I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
11
, N
o.
1
,
M
a
r
c
h
2022
, pp.
319
~
326
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
11
.i
1
.pp
319
-
326
319
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
Wi
k
i
se
n
se
b
ag c
r
e
at
i
on
u
si
n
g
m
u
l
t
i
l
i
n
gu
al
w
or
d
se
n
se
d
i
sam
b
i
gu
at
i
o
n
S
h
r
e
ya P
at
an
k
ar
1
, M
ad
h
u
r
a
P
h
ad
k
e
2
, S
a
t
is
h
D
e
van
e
3
1
,2
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
E
ngi
ne
e
r
i
ng, D
a
t
t
a
M
e
ghe
C
ol
l
e
ge
of
E
ngi
ne
e
r
i
ng, N
a
vi
M
um
ba
i
, I
ndi
a
3
D
e
pa
r
t
m
e
nt
of
I
nf
or
m
a
t
i
on
T
e
c
hnol
ogy, D
a
t
t
a
M
e
ghe
C
ol
l
e
ge
of
E
ngi
ne
e
r
i
ng, N
a
vi
M
um
ba
i
, I
ndi
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
J
ul
1
,
2021
R
e
vi
s
e
d
D
e
c
2
2
,
2021
A
c
c
e
pt
e
d
J
a
n
2
,
202
2
Perfor
mance
of
word
sense
disambig
uation
(WSD)
is
one
of
the
chall
enging
tasks
in
the
area
of
natural
language
processing
(NLP).
Generation
of
sense
annotated
corpus
for
multi
lingual
word
sense
disambi
guation
is
out
of
reach
for
most
languages
even
if
resources
are
available.
In
this
paper
we
p
ropose
an
unsupervised
method
using
word
and
sense
embedding
or
improvi
ng
the
performance
of
these
systems
using
untagged.
Corpora
and
create
tw
o
bags
namely
ontological
bag
and
wiki
sense
bag
to
generate
the
sense
s
with
highest
similarity.
Wiki
sense
bag
provides
external
know
ledge
to
the
system
required
to
boost
the
disambiguation
accuracy.
We
e
xplore
Word2Vec mo
del to
generate th
e sense
K
e
y
w
o
r
d
s
:
M
ul
ti
li
ngua
l
N
a
tu
r
a
l
la
ngua
ge
pr
oc
e
s
s
in
g
W
or
d s
e
ns
e
di
s
a
m
bi
gua
ti
on
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
S
hr
e
ya
P
a
ta
nka
r
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
E
ngi
ne
e
r
in
g, D
a
tt
a
M
e
ghe
C
ol
le
ge
of
E
ngi
ne
e
r
in
g
S
e
c
to
r
-
3,
A
ir
ol
i,
O
pp
K
ha
ndoba
T
e
m
pl
e
S
r
i
S
a
dgur
u
V
a
na
m
r
a
o
P
a
i
M
a
r
g,
N
a
vi
M
um
ba
i,
M
a
ha
r
a
s
ht
r
a
,
I
ndi
a
E
m
a
il
:
s
np.c
m
.dm
c
e
@
gm
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
I
nc
r
e
a
s
in
g
de
m
a
nd
s
by
th
e
us
e
r
to
a
c
c
e
s
s
te
xt
d
a
ta
in
v
a
r
io
us
la
ngua
ge
s
ope
n
s
up
th
e
door
s
of
m
ul
ti
li
ngua
l
na
tu
r
a
l
la
ngua
ge
pr
oc
e
s
s
in
g
(
N
L
P
)
a
nd
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
(
W
S
D
)
ha
s
pr
ove
d
to
be
a
ke
y
s
te
p
in
pe
r
f
or
m
a
nc
e
im
pr
ove
m
e
nt
of
m
a
ny
N
L
P
s
ys
te
m
s
.T
he
a
c
c
ur
a
c
y
of
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
s
ys
te
m
s
is
f
a
r
f
r
om
be
in
g
s
a
ti
s
f
a
c
to
r
y
a
nd
m
ul
ti
li
ngua
l
W
S
D
ha
s
not
a
c
hi
e
ve
d
s
a
ti
s
f
a
c
to
r
y
r
e
s
ul
ts
due
to
in
s
uf
f
ic
ie
nt
r
e
s
our
c
e
a
va
il
a
bi
li
ty
[
1]
.
T
he
a
va
il
a
bi
li
ty
of
m
u
lt
il
in
gua
l
di
c
ti
ona
r
ie
s
ha
s
e
nha
nc
e
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
us
in
g
m
ul
ti
li
ngua
l
c
ont
e
nt
w
hi
c
h
de
pi
c
ts
th
e
ne
e
d
f
or
m
ul
ti
li
ngua
l
W
S
D
[
2
]
.
I
t
a
ls
o
ope
ns
up
a
di
f
f
e
r
e
nt
w
a
y
of
a
ppr
oa
c
hi
ng
m
ul
ti
li
ngua
l
W
S
D
by
m
a
ki
ng
us
e
of
B
a
be
lNe
t,
a
w
id
e
ont
ol
ogi
c
a
l
s
tr
uc
tu
r
e
e
xpl
or
in
g
s
e
m
a
nt
ic
knowle
dge
.
T
hi
s
is
th
e
m
ot
iv
a
ti
on
f
or
w
or
ki
ng
on
m
ul
ti
li
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on by e
xpl
or
in
g t
he
a
va
il
a
bl
e
r
e
s
our
c
e
s
.
R
e
ly
in
g only on m
ul
ti
li
ngua
l
knowle
dge
-
ba
s
e
d s
ys
te
m
m
a
y ha
m
pe
r
t
he
gr
ow
th
of
W
S
D
s
ys
te
m
s
a
nd
th
ough
m
ul
ti
li
ngua
l
di
c
ti
ona
r
ie
s
pr
ovi
de
w
id
e
c
ove
r
a
ge
e
xpl
or
in
g
th
e
in
te
r
c
onne
c
te
d
ont
ol
ogy
s
tr
uc
tu
r
e
,
va
r
io
us
is
s
ue
s
s
ti
ll
r
e
m
a
in
to
be
s
e
e
n
s
uc
h
a
s
pr
ope
r
nouns
a
r
e
not
pa
r
t
of
th
e
di
c
ti
ona
r
y
a
nd
c
or
r
e
la
ti
on
be
twe
e
n
m
os
t
f
r
e
que
nt
w
or
ds
a
nd
r
a
r
e
c
ont
e
xt
ua
l
w
or
ds
la
c
k
di
c
ti
ona
r
y
c
ove
r
a
ge
.
E
xt
e
r
na
l
knowle
dge
in
te
r
m
s
of
r
a
w
te
xt
is
ne
e
d
e
d
w
hi
c
h
i
s
pr
ovi
de
d
u
s
in
g
w
or
d
a
nd
s
e
ns
e
e
m
be
ddi
ng
[
3]
.
O
ur
r
e
s
e
a
r
c
h
m
a
k
e
s
u
s
e
of
w
or
d
a
nd
s
e
ns
e
e
m
be
ddi
ng
s
to
c
r
e
a
te
a
s
e
m
a
nt
ic
w
or
d
c
lo
ud
by
de
s
ig
ni
ng
a
w
ik
i
ba
g
in
a
ddi
ti
on
to
th
e
s
e
ns
e
b
a
g.
W
ik
i
ba
g
i
s
de
s
ig
ne
d
us
in
g
W
ik
ip
e
di
a
a
s
it
is
th
e
l
a
r
ge
s
t
e
nc
yc
lo
pe
di
a
w
hi
c
h
c
ove
r
s
m
os
t
of
th
e
da
ta
ba
s
e
e
s
s
e
nt
ia
l
f
or
di
s
a
m
bi
gua
ti
on.
T
he
pa
pe
r
is
or
ga
ni
z
e
d
be
in
g
a
s
:
s
e
c
ti
on
2
pr
e
s
e
nt
s
th
e
li
te
r
a
tu
r
e
r
e
vi
e
w
w
hi
c
h
hi
ghl
ig
ht
s
th
e
r
e
s
e
a
r
c
h
w
or
k
of
va
r
io
us
r
e
s
e
a
r
c
he
r
s
,
s
e
c
ti
on
3
d
e
s
c
r
ib
e
s
th
e
pr
opos
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
11
, N
o.
1
,
M
a
r
c
h 20
22
:
319
-
326
320
m
e
th
odol
ogy
us
e
d
w
hi
c
h
in
c
lu
de
s
w
or
ki
ng
w
it
h
m
ul
ti
li
ngua
l
i
nput
,
m
ul
ti
li
ngua
l
di
c
ti
ona
r
y
B
a
be
lNe
t
a
nd
th
e
w
or
ki
ng of
W
S
D
e
ngi
ne
. S
e
c
ti
on 4 f
oc
us
e
s
on r
e
s
ul
ts
a
nd dis
c
u
s
s
io
ns
a
nd
s
e
c
ti
on 5
s
um
s
up with c
onc
lu
s
io
n.
2.
L
I
T
E
R
A
T
U
R
E
R
E
V
I
E
W
W
or
d2V
e
c
m
ode
l
[
4]
–
[
13]
pr
ovi
de
s
a
n
e
f
f
ic
ie
nt
to
o
l
f
or
e
s
ti
m
a
ti
ng
ve
c
to
r
m
ode
l
us
in
g
th
e
c
or
pus
.
A
s
e
ns
e
ba
g
w
a
s
c
r
e
a
te
d
[
14]
m
a
ki
ng
us
e
of
di
c
ti
ona
r
y
r
e
s
our
c
e
s
s
u
c
h
a
s
s
yns
e
t
m
e
m
be
r
s
,
e
xa
m
pl
e
s
e
nt
e
nc
e
s
,
hype
r
nym
y
a
nd
hyponymy
s
ubs
e
ts
.
A
s
ur
ve
y
w
a
s
pr
e
s
e
nt
e
d
on
W
S
D
[
15]
hi
ghl
ig
ht
in
g
th
e
m
ot
iv
a
ti
on
f
or
s
ol
vi
ng
th
e
a
m
bi
gui
ty
of
w
or
ds
a
nd
pr
ovi
di
ng
de
s
c
r
ip
ti
on
of
th
e
ta
s
k.
T
h
e
c
onc
e
pt
of
W
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
in
m
ul
ti
li
ngua
l
s
e
tt
in
g
[
16]
in
tr
oduc
e
s
by
m
a
ki
ng
us
e
of
la
r
ge
e
nc
yc
lo
pe
di
c
ont
ol
ogi
c
a
l
ne
twor
k
B
a
be
lNe
t.
P
r
e
c
is
io
n
a
c
hi
e
ve
d
w
a
s
54.3
%
w
he
n
te
s
te
d
on
S
e
m
E
va
l
2010
da
ta
s
e
t.
