I
AE
S
I
nte
rna
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
5
,
No
.
2
,
J
u
n
e
2016
,
p
p
.
72
~
79
I
SS
N:
2252
-
8938
72
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
J
A
I
Rule Ba
sed
a
nd
Ex
pectation
Ma
x
im
i
z
a
tion
alg
o
rithm
for
Ara
bic
-
Eng
lish
H
y
brid Ma
chine Tr
a
nsla
tion
Arw
a
Alqu
ds
i
*
,
Na
zlia
O
m
a
r
*
,
Ra
bh
a
W.
I
bra
hi
m
*
*
*
Kn
o
w
led
g
e
T
e
c
h
n
o
lo
g
y
Re
s
e
a
r
c
h
G
ro
u
p
(KT
),
Ce
n
ter f
o
r
A
I
T
e
c
h
n
o
lo
g
y
(C
A
I
T
),
F
a
c
u
lt
y
o
f
In
f
o
r
m
a
ti
o
n
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
lo
g
y
,
Un
iv
e
rsiti
Ke
b
a
n
g
sa
a
n
M
a
lay
sia
*
*
In
stit
u
te o
f
M
a
th
e
m
a
ti
c
a
l
S
c
ie
n
c
e
s,
Un
iv
e
rsity
o
f
M
a
la
y
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
r
5
,
2
0
1
6
R
ev
i
s
ed
Ma
y
9
,
2
0
1
6
A
cc
ep
ted
Ma
y
2
6
,
2
0
1
6
It
is
p
ra
c
ti
c
a
ll
y
i
m
p
o
ss
ib
le
f
o
r
p
u
re
m
a
c
h
in
e
tran
sla
ti
o
n
a
p
p
r
o
a
c
h
to
p
ro
c
e
ss
a
ll
o
f
tran
sla
ti
o
n
p
ro
b
lem
s;
h
o
we
v
e
r,
Ru
le
Ba
se
d
M
a
c
h
in
e
T
ra
n
s
latio
n
a
n
d
S
tatisti
c
a
l
M
a
c
h
in
e
tran
sla
t
io
n
(R
BM
T
a
n
d
S
M
T
)
u
se
d
if
f
e
re
n
t
a
rc
h
it
e
c
tu
re
s
f
o
r
p
e
rf
o
r
m
in
g
tran
sla
ti
o
n
tas
k
.
L
e
x
ica
l
a
n
a
l
y
s
e
r
a
n
d
sy
n
tac
ti
c
a
n
a
ly
s
e
r
a
r
e
so
lv
e
d
b
y
Ru
le
Ba
s
e
d
a
n
d
so
m
e
a
m
o
u
n
t
o
f
a
m
b
ig
u
it
y
is
le
f
t
to
b
e
so
lv
e
d
b
y
Ex
p
e
c
tatio
n
–
M
a
x
im
i
z
a
ti
o
n
(EM
)
a
lg
o
rit
h
m
,
w
h
ich
is
a
n
i
tera
ti
v
e
sta
ti
stic
a
lg
o
rit
h
m
f
o
r
f
in
d
in
g
m
a
x
i
m
u
m
li
k
e
li
h
o
o
d
.
I
n
t
h
is
p
a
p
e
r
w
e
h
a
v
e
p
ro
p
o
se
d
a
n
in
teg
ra
ted
Hy
b
rid
M
a
c
h
i
n
e
T
ra
n
sla
ti
o
n
(HMT
)
sy
st
e
m
.
T
h
e
g
o
a
l
is
to
c
o
m
b
in
e
th
e
b
e
st
p
ro
p
e
rti
e
s
o
f
e
a
c
h
a
p
p
ro
a
c
h
.
I
n
it
ially
,
A
r
a
b
ic
tex
t
is
k
e
y
e
d
in
to
RBM
T
;
th
e
n
t
h
e
o
u
t
p
u
t
w
il
l
b
e
e
d
it
e
d
b
y
EM
a
l
g
o
rit
h
m
to
g
e
n
e
ra
te
th
e
f
in
a
l
tran
sla
ti
o
n
o
f
En
g
li
sh
tex
t.
As
w
e
h
a
v
e
se
e
n
in
p
re
v
io
u
s
w
o
rk
s,
th
e
p
e
rf
o
r
m
a
n
c
e
a
n
d
e
n
h
a
n
c
e
m
e
n
t
o
f
EM
a
lg
o
rit
h
m
,
th
e
k
e
y
o
f
EM
a
lg
o
rit
h
m
p
e
rf
o
r
m
a
n
c
e
is
th
e
a
b
il
it
y
to
a
c
c
u
ra
tel
y
tran
s
f
o
r
m
a
f
re
q
u
e
n
c
y
f
ro
m
o
n
e
lan
g
u
a
g
e
to
a
n
o
th
e
r.
Re
su
lt
s
sh
o
w
in
g
th
a
t,
a
s
p
r
o
v
e
d
b
y
BL
EU
s
y
s
te
m
,
th
e
p
ro
p
o
se
d
m
e
th
o
d
c
a
n
su
b
sta
n
ti
a
ll
y
o
u
tp
e
rf
o
r
m
sta
n
d
a
rd
Ru
le
Ba
se
d
a
p
p
ro
a
c
h
a
n
d
E
M
a
lg
o
rit
h
m
in
term
s
o
f
fre
q
u
e
n
c
y
a
n
d
a
c
c
u
ra
c
y
.
T
h
e
re
s
u
lt
s
o
f
th
is
stu
d
y
h
a
v
e
b
e
e
n
sh
o
w
e
d
t
h
a
t
t
h
e
sc
o
re
o
f
HMT
sy
st
e
m
is
h
ig
h
e
r
th
a
n
S
M
T
s
y
ste
m
in
a
ll
c
a
s
e
s.
W
h
e
n
c
o
m
b
in
in
g
tw
o
a
p
p
r
o
a
c
h
e
s,
HMT
o
u
t
p
e
rf
o
r
m
e
d
S
M
T
in
Bleu
sc
o
re
.
K
ey
w
o
r
d
:
A
r
ab
ic
-
E
n
g
lis
h
Ma
c
h
i
n
e
T
r
an
s
latio
n
E
x
p
ec
tatio
n
–
Ma
x
i
m
izatio
n
(
E
M)
A
lg
o
r
it
h
m
H
y
b
r
id
Ma
ch
i
n
e
T
r
an
s
latio
n
Ma
ch
i
n
e
T
r
an
s
latio
n
Co
p
y
rig
h
t
©
2
0
1
6
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
A
r
w
a
A
lq
u
d
s
i
,
Kn
o
w
led
g
e
T
ec
h
n
o
lo
g
y
R
esea
r
ch
Gr
o
u
p
(
KT
)
,
C
en
ter
f
o
r
A
I
T
ec
h
n
o
lo
g
y
(
C
A
I
T
)
,
Facu
lt
y
o
f
I
n
f
o
r
m
a
tio
n
Scie
n
c
e
an
d
T
ec
h
n
o
lo
g
y
,
Un
i
v
er
s
iti Ke
b
a
n
g
s
aa
n
Ma
la
y
s
ia,
4
3
6
0
0
UKM
B
an
g
i,
Sela
n
g
o
r
,
Ma
la
y
s
ia.
E
m
ail:
ar
w
a.
alq
u
d
s
i8
1
@
g
m
ail
.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
er
e
ar
e
m
a
n
y
lan
g
u
ag
e
s
i
n
th
e
w
o
r
ld
an
d
it
is
d
i
f
f
icu
l
t
to
f
i
n
d
a
s
u
itab
le
m
ac
h
in
e
tr
an
s
lato
r
th
at
m
ee
t
s
h
u
m
an
r
eq
u
ir
e
m
e
n
t
s
to
tr
an
s
late
th
e
s
e
la
n
g
u
a
g
es.
I
n
t
h
e
s
a
m
e
t
i
m
e
tr
an
s
latio
n
p
r
o
ce
s
s
f
ac
es
a
n
u
m
b
er
o
f
p
r
o
b
lem
s
,
i
n
clu
d
i
n
g
lin
g
u
is
tic
p
r
o
b
lem
s
,
esp
ec
iall
y
in
ter
m
s
o
f
A
r
ab
ic
to
E
n
g
l
is
h
tr
an
s
l
atio
n
,
s
u
c
h
as
w
o
r
d
co
n
s
tr
u
ct
io
n
,
d
i
f
f
er
e
n
ce
s
in
g
r
a
m
m
ar
an
d
a
m
b
ig
u
o
u
s
w
o
r
d
s
.
Ma
n
y
d
if
f
er
e
n
t
ap
p
r
o
ac
h
es
h
av
e
b
ee
n
ap
p
lied
to
s
o
lv
e
A
r
ab
ic
m
ac
h
i
n
e
tr
an
s
la
tio
n
p
r
o
b
le
m
s
a
n
d
it
s
e
v
alu
a
t
io
n
[
4
]
[
5
]
.
T
h
ese
ap
p
r
o
ac
h
es
ar
e
d
is
c
u
s
s
ed
in
d
etail
b
y
[
1
]
.
T
h
e
R
u
le
B
ased
m
ac
h
i
n
e
tr
an
s
lat
io
n
s
y
s
te
m
i
s
o
n
e
o
f
t
h
ese
ap
p
r
o
ac
h
es,
d
en
o
tes
li
n
g
u
is
t
ic
r
u
le
b
et
w
ee
n
s
o
u
r
ce
la
n
g
u
a
g
e
a
n
d
tar
g
et
lan
g
u
ag
e
.
R
u
le
b
ased
h
as
co
n
tr
o
l
o
n
ex
p
lic
it
li
n
g
u
i
s
ti
c
k
n
o
w
led
g
e
th
a
t
it
ca
n
to
tall
y
an
a
l
y
ze
i
n
b
o
th
s
em
an
tic
an
d
s
y
n
ta
x
lev
el
s
.
T
o
cr
ea
te
th
e
lin
g
u
i
s
tic
r
u
le
t
h
e
R
u
le
B
ased
r
eq
u
ir
es
m
u
c
h
li
n
g
u
i
s
tic
k
n
o
w
led
g
e
s
o
th
at
it
n
ee
d
s
h
i
g
h
co
s
t
d
e
v
elo
p
m
e
n
t.
T
h
e
r
es
u
lt
o
f
R
u
le
B
ased
s
y
s
te
m
is
d
ep
en
d
s
o
n
t
h
e
ac
cu
r
ac
y
o
f
e
ac
h
le
v
el
a
n
d
th
e
o
u
tp
u
t
is
le
s
s
f
l
u
en
c
y
t
h
a
n
s
ta
tis
tica
l
s
y
s
t
e
m
.
T
h
e
y
ar
e
m
a
n
y
r
esear
ch
er
s
u
s
ed
r
u
le
b
a
s
ed
a
p
p
r
o
ac
h
as
t
h
eir
tr
a
n
s
latio
n
m
et
h
o
d
,
s
u
c
h
a
s
:
MT
f
o
r
R
o
m
an
ce
L
an
g
u
ag
e
s
to
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
R
u
le
B
a
s
ed
a
n
d
E
x
p
ec
ta
tio
n
Ma
ximiza
tio
n
a
lg
o
r
ith
fo
r
A
r
a
b
ic
-
E
n
g
lis
h
Hyb
r
id
.
