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.
7
,
No
.
1
,
Feb
r
u
ar
y
201
7
,
p
p
.
5
5
1
~
55
8
I
SS
N:
2088
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
7
i
1
.
p
p
5
5
1
-
5
5
8
551
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
JE
C
E
A
n I
m
pro
v
ed Si
mila
rity Ma
tching
b
a
sed Cluste
ring
Fra
m
ew
o
rk
for Sho
rt
and Sen
tence
L
ev
el Tex
t
M
.
J
o
hn
B
a
s
ha
1
,
K
.
P
.
K
a
liy
a
m
u
rt
hie
2
1
De
p
a
rtme
n
t
o
f
CS
E,
P
.
T
.
R
C
o
ll
e
g
e
o
f
En
g
in
e
e
rin
g
&
T
e
c
h
n
o
l
o
g
y
,
M
a
d
u
ra
i,
T
a
m
il
Na
d
u
6
2
5
0
0
8
,
I
n
d
ia
2
De
p
a
rtme
n
t
o
f
CS
E,
Bh
a
ra
th
Un
i
v
e
rsit
y
,
Ch
e
n
n
a
i
-
6
0
0
0
7
3
,
T
a
m
il
Na
d
u
,
I
n
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
1
9
,
2
0
1
6
R
ev
i
s
ed
Oct
2
1
,
2
0
1
6
A
cc
ep
ted
No
v
5
,
2
0
1
6
T
e
x
t
c
lu
ste
rin
g
p
la
y
s
a
k
e
y
ro
le
in
n
a
v
ig
a
ti
o
n
a
n
d
b
ro
w
sin
g
p
ro
c
e
ss
.
F
o
r
a
n
e
ff
ici
e
n
t
tex
t
c
lu
ste
rin
g
,
th
e
lar
g
e
a
m
o
u
n
t
o
f
in
f
o
rm
a
ti
o
n
is
g
ro
u
p
e
d
in
to
m
e
a
n
in
g
f
u
l
c
lu
ste
rs.
M
u
lt
i
p
le
te
x
t
c
lu
ste
rin
g
tec
h
n
i
q
u
e
s
d
o
n
o
t
a
d
d
re
ss
th
e
issu
e
s
su
c
h
a
s,
h
ig
h
ti
m
e
a
n
d
sp
a
c
e
c
o
m
p
lex
it
y
,
in
a
b
il
it
y
to
u
n
d
e
rsta
n
d
th
e
re
latio
n
a
l
a
n
d
c
o
n
tex
tu
a
l
a
tt
rib
u
te
s
o
f
th
e
w
o
rd
,
les
s
ro
b
u
stn
e
ss
,
ris
k
s
re
late
d
to
p
riv
a
c
y
e
x
p
o
su
re
,
e
tc.
T
o
a
d
d
re
ss
th
e
se
issu
e
s,
a
n
e
ff
icie
n
t
tex
t
b
a
se
d
c
lu
ste
rin
g
f
ra
m
e
w
o
rk
is
p
ro
p
o
se
d
.
T
h
e
Re
u
ters
d
a
tas
e
t
is
c
h
o
se
n
a
s
th
e
in
p
u
t
d
a
tas
e
t.
On
c
e
th
e
in
p
u
t
d
a
tas
e
t
is
p
re
p
ro
c
e
ss
e
d
,
th
e
sim
il
a
rit
y
b
e
tw
e
e
n
th
e
w
o
rd
s
a
re
c
o
m
p
u
ted
u
sin
g
th
e
c
o
sin
e
sim
il
a
rit
y
.
T
h
e
si
m
il
a
rit
ies
b
e
twe
e
n
th
e
c
o
m
p
o
n
e
n
ts
a
re
c
o
m
p
a
re
d
a
n
d
th
e
v
e
c
to
r
d
a
ta
is
c
re
a
ted
.
F
ro
m
th
e
v
e
c
to
r
d
a
ta
th
e
c
lu
ste
rin
g
p
a
rti
c
le
is
c
o
m
p
u
ted
.
T
o
o
p
ti
m
ize
th
e
c
lu
ste
ri
n
g
re
su
lt
s,
m
u
tatio
n
is
a
p
p
li
e
d
t
o
t
h
e
v
e
c
to
r
d
a
ta.
T
h
e
p
e
rf
o
rm
a
n
c
e
th
e
p
ro
p
o
se
d
tex
t
b
a
se
d
c
lu
ste
rin
g
f
r
a
m
e
w
o
rk
is
a
n
a
ly
z
e
d
u
sin
g
th
e
m
e
tri
c
s
su
c
h
a
s
M
e
a
n
S
q
u
a
re
Err
o
r
(M
S
E)
,
P
e
a
k
S
ig
n
a
l
No
ise
Ra
t
io
(P
S
NR)
a
n
d
P
ro
c
e
ss
in
g
ti
m
e
.
F
ro
m
th
e
e
x
p
e
rim
e
n
tal
re
su
lt
s,
it
is
f
o
u
n
d
th
a
t,
t
h
e
p
r
o
p
o
se
d
t
e
x
t
b
a
se
d
c
lu
ste
rin
g
f
ra
m
e
w
o
rk
p
ro
d
u
c
e
d
o
p
ti
m
a
l
M
S
E,
P
S
NR
a
n
d
p
ro
c
e
ss
in
g
ti
m
e
w
h
e
n
c
o
m
p
a
r
e
d
to
t
h
e
e
x
isti
n
g
F
u
z
z
y
C
-
M
e
a
n
s
(F
CM
)
a
n
d
P
a
irw
ise
Ra
n
d
o
m
S
w
a
p
(P
RS
)
m
e
t
h
o
d
s.
K
ey
w
o
r
d
:
Fu
zz
y
c
-
m
ea
n
s
(
F
C
M)
Me
an
s
q
u
ar
e
er
r
o
r
(
MSE
)
P
air
w
is
e
r
an
d
o
m
s
w
ap
(
P
R
S)
P
ar
ticle
s
w
ar
m
o
p
ti
m
izatio
n
(
P
SO)
P
ea
k
s
ig
n
al
n
o
i
s
e
r
atio
(
P
SNR
)
T
ex
t c
lu
s
ter
i
n
g
Co
p
y
rig
h
t
©
2
0
1
7
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
:
M.