I
n
2013,
A
z
iz
a
nd
S
pe
c
ia
[
17]
di
s
c
us
s
e
s
e
xpr
e
s
s
in
g m
e
a
ni
ngs
i
n t
e
r
m
s
of
pa
r
a
phr
a
s
e
s
.
T
he
r
ol
e
of
W
S
D
f
or
m
ul
ti
li
ngua
l
s
c
e
na
r
io
of
N
L
P
te
xt
w
a
s
s
ur
ve
ye
d
u
s
in
g
E
ngl
is
h
-
S
pa
ni
s
h
la
ngua
ge
s
[
18]
.
W
S
D
in
m
ul
ti
li
ngua
l
m
a
c
hi
ne
tr
a
n
s
la
ti
on
(
M
T
)
is
ba
s
e
d
on
th
e
c
on
c
e
pt
th
a
t
r
e
s
our
c
e
f
ul
l
la
ngua
ge
he
lp
s
a
r
e
s
our
c
e
lo
w
la
ngua
ge
by
pr
oj
e
c
ti
ng
pa
r
a
m
e
te
r
s
li
ke
s
e
ns
e
di
s
tr
ib
ut
io
ns
,
a
nd
c
or
pus
c
o
-
oc
c
ur
r
e
nc
e
s
[
19]
.
T
he
a
c
c
ur
a
c
y
obs
e
r
ve
d
w
a
s
75%
f
or
th
r
e
e
la
ngua
ge
s
w
it
h
dom
a
in
s
pe
c
if
ic
c
or
pus
.
W
S
D
in
N
L
P
a
ppl
ic
a
ti
ons
is
a
l
s
o
di
s
c
u
s
s
e
d
[
20]
.
C
r
os
s
-
li
ngua
l
W
S
D
s
ys
te
m
s
w
a
s
di
s
c
us
s
e
d
[
21]
,
a
nd
e
va
lu
a
te
d
on
S
e
m
E
va
l
2010
ta
s
k.
M
a
c
hi
ne
tr
a
ns
la
ti
on
is
one
of
th
e
im
por
ta
nt
a
ppl
ic
a
ti
o
ns
of
W
S
D
a
nd
is
di
s
c
us
s
e
d
[
22]
,
[
23]
.
A
s
ur
ve
y
of
te
xt
c
l
a
s
s
if
ic
a
ti
on
of
K
ur
di
s
h
la
ngua
g
e
is
be
a
ut
if
ul
ly
pr
e
s
e
nt
e
d
[
24]
–
[
27]
w
he
r
e
th
e
y
a
ppl
ie
d
s
te
m
m
e
r
a
lg
or
it
hm
to
f
in
d
th
e
s
te
m
to
pe
r
f
or
m
c
la
s
s
if
ic
a
ti
on.
W
S
D
ne
twor
k
a
ppr
oa
c
h,
s
e
nt
im
e
nt
a
na
ly
s
is
a
nd
s
ur
ve
y
i
s
e
xpl
or
e
d
[
28]
–
[
31]
.
I
t
is
obs
e
r
ve
d
t
ha
t
not
m
uc
h
w
or
k
is
r
e
por
te
d
on
W
S
D
in
m
ul
ti
li
ngua
l
s
e
tt
in
g
to
th
e
be
s
t
of
our
knowle
dge
a
nd
it
ne
e
ds
to
be
e
xpl
or
e
d
us
in
g
va
r
io
us
s
ta
te
of
th
e
a
r
t
W
S
D
m
e
th
ods
.
3.
P
R
O
P
O
S
E
D
M
E
T
H
O
D
O
L
O
G
Y
T
he
pr
opos
e
d
m
e
th
odol
ogy
is
pr
e
s
e
nt
e
d
in
th
e
F
ig
ur
e
1
a
nd
w
e
pr
e
s
e
nt
th
e
c
onc
e
pt
of
r
e
pr
e
s
e
nt
in
g
m
ul
ti
li
ngua
l
in
put
da
ta
in
s
e
c
ti
on
3.1.
I
t
in
c
lu
de
s
a
c
c
e
pt
in
g
m
ul
ti
li
ngua
l
in
put
w
hi
c
h
w
il
l
be
ne
f
it
th
e
e
ngi
ne
.
E
xt
e
r
na
l
knowle
dge
i
s
a
ls
o pr
ovi
de
d t
o t
he
s
y
s
te
m
us
in
g s
e
ns
e
e
m
be
ddi
ngs
.
F
ig
ur
e
1. P
r
opos
e
d m
ul
ti
li
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on (
W
S
D
)
f
r
a
m
e
w
or
k
3.1.
M
u
lt
il
in
gu
al
i
n
p
u
t
W
e
c
ons
id
e
r
he
r
e
in
put
f
r
om
va
r
io
us
la
ngua
ge
s
li
ke
G
e
r
m
a
n
a
n
d
F
r
e
nc
h
a
nd
m
a
ke
us
e
of
B
a
b
e
l
N
e
t
m
ul
ti
li
ngua
l
di
c
ti
ona
r
y
de
s
c
r
ib
e
d
in
s
e
c
ti
on
3.2.
T
hi
s
is
done
to
e
xpl
or
e
va
r
io
us
la
ngua
ge
s
a
nd
ta
ki
ng
he
lp
f
r
om
ot
he
r
la
ngua
ge
s
im
pr
ove
s
th
e
s
ys
te
m
a
c
c
ur
a
c
y.
A
m
b
ig
uous
w
or
d
in
one
la
ngua
ge
m
a
y
not
b
e
a
m
bi
guous
i
n ot
he
r
l
a
ngua
ge
a
nd t
hi
s
w
il
l
be
ne
f
it
t
he
s
ys
te
m
e
n
gi
ne
f
or
i
m
pr
ovi
ng t
he
a
c
c
ur
a
c
y.
3.2.
B
ab
e
lNe
t
T
he
B
a
b
e
lNe
t
is
a
hug
e
m
ul
ti
li
ngua
l
ont
ol
ogi
c
a
l
ne
twor
k
in
c
or
por
a
ti
ng
le
xi
c
a
l
s
e
m
a
nt
ic
a
nd
s
ynt
a
c
ti
c
knowle
dge
f
r
om
va
r
io
us
la
ngua
ge
s
[
1]
.
I
t
r
e
pr
e
s
e
nt
s
a
la
be
ll
e
d
gr
a
ph
s
pe
c
if
yi
ng
s
e
m
a
nt
ic
r
e
la
ti
ons
be
twe
e
n
va
r
io
us
node
s
a
nd
e
dge
s
.
I
t
c
om
bi
ne
s
th
e
knowle
dge
of
va
r
io
us
la
ngua
ge
W
or
dN
e
t
a
nd
la
r
ge
s
t
m
ul
ti
li
ngua
l
e
nc
yc
lo
pe
di
a
. S
e
c
ti
on
3.3 r
e
pr
e
s
e
nt
s
t
he
w
or
ki
ng of
W
S
D
w
it
h
th
e
a
lg
or
it
hm
f
or
t
he
s
a
m
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
W
ik
i
s
e
ns
e
bag c
r
e
at
io
n us
in
g m
ul
ti
li
ngual w
or
d
s
e
ns
e
di
s
am
bi
guat
io
n
(
Shr
e
y
a P
at
ank
ar
)
321
3.3.
Wor
d
s
e
n
s
e
d
is
a
m
b
ig
u
at
io
n
(
WS
D
)
e
n
gi
n
e
W
S
D
e
ngi
ne
ta
ke
s
th
e
m
ul
ti
li
ngua
l
in
put
by
e
xpl
or
in
g
va
r
io
us
la
ngua
ge
s
a
lt
oge
th
e
r
a
t
th
e
s
a
m
e
ti
m
e
.
I
t
c
om
bi
ne
s
th
e
tr
a
ns
la
ti
ons
of
ta
r
ge
t
w
or
d
a
nd
ot
he
r
c
ont
e
xt
w
or
ds
to
pr
oduc
e
m
or
e
a
c
c
ur
a
te
s
e
ns
e
pr
e
di
c
ti
ons
.
S
e
ns
e
di
s
a
m
bi
gua
ti
on
be
gi
ns
by
ga
th
e
r
in
g
th
e
da
ta
r
e
qui
r
e
d
f
or
di
s
a
m
bi
gua
ti
on
w
he
r
e
th
e
di
f
f
e
r
e
nt
s
e
ns
e
s
of
th
e
a
m
bi
guous
w
or
d
a
r
e
c
ol
le
c
te
d
in
S
r
e
pr
e
s
e
nt
e
d
a
s
s
ynonyms
e
t
f
r
om
th
e
B
a
be
lNe
t.
C
ont
e
xt
w
or
ds
a
r
e
c
ol
le
c
te
d
in
C
tx
a
nd
th
e
a
lg
or
it
hm
th
e
n
pr
oc
e
e
ds
by
pi
c
ki
ng
up
th
e
m
ul
ti
li
ngua
l
tr
a
ns
la
ti
ons
of
t
he
a
m
bi
guous
a
nd c
lu
e
w
or
ds
s
to
r
e
d i
n T
x a
nd T
y r
e
s
pe
c
ti
ve
ly
.
T
r
a
n
s
la
ti
ons
a
r
e
c
on
s
id
e
r
e
d
in
F
r
e
nc
h
a
nd
G
e
r
m
a
n
la
ngua
ge
s
a
s
f
or
e
ig
n
la
ngu
a
ge
s
a
r
e
e
xpl
or
e
d.
T
h
e
a
lg
or
it
hm
it
e
r
a
te
s
th
r
ough
e
a
c
h
s
yns
e
t
s
€ S
t
o c
ol
le
c
t
th
e
t
r
a
ns
la
ti
ons
of
e
a
c
h of
i
ts
s
e
n
s
e
s
[
7]
.
A
lg
or
it
hm
a
ls
o
it
e
r
a
te
s
th
r
ough
e
a
c
h c
ont
e
xt
w
or
d
c
i
€
C
tx
to
c
ol
le
c
t
th
e
tr
a
ns
la
ti
ons
in
T
y
in
s
e
ns
e
-
s
pe
c
if
ic
G
e
r
m
a
n
a
nd
F
r
e
nc
h
tr
a
ns
la
ti
ons
.
E
le
m
e
nt
ti
is
s
e
le
c
te
d
f
r
om
T
x
a
nd
e
le
m
e
nt
tj
is
s
e
le
c
te
d
f
r
om
T
y
a
nd
a
m
ul
ti
li
ngua
l
c
ont
e
xt
µ’
is
c
r
e
a
te
d
by
c
om
bi
ni
ng
ti
a
nd
tj
w
it
h
th
e
C
tx
.
