I
nte
rna
t
io
na
l
J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
,
p
p
.
3
3
0
7
~
3
3
1
3
I
SS
N:
2088
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
9
i
4
.
p
p
3
3
0
7
-
3313
3307
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e
.
co
m/
jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JE
C
E
O
pinio
n
m
i
ni
ng
u
sing
co
m
bina
t
io
na
l appro
a
ch
for dif
fere
n
t
do
ma
ins
J
y
o
t
i D
esh
m
u
kh
1
,
A
m
iy
a
K
um
a
r
T
ripa
t
hy
2
,
Dilendra
H
ir
a
n
3
1
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter E
n
g
in
e
e
rin
g
,
P
A
HER
Un
iv
e
rsit
y
,
In
d
ia
2
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter E
n
g
in
e
e
rin
g
,
Un
iv
e
rsity
o
f
M
u
m
b
a
i
,
In
d
i
a
3
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter A
p
p
li
c
a
ti
o
n
s,
P
A
HER
Un
iv
e
rsity
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
24
,
2
0
1
8
R
ev
i
s
ed
Mar
27
,
2
0
1
9
A
cc
ep
ted
A
p
r
8
,
2
0
1
9
A
n
in
c
re
a
se
in
u
se
o
f
w
e
b
p
ro
d
u
c
e
s
larg
e
c
o
n
ten
t
o
f
i
n
f
o
rm
a
ti
o
n
a
b
o
u
t
p
ro
d
u
c
ts.
On
li
n
e
re
v
iew
s
a
r
e
u
se
d
to
m
a
k
e
d
e
c
isio
n
b
y
p
e
o
p
les
.
Op
in
io
n
m
in
in
g
is
v
a
st
re
se
a
rc
h
a
re
a
in
wh
ich
d
if
f
e
re
n
t
t
y
p
e
s
o
f
re
v
ie
w
s
a
r
e
a
n
a
l
y
z
e
d
.
S
e
v
e
ra
l
issu
e
s
a
re
e
x
isti
n
g
in
th
is
a
re
a
.
Do
m
a
in
a
d
a
p
tatio
n
is
e
m
e
r
g
in
g
issu
e
in
o
p
i
n
i
o
n
m
in
in
g
.
L
a
b
li
n
g
o
f
d
a
ta
f
o
r
e
v
e
r
y
d
o
m
a
in
is
ti
m
e
c
o
n
su
m
in
g
a
n
d
c
o
stly
tas
k
.
He
n
c
e
th
e
n
e
e
d
a
rise
s
f
o
r
m
o
d
e
l
th
a
t
train
o
n
e
d
o
m
a
in
a
n
d
a
p
p
li
e
d
it
o
n
o
t
h
e
r
d
o
m
a
in
re
d
u
c
in
g
c
o
st
a
sw
e
ll
a
s
ti
m
e
.
T
h
is
is
c
a
ll
e
d
d
o
m
a
in
a
d
a
p
tatio
n
w
h
ich
is
a
d
d
re
ss
e
d
in
th
is
p
a
p
e
r.
Us
in
g
m
a
x
i
m
u
m
e
n
tro
p
y
a
n
d
c
lu
ste
rin
g
tec
h
n
iq
u
e
so
u
rc
e
d
o
m
a
in
s
d
a
ta
is
train
e
d
.
T
ra
in
e
d
d
a
ta
f
ro
m
so
u
rc
e
d
o
m
a
in
is
a
p
p
li
e
d
o
n
targ
e
t
d
a
ta
to
lab
e
li
n
g
p
u
r
p
o
se
A
re
s
u
lt
sh
o
w
s
m
o
d
e
ra
te
a
c
c
u
ra
c
y
f
o
r
5
f
o
ld
c
r
o
ss
v
a
li
d
a
ti
o
n
a
n
d
c
o
m
b
in
a
ti
o
n
o
f
so
u
rc
e
d
o
m
a
in
s f
o
r
Bli
tze
r
e
t
a
l
(2
0
0
7
)
m
u
lt
i
d
o
m
a
in
p
ro
d
u
c
t
d
a
tas
e
t.
K
ey
w
o
r
d
s
:
I
n
f
o
r
m
a
tio
n
s
y
s
te
m
s
Kn
o
w
led
g
e
d
is
co
v
er
y
Ma
ch
i
n
e
lear
n
i
n
g
Op
in
io
n
m
i
n
in
g
T
ex
t m
in
i
n
g
Co
p
y
rig
h
t
©
2
0
1
9
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
:
J
y
o
ti De
s
h
m
u
k
h
,
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
E
n
g
i
n
ee
r
in
g
,
P
A
HE
R
U
n
i
v
er
s
it
y
,
Ud
aip
u
r
,
R
aj
s
th
a
n
,
I
n
d
ia.
E
m
ail:
j
y
o
j
a2
0
0
7
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Op
in
io
n
m
i
n
in
g
i
s
e
m
er
g
in
g
ar
ea
o
f
r
esear
ch
as
p
o
p
u
lar
ity
&
a
v
ailab
ilit
y
o
f
r
e
v
ie
w
s
i
n
cr
ea
s
i
n
g
.
Op
in
io
n
m
in
i
n
g
is
u
s
ed
to
d
eter
m
i
n
e
th
e
p
o
lar
it
y
o
f
a
tex
t
s
u
c
h
a
s
p
o
s
itiv
e,
n
e
g
ati
v
e
o
r
n
eu
tr
al
.
Op
in
io
n
r
ep
r
esen
ts
t
h
e
i
n
d
iv
id
u
al
’
s
id
ea
s
,
j
u
d
g
m
e
n
t
s
,
as
s
es
s
m
e
n
t
s
,
b
elief
s
ab
o
u
t
s
p
ec
if
ic
to
p
ic.
E
x
tr
ac
tio
n
o
f
o
p
in
io
n
o
r
s
en
ti
m
e
n
t
i
s
v
er
y
i
m
p
o
r
tan
t
tas
k
i
n
b
u
s
i
n
es
s
a
n
d
ac
ad
e
m
i
c
w
o
r
ld
.
E
v
er
y
m
a
n
u
f
ac
t
u
r
er
w
a
n
ted
to
k
n
o
w
t
h
e
r
ev
ie
w
ab
o
u
t
th
e
ir
p
r
o
d
u
cts
[
1
]
.
Dec
is
io
n
m
a
k
i
n
g
a
f
f
ec
ts
b
ec
au
s
e
in
d
i
v
id
u
a
ls
ar
e
r
el
y
o
n
w
h
at
o
t
h
er
s
ar
e
th
i
n
k
i
n
g
.
T
h
er
ef
o
r
e,
s
en
ti
m
e
n
t
an
al
y
s
is
i
s
p
o
p
u
lar
s
tr
ea
m
,
wh
ich
e
x
tr
ac
ts
s
en
ti
m
en
t
s
an
d
a
n
al
y
ze
i
t [
2
-
3
]
.
C
las
s
i
f
icatio
n
o
f
d
o
cu
m
en
t
s
d
ep
en
d
in
g
o
n
p
o
lar
it
y
i
s
a
k
e
y
ac
tiv
it
y
i
n
o
p
in
io
n
m
in
i
n
g
.
Do
cu
m
e
n
t
s
ar
e
w
r
it
ten
in
p
o
s
iti
v
e,
n
eg
at
iv
e
o
r
i
n
n
e
u
tr
al
o
r
ien
ta
tio
n
.
W
o
r
d
s
p
o
lar
ity
is
i
m
p
o
ta
n
t
f
ea
t
u
r
e
i
n
o
p
in
io
n
m
i
n
in
g
.
A
s
w
o
r
d
s
ar
e
d
o
m
a
in
d
ep
en
d
en
t,
th
e
k
n
o
w
led
g
e
o
f
d
o
m
ai
n
is
r
eq
u
ir
ed
to
f
i
n
d
o
u
t
w
h
e
th
er
w
o
r
d
is
p
o
s
itiv
e
o
r
n
eg
ati
v
e.