J
o
h
n
B
ash
a
,
A
P
&
Hea
d
,
Dep
ar
t
m
en
t o
f
C
SE,
P
.
T
.
R
C
o
lleg
e
o
f
E
n
g
i
n
ee
r
in
g
&
T
ec
h
n
o
lo
g
y
,
Ma
d
u
r
ai,
T
am
il Na
d
u
6
2
5
0
0
8
,
I
n
d
ia
.
E
m
ail:
1.
I
NT
RO
D
UCT
I
O
N
T
ex
t
clu
s
ter
i
n
g
i
s
th
e
p
r
o
ce
s
s
o
f
m
a
n
ag
in
g
t
h
e
lar
g
e
a
m
o
u
n
t
o
f
d
ig
itall
y
s
to
r
ed
elec
tr
o
n
ic
d
ata.
T
h
e
h
ig
h
v
o
l
u
m
e
o
f
d
ata
is
u
s
ed
f
o
r
th
e
d
ata
an
aly
s
is
,
clas
s
i
f
icatio
n
an
d
r
etr
iev
al
tech
n
iq
u
es.
I
n
ca
s
e
o
f
th
e
p
r
o
to
ty
p
e
b
ased
clu
s
ter
i
n
g
,
t
h
e
s
eq
u
en
ce
o
f
t
h
e
p
r
o
to
t
y
p
e
s
ar
e
u
s
ed
f
o
r
f
i
n
d
in
g
t
h
e
b
e
s
t
f
it
d
ata
w
i
th
th
e
u
n
k
n
o
w
n
s
tr
u
ct
u
r
es.
T
o
r
ep
r
esen
t
t
h
e
cl
u
s
ter
s
in
k
-
m
ea
n
s
,
o
n
l
y
a
s
in
g
le
p
r
o
to
t
y
p
e
is
u
s
ed
.
Mu
ltip
le
r
ea
l
ap
p
licatio
n
s
u
s
e
t
h
e
p
r
o
to
ty
p
e
b
ased
clu
s
ter
in
g
b
ec
au
s
e,
it
p
r
o
v
id
es
less
co
m
p
u
tatio
n
a
l
an
d
m
e
m
o
r
y
s
p
ac
e.
Sev
er
al
o
t
h
er
m
e
th
o
d
s
h
a
v
e
b
ee
n
d
ev
elo
p
ed
,
w
h
ic
h
ar
e
b
ased
o
n
s
to
ch
ast
ic
g
lo
b
al
o
p
ti
m
izat
io
n
s
u
c
h
a
s
s
i
m
u
lated
an
n
ea
li
n
g
an
d
g
en
e
tic
alg
o
r
ith
m
s
.
B
u
t
th
e
s
e
m
et
h
o
d
s
p
r
o
v
id
e
a
h
ig
h
ti
m
e
co
m
p
lex
i
t
y
.
C
l
u
s
ter
in
g
alg
o
r
ith
m
a
n
d
cl
u
s
ter
v
al
id
it
y
ar
e
t
h
e
co
m
m
o
n
l
y
u
s
ed
co
r
r
elate
d
p
ar
ts
i
n
t
h
e
cl
u
s
ter
an
a
l
y
s
i
s
.
Ge
n
er
all
y
,
to
p
r
ev
en
t
t
h
e
in
itializatio
n
p
r
o
b
le
m
s
,
t
h
e
k
-
m
ea
n
s
al
g
o
r
ith
m
i
s
ex
ec
u
ted
m
an
y
ti
m
e
s
w
it
h
d
if
f
er
en
t
p
ar
a
m
e
ter
s
.
T
h
e
o
p
tim
al
s
o
l
u
tio
n
i
s
p
r
o
v
id
ed
as
th
e
r
esu
lt.
T
h
e
q
u
alit
y
o
f
th
e
cl
u
s
ter
i
n
g
i
s
co
m
p
u
ted
u
s
i
n
g
co
s
t
f
u
n
ctio
n
.
T
h
ec
ateg
o
r
izatio
n
o
f
th
e
d
ataset
d
ep
en
d
s
o
n
th
e
co
s
t
f
u
n
ctio
n
.
T
h
e
clu
s
ter
i
n
g
m
et
h
o
d
s
ar
e
class
if
ied
a
s
d
en
s
it
y
-
b
a
s
ed
m
eth
o
d
s
,
g
r
ap
h
b
ased
m
eth
o
d
s
,
g
r
id
b
ased
m
et
h
o
d
s
an
d
m
eth
o
d
s
f
o
r
h
i
g
h
d
i
m
en
s
io
n
a
l
s
p
ac
e
d
ata.
T
h
e
m
aj
o
r
is
s
u
es
i
n
t
h
e
ex
is
t
in
g
cl
u
s
ter
i
n
g
al
g
o
r
ith
m
s
in
cl
u
d
e,
h
i
g
h
p
r
o
ce
s
s
in
g
t
i
m
e
co
n
s
u
m
p
tio
n
an
d
h
ig
h
co
m
p
le
x
it
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
1
,
Feb
r
u
ar
y
2
0
1
7
:
55
1
–
558
552
Gra
n
a
d
o
s
,
et
a
l
[
1
]
p
r
o
p
o
s
ed
a
B
Or
j
ae
tech
n
iq
u
e
to
clu
s
ter
t
h
e
te
x
ts
b
ased
o
n
th
e
s
tr
i
n
g
co
m
p
r
es
s
io
n
.
I
t
ef
f
ec
ti
v
el
y
co
m
p
u
ted
th
e
d
is
to
r
tio
n
p
r
e
s
e
n
t
in
th
e
i
n
f
o
r
m
atio
n
o
f
t
h
e
tex
t.
W
h
en
th
e
p
r
o
p
o
s
ed
tech
n
iq
u
e
w
as
ap
p
lied
f
o
r
th
e
s
tr
u
ctu
r
al
d
ataset
s
,
th
e
s
tr
u
ct
u
r
e
o
f
th
e
d
ataset
w
a
s
co
m
p
le
tel
y
d
estro
y
ed
.
Lee,
et
a
l
[
2
]
s
u
g
g
ested
a
f
u
zz
y
b
ased
m
et
h
o
d
to
class
i
f
y
th
e
t
ex
t
p
r
esen
t
in
m
u
lti
ca
te
g
o
r
y
d
o
cu
m
en
t.