T
he
va
r
ia
bl
e
µ’
is
u
s
e
d
to
bu
il
d
a
gr
a
ph
G
=
{
V
,
E
}
by
c
om
put
in
g
th
e
pa
th
s
in
B
a
be
lNe
t
w
hi
c
h
c
onne
c
ts
th
e
s
yn
s
e
ts
of
ti
w
it
h
th
o
s
e
of
ot
he
r
w
or
ds
in
µ’
a
s
s
how
n
in
F
ig
ur
e
2. B
y
s
e
le
c
ti
ng a
t
e
a
c
h
s
t
e
p,
a
di
f
f
e
r
e
nt
e
le
m
e
nt
f
r
om
T
,
a
ne
w
gr
a
ph
is
c
r
e
a
te
d
w
he
r
e
di
f
f
e
r
e
nt
s
e
ts
of
B
a
be
l
s
yns
e
t
s
ge
t
a
c
ti
va
te
d by the
c
ont
e
xt
w
or
ds
i
n C
tx
. T
he
r
e
s
ul
t
of
t
hi
s
pr
oc
e
dur
e
i
s
a
s
ubgr
a
ph
of
B
a
be
lN
e
t
c
ont
a
in
in
g
th
e
s
e
ns
e
s
of
th
e
w
or
ds
in
th
e
c
ont
e
xt
a
nd
a
ll
e
dge
s
a
nd
in
te
r
m
e
di
a
te
s
e
ns
e
s
f
ound
in
B
a
be
lNe
t
a
lo
ng
a
ll
pa
th
s
c
onne
c
ti
ng
th
e
m
.
F
ig
ur
e
2
s
how
s
th
e
di
s
a
m
bi
gua
ti
on
gr
a
ph
c
r
e
a
te
d
to
di
s
a
m
bi
gua
te
th
e
E
ngl
is
h
la
ngua
g
e
ta
r
ge
t
w
or
d
‘
bank
’
.
I
n
th
e
gr
a
ph,
s
om
e
of
th
e
po
s
s
ib
le
s
e
ns
e
s
of
th
is
w
or
d
a
r
e
a
c
ti
va
te
d
in
c
lu
di
ng
th
e
c
or
r
e
c
t
s
e
ns
e
(
ba
nk
2
E
N
G
L
I
S
H
)
but
a
ls
o
r
e
la
te
d
ye
t
in
c
or
r
e
c
t
one
is
a
c
ti
va
te
d
(
ba
nk
9
EN
G
L
I
S
H
)
.
F
ig
ur
e
2.
D
is
a
m
bi
gua
ti
on gr
a
ph f
or
E
ngl
is
h l
a
ngua
ge
3.4.
S
c
or
in
g d
is
t
r
ib
u
t
io
n
S
c
or
in
g
di
s
tr
ib
ut
io
n
is
c
a
lc
ul
a
te
d
us
in
g
th
e
I
nve
r
s
e
pa
th
le
ngt
h s
um
m
e
a
s
ur
e
.
I
t
s
c
or
e
s
e
a
c
h s
e
ns
e
by
s
um
m
in
g
ove
r
th
e
in
ve
r
s
e
le
ngt
h
of
a
ll
pa
th
s
w
hi
c
h
c
onne
c
t
it
to
ot
he
r
s
e
n
s
e
s
in
th
e
gr
a
ph.
I
t
is
ve
r
y
us
e
f
ul
f
or
s
e
ns
e
di
s
a
m
bi
gua
ti
on a
nd i
m
pr
ove
s
t
h
e
a
c
c
ur
a
c
y.
s
c
o
r
e
j
=
1
ℎ
(
)
−
1
(
1)
W
he
r
e
pa
th
s
(
s
j)
i
s
t
he
s
e
t
of
s
im
pl
e
pa
th
s
c
onne
c
ti
ng s
j
to
th
e
s
e
ns
e
s
of
ot
he
r
c
ont
e
xt
w
or
ds
. L
e
ngt
h
(
p)
is
t
he
num
be
r
of
e
dge
s
in
th
e
pa
th
p
a
nd
e
a
c
h
pa
th
i
s
s
c
or
e
d
w
it
h
th
e
e
xpone
nt
ia
l
in
ve
r
s
e
de
c
a
y
of
th
e
pa
th
l
e
ngt
h.
S
c
or
e
s
a
r
e
c
a
lc
ul
a
te
d
a
nd
s
to
r
e
d
in
Δ
s
c
or
e
a
nd
in
th
e
f
in
a
l
s
te
p;
c
os
in
e
di
s
ta
nc
e
s
im
il
a
r
it
y
m
e
a
s
ur
e
is
c
a
lc
ul
a
te
d
to
f
in
d
th
e
m
a
xi
m
um
s
c
or
e
w
hi
c
h
de
te
r
m
in
e
s
th
e
c
l
os
e
ne
s
s
be
twe
e
n
th
e
a
m
bi
guous
w
or
d
a
nd
th
e
c
ont
e
xt
w
or
ds
. T
he
c
os
in
e
di
s
t
a
nc
e
f
or
m
ul
a
i
s
pr
e
s
e
nt
e
d
in
(
2)
:
C
o
s
(
S
,
SC
(
T
)
)
=
∑
∗
(
)
=
1
√
∑
2
=
0
∗
√
∑
(
)
2
=
1
(
2
)
w
he
r
e
S
is
ve
c
to
r
r
e
pr
e
s
e
nt
in
g
th
e
s
c
or
e
of
a
m
bi
guous
w
or
d
s
,
S
C
(
T
)
i
s
ve
c
to
r
r
e
pr
e
s
e
nt
in
g
th
e
s
c
or
e
of
c
ont
e
xt
w
or
ds
.
G
lo
ba
l
s
c
or
e
c
on
s
is
ts
of
s
e
le
c
ti
ng
th
e
hi
ghe
s
t
s
c
or
e
r
e
pr
e
s
e
nt
e
d
a
nd
a
s
a
r
e
s
ul
t
of
e
xe
c
ut
io
n
of
a
lg
or
it
hm
;
th
e
s
c
or
in
g
di
s
tr
ib
ut
io
n
w
hi
c
h
is
m
a
xi
m
um
is
r
e
tu
r
ne
d
to
s
e
le
c
t
th
e
be
s
t
di
s
a
m
bi
gua
ti
on
s
e
ns
e
.
S
e
c
ti
ons
3.5
a
nd
3.6
r
e
pr
e
s
e
nt
s
th
e
us
e
of
de
e
p
le
a
r
ni
ng
to
ol
s
to
r
e
pr
e
s
e
nt
th
e
di
c
ti
ona
r
y
f
r
a
m
e
w
or
k
in
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
11
, N
o.
1
,
M
a
r
c
h 20
22
:
319
-
326
322
num
e
r
ic
r
e
pr
e
s
e
nt
a
ti
on.
T
a
bl
e
1
r
e
pr
e
s
e
nt
s
th
e
s
c
or
in
g
di
s
tr
ib
ut
i
on
us
in
g
th
e
a
bov
e
f
or
m
ul
a
f
or
th
e
two
s
e
ns
e
s
of
ba
nk na
m
e
ly
bui
ld
in
g
s
e
ns
e
ba
nk
9
E
N
G
L
I
S
H
a
nd f
in
a
nc
ia
l
in
s
ti
t
ut
io
n ba
nk
2
E
N
G
L
I
S
H
.
T
a
bl
e
1.
S
c
or
in
gdi
s
tr
ib
ut
io
n
L
a
ngua
ge
ba
nk
2
E
N
G
L
I
S
H
ba
nk
9
E
N
G
L
I
S
H
ba
nkE
N
0.666666666
0.3333333333
ba
nkG
E
R
M
A
N
0.333333333
0
ba
nque
F
R
E
N
C
H
0.444444444
0
3.5.
S
yn
s
e
t
d
ic
t
io
n
ar
y f
r
am
e
w
o
r
k
O
ur
s
tu
dy
e
xpl
or
e
s
th
e
ont
ol
ogy
of
e
a
c
h
s
e
ns
e
de
f
in
it
io
n
f
r
o
m
th
e
di
c
ti
ona
r
y
na
m
e
ly
hype
r
nym
,
hyponym,
hol
onymy,
a
nd
gl
os
s
.
a
s
s
yn
s
e
t
m
e
m
be
r
s
a
lo
ne
a
r
e
not
s
uf
f
ic
ie
nt
f
or
id
e
nt
if
yi
ng
th
e
c
or
r
e
c
t
s
e
ns
e
.
S
om
e
of
s
yns
e
t
s
ha
ve
a
v
e
r
y
s
m
a
ll
num
be
r
of
s
yns
e
t
m
e
m
be
r
s
a
nd
th
e
ot
he
r
r
e
a
s
on
is
to
br
in
g
dow
n
to
pi
c
dr
if
t
w
hi
c
h
m
a
y
ha
ve
oc
c
ur
r
e
d
b
e
c
a
u
s
e
of
pol
y
s
e
m
ous
s
yns
e
t
m
e
m
be
r
s
.
I
t
is
a
ls
o
obs
e
r
ve
d
th
a
t
a
ddi
ng
gl
os
s
of
hype
r
nym
/h
yponym gi
ve
s
be
tt
e
r
pe
r
f
or
m
a
nc
e
c
om
pa
r
e
d t
o s
yns
e
t
m
e
m
be
r
s
of
hype
r
nym
/h
yponym
[
5]
.
3.6.
Wor
d
an
d
s
e
n
s
e
e
m
b
e
d
d
in
g
T
he
r
e
is
a
ne
e
d
to
br
in
g
th
e
c
lu
e
w
or
ds
a
nd
a
m
bi
guous
w
o
r
ds
to
ge
th
e
r
w
hi
c
h
is
done
u
s
in
g
w
or
d
e
m
be
ddi
ngs
.
I
t
r
e
pr
e
s
e
nt
s
e
m
b
e
ddi
ng
c
ont
in
uous
ve
c
to
r
s
pa
c
e
w
it
h
le
s
s
e
r
di
m
e
ns
io
n
s
a
nd
w
or
d
e
m
be
ddi
ng
a
r
e
tr
a
in
e
d
us
in
g
w
or
d2V
e
c
to
ol
[
4]
.
T
he
tr
a
in
in
g
pr
oc
e
e
ds
by
pr
e
s
e
nt
in
g
di
f
f
e
r
e
nt
c
ont
e
xt
-
ta
r
ge
t
w
or
ds
pa
ir
f
r
om
th
e
c
or
pus
th
us
pr
e
pa
r
in
g
a
n
e
ns
e
m
bl
e
m
ode
l
f
or
a
ll
th
e
a
m
bi
guous
w
or
d
s
in
th
e
vo
c
a
bul
a
r
y
a
s
pr
e
s
e
nt
e
d
in
F
ig
ur
e
3.