R
eu
s
a
b
ilit
y
o
f
k
n
o
w
led
g
e
o
f
o
n
e
d
o
m
a
in
in
an
o
t
h
er
d
o
m
ai
n
i
s
k
e
y
is
s
u
e
i
n
o
p
in
io
n
m
i
n
i
n
g
.
Do
a
m
in
ad
ap
tatio
n
ca
n
b
e
t
h
e
s
o
lu
tio
n
f
o
r
th
i
s
p
r
o
b
le
m
i
n
wh
ich
k
n
o
w
led
g
e
ca
n
b
e
tr
an
s
f
e
r
r
ed
f
r
o
m
o
n
e
d
o
m
a
in
to
an
o
t
h
er
r
ed
u
ci
n
g
co
s
t
in
ter
m
s
o
f
ti
m
e
an
d
m
o
n
e
y
.
C
o
n
s
id
er
th
e
p
r
o
b
le
m
,
w
h
er
e
task
i
s
to
au
to
m
at
icall
y
cla
s
s
i
f
y
t
h
e
r
ev
i
e
w
s
o
n
E
lectr
o
n
ics
d
o
m
ain
i
n
t
o
p
o
s
itiv
e
a
n
d
n
eg
ati
v
e
o
r
ien
t
atio
n
.
Fo
r
th
is
task
,
f
ir
s
t
w
e
h
a
v
e
to
co
llect
m
a
n
y
r
e
v
ie
w
s
o
f
t
h
e
d
o
m
ai
n
.
T
h
en
tr
ai
n
a
clas
s
if
ier
o
n
th
e
r
ev
ie
w
s
w
i
th
t
h
eir
s
u
b
s
eq
u
en
t
lab
els.
L
ar
g
e
a
m
o
u
n
t
o
f
r
e
v
ie
w
s
ar
e
n
ee
d
ed
to
m
ai
n
tai
n
g
o
o
d
class
i
f
ica
tio
n
p
er
f
o
r
m
an
ce
.
L
ab
eli
n
g
r
ev
ie
w
s
f
o
r
ea
ch
d
o
m
ai
n
i
s
ti
m
e
co
n
s
u
m
in
g
a
s
w
e
ll a
s
e
x
p
en
s
i
v
e
p
r
o
ce
s
s
.
He
n
ce
,
d
o
m
ai
n
ad
ap
tatio
n
n
ee
d
ar
is
es
w
h
ic
h
ca
n
u
s
es
k
n
o
w
led
g
e
o
f
o
n
e
d
o
m
ai
n
to
an
o
th
er
o
n
e
[
4
]
.
Stru
ctu
r
al
co
r
r
esp
o
n
d
en
ce
lear
n
i
n
g
(
SC
L
)
p
r
o
p
o
s
ed
to
ex
ten
d
s
tr
u
ctu
r
al
lear
n
i
n
g
.
SC
L
d
e
f
in
e
s
p
iv
o
t
f
ea
t
u
r
es,
w
h
ic
h
ar
e
co
m
m
o
n
to
b
o
th
s
o
u
r
ce
an
d
tar
g
et
d
o
m
ai
n
.
T
h
is
m
et
h
o
d
tr
ies to
f
in
d
th
e
co
r
r
elatio
n
b
et
w
ee
n
p
iv
o
t
f
ea
t
u
r
es,
an
d
n
o
n
-
p
iv
o
t
f
ea
tu
r
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
:
3
3
0
7
-
3313
3308
W
h
iteh
ea
d
et
al.
,
[
5
]
p
r
o
p
o
s
ed
a
m
et
h
o
d
f
o
r
b
u
ild
i
n
g
e
n
s
e
m
b
le
m
o
d
el
s
,
u
s
i
n
g
lex
ico
n
s
i
m
ilar
it
y
,
t
h
at
y
ie
ld
a
h
ig
h
class
i
f
icat
io
n
ac
cu
r
ac
y
f
o
r
d
o
m
ai
n
s
i
n
w
h
i
ch
n
o
tr
ain
i
n
g
w
a
s
p
er
f
o
r
m
ed
.
I
t
is
r
ep
o
r
ted
th
at
an
ad
j
u
s
ted
f
o
r
m
o
f
co
s
in
e
s
i
m
ilar
it
y
b
et
w
ee
n
d
o
m
a
in
le
x
i
co
n
s
ca
n
b
e
u
s
ed
to
p
r
ed
ict
w
h
ich
m
o
d
els
w
i
ll
b
e
ef
f
ec
tiv
e
i
n
a
n
e
w
tar
g
et
d
o
m
ain
.
J
iali
n
et
a
l.,
[
6
]
p
r
o
p
o
s
ed
a
g
e
n
er
al
f
r
a
m
e
w
o
r
k
f
o
r
cr
o
s
s
-
d
o
m
ai
n
s
en
t
i
m
e
n
t
class
i
f
icatio
n
.
Sp
ec
tr
al
f
ea
t
u
r
e
alig
n
m
e
n
t
(
S
F
A
)
cr
ea
tes
m
ea
n
in
g
f
u
l
clu
s
ter
s
w
i
th
t
h
e
h
elp
o
f
co
m
m
o
n
w
o
r
d
s
.
B
ip
ar
tite
g
r
ap
h
is
co
n
s
tr
u
cted
b
et
w
ee
n
co
m
m
o
n
o
r
d
o
m
a
in
i
n
d
ep
en
d
en
t
a
n
d
u
n
co
m
m
o
n
wo
r
d
s
o
f
b
o
th
s
o
u
r
ce
an
d
tar
g
et
d
o
m
ai
n
s
.
M
u
t
u
al
i
n
f
o
r
m
atio
n
u
s
ed
to
s
elec
t
co
m
m
o
n
w
o
r
d
s
an
d
b
in
ar
y
clas
s
if
ier
i
s
tr
ain
ed
f
o
r
class
i
f
icatio
n
.
E
x
p
er
i
m
e
n
tal
r
e
s
u
lt
s
s
h
o
w
s
ef
f
ec
ti
v
e
p
er
f
o
r
m
an
ce
o
f
ap
p
r
o
ac
h
o
n
b
o
th
d
o
c
u
m
e
n
t
a
n
d
s
en
te
n
ce
lev
el
cla
s
s
i
f
ica
tio
n
.
L
i
u
et
al.
,
[
7
]
p
r
o
p
o
s
es
a
m
et
h
o
d
f
o
r
co
-
ex
tr
ac
tin
g
o
p
in
io
n
tar
g
ets
an
d
o
p
in
io
n
w
o
r
d
s
b
y
u
s
i
n
g
a
w
o
r
d
ali
g
n
m
e
n
t
m
o
d
el.
Ma
in
f
o
c
u
s
i
s
o
n
d
etec
ti
n
g
o
p
in
io
n
r
elatio
n
s
b
et
w
ee
n
o
p
in
io
n
tar
g
et
s
an
d
o
p
in
io
n
w
o
r
d
s
.
A
s
co
m
p
ar
ed
to
p
r
ev
io
u
s
m
et
h
o
d
s
b
ased
o
n
n
ea
r
est
n
eig
h
b
o
r
r
u
les
an
d
s
y
n
tac
tic
p
atter
n
s
,
p
r
o
p
o
s
ed
m
et
h
o
d
ca
p
tu
r
es
o
p
in
io
n
r
elatio
n
s
m
o
r
e
p
r
ec
is
el
y
.
An
Op
in
io
n
R
elatio
n
Gr
ap
h
is
co
n
s
tr
u
cted
to
m
o
d
el
all
ca
n
d
id
ates
alo
n
g
w
i
th
a
g
r
ap
h
co
-
r
a
n
k
in
g
alg
o
r
it
h
m
to
esti
m
ate
th
e
co
n
f
id
e
n
ce
o
f
ea
ch
ca
n
d
id
ate.
T
h
e
ite
m
s
w
it
h
h
i
g
h
er
r
an
k
s
ar
e
ex
tr
ac
ted
o
u
t.
T
h
e
e
x
p
er
i
m
en
tal
r
es
u
lt
s
f
o
r
th
r
ee
d
atasets
w
it
h
d
i
f
f
er
e
n
t
lan
g
u
a
g
es a
n
d
d
if
f
er
en
t size
s
p
r
o
v
e
th
e
ef
f
ec
ti
v
e
n
es
s
o
f
t
h
e
p
r
o
p
o
s
ed
m
e
th
o
d
.