A
f
u
zz
y
r
elev
an
ce
m
ea
s
u
r
e
w
as
u
s
ed
to
co
n
v
er
t
th
e
h
i
g
h
d
i
m
e
n
s
io
n
al
d
o
cu
m
en
t
in
to
a
lo
w
d
i
m
en
s
io
n
al
d
o
cu
m
en
t.
T
h
e
p
r
o
p
o
s
ed
clu
s
ter
in
g
tec
h
n
iq
u
e
s
p
litt
ed
t
h
e
r
elev
a
n
ce
s
p
a
ce
in
to
m
u
lt
ip
le
s
u
b
r
eg
io
n
s
.
T
h
e
in
d
iv
id
u
a
l
s
u
b
r
eg
io
n
s
w
er
e
t
h
en
co
m
b
i
n
ed
to
cr
ea
te
th
e
in
d
iv
id
u
a
l
ca
teg
o
r
y
.
T
h
e
s
u
g
g
e
s
ted
clu
s
ter
in
g
m
eth
o
d
p
r
o
v
id
ed
o
p
tim
a
l
p
er
f
o
r
m
an
ce
a
n
d
s
p
ee
d
th
an
th
e
o
t
h
er
tex
t
cl
u
s
ter
i
n
g
tec
h
n
iq
u
es.
W
e
i,
et
a
l
[
3
]
p
r
o
p
o
s
ed
a
lex
ical
ch
ain
b
ased
w
o
r
d
n
et.
I
t
u
s
ed
t
h
eo
n
to
lo
g
y
h
ier
ar
ch
ica
l
s
tr
u
ct
u
r
e
to
d
eter
m
in
e
th
e
s
i
m
ilar
i
t
y
b
et
w
ee
n
th
e
ter
m
s
o
f
th
e
w
o
r
d
s
.
T
h
e
lex
ical
ch
ai
n
w
a
s
u
s
ed
to
o
b
tain
t
h
e
s
e
m
a
n
tic
r
ela
tio
n
s
h
ip
o
f
th
e
w
o
r
d
s
p
r
esen
t
i
n
th
e
te
x
t.
W
h
en
co
m
p
ar
ed
to
th
e
clas
s
ic
al
m
et
h
o
d
s
,
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
in
cr
ea
s
ed
th
e
p
er
f
o
r
m
an
ce
.
P
en
g
,
et
a
l
[
4
]
p
r
o
p
o
s
ed
a
n
o
v
el
C
F
u
-
tr
ee
b
ased
d
o
w
n
-
to
p
in
cr
e
m
e
n
tal
co
n
ce
p
tu
a
l
h
ier
ar
ch
ical
te
x
t
clu
s
ter
i
n
g
ap
p
r
o
ac
h
f
o
r
clu
s
te
r
in
g
t
h
e
tex
t
i
n
t
h
e
d
o
cu
m
e
n
t.
T
h
e
co
m
p
ar
is
o
n
v
ar
ia
tio
n
(
C
V)
cr
iter
io
n
d
ec
id
ed
w
h
et
h
er
to
m
er
g
e
o
r
s
p
lit
t
h
e
clu
s
ter
s
.
W
h
en
co
m
p
ar
ed
to
th
e
ex
is
ti
n
g
K
-
Me
an
s
al
g
o
r
ith
m
,
t
h
e
p
r
o
p
o
s
ed
tex
t
clu
s
ter
i
n
g
alg
o
r
ith
m
w
as
e
f
f
ici
en
t.
Yu
a
n
a
n
d
S
h
i
[
5
]
p
r
o
p
o
s
e
d
a
tex
t
clu
s
ter
in
g
alg
o
r
it
h
m
t
o
p
r
ev
en
t
th
e
is
s
u
e
s
in
th
e
d
i
v
is
io
n
b
ased
clu
s
ter
i
n
g
m
et
h
o
d
.
T
h
e
co
m
p
lex
f
ea
tu
r
es
s
u
c
h
as,
s
y
n
o
n
y
m
an
d
co
-
o
cc
u
r
r
in
g
w
o
r
d
s
w
er
e
o
b
tain
ed
f
r
o
m
th
e
m
u
lti
p
le
s
e
m
a
n
tic
in
f
o
r
m
at
io
n
.
U
s
i
n
g
th
e
d
iv
id
e
a
n
d
co
n
q
u
er
tec
h
n
iq
u
e,
t
h
e
iter
atio
n
en
d
ed
w
it
h
t
h
e
ex
p
ec
ted
cl
u
s
ter
n
u
m
b
er
.
B
y
d
y
n
a
m
ica
ll
y
u
p
d
atin
g
th
e
ce
n
ter
n
u
m
b
er
,
o
p
tim
a
l
clu
s
ter
i
n
g
r
esu
lt
s
w
er
e
o
b
tain
ed
.
B
h
a
r
th
i,
et
a
l
[
6
]
s
u
g
g
ested
a
t
h
r
ee
-
s
tag
e
d
i
m
e
n
s
io
n
r
ed
u
ctio
n
m
o
d
el
to
g
en
er
ate
an
in
f
o
r
m
ati
v
e
f
ea
tu
r
e
s
u
b
s
p
ac
e.
T
h
e
d
im
e
n
s
io
n
s
o
f
t
h
e
f
ea
t
u
r
e
s
p
ac
e
w
er
e
m
in
i
m
ized
.
T
h
e
to
tal
ex
ec
u
tio
n
ti
m
e
f
o
r
cr
ea
tin
g
t
h
e
clu
s
ter
an
d
cr
ea
tin
g
t
h
e
d
o
cu
m
e
n
t
cl
u
s
ter
was
s
i
g
n
if
ica
n
tl
y
r
ed
u
ce
d
.
S
o
n
g
,
et
a
l
[
7
]
p
r
o
p
o
s
e
d
a
n
o
v
el
h
y
b
r
id
s
e
m
a
n
tic
s
i
m
il
ar
it
y
m
ea
s
u
r
e
b
ased
f
u
zz
y
co
n
tr
o
l
Gen
e
tic
A
l
g
o
r
ith
m
(
G
A)
f
o
r
clu
s
ter
in
g
t
h
e
d
o
cu
m
en
ts
.