T
h
e
c
or
pus
e
n
s
e
m
bl
e
m
ode
l
of
ve
c
to
r
s
r
e
pr
e
s
e
nt
s
th
e
c
lo
s
e
ne
s
s
of
th
e
c
ont
e
xt
-
ta
r
ge
t
pa
ir
f
or
s
pe
c
if
ic
s
e
ns
e
a
nd
to
th
e
be
s
t
of
our
knowle
dge
,
th
is
is
th
e
f
ir
s
t
o
f
th
e
k
in
d
a
tt
e
m
pt
to
ge
ne
r
a
te
s
e
ns
e
s
pe
c
if
ic
w
or
d
ve
c
to
r
m
ode
l
w
hi
c
h
r
e
pr
e
s
e
nt
s
c
lo
s
e
pr
oxi
m
it
y
be
twe
e
n
th
e
c
ont
e
xt
w
or
d
s
a
nd
a
m
bi
guous
w
or
d i
n t
he
ve
c
to
r
s
pa
c
e
. S
e
c
ti
on 3.6 r
e
pr
e
s
e
nt
s
our
c
ont
r
ib
ut
io
n of
s
e
ns
e
ba
g
c
r
e
a
ti
on.
F
ig
ur
e
3. C
or
pus
ba
s
e
d e
n
s
e
m
bl
e
ve
c
to
r
m
ode
l
3.7. S
e
n
s
e
b
ag c
r
e
at
io
n
S
e
ns
e
s
pe
c
if
ic
ve
c
to
r
m
ode
l
is
r
e
pr
e
s
e
nt
e
d
by
e
xt
r
a
c
ti
ng
f
e
a
tu
r
e
s
f
r
om
th
e
le
xi
c
a
l
ont
ol
ogy
a
s
w
e
ll
a
s
e
nc
yc
lo
p
e
di
c
kno
w
le
dge
.
W
or
ds
a
r
e
r
e
pr
e
s
e
nt
e
d
by
r
e
tr
ie
vi
ng
th
e
c
ont
e
xt
w
or
ds
f
r
om
th
e
ont
ol
ogi
c
a
l
s
tr
uc
tu
r
e
of
e
a
c
h
s
e
n
s
e
s
uc
h
a
s
s
yn
s
e
t
m
e
m
be
r
s
,
gl
o
s
s
or
e
xa
m
pl
e
s
e
nt
e
nc
e
,
r
e
la
ti
ons
s
u
c
h
a
s
hyp
e
r
nym
or
hyponym.
W
or
d2V
e
c
m
ode
l
i
s
a
la
ye
r
e
d
n
e
ur
a
l
ne
twor
k
s
tr
uc
t
ur
e
th
a
t
p
r
oc
e
s
s
e
s
th
e
te
xt
by
c
onve
r
ti
ng
th
e
m
in
to
ve
c
to
r
s
;
a
num
e
r
ic
a
l
f
or
m
w
hi
c
h
br
in
gs
r
e
la
te
d
w
or
ds
to
ge
th
e
r
.
T
he
in
put
to
th
e
n
e
ur
a
l
ne
twor
k
is
w
in
dow
of
w
or
ds
,
hi
dde
n
la
y
e
r
c
om
pr
is
e
s
of
w
e
ig
ht
m
a
tr
ix
a
nd
out
put
is
ve
c
to
r
r
e
pr
e
s
e
nt
a
ti
on
of
w
or
ds
.
W
ik
i
s
e
ns
e
ba
g
i
s
a
ls
o
c
r
e
a
te
d
w
hi
c
h
is
ve
c
to
r
r
e
pr
e
s
e
nt
a
ti
on
of
W
ik
ip
e
di
a
of
a
m
bi
guous
w
or
d
s
.
T
hi
s
is
don
e
s
o
a
s
to
pr
ovi
de
a
ddi
ti
ona
l
w
or
ld
knowle
dge
to
th
e
W
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
e
ngi
ne
a
s
W
ik
i
s
e
n
s
e
ba
g
c
ove
r
s
m
a
xi
m
um
voc
a
bul
a
r
y
ne
e
de
d
to
br
in
g
c
ont
e
xt
-
ta
r
ge
t
pa
ir
s
c
lo
s
ur
e
in
th
e
ve
c
to
r
s
pa
c
e
.
W
ik
i
ba
g
c
r
e
a
ti
on i
s
r
e
pr
e
s
e
nt
e
d i
n F
ig
ur
e
4 a
nd S
im
il
a
r
it
y m
e
a
s
ur
e
i
s
c
a
l
c
ul
a
te
d i
n s
e
c
ti
on 3.8.
3.8. S
im
il
ar
it
y m
e
as
u
r
e
T
he
s
im
il
a
r
it
y
m
e
a
s
ur
e
is
c
a
lc
ul
a
te
d
by
c
ons
id
e
r
in
g
th
e
c
os
in
e
s
im
il
a
r
it
y
be
twe
e
n
th
e
w
or
d
r
e
pr
e
s
e
nt
a
ti
on
of
c
ont
e
xt
ve
c
to
r
a
nd
s
e
ns
e
ba
g
r
e
pr
e
s
e
nt
a
ti
on.
I
t
he
lp
s
to
ge
ne
r
a
te
a
s
im
il
a
r
it
y
s
c
or
e
w
hi
c
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
W
ik
i
s
e
ns
e
bag c
r
e
at
io
n us
in
g m
ul
ti
li
ngual w
or
d
s
e
ns
e
di
s
am
bi
guat
io
n
(
Shr
e
y
a P
at
ank
ar
)
323
he
lp
s
in
th
e
di
s
a
m
bi
gua
ti
on
pr
oc
e
s
s
.
C
o
s
in
e
s
im
il
a
r
it
y
m
e
a
s
u
r
e
ha
s
pr
ove
n
to
be
m
or
e
us
e
f
ul
in
th
e
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on pr
oc
e
s
s
.
C
o
s
(
v
e
c
(
w
)
,
v
e
c
(
SB
)
)
=
∑
∗
=
1
√
∑
2
=
1
∗
√
∑
2
=
1
(
3)
W
he
r
e
ve
c
(
w
)
is
th
e
w
or
d
e
m
be
ddi
ng
f
or
w
or
d
w
,
S
B
r
e
pr
e
s
e
nt
s
th
e
s
e
n
s
e
ba
g
a
nd
ve
c
(
S
B
)
is
th
e
s
e
ns
e
e
m
be
ddi
ng
r
e
pr
e
s
e
nt
in
g
th
e
c
om
bi
ne
d
s
c
or
e
of
ont
ol
ogy
ba
g
a
nd
th
e
w
ik
i
s
e
ns
e
ba
g.
S
e
ns
e
d
is
a
m
bi
gu
a
ti
on
(
S
D
)
is
pe
r
f
or
m
e
d
by
s
um
m
in
g
th
e
s
c
or
e
s
of
(
1)
-
(
3)
w
hi
c
h
r
e
pr
e
s
e
nt
s
m
ul
ti
li
ngua
l
W
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
s
im
il
a
r
it
y
s
c
or
e
,
w
or
d
e
m
be
ddi
ng
a
nd
s
e
ns
e
e
m
be
ddi
ng
s
c
or
e
s
of
ont
ol
ogy
ba
g
a
nd
w
ik
i
s
e
ns
e
ba
g
to
boo
s
t
th
e
di
s
a
m
bi
gua
ti
on
a
c
c
ur
a
c
y.
T
h
e
out
put
of
th
e
W
S
D
e
ngi
ne
r
e
s
ul
t
s
in
di
s
a
m
bi
gua
te
d
s
e
ns
e
w
hi
c
h
i
s
c
onve
r
te
d
in
to
ne
ut
r
a
l
la
ngua
g
e
c
ode
to
be
us
e
d
f
or
M
T
.
S
e
c
ti
on
3.9
r
e
pr
e
s
e
nt
s
th
e
f
or
m
a
ti
on
of
ne
ut
r
a
l
la
ngua
ge
c
ode
.
-
3.1438863e
-
03
2.5703609e
-
03
2.3864100e
-
03
-
1.6323227e
-
03
6.4692349
e
-
04
2.6351425e
-
03
-
4.3628053e
-
03
4.3827966e
-
03
-
4.7720312e
-
03
-
3.4716928e
-
03
3.4759669e
-
03
4.3763947e
-
03
3.2847153e
-
03
2.3355209e
-
03
2.8738815e
-
03
-
2.2687481e
-
03
-
4.8421333e
-
03
3.0184705e
-
03
-
2.1880846e
-
03
1.8512266e
-
03
1.6703347e
-
03
-
6.8748498e
-
04
-
6.2847714e
-
04
-
3.0067556e
-
03
3.0463885e
-
03
-
3.5307638e
-
03
2.7850315e
-
03
3.9292048e
-
04
-
2.6362720e
-
03
-
3.6856441e
-
03
2.7092642e
-
04
1.5298135e
-
04
-
4.8553180e
-
03
3.8366476e
-
03
-
2.4513335e
-
03
3.6468427e
-
03
2.3314022e
-
03
1.7899536e
-
03
-
4.3625557e
-
03
3.3640813e
-
03
-
1.8001328e
-
03
1.4276117e
-
03
-
1.1264355e
-
03
-
4.4314810e
-
03
4.2599617e
-
03
1.2551763e
-
03
3.8926408e
-
03
2.4237178e
-
04
-
4.3531498e
-
03
2.6536058e
-
03
-
3.3246232e
-
03
4.0993919e
-
03
F
ig
ur
e
4. W
ik
i
s
e
ns
e
ba
g
c
r
e
a
ti
on
3.9.
N
e
u
t
r
al
l
an
gu
age
c
od
e
W
or
ds
a
f
te
r
di
s
a
m
bi
gua
ti
on
a
r
e
c
onv
e
r
te
d
in
to
uni
que
r
e
pr
e
s
e
nt
a
ti
on
te
r
m
e
d
a
s
ne
ut
r
a
l
la
ngu
a
ge
c
ode
is
f
or
m
e
d
us
in
g
bi
na
r
y
c
om
bi
na
ti
on
of
30
-
bi
t
uni
que
c
ode
w
he
r
e
e
a
c
h
bi
t
r
e
pr
e
s
e
nt
s
s
ig
ni
f
ic
a
nt
in
f
or
m
a
ti
on
a
bout
th
e
di
s
a
m
bi
gua
te
d
pol
y
s
e
m
y
noun
r
e
pr
e
s
e
nt
e
d
in
T
a
bl
e
2.