.
.
.
(
A
r
w
a
A
lq
u
d
s
i
)
73
Sp
ain
[
1
9
]
,
A
r
ab
ic
to
E
n
g
lis
h
[
2
]
an
d
[
3
]
,
I
n
d
o
n
esian
to
Ma
la
y
s
ia
n
[
2
1
]
,
B
u
lg
ar
ian
to
Ma
ce
d
o
n
ian
[
2
2
]
,
E
n
g
l
is
h
to
San
k
r
it [
2
3
]
,
etc.
I
n
co
n
tr
ast,
th
e
Stati
s
tical
M
T
is
an
o
th
er
v
er
y
co
m
m
o
n
tr
an
s
lat
io
n
s
y
s
te
m
,
w
h
er
e
p
r
o
b
ab
ilit
y
r
u
le
s
ar
e
u
s
ed
,
s
u
ch
a
s
,
B
a
ye
's
r
u
le
o
f
th
e
E
x
p
ec
tatio
n
Ma
x
i
m
izatio
n
A
l
g
o
r
ith
m
(
E
M)
.
Statis
tic
Ma
c
h
in
e
T
r
an
s
latio
n
s
y
s
te
m
is
co
n
s
tr
u
cted
b
ased
o
n
p
ar
allel
co
r
p
o
r
a.
I
t
p
er
f
o
r
m
s
tr
ai
n
i
n
g
p
r
o
ce
s
s
o
n
t
h
e
p
ar
allel
co
r
p
o
r
a
to
lear
n
im
p
lied
k
n
o
wled
g
e
ex
is
tin
g
i
n
co
-
o
cc
u
r
r
en
c
e
s
tatis
tic.
T
r
an
s
latio
n
f
o
r
ce
r
tain
w
o
r
d
o
f
s
o
u
r
ce
lan
g
u
a
g
e
i
n
t
h
is
s
y
s
te
m
w
ill
b
e
f
o
u
n
d
b
y
lo
o
k
in
g
t
h
e
wo
r
d
in
tar
g
et
la
n
g
u
a
g
e
w
h
ic
h
f
r
eq
u
e
n
tl
y
o
cc
u
r
s
to
g
eth
er
w
i
th
th
e
m
in
p
ar
all
el
co
r
p
o
r
a.
Statis
tic
Ma
ch
i
n
e
T
r
an
s
latio
n
s
y
s
te
m
is
ab
le
to
p
r
o
d
u
ce
g
o
o
d
tr
an
s
latio
n
s
i
f
t
h
e
s
o
u
r
ce
s
e
n
te
n
ce
is
n
o
t si
m
ilar
to
an
y
s
en
te
n
ce
s
i
n
a
tr
ain
i
n
g
co
r
p
u
s
[
6
]
.
Ho
w
e
v
er
,
it
w
ill
h
a
v
e
p
r
o
b
lem
s
i
n
d
ea
lin
g
w
it
h
s
tr
u
ct
u
r
es
an
d
v
o
ca
b
u
lar
y
t
h
at
d
id
n
o
t
o
cc
u
r
in
t
h
e
tr
ain
i
n
g
d
ata.
B
y
u
s
i
n
g
a
co
m
b
in
ed
p
h
r
ase
tab
les
f
r
o
m
t
h
e
R
B
MT
s
y
s
te
m
s
a
s
w
ell
a
s
th
e
SMT
p
ar
allel
co
r
p
u
s
tab
le
,
th
e
h
y
b
r
id
s
y
s
te
m
ca
n
h
an
d
le
a
w
id
er
r
an
g
e
o
f
ex
p
lo
it
k
n
o
w
led
g
e
an
d
s
y
n
tactic
co
n
s
tr
u
ctio
n
s
th
at
t
h
e
R
B
MT
s
y
s
te
m
h
as
ab
o
u
t
th
e
s
p
ec
if
ic
v
o
ca
b
u
lar
y
o
f
t
h
e
s
o
u
r
ce
tex
t
[
2
5
]
.
L
i
n
g
u
i
s
ti
c
d
ata
f
r
o
m
R
B
MT
s
y
s
te
m
s
h
av
e
alr
ea
d
y
b
ee
n
u
s
ed
to
en
r
i
ch
SMT
s
y
s
te
m
s
[
2
6
-
2
9
]
.
Ma
n
y
r
esear
c
h
es
h
a
v
e
atte
m
p
ted
to
im
p
r
o
v
e
th
i
s
Statis
ti
ca
l
ap
p
r
o
ac
h
,
s
o
m
e
u
s
ed
wo
r
d
s
en
s
e
d
is
a
m
b
i
g
u
at
io
n
[
8
]
,
an
d
s
o
m
e
u
s
ed
w
o
r
d
ca
teg
o
r
izatio
n
an
d
g
r
a
m
m
atica
l
ca
teg
o
r
ies
to
h
an
d
le
th
e
er
r
o
r
[
7
]
.
[
9
]
Used
tw
o
d
ec
o
d
in
g
al
g
o
r
i
th
m
s
to
s
ea
r
ch
f
o
r
th
e
m
o
s
t
p
r
o
b
ab
le
tr
an
s
latio
n
o
f
an
i
n
p
u
t
tr
ee
.
E
x
p
ec
tatio
n
Ma
x
i
m
izatio
n
(
E
M)
alg
o
r
ith
m
w
as
u
s
ed
to
o
b
tain
t
h
e
p
r
o
b
ab
ilit
ies
f
r
o
m
a
T
r
ee
b
an
k
.
[
1
0
]
I
m
p
r
o
v
ed
th
e
n
u
m
b
er
o
f
r
elev
an
t
d
o
cu
m
e
n
t
s
r
etr
iev
ed
b
y
u
s
i
n
g
E
M
alg
o
r
ith
m
a
n
d
s
tu
d
ied
th
e
b
est
E
M
d
is
tan
ce
f
o
r
th
e
w
o
r
d
s
in
A
r
ab
ic
la
n
g
u
a
g
e
th
at
d
escr
ib
es th
e
s
i
m
ilar
it
y
b
et
w
e
en
th
e
s
e
w
o
r
d
s
.
I
f
w
e
co
u
ld
co
m
b
in
e
t
h
e
ad
v
a
n
tag
e
s
o
f
t
h
ese
ap
p
r
o
ac
h
es,
th
e
r
esu
lti
n
g
h
y
b
r
id
s
y
s
te
m
co
u
l
d
p
er
f
o
r
m
b
etter
tr
a
n
s
latio
n
t
h
a
n
an
y
o
th
er
s
y
s
te
m
[
1
1
]
.
T
h
er
e
ar
e
a
m
an
y
d
i
f
f
er
e
n
t
t
y
p
e
s
o
f
H
y
b
r
id
(
HM
T
)
,
s
in
ce
th
e
co
r
e
o
f
th
is
v
er
s
io
n
o
f
Ma
ch
i
n
e
T
r
an
s
latio
n
f
o
c
u
s
es
u
p
o
n
a
m
ix
of
ea
c
h
o
th
er
.
Sti
ll
a
n
u
m
b
er
o
f
r
esear
c
h
es
h
av
e
be
en
d
o
n
e
w
i
th
i
n
HM
T
.
T
h
e
g
o
al
o
f
H
y
b
r
id
ap
p
r
o
ac
h
is
ef
f
ec
ti
v
el
y
to
o
b
tain
m
o
r
e
ac
cu
r
ate
r
esu
lt
s
t
h
an
o
th
er
ex
is
ti
n
g
ap
p
r
o
ac
h
es
.
T
h
e
m
o
ti
v
atio
n
i
s
e
m
p
lo
y
i
n
g
d
if
f
er
e
n
t
m
ac
h
i
n
e
tr
an
s
latio
n
p
ar
ad
ig
m
s
,
w
h
ic
h
i
m
p
lies
t
h
at
a
s
m
ar
t
co
m
b
i
n
at
io
n
o
f
th
eir
o
u
tp
u
t
w
o
u
ld
r
etu
r
n
an
o
v
er
all
g
o
o
d
tr
an
s
latio
n
[
1
6
]
.
T
r
an
s
latio
n
p
r
o
ce
s
s
o
f
HM
T
is
co
m
p
lete
d
b
y
co
u
p
lin
g
t
w
o
o
r
m
o
r
e
s
y
s
te
m
s
th
a
t
ar
e
e
m
p
lo
y
ed
to
s
o
lv
e
p
r
o
b
le
m
s
w
it
h
ce
r
tain
p
ar
ts
.
A
cc
o
r
d
in
g
to
p
r
esen
t
r
eq
u
ir
e
m
e
n
ts
,
t
h
e
m
o
s
t
p
o
p
u
lar
co
m
b
i
n
atio
n
s
co
m
p
r
i
s
e
R
B
MT
v
s
.
SMT
[
2
9
]
.
[
1
2
]
n
o
r
m
al
ized
th
e
d
ialec
tal
w
o
r
d
s
in
a
h
y
b
r
id
m
ac
h
i
n
e
tr
an
s
latio
n
(
r
u
le
b
ased
an
d
s
tatis
t
ical)
s
y
s
te
m
,
b
y
p
er
f
o
r
m
in
g
a
co
m
b
in
atio
n
o
f
m
o
r
p
h
e
m
e
-
le
v
el
m
a
p
p
in
g
s
a
n
d
c
h
ar
ac
ter
.
T
h
ey
tr
an
s
lated
t
h
e
A
r
ab
i
c
to
E
n
g
li
s
h
u
s
i
n
g
a
h
y
b
r
id
MT
.
I
n
ter
m
s
o
f
B
L
E
U
s
y
s
te
m
b
y
m
ea
s
u
r
i
n
g
an
d
co
m
p
ar
in
g
t
h
e
r
esu
l
ts
t
h
e
au
t
h
o
r
p
r
o
v
ed
th
e
f
ea
s
ib
ilit
y
o
f
t
h
e
H
MT
a
p
p
r
o
ac
h
.
On
th
e
o
th
er
h
a
n
d
th
e
ad
v
a
n
ta
g
es
o
f
t
h
e
r
u
le
b
ased
an
d
ex
a
m
p
le
b
ased
ap
p
r
o
ac
h
es
u
s
ed
to
s
u
g
g
est
a
f
o
r
m
o
f
J
ap
an
ese
-
to
-
E
n
g
l
is
h
m
ac
h
i
n
e
tr
a
n
s
lat
io
n
t
h
at
ca
n
b
e
u
s
ed
w
it
h
ex
is
t
in
g
tech
n
o
lo
g
y
[
1
3
]
.
T
w
o
w
a
y
s
p
r
ese
n
ted
b
y
[
1
4
]
t
o
co
m
b
i
n
e
r
u
le
-
b
ased
an
d
s
tati
s
tical
ap
p
r
o
ac
h
es
to
E
n
g
l
is
h
-
Ger
m
a
n
m
ac
h
i
n
e
tr
an
s
latio
n
b
y
i
n
te
g
r
ati
n
g
e
x
is
tin
g
i
m
p
le
m
en
ta
ti
o
n
s
in
to
a
lar
g
er
ar
ch
itect
u
r
e.