T
h
e
Se
m
a
n
tic
Sp
ac
e
Mo
d
el
(
SS
M)
w
a
s
u
s
ed
a
s
t
h
e
co
r
p
u
s
-
b
ased
m
et
h
o
d
.
T
h
e
r
ed
u
ctio
n
i
n
th
e
d
i
m
en
s
io
n
s
o
f
t
h
e
S
SM
w
as
u
s
ed
to
o
b
tain
th
e
tr
u
e
r
elatio
n
s
h
ip
b
et
w
ee
n
t
h
e
d
o
cu
m
e
n
ts
.
T
h
e
t
h
esa
u
r
u
s
b
a
s
ed
m
et
h
o
d
w
as
co
m
b
in
ed
w
it
h
t
h
e
SS
M
to
p
r
o
v
id
e
th
e
s
e
m
a
n
tic
s
i
m
i
lar
it
y
m
ea
s
u
r
e.
W
h
e
n
co
m
p
ar
ed
to
th
e
tr
ad
itio
n
al
G
A
,
t
h
e
p
r
o
p
o
s
ed
h
y
b
r
id
s
e
m
a
n
tic
s
tr
ate
g
y
p
r
o
v
id
ed
o
p
tim
al
p
er
f
o
r
m
a
n
ce
.
Go
n
g
,
et
a
l
[
8
]
p
r
o
p
o
s
ed
a
v
alid
it
y
in
d
e
x
b
as
ed
m
eth
o
d
to
ad
d
r
ess
t
h
e
is
s
u
es
o
f
t
h
e
ad
ap
tiv
e
f
ea
t
u
r
e
s
ele
ctio
n
f
o
r
cl
u
s
ter
i
n
g
th
e
tex
t
s
tr
ea
m
.
T
h
e
th
r
es
h
o
ld
o
f
th
e
clu
s
ter
v
alid
i
n
d
ex
w
as
u
s
ed
to
r
eselect
th
e
f
ea
tu
r
e
s
f
o
r
cr
ea
tin
g
a
v
alid
clu
s
ter
.
T
h
e
q
u
alit
y
o
f
th
e
p
r
o
p
o
s
ed
clu
s
ter
in
g
al
g
o
r
ith
m
w
a
s
h
i
g
h
.
Ya
o
,
et
a
l
[
9
]
p
r
o
p
o
s
e
d
a
k
-
m
ea
n
s
b
ased
C
h
i
n
e
s
e
tex
t
cl
u
s
ter
in
g
al
g
o
r
ith
m
to
clu
s
ter
th
e
te
x
t.
T
h
e
av
er
ag
e
s
i
m
ilar
it
y
p
ar
a
m
eter
w
a
s
u
s
ed
to
o
b
tain
t
h
e
s
i
m
ilar
it
y
th
r
es
h
o
ld
v
al
u
e.
I
n
it
iall
y
,
t
h
e
o
r
ig
i
n
al
cl
u
s
ter
ce
n
te
r
th
at
w
as
ab
o
v
e
th
e
t
h
r
es
h
o
ld
v
al
u
e
w
as
c
h
o
s
e
n
as th
e
ca
n
d
id
ate
co
llectio
n
,
th
en
t
h
e
cl
u
s
ter
L
in
,
et
a
l
[
1
0
]
p
r
o
p
o
s
ed
a
n
o
v
el
s
i
m
ilar
it
y
m
ea
s
u
r
e
to
co
m
p
u
te
t
h
e
s
i
m
ilar
it
y
b
et
w
ee
n
t
w
o
d
o
cu
m
en
ts
.
T
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
co
n
s
id
er
ed
th
e
s
i
tu
at
io
n
s
,
s
u
c
h
as,
f
ea
tu
r
e
s
in
b
o
t
h
th
e
d
o
cu
m
e
n
t
s
,
f
ea
t
u
r
es
i
n
o
n
l
y
o
n
e
d
o
cu
m
e
n
t
a
n
d
f
ea
t
u
r
es
ab
s
en
t
i
n
b
o
th
th
e
d
o
cu
m
en
ts
.
I
f
b
o
th
t
h
e
d
o
cu
m
en
ts
h
ad
t
h
e
f
ea
t
u
r
es,
th
e
s
i
m
ilar
it
y
b
et
w
ee
n
t
h
e
m
w
a
s
i
n
cr
ea
s
ed
.
I
f
o
n
l
y
o
n
e
d
o
cu
m
e
n
t
h
ad
t
h
e
f
ea
t
u
r
es,
t
h
en
a
f
i
x
ed
v
a
lu
e
w
a
s
ch
o
s
en
a
s
t
h
e
s
i
m
i
lar
it
y
.
I
f
n
o
n
e
o
f
th
e
d
o
cu
m
e
n
ts
h
ad
th
e
f
ea
tu
r
es,
th
e
s
i
m
ilar
it
y
v
alu
e
w
as
f
o
u
n
d
t
o
b
e
ab
s
en
t.
W
h
en
co
m
p
ar
e
d
to
th
e
o
th
er
m
ea
s
u
r
e
s
,
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
p
r
o
d
u
ce
d
o
p
tim
a
l r
esu
l
ts
.
Liu
,
et
a
l
[
1
1
]
p
r
o
p
o
s
ed
a
s
em
an
tic
tr
ee
b
ased
tex
t
clu
s
ter
in
g
al
g
o
r
ith
m
f
o
r
clu
s
ter
in
g
t
h
e
p
ar
allel
tex
ts
.
T
h
e
p
ar
allel
alg
o
r
it
h
m
s
w
er
e
u
s
ed
to
m
in
i
m
ize
t
h
e
t
i
m
e
co
m
p
lex
i
t
y
.
I
t
in
itiated
t
h
e
p
r
o
ce
s
s
es
at
t
h
e
s
a
m
e
t
i
m
e.
T
h
e
m
aster
p
r
o
ce
s
s
p
er
f
o
r
m
ed
t
h
e
d
ata
p
ar
titi
o
n
i
n
g
,
i
n
f
o
r
m
atio
n
co
llect
io
n
an
d
cl
u
s
ter
i
n
g
p
r
o
ce
s
s
es.