I
n
f
u
t
u
r
e
wo
r
k
,
au
th
o
r
s
p
la
n
n
ed
to
co
n
s
id
er
ad
d
itio
n
al
ty
p
e
s
o
f
r
elatio
n
s
b
et
w
e
e
n
w
o
r
d
s
,
s
u
c
h
as
to
p
ical
r
elatio
n
s
,
in
Op
i
n
i
o
n
R
elatio
n
Gr
ap
h
.
B
ala
m
u
r
ali
A
.
R
.
et
al.
,
[
8
]
p
r
o
p
o
s
es
ap
p
r
o
ac
h
f
o
r
cr
o
s
s
d
o
m
a
in
s
e
n
ti
m
en
t
ta
g
g
in
g
.
A
m
et
h
o
d
f
o
r
cr
ea
tin
g
h
ig
h
i
n
-
d
o
m
ai
n
class
if
ier
u
s
i
n
g
s
i
m
p
le
lo
w
lev
el
f
ea
t
u
r
es is
i
n
tr
o
d
u
ce
d
.
A
g
en
er
ic
cla
s
s
i
f
ier
b
ased
o
n
m
eta
-
clas
s
i
f
icatio
n
ap
p
r
o
ac
h
co
u
p
led
w
it
h
t
h
is
h
ig
h
in
-
d
o
m
a
in
class
i
f
ier
i
s
u
s
ed
to
cr
ea
te
lab
eled
d
ata
f
o
r
a
n
e
w
d
o
m
ai
n
f
r
o
m
d
o
m
ai
n
s
h
a
v
i
n
g
lab
eled
d
ata.
R
es
u
lt
s
s
h
o
w
ed
co
n
s
id
er
ab
le
i
m
p
r
o
v
e
m
e
n
t
i
n
cr
o
s
s
d
o
m
ai
n
s
en
t
i
m
e
n
t
ta
g
g
in
g
ac
c
u
r
ac
y
i
f
d
o
m
a
in
s
ar
e
s
i
m
ilar
.
I
n
ca
s
e
o
f
d
is
s
i
m
ilar
d
o
m
ain
s
s
y
s
te
m
e
x
ce
ed
s
th
e
b
aseli
n
e
ac
cu
r
ac
ies
b
y
s
u
b
s
ta
n
tial
m
ar
g
in
s
.
B
o
lleg
ala
D.
et
al.
,
[
9
]
p
r
o
p
o
s
ed
a
m
et
h
o
d
cr
ea
tin
g
t
h
esau
r
u
s
w
h
ic
h
i
s
r
ec
ep
ti
v
e
to
s
en
ti
m
e
n
t
w
o
r
d
s
f
r
o
m
d
if
f
er
e
n
t
d
o
m
ai
n
s
.
Au
th
o
r
u
s
ed
b
o
th
lab
eled
a
n
d
u
n
lab
el
ed
d
ata.
C
r
ea
ted
lex
ico
n
v
o
ca
b
u
lar
y
w
a
s
ex
p
a
n
d
ed
at
tr
ain
an
d
test
ti
m
e
s
i
n
a
class
i
f
ier
.
P
r
o
p
o
s
ed
m
eth
o
d
co
m
p
ar
ed
w
i
th
m
an
y
b
ase
li
n
e
m
et
h
o
d
s
w
h
ic
h
r
ev
ea
l
a
g
o
o
d
p
er
f
o
r
m
an
ce
.
Sh
o
u
s
h
a
n
et
al.
,
[
1
0
]
p
r
o
p
o
s
ed
ac
tiv
e
lear
n
in
g
in
w
h
ic
h
s
o
u
r
c
e
an
d
tar
g
et
cla
s
s
i
f
ier
s
ar
e
tr
ain
ed
s
ep
ar
atel
y
.
Usi
n
g
Q
u
er
y
by
C
o
m
m
i
ttee
(
QB
C
)
s
elec
tio
n
s
tr
ateg
y
in
f
o
r
m
at
iv
e
s
a
m
p
les
ar
e
s
elec
te
d
an
d
class
if
icatio
n
d
ec
is
io
n
m
ad
e
b
y
co
m
b
i
n
in
g
cla
s
s
i
f
ier
s
.
L
ab
el
p
r
o
p
ag
atio
n
i
s
u
s
ed
to
tr
ain
b
o
t
h
class
i
f
ier
s
.
R
es
u
l
t
d
em
o
n
s
tr
ate
s
s
i
g
n
i
f
ican
tl
y
o
u
tp
er
f
o
r
m
s
t
h
a
n
th
e
b
a
s
elin
e
m
et
h
o
d
s
.
L
i
k
e
e
n
s
e
m
b
le
clas
s
if
ier
s
g
r
ap
h
b
ased
m
et
h
o
d
o
lo
g
y
also
u
s
ed
f
o
r
d
o
m
a
in
ad
ap
tatio
n
.
I
n
d
er
j
it
S.
et
al.
,
[
1
1
]
p
r
o
p
o
s
ed
th
e
g
r
ap
h
b
ased
d
o
m
ai
n
ad
ap
tatio
n
m
et
h
o
d
.
Si
m
ilar
it
y
g
r
ap
h
co
n
s
tr
u
cted
b
et
w
ee
n
f
ea
tu
r
es
f
r
o
m
all
d
o
m
a
in
s
,
i
f
th
ese
f
ea
t
u
r
es
ar
e
s
i
m
ilar
t
h
en
ed
g
e
ex
i
s
t
b
et
wee
n
t
h
e
m
.
A
ll
lab
eled
f
ea
tu
r
es
u
s
ed
i
n
m
etr
ic
-
lear
n
in
g
a
l
g
o
r
ith
m
s
.
Gr
ap
h
i
s
co
n
s
tr
u
cted
u
s
in
g
d
ata
-
d
ep
en
d
en
t
m
etr
ic
a
n
d
w
ei
g
h
t
is
ca
lcu
lated
f
o
r
ea
ch
ed
g
e.
An
e
x
p
er
i
m
e
n
tal
r
es
u
lt
d
em
o
n
s
tr
ate
s
r
ed
u
ct
io
n
in
c
lass
i
f
icatio
n
er
r
o
r
.
S.
B
h
at
t
et
a
l.,
[
12
]
p
r
o
p
o
s
es
an
alg
o
r
ith
m
to
ad
ap
t
class
i
f
icatio
n
m
o
d
el
b
y
iter
ati
v
el
y
lear
n
i
n
g
d
o
m
ai
n
s
p
ec
i
f
ic
f
ea
tu
r
es
f
r
o
m
t
h
e
u
n
lab
eled
test
d
ata.
Mo
r
eo
v
er
,
th
is
ad
ap
tatio
n
tr
an
s
p
ir
es
i
n
a
s
i
m
ilar
it
y
a
w
ar
e
m
a
n
n
er
b
y
i
n
te
g
r
atin
g
s
i
m
ilar
it
y
b
et
wee
n
d
o
m
ai
n
s
i
n
t
h
e
ad
ap
tatio
n
s
etti
n
g
.
C
r
o
s
s
-
d
o
m
ai
n
clas
s
i
f
icatio
n
ex
p
er
i
m
e
n
ts
o
n
d
i
f
f
er
e
n
t
d
ataset
s
,
i
n
c
lu
d
in
g
a
r
ea
l
w
o
r
ld
d
ataset,
d
e
m
o
n
s
tr
ate
e
f
f
icac
y
o
f
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
o
v
e
r
s
tate
-
of
-
t
h
e
ar
t.
Ma
n
y
o
p
en
s
o
u
r
ce
le
x
ico
n
s
ar
e
av
ailab
le
w
h
ic
h
s
er
v
e
a
s
a
d
atab
ase
f
o
r
ex
tr
ac
ti
n
g
t
h
e
p
o
lar
it
y
v
a
lu
e
s
o
f
o
p
in
io
n
w
o
r
d
s
.