N
e
ut
r
a
l
l
a
ngua
ge
c
ode
is
uni
que
in
th
e
s
e
ns
e
th
a
t
it
c
ove
r
s
a
ll
th
e
in
f
or
m
a
ti
on
ot
he
r
th
a
n
s
e
ns
e
id
e
nt
if
ic
a
ti
on
a
nd
pa
r
ts
of
s
pe
e
c
h.
R
e
s
ul
ts
a
r
e
pr
e
s
e
nt
e
d i
n t
he
ne
xt
s
e
c
ti
on.
T
a
bl
e
2. N
e
ut
r
a
l
la
ngua
g
e
c
ode
N
oun
C
ode
B
a
nk
0001000110101100101011110011xx
000
-
pa
r
t
s
of
s
pe
e
c
h
0001
-
uni
que
i
de
nt
i
f
i
c
a
t
i
on
1101
-
T
ype
of
noun
011
-
num
be
r
001
-
ge
nde
r
11110000
-
t
e
ns
e
s
xx
-
r
e
s
e
r
ve
d bi
t
s
4.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
W
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
f
r
a
m
e
w
or
k
c
om
pr
is
e
s
of
m
ul
ti
li
ng
ua
l
in
put
a
nd
e
va
lu
a
ti
on
is
pe
r
f
or
m
e
d
on
a
m
a
nua
ll
y
c
r
e
a
te
d
c
or
pus
f
or
E
ngl
is
h
la
ngua
ge
c
on
s
is
ti
ng
of
25
pol
ys
e
m
ous
noun
s
,
f
or
E
ngl
is
h
le
xi
c
a
l
s
a
m
pl
e
ta
s
k.
E
xpe
r
im
e
nt
s
w
e
r
e
pe
r
f
or
m
e
d
w
it
h
5000
in
s
ta
nc
e
s
out
of
w
hi
c
h
70%
w
a
s
u
s
e
d
f
or
t
r
a
in
in
g
a
nd
30%
f
or
te
s
ti
ng.
T
e
s
t
in
s
ta
nc
e
s
w
e
r
e
a
l
s
o
c
ol
le
c
te
d
f
r
om
va
r
io
us
s
e
a
r
c
h
e
ngi
n
e
s
books
a
nd
th
e
a
c
c
ur
a
c
y
obs
e
r
ve
d
f
or
m
ul
ti
li
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
is
40%
a
s
c
om
pa
r
e
d
to
25%
obs
e
r
ve
d
f
or
m
onol
in
gua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on.
T
a
bl
e
3
pr
e
s
e
nt
s
c
om
pa
r
is
on
of
th
e
two
s
ys
te
m
s
a
nd
r
e
s
ul
ts
a
r
e
pr
e
s
e
nt
e
d
f
or
1
0
pol
ys
e
m
y
nouns
.
F
or
s
im
pl
ic
it
y
w
e
c
ons
id
e
r
two
s
e
ns
e
s
e
a
c
h
f
or
pol
ys
e
m
y
nouns
.
T
he
s
ys
te
m
w
a
s
te
s
t
e
d
us
in
g
m
ul
ti
li
ngua
l
a
ppr
oa
c
h
a
nd
obs
e
r
ve
d
a
c
c
ur
a
c
y
w
a
s
im
pr
ove
d
by
15
%
.
T
he
ove
r
a
ll
a
c
c
ur
a
c
y
obs
e
r
ve
d
w
a
s
40%
. O
bs
e
r
va
ti
ons
a
nd f
in
di
ngs
a
r
e
pr
e
s
e
nt
e
d i
n
s
e
c
ti
on 4.
1.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
11
, N
o.
1
,
M
a
r
c
h 20
22
:
319
-
326
324
T
a
bl
e
3. M
onol
in
gua
l
v
e
r
s
us
m
ul
ti
li
ngua
l
w
or
d s
e
ns
e
di
s
a
m
bi
g
ua
ti
on
E
ngl
i
s
h
S
e
ns
e
A
c
c
ur
a
c
y i
n %
f
or
M
onol
i
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
t
i
on
A
c
c
ur
a
c
y i
n %
f
or
M
ul
t
i
l
i
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
t
i
on
C
hi
ps
S
i
l
i
c
on c
hi
p
25
45
W
a
f
e
r
s
24
40
T
a
bl
e
F
ur
ni
t
ur
e
30
35
R
ow
/
c
ol
um
n
32
43
B
a
t
M
a
m
m
a
l
25
45
S
por
t
s
27
45
B
a
nk
F
i
na
nc
e
32
45
R
i
ve
r
ba
nk
32
47
T
a
nk
M
i
l
i
t
a
r
y t
a
nk
25
44
P
l
a
nt
I
ndus
t
r
y pl
a
nt
35
44
T
r
e
e
35
47
S
t
oc
k
C
a
pi
t
a
l
30
43
S
t
or
a
ge
29
40
P
a
l
m
H
a
nd
28
44
N
a
m
e
of
t
r
e
e
26
43
A
c
c
ount
B
a
nk a
c
c
ount
35
43
W
r
i
t
e
up
35
45
4.1.
O
b
s
e
r
vat
io
n
s
an
d
f
in
d
in
gs
T
he
pr
obl
e
m
of
s
im
il
a
r
s
c
or
e
f
a
c
e
d
in
m
onol
in
gua
l
a
ppr
oa
c
h
w
a
s
e
li
m
in
a
te
d
us
in
g
m
ul
ti
li
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on.
O
b
s
e
r
ve
d
a
c
c
ur
a
c
y
is
40%
w
hi
c
h
i
s
f
a
r
le
s
s
th
a
n
th
e
ba
s
e
li
ne
a
c
c
ur
a
c
y
obs
e
r
ve
d
f
or
m
os
t
f
r
e
que
nt
s
e
ns
e
.
I
t
is
a
ls
o
ob
s
e
r
ve
d
th
a
t
pr
ope
r
nouns
li
ke
M
a
dhur
a
,
S
hr
e
ya
s
f
r
om
our
in
s
ta
nc
e
s
w
e
r
e
not
pa
r
t
of
th
e
di
c
ti
ona
r
y
de
f
in
it
io
ns
w
hi
c
h
f
a
il
e
d
to
ge
ne
r
a
te
pr
ope
r
s
c
or
e
s
.
A
ls
o,
di
c
ti
ona
r
y
de
f
in
it
io
n
be
in
g
s
hor
t
la
c
ks
s
tr
ong c
lu
e
s
w
hi
c
h f
a
il
t
he
di
s
a
m
bi
gua
ti
on a
c
c
ur
a
c
y.
F
e
a
tu
r
e
s
of
B
a
be
lNe
t
s
e
ns
e
s
a
r
e
e
xt
r
a
c
te
d
f
r
om
th
e
s
yns
e
t
(
S
)
,
gl
os
s
of
s
yns
e
t
m
e
m
be
r
(
G
)
,
hype
r
nym
y
(
H
)
,
hyponymy
(
H
P
)
,
s
yns
e
t
gl
os
s
of
h
ype
r
nym
y
-
hyponym
y
r
e
la
ti
on
(
H
G
)
,
hol
onymy
(
H
O
)
a
nd
gl
os
s
of
hol
onymy
(
H
O
G
)
.
W
e
te
s
te
d
th
e
s
e
f
e
a
tu
r
e
s
on
2000
i
ns
ta
nc
e
s
a
nd
r
e
s
ul
t
s
a
r
e
r
e
pr
e
s
e
nt
e
d
by
ta
ki
ng
th
e
m
a
xi
m
um
of
th
e
gl
oba
l
s
c
or
e
s
r
e
c
e
iv
e
d
r
e
pr
e
s
e
nt
e
d
in
T
a
bl
e
4.
I
t
is
obs
e
r
ve
d
f
r
om
th
e
T
a
bl
e
4
th
a
t
c
om
bi
ni
ng
a
l
l
th
e
f
e
a
tu
r
e
s
of
B
a
be
lN
e
t
s
e
n
s
e
s
to
ge
th
e
r
gi
ve
s
u
s
a
n
im
pr
ove
d
a
c
c
ur
a
c
y
of
50%
.
I
t
s
how
s
th
a
t
c
om
bi
ni
ng
a
ll
th
e
f
e
a
tu
r
e
s
to
ge
th
e
r
yi
e
ld
s
s
ig
ni
f
ic
a
nt
im
pr
ove
m
e
nt
in
th
e
di
s
a
m
bi
gua
ti
on
pr
oc
e
s
s
.
M
ul
ti
li
ngua
l
a
ppr
oa
c
h
im
pl
e
m
e
nt
s
gr
a
ph
-
ba
s
e
d
di
s
a
m
bi
gua
ti
on
a
nd
w
e
obs
e
r
ve
d
th
a
t
m
a
ny
c
lu
e
w
or
ds
f
r
om
th
e
c
ont
e
xt
w
e
r
e
not
in
c
lo
s
e
pr
oxi
m
it
y
w
it
h
th
e
a
m
bi
guou
s
w
or
ds
.
M
a
ny
w
or
ds
c
lo
s
e
ly
r
e
l
a
te
d
a
r
e
a
t
di
s
ta
nc
e
f
r
om
one
a
not
he
r
a
nd
th
is
be
in
g
one
of
th
e
im
por
ta
nt
f
in
di
ngs
r
e
s
ul
ts
in
le
s
s
s
c
or
e
w
hi
c
h
a
f
f
e
c
ts
th
e
di
s
a
m
b
ig
ua
ti
on
pr
oc
e
s
s
.
W
or
ds
in
s
im
il
a
r
c
ont
e
xt
n
e
e
ds
to
c
o
m
e
c
lo
s
e
f
or
im
pr
ove
th
e
a
c
c
ur
a
c
y.
W
or
d
a
nd
s
e
ns
e
e
m
be
ddi
ngs
a
r
e
pr
e
s
e
nt
e
d i
n s
e
c
ti
on 4.2.
T
a
bl
e
4. S
yns
e
t
di
c
ti
ona
r
y f
r
a
m
e
w
or
k
F
e
a
t
ur
e
s
G
l
oba
l
s
c
or
e
A
c
c
ur
a
c
y i
n %
S
0.0869
24
S
+G
0.1923
27
S
+G
+H
0.1666
33
S
+G
+H
+H
P
0.0588
38
S
+G
+H
+H
P
+H
G
0.3333
42
S
+G
+H
+H
P
+H
G
+H
O
0.0526
47
S
+G
+H
+H
P
+H
G
+H
O
+H
O
G
0.5238
50
4.2.
Wor
d
an
d
s
e
n
s
e
e
m
b
e
d
d
in
gs
W
e
e
va
lu
a
te
d
our
a
ppr
oa
c
h
f
or
te
s
ti
ng
th
e
s
ys
te
m
on
w
or
d
a
nd
s
e
ns
e
e
m
be
ddi
ng
s
s
e
p
a
r
a
te
ly
a
nd
th
e
n
c
om
bi
ni
ng
th
e
two
r
e
s
ul
ts
f
or
di
s
a
m
bi
gua
ti
on
pr
oc
e
s
s
.