I
n
th
is
p
ap
er
w
e
h
a
v
e
d
escr
ib
e
d
th
e
p
r
o
ce
s
s
o
f
d
e
v
elo
p
in
g
A
r
ab
ic
-
En
g
lis
h
H
MT
s
y
s
te
m
as
a
w
a
y
to
i
m
p
r
o
v
e
t
h
e
p
er
f
o
r
m
an
ce
o
f
m
ac
h
in
e
tr
an
s
latio
n
.
W
e
h
a
v
e
co
m
b
i
n
e
d
R
B
MT
w
it
h
S
MT
u
s
in
g
U
n
ited
Natio
n
s
p
ar
allel
co
r
p
u
s
.
T
h
e
r
em
ai
n
i
n
g
o
f
t
h
i
s
p
ap
er
is
s
tr
u
ctu
r
ed
a
s
f
o
llo
w
s
:
Sectio
n
2
w
ill
d
esc
r
ib
e
th
e
ar
ch
itect
u
r
e
o
f
A
r
ab
ic
-
E
n
g
li
s
h
h
y
b
r
id
m
ac
h
in
e
tr
an
s
latio
n
s
y
s
te
m
;
Secti
o
n
3
w
ill
d
escr
ib
e
t
h
e
i
m
p
le
m
en
tatio
n
o
f
HM
T
;
Sectio
n
4
w
i
ll
p
r
ese
n
t
t
h
e
e
x
p
er
i
m
en
t
r
es
u
lt
to
g
et
h
er
w
it
h
it
s
an
al
y
s
is
;
an
d
Sectio
n
5
w
ill
g
i
v
e
a
co
n
clu
s
io
n
ab
o
u
t th
is
r
esear
c
h
.
2.
ARAB
I
C
-
E
N
G
L
I
SH
H
YB
R
I
D
M
ACH
I
NE
T
RAN
SL
AT
I
O
N
SYS
T
E
M
An
i
m
p
o
r
tan
t
tr
en
d
o
v
er
t
h
e
last
y
ea
r
s
lie
s
i
n
a
f
o
c
u
s
s
h
i
f
t
to
w
ar
d
s
h
y
b
r
id
m
ac
h
i
n
e
tr
an
s
latio
n
s
y
s
te
m
s
.
T
h
e
ai
m
o
f
th
e
s
e
s
y
s
te
m
s
is
co
m
b
in
i
n
g
o
f
r
eso
u
r
ce
s
an
d
tech
n
iq
u
es
f
r
o
m
d
i
f
f
er
en
t
tech
n
o
lo
g
ica
l
b
ac
k
g
r
o
u
n
d
s
,
e.
g
.
,
r
u
le
b
ased
an
d
s
ta
tis
tical
ap
p
r
o
ac
h
es
[
2
4
]
.
A
r
ab
ic
-
E
n
g
lis
h
HM
T
s
y
s
te
m
i
s
p
r
esen
ted
i
n
t
h
i
s
p
ap
er
.
T
h
is
s
y
s
te
m
co
n
s
is
t
s
o
f
t
w
o
m
ai
n
co
m
p
o
n
e
n
t
s
.
R
u
le
b
ased
co
m
p
o
n
en
t
an
d
Sta
tis
tica
l
co
m
p
o
n
en
t.
R
B
MT
P
a
r
s
er
m
ap
s
t
h
e
A
r
ab
ic
r
u
les
i
n
to
E
n
g
l
is
h
r
u
les
a
n
d
SMT
tech
n
i
q
u
es h
a
n
d
le
th
e
la
n
g
u
a
g
e
a
m
b
ig
u
it
y
u
s
in
g
co
r
p
u
s
.
2
.
1
.
RB
M
T
Co
m
po
ne
nt
R
u
le
B
ased
h
as
it
s
o
r
ig
i
n
in
tr
an
s
f
er
s
y
s
te
m
m
ac
h
i
n
e
tr
an
s
la
tio
n
w
h
er
e
it
in
li
k
e
n
es
s
u
s
e
s
r
u
les
[
1
5
]
,
an
d
is
b
ased
o
n
lin
g
u
i
s
tic
i
n
f
o
r
m
atio
n
ab
o
u
t
t
h
e
s
o
u
r
ce
an
d
tar
g
et
lan
g
u
a
g
es
m
o
s
tl
y
e
x
tr
ac
ted
f
r
o
m
di
ctio
n
ar
ies
.
R
u
le
b
ased
m
ac
h
in
e
tr
a
n
s
la
tio
n
s
y
s
te
m
is
a
k
n
o
w
i
n
g
s
y
s
te
m
,
b
ec
au
s
e
it
is
b
ased
o
n
tr
an
s
latio
n
r
u
les
r
at
h
er
th
a
n
a
d
ictio
n
ar
y
.
W
h
en
t
h
e
s
tr
u
ct
u
r
e
o
f
t
h
e
s
o
u
r
ce
s
e
n
ten
ce
m
atc
h
e
s
o
n
e
o
f
th
e
r
u
les,
i
t
is
tr
an
s
lated
d
ir
ec
tl
y
u
s
i
n
g
a
d
i
ctio
n
ar
y
.
I
t
g
o
es
f
r
o
m
th
e
s
o
u
r
ce
s
e
n
te
n
ce
to
a
m
o
r
p
h
o
lo
g
ical
a
n
al
y
s
i
s
a
n
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
5
,
No
.
2
,
J
u
n
e
2
0
1
6
:
72
–
79
74
s
y
n
tactic
a
n
al
y
s
i
s
to
p
r
o
d
u
ce
a
n
e
w
s
o
r
t
o
f
s
en
te
n
ce
s
tr
u
ct
u
r
e
b
ased
o
n
r
u
le
o
f
t
h
e
s
tr
u
ctu
r
e
s
o
u
r
ce
s
en
te
n
ce
,
f
r
o
m
th
is
it
tr
an
s
lates
to
th
e
t
ar
g
et
lan
g
u
a
g
e
b
ased
o
n
r
u
le
o
f
th
e
s
tr
u
ctu
r
e
tar
g
et
la
n
g
u
a
g
e
an
d
f
r
o
m
t
h
es
e
s
tep
s
a
b
etter
tr
an
s
latio
n
is
p
r
o
d
u
ce
d
to
cr
ea
te
th
e
f
in
a
l step
o
f
th
e
tr
a
n
s
la
tio
n
.
T
h
is
r
esear
ch
u
s
e
s
o
u
r
R
B
MT
(
A
E
-
T
B
M
T
)
d
ev
elo
p
ed
in
p
r
ev
io
u
s
w
o
r
k
[
2
]
.
B
asicall
y
,
t
h
e
tr
an
s
latio
n
p
r
o
ce
s
s
o
f
A
E
-
T
B
MT
c
o
n
s
is
t
s
o
f
s
i
x
m
ain
p
h
a
s
es:
f
ir
s
t,
tex
t
i
n
th
e
s
o
u
r
ce
lan
g
u
a
g
e
is
tr
a
n
s
f
er
r
ed
to
to
k
en
izer
to
d
i
v
id
e
th
e
tex
t
in
to
to
k
e
n
s
.
Seco
n
d
,
s
tar
t
m
o
r
p
h
o
lo
g
ical
an
a
l
y
s
is
to
p
r
o
v
id
e
m
o
r
p
h
o
-
s
y
n
tactic
in
f
o
r
m
atio
n
.
T
h
ir
d
,
t
h
e
s
y
n
t
ac
tic
p
ar
s
er
b
u
ild
s
a
s
y
n
tac
t
ic
r
elev
a
n
t
tr
ee
,
w
h
ich
r
ep
r
esen
t
s
r
elatio
n
s
h
ip
s
b
et
w
ee
n
t
h
e
w
o
r
d
s
o
f
t
h
e
p
h
r
ase.
Fo
r
th
,
le
x
ical
tr
an
s
f
er
w
i
ll
m
ap
A
r
ab
ic
lex
ica
l
ele
m
en
ts
to
t
h
eir
E
n
g
l
is
h
eq
u
iv
ale
n
t.
I
t
w
i
ll
also
m
ap
Ar
ab
ic
m
o
r
p
h
o
lo
g
ical
f
ea
tu
r
e
s
t
o
th
e
co
r
r
esp
o
n
d
in
g
s
et
o
f
E
n
g
lis
h
f
ea
t
u
r
es.
Fi
f
t
h
,
s
tr
u
ct
u
r
e
tr
an
s
f
er
w
i
ll
m
ap
th
e
A
r
ab
ic
d
ep
en
d
en
c
y
tr
ee
to
th
e
eq
u
iv
alen
t
E
n
g
lis
h
s
y
n
tact
ic
s
tr
u
ctu
r
e.
Fi
n
all
y
,
A
r
ab
ic
s
y
n
th
e
s
is
er
w
il
l
s
y
n
t
h
esis
t
h
e
in
f
lecte
d
E
n
g
l
is
h
w
o
r
d
-
f
o
r
m
b
ased
o
n
t
h
e
m
o
r
p
h
o
l
o
g
ical
f
ea
t
u
r
es
a
n
d
tr
av
er
s
es t
h
e
s
y
n
tactic
tr
ee
to
p
r
o
d
u
ce
th
e
s
u
r
f
ac
e
E
n
g
lis
h
p
h
r
ase.
2
.
1
.
1
.
T
o
k
eniza
t
io
n
T
h
is
an
i
m
p
o
r
tan
t
s
tep
f
o
r
a
s
y
n
tac
tic
p
ar
s
er
to
c
o
n
s
tr
u
ct
a
p
h
r
ase
s
tr
u
ct
u
r
e
tr
ee
f
r
o
m
s
y
n
tactic
u
n
i
ts
.
Af
ter
i
n
s
er
ti
n
g
t
h
e
s
o
u
r
ce
s
en
t
en
ce
in
t
h
e
s
y
s
te
m
t
h
e
to
k
e
n
i
z
er
d
iv
id
es th
e
te
x
t i
n
to
to
k
en
s
.
T
h
e
to
k
en
ca
n
b
e
a
w
o
r
d
,
a
p
ar
t o
f
a
w
o
r
d
,
o
r
a
p
u
n
ctu
a
tio
n
m
ar
k
.
A
to
k
e
n
izer
r
e
q
u
ests
to
k
n
o
w
t
h
e
w
h
ite
s
p
ac
es a
n
d
p
u
n
ct
u
atio
n
m
ar
k
s
.
2
.
1
.
2
.
M
o
rpho
lo
g
ica
l a
na
ly
s
is
Af
ter
t
h
e
to
k
e
n
izatio
n
p
r
o
ce
s
s
,
t
h
e
m
o
r
p
h
o
lo
g
ical
an
a
l
y
s
er
w
i
ll
p
r
o
v
id
e
t
h
e
m
o
r
p
h
o
lo
g
ical
in
f
o
r
m
atio
n
ab
o
u
t
w
o
r
d
s
.
I
t
p
r
o
v
id
es
th
e
g
r
a
m
m
atica
l
cla
s
s
o
f
t
h
e
w
o
r
d
s
(
p
ar
ts
o
f
s
p
ee
c
h
)
an
d
cr
ea
tes
th
e
A
r
ab
ic
w
o
r
d
i
n
its
r
i
g
h
t
f
o
r
m
,
d
ep
en
d
in
g
o
n
t
h
e
m
o
r
p
h
o
lo
g
ic
al
f
ea
t
u
r
es.
