o
w
led
g
e
g
ate
an
d
an
atte
n
tio
n
g
at
e
co
n
tr
o
l
th
e
q
u
an
tit
y
o
f
i
n
f
o
r
m
atio
n
f
r
o
m
v
ar
io
u
s
s
o
u
r
ce
s
to
ass
is
t
in
d
ec
o
d
in
g
an
d
co
n
s
tr
u
ctin
g
tar
g
et
w
o
r
d
s
.
T
h
is
ap
p
r
o
ac
h
p
r
o
v
es
ef
f
icie
n
t
in
e
n
h
a
n
ci
n
g
NM
T
p
er
f
o
r
m
an
ce
b
y
le
v
er
ag
in
g
p
r
io
r
tr
an
s
lat
io
n
k
n
o
w
led
g
e
f
r
o
m
t
h
e
s
o
u
r
ce
s
id
e
o
f
NM
T
'
s
tr
ain
in
g
p
ip
eli
n
e.
I
t
en
ab
le
s
NM
T
to
ef
f
ec
ti
v
el
y
i
n
co
r
p
o
r
ate
p
ast
tr
an
s
latio
n
in
f
o
r
m
atio
n
a
n
d
cr
o
s
s
-
lan
g
u
ag
e
tr
an
s
latio
n
d
ata,
r
esu
lti
n
g
in
i
m
p
r
o
v
ed
tr
an
s
la
t
io
n
ac
c
u
r
ac
y
an
d
q
u
alit
y
.
T
o
en
h
an
ce
NM
T
m
o
d
el
p
er
f
o
r
m
a
n
ce
f
o
r
co
n
s
tr
ain
ed
r
eso
u
r
ce
s
an
d
lan
g
u
a
g
e
p
air
s
,
a
n
o
v
el
s
tr
ateg
y
is
p
r
o
p
o
s
ed
f
o
r
s
tan
d
ar
d
izin
g
NM
T
m
o
d
el
tr
ain
in
g
[
6
]
.
T
h
is
ap
p
r
o
ac
h
in
v
o
lv
e
s
tr
ain
i
n
g
th
e
m
o
d
el
to
p
r
ed
ict
tar
g
et
tr
ain
i
n
g
te
x
ts
u
s
i
n
g
w
o
r
d
an
d
s
en
ten
ce
e
m
b
ed
d
i
n
g
s
as
w
e
ll
as
ca
teg
o
r
ical
o
u
tp
u
ts
(
i.e
.
,
w
o
r
d
s
eq
u
en
ce
s
)
.
B
y
p
r
e
-
tr
ai
n
i
n
g
wo
r
d
an
d
p
h
r
ase
e
m
b
ed
d
in
g
s
o
n
s
u
b
s
ta
n
tial
m
o
n
o
lin
g
u
al
d
at
a
co
r
p
o
r
a
,
th
e
m
o
d
el
g
ain
s
t
h
e
ab
ilit
y
to
g
en
er
alize
b
ey
o
n
d
th
e
tr
an
s
latio
n
tr
ain
i
n
g
s
et,
i
m
p
r
o
v
i
n
g
tr
a
n
s
lat
io
n
ac
c
u
r
ac
y
.
A
u
n
iq
u
e
i
m
p
r
o
v
e
m
e
n
t
to
g
en
er
ativ
e
ad
v
er
s
ar
ial
n
et
w
o
r
k
s
-
n
eu
r
al
m
ac
h
in
e
tr
an
s
latio
n
(
G
A
N
-
NM
T
)
is
s
u
g
g
es
ted
b
y
i
n
co
r
p
o
r
atin
g
d
ee
p
r
ein
f
o
r
ce
m
en
t le
ar
n
i
n
g
-
b
ased
atten
tio
n
o
p
ti
m
izatio
n
in
t
o
th
e
g
e
n
er
ato
r
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
S
elf
-
a
tten
tio
n
en
co
d
er
-
d
ec
o
d
e
r
w
ith
mo
d
el
a
d
a
p
ta
tio
n
fo
r
tr
a
n
s
liter
a
tio
n
a
n
d
…
(
S
h
a
n
th
a
l
a
N
a
g
a
r
a
ja
)
245
a
co
n
v
o
lu
t
io
n
al
n
eu
r
al
n
et
w
o
r
k
in
to
th
e
d
is
cr
i
m
i
n
ato
r
[
2
3
]
.
T
h
is
en
h
a
n
ce
m
e
n
t
ad
d
r
ess
e
s
th
e
ch
a
llen
g
e
o
f
u
n
u
s
u
al
ter
m
s
in
lo
w
r
eso
u
r
ce
lan
g
u
ag
e
s
(
L
R
L
s
)
an
d
i
m
p
r
o
v
es
t
h
e
ass
i
m
ilat
io
n
o
f
s
o
u
r
ce
s
e
n
ten
c
e
r
ep
r
esen
tatio
n
s
.
A
d
d
itio
n
al
l
y
,
a
n
o
v
el
j
o
in
t
em
b
ed
d
in
g
o
f
s
u
b
w
o
r
d
s
an
d
s
u
b
-
p
h
o
n
etic
r
ep
r
esen
tatio
n
s
o
f
s
en
te
n
ce
s
i
s
u
ti
lized
as
G
A
N
i
n
p
u
t,
en
ab
li
n
g
th
e
m
o
d
el
to
le
ar
n
s
u
p
er
io
r
r
ep
r
esen
tatio
n
s
a
n
d
g
en
er
ate
co
n
tex
t
v
ec
to
r
s
m
o
r
e
s
u
itab
le
f
o
r
L
R
L
s
th
a
n
tr
ad
itio
n
al
tec
h
n
iq
u
es.
A
m
u
lti
-
ta
s
k
m
u
lti
-
s
ta
g
e
tr
an
s
it
io
n
al
(
MM
T
)
tr
ain
in
g
f
r
a
m
e
w
o
r
k
is
p
r
o
p
o
s
ed
[
2
4
]
,
u
tili
zi
n
g
a
b
ilin
g
u
al
co
n
v
er
s
atio
n
tr
an
s
la
tio
n
d
ataset
an
d
e
x
tr
a
m
o
n
o
l
in
g
u
al
co
n
v
er
s
atio
n
s
.
T
h
is
f
r
a
m
e
w
o
r
k
in
v
o
lv
e
s
th
r
ee
s
tep
s
:
s
e
n
te
n
ce
-
le
v
el
p
r
e
-
tr
ain
in
g
o
n
a
s
izab
le
p
ar
all
el
co
r
p
u
s
,
in
ter
m
ed
iate
tr
ai
n
i
n
g
w
it
h
ad
d
itio
n
al
m
o
n
o
li
n
g
u
al
co
n
v
er
s
atio
n
s
a
n
d
u
n
iq
u
e
ta
s
k
s
(
u
tter
an
ce
an
d
s
p
ea
k
er
d
etec
tio
n
)
to
m
o
d
el
co
n
v
er
s
atio
n
co
h
er
en
ce
an
d
s
p
ea
k
er
ch
ar
ac
ter
is
tics
,
a
n
d
co
n
tex
t
-
a
w
ar
e
f
in
e
-
t
u
n
in
g
w
i
th
a
g
r
ad
u
al
tr
an
s
itio
n
.
T
h
is
in
cr
e
m
e
n
tal
tr
an
s
itio
n
s
tr
ate
g
y
s
m
o
o
t
h
l
y
s
w
i
tch
e
s
m
o
n
o
lin
g
u
al
co
n
v
er
s
at
io
n
s
to
m
u
ltil
i
n
g
u
al
o
n
es,
f
ac
ilit
at
in
g
a
m
o
r
e
r
ef
i
n
ed
tr
ai
n
in
g
p
r
o
ce
s
s
.
T
o
aid
NM
T
s
y
s
te
m
s
,
a
s
tr
ai
g
h
t
f
o
r
w
ar
d
a
n
d
u
s
ef
u
l
m
o
d
el
o
f
th
e
p
o
te
n
tial
co
s
t
o
f
ea
ch
tar
g
et
w
o
r
d
i
s
in
tr
o
d
u
ce
d
[
2
5
]
.
T
h
is
m
o
d
el
le
ar
n
s
a
r
ep
r
esen
tat
io
n
o
f
f
u
t
u
r
e
co
s
ts
b
ased
o
n
t
h
e
p
r
ev
io
u
s
l
y
cr
ea
ted
tar
g
et
ter
m
an
d
its
co
n
tex
t,
w
h
ich
a
s
s
i
s
ts
in
NM
T
m
o
d
el
tr
ain
in
g
.
D
u
r
i
n
g
d
ec
o
d
in
g
,
t
h
e
lear
n
ed
r
ep
r
esen
tat
io
n
o
f
f
u
t
u
r
e
co
s
ts
in
t
h
e
cu
r
r
e
n
t
ti
m
e
p
h
ase
is
u
til
ized
to
p
r
o
d
u
ce
th
e
n
e
x
t
tar
g
et
w
o
r
d
,
en
h
an
ci
n
g
tr
an
s
l
atio
n
ac
cu
r
ac
y
.
R
esear
ch
f
o
cu
s
es
o
n
d
ee
p
lear
n
in
g
-
b
ased
Hin
d
i
-
to
-
E
n
g
li
s
h
tr
an
s
liter
at
io
n
a
n
d
tr
an
s
lat
io
n
in
C
L
I
R
.
Ob
j
ec
tiv
e:
i
m
p
r
o
v
e
i
n
f
o
r
m
ati
o
n
ac
ce
s
s
a
n
d
cr
o
s
s
-
c
u
lt
u
r
al
c
o
m
m
u
n
icatio
n
.
De
v
elo
p
ac
cu
r
ate,
ef
f
icien
t
C
L
I
R
s
y
s
te
m
s
u
s
i
n
g
N
L
P
ad
v
an
c
e
m
en
ts
.
B
en
e
f
it
i
n
d
u
s
tr
ie
s
(
b
u
s
in
e
s
s
,
h
ea
lth
ca
r
e,
an
d
e
d
u
ca
tio
n
)
n
ee
d
i
n
g
m
u
ltil
i
n
g
u
al
i
n
f
o
r
m
atio
n
.
I
n
v
esti
g
a
te
d
ee
p
lear
n
in
g
m
o
d
el
ad
ap
tab
ilit
y
an
d
d
o
m
ai
n
-
s
p
ec
if
ic
ef
f
ec
ti
v
e
n
es
s
.
E
n
h
a
n
ce
in
ter
c
u
lt
u
r
al
u
n
d
er
s
t
an
d
in
g
an
d
aid
tr
av
el,
h
o
s
p
it
alit
y
,
an
d
co
m
m
er
cial
s
ec
to
r
s
.
T
r
an
s
latio
n
an
d
tr
an
s
liter
at
io
n
e
n
ab
le
m
ar
k
et
ex
p
an
s
io
n
a
n
d
ec
o
n
o
m
ic
g
r
o
w
t
h
i
n
Hi
n
d
i
an
d
Kan
n
ad
a
l
an
g
u
a
g
es.
O
v
er
all,
r
esear
ch
ai
m
s
to
b
r
ea
k
lan
g
u
a
g
e
b
ar
r
ier
s
,
f
o
s
ter
cr
o
s
s
-
li
n
g
u
al
co
ll
ab
o
r
atio
n
,
an
d
f
ac
ilit
ate
e
f
f
ic
ien
t
in
f
o
r
m
atio
n
r
etr
ie
v
al.
No
v
el
s
el
f
-
atte
n
tio
n
e
n
co
d
er
-
d
ec
o
d
er
w
it
h
m
o
d
el
ad
ap
tatio
n
(
S
A
E
DM
)
:
t
h
e
p
r
o
p
o
s
ed
in
tr
o
d
u
ce
s
a
n
eu
r
al
n
et
w
o
r
k
ar
ch
itec
tu
r
e
t
h
at
u
tili
ze
s
s
e
lf
-
atte
n
tio
n
to
c
o
n
s
tr
u
ct
an
en
co
d
er
-
d
ec
o
d
er
w
it
h
o
u
t
t
h
e
n
e
ed
f
o
r
iter
atio
n
s
o
r
co
n
v
o
l
u
tio
n
al
o
p
er
atio
n
s
.
T
h
is
n
o
v
el
ap
p
r
o
ac
h
en
h
a
n
ce
s
t
h
e
e
f
f
icie
n
c
y
a
n
d
ef
f
ec
tiv
e
n
e
s
s
o
f
t
h
e
en
co
d
er
-
d
ec
o
d
er
,
allo
w
i
n
g
f
o
r
m
o
r
e
ac
cu
r
ate
an
d
f
a
s
ter
p
r
o
c
ess
i
n
g
o
f
w
o
r
d
an
d
s
eq
u
e
n
ce
e
m
b
ed
d
i
n
g
s
.
I
m
p
r
o
v
ed
tr
an
s
liter
atio
n
a
n
d
tr
an
s
latio
n
p
er
f
o
r
m
an
ce
:
t
h
e
s
t
u
d
y
d
e
m
o
n
s
tr
ates
t
h
e
s
u
p
er
io
r
it
y
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
o
v
er
e
x
is
ti
n
g
s
y
s
te
m
s
i
n
tr
an
s
liter
atio
n
an
d
tr
an
s
latio
n
tas
k
s
f
o
r
Kan
n
ad
a
a
n
d
Hi
n
d
i
lan
g
u
a
g
es.