W
or
d
e
m
be
ddi
ngs
a
r
e
ta
ke
n
f
r
om
th
e
r
a
w
c
or
pus
a
nd
m
a
ke
us
e
of
ge
ns
im
w
or
d2V
e
c
m
ode
l
f
or
our
s
tu
dy.
W
e
c
o
m
pa
r
e
d
our
w
or
k
w
it
h
ot
he
r
s
ta
te
of
th
e
a
r
t
m
e
th
ods
in
te
r
m
s
of
pr
e
c
is
io
n
a
nd
r
e
c
a
ll
r
e
pr
e
s
e
nt
e
d
in
T
a
bl
e
5.
I
t
is
obs
e
r
ve
d
th
a
t
our
a
ppr
oa
c
h
w
it
h
w
or
d
e
m
be
ddi
ngs
c
a
m
e
c
lo
s
e
to
ba
s
e
li
ne
a
c
c
ur
a
c
y
a
nd
uns
upe
r
vi
s
e
d
m
os
t
f
r
e
que
nt
s
e
ns
e
(
U
M
F
S
)
a
ppr
oa
c
h.
O
ur
a
pp
r
oa
c
h
gi
ve
s
a
f
e
a
s
ib
le
w
a
y
to
e
xt
r
a
c
t
pr
e
dom
in
a
nt
s
e
n
s
e
s
in
a
n
uns
up
e
r
vi
s
e
d
s
e
tu
p.
O
ur
a
ppr
oa
c
h
i
s
dom
a
in
in
de
pe
nde
nt
s
o
th
a
t
it
c
a
n
be
e
a
s
il
y
a
da
pt
e
d
to
a
dom
a
in
s
pe
c
if
ic
c
or
pus
.
T
o
ge
t
th
e
dom
a
in
s
pe
c
if
ic
w
or
d
a
nd
s
e
n
s
e
e
m
be
ddi
ng
s
,
w
e
s
im
pl
y
ha
ve
to
r
un
th
e
w
or
d
2ve
c
pr
ogr
a
m
on
th
e
dom
a
in
s
pe
c
if
ic
c
or
pu
s
.
A
ls
o,
our
a
ppr
oa
c
h
is
la
ngua
ge
in
de
pe
nde
nt
a
nd
por
ta
bl
e
a
c
r
o
s
s
m
obi
le
de
vi
c
e
s
a
s
s
m
a
r
t
phone
s
b
e
in
g
th
e
m
os
t
pr
e
f
e
r
r
e
d m
ode
of
c
om
m
uni
c
a
ti
on. C
onc
lu
s
io
n i
s
s
um
m
e
d
up i
n t
he
ne
xt
s
e
c
ti
on.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
W
ik
i
s
e
ns
e
bag c
r
e
at
io
n us
in
g m
ul
ti
li
ngual w
or
d
s
e
ns
e
di
s
am
bi
guat
io
n
(
Shr
e
y
a P
at
ank
ar
)
325
T
a
bl
e
5. P
e
r
f
or
m
a
nc
e
c
om
pa
r
is
on of
s
e
n
s
e
e
m
be
ddi
ng
s
w
it
h ot
he
r
m
e
th
ods
S
ys
t
e
m
P
r
e
c
i
s
i
on
R
e
c
a
l
l
M
os
t
f
r
e
que
nt
s
e
n
s
e
ba
s
e
l
i
ne
0.552
0.552
L
e
s
k a
l
gor
i
t
hm
0.097
0.053
A
da
pt
e
d l
e
s
k
0.240
0.234
U
M
F
S
(
B
hi
nga
r
di
ve
)
0.433
0.432
M
ul
t
i
l
i
ngua
l
W
S
D
w
i
t
h
0.489
0.489
w
or
d a
nd
s
e
ns
e
e
m
b
e
ddi
ngs
5.
C
O
N
C
L
U
S
I
O
N
I
n
th
is
r
e
s
e
a
r
c
h
w
or
k,
w
e
pr
e
s
e
nt
e
d
m
ul
ti
li
ngua
l
a
ppr
oa
c
h
to
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
a
nd
u
s
e
d
B
a
be
lNe
t
a
s
m
ul
ti
li
ngua
l
le
xi
c
on f
or
di
s
a
m
bi
gua
ti
on. M
ul
ti
li
ngua
l
w
or
d s
e
ns
e
di
s
a
m
bi
gua
ti
on e
xpl
oi
ts
gr
a
ph
-
ba
s
e
d
m
e
th
od
to
c
ol
le
c
t
e
vi
de
nc
e
s
f
r
om
tr
a
ns
la
ti
ons
in
va
r
i
ous
la
ngua
ge
s
.
W
e
a
l
s
o
e
xpl
or
e
d
th
e
s
yn
s
e
t
di
c
ti
ona
r
y
f
r
a
m
e
w
or
k
by
m
a
ki
ng
us
e
of
f
e
a
tu
r
e
s
f
r
om
B
a
b
e
lNe
t
di
c
ti
ona
r
y.
W
e
c
r
e
a
te
d
s
e
p
a
r
a
te
m
ode
l
f
or
e
a
c
h
a
m
bi
guous
w
or
d
s
e
ns
e
a
nd
m
a
d
e
a
n
e
n
s
e
m
bl
e
of
th
e
w
or
d2V
e
c
m
ode
ls
f
or
di
s
a
m
bi
gua
ti
on
pur
pos
e
us
in
g
w
or
d
e
m
be
ddi
ngs
.
O
ur
r
e
s
e
a
r
c
h
c
ont
r
ib
ut
io
n
in
c
lu
de
s
s
e
ns
e
b
a
g
c
r
e
a
ti
on
by
u
s
in
g
th
e
ont
ol
ogi
c
a
l
f
e
a
tu
r
e
s
of
th
e
B
a
be
lNe
t
le
xi
c
on
a
nd
e
n
c
yc
lo
pe
di
c
knowl
e
dge
f
r
om
W
ik
ip
e
di
a
.
I
t
is
obs
e
r
ve
d
th
a
t
m
ul
ti
li
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
a
c
hi
e
ve
d
good
r
e
s
ul
ts
in
c
om
pa
r
is
on
to
m
onol
in
gua
l
s
y
s
te
m
a
s
a
ddi
ti
ona
l
knowle
dge
f
r
om
va
r
io
us
la
ngu
a
ge
s
he
lp
to
boo
s
t
t
he
a
c
c
ur
a
c
y.
T
h
e
r
e
s
ul
ts
a
ls
o
s
how
th
a
t
our
m
e
th
od
of
m
ul
ti
li
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
ti
on
w
it
h
s
e
n
s
e
e
m
be
ddi
ng
im
pr
ove
s
th
e
a
c
c
ur
a
c
y
of
th
e
s
ys
te
m
.
T
he
a
ppr
oa
c
h
is
ope
n
to
e
xpl
or
e
ot
he
r
la
ngua
ge
s
.
W
e
w
il
l
e
xpl
or
e
our
a
ppr
oa
c
h
f
or
ot
he
r
pa
r
ts
o
f
s
pe
e
c
h
a
nd
ot
he
r
l
a
ngua
ge
s
e
s
pe
c
ia
ll
y
I
ndi
a
n
la
ngua
ge
s
li
ke
M
a
r
a
th
i,
H
in
di
,
a
nd
B
a
ngl
a
.
W
e
pl
a
n
in
th
e
ne
a
r
f
ut
ur
e
to
c
r
e
a
te
ge
ne
r
a
li
z
e
d
s
e
n
s
e
r
e
pr
e
s
e
nt
a
ti
on
f
or
m
ul
ti
pl
e
la
ngua
ge
s
s
o a
s
to
pr
ovi
de
a
ge
ne
r
a
l
f
r
a
m
e
w
or
k
f
or
knowle
dge
r
ic
h m
ul
ti
li
ngua
l
w
or
d s
e
ns
e
di
s
a
m
bi
gua
ti
o
n.
R
E
F
E
R
E
N
C
E
S
[
1]
R
. N
a
vi
gl
i
a
nd S
. P
. P
onz
e
t
t
o, “
J
oi
ni
ng f
or
c
e
s
pa
y
s
of
f
:
m
ul
t
i
l
i
ngua
l
j
oi
nt
w
or
d s
e
ns
e
di
s
a
m
bi
gua
t
i
on,”
i
n
P
r
oc
e
e
di
ngs
of
t
h
e
2012
J
oi
nt
C
onf
e
r
e
n
c
e
on
E
m
pi
r
i
c
al
M
e
t
hods
i
n
N
at
ur
al
L
anguage
P
r
o
c
e
s
s
i
ng
an
d
C
om
put
at
i
onal
N
at
ur
al
L
anguage
L
e
ar
ni
ng
,
J
ul
.
2012, pp. 1399
–
1410.
[
2]
D
.
O
,
S
.
K
w
on,
K
.
K
i
m
,
a
nd
Y
.
K
o,
“
W
or
d
s
e
n
s
e
di
s
a
m
bi
gua
t
i
on
ba
s
e
d
on
w
or
d
s
i
m
i
l
a
r
i
t
y
c
a
l
c
ul
a
t
i
on
us
i
ng
w
or
d
ve
c
t
or
r
e
pr
e
s
e
nt
a
t
i
on
f
r
om
a
know
l
e
dge
-
ba
s
e
d
gr
a
ph,”
i
n
P
r
oc
e
e
d
i
ng
s
of
t
he
2
7t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
C
o
m
put
at
i
onal
L
i
ngui
s
t
i
c
s
, A
ug. 2018, pp. 2704
–
2714.
[
3]
Y
.
W
a
ng,
M
.
W
a
ng,
a
nd
H
.
F
uj
i
t
a
,
“
W
or
d
S
e
ns
e
D
i
s
a
m
bi
gua
t
i
on:
A
c
o
m
pr
e
he
ns
i
ve
know
l
e
dge
e
xpl
oi
t
a
t
i
on
f
r
a
m
e
w
or
k,”
K
now
l
e
dge
-
B
as
e
d Sy
s
t
e
m
s
, vol
. 190, p. 105030, F
e
b. 2020, doi
:
10.1016/
j
.knos
ys
.2019.105030.
[
4]
T
.
M
i
kol
ov,
K
.
C
he
n,
G
.
C
or
r
a
do,
a
nd
J
.