2
.
1
.
3
.
L
ex
ico
n
I
n
th
i
s
s
y
s
te
m
t
h
e
lex
ico
n
is
ac
co
u
n
tab
le
f
o
r
in
f
er
r
i
n
g
m
o
r
p
h
o
lo
g
ical
an
d
clas
s
i
f
y
in
g
v
er
b
s
,
n
o
u
n
s
,
ad
v
er
b
an
d
ad
j
ec
tiv
es
w
h
e
n
n
ee
d
ed
.
I
t
is
th
e
m
ai
n
le
x
ico
n
tr
an
s
latio
n
;
t
h
e
s
o
u
r
ce
la
n
g
u
ag
e
s
ea
r
ch
e
s
i
n
a
d
ictio
n
ar
y
an
d
th
e
n
ch
o
o
s
es
t
h
e
tr
an
s
la
tio
n
.
A
lex
ico
n
p
r
o
v
id
es
th
e
s
p
ec
i
f
ic
d
etails
ab
o
u
t
ev
er
y
i
n
d
iv
id
u
a
l
lex
ical
e
n
tr
y
(
i.e
.
w
o
r
d
o
r
p
h
r
a
s
e)
in
t
h
e
v
o
ca
b
u
lar
y
o
f
t
h
e
la
n
g
u
a
g
e
co
n
ce
r
n
ed
.
L
ex
ico
n
co
n
tai
n
s
g
r
a
m
m
atica
l
in
f
o
r
m
atio
n
w
h
ic
h
u
s
u
all
y
h
a
v
e
ab
b
r
ev
iated
f
o
r
m
:
‘
n
’
f
o
r
n
o
u
n
,
‘
v
’
f
o
r
v
er
b
,
‘
p
r
o
n
’
f
o
r
p
r
o
n
o
u
n
,
‘
d
et’
f
o
r
d
eter
m
in
er
,
‘
p
r
ep
’
f
o
r
p
r
ep
o
s
itio
n
,
’
ad
j
’
f
o
r
ad
j
ec
tiv
e,
‘
ad
v
’
f
o
r
ad
v
er
b
,
an
d
‘
co
n
j
’
f
o
r
co
n
j
u
n
ctio
n
.
T
h
e
lex
ico
n
m
u
s
t
co
n
tai
n
i
n
f
o
r
m
at
io
n
ab
o
u
t
all
th
e
d
if
f
er
en
t
w
o
r
d
s
th
at
ca
n
b
e
u
s
ed
.
I
f
t
h
e
w
o
r
d
is
am
b
i
g
u
o
u
s
,
it
w
il
l b
e
d
escr
ib
ed
b
y
m
u
ltip
le
en
tr
ies i
n
th
e
le
x
ico
n
,
o
n
e
f
o
r
ea
ch
d
if
f
er
en
t
u
s
e.
2
.
1
.
4
.
P
a
rsin
g
T
h
e
p
ar
s
er
d
iv
id
es
th
e
s
e
n
te
n
ce
in
to
s
m
aller
s
e
ts
d
ep
en
d
in
g
o
n
th
eir
s
y
n
tactic
f
u
n
cti
o
n
s
in
t
h
e
s
en
te
n
ce
.
T
h
er
e
ar
e
f
o
u
r
t
y
p
e
s
o
f
p
h
r
ases
i.e
.
Ver
b
P
h
r
ase
(
VP
)
,
No
u
n
P
h
r
ase
(
NP
)
,
Ad
j
ec
tiv
e/A
d
v
er
b
ial
P
h
r
ase
(
A
P
)
,
an
d
P
r
ep
o
s
itio
n
al
P
h
r
ase
(
P
P
)
.
A
f
ter
th
e
p
ar
s
i
n
g
p
r
o
ce
s
s
t
h
e
s
e
n
te
n
ce
is
r
ep
r
esen
ted
i
n
a
p
h
r
as
e
s
tr
u
ct
u
r
e
tr
ee
.
Fig
u
r
e
.1
s
h
o
w
t
h
e
p
h
r
ase
s
tr
u
ct
u
r
e
tr
ee
f
o
r
th
e
s
en
te
n
ce
سي
ئ
ر
لا
م
ا
يك
ي
ر
ر
ضح
ةمق
لا
(
US
P
r
esid
en
t
atten
d
ed
th
e
s
u
m
m
it
(
.
Fig
u
r
e
1
.
P
h
r
ase
s
tr
u
ct
u
r
e
tr
ee
2
.
1
.
5
.
Sy
nta
ct
ic
rules
A
s
e
t
o
f
A
r
ab
ic
an
d
E
n
g
lis
h
r
u
les
ar
e
f
ed
i
n
to
th
e
s
y
s
te
m
.
I
n
th
i
s
s
tep
t
h
e
r
eo
r
d
er
in
g
p
r
o
c
ess
w
ill
b
e
f
o
u
n
d
w
h
ic
h
w
ill b
e
b
ased
o
n
th
e
o
r
d
er
o
f
w
o
r
d
s
in
a
s
e
n
te
n
ce
,
an
d
h
o
w
t
h
e
w
o
r
d
s
ar
e
g
r
o
u
p
ed
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
R
u
le
B
a
s
ed
a
n
d
E
x
p
ec
ta
tio
n
Ma
ximiza
tio
n
a
lg
o
r
ith
fo
r
A
r
a
b
ic
-
E
n
g
lis
h
Hyb
r
id
.
.
.
.
(
A
r
w
a
A
lq
u
d
s
i
)
75
2
.
1
.
6
.
Ag
re
e
m
e
nt
rules
Af
ter
s
y
n
tact
ic
r
u
les
t
h
e
ag
r
ee
m
en
t
r
u
les
ap
p
lied
w
h
ich
ar
e
r
esp
o
n
s
ib
le
ab
o
u
t
th
e
ad
d
itio
n
s
o
f
p
r
ef
i
x
an
d
s
u
f
f
i
x
in
t
h
e
s
e
n
te
n
ce
s
.
Fi
g
u
r
e
.
2
s
h
o
w
s
a
n
e
x
a
m
p
le
o
f
A
r
ab
ic
r
u
le
s
an
d
th
e
ir
eq
u
iv
al
en
t in
E
n
g
lis
h
S
:
[
NP VP
]
=
=
S
: [
NP:
[
N
1
N2
]
VP
:[
V
P
1 N3
P
2 N4 Prop
]
Ar
a
bic r
ule
S
: [
NP V
P
]
=
=
S
:
[N
P
: [
N2 N
1
]
VP
:[
V
P
1
N3 P2 P
rop N
4
]
]
Eng
li
sh rule
F
ig
u
r
e
.
2
A
r
ab
ic
r
u
les a
n
d
th
ei
r
eq
u
iv
ale
n
t in
E
n
g
lis
h
T
h
er
e
is
a
s
ig
n
i
f
ica
n
t
d
i
f
f
er
en
ce
b
et
w
ee
n
A
r
ab
ic
r
u
le
s
a
n
d
E
n
g
l
is
h
r
u
les.
A
r
ab
ic
s
e
n
ten
c
e
s
tr
u
c
tu
r
e
f
o
llo
w
s
S
u
b
j
ec
t
-
Ob
j
ec
t
-
Ver
b
o
r
Su
b
j
ec
t
-
Ver
b
-
Ob
j
ec
t
o
r
Ver
b
-
Su
b
j
ec
t
-
Ob
j
ec
t
o
r
Ver
b
-
Ob
j
ec
t
-
Su
b
j
ec
t
w
h
er
ea
s
,
E
n
g
lis
h
f
o
llo
w
s
s
t
r
u
ctu
r
e
a
s
S
u
b
j
ec
t
-
Ver
b
-
Ob
j
ec
t.
T
h
is
p
ar
t
is
r
e
s
p
o
n
s
ib
l
e
to
p
er
f
o
r
m
r
u
les
m
atc
h
in
g
f
o
r
th
e
tr
a
n
s
lat
io
n
.
R
u
le
b
a
s
ed
w
ill
m
atch
th
o
s
e
g
ap
s
an
d
tr
a
n
s
late
th
e
s
en
te
n
ce
.
T
o
p
er
f
o
r
m
t
h
e
tr
an
s
latio
n
p
r
o
ce
s
s
it
n
ee
d
s
lo
ts
o
f
r
u
les.
A
d
d
i
n
g
m
o
r
e
r
u
le
s
to
t
h
e
s
y
s
te
m
ca
n
en
h
a
n
ce
t
h
e
ac
c
u
r
ac
y
o
f
th
e
tr
an
s
lated
o
u
tp
u
t
.
2
.
2
.
SM
T
Co
m
po
nent
E
x
p
ec
tatio
n
Ma
x
i
m
izat
io
n
Al
g
o
r
ith
m
(
E
M)
i
s
a
n
iter
ativ
e
s
t
atis
tical
alg
o
r
it
h
m
f
o
r
f
in
d
i
n
g
m
ax
i
m
u
m
lik
eli
h
o
o
d
o
f
m
i
s
s
ed
d
ata
o
r
m
is
s
ed
tr
an
s
la
tio
n
s
;
ex
p
ec
t
it
is
t
h
e
m
o
s
t
ac
cu
r
ate
o
n
e,
u
s
ed
as
t
h
e
SMT
co
m
p
o
n
e
n
t
s
y
s
te
m
in
t
h
i
s
w
o
r
k
.
T
h
e
E
x
p
ec
tatio
n
Ma
x
i
m
izatio
n
al
g
o
r
ith
m
co
n
tain
s
t
w
o
p
r
o
ce
s
s
es
in
ea
ch
o
f
t
h
e
iter
atio
n
:
T
h
e
E
-
s
tep
(
E
x
p
ec
tatio
n
)
,
an
d
th
e
M
-
s
tep
(
Ma
x
i
m
izatio
n
)
.
I
n
E
-
s
tep
u
s
e
cu
r
r
en
t
o
b
s
er
v
atio
n
s
an
d
p
ar
a
m
eter
s
to
co
m
p
u
te
th
e
p
r
o
b
ab
ilit
y
o
f
all
p
o
s
s
ib
le
co
llo
ca
tio
n
o
f
t
h
e
d
ata
.
T
h
is
s
tep
is
ac
h
iev
ed
b
y
u
s
in
g
t
h
e
co
n
d
itio
n
al
ex
p
ec
tatio
n
.
)
(
)
|
,
(
1
j
j
J
j
c
e
a
t
E
A
C
P
C
j
j
J
j
j
j
J
j
A
c
l
e
a
t
e
a
t
E
A
C
P
E
A
C
P
A
E
C
P
)
(
)
(
)
,
(
)
,
(
)
,
(
1
1
a.
A
r
ef
er
s
to
in
p
u
t
w
o
r
d
s
(
A
r
ab
ic
w
o
r
d
s
)
b.
E
r
ef
er
s
to
o
u
tp
u
t
w
o
r
d
s
(
E
n
g
lis
h
w
o
r
d
s
)
c.