B
y
in
co
r
p
o
r
atin
g
lex
ical
w
ei
g
h
t
in
g
an
d
f
i
n
e
-
tu
n
in
g
tech
n
iq
u
es,
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
ac
h
ie
v
es
s
u
b
s
ta
n
tial
i
m
p
r
o
v
e
m
en
t
s
in
ac
cu
r
ac
y
,
s
h
o
w
ca
s
in
g
its
p
o
ten
tial
in
e
n
h
a
n
ci
n
g
cr
o
s
s
-
l
in
g
u
al
i
n
f
o
r
m
atio
n
r
etr
iev
al
an
d
co
m
m
u
n
ica
tio
n
.
Stu
d
e
n
t
-
teac
h
er
m
o
d
el
f
o
r
m
o
d
el
c
o
m
p
r
es
s
io
n
an
d
d
o
m
ai
n
ad
ap
tatio
n
:
t
h
e
r
esear
ch
in
tr
o
d
u
ce
s
a
s
tu
d
e
n
t
-
teac
h
er
m
o
d
el
ap
p
r
o
a
ch
f
o
r
m
o
d
el
co
m
p
r
ess
io
n
an
d
d
o
m
ai
n
ad
ap
tatio
n
.
T
h
is
te
ch
n
iq
u
e
f
ac
ili
tates
ef
f
ec
tiv
e
k
n
o
w
led
g
e
tr
an
s
f
er
b
etw
ee
n
m
o
d
els,
en
ab
li
n
g
th
e
s
y
s
te
m
to
ad
ap
t
to
d
if
f
er
en
t
d
o
m
ai
n
s
a
n
d
lan
g
u
a
g
es.
T
h
e
u
s
e
o
f
s
t
u
d
en
t
-
teac
h
er
m
o
d
els
e
n
h
a
n
ce
s
t
h
e
m
o
d
el's
ab
ilit
y
to
p
er
f
o
r
m
i
n
v
ar
io
u
s
li
n
g
u
i
s
tic
co
n
tex
t
s
,
m
a
k
i
n
g
i
t a
v
er
s
atile
an
d
ad
ap
tab
le
s
o
lu
tio
n
.
I
n
teg
r
atio
n
o
f
b
in
ar
y
cla
s
s
i
f
ie
r
f
o
r
co
n
te
n
t
o
r
ig
in
al
it
y
:
t
h
e
i
n
co
r
p
o
r
atio
n
o
f
a
b
i
n
ar
y
c
las
s
if
ier
w
it
h
th
e
au
to
-
en
co
d
er
en
s
u
r
es
co
n
t
en
t
o
r
ig
i
n
alit
y
a
n
d
m
it
ig
ate
s
t
h
e
r
is
k
o
f
p
la
g
iar
is
m
.
T
h
is
co
n
tr
ib
u
tio
n
ad
d
s
an
ex
tr
a
la
y
er
o
f
cr
ed
ib
ilit
y
an
d
r
eliab
ilit
y
to
t
h
e
m
o
d
el,
m
a
k
in
g
i
t
s
u
itab
le
f
o
r
ap
p
licatio
n
s
w
h
er
e
co
n
te
n
t
au
th
e
n
tic
it
y
i
s
es
s
en
t
ial,
s
u
c
h
as in
ac
ad
e
m
ic
r
esear
ch
,
leg
a
l d
o
cu
m
en
ts
,
an
d
o
f
f
icia
l c
o
m
m
u
n
ica
tio
n
s
.
T
h
e
r
esear
ch
w
o
r
k
in
th
i
s
p
ap
er
is
o
r
g
an
ized
as
f
o
llo
w
s
: i
n
t
h
e
s
ec
tio
n
1
,
a
b
r
ief
i
n
tr
o
d
u
cti
o
n
is
g
i
v
en
ab
o
u
t
th
e
ch
alle
n
g
e
s
ac
r
o
s
s
tex
t
-
p
r
e
-
p
r
o
ce
s
s
in
g
,
h
o
w
t
h
e
tr
an
s
liter
atio
n
an
d
tr
an
s
la
tio
n
m
o
d
el
s
h
a
v
e
b
ee
n
b
u
ilt
t
h
at
o
v
er
co
m
e
s
t
h
e
ch
all
en
g
e
s
i
n
v
ar
io
u
s
la
n
g
u
a
g
es,
a
n
d
th
e
b
r
ea
k
t
h
r
o
u
g
h
s
i
n
v
o
l
v
e
d
in
p
r
o
ce
s
s
in
g
t
h
e
Hin
d
i
an
d
K
a
n
n
ad
a
la
n
g
u
a
g
e.
I
n
s
ec
tio
n
2
a
n
e
u
r
al
n
e
t
w
o
r
k
w
it
h
o
u
t
iter
atio
n
s
o
r
co
n
v
o
lu
tio
n
al
o
p
er
atio
n
s
is
d
ev
elo
p
ed
th
at
f
o
cu
s
es
o
n
a
s
elf
-
atte
n
tio
n
m
ec
h
a
n
is
m
to
b
u
ild
an
en
co
d
er
-
d
ec
o
d
er
m
o
d
el
.
I
n
s
ec
tio
n
3
th
e
d
ataset
d
etails an
d
r
es
u
lts
f
o
r
tr
an
s
liter
atio
n
an
d
tr
an
s
latio
n
m
o
d
el
s
ar
e
s
h
o
w
n
.
T
h
e
last
s
e
ctio
n
is
co
n
cl
u
s
io
n
.
2.
P
RO
P
O
SE
D
M
E
T
H
O
D
T
h
e
p
r
o
v
id
ed
in
f
o
r
m
a
tio
n
d
escr
ib
es
a
n
eu
r
al
n
et
w
o
r
k
ar
ch
itect
u
r
e
th
at
e
m
p
lo
y
s
s
el
f
-
atten
tio
n
m
ec
h
a
n
i
s
m
s
a
n
d
d
o
m
ai
n
ad
ap
tatio
n
tech
n
iq
u
es.
T
h
e
n
et
wo
r
k
tak
es
w
o
r
d
o
r
s
eq
u
e
n
ce
e
m
b
ed
d
i
n
g
s
as
in
p
u
t
an
d
co
n
s
is
t
s
o
f
m
u
lt
ip
le
e
m
b
e
d
d
ed
lay
er
s
w
it
h
s
elf
-
atte
n
tio
n
a
n
d
m
u
lt
i
-
la
y
er
p
er
ce
p
tr
o
n
(
ML
P
)
co
m
p
o
n
e
n
ts
.
I
t
u
tili
ze
s
r
ec
u
r
s
iv
e
tr
an
s
f
er
an
d
au
to
-
en
co
d
er
w
it
h
m
as
k
e
d
la
y
er
s
to
h
an
d
le
co
m
p
le
x
d
ata
an
d
o
v
er
co
m
e
g
r
ad
ien
t
v
an
is
h
i
n
g
.
Mo
d
el
co
m
p
r
e
s
s
io
n
is
ac
h
ie
v
ed
u
s
i
n
g
s
tu
d
e
n
t
-
teac
h
er
m
o
d
els,
w
h
ile
d
o
m
ai
n
ad
ap
tatio
n
en
s
u
r
es
g
o
o
d
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
d
if
f
er
en
t
d
o
m
ai
n
s
.
F
ig
u
r
e
1
s
h
o
w
s
t
h
e
p
r
o
p
o
s
ed
w
o
r
k
f
lo
w
.
T
h
e
s
y
s
te
m
d
esig
n
in
v
o
lv
e
s
ch
ar
ac
ter
-
le
v
el
w
o
r
d
-
to
-
w
o
r
d
e
m
b
ed
d
in
g
ar
ch
itect
u
r
e
w
it
h
atten
tio
n
m
ec
h
an
i
s
m
s
.
B
ea
m
s
ea
r
ch
a
n
d
a
b
in
ar
y
cla
s
s
i
f
ier
ar
e
u
tili
ze
d
f
o
r
NL
P
tas
k
s
a
n
d
co
n
ten
t
o
r
i
g
in
a
lit
y
as
s
u
r
a
n
ce
,
r
esp
ec
tiv
el
y
.
T
h
is
co
m
p
r
e
h
en
s
iv
e
ar
c
h
itect
u
r
e
ai
m
s
to
p
r
o
ce
s
s
an
d
ad
ap
t to
v
ar
io
u
s
d
o
m
ai
n
s
e
f
f
icien
tl
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2089
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
1
,
Ma
r
c
h
2025
:
2
4
3
-
253
246
Fig
u
r
e
1
illu
s
tr
ates
a
m
ac
h
i
n
e
lear
n
in
g
p
ip
elin
e
f
o
r
p
r
o
ce
s
s
in
g
n
at
u
r
al
lan
g
u
ag
e
d
ata,
s
ta
r
tin
g
w
i
th
th
e
in
p
u
t
o
f
w
o
r
d
s
eq
u
en
ce
e
m
b
ed
d
in
g
s
-
v
ec
to
r
r
ep
r
esen
ta
tio
n
s
o
f
te
x
t.
A
n
e
n
co
d
er
co
m
p
o
n
en
t
p
r
o
ce
s
s
es
th
ese
e
m
b
ed
d
in
g
s
to
co
n
d
en
s
e
t
h
e
i
n
f
o
r
m
atio
n
i
n
to
a
m
o
r
e
m
an
a
g
ea
b
le
f
o
r
m
f
o
r
t
h
e
m
o
d
el
to
h
an
d
le.
Fo
llo
w
i
n
g
t
h
is
,
a
d
ec
o
d
er
tak
es
th
e
en
co
d
ed
d
ata
an
d
co
n
s
tr
u
cts
a
n
o
u
tp
u
t
s
eq
u
e
n
ce
,
wh
ich
ca
n
b
e
tex
t,
a
s
u
m
m
ar
y
,
tr
an
s
latio
n
,
o
r
an
o
th
er
f
o
r
m
d
ep
en
d
i
n
g
o
n
th
e
a
p
p
licatio
n
.
T
h
e
m
o
d
el
ad
ap
tatio
n
s
ta
g
e
s
u
g
g
e
s
t
s
th
at
t
h
e
b
ase
m
o
d
el
ca
n
b
e
f
i
n
e
-
tu
n
ed
to
b
etter
f
i
t
s
p
e
cif
ic
ta
s
k
s
o
r
n
e
w
d
ata
d
o
m
ain
s
,
e
n
h
an
ci
n
g
it
s
p
er
f
o
r
m
a
n
ce
o
n
ta
s
k
s
d
i
f
f
er
e
n
t
f
r
o
m
t
h
e
o
n
e
s
it
w
as
in
itiall
y
tr
ain
ed
o
n
.
Fi
n
all
y
,
i
n
-
m
o
d
el
ad
ap
tatio
n
an
d
o
u
t
-
m
o
d
el
ad
ap
tatio
n
i
m
p
l
y
t
h
at
f
i
n
e
-
t
u
n
i
n
g
ca
n
o
cc
u
r
w
it
h
in
th
e
ex
i
s
tin
g
m
o
d
el
s
tr
u
c
tu
r
e
o
r
th
r
o
u
g
h
ex
ter
n
al
m
o
d
i
f
icat
io
n
s
to
t
h
e
ar
ch
itect
u
r
e,
r
esp
ec
tiv
el
y
.
Fig
u
r
e
1
.
P
r
o
p
o
s
ed
w
o
r
k
f
lo
w
T
h
e
ab
o
v
e
n
et
w
o
r
k
i
s
a
u
n
iq
u
e
t
y
p
e
o
f
n
eu
r
al
n
et
w
o
r
k
th
at
d
o
es
n
o
t
in
v
o
lv
e
ite
r
atio
n
s
o
r
co
n
v
o
lu
tio
n
al
o
p
er
atio
n
s
.
I
n
s
t
ea
d
,
it
r
elies
o
n
a
s
elf
-
at
ten
t
i
o
n
m
ec
h
an
i
s
m
to
co
n
s
tr
u
ct
a
n
au
to
-
en
co
d
er
.
T
h
e
in
p
u
t
p
r
o
v
id
ed
to
th
e
n
et
w
o
r
k
ca
n
b
e
eit
h
er
a
w
o
r
d
e
m
b
ed
d
i
n
g
o
r
a
s
eq
u
en
ce
e
m
b
ed
d
in
g
.
T
h
e
e
n
co
d
er
d
ec
o
d
er
c
o
n
s
is
t
s
o
f
A
e
m
b
ed
d
ed
la
y
er
s
,
m
u
lt
ip
le
s
el
f
-
atte
n
t
io
n
la
y
er
s
,
co
n
v
o
lu
tio
n
s
,
a
n
d
m
as
k
ed
m
u
l
tip
le
s
elf
-
atte
n
tio
n
la
y
er
s
f
o
r
mo
p
.
T
h
e
s
elf
-
atte
n
tio
n
la
y
er
s
alo
n
g
w
it
h
th
e
tex
t
is
n
eit
h
er
cr
ea
ted
n
o
r
m
as
k
ed
.