D
e
a
n,
“
E
f
f
i
c
i
e
nt
e
s
t
i
m
a
t
i
on
of
w
or
d
r
e
pr
e
s
e
nt
a
t
i
ons
i
n
ve
c
t
or
s
pa
c
e
,”
1s
t
I
nt
e
r
nat
i
ona
l
C
onf
e
r
e
nc
e
on L
e
ar
ni
ng R
e
pr
e
s
e
nt
at
i
ons
, I
C
L
R
2013
-
W
or
k
s
hop T
r
ac
k
P
r
o
c
e
e
di
ngs
, J
a
n. 2013.
[
5]
S
.
B
hi
nga
r
di
ve
,
S
.
S
ha
i
kh,
a
nd
P
.
B
h
a
t
t
a
c
ha
r
yya
,
“
N
e
i
ghbor
s
he
l
p:
bi
l
i
ngua
l
u
ns
up
e
r
vi
s
e
d
W
S
D
u
s
i
ng
c
ont
e
xt
,”
2013,
vol
.
2,
pp
.
538
–
542.
[
6]
R
.
F
u,
J
.
G
uo,
B
.
Q
i
n,
W
.
C
he
,
H
.
W
a
ng,
a
nd
T
.
L
i
u,
“
L
e
a
r
ni
ng
s
e
m
a
nt
i
c
hi
e
r
a
r
c
hi
e
s
:
a
c
ont
i
nuous
ve
c
t
or
s
pa
c
e
a
ppr
oa
c
h,
”
I
E
E
E
T
r
ans
ac
t
i
ons
on
A
udi
o,
Spe
e
c
h
and
L
anguage
P
r
oc
e
s
s
i
ng
,
vol
.
23,
no.
3,
pp.
461
–
471,
M
a
r
.
2015,
doi
:
10.1109/
T
A
S
L
P
.2014.2377580.
[
7]
K
.
T
a
ghi
pour
a
nd
H
.
T
.
N
g,
“
S
e
m
i
-
s
upe
r
vi
s
e
d
w
or
d
s
e
n
s
e
di
s
a
m
bi
gua
t
i
on
us
i
ng
w
or
d
e
m
be
ddi
ngs
i
n
ge
n
e
r
a
l
a
nd
s
pe
c
i
f
i
c
dom
a
i
ns
,”
i
n
N
A
A
C
L
H
L
T
2015
-
2015
C
onf
e
r
e
n
c
e
of
t
h
e
N
o
r
t
h
A
m
e
r
i
c
a
n
C
hapt
e
r
of
t
he
A
s
s
oc
i
at
i
on
f
o
r
C
om
put
at
i
ona
l
L
i
ngui
s
t
i
c
s
:
H
um
an L
anguage
T
e
c
hnol
ogi
e
s
, P
r
oc
e
e
di
ngs
of
t
he
C
onf
e
r
e
nc
e
, 2
015, pp. 314
–
323, doi
:
10.3115/
v1/
n15
-
1035.
[
8]
X
.
C
he
n,
Z
.
L
i
u,
a
nd
M
.
S
un,
“
A
uni
f
i
e
d
m
ode
l
f
or
w
or
d
s
e
ns
e
r
e
pr
e
s
e
nt
a
t
i
on
a
nd
di
s
a
m
bi
gua
t
i
on,”
i
n
E
M
N
L
P
2014
-
2014
C
onf
e
r
e
nc
e
on
E
m
pi
r
i
c
al
M
e
t
hods
i
n
N
at
ur
al
L
anguage
P
r
oc
e
s
s
i
ng,
P
r
oc
e
e
di
ngs
of
t
he
C
onf
e
r
e
nc
e
,
2014,
pp.
1025
–
1035,
doi
:
10.3115/
v1/
d14
-
1110.
[
9]
I
.
I
a
c
oba
c
c
i
,
M
.
T
.
P
i
l
e
hva
r
,
a
nd
R
.
N
a
vi
gl
i
,
“
E
m
be
ddi
ngs
f
or
w
or
d
s
e
n
s
e
di
s
a
m
bi
gua
t
i
on:
a
n
e
va
l
ua
t
i
on
s
t
udy,”
i
n
54t
h
A
nnual
M
e
e
t
i
ng
of
t
he
A
s
s
oc
i
at
i
on
f
or
C
om
put
at
i
onal
L
i
ngui
s
t
i
c
s
,
A
C
L
2016
-
L
ong
P
ape
r
s
,
2016,
vol
.
2,
pp.
897
–
907,
doi
:
10.18653/
v1/
p16
-
1085.
[
10]
A
.
T
r
a
s
k,
P
.
M
i
c
ha
l
a
k,
a
nd
J
.
L
i
u,
“
s
e
ns
e
2ve
c
-
A
f
a
s
t
a
nd
a
c
c
ur
a
t
e
m
e
t
hod
f
or
w
o
r
d
s
e
ns
e
di
s
a
m
bi
gua
t
i
on
i
n
ne
ur
a
l
w
or
d
e
m
be
ddi
ngs
,”
2015.
[
11]
H
.
S
uga
w
a
r
a
,
H
.
T
a
ka
m
ur
a
,
R
.
S
a
s
a
no,
a
nd
M
.
O
kum
ur
a
,
“
C
ont
e
xt
r
e
pr
e
s
e
nt
a
t
i
on
w
i
t
h
w
or
d
e
m
be
ddi
ngs
f
or
W
S
D
,”
i
n
C
om
m
uni
c
at
i
ons
i
n C
om
put
e
r
and I
nf
or
m
at
i
on Sc
i
e
nc
e
, vol
. 593, S
pr
i
nge
r
S
i
nga
por
e
,
2016, pp. 108
–
119.
[
12]
L
. S
pe
c
i
a
, M
. S
t
e
ve
n
s
on, a
nd
M
. N
une
s
, “
L
e
a
r
ni
ng e
xpr
e
s
s
i
ve
m
ode
l
s
f
or
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
t
i
on,”
2007.
[
13]
S
. T
ha
t
e
r
, H
. F
ür
s
t
e
na
u, a
nd M
. P
i
nka
l
, “
W
or
d m
e
a
ni
ng
i
n c
ont
e
xt
:
a
s
i
m
pl
e
a
n
d e
f
f
e
c
t
i
ve
ve
c
t
or
m
ode
l
,”
A
F
N
L
P
. pp. 1134
–
11
43,
2011.
[
14]
S
. B
hi
nga
r
di
ve
, D
. S
i
ngh,
M
.
V
R
udr
a
,
H
.
R
e
dka
r
, a
nd
P
. B
h
a
t
t
a
c
ha
r
yya
,
“
U
ns
upe
r
vi
s
e
d m
os
t
f
r
e
que
nt
s
e
n
s
e
d
e
t
e
c
t
i
on u
s
i
ngw
or
d
e
m
be
ddi
ngs
,”
i
n
N
A
A
C
L
H
L
T
2015
-
2015
C
onf
e
r
e
nc
e
of
t
he
N
or
t
h
A
m
e
r
i
c
an
C
hapt
e
r
of
t
he
A
s
s
oc
i
at
i
on
f
or
C
om
put
at
i
onal
L
i
ngui
s
t
i
c
s
:
H
um
an L
anguage
T
e
c
hnol
ogi
e
s
, P
r
oc
e
e
di
ngs
of
t
he
C
onf
e
r
e
nc
e
, 2
015, pp. 1238
–
1243, doi
:
10.3115/
v1/
n15
-
1132.
[
15]
R
.
N
a
vi
gl
i
,
“
W
or
d
s
e
n
s
e
di
s
a
m
bi
gua
t
i
on:
a
s
ur
ve
y,”
A
C
M
C
om
put
i
ng
Sur
v
e
y
s
,
vol
.
41,
no.
2,
pp.
1
–
69,
F
e
b.
2009,
doi
:
1
0.1145/
1459352.1459355.
[
16]
R
. N
a
vi
gl
i
a
nd S
. P
onz
e
t
t
o, “
M
ul
t
i
l
i
ngua
l
W
S
D
w
i
t
h j
us
t
a
f
e
w
l
i
ne
s
of
c
ode
:
t
h
e
B
a
be
l
N
e
t
A
P
I
,”
2012, pp. 67
–
72.
[
17]
W
. A
z
i
z
a
nd L
. S
pe
c
i
a
, “
M
ul
t
i
l
i
ngua
l
W
S
D
-
l
i
ke
c
ons
t
r
a
i
nt
s
f
or
pa
r
a
phr
a
s
e
e
xt
r
a
c
t
i
on,”
2013, pp. 202
–
211.
[
18]
A
. M
ont
oyo, R
.
R
om
e
r
o, S
. V
á
z
que
z
,
C
. C
a
l
l
e
, a
nd S
.
S
ol
e
r
, “
T
he
r
ol
e
of
W
S
D
f
or
m
ul
t
i
l
i
ngua
l
na
t
ur
a
l
l
a
ngua
ge
a
ppl
i
c
a
t
i
ons
,”
i
n
L
e
c
t
ur
e
N
ot
e
s
i
n
C
om
put
e
r
Sc
i
e
nc
e
(
i
nc
l
udi
ng
s
ubs
e
r
i
e
s
L
e
c
t
ur
e
N
ot
e
s
i
n
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
and
L
e
c
t
ur
e
N
ot
e
s
i
n
B
i
oi
nf
or
m
a
t
i
c
s
)
, vol
. 2448, S
pr
i
nge
r
B
e
r
l
i
n H
e
i
de
l
be
r
g, 2002, pp. 41
–
48.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
11
, N
o.
1
,
M
a
r
c
h 20
22
:
319
-
326
326
[
19]
M
.
M
.
K
ha
pr
a
,
S
.
S
ha
h,
P
.
K
e
di
a
,
a
nd
P
.
B
ha
t
t
a
c
ha
r
yya
,
“
P
r
oj
e
c
t
i
ng
pa
r
a
m
e
t
e
r
s
f
or
m
ul
t
i
l
i
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
t
i
on,”
i
n
E
M
N
L
P
2009
-
P
r
oc
e
e
di
ngs
of
t
he
2009
C
onf
e
r
e
nc
e
on
E
m
pi
r
i
c
al
M
e
t
hods
i
n
N
at
ur
al
L
anguage
P
r
oc
e
s
s
i
ng:
A
M
e
e
t
i
ng
o
f
SI
G
D
A
T
,
a
Spe
c
i
al
I
nt
e
r
e
s
t
G
r
oup
of
A
C
L
,
H
e
l
d
i
n
C
onj
un
c
t
i
on
w
i
t
h
A
C
L
-
I
J
C
N
L
P
2009
,
2009,
pp.
459
–
467,
doi
:
10.3115/
1699510.1699570.
[
20]
P
. R
e
s
ni
k, “
W
S
D
i
n N
L
P
a
ppl
i
c
a
t
i
ons
,”
S
pr
i
nge
r
N
e
t
he
r
l
a
nds
, 2007, p
p. 299
–
3
37.