Ho
w
ev
er
,
t
h
ese
g
e
n
er
ic
p
o
lar
it
y
lex
ico
n
s
r
ev
ea
l
t
h
e
g
e
n
er
al
s
e
n
ti
m
e
n
t
o
f
o
p
in
io
n
w
o
r
d
s
.
An
o
p
in
io
n
w
o
r
d
co
u
ld
b
e
c
o
n
tex
t
d
ep
en
d
en
t
o
r
d
o
m
ai
n
s
p
ec
if
ic.
T
h
e
w
o
r
d
li
k
e
“
s
m
a
ll”
m
a
y
r
ep
r
ese
n
t
a
n
eg
at
iv
e
o
r
ie
n
tatio
n
in
a
h
o
tel
d
o
m
ai
n
b
u
t
if
u
s
ed
in
m
o
b
ile
ap
p
licatio
n
s
it
i
s
a
p
o
s
iti
v
e.
Sa
m
e
w
a
y
“
f
r
ee
zi
n
g
”
is
g
o
o
d
f
o
r
a
r
ef
r
ig
er
ato
r
b
u
t
n
eg
at
iv
e
f
o
r
s
o
f
t
w
ar
e
ap
p
licat
io
n
s
.
T
h
e
v
ar
iatio
n
o
f
o
p
in
io
n
p
o
s
s
ess
b
y
a
s
a
m
e
w
o
r
d
i
n
d
i
f
f
er
e
n
t
d
o
m
a
in
s
r
est
r
icts
t
h
e
u
s
a
g
e
o
f
g
e
n
er
ic
le
x
i
co
n
s
a
s
it
co
n
tai
n
s
g
en
er
alize
p
o
lar
ity
o
f
a
w
o
r
d
.
So
n
ee
d
o
f
d
o
m
ai
n
ad
ap
tab
le
lex
ico
n
s
e
m
er
g
es.
Do
m
ai
n
a
d
ap
tab
ilit
y
is
m
aj
o
r
is
s
u
e
i
n
s
en
ti
m
e
n
t
a
n
al
y
s
i
s
w
h
ic
h
h
as
b
ee
n
ad
d
r
ess
e
d
in
p
r
o
p
o
s
ed
f
r
am
e
w
o
r
k
.
A
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
attem
p
ts
i
n
b
u
ild
in
g
a
class
if
ier
w
h
ic
h
u
s
es
m
a
x
i
m
u
m
en
tr
o
p
y
cla
s
s
i
f
ier
w
i
th
clu
s
ter
i
n
g
b
ased
o
n
p
o
in
t
w
is
e
m
u
tu
al
in
f
o
r
m
a
tio
n
b
et
w
ee
n
w
o
r
d
s
.
2.
P
RO
P
O
SE
D
M
E
T
H
O
D
Op
in
io
n
le
x
ico
n
is
a
w
o
r
d
o
r
g
r
o
u
p
o
f
w
o
r
d
s
in
r
ev
ie
w
.
I
d
en
tif
icatio
n
o
f
o
p
in
ian
ated
w
o
r
d
s
o
r
lex
ico
n
s
is
an
i
m
p
o
r
tan
t
ta
s
k
.
I
n
p
r
o
p
o
s
ed
m
et
h
o
d
d
if
f
er
e
n
t
t
ask
s
ar
e
d
is
cu
s
s
ed
.
Data
co
llectio
n
is
cr
u
cial
ta
s
k
as
lar
g
e
d
ata
i
s
a
v
aila
b
le
o
n
l
in
e.
Fo
r
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
Am
az
o
n
s
m
u
lti
p
r
o
d
u
ct
r
ev
ie
w
d
ata
s
et
is
u
s
ed
.
Af
ter
d
ata
co
llectio
n
clea
n
i
n
g
o
f
d
ata
is
n
ec
ess
ar
y
.
Sto
p
w
o
r
d
r
em
o
v
al
m
et
h
o
d
is
ap
p
lied
o
n
d
ata
c
o
llected
.
P
ar
t
o
f
s
p
ee
ch
ta
g
g
i
n
g
is
u
s
e
d
to
tag
w
o
r
d
s
a
s
ad
j
ec
tiv
e,
a
d
v
er
b
.
Af
ter
t
h
is
p
r
ep
r
o
ce
s
s
in
g
s
tep
class
if
ier
is
ap
p
lied
o
n
clea
n
ed
d
ata.
Ma
x
im
u
m
e
n
tr
o
p
y
clas
s
if
ier
is
u
s
ed
alo
n
g
w
it
h
cl
u
s
ter
i
n
g
as s
h
o
wn
in
Fig
u
r
e
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
Op
in
io
n
min
in
g
u
s
in
g
c
o
mb
in
a
tio
n
a
l a
p
p
r
o
a
c
h
fo
r
d
iffer
en
t d
o
ma
in
s
(
Jy
o
ti De
s
h
mu
kh
)
3309
Fig
u
r
e
1
.
Flo
w
o
f
p
r
o
p
o
s
ed
m
eth
o
d
I
n
p
u
t:
So
u
r
ce
an
d
tar
g
et
d
o
m
ain
s
p
r
ep
r
o
ce
s
s
ed
r
ev
ie
w
d
o
c
u
m
e
n
t
s
.
Fo
r
all
f
ea
t
u
r
es
ex
tr
a
cted
f
r
o
m
in
p
u
t d
atase
t
1.
I
n
itialize
1
n
1
t
o
i
f
o
r
0
i
(
w
h
er
e
n
is
to
tal
n
u
m
b
er
o
f
f
ea
t
u
r
es)
2.
R
ep
ea
t u
n
til co
n
v
er
g
e
n
ce
.
C
al
cu
late
t
h
e
p
r
o
b
ab
ilit
y
o
f
cla
s
s
c
'
e
x
p
(
,
)
(,
e
x
p
(
'
,
)
d
e
f
ii
i
ii
i
cC
f
c
d
P
c
d
f
c
d
…….
.
(
1
)
w
h
er
e,
i
i
(
is
c
a
l
c
ul
a
te
d
by
Ite
r
a
tive
s
c
a
l
in
g
a
l
gor
ithm
)
i
is
in
cr
e
m
en
t q
u
an
t
it
y
)
,
(
c
d
f
i
=
tf
-
id
f
va
lu
e
o
f
ea
c
h
w
o
r
d
co
n
s
id
er
ed
as
f
ea
t
u
r
e.
3.
Fo
r
all
class
i
f
ied
w
o
r
d
s
w
ei
g
h
t is ass
i
g
n
ed
u
s
in
g
P
o
in
t
w
is
e
Mu
t
u
al
I
n
f
o
r
m
atio
n
(
P
MI
)
w
i
t
h
(
2
)
(
2
)
4
.
C
o
m
m
o
n
a
n
d
u
n
co
m
m
o
n
w
o
r
d
s
f
r
o
m
s
o
u
r
ce
an
d
ta
r
g
et
d
o
m
ai
n
s
ar
e
cl
u
s
ter
ed
u
s
in
g
w
eig
h
t
v
al
u
e.
Ou
tp
u
t: p
o
s
itiv
e
a
n
d
n
e
g
ati
v
e
class
i
f
ied
d
o
cu
m
e
n
t
s
an
d
w
o
r
d
clu
s
ter
s
.
I
ter
ativ
e
Scali
n
g
Alg
o
r
it
h
m
:
I
n
p
u
t:
f
ea
tu
r
e
s
f
u
n
ctio
n
s
n
f
f
f
,.......
,
2
1
,
E
m
p
ir
ical
Dis
tr
ib
u
t
io
n
)
,
(
~
d
c
p
.
Ou
tp
u
t: Op
ti
m
a
l p
ar
a
m
eter
v
a
lu
es,
o
p
ti
m
a
l
m
o
d
el
p
*
.
1.
Star
t
w
ith
}
,.
...
,
2
,
1
{
0
n
i
a
l
l
f
or
i
2.
Do
f
o
r
ea
ch
}
,.
...,
2
,
1
{
n
i
a.
L
et
i
b
e
th
e
s
o
lu
tio
n
to
)
(
)
,
(
f
e
x
p(
)
,
(
)
|
(
)
(
~
#
~
,
i
i
i
d
c
f
p
d
c
d
c
f
c
d
p
c
p
w
h
er
e,
)
,
(
)
,
(
,
1
#
d
c
f
d
c
f
i
d
c
i
b.