An
in
p
u
t
is
p
r
o
v
id
ed
as
x
p
w
it
h
th
e
s
eq
u
en
ce
,
e
m
b
ed
d
in
g
t
h
e
tr
a
n
s
f
o
r
m
a
tio
n
v
ia
a
w
ei
g
h
t
m
atr
i
x
s
u
b
j
ec
ted
to
L
p
,
th
e
p
r
o
j
ec
tio
n
m
atr
i
x
d
en
o
ted
as
T
p
w
ith
th
e
f
ea
tu
r
e
m
atr
i
x
s
h
o
w
n
b
y
N
p
,
s
elf
-
atte
n
tio
n
m
ec
h
an
is
m
i
s
later
ap
p
lied
o
n
L
p
,
T
p
,
N
p
ir
r
esp
ec
tiv
e,
So
f
t
m
a
x
f
u
n
c
tio
n
d
en
o
ted
b
y
ℋ
.
α
(
L
p
,
T
p
,
N
p
)
=
s
e
g
(
L
p
+
1
p
+
1
…
…
.
.
L
X
p
+
1
)
R
p
(
1
)
L
x
p
+
1
=
ℋ
L
x
p
L
x
Z
(
l
m
)
−
2
(
N
x
p
)
(
2
)
L
x
p
,
T
x
p
,
N
x
p
=
L
p
R
x
l
,
T
p
R
x
t
,
N
p
R
x
n
,
(
3
)
Hen
ce
f
o
r
th
L
x
p
,
T
x
p
,
N
x
p
r
ep
r
esen
ts
th
e
x
−
th
q
u
er
y
w
it
h
th
e
f
ea
t
u
r
e
m
atr
i
x
f
o
r
th
e
p
−
th
la
y
er
.
{
R
x
l
,
R
x
t
,
R
x
n
}
∈
S
m
d
d
ep
icts
th
e
v
ar
iab
le
m
atr
i
x
m
d
r
ep
r
esen
tin
g
t
h
e
d
i
m
en
s
io
n
f
o
r
th
e
m
o
d
el
ir
r
esp
ec
ti
v
el
y
.
A
M
L
P
is
a
f
u
ll
y
co
n
n
ec
ted
n
e
t
w
o
r
k
w
i
th
t
h
e
ac
ti
v
atio
n
f
u
n
c
tio
n
th
at
i
s
ap
p
lied
o
n
ea
ch
f
u
n
ctio
n
.
L
n
+
1
=
β
(
α
(
L
p
,
T
p
,
N
p
)
)
+
L
p
(
4
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
S
elf
-
a
tten
tio
n
en
co
d
er
-
d
ec
o
d
e
r
w
ith
mo
d
el
a
d
a
p
ta
tio
n
fo
r
tr
a
n
s
liter
a
tio
n
a
n
d
…
(
S
h
a
n
th
a
l
a
N
a
g
a
r
a
ja
)
247
Ho
w
e
v
er
,
L
n
+
1
is
t
h
e
p
r
eli
m
in
ar
y
s
ta
g
e
w
it
h
r
ele
v
a
n
t
f
ea
t
u
r
e
i
n
f
o
r
m
atio
n
h
o
w
e
v
er
L
n
is
ad
d
ed
to
d
ev
elo
p
th
e
co
n
n
ec
tio
n
s
t
h
at
o
v
er
co
m
es
g
r
ad
ie
n
t
v
an
is
h
i
n
g
.
T
h
is
p
r
o
ce
s
s
s
eq
u
en
ci
n
g
is
d
ep
icted
as
I
α
,
d
en
o
ted
b
y
t
h
e
s
o
u
r
ce
as
M
r
+
1
.
T
h
e
α
u
tili
s
es d
if
f
er
e
n
t la
y
er
s
to
lear
n
f
r
o
m
t
h
e
s
o
u
r
ce
.
M
r
+
1
=
I
α
p
+
1
(
L
p
,
T
p
,
N
p
)
(
5
)
T
h
e
α
u
tili
s
e
s
d
if
f
er
e
n
t
la
y
er
s
to
lear
n
f
r
o
m
th
e
s
o
u
r
ce
.
H
en
ce
f
o
r
t
h
th
is
[
…
]
z
(
z
∈
{
1
,
2
,
…
…
Z
}
)
r
ep
r
esen
ts
th
e
Z
s
i
m
i
lar
lay
er
s
s
tack
ed
u
p
w
it
h
ea
ch
o
th
er
.
T
h
e
o
u
tco
m
e
L
z
f
o
r
th
e
Z
−
th
atten
ti
o
n
la
y
er
d
en
o
ted
b
y
t
h
e
f
in
al
o
u
tco
m
e
tr
an
s
m
itted
.
T
h
e
tr
an
s
f
er
o
f
th
e
a
u
to
-
en
co
d
er
b
y
a
tr
an
s
la
tio
n
m
o
d
el
w
h
ic
h
p
r
ed
icts
th
e
tar
g
et.
T
h
e
d
if
f
er
en
ce
b
et
w
ee
n
t
h
e
a
u
to
-
e
n
co
d
er
b
y
t
h
e
m
as
k
ed
la
y
er
t
h
at
r
e
p
r
esen
ts
t
h
e
o
u
tp
u
t
d
esig
n
ed
d
y
n
a
m
icall
y
.
T
h
e
o
u
tp
u
t is d
ec
o
d
ed
w
h
ic
h
est
i
m
a
t
es th
e
p
r
o
b
ab
ilit
y
o
f
l
og
e
(
y
m
|
y
<
m
,
ℋ
)
.
[
L
z
=
i
α
z
(
L
z
−
1
,
T
z
−
1
,
N
z
−
1
)
]
z
(
6
)
2
.
1
.
E
ntr
o
py
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
h
er
e
b
ec
o
m
e
ac
cu
s
to
m
ed
b
y
f
in
e
-
t
u
n
in
g
th
e
m
o
d
el
a
n
d
o
u
t
-
m
o
d
el
t
r
ain
in
g
t
o
co
n
ce
n
tr
ate
o
n
tr
ai
n
i
n
g
a
n
d
p
er
f
o
r
m
t
h
e
m
o
d
el
co
m
p
r
ess
io
n
.
T
h
is
in
c
lu
d
es
t
w
o
-
s
u
b
s
ec
t
io
n
s
k
n
o
w
n
as
t
h
e
s
tu
d
e
n
t a
n
d
teac
h
er
m
o
d
el.
T
h
e
s
tu
d
en
t lo
s
s
i
s
d
eter
m
i
n
ed
b
y
t
w
o
co
m
p
o
n
e
n
ts
.
−
T
h
e
lo
s
s
ass
o
ciate
d
w
it
h
th
e
p
r
o
b
ab
ilit
y
a
n
d
p
r
ed
ictio
n
o
f
th
e
lab
el
b
y
i
n
co
r
p
o
r
atin
g
n
e
g
ati
v
e
lo
g
en
tr
o
p
y
f
u
n
ctio
n
.
E
n
tr
opy
(
δ
d
;
F
)
=
∑
∑
−
U
(
w
m
=
1
(
c
,
d
∈
F
)
y
m
)
∗
l
og
e
(
y
m
|
y
<
m
,
δ
d
)
(
7
)
−
T
h
e
m
o
d
el
co
m
p
r
es
s
io
n
co
m
p
u
tes
t
h
e
lo
s
s
in
b
et
w
ee
n
t
h
e
o
u
tp
u
t
p
r
o
b
ab
ilit
y
an
d
t
h
e
s
t
u
d
en
t
teac
h
er
m
o
d
el.
E
n
tr
opy
m
(
δ
d
;
F
,
δ
d
)
=
∑
∑
−
l
(
w
m
=
1
(
c
,
d
∈
F
)
y
m
|
y
<
m
,
δ
d
)
∗
l
og
e
(
y
m
|
y
<
m
,
δ
d
)
(
8
)
T
h
e
n
(
y
m
|
y
<
m
,
δ
d
^
)
w
it
h
δ
d
r
ep
r
esen
ts
th
e
p
ar
a
m
eter
f
o
r
th
e
s
tu
d
e
n
t
m
o
d
e
l.
δ
d
^
=
δ
d
∗
,
th
e
m
et
h
o
d
to
av
er
ag
e
δ
d
=
1
z
∑
δ
d
(
x
)
z
,
th
e
w
e
i
g
h
t
ap
p
r
o
ac
h
is
r
ep
r
esen
ted
as
δ
d
(
z
)
=
∑
g
−
p
(
S
z
δ
d
(
z
)
)
)
δ
d
(
z
)
,
w
h
er
ein
[
g
-
p
]
r
ep
r
esen
ts
th
e
n
o
r
m
al
ized
f
u
n
ctio
n
f
o
r
z
−
th
ev
alu
a
tio
n
f
o
r
th
e
p
ar
am
eter
δ
d
(
z
)
th
at
r
ep
r
esen
t
s
an
en
s
e
m
b
le
m
o
d
el.
T
h
e
av
er
ag
in
g
an
d
w
ei
g
h
ted
-
a
v
er
ag
i
n
g
ap
p
r
o
ac
h
in
th
e
s
t
u
d
en
t
m
o
d
e
l
is
ap
p
lied
to
g
ain
n
ec
es
s
ar
y
i
n
f
o
r
m
atio
n
b
y
ac
c
u
m
u
latio
n
o
f
in
f
o
r
m
at
io
n
b
y
t
h
e
iter
atio
n
s
o
f
t
h
e
teac
h
er
m
o
d
el.
2
.
2
.
M
o
del a
da
pta
t
i
o
n
T
h
e
in
-
m
o
d
el
a
n
d
o
u
t
-
m
o
d
el
c
o
m
p
u
te
t
h
e
p
r
e
-
tr
ain
i
n
g
o
f
th
e
m
o
d
el.
I
n
ea
c
h
le
v
el
th
e
m
o
d
el
v
al
u
e
i
s
th
e
ad
d
ed
ad
v
an
ta
g
e
f
o
r
th
e
f
o
r
m
er
iter
atio
n
f
o
r
t
h
e
p
u
r
p
o
s
e
o
f
in
-
m
o
d
el
p
ar
a
m
e
ter
s
.
T
h
e
p
r
o
ce
s
s
is
iter
ated
to
ac
h
iev
e
m
u
t
u
al
tr
a
n
s
f
er
o
f
d
ata.
T
h
e
in
-
m
o
d
el
a
n
d
o
u
t
-
m
o
d
el
f
ea
tu
r
e
s
ar
e
tr
a
n
s
m
it
ted
ac
r
o
s
s
ea
ch
o
t
h
er
at
m
o
d
el
lev
el
to
tr
an
s
m
it
d
ata
h
en
ce
f
o
r
t
h
en
s
u
r
i
n
g
g
o
o
d
p
er
f
o
r
m
an
ce
.
T
h
e
m
o
d
el
e
f
f
icie
n
c
y
is
ev
a
lu
ated
ac
r
o
s
s
s
o
u
r
ce
an
d
tar
g
et
d
o
m
ai
n
d
ata
F
d
,
F
h
is
p
ar
titi
o
n
ed
i
n
to
t
h
e
tr
ain
i
n
g
s
ets
F
d
n
,
F
h
n
an
d
m
o
d
el
an
d
t
h
e
e
v
alu
a
tio
n
p
air
co
n
s
is
t
o
f
F
d
co
m
,
F
h
co
m
th
at
is
f
u
r
t
h
er
r
esp
o
n
s
ib
le
to
tr
ain
an
d
ev
a
lu
a
te
th
e
m
o
d
el
as s
h
o
w
n
in
A
l
g
o
r
ith
m
1
.
A
l
g
o
r
ith
m
1
.
C
o
n
s
is
ts
o
f
t
h
e
m
o
d
el
ad
ap
tatio
n
al
g
o
r
ith
m
f
o
r
th
e
p
r
o
p
o
s
ed
m
o
d
el
th
at
co
n
s
is
t
s
o
f
t
w
o
s
ta
g
es
I
n
th
e
i
n
itial
s
tag
e,
t
h
e
m
ai
n
ai
m
is
to
co
m
p
lete
th
e
i
n
iti
aliza
tio
n
f
o
r
th
e
i
n
-
m
o
d
el
an
d
o
u
t
-
m
o
d
el
m
o
d
e
l
p
ar
am
eter
s
.
-
T
h
e
µ
f
u
n
ctio
n
is
r
esp
o
n
s
ib
le
to
tr
ain
th
e
m
o
d
el
f
o
r
th
e
o
b
j
ec
tiv
e
f
u
n
ctio
n
F
d
n
an
d
th
e
p
ar
am
eter
u
s
ed
alo
n
g
w
it
h
t
h
is
δ
h
(
p
+
1
)
,
in
itialized
b
y
E
n
tr
opy
m
(
δ
d
;
F
d
n
)
re
tain
ed
f
o
r
th
e
s
o
u
r
ce
.
I
n
th
i
s
s
tep
th
e
r
ec
u
r
s
i
v
e
tr
an
s
f
er
o
f
d
ata
is
estab
li
s
h
ed
i
n
b
etw
ee
n
in
-
m
o
d
el
an
d
o
u
t
-
m
o
d
el
ad
ap
tatio
n
.