T
h
e
s
la
v
e
p
r
o
ce
s
s
ca
lcu
lated
t
h
e
w
o
r
d
f
r
eq
u
en
c
y
.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
p
r
o
d
u
ce
d
ac
c
u
r
ate
r
esu
lt
s
w
i
th
les
s
ti
m
e
co
m
p
lex
it
y
.
L
i,
et
a
l
[
1
2
]
s
u
g
g
est
ed
a
Fu
zz
y
Ma
h
ala
n
o
b
is
d
is
tan
ce
s
b
ased
tex
t
clu
s
ter
i
n
g
al
g
o
r
ith
m
to
i
n
cr
ea
s
e
th
e
p
r
ec
is
io
n
a
n
d
e
f
f
icien
c
y
o
f
th
e
d
ataset.
T
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
w
a
s
f
o
u
n
d
to
b
e
m
o
r
e
v
alid
th
a
n
t
h
e
tr
ad
i
tio
n
al
f
u
z
z
y
p
ar
titi
o
n
i
n
g
te
x
t
c
lu
s
ter
i
n
g
alg
o
r
it
h
m
s
.
N
g
u
ye
n
,
et
a
l
[
1
3
]
an
aly
ze
d
th
e
q
u
alit
y
is
s
u
e
s
o
f
th
e
cl
u
s
t
er
in
g
r
es
u
lts
.
T
h
e
ex
te
n
d
ed
Se
m
an
tic
E
v
al
u
atio
n
b
y
E
x
p
lo
r
atio
n
(
SEE
)
m
e
th
o
d
w
a
s
u
s
ed
to
r
etr
iev
e
th
e
I
N
FON
A
d
o
cu
m
e
n
ts
.
Ga
o
,
et
a
l
[
1
4
]
p
r
o
p
o
s
ed
a
g
en
etic
a
lg
o
r
ith
m
b
ased
te
x
t
clu
s
ter
i
n
g
.
I
t
in
te
g
r
ated
th
e
l
aten
t
s
e
m
a
n
tic
a
n
al
y
s
i
s
.
W
h
e
n
co
m
p
ar
ed
to
th
e
s
i
n
g
le
cl
u
s
ter
in
g
m
eth
o
d
,
th
e
p
r
o
p
o
s
ed
clu
s
ter
in
g
al
g
o
r
ith
m
p
r
o
v
id
ed
o
p
tim
a
l
clu
s
ter
in
g
s
o
lu
tio
n
s
.
S
h
i,
et
a
l
[
1
5
]
p
r
o
p
o
s
ed
a
p
aten
ted
tex
t
clu
s
ter
i
n
g
al
g
o
r
i
th
m
n
a
m
ed
,
C
l
u
s
ter
i
n
g
b
y
Gen
et
ic
A
l
g
o
r
ith
m
Mo
d
el
(
C
G
A
M)
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
in
te
g
r
ated
th
e
f
it
n
es
s
f
u
n
ctio
n
i
n
th
e
Ge
n
etic
Alg
o
r
it
h
m
(
GA
)
an
d
co
n
v
er
g
en
ce
cr
iter
i
o
n
i
n
t
h
e
K
-
Me
a
n
s
alg
o
r
ith
m
.
W
h
en
co
m
p
ar
ed
to
th
e
tr
ad
itio
n
al
GA
an
d
K
-
Me
an
s
,
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
o
b
tai
n
ed
o
p
tim
a
l
clu
s
ter
i
n
g
r
esu
lt
s
.
Su
m
m
ar
iza
tio
n
o
f
d
o
cu
m
e
n
ts
b
ased
o
n
th
e
s
a
m
e
to
p
ics
p
la
y
t
h
e
m
aj
o
r
r
o
le
in
th
e
q
u
ick
u
n
d
er
s
ta
n
d
in
g
a
n
d
cr
ea
tio
n
o
f
lea
g
al
j
u
d
g
e
m
en
t
s
b
et
w
ee
n
th
e
d
o
cu
m
en
t
s
an
d
to
p
ics.
V
e
n
ka
tesh
et
a
l
[
16
]
u
tili
ze
d
th
e
h
ier
ar
c
h
ical
L
ate
n
t
D
ir
ich
let
A
llo
ca
tio
n
(
h
L
D
A
)
u
s
in
g
s
i
m
ilar
it
y
m
ea
s
u
r
e
b
et
w
ee
n
to
p
ics
an
d
d
o
cu
m
en
ts
an
d
to
f
i
n
d
th
e
s
u
m
m
ar
izatio
n
o
f
ea
ch
d
o
cu
m
en
t
u
s
i
n
g
t
h
e
s
a
m
e
to
p
ics.
T
h
e
p
r
o
ce
s
s
in
g
o
v
er
h
ea
d
is
h
ig
h
d
u
e
to
t
h
e
m
o
r
e
n
u
m
b
er
o
f
p
o
s
e
tag
g
er
s
,
p
r
o
ce
s
s
i
n
g
to
o
ls
a
n
d
d
iv
er
s
e
c
h
o
ices
o
f
n
at
u
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
s
ce
n
ar
io
s
i
n
cl
u
s
t
er
in
g
al
g
r
o
tih
m
s
.
B
a
n
o
et
a
l
[
17
]
cr
ea
ted
th
e
lar
g
e
s
ca
le
co
r
p
u
s
w
it
h
t
h
e
an
n
o
tatio
n
o
f
d
is
ea
s
e
n
a
m
e
s
t
h
at
tr
ai
n
th
e
p
r
o
b
ab
ilis
tic
n
eu
r
al
n
et
w
o
r
k
m
o
d
el.
T
h
e
y
e
m
p
lo
y
ed
t
h
e
co
n
te
x
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
A
n
I
mp
r
o
ve
d
S
imila
r
ity
Ma
tch
in
g
B
a
s
ed
C
lu
s
teri
n
g
F
r
a
mewo
r
k
fo
r
S
h
o
r
t a
n
d
S
en
te
n
ce
…
(
M.
Jo
h
n
B
a
s
h
a
)
553
r
an
k
b
ased
h
ier
ar
ch
ical
cl
u
s
t
er
in
g
m
eth
o
d
an
d
o
p
ti
m
al
r
u
le
f
il
ter
in
g
alg
o
r
it
h
m
to
r
em
o
v
e
t
h
e
u
n
w
an
te
d
s
p
ec
ial
ch
ar
ac
ter
s
i
n
th
e
d
atas
ets.