J
,
th
e
n
u
m
b
er
o
f
w
o
r
d
s
in
E
:
E
=
e
1
, e
2
, …e
J
d.
c
1
-
to
-
m
a
n
y
co
llo
ca
tio
n
C
:
C
=
c
1
,
c
2
, …
c
J
e.
Fo
r
ea
ch
p
o
s
itio
n
in
A
,
g
e
n
er
at
e
a
w
o
r
d
c
j
f
r
o
m
th
e
co
llo
ca
tio
n
w
o
r
d
in
E:
j
c
e
I
n
M
-
s
tep
co
llo
ca
tio
n
p
r
o
b
a
b
ilit
y
w
h
ic
h
ac
h
ie
v
ed
in
E
-
s
tep
ar
e
u
s
ed
to
re
-
esti
m
ate
th
e
v
al
u
es
o
f
all
p
ar
am
eter
s
u
n
til
co
n
v
er
g
e.
T
h
e
esti
m
ate
o
f
t
h
e
m
i
s
s
i
n
g
d
ata
f
r
o
m
t
h
e
e
x
p
ec
tatio
n
s
tep
is
u
s
ed
i
n
s
tead
o
f
th
e
ac
tu
al
m
is
s
i
n
g
d
ata.
I
ter
atin
g
t
h
ese
s
tep
s
w
ill
lead
to
a
co
n
v
e
r
g
en
ce
t
h
at
est
i
m
a
tes
t
h
e
m
a
x
i
m
u
m
l
ik
eli
h
o
o
d
o
f
tr
an
s
latio
n
.
T
h
e
E
M
al
g
o
r
ith
m
i
s
d
is
c
u
s
s
ed
in
d
etail
b
y
[
1
7
]
.
A
co
m
m
o
n
tec
h
n
iq
u
e
f
o
r
m
ax
i
m
u
m
l
ik
eli
h
o
o
d
esti
m
atio
n
o
f
m
o
d
el
p
ar
a
m
eter
s
in
t
h
e
p
r
esen
ce
o
f
m
i
s
s
i
n
g
d
a
ta
is
E
M
alg
o
r
ith
m
[
1
8
]
.
3.
I
M
P
L
E
M
E
NT
AT
I
O
N
T
h
e
ex
p
er
im
e
n
t
s
h
a
v
e
b
ee
n
c
o
n
d
u
c
ted
o
n
A
r
ab
ic
to
E
n
g
li
s
h
.
H
y
b
r
id
ap
p
r
o
ac
h
r
elies
o
n
s
tr
u
ct
u
r
es
o
u
tp
u
t b
y
t
h
e
co
m
p
o
n
e
n
t
R
u
le
B
ased
an
d
Statis
tical
s
y
s
te
m
s
.
R
u
le
b
ased
m
ac
h
i
n
e
tr
an
s
lati
o
n
ap
p
r
o
ac
h
g
en
er
ates
ca
n
d
i
d
ate
s
en
ten
ce
s
f
o
r
ea
ch
i
n
p
u
t
s
en
te
n
ce
.
Sin
ce
o
n
e
w
o
r
d
o
f
A
r
ab
ic
ca
n
h
a
v
e
m
an
y
m
ea
n
in
g
s
in
E
n
g
lis
h
,
t
h
u
s
it
i
s
i
m
p
o
r
tan
t
to
i
d
en
tify
t
h
e
s
u
itab
le
m
ea
n
in
g
o
f
a
w
o
r
d
f
o
r
th
e
s
o
u
r
ce
s
en
te
n
ce
.
T
h
er
ef
o
r
e
th
is
g
en
er
ate
s
ca
n
d
id
ate
s
en
te
n
ce
s
w
it
h
ea
ch
m
ea
n
i
n
g
o
f
a
w
o
r
d
.
I
t
cr
ea
tes
tr
an
s
lat
ed
E
n
g
lis
h
ca
n
d
id
ates
s
e
n
ten
ce
s
w
h
ic
h
ar
e
g
r
a
m
m
at
ic
all
y
co
r
r
ec
t
b
u
t
d
u
e
to
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
5
,
No
.
2
,
J
u
n
e
2
0
1
6
:
72
–
79
76
a
m
b
ig
u
o
u
s
m
ea
n
i
n
g
s
o
f
w
o
r
d
s
th
er
e
m
a
y
b
e
m
ea
n
i
n
g
les
s
s
e
n
ten
ce
s
.
E
M
alg
o
r
ith
m
w
ill
c
h
o
o
s
e
m
o
s
t
s
u
i
tab
le
s
en
te
n
ce
f
r
o
m
th
o
s
e
ca
n
d
id
ate
s
en
te
n
ce
s
.
T
h
is
is
th
e
p
lace
w
h
er
e
SMT
t
ec
h
n
iq
u
es
co
m
e
to
th
e
w
o
r
k
.
F
r
o
m
R
u
le
b
ased
p
ar
t
g
en
er
ate
t
r
an
s
lated
ca
n
d
id
ate
s
en
te
n
ce
s
f
o
r
th
e
s
o
u
r
ce
s
en
te
n
ce
s
b
ased
o
n
th
e
n
u
m
b
er
o
f
a
m
b
i
g
u
o
u
s
w
o
r
d
s
an
d
th
e
n
u
m
b
er
o
f
a
m
b
ig
u
o
u
s
m
ea
n
in
g
s
o
f
ea
ch
w
o
r
d
.
B
ased
o
n
t
h
e
p
r
ese
n
ce
o
f
a
m
b
i
g
u
o
u
s
m
ea
n
in
g
s
o
f
th
e
w
o
r
d
s
,
t
h
er
e
m
a
y
b
e
m
an
y
ca
n
d
id
ate
s
e
n
te
n
ce
s
f
o
r
s
o
u
r
ce
s
en
t
en
ce
.
T
h
en
w
e
u
s
e
a
co
r
p
u
s
to
m
atc
h
m
o
s
t
s
u
i
t
ab
le
s
en
te
n
ce
f
r
o
m
th
e
g
iv
e
n
ca
n
d
id
ate
s
e
n
te
n
ce
s
we
h
av
e
u
s
ed
Un
i
ted
Nati
o
n
s
(
A
r
ab
ic
-
E
n
g
lis
h
)
p
ar
allel
co
r
p
u
s
[
2
0
]
s
a
m
e
tr
ain
i
n
g
a
n
d
test
d
ata
s
p
lit
was
u
s
ed
as
i
n
[
31
]
:
1
,
0
0
0
,
0
0
0
tr
ain
in
g
s
en
ten
ce
p
air
s
an
d
test
ed
o
n
9
9
4
test
s
en
te
n
ce
s
.
T
h
en
w
e
u
s
e
E
M
al
g
o
r
ith
m
to
m
atc
h
o
t
h
er
r
elate
d
w
o
r
d
s
to
ea
ch
w
o
r
d
i
n
t
h
e
ca
n
d
id
ate
s
e
n
ten
ce
,
b
y
u
s
i
n
g
t
h
o
s
e
r
elate
d
w
o
r
d
s
w
e
cr
ea
te
n
e
w
s
e
n
te
n
ce
s
f
o
r
ea
ch
ca
n
d
id
ate
s
e
n
ten
ce
.
T
h
e
n
f
in
d
th
e
p
r
o
b
ab
ilit
y
o
f
ea
ch
s
e
n
ten
ce
b
y
m
atc
h
i
n
g
w
it
h
th
e
co
r
p
u
s
.
A
ll
o
f
th
e
s
e
p
r
o
b
ab
ilit
y
ca
lc
u
latio
n
s
w
ill
b
e
co
n
s
id
er
ed
to
s
elec
t
th
e
m
o
s
t
s
u
i
tab
le
s
en
t
en
ce
.
T
h
en
w
e
s
elec
t
t
h
e
co
r
r
ec
t
s
e
n
te
n
ce
as
th
e
ca
n
d
id
ate
s
e
n
ten
ce
w
h
ich
h
as
h
ig
h
es
t
p
r
o
b
ab
ilit
y
b
e
ca
u
s
e
it
h
as
a
h
i
g
h
p
o
s
s
ib
ilit
y
to
b
e
a
m
ea
n
i
n
g
f
u
l
an
d
co
r
r
ec
t
s
en
ten
ce
.
T
h
e
ar
ch
itect
u
r
e
o
f
th
i
s
HM
T
s
y
s
te
m
is
il
lu
s
tr
ated
o
n
Fig
u
r
e
.
3.
T
h
e
d
ata
d
r
iv
en
R
B
MT
an
d
SMT
m
et
h
o
d
s
ar
e
r
o
b
u
s
t.
T
h
e
f
ea
tu
r
e
m
ak
e
s
s
u
ch
s
y
s
te
m
s
v
er
y
attr
ac
tiv
e
a
s
t
h
e
y
al
w
a
y
s
p
r
o
d
u
ce
tr
a
n
s
latio
n
,
ir
r
esp
ec
ti
v
e
o
f
t
h
e
i
n
p
u
t
s
t
r
in
g
.
;
I
f
a
R
B
M
T
s
y
s
te
m
d
o
es
n
o
t
f
i
n
d
a
s
eq
u
en
ce
o
f
r
u
les
w
h
ic
h
ca
n
b
e
ap
p
lied
s
u
cc
e
s
s
f
u
ll
y
to
th
e
i
n
p
u
t
,
th
e
n
t
h
e
SMT
w
i
ll
b
e
id
en
ti
f
ied
a
n
d
p
r
o
d
u
ce
d
.
Ho
w
e
v
er
,
s
tati
s
tical
s
y
s
te
m
is
n
o
t
g
o
o
d
at
m
o
d
elli
n
g
lin
g
u
i
s
tic
p
h
e
n
o
m
e
n
a
s
u
c
h
as
w
o
r
d
o
r
d
er
an
d
ag
r
ee
m
e
n
t.
I
n
co
n
tr
ast,
R
B
MT
s
y
s
te
m
,
ca
n
h
a
n
d
le
li
n
g
u
i
s
tic
p
h
en
o
m
e
n
a
,
s
u
c
h
a
s
,
w
o
r
d
o
r
d
er
an
d
u
s
in
g
h
a
n
d
-
w
r
itte
n
r
u
les
an
d
d
ictio
n
ar
ies
.
T
h
u
s
,
t
h
e
ad
v
a
n
ta
g
e
o
f
co
m
b
in
i
n
g
t
h
e
p
o
s
iti
v
e
ele
m
e
n
t
s
o
f
t
h
e
r
u
le
b
ased
ap
p
r
o
ac
h
an
d
s
tatis
tical
ap
p
r
o
ac
h
to
MT
a
r
e
clea
r
:
a
co
m
b
i
n
ed
m
o
d
el
h
as
t
h
e
p
o
s
s
ib
ilit
y
to
b
e
r
o
b
u
s
t,
h
ig
h
l
y
ac
cu
r
ate,
co
s
t
e
f
f
ec
tiv
e
to
b
u
i
ld
an
d
ad
ap
tab
le.
C
o
m
b
in
i
n
g
r
u
les
w
i
th
lin
g
u
is
tic
i
n
f
o
r
m
ati
o
n
an
d
a
s
tatis
tical
tr
an
s
latio
n
m
o
d
el
m
i
g
h
t
r
es
u
lt
as
a
h
y
b
r
id
m
o
d
el.