-
T
h
e
ƿ
f
u
n
ctio
n
tr
an
s
f
er
s
th
e
m
o
d
el,
it
s
m
a
in
g
o
al
is
to
u
ti
lize
th
i
s
w
it
h
s
el
f
-
k
n
o
w
led
g
e
f
u
n
ctio
n
lo
s
s
d
en
o
ted
as
E
n
tr
o
py
(
δ
h
(
t
−
1
)
;
F
d
n
)
an
d
th
e
E
n
tr
o
py
m
(
δ
d
(
t
−
1
)
;
F
h
n
)
,
u
s
ed
alo
n
g
t
h
e
tr
ai
n
in
g
s
et
s
F
h
n
.
-
T
h
e
m
o
d
el
tr
an
s
f
er
f
o
r
in
-
d
o
m
ain
p
ar
am
e
ter
s
et
δ
d
is
in
itialize
d
th
r
o
u
g
h
t
h
e
p
r
ev
io
u
s
o
u
t
-
d
o
m
ai
n
m
o
d
el
p
ar
am
eter
as
δ
h
(
t
−
1
)
.
T
h
e
in
itial
izatio
n
is
ca
r
r
ied
o
u
t
w
h
ile
o
p
ti
m
iz
in
g
th
e
m
o
d
el
is
p
er
f
o
r
m
ed
t
h
r
o
u
g
h
th
e
in
-
m
o
d
el
an
d
r
ep
ea
ted
f
o
r
s
o
u
r
ce
d
o
m
ain
.
-
T
h
e
β
m
o
d
el
i
s
u
tili
s
ed
f
o
r
th
e
e
v
al
u
atio
n
p
u
r
p
o
s
e
v
ia
th
e
e
n
s
e
m
b
le
ac
ti
v
atio
n
f
u
n
c
tio
n
u
s
ed
f
o
r
ev
alu
a
tio
n
p
u
r
p
o
s
e
f
o
r
en
s
u
r
i
n
g
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
δ
d
(
t
)
f
o
r
b
u
ild
in
g
a
s
et
o
f
F
d
co
m
b
y
th
e
en
s
e
m
b
le
p
ar
am
eter
d
ep
icted
as
δ
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2089
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
1
,
Ma
r
c
h
2025
:
2
4
3
-
253
248
I
n
p
u
t
:
T
r
ain
in
g
o
f
{
F
d
n
,
F
h
n
}
,
r
ep
r
esen
ts
th
e
l
is
ts
as,
{
F
d
co
m
,
F
h
co
m
}
,
w
it
h
lev
e
l
T
.
Step
1
:
tr
a
in
In
−
mode
l
Step
2
:
δ
d
(
p
+
1
)
←
tr
(
E
n
tr
opy
(
δ
d
;
F
d
n
)
Step
3
:
tr
a
in
out
−
mode
l
Step
4
:
δ
h
(
p
+
1
)
←
E
n
tr
opy
(
δ
h
;
F
h
n
)
Step
5
:
I
n
itialize
in
−
mode
l
an
d
out
−
mode
l
en
s
e
m
b
le
m
o
d
e
l p
ar
am
eter
s
Step
6
:
δ
d
←
δ
d
(
p
+
1
)
,
δ
h
←
δ
h
(
p
+
1
)
Step
7
:
f
o
r
t
=
1
,
2
,
…
T
do
Step
8
:
in
−
mode
l
tr
ain
in
g
=T
r
an
s
f
er
tr
ain
i
n
g
m
o
d
el
an
d
co
m
p
u
tatio
n
Step
9
:
δ
d
(
t
)
←
ϑ
E
n
tr
o
py
(
δ
h
(
t
−
1
)
;
F
d
n
)
E
n
tr
op
y
m
(
δ
h
(
t
−
1
)
;
D
b
l
τ
b
)
)
Step
1
0
:
δ
d
←
β
(
F
d
co
m
,
δ
d
(
t
)
Step
1
1
:
tr
a
in
out
−
mode
l
=T
r
an
s
f
er
tr
ain
i
n
g
m
o
d
el
an
d
co
m
p
u
tatio
n
Step
1
2
:
δ
h
(
t
)
←
(
L
oss
(
δ
h
(
t
−
1
)
;
F
h
n
)
E
n
tr
opy
m
(
δ
d
(
t
−
1
)
;
F
h
n
δ
h
)
)
Step
1
3
:
δ
h
←
β
(
F
h
co
m
,
δ
h
(
t
)
)
Step
1
4
:
en
d
f
o
r
O
u
tp
u
t
:
i
n
m
o
d
el
tr
ai
n
in
g
δ
d
; o
u
t
m
o
d
el
tr
ain
in
g
δ
h
2
.
3
.
Sy
s
t
em
de
s
ig
n
Neu
r
al
n
et
w
o
r
k
s
b
ased
o
n
wo
r
d
s
ca
n
o
f
f
er
an
en
d
-
to
-
e
n
d
s
o
lu
tio
n
to
t
h
e
co
m
p
le
x
it
ies
ass
o
ciate
d
w
it
h
t
h
e
h
u
g
e
n
u
m
b
er
o
f
wo
r
d
s
.
Ho
w
e
v
er
,
ch
ar
ac
ter
-
le
v
el
m
et
h
o
d
s
ca
n
also
b
e
u
s
e
d
to
ev
alu
ate
th
e
co
m
p
le
x
it
y
a
s
s
o
ciat
ed
w
it
h
n
o
is
e,
alter
atio
n
s
,
an
d
er
r
o
r
s
.
T
h
ese
m
e
th
o
d
s
u
s
e
w
o
r
d
-
to
-
w
o
r
d
m
o
d
el
e
m
b
ed
d
in
g
s
to
ass
e
s
s
t
h
e
co
m
p
lex
it
y
o
f
tex
t.
2
.
3
.
1
.
P
re
-
pro
ce
s
s
ing
a
nd
po
s
t
-
pro
ce
s
s
ing
I
n
p
u
t
p
r
e
-
p
r
o
ce
s
s
i
n
g
:
w
h
e
n
a
n
in
p
u
t
w
o
r
d
is
u
tter
ed
,
it
is
n
o
r
m
alize
d
b
y
co
n
v
er
ti
n
g
a
l
l
let
ter
s
to
lo
w
er
ca
s
e,
r
e
m
o
v
i
n
g
r
ep
ea
ted
ch
ar
ac
ter
s
,
tr
an
s
f
o
r
m
in
g
d
iac
r
itics
in
to
s
tan
d
ar
d
7
-
b
it
Am
e
r
ican
s
ta
n
d
ar
d
co
d
e
f
o
r
in
f
o
r
m
atio
n
i
n
ter
ch
a
n
g
e
(
A
S
C
I
I
)
,
an
d
co
n
v
er
ti
n
g
e
m
o
j
is
an
d
e
m
o
tico
n
s
in
v
o
lv
i
n
g
p
u
n
ct
u
atio
n
i
n
to
h
as
h
ta
g
s
.
Du
r
i
n
g
tr
ain
i
n
g
,
f
o
r
eig
n
w
o
r
d
s
ar
e
tag
g
ed
as
h
a
s
h
ta
g
s
a
n
d
th
e
o
u
tp
u
t
is
ali
g
n
ed
w
i
th
t
h
e
in
p
u
t
th
r
o
u
g
h
th
e
s
e
h
as
h
ta
g
s
.
T
h
is
en
s
u
r
es
t
h
at
th
e
m
o
d
el
lear
n
s
to
id
en
ti
f
y
f
o
r
eig
n
w
o
r
d
s
an
d
tr
an
s
f
er
th
e
m
in
to
h
as
h
ta
g
s
t
h
at
ar
e
id
en
tical
to
t
h
e
in
p
u
t.
A
d
d
itio
n
all
y
,
e
m
o
j
is
,
em
o
tico
n
s
,
a
n
d
p
u
n
c
t
u
atio
n
s
ar
e
co
n
v
er
ted
in
to
h
as
h
ta
g
s
d
u
r
i
n
g
tr
ai
n
i
n
g
a
n
d
p
r
ed
ictio
n
.
A
f
ter
tr
ain
i
n
g
,
a
p
o
s
t
-
p
r
o
ce
s
s
i
n
g
s
tep
co
n
v
er
t
s
t
h
e
h
a
s
h
ta
g
s
b
ac
k
to
w
o
r
d
s
in
t
h
e
s
o
u
r
ce
.
I
f
th
e
i
n
p
u
t
an
d
o
u
tp
u
t
ar
e
alig
n
ed
,
th
i
s
s
tep
is
p
er
f
o
r
m
ed
b
ef
o
r
e
r
em
o
v
in
g
t
h
e
to
k
e
n
s
[
+]
an
d
[
-
]
.
Ho
w
ev
er
,
i
n
th
e
f
in
al
o
u
tp
u
t,
th
e
w
o
r
d
s
alo
n
g
w
it
h
th
e
[
+]
to
k
en
ar
e
m
er
g
ed
an
d
th
e
[
-
]
to
k
en
s
ar
e
r
ep
lace
d
w
i
th
a
w
h
ite
s
p
ac
e
t
h
at
s
p
lits
a
w
o
r
d
i
n
to
m
u
ltip
le
w
o
r
d
s
.
2
.
3
.
2
.
Sy
s
t
e
m
a
rc
hite
ct
ure
A
c
h
ar
ac
ter
-
le
v
el
w
o
r
d
-
to
-
wo
r
d
em
b
ed
d
in
g
ar
ch
i
tectu
r
e
is
a
t
y
p
e
o
f
n
e
u
r
al
n
et
w
o
r
k
th
at
u
s
es
ch
ar
ac
ter
-
le
v
el
e
m
b
ed
d
in
g
to
r
ep
r
esen
t
w
o
r
d
s
.
T
h
is
t
y
p
e
o
f
ar
ch
itect
u
r
e
is
o
f
te
n
u
s
ed
f
o
r
task
s
s
u
c
h
as
m
ac
h
in
e
tr
an
s
latio
n
an
d
tex
t
class
i
f
icatio
n
.
T
h
e
m
o
d
el
J
(
c
|
d
)
th
at
g
en
er
ates
a
n
in
p
u
t
d
f
o
r
tar
g
et
c
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
co
n
s
is
t
s
o
f
a
n
atte
n
tio
n
m
ec
h
an
i
s
m
w
h
ic
h
co
n
s
is
t
o
f
g
a
ted
r
ec
u
r
r
en
t
u
n
it
(
GR
U
)
.
T
h
e
in
itial
s
tag
e
i
n
th
e
p
r
o
p
o
s
ed
m
o
d
el
w
h
ic
h
co
n
s
i
s
t
s
o
f
r
ec
u
r
r
en
t
n
eu
r
al
n
et
w
o
r
k
s
an
d
n
o
n
-
r
ec
u
r
r
en
t
co
n
n
ec
tio
n
s
ess
e
n
tial
f
o
r
tr
ain
in
g
p
u
r
p
o
s
e.
T
h
e
So
f
t
m
ax
la
y
er
f
o
r
t
h
e
p
r
o
p
o
s
ed
m
o
d
el’
s
o
u
tp
u
t
to
t
h
e
f
i
n
al
s
eq
u
en
c
e
o
u
tp
u
t
c
.
T
h
e
en
tr
o
p
y
f
u
n
ctio
n
e
v
alu
a
ted
f
o
r
lo
s
s
f
o
r
ti
m
e
av
er
ag
e
o
v
er
c
x
.
B
ea
m
s
ea
r
ch
i
s
a
te
ch
n
iq
u
e
u
s
ed
i
n
NL
P
to
f
in
d
th
e
m
o
s
t
li
k
el
y
s
e
q
u
en
ce
o
f
w
o
r
d
s
in
a
s
e
n
ten
ce
.
I
t
w
o
r
k
s
b
y
k
ee
p
i
n
g
tr
ac
k
o
f
a
f
ix
ed
n
u
m
b
er
o
f
ca
n
d
id
ate
s
eq
u
en
ce
s
,
an
d
th
en
at
ea
ch
s
tep
,
p
r
e
d
ictin
g
t
h
e
n
ex
t
w
o
r
d
in
th
e
s
eq
u
e
n
ce
w
it
h
th
e
h
ig
h
es
t
lo
g
-
lik
eli
h
o
o
d
.
T
h
e
b
ea
m
s
ize
is
a
h
y
p
er
-
p
ar
a
m
eter
th
at
co
n
tr
o
ls
h
o
w
m
a
n
y
ca
n
d
id
ate
s
eq
u
e
n
ce
s
ar
e
k
ep
t tr
ac
k
o
f
.
2
.
3
.
3
.
B
ina
ry
cl
a
s
s
if
ier
As
m
e
n
tio
n
ed
ea
r
lier
,
an
au
to
en
co
d
er
is
en
h
an
ce
d
w
ith
a
b
in
ar
y
clas
s
if
ier
b
y
in
teg
r
ati
n
g
an
atten
tio
n
m
ec
h
a
n
i
s
m
.
T
h
is
atte
n
tio
n
m
ec
h
an
i
s
m
i
s
e
m
p
lo
y
ed
to
cr
ea
te
a
task
-
s
p
ec
if
ic
co
n
te
x
t
u
al
r
ep
r
esen
tatio
n
.