[
21]
C
.
S
i
l
be
r
e
r
a
nd
S
.
P
onz
e
t
t
o,
“
U
H
D
:
C
r
os
s
-
l
i
ngua
l
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
t
i
on
us
i
ng
m
ul
t
i
l
i
ngua
l
c
o
-
oc
c
ur
r
e
nc
e
gr
a
phs
,”
pp.
134
–
137, 2010.
[
22]
D
.
V
i
c
kr
e
y,
L
.
B
i
e
w
a
l
d,
M
.
T
e
ys
s
i
e
r
,
a
nd
D
.
K
ol
l
e
r
,
“
W
or
d
-
s
e
ns
e
di
s
a
m
bi
gua
t
i
on
f
or
m
a
c
hi
ne
t
r
a
ns
l
a
t
i
on,”
i
n
H
L
T
/
E
M
N
L
P
2005
-
H
um
an L
anguage
T
e
c
hnol
ogy
C
onf
e
r
e
nc
e
and C
onf
e
r
e
nc
e
on
E
m
pi
r
i
c
al
M
e
t
hods
i
n N
at
ur
al
L
anguage
P
r
oc
e
s
s
i
ng, P
r
oc
e
e
di
ng
s
of
t
he
C
onf
e
r
e
nc
e
, 2005, pp. 771
–
778, doi
:
10.3115/
1220575.1220672.
[
23]
R
.
M
a
r
vi
n
a
nd
P
.
K
oe
hn,
“
E
xpl
or
i
ng
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
t
i
on
a
bi
l
i
t
i
e
s
of
ne
ur
a
l
m
a
c
hi
ne
t
r
a
ns
l
a
t
i
on
s
ys
t
e
m
s
,”
A
M
T
A
,
vol
.
1,
pp. 125
–
131, M
a
r
. 2018, doi
:
10.2/
J
Q
U
E
R
Y
.M
I
N
.J
S
.
[
24]
T
.
A
.
R
a
s
hi
d,
A
.
M
.
M
us
t
a
f
a
,
a
nd
A
.
M
.
S
a
e
e
d,
“
A
R
obu
s
t
C
a
t
e
gor
i
z
a
t
i
on
S
ys
t
e
m
f
or
K
ur
di
s
h
S
or
a
ni
T
e
xt
D
o
c
um
e
nt
s
,
”
In
f
or
m
at
i
on T
e
c
hnol
ogy
J
our
nal
, vol
. 16, no. 1, pp. 27
–
34, D
e
c
. 2016, doi
:
10.
3923/
i
t
j
.2017.27.34.
[
25]
A
.
M
.
S
a
e
e
d,
T
.
A
.
R
a
s
hi
d,
A
.
M
.
M
u
s
t
a
f
a
,
R
.
A
.
A
.
-
R
.
A
gha
,
A
.
S
.
S
ha
m
s
a
l
di
n,
a
nd
N
.
K
.
A
l
-
S
a
l
i
hi
,
“
A
n
e
va
l
u
a
t
i
on
of
R
e
be
r
s
t
e
m
m
e
r
w
i
t
h
l
onge
s
t
m
a
t
c
h
s
t
e
m
m
e
r
t
e
c
hni
que
i
n
K
ur
di
s
h
S
or
a
ni
t
e
xt
c
l
a
s
s
i
f
i
c
a
t
i
on,”
I
r
an
J
our
nal
of
C
om
put
e
r
Sc
i
e
nc
e
,
vol
.
1,
no. 2, pp. 99
–
107, J
a
n. 2018, doi
:
10.1007/
s
42044
-
018
-
0007
-
4.
[
26]
A
.
M
.
M
u
s
t
a
f
a
a
nd
T
.
A
.
R
a
s
hi
d,
“
K
ur
di
s
h
s
t
e
m
m
e
r
pr
e
-
pr
oc
e
s
s
i
ng
s
t
e
ps
f
or
i
m
pr
ovi
ng
i
nf
o
r
m
a
t
i
on
r
e
t
r
i
e
va
l
,”
J
our
nal
of
I
nf
or
m
at
i
on Sc
i
e
nc
e
, vol
. 44, no. 1, pp. 15
–
27, J
a
n. 2018, doi
:
10.1177/
0165551516683617.
[
27]
T
. A
. R
a
s
hi
d, A
.
M
.
M
us
t
a
f
a
,
a
nd A
.
M
.
S
a
e
e
d,
“
A
ut
om
a
t
i
c
kur
di
s
h
t
e
xt
c
l
a
s
s
i
f
i
c
a
t
i
on us
i
ng K
D
C
4007
da
t
a
s
e
t
,”
i
n
L
e
c
t
ur
e
N
ot
e
s
on D
at
a
E
ngi
ne
e
r
i
ng and C
om
m
uni
c
at
i
ons
T
e
c
hnol
ogi
e
s
, vol
. 6, S
pr
i
nge
r
I
nt
e
r
na
t
i
ona
l
P
ubl
i
s
hi
ng, 2018, pp. 187
–
198.
[
28]
Y
.
C
hoi
,
J
.
W
i
e
be
,
a
nd
R
.
M
i
ha
l
c
e
a
,
“
C
oa
r
s
e
-
gr
a
i
ne
d
+/
-
e
f
f
e
c
t
w
or
d
s
e
ns
e
di
s
a
m
bi
gua
t
i
on
f
or
i
m
pl
i
c
i
t
s
e
nt
i
m
e
nt
a
na
l
ys
i
s
,
”
I
E
E
E
T
r
ans
ac
t
i
ons
on A
f
f
e
c
t
i
v
e
C
o
m
put
i
ng
, vol
. 8, no. 4, pp. 471
–
479, O
c
t
. 2017, doi
:
10.1109/
T
A
F
F
C
.2017.2734085.
[
29]
R
.
R
.
K
a
r
w
a
a
nd
M
.
B
.
C
ha
nda
k,
“
W
or
d
s
e
n
s
e
di
s
a
m
bi
gua
t
i
on:
hybr
i
d
a
ppr
oa
c
h
w
i
t
h
a
nnot
a
t
i
on
up
t
o
c
e
r
t
a
i
n
l
e
ve
l
–
a
r
e
vi
e
w
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
E
ngi
ne
e
r
i
ng
T
r
e
nds
and
T
e
c
hnol
ogy
,
vol
.
18,
no.
7,
pp.
328
–
330,
D
e
c
.
2014,
doi
:
10.14445/
22315381/
i
j
e
t
t
-
v18p267.
[
30]
E
.
A
.
C
or
r
ê
a
,
A
.
A
.
L
ope
s
,
a
nd
D
.
R
.
A
m
a
nc
i
o,
“
W
or
d
s
e
n
s
e
di
s
a
m
bi
gua
t
i
on:
a
c
om
pl
e
x
ne
t
w
or
k
a
ppr
oa
c
h,”
I
nf
or
m
at
i
on
Sc
i
e
nc
e
s
, vol
. 442
–
443, pp. 103
–
113,
M
a
y 2018, doi
:
10.1016/
j
.i
ns
.2018.02.04
7.
[
31]
A
. H
. A
l
i
w
y a
nd H
. A
. T
a
he
r
, “
W
or
d s
e
n
s
e
di
s
a
m
bi
gua
t
i
on:
s
ur
ve
y s
t
udy,”
J
ou
r
nal
of
C
om
put
e
r
S
c
i
e
nc
e
, vol
. 15, no. 7, pp. 1004
–
1011, J
ul
. 2019, doi
:
10.3844/
j
c
s
s
p.2019.1004.1011.
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Shreya
Nandkumar
Patankar
is
pursuing
Ph.D
.
from
Mumbai
University
in
the
area
of
Natural
language
processing
and
is
currently
working
as
Assistant
Professor
at
Datta
Meghe
College
of
Engineering,
Airoli,
Navi
Mumbai,
Indi
a
with
19
years
of
teaching
experie
nce.
Her
resear
ch
is
mainly
focused
on
wor
d
sense
disambiguation,
Natural
language
processing,
Artificial
Intelligence
and
machine
lear
ning.
She
is
good
in
programming
languages
like
java
and
subjects
such
as
algorithms,
ope
rating
systems,
data
structures
and
database.
She
has
published
10
papers
in
internati
onal
conference,
7
internationa
l
journals,
4
in
national
confere
nces
and
few
pap
ers
are
indexed
in
Scopus
database
and
Elsevier.
One
of
her
paper
featured
as
top
10
downloa
ded
papers
in
SSRN
digital
library.
One
Masters
candidate
is
actively
in
volved
under
h
er
g
uidance.
She
can
be
contacted
at email
:
shreya.pata
nkar@
dmce.ac.in
.
Madhu
ra
Mandar
Phadk
e
is
pursuing
Ph.D
.
from
Mumbai
University
in
the
area
of
Natural
language
processing
and
is
currently
working
as
Assistant
Professor
at
Datta
Meghe
College
of
Engineering,
Airoli,
Navi
Mumbai,
India.
She
has
21
years
of
teaching
experie
nce.
She
is
good
in
various
subjects
such
as
big
data
analysis,
cryptograp
hy
and
system
securi
ty,
machine
learning,
security
and
dat
abase.
Her
research
is
mainly
focused
on
machine
translation
using
machine
learning.
S
he
has
published
11
papers
in
international
conference,
4
international
journals,
13
in
n
ational
conferences.
Her
work
was
appreciat
ed
at
one
of
the
NLP
conference.
One
Ma
sters
candidate
has
successfully
completed
her
work
under
her
guidance.
She
can
be
contacted
at
email:
madhura.ph
adke@dmce.ac.in
.
Dr.
Satish.
R.
Devane
is
an
Academician
and
completed
his
Ph.
D
.
degree
from
Indian
Institute
of
Technology
(IIT.
He
is
currently
worki
ng
as
Principal
at
Karmaveer
Baburao
Ganpatrao
Thakare
College
Of
Engineering,
Na
shik.
NaviMumbai.
He
is
having
34
years
of
teaching
experience
,
one
year
Industry
and
4
years
of
Research
experience
and
is
proficient
in
many
technical
areas
such
as
E
-
com
merce,
networking,
Artificial
Intelligence,
Data
Mining
etc.
His
research
area
includes
security,
computer
networks,
natural
language
processing.
He
has
published
various
research
papers
in
internationa
l
confere
nces
and
Journals
out
of
which
few
pap
ers
are
indexed
in
Scopus
database.
Four
PhD
awarded,
and
various
candidates
received
their
Masters
degree
under
his guidance.
H
e can be con
tacted at
ema
il:
satish@
dmce.ac.in.
Evaluation Warning : The document was created with Spire.PDF for Python.