T
h
e
m
o
ti
v
atio
n
s
f
o
r
ad
o
p
tin
g
h
y
b
r
id
m
o
d
el
ar
e
p
r
ec
is
el
y
as
m
en
tio
n
ed
b
ef
o
r
e:
it c
o
m
b
i
n
es
th
e
r
o
b
u
s
t
n
ess
o
f
R
B
MT
an
d
SMT
ap
p
r
o
ac
h
es
.
Fig
u
r
e
3
.
T
h
e
A
r
ch
itect
u
r
e
o
f
HM
T
s
y
s
te
m
4.
E
XP
E
R
I
M
E
NT
A
L
RE
SUL
T
We
h
av
e
ev
al
u
ated
o
u
r
s
y
s
te
m
u
s
in
g
B
L
E
U
[
3
0
]
s
co
r
es
a
g
ain
s
t
t
w
o
r
ef
er
en
ce
h
u
m
a
n
t
r
an
s
latio
n
.
B
leu
s
co
r
e
f
o
r
b
o
th
o
f
t
h
e
s
y
s
te
m
is
ca
lcu
la
ted
an
d
it
is
d
es
cr
ib
ed
in
T
a
b
le
1
.
T
h
e
tab
le
s
h
o
w
s
t
h
e
v
al
u
e
s
o
f
B
L
E
U
o
b
tain
ed
f
o
r
p
h
r
ase
le
n
g
th
:
1
-
g
r
a
m
,
2
-
gr
am
,
3
-
g
r
a
m
,
an
d
4
-
g
r
a
m,
r
esp
ec
tiv
el
y
.
No
te
th
at
B
L
E
U
i
s
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
R
u
le
B
a
s
ed
a
n
d
E
x
p
ec
ta
tio
n
Ma
ximiza
tio
n
a
lg
o
r
ith
fo
r
A
r
a
b
ic
-
E
n
g
lis
h
Hyb
r
id
.
.
.
.
(
A
r
w
a
A
lq
u
d
s
i
)
77
b
et
w
ee
n
0
an
d
1
(
0
≤
B
le
u
<1
)
.
W
h
en
B
L
E
U
v
a
lu
e
i
s
clo
s
e
to
1
,
w
h
ic
h
m
ea
n
s
t
h
e
q
u
al
it
y
o
f
tr
an
s
latio
n
is
b
etter
an
d
clo
s
e
to
t
h
e
m
a
n
u
al
tr
an
s
latio
n
.
I
n
t
h
is
ev
a
lu
atio
n
1
ca
n
d
id
ate
f
ile
(
r
e
p
r
esen
t
o
u
r
s
y
s
te
m
tr
an
s
latio
n
)
a
n
d
t
w
o
r
e
f
er
en
c
es
f
iles
(
r
ep
r
esen
t
2
d
i
f
f
er
en
t
m
a
n
u
al
tr
a
n
s
lat
io
n
)
h
av
e
b
ee
n
u
s
ed
.
I
t
ca
n
b
e
clea
r
l
y
s
ee
n
t
h
at
s
co
r
e
o
f
HM
T
s
y
s
te
m
is
h
ig
h
er
th
a
n
SMT
an
d
R
B
MT
s
y
s
te
m
s
i
n
all
ca
s
es.
W
h
en
co
m
p
ar
ed
th
r
ee
ap
p
r
o
ac
h
es,
HM
T
o
u
tp
er
f
o
r
m
ed
SMT
an
d
R
B
MT
in
all
B
L
E
U
s
co
r
e
.
Af
ter
p
er
f
o
r
m
i
n
g
a
n
al
y
s
i
s
to
w
ar
d
th
e
o
u
tp
u
t
o
f
R
B
MT
,
w
e
f
o
u
n
d
th
at
,
co
m
p
r
eh
e
n
s
i
v
e
r
eo
r
d
e
r
in
g
r
u
les
p
la
y
a
n
i
m
p
o
r
tan
t
r
o
le
i
n
t
h
e
q
u
alit
y
o
f
tr
an
s
latio
n
.
As
HM
T
s
y
s
te
m
y
ield
ed
a
g
o
o
d
im
p
r
o
v
e
m
e
n
t
i
n
B
L
E
U
p
o
in
ts
o
v
er
tr
an
s
lati
n
g
o
f
SMT
s
y
s
te
m
.
I
n
ad
d
itio
n
,
m
o
r
e
d
ata
tr
ai
n
i
n
g
m
a
k
es
th
e
o
u
tp
u
t
o
f
SMT
m
o
r
e
ac
cu
r
ate.
Fig
u
r
e
.
4
illu
s
tr
ates
h
o
w
HM
T
s
y
s
te
m
tr
a
n
s
lat
io
n
is
clo
s
er
th
an
SMT
s
y
s
te
m
tr
a
n
s
lat
io
n
to
m
an
u
al
tr
an
s
latio
n
w
it
h
p
h
r
ase
le
n
g
th
:
1
-
g
r
a
m
,
2
-
g
r
a
m
,
3
-
g
r
a
m
,
a
n
d
4
-
g
r
a
m
,
r
esp
ec
ti
v
el
y
.
W
e
b
eliev
e
t
h
at
,
a
g
o
o
d
tr
an
s
latio
n
co
u
ld
b
e
ac
h
ie
v
ed
w
h
e
n
R
B
MT
is
co
m
b
i
n
ed
w
it
h
SMT
,
as
R
B
MT
s
o
lv
es
w
o
r
d
o
r
d
er
in
g
p
r
o
b
lem
w
h
e
n
tr
an
s
lated
f
r
o
m
A
r
ab
ic
t
o
E
n
g
lis
h
,
an
d
SMT
s
o
lv
e
s
th
e
a
m
b
ig
u
it
y
p
r
o
b
lem
.
T
ab
le
1
B
lu
e
ev
alu
atio
n
r
es
u
lt
s
o
f
HM
T
an
d
SMT
P
h
r
a
se
l
e
n
g
t
h
n
-
g
r
a
m
H
M
T
S
M
T
R
B
M
T
1
-
g
r
a
m
0
.
8
8
0
.
7
3
0
.
7
1
2
-
g
r
a
m
0
.
8
0
0
.
6
1
0
.
5
7
3
-
g
r
a
m
0
.
6
6
0
.
5
6
0
.
5
0
4
-
g
r
a
m
0
.
5
1
0
.
3
7
0
.
3
3
Fig
u
r
e
4
.
B
leu
s
co
r
e
o
f
HM
T
,
SMT
an
d
R
B
MT
w
it
h
p
h
r
ase
l
en
g
t
h
1
-
g
r
a
m,
2
-
g
r
a
m,
3
-
g
r
a
m
,
an
d
4
-
g
r
a
m
.
5.
CO
NCLU
SI
O
N
I
n
th
is
w
o
r
k
w
e
h
ad
p
r
esen
te
d
a
u
n
iq
u
e
co
m
p
le
m
e
n
tar
y
wa
y
to
co
m
b
i
n
e
r
u
le
-
b
a
s
ed
an
d
s
tati
s
tical
ap
p
r
o
ac
h
es
to
HM
T
,
as
it
in
t
er
w
ea
v
es
t
h
e
p
h
ilo
s
o
p
h
ies
o
f
th
e
r
u
le
b
ased
,
an
d
s
tatis
tica
l
ap
p
r
o
ac
h
es
in
a
n
in
te
g
r
ated
f
r
a
m
e
w
o
r
k
.
T
h
e
g
o
al
o
f
co
m
b
i
n
ed
th
o
s
e
ap
p
r
o
ac
h
es
ef
f
ec
t
iv
el
y
to
o
b
tain
m
o
r
e
ac
cu
r
ate
r
esu
lts
th
an
o
t
h
er
ex
i
s
ti
n
g
ap
p
r
o
ac
h
es
.
T
h
is
m
o
d
el
h
as
th
e
ca
p
ac
i
t
y
to
co
m
b
i
n
e
th
e
li
n
g
u
i
s
tic
co
m
p
l
ex
it
y
o
f
r
u
le
b
ased
m
o
d
els
o
f
t
r
an
s
latio
n
w
it
h
t
h
e
r
o
b
u
s
t
n
es
s
an
d
ad
ap
tab
ilit
y
o
f
s
ta
tis
tica
l
m
et
h
o
d
.
T
h
e
m
o
d
el
also
h
elp
s
to
ad
d
r
ess
lan
g
u
ag
e
a
m
b
ig
u
it
y
p
r
o
b
le
m
w
h
ic
h
is
a
o
n
e
o
f
t
h
e
b
ig
g
es
t
c
h
alle
n
g
e
u
n
d
er
R
B
MT
an
d
s
o
lv
e
l
ex
ical
a
n
al
y
s
i
s
a
n
d
s
y
n
tactic
an
al
y
s
is
r
eq
u
ir
e
m
e
n
t
p
r
o
b
lem
i
n
S
MT
.
T
h
u
s
,
w
e
h
a
d
p
r
esen
ted
th
e
i
m
p
le
m
e
n
ta
tio
n
d
etail
s
o
f
o
u
r
h
y
b
r
id
s
y
s
te
m
,
w
h
ic
h
w
a
s
i
n
s
p
ir
ed
b
y
r
u
le
b
ased
w
it
h
th
e
E
M
alg
o
r
ith
m
f
o
r
s
tat
is
tical
m
e
t
h
o
d
;
th
e
s
y
s
te
m
h
as
d
o
cu
m
en
ted
lar
g
er
s
ca
le,
m
o
r
e
tr
an
s
latio
n
s
an
d
co
m
p
le
x
ex
p
er
im
e
n
t
s
.
T
h
is
e
m
p
ir
ical
ev
a
lu
atio
n
s
h
o
w
ed
f
o
r
th
e
A
r
ab
ic
to
E
n
g
li
s
h
Un
i
ted
Natio
n
s
p
ar
allel
co
r
p
u
s
.
T
h
e
m
o
ti
v
atio
n
b
eh
i
n
d
th
i
s
r
esear
ch
is
co
m
b
i
n
i
n
g
th
e
ad
v
a
n
tag
e
o
f
in
f
o
r
m
a
tio
n
pr
ese
n
t
i
n
ea
ch
o
f
th
e
MT
s
y
s
te
m
to
g
et
b
etter
t
r
an
s
latio
n
r
esu
lt.
E
v
al
u
atio
n
b
y
u
s
i
n
g
B
le
u
s
co
r
e
in
d
icato
r
s
h
o
w
s
t
h
at:
1
)
.
T
h
e
s
ize
o
f
th
e
tr
ai
n
in
g
d
ata
ef
f
e
cts
th
e
s
tat
is
tical
m
o
d
el
o
n
SMT
an
d
HM
T
s
y
s
te
m
,
s
o
ad
d
in
g
m
o
r
e
tr
ain
in
g
co
r
p
u
s
ca
n
i
m
p
r
o
v
e
th
e
p
er
f
o
r
m
an
ce
HM
T
s
y
s
te
m
.
2
)
.
HM
T
s
y
s
te
m
o
u
tp
er
f
o
r
m
s
SMT
an
d
R
B
MT
s
y
s
te
m
s
i
n
all
ca
s
es.