T
h
e
p
u
r
p
o
s
e
o
f
th
is
a
u
g
m
e
n
tatio
n
i
s
to
av
o
id
p
lag
iar
is
m
an
d
en
s
u
r
e
th
e
o
r
ig
in
al
it
y
o
f
th
e
co
n
ten
t
I
x
(
h
)
o
f
(
h
)
.
I
x
(
h
)
=
∑
ϑ
v
u
v
g
v
=
1
ϑ
v
=
ex
p
(
u
v
)
∑
ex
p
(
u
v
′
)
g
v
′
(
9
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
S
elf
-
a
tten
tio
n
en
co
d
er
-
d
ec
o
d
e
r
w
ith
mo
d
el
a
d
a
p
ta
tio
n
fo
r
tr
a
n
s
liter
a
tio
n
a
n
d
…
(
S
h
a
n
th
a
l
a
N
a
g
a
r
a
ja
)
249
u
v
=
(
j
s
)
y
ta
n
h
(
N
s
i
v
)
Her
e
j
s
an
d
N
s
ar
e
th
e
r
elate
d
co
n
s
tr
ai
n
ts
.
He
n
ce
a
b
i
n
ar
y
cla
s
s
i
f
ier
f
o
r
I
x
(
h
)
is
s
h
o
w
n
a
s
i
n
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
th
is
class
i
f
ie
r
is
tr
ain
ed
to
m
ax
i
m
ize
th
e
en
tr
o
p
y
.
I
n
th
is
co
n
tex
t,
th
e
w
ei
g
h
ts
ar
e
u
s
ed
to
d
eter
m
in
e
th
e
i
m
p
o
r
tan
ce
o
r
r
elev
an
ce
o
f
v
ar
io
u
s
tar
g
et
w
o
r
d
s
w
it
h
i
n
t
h
e
s
e
n
te
n
ce
.
T
o
ac
h
iev
e
th
i
s
,
t
h
e
class
i
f
ier
is
e
m
p
lo
y
ed
to
ca
lcu
late
atten
tio
n
al
w
ei
g
h
ts
f
o
r
th
e
tar
g
et
w
o
r
d
s
d
u
r
i
n
g
th
e
m
o
d
e
l tr
ain
i
n
g
p
r
o
ce
s
s
.
β
ie
i
(
h
;
γ
jk
j
)
=
l
ogs
(
d
|
h
;
γ
jk
j
)
(
1
0
)
β
ps
(
k
,
h
;
γ
ps
)=
∑
(
1
+
g
v
=
1
Ƥ
v
)
l
ogs
(
h
v
|
v
,
h
<
v
;
γ
ps
)
(
1
1
)
Her
e
Ƥ
v
d
ep
icts
th
e
at
ten
tio
n
w
eig
h
t
ass
o
ciate
d
alo
n
g
t
h
e
h
v
to
o
b
tain
(
1
1
)
an
d
γ
ps
r
ep
r
esen
ts
t
h
e
v
ar
iab
le
r
elate
d
to
ps
.
T
h
e
m
o
d
el
d
ev
elo
p
m
e
n
t
m
o
d
if
ies
t
h
e
m
ag
n
it
u
d
e
o
f
t
h
e
u
p
d
ated
p
ar
am
eter
w
h
ile
p
r
eser
v
in
g
it
s
d
ir
ec
tio
n
.
T
h
is
en
s
u
r
es
t
h
at
ta
s
k
-
s
p
ec
i
f
ic
w
o
r
d
s
an
d
tas
k
-
s
h
ar
ed
w
o
r
d
s
ar
e
co
n
s
is
ten
tl
y
u
p
d
ated
d
u
r
in
g
tr
ain
i
n
g
.
T
h
e
p
r
im
ar
y
b
en
ef
it
o
f
in
co
r
p
o
r
atin
g
lex
ical
w
ei
g
h
ti
n
g
is
t
h
e
ab
ilit
y
to
tr
ain
at
th
e
w
o
r
d
lev
el
r
ath
er
th
an
t
h
e
s
en
te
n
ce
lev
e
l.
T
h
is
ap
p
r
o
ac
h
is
ad
v
an
tag
eo
u
s
as
it
r
ed
u
ce
s
th
e
ti
m
e
r
eq
u
i
r
ed
f
o
r
tr
ain
in
g
th
e
m
o
d
el
w
h
ile
m
ai
n
tai
n
i
n
g
e
f
f
ec
tiv
e
n
ess
.
3.
RE
SU
L
T
E
VA
L
UA
T
I
O
N
T
h
is
s
ec
tio
n
p
r
esen
t
s
t
h
e
a
n
al
y
s
i
s
o
f
r
e
s
u
l
ts
o
b
tai
n
ed
u
s
in
g
th
e
S
A
E
DM
m
o
d
el
f
o
r
tr
a
n
s
latio
n
a
n
d
tr
an
s
liter
at
io
n
.
T
h
e
m
o
d
el
's
p
er
f
o
r
m
a
n
ce
i
s
e
v
al
u
ated
b
ased
o
n
ac
c
u
r
ac
y
,
a
n
d
a
co
m
p
ar
at
iv
e
s
t
u
d
y
is
co
n
d
u
cted
w
it
h
o
th
er
tr
a
n
s
l
it
er
atio
n
m
et
h
o
d
s
.
T
h
e
m
ai
n
o
b
j
ec
tiv
e
is
to
i
m
p
r
o
v
e
i
n
f
o
r
m
atio
n
tr
a
n
s
f
e
r
i
n
Kan
n
ad
a
an
d
Hin
d
i
lan
g
u
a
g
es
b
y
en
h
a
n
ci
n
g
th
e
m
o
d
el
's
ef
f
ec
ti
v
e
n
ess
.
Deta
il
s
o
f
th
e
d
ataset
u
s
ed
f
o
r
tr
an
s
liter
at
io
n
a
n
d
tr
an
s
latio
n
ar
e
p
r
o
v
id
ed
.
Si
m
u
la
tio
n
s
a
r
e
co
n
d
u
cted
o
n
a
n
I
NT
E
L
C
o
r
e
i7
p
r
o
ce
s
s
o
r
,
u
tili
zi
n
g
P
y
t
h
o
n
an
d
d
ee
p
l
ea
r
n
in
g
lib
r
ar
ies,
w
i
t
h
8
GB
r
an
d
o
m
ac
ce
s
s
m
e
m
o
r
y
(
R
A
M
)
an
d
a
6
4
-
b
it
W
in
d
o
w
s
OS.
T
h
e
s
tu
d
y
a
i
m
s
to
d
e
m
o
n
s
tr
ate
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
'
s
e
f
f
icien
c
y
i
n
ac
h
iev
in
g
b
etter
tr
an
s
liter
at
io
n
an
d
tr
an
s
latio
n
r
esu
lt
s
.
3
.
1
.
Da
t
a
s
et
d
et
a
ils
T
h
e
Dak
s
h
i
n
a
d
ataset
[
2
6
]
is
a
co
m
p
r
eh
en
s
i
v
e
a
n
d
d
iv
er
s
e
co
llectio
n
o
f
I
n
d
ian
lan
g
u
ag
e
s
,
d
esig
n
ed
to
s
u
p
p
o
r
t
NL
P
r
esear
c
h
i
n
I
n
d
ia
n
la
n
g
u
a
g
es.
I
t
in
c
l
u
d
es
te
x
t
f
r
o
m
v
ar
io
u
s
d
o
m
ai
n
s
a
n
d
s
o
u
r
ce
s
,
en
co
m
p
as
s
in
g
Hi
n
d
i
an
d
Ka
n
n
ad
a
lan
g
u
a
g
es,
a
m
o
n
g
o
th
er
s
.
T
h
e
d
ataset
is
ai
m
ed
at
f
ac
ili
tatin
g
N
L
P
s
tu
d
ie
s
f
o
r
I
n
d
ian
la
n
g
u
a
g
es.
R
esear
c
h
er
s
ca
n
ac
ce
s
s
te
x
t
f
r
o
m
d
if
f
er
en
t
g
en
r
es,
s
u
ch
as
n
e
w
s
s
t
o
r
ies,
s
o
cial
m
ed
ia
co
n
ten
t,
an
d
o
f
f
icial
d
o
cu
m
e
n
ts
.
T
h
e
av
ailab
ilit
y
o
f
tex
t
i
n
b
o
th
its
o
r
ig
in
al
s
cr
ip
t
an
d
tr
an
s
liter
ated
f
o
r
m
en
s
u
r
es
ac
ce
s
s
ib
il
it
y
e
v
en
f
o
r
th
o
s
e
u
n
f
a
m
iliar
w
i
th
t
h
e
s
cr
ip
t.
Mo
r
e
o
v
er
,
th
e
d
ataset
p
r
o
v
id
es
an
n
o
tatio
n
s
f
o
r
n
a
m
ed
e
n
tit
ies,
p
ar
t
-
of
-
s
p
ee
c
h
id
en
tifie
r
s
,
an
d
s
en
ten
ce
b
o
u
n
d
ar
ies,
m
a
k
in
g
i
t
v
a
lu
ab
l
e
f
o
r
in
f
o
r
m
a
tio
n
ex
tr
ac
tio
n
a
n
d
m
ac
h
i
n
e
tr
a
n
s
latio
n
tas
k
s
.
O
v
er
all,
th
e
Da
k
s
h
i
n
a
d
ataset
s
er
v
es
as
a
v
a
lu
ab
le
r
eso
u
r
ce
f
o
r
ac
ad
em
ic
r
esear
ch
er
s
e
x
p
lo
r
in
g
I
n
d
ian
la
n
g
u
a
g
es l
ik
e
Hi
n
d
i
an
d
Kan
n
ad
a
an
d
th
eir
ap
p
lica
tio
n
s
i
n
N
L
P
.
T
h
e
w
eb
an
d
tr
an
s
latio
n
(
W
A
T
)
co
llectio
n
in
co
r
p
o
r
at
es
p
ar
allel
co
r
p
o
r
a
f
o
r
m
u
lt
ip
le
I
n
d
ian
lan
g
u
a
g
es,
in
c
lu
d
i
n
g
Hi
n
d
i
an
d
Kan
n
ad
a.
T
h
e
m
o
s
t
r
ec
en
t
v
er
s
io
n
,
W
A
T
2
0
2
1
in
r
esear
ch
[
2
7
]
,
w
a
s
r
elea
s
ed
in
2
0
2
1
an
d
is
p
ar
t
o
f
th
e
w
o
r
k
s
h
o
p
o
n
Asi
an
tr
a
n
s
la
ti
o
n
(
W
A
T
)
jo
in
t
in
it
iativ
e.
T
h
is
d
atase
t
in
cl
u
d
es
m
u
ltil
i
n
g
u
al
tr
an
s
latio
n
d
ata
a
n
d
co
r
r
esp
o
n
d
in
g
d
etails
i
n
E
n
g
l
is
h
.
T
h
e
co
m
p
ilatio
n
co
n
s
i
s
ts
o
f
ap
p
r
o
x
i
m
ate
l
y
1
.
5
m
illi
o
n
p
h
r
ase
p
air
s
ex
tr
ac
ted
f
r
o
m
d
i
v
er
s
e
s
o
u
r
ce
s
s
u
c
h
as
n
e
w
s
,
co
m
m
u
n
ic
atio
n
,
in
f
o
r
m
atio
n
tech
n
o
lo
g
y
,
le
g
al,
an
d
s
cie
n
ti
f
ic
co
n
te
n
t.
E
ac
h
s
e
n
te
n
ce
in
th
e
d
ataset
h
a
s
b
ee
n
to
k
en
ize
d
,
n
o
r
m
alize
d
,
an
d
alig
n
ed
at
th
e
s
e
n
te
n
ce
lev
el,
e
n
s
u
r
in
g
it
s
u
s
ab
ilit
y
f
o
r
v
ar
io
u
s
lan
g
u
a
g
e
-
p
r
o
ce
s
s
i
n
g
ta
s
k
s
.
3
.
1
.
1
.
T
ra
ns
lite
ra
t
io
n
T
h
e
ev
alu
atio
n
o
f
tr
a
n
s
l
iter
at
io
n
m
o
d
els
u
til
izes
p
u
b
licl
y
av
ailab
le
tr
an
s
liter
atio
n
co
r
p
o
r
a,
w
it
h
a
m
aj
o
r
p
o
r
tio
n
o
f
th
e
d
ata
o
b
tain
ed
f
r
o
m
th
e
Da
k
s
h
i
n
a
co
r
p
u
s
[
2
6
]
.
Fig
u
r
e
2
s
h
o
w
s
t
h
e
ac
cu
r
ac
y
v
alu
e
f
o
r
tr
an
s
liter
at
io
n
i
n
Kan
n
ad
a
lan
g
u
a
g
e.
Fi
g
u
r
e
3
s
h
o
w
s
t
h
e
ac
c
u
r
ac
y
v
alu
e
f
o
r
tr
an
s
l
iter
atio
n
in
Hi
n
d
i
lan
g
u
ag
e.