Up
d
ate
th
e
v
al
u
e
o
f
i
ac
co
r
d
in
g
to
i
i
i
3.
Go
to
s
tep
2
if
n
o
t
all
th
e
i
h
a
v
e
co
n
v
er
g
ed
.
)
,
(
#
d
c
f
is
th
e
to
tal
n
u
m
b
er
o
f
f
ea
tu
r
es
w
h
ic
h
ar
e
ac
tiv
e
f
o
r
p
ar
ticu
lar
)
,
(
d
c
p
air
.
I
ter
ativ
e
s
ca
lin
g
alg
o
r
it
h
m
is
u
s
ed
to
ca
lcu
late
v
alu
e
[
1
3
]
.
)
)
2
(
)
1
(
)
2
1
(
(
2
l
o
g
)
2
,
1
(
W
o
r
d
p
W
o
r
d
p
W
o
r
d
W
o
r
d
p
W
o
r
d
W
o
r
d
P
M
I
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
:
3
3
0
7
-
3313
3310
3.
RE
SU
L
T
S
A
ND
AN
AL
Y
SI
S
B
litzer
et
al.
,
[
1
4
]
Mu
lti
Do
m
ain
d
ataset
is
u
s
ed
f
o
r
ev
alu
ati
o
n
o
f
p
r
o
p
o
s
ed
m
et
h
o
d
.
R
esu
lts
o
f
ea
ch
s
tep
r
ec
o
r
d
ed
in
f
o
llo
w
i
n
g
s
ec
tio
n
.
P
r
ep
r
o
ce
s
s
ed
d
ata
is
u
s
e
d
f
o
r
class
i
f
icat
io
n
a
n
d
clu
s
tei
n
g
p
r
o
ce
s
s
.
I
n
f
ir
s
t
ex
p
er
i
m
e
n
t
s
o
u
r
ce
d
o
m
ai
n
i
s
d
iv
id
ed
in
to
5
p
ar
ts
a
n
d
ta
r
g
e
t
d
o
m
ai
n
ta
k
e
n
as
it i
s
.
Seco
n
d
ex
p
er
i
m
e
n
t d
o
n
e
o
n
co
m
b
i
n
atio
n
o
f
s
o
u
r
ce
d
o
m
ai
n
s
.
T
h
e
Mu
lti
-
Do
m
ain
Se
n
ti
m
e
n
t
Data
s
et
co
n
tai
n
s
p
r
o
d
u
ct
r
e
v
ie
w
s
ta
k
en
f
r
o
m
Am
az
o
n
.
co
m
f
r
o
m
m
an
y
p
r
o
d
u
ct
d
o
m
ain
s
[
1
4
]
.
T
h
is
d
ataset
co
n
tai
n
s
t
h
r
ee
t
y
p
es
o
f
f
iles
p
o
s
iti
v
e,
n
e
g
a
ti
v
e
an
d
u
n
lab
eled
i
n
XM
L
f
o
r
m
at.
E
ac
h
l
in
e
i
n
f
o
r
m
o
f
:
f
ea
t
u
r
e:<
co
u
n
t>
.
.
.
.
f
ea
tu
r
e:<
c
o
u
n
t>
#
lab
el#
:<lab
el>
.
T
h
ese
f
iles
ar
e
e
x
tr
ac
ted
u
s
i
n
g
XM
L
f
i
le
s
p
litt
er
an
d
r
e
v
ie
w
s
ar
e
w
r
itte
n
i
n
to
tex
t
f
ile
as sh
o
w
n
i
n
Fi
g
u
r
e
2
.
Fig
u
r
e
2
.
XM
L
f
ile
s
p
litt
er
T
h
e
d
ataset
co
n
tain
s
1
0
0
0
p
o
s
iti
v
e
f
ile
s
an
d
1
0
0
0
n
eg
ati
v
e
f
iles
f
o
r
ea
ch
d
o
m
ai
n
.
On
t
h
is
d
ataset
p
r
ep
r
o
ce
s
s
in
g
s
tep
is
ap
p
lied
to
r
em
o
v
e
n
o
i
s
y
d
ata.
I
n
t
h
is
p
h
a
s
e,
p
r
e
-
p
r
o
ce
s
s
in
g
is
d
o
n
e
to
elim
i
n
ate
u
n
n
ec
es
s
ar
y
w
o
r
d
s
ca
lled
a
s
s
to
p
w
o
r
d
s
.
T
h
is
is
i
m
p
o
r
t
a
n
t
b
ec
au
s
e
th
e
ir
r
elev
a
n
t
d
ata
f
r
o
m
t
h
e
r
ev
ie
w
s
co
u
ld
b
e
elim
in
ated
.
T
h
is
eli
m
in
ate
s
th
e
p
r
o
ce
s
s
in
g
o
v
er
h
ea
d
s
o
f
a
lar
g
e
a
m
o
u
n
t
o
f
te
x
t
u
al
d
ata.
Mo
s
t
o
f
th
e
E
n
g
l
is
h
s
e
n
te
n
ce
s
in
cl
u
d
e
w
o
r
d
s
lik
e
“
a,
a
n
,
o
f
,
th
e,
I
,
i
t,
y
o
u
,
an
d
,
etc”
.
S
u
c
h
w
o
r
d
s
d
o
n
o
t
ca
r
r
y
p
ar
ticu
lar
m
ea
n
in
g
.
I
n
f
o
r
m
at
io
n
e
x
tr
ac
ti
o
n
f
r
o
m
n
at
u
r
al
la
n
g
u
a
g
e
ca
n
b
e
d
o
n
e
effec
tiv
el
y
a
n
d
clea
r
l
y
b
y
a
v
o
id
in
g
t
h
o
s
e
w
o
r
d
s
w
h
ich
o
cc
u
r
s
v
er
y
o
f
te
n
.
T
o
r
em
o
v
e
s
to
p
w
o
r
d
s
f
r
o
m
s
e
n
te
n
ce
s
te
x
t
file
i
s
u
s
ed
w
h
ic
h
co
n
s
is
t
s
o
f
l
is
t
o
f
E
n
g
lis
h
s
to
p
w
o
r
d
s
as
s
h
o
wn
in
Fig
u
r
e
3
.
Fig
u
r
e
3
.
Ou
tp
u
t o
f
d
ata
p
r
ep
r
o
ce
s
s
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
Op
in
io
n
min
in
g
u
s
in
g
c
o
mb
in
a
tio
n
a
l a
p
p
r
o
a
c
h
fo
r
d
iffer
en
t d
o
ma
in
s
(
Jy
o
ti De
s
h
mu
kh
)
3311
Af
ter
s
to
p
w
o
r
d
r
e
m
o
v
a
l,
u
s
in
g
p
ar
s
er
p
ar
t
o
f
s
p
ee
ch
o
f
s
en
te
n
ce
li
k
e
n
o
u
n
,
ad
j
ec
tiv
e,
ad
v
er
b
,
v
er
b
,
etc.
ar
e
ex
tr
ac
ted
.
P
a
r
s
in
g
is
v
ital
s
tep
as
it
g
iv
e
s
o
p
in
i
o
n
w
o
r
d
s
as
a
n
o
u
tp
u
t.
Sen
ten
ce
P
ar
s
in
g
i
n
v
o
l
v
es
ass
i
g
n
in
g
d
i
ff
er
en
t
p
ar
ts
o
f
s
p
ee
ch
tag
s
s
u
ch
a
s
n
o
u
n
,
p
r
ep
o
s
itio
n
,
v
er
b
,
ad
j
ec
tiv
e
an
d
ad
v
er
b
s
to
a
g
iv
en
tex
t
ar
e
k
n
o
w
n
a
s
P
ar
t
-
o
f
-
Sp
ee
c
h
(
P
OS)
tag
g
i
n
g
.
T
h
e
p
ar
t
-
of
-
s
p
ee
ch
is
a
ca
te
g
o
r
y
u
s
ed
in
lin
g
u
i
s
tic
s
th
at
is
d
efin
ed
b
y
a
s
y
n
tact
ic
o
r
m
o
r
p
h
o
lo
g
ical
b
eh
av
io
r
o
f
a
w
o
r
d
.