W
e
h
ad
id
en
ti
f
ie
d
th
at
h
y
b
r
id
s
o
lu
tio
n
s
ten
d
to
co
m
b
in
e
t
h
e
ad
v
a
n
ta
g
es
o
f
th
e
in
d
i
v
id
u
a
l
ap
p
r
o
ac
h
es
to
ac
h
iev
e
an
o
v
er
all
b
etter
tr
an
s
latio
n
.
T
h
e
ap
p
r
o
ac
h
is
m
o
s
t
u
s
e
f
u
l
to
a
d
d
r
ess
o
n
e
o
f
R
u
le
-
B
ased
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
5
,
No
.
2
,
J
u
n
e
2
0
1
6
:
72
–
79
78
MT
g
r
ea
test
c
h
alle
n
g
e
s
–
tr
a
n
s
lat
io
n
a
m
b
i
g
u
i
t
y
.
W
h
e
n
a
w
o
r
d
/p
h
r
ase
ca
n
h
a
v
e
m
o
r
e
th
an
o
n
e
m
ea
n
i
n
g
,
s
tatis
t
ics ca
n
h
elp
id
en
ti
f
y
t
h
e
m
o
s
t s
u
itab
le
o
p
tio
n
.
RE
F
E
R
E
NC
E
S
[1
]
A
lq
u
d
si
A
,
Om
a
r
N,
S
h
a
k
e
r
K.
Ara
b
ic M
a
c
h
in
e
T
ra
n
sla
ti
o
n
:
a
S
u
r
v
e
y
,
A
rti
f
icia
l
In
telli
g
e
n
c
e
Re
v
ie
w
.
2
0
1
2
:
1
-
20.
[2
]
Ha
te
m
A
,
O
m
a
r
N.
S
y
n
ta
c
ti
c
re
o
r
d
e
rin
g
f
o
r
A
ra
b
ic
-
En
g
li
sh
p
h
ra
se
-
b
a
se
d
m
a
c
h
in
e
tran
sla
ti
o
n
.
Da
tab
a
se
T
h
e
o
r
y
a
n
d
A
p
p
li
c
a
ti
o
n
,
B
io
-
S
c
ien
c
e
a
n
d
Bi
o
-
T
e
c
h
n
o
lo
g
y
.
S
p
rin
g
e
r L
e
c
tu
re
N
o
tes
in
C
o
mp
u
ter
S
c
ien
c
e
.
1
1
8
,
2
0
1
0
:
1
9
8
-
2
0
6
.
[3
]
Ha
te
m
,
A
O
m
a
r,
N
S
h
a
k
e
r,
K.
M
o
rp
h
o
l
o
g
ica
l
a
n
a
lys
is
fo
r
ru
le
b
a
se
d
m
a
c
h
in
e
tr
a
n
sl
a
ti
o
n
,
S
e
ma
n
ti
c
T
e
c
h
n
o
l
o
g
y
an
d
In
f
o
rm
a
ti
o
n
Retrie
v
a
l
(
S
T
AIR
).
I
n
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
,
2
0
1
1
:
2
6
0
-
2
6
3
.
[4
]
A
r
wa
A
lq
u
d
si,
Na
z
li
a
h
Om
a
r,
Ra
b
h
a
W
,
Ib
ra
h
im
,
A
N
e
w
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
Ev
a
lu
a
ti
o
n
M
e
tri
c
Ba
se
d
On
T
h
e
G
e
o
m
e
tri
c
M
e
a
n
,
(2
0
1
5
)
.
T
o
a
p
p
e
a
r.
[5
]
Ra
b
h
a
W
.
Ib
ra
h
im
,
A
r
wa
A
lq
u
d
s
i
a
n
d
Na
z
li
a
h
Om
a
r,
A
Ne
w
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
Ev
a
lu
a
ti
o
n
M
e
tr
ic
Util
izin
g
t
h
e
Ho
ld
e
r
M
e
a
n
,
(2
0
1
5
).
T
o
a
p
p
e
a
r
.
[6
]
Ch
a
ro
e
n
p
o
rn
sa
w
a
t,
P
S
o
r
n
lertlam
v
a
n
ich
,
V
Ch
a
ro
e
n
p
o
r
n
,
T
.
Imp
ro
v
in
g
T
ra
n
sla
ti
o
n
Qu
a
li
ty o
f
Ru
le
-
b
a
se
d
M
a
c
h
in
e
T
ra
n
sl
a
ti
o
n
.
I
n
:
P
ro
c
e
e
d
in
g
s o
f
C
OL
IN
G
W
o
rk
sh
o
p
o
n
M
a
c
h
i
n
e
T
ra
n
sla
ti
o
n
in
A
sia
,
2
0
0
2
:
3
5
1
-
3
5
6
.
[7
]
F
a
rrú
s,
M
M
a
riñ
o
,
JB
P
o
c
h
,
M
He
rn
á
n
d
e
z
,
A
He
n
ríq
u
e
z
,
C
F
o
n
o
ll
o
sa
,
JA
R
Co
sta
-
Ju
ss
à
,
M
R.
Ov
e
rc
o
m
in
g
S
tatisti
c
a
l
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
L
i
m
it
a
ti
o
n
s:
Err
o
r
A
n
a
ly
sis
a
n
d
P
r
o
p
o
se
d
S
o
l
u
ti
o
n
s
f
o
r
th
e
Ca
tala
n
---
S
p
a
n
is
h
L
a
n
g
u
a
g
e
P
a
ir.
I
n
:
J
o
u
rn
a
l
L
a
n
g
u
a
g
e
Res
o
u
rc
e
s a
n
d
Eva
l
u
a
t
io
n
.
2
0
1
1
;
4
5
(2
)
:1
8
1
-
2
0
8
.
[8
]
Ca
rp
u
a
t,
M
W
u
,
D
.
Imp
ro
v
i
n
g
S
ta
t
ist
ica
l
M
a
c
h
in
e
T
r
a
n
sl
a
ti
o
n
u
si
n
g
W
o
r
d
S
e
n
se
Disa
m
b
ig
u
a
ti
o
n
.
I
n
:
Jo
i
n
t
Co
n
f
e
re
n
c
e
o
n
E
m
p
iri
c
a
l
M
e
th
o
d
s
in
Na
tu
ra
l
Lan
g
u
a
g
e
P
ro
c
e
ss
in
g
a
n
d
Co
m
p
u
tatio
n
a
l
Na
tu
ra
l
L
a
n
g
u
a
g
e
L
e
a
rn
in
g
(EM
NL
P
-
Co
NL
L
),
2
0
0
5
:
6
1
-
7
2
.
[9
]
Bo
jar
O,
M
Cm
e
jre
k
,
M
a
th
e
m
a
ti
c
a
l
M
o
d
e
l
o
f
T
re
e
T
r
a
n
sf
o
r
m
a
ti
o
n
s.
ˇ
P
r
o
jec
t
E
u
ro
m
a
tri
x
–
De
li
v
e
ra
b
le
3
.
2
,
P
ra
g
u
e
,
Cz
e
c
h
Re
p
u
b
l
ic 2
0
0
7
.
[1
0
]
S
h
a
a
lan
,
K
A
l
-
S
h
e
ik
h
,
S
Oro
u
m
c
h
ian
,
F
.
2
0
1
2
.
Qu
e
ry
Exp
a
n
si
o
n
Ba
se
d
-
o
n
S
imil
a
rity
o
f
T
e
rm
s
fo
r
Imp
ro
v
in
g
Ara
b
ic I
n
fo
rm
a
ti
o
n
Retrie
v
a
l
.
In
t
e
ll
ig
e
n
t
In
f
o
rm
a
ti
o
n
P
ro
c
e
ss
in
g
VI.
2
0
1
2
:
1
6
7
–
1
7
6
.
[1
1
]
Ch
e
n
,
K,
Ch
e
n
,
H
.
A
h
y
b
rid
a
p
p
ro
a
c
h
to
ma
c
h
i
n
e
tra
n
sl
a
ti
o
n
s
y
ste
m
d
e
sig
n
.
In
Co
m
p
.
L
in
g
u
ist.
a
n
d
Ch
i
n
e
se
L
a
n
g
u
a
g
e
P
ro
c
e
ss
in
g
2
3
,
1
9
9
7
:
2
4
1
–
2
6
5
.
[1
2
]
Ha
ss
a
n
S
a
wa
f
.
2
0
1
0
.
Ara
b
ic
d
i
a
l
e
c
t
h
a
n
d
li
n
g
in
h
y
b
ri
d
ma
c
h
in
e
t
ra
n
sla
t
io
n
.
I
n
P
r
o
c
e
e
d
in
g
s
o
f
th
e
Co
n
f
e
re
n
c
e
o
f
th
e
A
ss
o
c
iatio
n
f
o
r
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
i
n
t
h
e
Am
e
rica
s (
A
M
TA
),
De
n
v
e
r,
Co
lo
ra
d
o
.
[1
3
]
S
h
irai,
S
,
e
t
a
l.
A
Hy
b
rid
Ru
le
a
n
d
Ex
a
m
p
le
-
b
a
se
d
M
e
th
o
d
f
o
r
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
,
NT
T
Co
m
m
in
i
c
a
ti
o
n
S
c
h
ien
c
e
L
a
b
o
ra
to
ries
,
p
p
.
1
-
5.
[1
4
]
Ei
se
le,
A
F
e
d
e
r
m
a
n
n
,
C
Us
z
k
o
re
it
,
H
S
a
in
t
-
Am
a
n
d
,
H
Ka
y
,
M
Je
ll
in
g
h
a
u
s,
M
Hu
n
sic
k
e
r,
S
He
rrm
a
n
n
,
T
Yu
Ch
e
n
.
Hy
b
rid
ma
c
h
in
e
tra
n
sla
ti
o
n
a
rc
h
it
e
c
tu
re
s
wit
h
in
a
n
d
b
e
y
o
n
d
t
h
e
e
u
ro
ma
trix
p
ro
jec
t
.
I
n
P
ro
c
e
e
d
in
g
s
o
f
Eu
ro
p
e
a
n
A
s
so
c
iatio
n
o
f
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
(Ha
m
b
u
rg
),
2
0
0
8
:
27
–
3
4
.
[1
5
]
S
´
a
n
c
h
e
z
-
M
a
rt
´
ın
e
z
,
F
,
Ne
y
,
H
.
Us
in
g
a
li
g
n
me
n
t
tem
p
la
tes
to
in
f
e
r
sh
a
ll
o
w
-
tra
n
sfe
r
ma
c
h
in
e
tra
n
sla
ti
o
n
ru
les
.
In
L
e
c
tu
re
No
tes
in
Co
mp
u
ter
S
c
i
e
n
c
e
4
1
3
9
,
P
ro
c
e
e
d
in
g
s
o
f
F
in
TAL
,
5
th
In
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
N
a
tu
ra
l
L
a
n
g
u
a
g
e
P
ro
c
e
ss
in
g
,
2
0
0
6
:
7
5
6
–
7
6
7
.
[1
6
]
F
e
d
e
rm
a
n
n
,
C.
Hy
b
rid
M
a
c
h
i
n
e
T
ra
n
sl
a
ti
o
n
Us
i
n
g
J
o
in
t,
B
in
a
rise
d
Fea
t
u
re
Vec
to
rs
.