T
h
e
o
b
tain
ed
r
esu
lts
ar
e
th
e
n
co
m
p
ar
ed
b
et
w
ee
n
t
h
e
e
x
is
ti
n
g
s
y
s
te
m
a
n
d
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
f
o
r
th
e
H
in
d
i
an
d
Kan
n
ad
a
la
n
g
u
a
g
e,
u
s
i
n
g
d
ata
f
r
o
m
t
h
e
Da
k
s
h
i
n
a
co
r
p
u
s
.
T
h
e
r
esu
lt
ev
al
u
ated
i
s
r
ep
r
esen
ted
i
n
th
e
f
o
llo
w
in
g
w
h
ich
s
h
o
w
s
t
h
at
t
h
e
ac
cu
r
ac
y
f
o
r
th
e
ex
i
s
ti
n
g
s
y
s
te
m
[
2
7
]
,
[
2
8
]
is
7
7
.
1
8
an
d
th
e
P
o
s
tScr
ip
t
(
PS
)
v
alu
e
g
e
n
er
ates
a
v
alu
e
o
f
8
5
.
8
7
%
f
o
r
Kan
n
ad
a
la
n
g
u
a
g
e
an
d
f
o
r
Hi
n
d
i
la
n
g
u
a
g
e
t
h
e
e
x
is
t
in
g
s
y
s
te
m
[
2
7
]
g
en
er
ate
s
a
v
al
u
e
o
f
6
0
.
5
6
an
d
th
e
P
S g
en
er
ate
s
a
v
al
u
e
o
f
8
5
.
5
6
%.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2089
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
1
,
Ma
r
c
h
2025
:
2
4
3
-
253
250
F
ig
u
r
e
2
.
A
cc
u
r
ac
y
v
alu
e
f
o
r
tr
an
s
liter
atio
n
in
Kan
n
ad
a
lan
g
u
a
g
e
Fig
u
r
e
3
.
A
cc
u
r
ac
y
v
alu
e
f
o
r
tr
an
s
liter
atio
n
in
H
in
d
i
lan
g
u
a
g
e
3
.
1
.
2
.
T
ra
ns
la
t
io
n
T
h
e
ev
alu
atio
n
o
f
m
o
d
els
in
t
h
is
s
t
u
d
y
is
co
n
d
u
c
ted
u
s
i
n
g
b
ilin
g
u
al
ev
a
lu
at
io
n
u
n
d
er
s
tu
d
y
(
B
L
E
U
)
s
co
r
es,
an
d
th
e
ass
e
s
s
m
e
n
t
an
n
o
tatio
n
s
in
cl
u
d
e
Sacr
eB
L
E
U
s
ig
n
at
u
r
es
f
o
r
b
o
th
I
n
d
ic
-
E
n
g
lis
h
2
1
an
d
E
n
g
li
s
h
-
I
n
d
ic2
2
tr
an
s
latio
n
s
,
en
s
u
r
in
g
co
n
s
is
te
n
c
y
,
an
d
r
ep
r
o
d
u
cib
ilit
y
ac
r
o
s
s
m
o
d
els.
Fi
g
u
r
e
4
s
h
o
w
s
th
e
ac
c
u
r
ac
y
co
m
p
ar
is
o
n
f
o
r
tr
an
s
la
tio
n
i
n
Kan
n
ad
a
lan
g
u
a
g
e.
Fig
u
r
e
5
s
h
o
w
s
th
e
ac
c
u
r
ac
y
co
m
p
ar
is
o
n
f
o
r
tr
an
s
lat
io
n
i
n
Hin
d
i
lan
g
u
a
g
e.
T
h
e
p
u
b
licly
av
ailab
le
tr
an
s
lat
io
n
co
r
p
o
r
a
a
r
e
co
m
p
iled
,
w
it
h
a
s
ig
n
i
f
ica
n
t
p
o
r
ti
o
n
o
f
th
e
d
ata
s
o
u
r
ce
d
f
r
o
m
W
A
T
2
0
2
1
[
2
7
]
.
T
h
e
r
esu
lts
o
b
tain
ed
ar
e
t
h
en
co
m
p
ar
ed
b
et
w
ee
n
t
h
e
e
x
is
t
i
n
g
s
y
s
te
m
an
d
t
h
e
p
r
o
p
o
s
ed
s
y
s
te
m
f
o
r
th
e
Ka
n
n
ad
a
lan
g
u
ag
e
f
r
o
m
W
A
T
2
0
2
1
.
m
B
AR
T
in
r
esear
ch
[
2
9
]
g
e
n
er
ates
an
ac
c
u
r
ac
y
v
alu
e
o
f
4
.
3
,
Go
o
g
le
(
GOOG
)
in
r
esear
ch
[
3
0
]
g
en
er
ates
an
ac
cu
r
ac
y
v
al
u
e
o
f
2
5
.
9
,
Mi
cr
o
s
o
f
t
(
MSFT
)
in
r
esear
ch
[
3
1
]
g
en
er
ates
a
v
alu
e
o
f
2
5
.
4
,
T
F
in
r
esear
ch
[
3
2
]
g
en
er
ates
a
v
alu
e
o
f
2
6
.
8
,
M
T
5
in
r
esear
ch
[
3
3
]
g
en
er
ate
s
a
v
alu
e
o
f
2
8
.
5
w
h
er
e
as
th
e
ex
is
ti
n
g
s
y
s
te
m
g
en
er
ates
a
v
alu
e
o
f
3
6
.
2
an
d
th
e
p
r
o
p
o
s
ed
Natio
n
al
Nativ
e
T
itle
T
r
ib
u
n
al
-
P
o
s
tScr
i
p
t
(
NN
T
T
-
PS
)
g
en
er
ates
a
v
al
u
e
o
f
5
6
.
6
9
f
o
r
Kan
n
ad
a
lan
g
u
ag
e.
T
h
e
OP
US
in
r
esear
ch
[
2
9
]
m
et
h
o
d
g
en
er
at
es
a
v
alu
e
o
f
1
3
.
3
,
m
B
AR
T
in
r
esear
ch
[
2
9
]
ge
n
er
ates
an
ac
cu
r
ac
y
v
alu
e
o
f
3
3
.
1
,
GOOG
[
3
0
]
g
en
er
ates
a
n
ac
cu
r
ac
y
v
a
lu
e
o
f
3
6
.
7
.
MSFT
[
3
1
]
g
en
er
ates
a
v
alu
e
o
f
3
8
,
T
F
[
3
2
]
g
en
er
ates
a
v
alu
e
o
f
3
8
.
8
,
M
T
5
[
3
3
]
g
en
er
ate
s
a
v
al
u
e
o
f
3
9
.
2
w
h
er
ea
s
t
h
e
ex
i
s
ti
n
g
s
y
s
te
m
g
en
er
ates
a
v
alu
e
o
f
4
0
.
3
[
3
2
]
an
d
th
e
p
r
o
p
o
s
ed
N
NT
T
-
P
S g
en
er
ates a
v
al
u
e
o
f
7
7
.
5
4
9
7
% f
o
r
Hin
d
i
lan
g
u
a
g
e.
Fig
u
r
e
4
.
A
cc
u
r
ac
y
co
m
p
ar
is
o
n
f
o
r
tr
an
s
latio
n
i
n
Ka
n
n
ad
a
la
n
g
u
a
g
e
7
7
.
1
8
8
5
.
8
7
70
75
80
85
90
Ka
n
n
a
d
a
(ES
)
Ka
n
n
a
d
a
(P
S
)
A
cc
u
r
ac
y
v
alu
e
Me
th
o
d
T
r
an
sl
i
t
erat
i
on
ka
nn
ad
a
0
20
40
60
80
1
0
0
h
in
d
i E
S
h
in
d
i P
S
A
cc
u
r
ac
y
v
alu
e
Me
th
o
d
T
r
a
n
s
l
i
t
e
r
a
t
i
o
n
H
i
n
d
i
0
10
20
30
40
50
60
M
b
a
rt
M
S
F
T
GO
O
G
TF
M
t5
IT
P
S
A
cc
u
r
ac
y
v
alu
e
Me
th
o
d
o
lo
g
y
T
r
an
s
latio
n
Kan
n
ad
a
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
S
elf
-
a
tten
tio
n
en
co
d
er
-
d
ec
o
d
e
r
w
ith
mo
d
el
a
d
a
p
ta
tio
n
fo
r
tr
a
n
s
liter
a
tio
n
a
n
d
…
(
S
h
a
n
th
a
l
a
N
a
g
a
r
a
ja
)
251
Fig
u
r
e
5
.
A
cc
u
r
ac
y
co
m
p
ar
is
o
n
f
o
r
tr
an
s
latio
n
i
n
Hi
n
d
i la
n
g
u
ag
e
3
.
2
.
Co
m
pa
ra
t
iv
e
a
na
ly
s
is
T
ab
le
1
s
h
o
w
s
a
co
m
p
ar
ati
v
e
an
al
y
s
i
s
o
f
t
h
e
ex
is
tin
g
s
y
s
te
m
a
n
d
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
f
o
r
tr
an
s
liter
at
io
n
an
d
tr
an
s
lat
io
n
task
s
i
n
Kan
n
ad
a
an
d
Hin
d
i
lan
g
u
ag
e
s
.
T
h
e
r
esu
lts
in
d
ic
ate
th
e
p
er
ce
n
tag
e
i
m
p
r
o
v
e
m
en
t
ac
h
iev
ed
b
y
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
o
v
er
th
e
ex
i
s
tin
g
s
y
s
te
m
.
Fo
r
tr
an
s
liter
ati
o
n
in
Kan
n
ad
a,
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
d
e
m
o
n
s
tr
ate
s
a
s
ig
n
i
f
ican
t
i
m
p
r
o
v
e
m
e
n
t
o
f
1
0
.
6
6
%
in
ac
cu
r
ac
y
,
ac
h
ie
v
in
g
a
n
i
m
p
r
ess
iv
e
ac
cu
r
ac
y
r
ate
o
f
8
5
.
8
7
%,
co
m
p
ar
ed
to
th
e
ex
is
tin
g
s
y
s
te
m
's
ac
cu
r
ac
y
o
f
7
7
.
1
8
%.
I
n
th
e
ca
s
e
o
f
tr
an
s
liter
atio
n
f
o
r
Hi
n
d
i,
th
e
p
r
o
p
o
s
ed
s
y
s
t
e
m
o
u
tp
er
f
o
r
m
s
t
h
e
ex
is
ti
n
g
s
y
s
te
m
b
y
a
r
e
m
ar
k
ab
le
3
4
.
2
2
%.
T
h
e
p
r
o
p
o
s
e
d
s
y
s
te
m
ac
h
ie
v
e
s
an
ac
cu
r
ac
y
o
f
8
5
.
5
6
%,
w
h
ile
t
h
e
e
x
is
ti
n
g
s
y
s
te
m
la
g
s
w
i
th
a
n
ac
c
u
r
ac
y
o
f
6
0
.
5
6
%.
Mo
v
in
g
o
n
to
tr
an
s
latio
n
in
Kan
n
ad
a,
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
s
h
o
w
s
r
e
m
ar
k
ab
le
p
r
o
g
r
es
s
,
ac
h
ie
v
i
n
g
an
i
m
p
r
o
v
e
m
e
n
t
o
f
4
4
.
1
2
%.
I
t
attain
s
an
ac
cu
r
ac
y
r
ate
o
f
5
6
.
6
9
%,
w
h
ile
t
h
e
ex
is
t
in
g
s
y
s
te
m
f
all
s
s
h
o
r
t
w
ith
an
ac
cu
r
ac
y
o
f
3
6
.
2
%.
Si
m
ilar
l
y
,
f
o
r
tr
an
s
latio
n
in
Hin
d
i,
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
ex
ce
ls
w
it
h
a
r
e
m
ar
k
ab
le
5
4
.
6
3
%
i
m
p
r
o
v
e
m
en
t.
I
t
ac
h
iev
e
s
an
i
m
p
r
e
s
s
i
v
e
ac
cu
r
ac
y
r
ate
o
f
7
6
.
8
9
7
%,
w
h
ile
t
h
e
ex
i
s
ti
n
g
s
y
s
te
m
tr
ai
ls
w
it
h
a
n
ac
cu
r
ac
y
o
f
4
3
.
9
%.
T
h
e
co
m
p
ar
ativ
e
an
al
y
s
is
clea
r
l
y
d
e
m
o
n
s
tr
ates
t
h
e
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
i
n
all
e
v
al
u
ated
task
s
,
s
h
o
w
ca
s
in
g
it
s
ef
f
ec
ti
v
en
e
s
s
i
n
i
m
p
r
o
v
i
n
g
tr
an
s
liter
ati
o
n
an
d
tr
an
s
la
tio
n
ac
cu
r
ac
y
f
o
r
b
o
th
Kan
n
ad
a
a
n
d
Hin
d
i la
n
g
u
a
g
es.
T
ab
le
1
.
C
o
m
p
ar
ativ
e
a
n
al
y
s
i
s
D
a
t
a
se
t
Ex
i
st
i
n
g
sy
st
e
m
P
r
o
p
o
se
d
sy
st
e
m
I
mp
r
o
v
i
sat
i
o
n
i
n
(
%
)
T
r
a
n
sl
i
t
e
r
a
t
i
o
n
(
K
a
n
n
a
d
a
)
7
7
.
1
8
8
5
.