T
h
e
tr
a
d
itio
n
al
E
n
g
li
s
h
lan
g
u
a
g
e
g
r
a
m
m
ar
class
i
fi
e
s
P
OS
in
th
e
f
o
llo
w
in
g
ca
teg
o
r
ies:
v
er
b
,
n
o
u
n
,
ad
j
e
ctiv
e,
ad
v
er
b
,
p
r
o
n
o
u
n
,
p
r
ep
o
s
itio
n
,
co
n
j
u
n
c
tio
n
an
d
in
ter
j
ec
tio
n
.
T
h
e
r
ea
s
o
n
wh
y
P
OS
ta
g
g
in
g
i
s
s
o
i
m
p
er
ati
v
e
to
in
f
o
r
m
a
tio
n
ex
tr
ac
tio
n
is
th
e
f
ac
t
th
at,
ea
c
h
ca
teg
o
r
y
p
la
y
s
a
s
p
ec
i
fi
c
r
o
le
w
it
h
i
n
a
s
en
ten
ce
.
No
u
n
s
g
iv
e
n
a
m
es
to
o
b
j
ec
ts
,
o
r
en
titi
es
f
r
o
m
r
ev
ie
w
s
.
An
ad
j
ec
tiv
e
d
escr
ib
es
o
p
in
io
n
.
A
l
s
o
,
s
o
m
e
ad
v
er
b
s
ca
n
p
la
y
k
e
y
r
o
le
as
an
ad
j
ec
tiv
e.
Firstl
y
,
te
x
t
r
ev
ie
w
is
di
v
id
ed
in
to
s
en
te
n
ce
s
.
S
tan
f
o
r
d
p
a
r
s
er
is
u
s
ed
to
g
en
er
at
e
th
e
P
OS
tag
g
in
g
o
f
ea
c
h
w
o
r
d
p
r
esen
t
in
t
h
e
s
en
te
n
ce
.
I
t is v
er
y
ess
e
n
tia
l a
s
it h
elp
s
i
n
fi
n
d
i
n
g
g
e
n
er
al
lan
g
u
a
g
e
p
atter
n
s
as s
h
o
w
n
i
n
Fi
g
u
r
e
4
.
Fig
u
r
e
4
.
Ou
tp
u
t o
f
P
OS ta
g
g
i
n
g
Fro
m
ea
ch
d
o
m
a
i
n
1
0
0
0
r
ev
ie
w
f
ile
s
i.e
.
5
0
0
n
eg
ativ
e
a
n
d
5
0
0
p
o
s
itiv
e
f
iles
ar
e
tak
en
f
o
r
ex
p
er
i
m
e
n
t.
Usi
n
g
cr
o
s
s
v
al
id
atio
n
tec
h
n
i
q
u
e
So
u
r
ce
d
o
m
ai
n
i
s
d
iv
id
e
d
in
to
5
p
ar
ts
a
n
d
ea
c
h
p
ar
t
is
ta
k
en
a
s
i
n
p
u
t.
T
ar
g
et
d
o
m
ai
n
i
s
f
u
ll
y
ta
k
en
.
R
es
u
lt
s
f
r
o
m
ea
c
h
p
ar
t
ar
e
av
er
ag
e
d
.
T
ab
le
1
s
h
o
w
s
ac
cu
r
ac
y
f
r
o
m
s
a
m
e
tech
n
iq
u
e
f
o
r
1
2
class
if
icatio
n
task
s
.
T
ab
le
1.
A
n
al
y
s
i
s
o
f
A
cc
u
r
ac
y
f
r
o
m
cr
o
s
s
v
a
lid
atio
n
tec
h
n
iq
u
e
W
h
en
s
ize
o
f
tr
ain
in
g
d
ata
is
less
t
h
an
te
s
ti
n
g
d
ata
ac
cu
r
ac
y
v
al
u
es
ar
e
d
ec
r
ea
s
i
n
g
co
m
p
ar
e
to
f
u
l
l
tr
ain
i
n
g
d
ata.
A
cc
u
r
ac
y
g
etti
n
g
a
f
ter
d
i
v
id
in
g
tar
g
et
d
o
m
ain
in
to
p
ar
ts
an
d
tak
in
g
So
u
r
ce
d
o
m
ai
n
a
s
f
u
ll
f
o
r
ea
ch
clas
s
i
f
icatio
n
ta
s
k
is
a
v
e
r
ag
e
9
8
.
4
%.
Kitc
h
en
d
o
m
a
in
an
d
DVD
d
o
m
ai
n
ar
e
s
h
o
w
in
g
av
er
a
g
e
6
7
.
8
7
%
ac
cu
r
ac
y
.
Mo
s
t
o
f
th
e
w
o
r
d
s
f
r
o
m
t
h
ese
d
o
m
ai
n
s
ar
e
s
i
m
ilar
.
E
lectr
o
n
ics
an
d
DVD
d
o
m
ai
n
ar
e
s
h
o
w
in
g
les
s
ac
cu
r
ac
y
5
6
.
6
6
%.
T
r
ain
in
g
d
ata
is
r
an
d
o
m
l
y
d
iv
id
ed
.
Fo
ld
4
s
h
o
w
in
g
7
4
.
7
%
ac
cu
r
ac
y
f
o
r
B
o
o
k
to
E
lec
tr
o
n
ics
d
o
m
ai
n
w
h
er
ea
s
f
o
ld
5
s
h
o
w
s
7
3
.
4
%
ac
cu
r
ac
y
f
o
r
E
lectr
o
n
ics
to
Ki
tch
e
n
d
o
m
ai
n
as
s
h
o
w
n
i
n
S
o
u
r
c
e
→
T
a
r
g
e
t
A
c
c
u
r
a
c
y
(
%)
S
o
u
r
c
e
1
A
c
c
u
r
a
c
y
(
%)
S
o
u
r
c
e
2
A
c
c
u
r
a
c
y
(
%)
S
o
u
r
c
e
3
A
c
c
u
r
a
c
y
(
%)
S
o
u
r
c
e
4
A
c
c
u
r
a
c
y
(
%)
S
o
u
r
c
e
5
A
c
c
u
r
a
c
y
(
%)
A
v
e
r
a
g
e
B
→
D
6
1
.
0
6
4
.
7
6
0
.
9
6
4
.
6
6
8
.
2
6
3
.
8
8
B
→
E
6
4
.
0
6
3
.
6
6
4
.
1
7
4
.
7
6
5
.
2
6
6
.
3
2
B
→
K
6
1
.
0
6
0
.
5
6
0
.
5
6
5
.
1
0
6
7
.
6
0
6
2
.
9
4
D
→
B
6
9
.
5
6
8
.
6
0
6
3
.
6
6
1
.
3
6
3
.
8
0
6
5
.
3
6
D
→
E
6
9
.
5
6
8
.
6
6
4
.
0
5
8
.
9
6
5
.
9
6
5
.
3
8
D
→
K
7
1
.
8
7
0
.
7
6
1
.
9
6
2
.
0
6
6
.
6
6
6
.
6
E
→
B
6
3
.
4
5
0
.
3
6
1
.
0
5
2
.
7
6
1
.
7
5
7
.
6
6
E
→
D
6
1
.
0
5
0
.
2
5
9
.
3
5
4
.
9
5
7
.
9
5
6
.
6
6
E
→
K
6
7
.
1
5
0
.
5
6
7
.
6
6
4
.
1
7
3
.
4
6
4
.
5
4
K
→
B
5
8
.
4
6
7
.
8
6
0
.
4
5
2
.
7
5
2
.
0
5
8
.
2
6
K
→
D
6
8
.
6
7
1
.
3
9
6
8
.
7
6
7
.
3
0
6
3
.
4
6
7
.
8
7
K
→
E
5
7
.
9
9
6
1
.
3
5
7
.
9
5
5
.
3
5
5
.
3
5
7
.
5
5
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
:
3
3
0
7
-
3313
3312
Fig
u
r
e
5
.