T
h
e
s
en
ten
ce
clu
s
ter
in
g
is
u
s
e
d
in
m
u
ltip
le
ap
p
licatio
n
s
s
u
c
h
as,
class
i
f
icatio
n
an
d
ca
teg
o
r
izatio
n
o
f
th
e
d
o
cu
m
e
n
t
s
,
au
to
m
at
ic
s
u
m
m
ar
y
g
e
n
er
atio
n
,
o
r
g
a
n
izi
n
g
th
e
d
o
cu
m
en
t
s
,
etc.
I
n
tex
t
p
r
o
ce
s
s
in
g
,
th
e
s
en
te
n
ce
clu
s
ter
i
n
g
is
u
s
ed
f
o
r
th
e
tex
t
m
in
in
g
p
r
o
ce
s
s
.
T
h
e
s
ize
o
f
th
e
cl
u
s
ter
is
u
n
iq
u
e
f
o
r
ea
ch
clu
s
ter
.
T
h
e
ex
is
t
in
g
s
en
te
n
ce
cl
u
s
ter
in
g
al
g
o
r
ith
m
s
cr
e
ate
m
u
l
tip
le
is
s
u
e
s
,
s
u
ch
as,
co
m
p
le
x
it
y
,
s
en
s
iti
v
it
y
,
i
n
s
tab
ilit
y
,
etc
.
C
o
m
p
ar
ed
to
th
e
s
en
te
n
ce
cl
u
s
ter
in
g
,
th
e
cl
u
s
ter
i
n
g
o
f
th
e
s
h
o
r
t
tex
ts
ar
e
v
er
y
d
i
f
f
ic
u
lt.
As
th
e
s
h
o
r
t
tex
t
s
in
th
e
co
m
m
er
cial
p
r
o
d
u
cts,
n
e
w
,
F
A
Q
s
an
d
s
cie
n
ti
f
ic
ab
s
tr
ac
ts
ar
e
w
id
el
y
u
s
ed
b
y
th
e
u
s
er
s
in
r
ea
l
lif
e,
t
h
e
clu
s
ter
i
n
g
o
f
th
e
s
h
o
r
t
tex
t
s
d
em
a
n
d
s
f
o
c
u
s
.
I
n
t
h
is
p
ap
er
,
th
e
p
r
o
p
o
s
ed
tex
t
b
ased
cl
u
s
ter
in
g
f
r
a
m
e
w
o
r
k
clu
s
ter
s
th
e
s
e
n
te
n
ce
s
a
s
w
ell
as th
e
s
h
o
r
t te
x
t
s
.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
i
s
e
x
ec
u
ted
till
t
h
e
d
u
p
licate
c
lu
s
te
r
s
ar
e
r
e
m
o
v
ed
.
A
f
ter
th
e
r
e
m
o
v
al
o
f
t
h
e
u
n
w
an
ted
w
o
r
d
s
,
t
h
e
p
r
o
p
o
s
e
d
s
y
s
te
m
c
h
ec
k
s
all
th
e
w
o
r
d
s
in
t
h
e
d
o
cu
m
en
t
f
o
r
f
in
d
i
n
g
t
h
e
ex
ac
t
w
o
r
d
.
T
h
e
s
i
m
ilar
ities
b
et
w
ee
n
th
e
s
e
n
t
en
ce
s
ar
e
u
s
ed
to
f
i
n
d
th
e
r
atio
o
f
th
e
s
i
m
ilar
it
y
o
f
th
e
w
o
r
d
s
.
Do
cu
m
e
n
t
clu
s
ter
i
n
g
is
a
n
au
to
m
atic
a
n
a
l
y
tic
p
r
o
ce
s
s
t
h
at
as
s
ig
n
s
d
o
c
u
m
e
n
t
s
to
u
n
k
n
o
w
n
ca
teg
o
r
ie
s
.
I
n
th
i
s
tas
k
,
o
n
l
y
th
e
in
h
er
e
n
t
s
tr
u
ctu
r
e
o
f
d
ata
is
co
n
s
id
er
ed
;
th
er
e
f
o
r
e,
it
is
m
o
r
e
d
i
f
f
icu
l
t
th
a
n
s
u
p
er
v
is
ed
tex
t
ca
teg
o
r
izatio
n
b
ec
au
s
e
n
o
in
f
o
r
m
atio
n
ab
o
u
t
co
r
r
ec
tly
ca
te
g
o
r
ized
ex
a
m
p
les
is
p
r
o
v
id
ed
in
ad
v
an
ce
.
T
o
o
v
er
co
m
e
th
i
s
d
if
f
ic
u
lt
y
,
i
n
th
i
s
p
ap
er
th
e
C
L
UDI
P
SO
b
ased
clu
s
ter
in
g
is
p
r
o
p
o
s
ed
.
T
h
e
k
ey
ad
v
a
n
ta
g
e
o
f
C
L
UDI
P
SO
is
th
e
cr
ea
tio
n
o
f
r
ea
l
n
u
m
b
er
v
ec
to
r
s
f
o
r
ea
ch
p
ar
ticle.
T
h
e
v
ec
to
r
s
r
ep
r
esen
t
th
e
s
ea
r
c
h
s
p
ac
e
d
e
f
in
ed
b
y
t
h
e
v
ar
iab
les co
r
r
esp
o
n
d
in
g
to
t
h
e
p
r
o
b
lem
to
s
o
lv
e.
T
h
e
r
em
ai
n
d
er
o
f
t
h
e
p
ap
er
is
s
y
s
te
m
at
ized
as
f
o
llo
w
s
:
Sectio
n
I
I
d
escr
ib
es
t
h
e
e
x
is
ti
n
g
te
x
t
clu
s
ter
i
n
g
tech
n
iq
u
e
s
.
Sectio
n
I
I
I
illu
s
tr
ates
th
e
p
r
o
p
o
s
ed
t
ex
t
b
ased
clu
s
ter
i
n
g
f
r
a
m
e
w
o
r
k
an
d
s
ec
t
io
n
I
V
d
escr
ib
es
th
e
p
er
f
o
r
m
a
n
ce
r
esu
lt
s
o
f
t
h
e
p
r
o
p
o
s
ed
tech
n
iq
u
e.