I
n
:
P
ro
c
e
e
d
i
n
g
s
o
f
th
e
2
0
t
h
Co
n
f
e
re
n
c
e
o
f
th
e
A
s
so
c
iatio
n
f
o
r
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
in
th
e
A
m
e
r
ica
s (
A
M
TA
2
0
1
2
)
.
2
0
1
2
:
1
1
3
–
1
1
8
.
[1
7
]
G
e
o
ff
re
y
M
c
L
a
c
h
lan
a
n
d
T
h
ri
y
a
m
b
a
k
a
m
Krish
n
a
n
.
T
h
e
EM
A
lg
o
rit
h
m
a
n
d
Ex
ten
sio
n
s.
J
o
h
n
W
il
e
y
&
S
o
n
s,
Ne
w
Yo
rk
,
1
9
9
6
.
[1
8
]
AP
De
m
p
ste
r,
NM
L
a
ird
,
DB
Ru
b
in
.
M
a
x
im
u
m
li
k
e
li
h
o
o
d
f
ro
m
in
c
o
m
p
lete
d
a
ta
v
ia
t
h
e
EM
a
lg
o
rit
h
m
.
J.
o
f
t
h
e
RS
S
o
c
iety
se
rie
s B,
1
9
7
7
;
3
9
:1
–
38
.
[1
9
]
Co
rb
i
-
Be
ll
o
t,
A
M
,
F
o
rc
a
d
a
,
M
L
Ortiz
-
Ro
jas
,
S
P
e
re
z
-
Ortiz,
JA
R
a
m
irez
-
S
a
n
c
h
e
z
,
G
S
a
n
c
h
e
z
-
M
a
r
ti
n
e
z
,
F
leg
ria,
I
M
a
y
o
r,
A
S
a
ra
so
la,
K
.
An
Op
e
n
-
S
o
u
rc
e
S
h
a
ll
o
w
-
T
ra
n
sfe
r M
a
c
h
i
n
e
T
ra
n
sl
a
ti
o
n
En
g
i
n
e
fo
r th
e
Ro
m
a
n
c
e
L
a
n
g
u
a
g
e
s
o
f
S
p
a
in
.
In
:
P
ro
c
e
e
d
i
n
g
s o
f
th
e
T
e
n
th
C
o
n
f
e
re
n
c
e
o
f
th
e
Eu
ro
p
e
a
n
.
[2
0
]
Ra
fa
lo
v
it
c
h
,
R
Da
le
.
Un
i
ted
n
a
ti
o
n
s
g
e
n
e
ra
l
a
ss
e
mb
ly
re
so
lu
ti
o
n
s
:
A
six
-
la
n
g
u
a
g
e
p
a
ra
ll
e
l
c
o
rp
u
s
.
P
ro
c
e
e
d
in
g
s
o
f
th
e
M
T
S
u
m
m
it
X
III,
2
0
0
9
:
2
9
2
–
2
9
9
.
[2
1
]
L
a
ra
sa
ti
,
S
D
Ku
b
o
ň
,
V
.
A
S
t
u
d
y
o
f
I
n
d
o
n
e
sia
n
-
to
-
M
a
l
a
y
sia
n
M
T
S
y
ste
m
.
In
:
P
ro
c
e
e
d
i
n
g
s
o
f
th
e
4
th
In
tern
a
ti
o
n
a
l
M
AL
IND
O
W
o
rk
sh
o
p
,
Ja
k
a
rta (2
0
1
0
).
[2
2
]
Ra
n
g
e
lo
v
,
T
.
Ru
le
-
b
a
se
d
M
a
c
h
in
e
T
ra
n
sl
a
ti
o
n
b
e
twee
n
Bu
l
g
a
ri
a
n
a
n
d
M
a
c
e
d
o
n
ia
n
.
I
n
:
P
ro
c
e
e
d
in
g
s
o
f
th
e
S
e
c
o
n
d
In
tern
a
ti
o
n
a
l
W
o
rk
sh
o
p
o
n
F
re
e
/
Op
e
n
-
S
o
u
rc
e
Ru
le
-
Ba
se
d
M
a
c
h
i
n
e
T
ra
n
sla
ti
o
n
,
2
0
1
1
:
53
-
59
.
[2
3
]
Ba
rk
a
d
e
,
V
M
De
v
a
le,
P
R.
E
n
g
li
sh
t
o
S
a
n
k
rit
M
a
c
h
i
n
e
T
ra
n
sla
ti
o
n
S
e
m
a
n
ti
c
M
a
p
p
e
r.
I
n
:
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
En
g
i
n
e
e
rin
g
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
.
2
0
1
0
;
2
(
1
0
):
5
3
13
-
5
3
1
8
.
[2
4
]
Ch
risti
a
n
F
e
d
e
rm
a
n
n
,
Hy
b
rid
M
a
c
h
in
e
T
r
a
n
sl
a
ti
o
n
Us
in
g
J
o
in
t
,
B
in
a
rise
d
Fea
t
u
re
Vec
to
rs
.
In
P
r
o
c
e
e
d
in
g
s
o
f
th
e
T
e
n
th
Bien
n
ial
C
o
n
f
e
re
n
c
e
o
f
th
e
A
s
so
c
iatio
n
f
o
r
M
a
c
h
i
n
e
T
ra
n
sl
a
ti
o
n
in
t
h
e
Am
e
rica
s,
2
0
1
2
:
1
1
3
-
1
1
8
.
[2
5
]
Ei
se
le,
A
F
e
d
e
r
m
a
n
n
,
C
S
a
in
t
-
A
m
a
n
d
,
H
Je
ll
in
g
h
a
u
s,
M
He
rrm
a
n
n
,
T
Ch
e
n
,
Y.
Us
in
g
mo
se
s
to
i
n
t
e
g
ra
te
mu
lt
i
p
le
ru
le
-
b
a
se
d
m
a
c
h
i
n
e
tra
n
sl
a
ti
o
n
e
n
g
i
n
e
s
in
t
o
a
h
y
b
rid
sy
ste
m.
In
:
P
ro
c
e
e
d
in
g
s
o
f
th
e
T
h
ird
W
o
rk
sh
o
p
o
n
S
tatisti
c
a
l
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
,
Co
l
u
m
b
u
s,
Oh
io
,
A
CL
.
2
0
0
8
:
1
7
9
–
1
8
2
.
[2
6
]
Ty
e
rs,
F
M
.
R
u
le
-
b
a
se
d
a
u
g
me
n
ta
ti
o
n
o
f
tra
in
in
g
d
a
t
a
i
n
Bre
t
o
n
–
Fre
n
c
h
st
a
ti
stica
l
ma
c
h
in
e
tra
n
sla
ti
o
n
.
I
n
P
r
o
c
e
e
d
in
g
s o
f
th
e
1
3
t
h
A
n
n
u
a
l
C
o
n
f
e
re
n
c
e
o
f
th
e
Eu
ro
p
e
a
n
A
ss
o
c
iatio
n
o
f
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
,
2
0
0
9
:
2
1
3
–
2
1
7
.
[2
7
]
S
c
h
w
e
n
k
,
H
A
b
d
u
l
-
Ra
u
f
,
S
Ba
rra
u
lt
,
L
S
e
n
e
ll
a
rt,
J.
S
M
T
a
n
d
S
P
E
M
a
c
h
in
e
T
ra
n
sl
a
ti
o
n
S
y
ste
ms
fo
r
W
M
T
’0
9
.
In
P
r
o
c
e
e
d
in
g
s
o
f
th
e
F
o
u
rth
W
o
rk
sh
o
p
o
n
S
tatisti
c
a
l
M
a
c
h
i
n
e
T
ra
n
s
latio
n
,
S
tatM
T
’0
9
,
2
0
0
9
:
1
3
0
–
1
3
4
,
S
tr
o
u
d
sb
u
rg
,
P
A
,
USA
.
A
ss
o
c
iatio
n
f
o
r
Co
m
p
u
tatio
n
a
l
L
in
g
u
isti
c
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
R
u
le
B
a
s
ed
a
n
d
E
x
p
ec
ta
tio
n
Ma
ximiza
tio
n
a
lg
o
r
ith
fo
r
A
r
a
b
ic
-
E
n
g
lis
h
Hyb
r
id
.
.
.
.
(
A
r
w
a
A
lq
u
d
s
i
)
79
[2
8
]
S
a
n
c
h
e
z
-
Ca
rtag
e
n
a
,
VM
,
A
n
c
h
e
z
-
M
a
rt,
F
S
,
P
e
re
z
-
Ortiz,
JA
.
In
teg
ra
ti
n
g
sh
a
ll
o
w
-
´
tra
n
sfe
r ru
les
in
t
o
p
h
r
a
se
-
b
a
se
d
sta
ti
stica
l
m
a
c
h
i
n
e
tra
n
sla
ti
o
n
.
I
n
P
ro
c
e
e
d
i
n
g
s
o
f
th
e
X
III
M
a
c
h
i
n
e
T
ra
n
sla
ti
o
n
S
u
m
m
it
,
2
0
1
1
:
5
6
2
–
5
6
9
,
X
iam
e
n
,
Ch
in
a
,
S
e
p
tem
b
e
r.
[2
9
]
Xu
a
n
,
H
,
L
i,
W
,
T
a
n
g
,
G
.
A
n
A
d
v
a
n
c
e
d
Re
v
ie
w
o
f
H
y
b
rid
M
a
c
h
in
e
T
ra
n
sla
ti
o
n
(HMT
).
P
r
o
c
e
d
ia
En
g
in
e
e
rin
g
,
2
0
1
2
;
2
9
:
3
0
1
7
-
3
0
2
2
.
[3
0
]
P
a
p
in
e
n
i,
KA
,
Ro
u
k
o
s,
S
,
War
d
,
T
,
Zh
u
,
W
J.
2
0
0
1
.
BL
EU:
A
m
e
th
o
d
f
o
r
a
u
to
m
a
ti
c
e
v
a
lu
a
ti
o
n
o
f
m
a
c
h
in
e
tran
sla
ti
o
n
.
T
e
c
h
n
ica
l
Re
p
o
rt
R
C2
2
1
7
6
(W
0
1
0
9
–
0
2
2
),
IBM
Re
se
a
rc
h
Div
isio
n
,
T
h
o
m
a
s
J.
W
a
tso
n
Re
se
a
rc
h
Ce
n
ter.
[3
1
]
Ba
n
g
a
lo
re
,
P
Ha
f
f
n
e
r,
S
Ka
n
th
a
k
.
S
tatisti
c
a
l
m
a
c
h
in
e
tran
sla
ti
o
n
th
ro
u
g
h
g
lo
b
a
l
lex
ica
l
se
lec
ti
o
n
a
n
d
se
n
ten
c
e
re
c
o
n
stru
c
ti
o
n
.
I
n
4
5
t
h
A
n
n
u
a
l
M
e
e
ti
n
g
o
f
th
e
A
ss
o
c
iatio
n
o
f
Co
m
p
u
tatio
n
a
l
L
in
g
u
ist
ics
,
2
0
0
7
:
152
–
1
5
9
,
P
ra
g
u
e
,
Cz
e
c
h
Re
p
u
b
li
c
,
Ju
n
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.