8
7
1
0
.
6
5
9
3
T
r
a
n
sl
i
t
e
r
a
t
i
o
n
(
H
i
n
d
i
)
6
0
.
5
6
8
5
.
5
6
3
4
.
2
1
8
5
T
r
a
n
sl
a
t
i
o
n
(
K
a
n
n
a
d
a
)
3
6
.
2
5
6
.
6
9
4
4
.
1
1
6
7
T
r
a
n
sl
a
t
i
o
n
(
H
i
n
d
i
)
4
3
.
9
7
6
.
8
9
7
5
4
.
6
3
2
2
4.
CO
NCLU
SI
O
N
I
n
co
n
clu
s
io
n
,
t
h
is
r
esear
c
h
i
n
tr
o
d
u
ce
s
a
n
o
v
el
S
A
E
DM
a
r
ch
itect
u
r
e
th
at
le
v
er
ag
es
t
h
e
p
o
w
er
o
f
s
elf
-
atte
n
tio
n
to
co
n
s
tr
u
ct
a
n
au
to
-
e
n
co
d
er
,
av
o
id
in
g
t
h
e
n
e
ed
f
o
r
iter
atio
n
s
o
r
co
n
v
o
lu
tio
n
al
o
p
er
atio
n
s
.
T
h
e
m
o
d
el
e
x
h
ib
its
e
x
ce
p
tio
n
al
p
er
f
o
r
m
a
n
ce
i
n
tr
an
s
lit
er
atio
n
an
d
tr
an
s
latio
n
tas
k
s
f
o
r
Kan
n
ad
a
an
d
Hin
d
i
lan
g
u
a
g
es,
ac
h
iev
i
n
g
s
i
g
n
if
ic
an
t
i
m
p
r
o
v
e
m
e
n
ts
o
v
er
th
e
e
x
is
t
in
g
s
y
s
te
m
s
.
I
n
co
r
p
o
r
atin
g
lex
ica
l
w
eig
h
ti
n
g
an
d
f
in
e
-
t
u
n
in
g
tech
n
iq
u
es,
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
e
f
f
icie
n
tl
y
tr
ain
s
at
th
e
w
o
r
d
lev
el,
r
ed
u
cin
g
tr
ai
n
in
g
ti
m
e
w
h
ile
m
ai
n
tai
n
i
n
g
ac
c
u
r
ac
y
.
Ov
er
all,
t
h
e
r
esear
ch
d
i
s
p
la
y
s
t
h
e
p
o
ten
tia
l
o
f
t
h
e
p
r
o
p
o
s
ed
SA
E
DM
i
n
ad
v
an
ci
n
g
N
L
P
tas
k
s
.
T
h
e
co
m
b
in
at
io
n
o
f
s
e
lf
-
atte
n
tio
n
,
f
i
n
e
-
tu
n
i
n
g
,
d
o
m
ain
ad
ap
tatio
n
,
an
d
b
in
ar
y
clas
s
i
f
ier
co
n
tr
ib
u
tes
to
a
r
o
b
u
s
t
an
d
e
f
f
icie
n
t
s
y
s
te
m
f
o
r
h
an
d
li
n
g
d
iv
er
s
e
li
n
g
u
i
s
tic
ch
al
len
g
es.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el,
f
ac
ilit
ate
s
cr
o
s
s
-
c
u
lt
u
r
al
u
n
d
er
s
tan
d
i
n
g
,
a
n
d
i
m
p
r
o
v
in
g
t
h
e
a
cc
ess
ib
ilit
y
o
f
i
n
f
o
r
m
a
tio
n
i
n
d
iv
er
s
e
la
n
g
u
a
g
es.
As
t
h
e
w
o
r
ld
b
ec
o
m
es
in
cr
ea
s
in
g
l
y
i
n
ter
co
n
n
ec
ted
,
th
e
e
f
f
ec
tiv
e
n
ess
o
f
s
u
c
h
m
o
d
els
in
h
a
n
d
lin
g
co
m
p
le
x
lan
g
u
a
g
e
tas
k
s
w
ill
p
la
y
a
cr
u
cial
r
o
le
in
p
r
o
m
o
ti
n
g
e
f
f
ec
tiv
e
cr
o
s
s
-
lan
g
u
ag
e
co
m
m
u
n
icati
o
n
an
d
in
f
o
r
m
a
tio
n
ac
ce
s
s
.
0
10
20
30
40
50
60
70
80
90
OPUS
M
b
art
GO
O
G
M
S
F
T
TF
M
t5
IT(
ES
)
NN
TT
(P
S
)
A
cc
u
r
ac
y
v
alu
e
Meth
o
d
o
lo
g
y
T
r
an
s
latio
n
Hin
d
i
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2089
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
1
,
Ma
r
c
h
2025
:
2
4
3
-
253
252
RE
F
E
R
E
NC
E
S
[
1
]
Y
.
Z
h
a
o
a
n
d
H
.
L
i
u
,
“
D
o
c
u
me
n
t
-
l
e
v
e
l
n
e
u
r
a
l
m
a
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
w
i
t
h
r
e
c
u
r
r
e
n
t
c
o
n
t
e
x
t
st
a
t
e
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
1
,
p
p
.
2
7
5
1
9
–
2
7
5
2
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
3
.
3
2
4
7
5
0
8
.
[
2
]
K
.
M
r
i
n
a
l
i
n
i
,
P
.
V
i
j
a
y
a
l
a
k
sh
mi
,
a
n
d
T
.
N
a
g
a
r
a
j
a
n
,
“
S
B
S
i
m:
a
se
n
t
e
n
c
e
-
b
e
r
t
si
mi
l
a
r
i
t
y
-
b
a
se
d
e
v
a
l
u
a
t
i
o
n
me
t
r
i
c
f
o
r
i
n
d
i
a
n
l
a
n
g
u
a
g
e
n
e
u
r
a
l
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
sy
st
e
ms,”
I
EEE/
A
C
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
ss
i
n
g
,
v
o
l
.
3
0
,
p
p
.
1
3
9
6
–
1
4
0
6
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
T
A
S
L
P
.
2
0
2
2
.
3
1
6
1
1
6
0
.
[
3
]
A
.
K
u
m
a
r
,
A
.
P
r
a
t
a
p
,
a
n
d
A
.
K
.
S
i
n
g
h
,
“
G
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
u
r
a
l
mac
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
f
o
r
p
h
o
n
e
t
i
c
l
a
n
g
u
a
g
e
s
v
i
a
r
e
i
n
f
o
r
c
e
me
n
t
l
e
a
r
n
i
n
g
,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
Em
e
r
g
i
n
g
T
o
p
i
c
s
i
n
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
T
ETCI)
,
v
o
l
.
7
,
n
o
.
1
,
p
p
.
1
9
0
–
1
9
9
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
T
ET
C
I
.
2
0
2
2
.
3
2
0
9
3
9
4
.
[
4
]
S
.
B
h
a
t
i
a
,
A
.
K
u
mar,
a
n
d
M
.
M
.
K
h
a
n
,
“
R
o
l
e
o
f
g
e
n
e
t
i
c
a
l
g
o
r
i
t
h
m
i
n
o
p
t
i
mi
z
a
t
i
o
n
o
f
H
i
n
d
i
w
o
r
d
se
n
se
d
i
sam
b
i
g
u
a
t
i
o
n
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
7
5
6
9
3
–
7
5
7
0
7
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
1
9
0
4
0
6
.
[
5
]
S
.
S
a
i
n
i
a
n
d
V
.
S
a
h
u
l
a
,
“
N
e
u
r
a
l
mac
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
f
o
r
E
n
g
l
i
sh
t
o
H
i
n
d
i
,
”
i
n
2
0
1
8
F
o
u
rt
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
f
o
rm
a
t
i
o
n
Re
t
ri
e
v
a
l
a
n
d
K
n
o
w
l
e
d
g
e
Ma
n
a
g
e
m
e
n
t
(
C
A
MP
)
,
I
EEE,
M
a
r
.
2
0
1
8
,
p
p
.
1
–
6
.
d
o
i
:
1
0
.
1
1
0
9
/
I
N
F
R
K
M
.
2
0
1
8
.
8
4
6
4
7
8
1
.
[
6
]
B
.
Z
h
a
n
g
,
D
.
X
i
o
n
g
,
a
n
d
J.
S
u
,
“
N
e
u
r
a
l
mac
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
w
i
t
h
d
e
e
p
a
t
t
e
n
t
i
o
n
,
”
I
EE
E
T
ra
n
sa
c
t
i
o
n
s
o
n
P
a
t
t
e
rn
A
n
a
l
y
si
s
a
n
d
Ma
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
4
2
,
n
o
.
1
,
p
p
.
1
5
4
–
1
6
3
,
Ja
n
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
T
P
A
M
I
.
2
0
1
8
.
2
8
7
6
4
0
4
.
[
7
]
Y
.
N
i
s
h
i
m
u
r
a
,
K
.
S
u
d
o
h
,
G
.
N
e
u
b
i
g
,
a
n
d
S
.
N
a
k
a
m
u
r
a
,
“
M
u
l
t
i
-
so
u
r
c
e
n
e
u
r
a
l
mac
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
w
i
t
h
m
i
ssi
n
g
d
a
t
a
,
”
I
EEE
/
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
.
2
8
,
p
p
.
5
6
9
–
5
8
0
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
T
A
S
L
P
.
2
0
1
9
.
2
9
5
9
2
2
4
.
[
8
]
H
.
M
o
o
n
,
C
.
P
a
r
k
,
S
.
Eo
,
J
.
S
e
o
,
a
n
d
H
.
L
i
m,
“
A
n
e
mp
i
r
i
c
a
l
st
u
d
y
o
n
a
u
t
o
mat
i
c
p
o
st
e
d
i
t
i
n
g
f
o
r
n
e
u
r
a
l
mac
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
,
”
I
E
EE
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
1
2
3
7
5
4
–
1
2
3
7
6
3
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
1
0
9
9
0
3
.
[
9
]
Y
.
F
a
n
,
F
.
T
i
a
n
,
Y
.
X
i
a
,
T
.
Q
i
n
,
X
.
-
Y
.
L
i
,
a
n
d
T
.
-
Y
.
L
i
u
,
“
S
e
a
r
c
h
i
n
g
b
e
t
t
e
r
a
r
c
h
i
t
e
c
t
u
r
e
s
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
,
”
I
EEE/
A
C
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
.
2
8
,
p
p
.
1
5
7
4
–
1
5
8
5
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
T
A
S
L
P
.
2
0
2
0
.
2
9
9
5
2
7
0
.
[
1
0
]
O
.
S
e
n
e
t
a
l
.
,
“
B
a
n
g
l
a
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ssi
n
g
:
a
c
o
m
p
r
e
h
e
n
si
v
e
a
n
a
l
y
si
s
o
f
c
l
a
ssi
c
a
l
,
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
-
b
a
se
d
me
t
h
o
d
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
3
8
9
9
9
–
3
9
0
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
6
5
5
6
3
.
[
1
1
]
J.
A
.
O
v
i
,
M
.
A
.
I
sl
a
m,
a
n
d
M
.
R
.
K
a
r
i
m,
“
B
a
N
e
P
:
a
n
e
n
d
-
to
-
e
n
d
n
e
u
r
a
l
n
e
t
w
o
r
k
b
a
se
d
mo
d
e
l
f
o
r
b
a
n
g
l
a
p
a
r
t
s
-
of
-
sp
e
e
c
h
t
a
g
g
i
n
g
,
”
I
EE
E
A
c
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
1
0
2
7
5
3
–
1
0
2
7
6
9
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
2
0
8
2
6
9
.
[
1
2
]
U
.
K
.
A
c
h
a
r
j
e
e
,
M
.
A
r
e
f
i
n
,
K
.
M
.
H
o
sse
n
,
M
.
N
.
U
d
d
i
n
,
M
.
A
.
U
d
d
i
n
,
a
n
d
L
.
I
sl
a
m,
“
S
e
q
u
e
n
c
e
-
to
-
se
q
u
e
n
c
e
l
e
a
r
n
i
n
g
-
b
a
se
d
c
o
n
v
e
r
si
o
n
o
f
p
se
u
d
o
-
c
o
d
e
t
o
so
u
r
c
e
c
o
d
e
u
s
i
n
g
n
e
u
r
a
l
t
r
a
n
sl
a
t
i
o
n
a
p
p
r
o
a
c
h
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
2
6
7
3
0
–
2
6
7
4
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
1
5
5
5
5
8
.
[
1
3
]
Q
.
D
u
,
N
.
X
u
,
Y
.
L
i
,
T
.
X
i
a
o
,
a
n
d
J.
Z
h
u
,
“
T
o
p
o
l
o
g
y
-
s
e
n
si
t
i
v
e
n
e
u
r
a
l
a
r
c
h
i
t
e
c
t
u
r
e
se
a
r
c
h
f
o
r
l
a
n
g
u
a
g
e
mo
d
e
l
i
n
g
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
1
0
7
4
1
6
–
1
0
7
4
2
3
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
1
0
1
2
5
5
.
[
1
4
]
O
.
F
i
r
a
t
,
K
.
C
h
o
,
a
n
d
Y
.