Sectio
n
V
ill
u
s
tr
ates
th
e
co
n
clu
s
io
n
o
f
t
h
i
s
p
ap
er
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
I
t is co
m
p
o
s
ed
o
f
f
o
llo
w
i
n
g
p
r
o
ce
s
s
es to
ac
h
ie
v
e
t
h
e
r
ed
u
cti
o
n
in
p
r
o
ce
s
s
i
n
g
ti
m
e
a
n
d
MS
E
.
a.
P
r
ep
r
o
ce
s
s
in
g
b.
Si
m
i
lar
it
y
C
o
m
p
u
tatio
n
c.
Vec
to
r
Data
Fo
r
m
at
io
n
d.
C
lu
s
ter
i
n
g
p
ar
ticle
e.
Mu
tatio
n
Fig
u
r
e
1
.
Ov
er
all
Flo
w
o
f
t
h
e
P
r
o
p
o
s
ed
T
ex
t B
ased
C
lu
s
ter
i
n
g
Fra
m
e
w
o
r
k
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
1
,
Feb
r
u
ar
y
2
0
1
7
:
55
1
–
558
554
Fig
u
r
e
1
s
h
o
w
s
th
e
w
o
r
k
f
lo
w
o
f
t
h
e
p
r
o
p
o
s
ed
tex
t
-
b
ased
clu
s
ter
i
n
g
f
r
a
m
e
w
o
r
k
th
at
i
n
clu
d
es
th
e
s
eq
u
en
tial
p
r
o
ce
s
s
es.
I
n
it
iall
y
,
th
e
d
ataset
is
p
ass
ed
to
th
e
p
r
ep
r
o
ce
s
s
in
g
b
lo
ck
w
h
er
e
t
h
e
d
ata
w
ar
e
h
o
u
s
e
en
tr
ies
a
n
d
t
h
e
tab
le
d
ec
lar
ati
o
n
ar
e
d
is
c
u
s
s
ed
in
d
etai
l.
T
h
en
,
t
h
e
co
s
i
n
e
s
i
m
ilar
it
y
b
et
wee
n
t
h
e
to
p
ics
a
n
d
d
o
cu
m
en
ts
i
n
t
h
e
d
ataset
is
co
m
p
u
ted
.
B
ased
o
n
th
e
s
i
m
ilar
i
t
y
v
a
lu
e
s
,
th
e
v
ec
to
r
s
t
h
at
r
ep
r
esen
t
t
h
e
d
ata
ar
e
co
m
p
u
ted
b
ased
o
n
th
e
r
an
k
i
n
g
v
al
u
es.
T
h
en
,
t
h
e
P
ar
ticle
Sw
ar
m
Op
ti
m
izatio
n
(
P
SO)
alg
o
r
ith
m
i
s
u
s
ed
as th
e
clu
s
ter
i
n
g
to
o
l to
f
in
d
t
h
e
o
p
tim
alit
y
.
2
.
1
.
P
re
pro
ce
s
s
ing
E
ac
h
v
a
lu
e
o
f
t
h
e
i
n
p
u
t
d
at
aset
is
p
r
ep
r
o
ce
s
s
ed
an
d
t
h
e
r
esu
lta
n
t
d
ata
s
et
is
s
to
r
ed
b
ac
k
i
n
t
h
e
d
atab
ase.
T
h
e
d
atab
ase
lo
a
d
u
tili
t
y
r
ea
d
s
th
e
u
s
er
-
p
r
o
v
id
ed
d
ata
an
d
s
to
r
es
th
e
m
in
t
h
e
tab
le.
T
h
e
in
p
u
t
R
eu
ter
’
s
d
ataset
co
n
tai
n
s
t
h
e
u
s
er
p
r
o
v
id
ed
d
ata.
T
h
e
d
atab
ase
lo
ad
u
tili
t
y
s
u
p
p
o
r
ts
f
o
u
r
d
if
f
er
e
n
t
f
o
r
m
a
ts
o
f
f
iles
.
B
e
f
o
r
e
th
e
d
ata
lo
ad
in
g
,
th
e
tab
le
m
u
s
t
b
e
d
ef
i
n
ed
.
T
h
e
d
ata
w
ar
eh
o
u
s
e
s
to
r
es
t
h
e
p
r
iv
ate
d
ata
an
d
also
m
ak
e
s
t
h
e
ed
g
e
d
ec
is
io
n
s
u
p
p
o
r
t s
y
s
te
m
.
T
h
e
k
e
y
ai
m
o
f
th
e
d
ata
w
ar
eh
o
u
s
e
is
to
co
llect
m
u
ltip
le
in
f
o
r
m
a
tio
n
f
r
o
m
v
ar
io
u
s
s
o
u
r
ce
s
t
h
at
f
o
llo
w
s
d
if
f
er
en
t p
lat
f
o
r
m
.
T
h
e
co
llected
v
ar
iab
le
d
ata
ar
e
u
n
ited
f
o
r
p
er
f
o
r
m
in
g
t
h
e
b
u
s
i
n
ess
d
ec
i
s
io
n
s
.
2
.
2
.
Si
m
ila
rit
y
Co
m
pu
t
a
t
io
n
B
ased
o
n
th
e
i
n
p
u
t
d
ata
s
et,
t
h
e
s
i
m
ilar
it
y
is
ca
lc
u
lated
.
F
u
r
t
h
er
d
ev
iatio
n
i
n
th
e
in
p
u
t
d
ata
s
et
ar
e
also
co
n
s
id
er
ed
to
r
ef
i
n
e
t
h
e
d
ata
i
n
t
h
e
f
iles
.
I
n
t
h
i
s
p
ap
er
,
th
e
v
ec
to
r
s
i
m
ilar
i
t
y
is
ac
co
m
p
lis
h
ed
u
s
i
n
g
t
h
e
co
s
i
n
e
s
i
m
ilar
it
y
.