Fi
g
u
r
e
6
s
h
o
w
s
t
h
e
ef
f
ec
t
o
f
co
m
b
i
n
in
g
m
u
ltip
le
s
o
u
r
ce
d
o
m
ai
n
s
.
W
e
s
ee
t
h
at
th
e
co
m
b
i
n
atio
n
o
f
DVD
a
n
d
E
lectr
o
n
ics
as
w
ell
as
B
o
o
k
an
d
DVD
a
s
s
o
u
r
ce
d
o
m
a
in
g
i
v
e
s
h
ig
h
es
t
ac
c
u
r
ac
y
.
Oth
er
o
b
s
er
v
atio
n
s
h
o
w
s
th
at
w
h
e
n
w
e
u
s
e
t
w
o
s
o
u
r
ce
d
o
m
ai
n
s
is
al
w
a
y
s
g
r
ea
t
er
th
an
t
h
e
ac
c
u
r
ac
y
i
f
w
e
u
s
e
s
in
g
le
d
o
m
ai
n
a
s
a
s
o
u
r
ce
.
E
x
p
er
i
m
e
n
t
w
as d
o
n
e
f
o
r
4
0
0
f
iles
.
Fig
u
r
e
5
.
5
Fo
ld
v
alid
atio
n
Fig
u
r
e
6
.
A
cc
u
r
ac
y
an
a
l
y
s
is
o
f
co
m
b
i
n
atio
n
o
f
s
o
u
r
ce
d
o
m
ain
s
C
o
m
b
i
n
atio
n
o
f
B
o
o
k
a
n
d
Kitch
e
n
a
s
s
o
u
r
ce
ap
p
lied
o
n
DV
D
d
o
m
ai
n
a
s
tar
g
et
s
h
o
w
i
n
g
les
s
ac
cu
r
ac
y
th
at
is
7
5
.
2
5
%
co
m
p
ar
ed
to
o
th
er
s
.
A
l
s
o
E
lectr
o
n
i
cs
an
d
Kitc
h
en
,
B
o
o
k
a
n
d
E
le
ctr
o
n
ics
a
s
a
s
o
u
r
ce
an
d
DVD
as
a
tar
g
et
p
r
o
d
u
ci
n
g
7
1
.
5
%
an
d
7
9
.
7
5
%
ac
cu
r
ac
y
r
esp
ec
tiv
e
l
y
.
Fro
m
o
b
s
er
v
atio
n
s
ac
c
u
r
ac
y
f
o
r
th
ese
d
o
m
ai
n
s
i
s
les
s
co
m
p
ar
ed
to
o
th
er
co
m
b
in
atio
n
s
an
d
DVD
as
tar
g
et.
I
t
s
tates
t
h
at
w
o
r
d
s
f
r
o
m
s
o
u
r
ce
d
o
m
ai
n
s
ar
e
n
o
t
m
atc
h
i
n
g
w
i
th
tar
g
et
d
o
m
ain
as
s
h
o
w
n
i
n
Fi
g
u
r
e
6
.
E
f
f
icie
n
c
y
o
f
m
o
d
el
is
d
ep
en
d
en
t
o
n
d
o
m
ai
n
s
s
i
m
ilar
it
y
.
4
.
CO
NCLUS
I
O
N
T
h
e
co
m
m
it
m
en
t
o
f
th
is
p
ap
er
is
to
ap
p
ly
th
e
s
e
m
i
s
u
p
er
v
i
s
ed
m
et
h
o
d
to
b
u
ild
d
o
m
ain
ad
ap
tatio
n
m
o
d
el
to
ex
tr
ac
t
f
ea
t
u
r
es
.
P
r
o
p
o
s
ed
ap
p
r
o
ac
h
u
tili
ze
s
m
a
x
i
m
u
m
e
n
tr
o
p
y
cla
s
s
i
f
ier
to
clas
s
if
y
r
ev
ie
w
s
i
n
to
t
wo
class
es
p
o
s
itiv
e
an
d
n
eg
a
tiv
e.
C
l
u
s
ter
i
n
g
is
ap
p
led
o
n
cla
s
s
if
ied
w
o
r
d
s
u
s
i
n
g
p
o
i
n
t
w
i
s
e
m
u
tu
al
in
f
o
r
m
a
tio
n
.
T
h
e
ex
p
er
ie
m
n
atl
r
e
s
u
l
ts
d
eli
v
er
g
o
o
d
ac
cu
r
ac
y
v
alu
e
f
o
r
5
f
o
ld
cr
o
s
s
v
alid
atio
n
an
d
co
m
b
in
at
io
n
o
f
s
o
u
r
ce
d
o
m
ai
n
s
.
B
litze
r
et
al.
[
1
4
]
m
u
ltip
r
o
d
u
ct
d
ataset
is
u
s
ed
f
o
r
e
x
p
er
i
m
e
n
ts
.
I
n
t
h
e
p
r
o
p
o
s
ed
f
r
a
m
e
w
o
r
k
,
clu
s
ter
i
n
g
f
o
r
o
n
l
y
u
n
i
g
r
a
m
s
is
u
s
ed
.
B
ig
r
a
m
s
ca
n
b
e
u
s
e
d
in
f
u
tu
r
e.
No
n
w
o
r
d
f
ea
t
u
r
es
ca
n
b
e
in
cl
u
d
ed
in
ap
p
r
o
ac
h
.
RE
F
E
R
E
NC
E
S
[
1
]
A
ciar
S.,
Z
h
an
g
D.
,
Si
m
o
f
f
S
.
,
Deb
en
h
a
m
J
.
,
"
I
n
f
o
r
m
ed
R
ec
o
m
m
e
n
d
er
:
B
asi
n
g
R
ec
o
m
m
en
d
atio
n
s
o
n
C
o
n
s
u
m
er
P
r
o
d
u
ct
R
ev
ie
w
s
,
"
I
n
I
n
tellig
en
t
S
ystems
,
I
E
E
E
,
Vo
l.2
2
,
I
s
s
u
e
0
3
,
p
p
.
3
9
-
4
7
,
2
0
0
7
.
[
2
]
B
in
g
L
i
u
,
“
Se
n
ti
m
e
n
t
An
al
y
s
is
&
Op
i
n
io
n
Mi
n
in
g
”,
Mo
r
g
a
n
&
C
la
yp
o
o
l
P
u
b
lis
h
ers
,
Kin
d
le
E
d
itio
n
,
2012.
[
3
]
B
er
m
i
n
g
h
a
m
A
.
C
o
n
w
a
y
M.
,
Mc
I
n
er
n
e
y
L
.
,
O
'
Har
e
N.
,
an
d
S
m
ea
to
n
A
.
,
“Co
m
b
i
n
i
n
g
So
cial
Net
w
o
r
k
An
al
y
s
i
s
a
n
d
Se
n
ti
m
e
n
t
An
al
y
s
i
s
to
E
x
p
lo
r
e
th
e
P
o
ten
tial
f
o
r
On
lin
e
R
ad
icali
s
atio
n
,
”
I
n
P
r
o
c.
o
f
I
n
t’
l
C
o
n
f.
o
n
A
d
va
n
ce
s
in
S
o
cia
l
N
etw
o
r
k
A
n
a
lysi
s
a
n
d
Min
in
g
,
A
t
h
e
n
s
,
Gr
ee
ce
,
J
u
l
y
2
0
-
2
2
,
p
p
.
2
3
1
-
2
3
6
,
2009.
[
4
]
B
litzer
J
,
R
.
Mc
Do
n
ald
,
F.
P
e
r
er
ia,
“
Do
m
ai
n
A
d
ap
tatio
n
with
C
o
r
r
esp
o
n
d
en
ce
L
ea
r
n
i
n
g
,
”
E
MN
L
P
’
0
6
,
In
P
r
o
ce
ed
in
g
s
o
f
th
e
2
0
0
6
C
o
n
feren
ce
o
n
E
mp
ir
ica
l
Meth
o
d
s
in
N
a
tu
r
a
l
La
n
g
u
a
g
e
P
r
o
ce
s
s
in
g
,
Stro
u
d
s
b
u
r
g
,
P
A
,
US
A
,
P
ag
e
s
1
2
0
-
1
2
8
,
2
0
0
6
.