B
e
n
g
i
o
,
“
M
u
l
t
i
-
w
a
y
,
mu
l
t
i
l
i
n
g
u
a
l
n
e
u
r
a
l
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
w
i
t
h
a
s
h
a
r
e
d
a
t
t
e
n
t
i
o
n
me
c
h
a
n
i
sm,
”
in
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
1
6
C
o
n
f
e
re
n
c
e
o
f
t
h
e
N
o
r
t
h
Am
e
r
i
c
a
n
C
h
a
p
t
e
r
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
st
i
c
s:
H
u
m
a
n
L
a
n
g
u
a
g
e
T
e
c
h
n
o
l
o
g
i
e
s
,
Ja
n
.
2
0
1
6
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
N
1
6
-
1
1
0
1
.
[
1
5
]
S
.
H
.
S
r
e
e
d
h
a
r
a
,
V
.
K
u
mar,
a
n
d
S
.
S
a
l
ma,
“
Ef
f
i
c
i
e
n
t
b
i
g
d
a
t
a
c
l
u
s
t
e
r
i
n
g
u
si
n
g
a
d
h
o
c
f
u
z
z
y
c
me
a
n
s
a
n
d
a
u
t
o
-
e
n
c
o
d
e
r
C
N
N
,
”
i
n
L
e
c
t
u
re
N
o
t
e
s
i
n
N
e
t
w
o
r
k
s
a
n
d
S
y
st
e
m
s
,
S
p
r
i
n
g
e
r
,
S
i
n
g
a
p
o
r
e
,
2
0
2
3
,
p
p
.
3
5
3
–
3
6
8
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
19
-
7
4
0
2
-
1
_
2
5
.
[
1
6
]
F
.
A
q
l
a
n
,
X
.
F
a
n
,
A
.
A
l
q
w
b
a
n
i
,
a
n
d
A
.
A
l
-
M
a
n
so
u
b
,
“
A
r
a
b
i
c
–
C
h
i
n
e
se
n
e
u
r
a
l
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
:
r
o
man
i
z
e
d
a
r
a
b
i
c
a
s
s
u
b
w
o
r
d
u
n
i
t
f
o
r
a
r
a
b
i
c
-
so
u
r
c
e
d
t
r
a
n
sl
a
t
i
o
n
,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
7
,
p
p
.
1
3
3
1
2
2
–
1
3
3
1
3
5
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
9
.
2
9
4
1
1
6
1
.
[
1
7
]
S
.
A
.
A
l
maa
y
t
a
h
a
n
d
S
.
A
.
A
l
z
o
b
i
d
y
,
“
C
h
a
l
l
e
n
g
e
s
i
n
r
e
n
d
e
r
i
n
g
a
r
a
b
i
c
t
e
x
t
t
o
e
n
g
l
i
s
h
u
s
i
n
g
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
:
a
sy
st
e
mat
i
c
l
i
t
e
r
a
t
u
r
e
r
e
v
i
e
w
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
1
,
p
p
.
9
4
7
7
2
–
9
4
7
7
9
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
3
.
3
3
0
9
6
4
2
.
[
1
8
]
A
.
S
o
b
h
y
,
M
.
H
e
l
m
y
,
M
.
K
h
a
l
i
l
,
S
.
El
masr
y
,
Y
.
B
o
u
l
e
s,
a
n
d
N
.
N
e
g
i
e
d
,
“
A
n
a
i
b
a
se
d
a
u
t
o
mat
i
c
t
r
a
n
sl
a
t
o
r
f
o
r
a
n
c
i
e
n
t
h
i
e
r
o
g
l
y
p
h
i
c
l
a
n
g
u
a
g
e
—
f
r
o
m
sca
n
n
e
d
i
mag
e
s
t
o
E
n
g
l
i
sh
t
e
x
t
,
”
I
E
EE
Ac
c
e
ss
,
v
o
l
.
1
1
,
p
p
.
3
8
7
9
6
–
3
8
8
0
4
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
3
.
3
2
6
7
9
8
1
.
[
1
9
]
J.
Z
a
k
r
a
o
u
i
,
M
.
S
a
l
e
h
,
S
.
A
l
-
M
a
a
d
e
e
d
,
a
n
d
J.
M
.
A
l
j
a
’
a
m,
“
A
r
a
b
i
c
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
:
a
s
u
r
v
e
y
w
i
t
h
c
h
a
l
l
e
n
g
e
s
a
n
d
f
u
t
u
r
e
d
i
r
e
c
t
i
o
n
s,
”
I
E
EE
A
c
c
e
ss
,
v
o
l
.
9
,
p
p
.
1
6
1
4
4
5
–
1
6
1
4
6
8
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
1
3
2
4
8
8
.
[
20]
M
.
S
h
a
h
r
o
z
,
M
.
F
.
M
u
s
h
t
a
q
,
A
.
M
e
h
mo
o
d
,
S
.
U
l
l
a
h
,
a
n
d
G
.
S
.
C
h
o
i
,
“
R
U
T
U
T
:
r
o
man
u
r
d
u
t
o
u
r
d
u
t
r
a
n
sl
a
t
o
r
b
a
se
d
o
n
c
h
a
r
a
c
t
e
r
su
b
s
t
i
t
u
t
i
o
n
r
u
l
e
s a
n
d
u
n
i
c
o
d
e
m
a
p
p
i
n
g
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
8
,
p
p
.
1
8
9
8
2
3
–
1
8
9
8
4
1
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
3
0
3
1
3
9
3
.
[
2
1
]
S
.
K
a
u
sar,
B
.
T
a
h
i
r
,
a
n
d
M
.
A
.
M
e
h
mo
o
d
,
“
P
r
o
S
O
U
L
:
a
f
r
a
mew
o
r
k
t
o
i
d
e
n
t
i
f
y
p
r
o
p
a
g
a
n
d
a
f
r
o
m
o
n
l
i
n
e
u
r
d
u
c
o
n
t
e
n
t
,
”
I
EE
E
Ac
c
e
ss
,
v
o
l
.
8
,
p
p
.
1
8
6
0
3
9
–
1
8
6
0
5
4
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
3
0
2
8
1
3
1
.
[
2
2
]
K
.
C
h
e
n
,
R
.
W
a
n
g
,
M
.
U
t
i
y
a
ma,
a
n
d
E.
S
u
mi
t
a
,
“
I
n
t
e
g
r
a
t
i
n
g
p
r
i
o
r
t
r
a
n
sl
a
t
i
o
n
k
n
o
w
l
e
d
g
e
i
n
t
o
n
e
u
r
a
l
mac
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
,
”
I
EEE/
A
C
M
T
ra
n
s
a
c
t
i
o
n
s
o
n
Au
d
i
o
,
S
p
e
e
c
h
,
a
n
d
L
a
n
g
u
a
g
e
P
ro
c
e
ssi
n
g
,
v
o
l
.
3
0
,
p
p
.
3
3
0
–
3
3
9
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
T
A
S
L
P
.
2
0
2
1
.
3
1
3
8
7
1
4
.
[
2
3
]
P
.
P
u
j
a
r
,
A
.
K
u
mar,
a
n
d
V
.
K
u
mar
,
“
Ef
f
i
c
i
e
n
t
p
l
a
n
t
l
e
a
f
d
e
t
e
c
t
i
o
n
t
h
r
o
u
g
h
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
b
a
se
d
o
n
c
o
r
n
l
e
a
f
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
I
A
ES
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
I
J
-
AI
)
,
v
o
l
.
1
3
,
n
o
.
1
,
p
.
1
1
3
9
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
a
i
.
v
1
3
.
i
1
.
p
p
1
1
3
9
-
1
1
4
8
.
[
2
4
]
Z
.
T
a
n
,
Z
.
Y
a
n
g
,
M
.
Z
h
a
n
g
,
Q
.
L
i
u
,
M
.
S
u
n
,
a
n
d
Y
.
L
i
u
,
“
D
y
n
a
mi
c
mu
l
t
i
-
b
r
a
n
c
h
l
a
y
e
r
s
f
o
r
o
n
-
d
e
v
i
c
e
n
e
u
r
a
l
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
,
”
I
EEE/
A
C
M
T
ra
n
s
a
c
t
i
o
n
s
o
n
Au
d
i
o
,
S
p
e
e
c
h
,
a
n
d
L
a
n
g
u
a
g
e
P
ro
c
e
ssi
n
g
,
v
o
l
.
3
0
,
p
p
.
9
5
8
–
9
6
7
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
T
A
S
L
P
.
2
0
2
2
.
3
1
5
3
2
5
7
.
[
2
5
]
G
.
H
.
N
g
o
,
M
.
N
g
u
y
e
n
,
a
n
d
N
.
F
.
C
h
e
n
,
“
P
h
o
n
o
l
o
g
y
-
a
u
g
me
n
t
e
d
st
a
t
i
s
t
i
c
a
l
f
r
a
me
w
o
r
k
f
o
r
mac
h
i
n
e
t
r
a
n
sl
i
t
e
r
a
t
i
o
n
u
si
n
g
l
i
m
i
t
e
d
l
i
n
g
u
i
s
t
i
c
r
e
so
u
r
c
e
s,”
I
EEE
/
A
C
M
T
r
a
n
sa
c
t
i
o
n
s
o
n
Au
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
.
2
7
,
n
o
.
1
,
p
p
.
1
9
9
–
2
1
1
,
Ja
n
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
T
A
S
L
P
.
2
0
1
8
.
2
8
7
5
2
6
9
.
[
2
6
]
B
.
R
o
a
r
k
e
t
a
l
.
,
“
P
r
o
c
e
ssi
n
g
so
u
t
h
a
s
i
a
n
l
a
n
g
u
a
g
e
s
w
r
i
t
t
e
n
i
n
t
h
e
l
a
t
i
n
scr
i
p
t
:
t
h
e
D
a
k
s
h
i
n
a
d
a
t
a
se
t
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
T
w
e
l
f
t
h
L
a
n
g
u
a
g
e
Re
s
o
u
rc
e
s
a
n
d
Ev
a
l
u
a
t
i
o
n
C
o
n
f
e
re
n
c
e
,
J
u
l
.
2
0
2
0
.
[
2
7
]
G
.
R
a
me
sh
e
t
a
l
.
,
“
S
a
ma
n
a
n
t
a
r
:
t
h
e
l
a
r
g
e
st
p
u
b
l
i
c
l
y
a
v
a
i
l
a
b
l
e
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
f
o
r
1
1
I
n
d
i
c
l
a
n
g
u
a
g
e
s,”
T
ra
n
s
a
c
t
i
o
n
s
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
st
i
c
s
,
v
o
l
.
1
0
,
p
p
.
1
4
5
–
1
6
2
,
F
e
b
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
6
2
/
t
a
c
l
_
a
_
0
0
4
5
2
.
[
2
8
]
J.
T
i
e
d
e
man
n
a
n
d
S
.
T
h
o
t
t
i
n
g
a
l
,
“
O
P
U
S
-
mt
–
b
u
i
l
d
i
n
g
o
p
e
n
t
r
a
n
sl
a
t
i
o
n
se
r
v
i
c
e
s fo
r
t
h
e
w
o
r
l
d
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
2
n
d
An
n
u
a
l
C
o
n
f
e
re
n
c
e
o
f
t
h
e
Eu
r
o
p
e
a
n
Asso
c
i
a
t
i
o
n
f
o
r
M
a
c
h
i
n
e
T
r
a
n
sl
a
t
i
o
n
,
L
i
sb
o
a
,
P
o
r
t
u
g
a
l
:
Eu
r
o
p
e
a
n
A
sso
c
i
a
t
i
o
n
f
o
r
M
a
c
h
i
n
e
T
r
a
n
sl
a
t
i
o
n
,
2
0
2
0
,
p
p
.
4
7
9
–
4
8
0
.
[
2
9
]
Y
.
T
a
n
g
e
t
a
l
.
,
“
M
u
l
t
i
l
i
n
g
u
a
l
t
r
a
n
sl
a
t
i
o
n
w
i
t
h
e
x
t
e
n
si
b
l
e
mu
l
t
i
l
i
n
g
u
a
l
p
r
e
t
r
a
i
n
i
n
g
a
n
d
f
i
n
e
t
u
n
i
n
g
,
”
a
rX
i
v
,
A
u
g
.
2
0
2
0
,
d
o
i
:
1
0
.
4
8
5
5
0
/
a
r
X
i
v
.
2
0
0
8
.
0
0
4
0
1
,
[
3
0
]
M
.
Jo
h
n
so
n
e
t
a
l
.
,
“
G
o
o
g
l
e
’
s
mu
l
t
i
l
i
n
g
u
a
l
n
e
u
r
a
l
ma
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
sy
st
e
m:
e
n
a
b
l
i
n
g
z
e
r
o
-
sh
o
t
t
r
a
n
sl
a
t
i
o
n
,
”
T
r
a
n
s
a
c
t
i
o
n
s
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
,
v
o
l
.
5
,
N
o
v
.
2
0
1
6
.
[
3
1
]
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
,
”
in
3
1
s
t
C
o
n
f
e
r
e
n
c
e
o
n
N
e
u
ra
l
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ss
i
n
g
S
y
st
e
m
s
(
N
I
PS
2
0
1
7
)
,
L
o
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.