[
5
]
Ma
tth
e
w
W
h
ite
h
ea
d
,
L
ar
r
y
Yae
g
er
,
“
B
u
ild
i
n
g
a
Gen
er
al
P
u
r
p
o
s
e
C
r
o
s
s
-
Do
m
ai
n
Se
n
t
i
m
e
n
t
Mi
n
i
n
g
Mo
d
el
,
”
C
o
mp
u
ter
S
cien
ce
a
n
d
I
n
fo
r
ma
tio
n
E
n
g
i
n
ee
r
in
g
,
2
0
0
9
W
R
I
W
o
r
ld
C
o
n
g
r
es
s
o
n
,
Vo
lu
m
e
-
4
,
p
p
.
4
7
2
-
476,
L
o
s
An
g
eles,
C
A,
2
0
0
9
.
[
6
]
P
an
Sin
n
o
J
ialin
,
Ni
Xiao
ch
u
a
n
,
J
ian
-
tao
Su
n
,
Qian
g
y
an
g
,
a
n
d
Z
h
en
g
C
h
e
n
,
“
C
r
o
s
s
Do
m
a
in
Se
n
ti
m
en
t
C
las
s
i
f
icatio
n
v
ia
Sp
ec
tr
al
Fe
atu
r
e
A
li
g
n
m
en
t
,
”
A
C
M,
1
9
th
I
n
ter
n
atio
n
al
W
o
r
ld
W
id
e
W
eb
C
o
n
feren
ce
,
R
aleig
h
,
No
r
th
C
ar
o
lin
a,
US
A
,
2
0
1
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
Op
in
io
n
min
in
g
u
s
in
g
c
o
mb
in
a
tio
n
a
l a
p
p
r
o
a
c
h
fo
r
d
iffer
en
t d
o
ma
in
s
(
Jy
o
ti De
s
h
mu
kh
)
3313
[
7
]
Kan
g
L
i
u
,
L
i
h
e
n
g
X
u
,
an
d
J
u
n
Z
h
ao
,
“
C
o
-
E
x
tr
ac
tin
g
Op
in
i
o
n
T
ar
g
ets
an
d
Op
in
io
n
W
o
r
d
s
f
r
o
m
On
lin
e
R
ev
ie
w
s
B
ased
o
n
t
h
e
W
o
r
d
A
li
g
n
m
en
t
Mo
d
el
,
”
I
E
E
E
tr
a
n
s
a
ctio
n
s
o
n
K
n
o
w
led
g
e
Da
t
a
E
n
g
in
ee
r
in
g
,
v
o
l.
2
7
,
n
o
.
3
,
Ma
r
ch
,
2
0
1
5
.
[
8
]
B
ala
m
u
r
ali
A
R
,
Deb
r
aj
Ma
n
n
a,
P
u
s
h
p
ak
B
h
attac
h
ar
y
y
a,
“
C
r
o
s
s
-
Do
m
ai
n
Sen
t
i
m
e
n
t
T
ag
g
in
g
Usi
n
g
Me
ta
-
C
las
s
if
ier
an
d
a
Hi
g
h
A
cc
u
r
ac
y
I
n
-
Do
m
ai
n
C
las
s
if
ier
,
”
P
r
o
ce
ed
in
g
s
o
f
I
C
O
N
2
0
1
0
:
8
th
I
n
tern
a
tio
n
a
l Co
n
feren
ce
o
n
N
a
tu
r
a
l La
n
g
u
a
g
e
P
r
o
ce
s
s
in
g
,
2
0
1
0
.
[
9
]
B
o
lleg
ala
D.
W
eir
D.
,
an
d
C
ar
r
o
ll
J
.
,
"
C
r
o
s
s
-
Do
m
ai
n
S
en
ti
m
en
t
C
la
s
s
i
f
icatio
n
u
s
in
g
a
Sen
ti
m
en
t
Sen
s
iti
v
e
T
h
esau
r
u
s
,
"
I
n
K
n
o
w
led
g
e
a
n
d
Da
ta
E
n
g
in
ee
r
i
n
g
,
I
E
E
E
Tr
a
n
s
a
ctio
n
s
,
Vo
l.
2
5
I
s
s
u
e:
8
,
p
p
.
1
7
1
9
-
1
7
3
1
,
2
0
1
3
.
[
1
0
]
L
i
Sh
o
u
s
h
an
,
Yu
n
x
ia
X
u
e,
Z
h
o
n
g
q
in
g
W
an
g
,
G
u
o
d
o
n
g
Z
h
o
u
,
“A
ct
iv
e
L
ea
r
n
i
n
g
f
o
r
C
r
o
s
s
-
d
o
m
ai
n
Sen
ti
m
e
n
t C
lass
if
ica
tio
n
,
”
In
I
JCAI
,
p
p
.
2
1
2
7
-
2
1
3
3
,
2
0
1
3
.
[
1
1
]
Dh
illo
n
I
n
d
er
j
it
S.,
S
u
b
r
a
m
a
n
y
a
m
Ma
llela,
Ku
m
ar
R
a
h
u
l,
“A
Di
v
is
i
v
e
I
n
f
o
r
m
atio
n
-
T
h
e
o
r
etic
Featu
r
e
C
lu
s
ter
i
n
g
Alg
o
r
it
h
m
f
o
r
T
ex
t
C
las
s
if
icatio
n
,
”
In
J
o
u
r
n
a
l
o
f
Ma
ch
in
e
Lea
r
n
in
g
R
esea
r
ch
,
p
p
.
1
2
6
5
-
1
2
8
7
,
2
0
0
3
.
[
1
2
]
Hi
m
a
n
s
h
u
S.
B
h
at
t,
De
ep
a
l
i
Se
m
w
a
l,
S
h
o
u
r
y
a
R
o
y
,
“An
I
te
r
a
ti
v
e
Si
m
i
lar
i
t
y
b
a
s
e
d
A
d
ap
t
at
io
n
T
ec
h
n
iq
u
e
f
o
r
C
r
o
s
s
Do
m
ai
n
T
e
x
t
C
l
a
s
s
i
f
ic
at
io
n
,
”
P
r
o
c
ee
d
in
g
s
o
f
th
e
1
9
th
C
o
n
fe
r
en
ce
o
n
C
o
mp
u
t
a
t
io
n
a
l
La
n
g
u
a
g
e
L
ea
r
n
in
g
,
p
p
.
5
2
-
6
1
,
B
e
ij
i
n
g
,
C
h
i
a
n
a
,
J
u
l
y
3
0
-
3
1
,
2
0
1
5
.
[
1
3
]
Nig
a
m
,
Ka
m
al,
J
o
h
n
L
a
f
f
er
t
y
,
A
n
d
r
e
w
Mc
C
al
lu
m
,
“
Usi
n
g
m
ax
i
m
u
m
e
n
tr
o
p
y
f
o
r
tex
t
c
lass
i
f
icatio
n
,
”
In
I
JCAI
-
9
9
w
o
r
ksh
o
p
o
n
ma
c
h
in
e
lea
r
n
in
g
f
o
r
in
fo
r
ma
tio
n
f
ilter
in
g
,
v
o
l.
1
,
p
p
.
6
1
-
6
7
.
1
9
9
9
.
[
1
4
]
B
litzer
J
.
,
M
.
Dr
ed
ze
,
F.
P
er
eir
a,
“
B
io
g
r
ap
h
ies,
b
o
lly
w
o
o
d
,
b
o
o
m
-
b
o
x
e
s
an
d
b
le
n
d
er
s
:
Do
m
ai
n
ad
ap
tatio
n
f
o
r
s
en
ti
m
e
n
t
cla
s
s
i
f
icatio
n
,
”
I
n
P
r
o
ce
ed
in
g
s
o
f
th
e
4
5
th
A
n
n
u
a
l
Meetin
g
o
f
th
e
A
s
s
o
cia
tio
n
o
f
C
o
mp
u
ta
tio
n
a
l Lin
g
u
is
tics
,
p
p
.
4
3
2
-
4
3
9
,
P
r
ag
u
e,
C
ze
ch
R
ep
u
b
lic,
2
0
0
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.