C
o
s
i
n
e
s
i
m
ilar
it
y
is
u
s
ed
to
esti
m
ate
th
e
s
i
m
ilar
i
t
y
b
et
w
ee
n
t
h
e
v
ec
o
r
s
o
f
an
i
n
n
er
p
r
o
d
u
ct
s
p
ac
e
an
d
m
ea
s
u
r
es
th
e
co
s
in
e
o
f
th
e
an
g
le
b
et
w
ee
n
t
h
e
m
.
C
o
s
in
e
s
i
m
ilar
it
y
is
co
m
m
o
n
l
y
u
s
ed
f
o
r
th
e
p
o
s
iti
v
e
s
p
ac
e
w
h
o
s
e
o
u
tco
m
e
al
w
a
y
s
lie
s
b
et
w
ee
n
[
0
,
1
]
.
C
o
s
i
n
e
s
i
m
ilar
it
y
i
s
m
o
s
t
s
u
ited
f
o
r
t
h
e
h
i
g
h
-
d
i
m
e
n
s
io
n
al
p
o
s
iti
v
e
s
p
ac
es.I
ts
m
er
its
ar
e
u
s
ed
i
n
th
e
f
ield
o
f
d
ata
m
i
n
i
n
g
f
o
r
m
ea
s
u
r
in
g
th
e
co
h
e
s
io
n
b
et
wee
n
t
h
e
cl
u
s
ter
s
.
T
h
e
tech
n
iq
u
e
is
also
u
s
ed
to
m
ea
s
u
r
e
co
h
esio
n
w
it
h
in
cl
u
s
ter
s
in
th
e
f
ield
o
f
d
ata
m
in
in
g
.
T
h
e
C
o
s
in
e
Si
m
ilar
it
y
o
f
t
w
o
v
ec
to
r
s
d
1
an
d
d
2
is
ca
lcu
lated
as d
ep
icted
in
(
1
)
,
(
)
(
)
‖
‖
‖
‖
(
1
)
w
h
er
e,
(
)
[
]
[
]
[
]
[
]
‖
‖
(
[
]
[
]
2
.
3
.
Vec
t
o
r
Da
t
a
F
o
rm
a
t
io
n
T
h
e
ter
m
v
ec
to
r
is
a
n
alg
eb
r
ai
c
m
o
d
el
f
o
r
r
ep
r
esen
ti
n
g
te
x
t
d
o
cu
m
en
ts
as
v
ec
to
r
s
o
f
id
en
t
if
ier
s
.
I
t
i
s
u
s
ed
in
i
n
f
o
r
m
atio
n
f
ilter
i
n
g
,
in
f
o
r
m
at
io
n
r
etr
iev
al,
in
d
e
x
in
g
an
d
r
elev
a
n
c
y
r
an
k
i
n
g
s
.
E
ac
h
d
im
e
n
s
io
n
co
r
r
esp
o
n
d
s
to
a
s
ep
ar
ate
ter
m
.
I
f
a
ter
m
o
cc
u
r
s
in
t
h
e
d
o
cu
m
en
t,
its
v
al
u
e
in
th
e
v
ec
to
r
is
n
o
n
-
ze
r
o
o
th
er
w
i
s
e
ze
r
o
.
2
.
4
.
Clus
t
er
ing
P
a
rt
icle
T
h
e
clu
s
ter
i
n
g
o
f
t
h
e
v
ec
t
o
r
s
is
p
er
f
o
r
m
ed
u
s
i
n
g
t
h
e
P
SO
alg
o
r
ith
m
.
W
h
en
th
e
s
ize
a
n
d
d
i
m
en
s
io
n
al
it
y
o
f
th
e
d
ataset
is
lar
g
e,
th
e
tr
ad
itio
n
al
P
SO
i
s
n
o
t
a
b
est
o
p
tio
n
,
h
e
n
ce
i
n
th
is
p
ap
er
,
a
n
e
w
v
er
s
io
n
o
f
t
h
e
P
SO
n
a
m
ed
,
C
L
UDI
P
SO
i
s
p
r
o
p
o
s
ed
.
T
h
r
ee
s
p
ec
if
ic
c
h
ar
ac
ter
is
tic
s
o
f
t
h
e
C
L
UI
DI
P
SO
m
a
k
es
it
s
u
itab
le
f
o
r
h
an
d
li
n
g
t
h
e
la
r
g
er
d
atasets
.
T
h
e
ch
ar
ac
ter
i
s
tics
i
n
cl
u
d
e,
n
e
w
r
ep
r
esen
tati
o
n
o
f
p
ar
ticles
f
o
r
r
ed
u
cin
g
th
e
d
i
m
e
n
s
io
n
alit
y
,
r
ed
u
ce
th
e
co
m
p
u
tatio
n
al
ti
m
e
an
d
in
cr
ea
s
in
g
th
e
s
p
ee
d
o
f
th
e
s
ilh
o
u
e
tte
co
m
p
u
tatio
n
.
B
ased
o
n
th
e
s
i
m
ilar
it
y
d
is
ta
n
ce
th
r
e
s
h
o
ld
v
al
u
e,
o
p
ti
m
al
cl
u
s
ter
s
ar
e
g
e
n
er
a
ted
.
2
.
5
.
M
ut
a
t
io
n
T
h
e
m
u
ta
tio
n
p
r
o
ce
s
s
is
u
s
e
d
to
u
p
d
ate
th
e
p
ar
ticle’
s
p
o
s
itio
n
.
T
h
e
tr
ad
itio
n
al
P
SO
i
s
u
s
ed
f
o
r
s
o
lv
i
n
g
o
n
l
y
t
h
e
co
n
tin
u
o
u
s
p
r
o
b
lem
s
,
b
u
t
th
e
p
r
o
p
o
s
ed
m
u
tatio
n
p
r
o
ce
s
s
is
n
o
t
d
ep
en
d
en
t
o
n
th
e
p
o
s
itio
n
o
f
th
e
p
ar
ticles
an
d
f
u
r
th
er
at
ea
ch
iter
atio
n
,
t
h
e
p
o
s
itio
n
u
p
d
a
tin
g
p
r
o
ce
s
s
i
s
ca
r
r
ied
o
u
t
i
n
all
th
e
d
i
m
e
n
s
io
n
s
.
T
o
co
m
p
u
te
t
h
e
d
i
m
en
s
io
n
at
w
h
ic
h
th
e
p
ar
ticle
is
u
p
d
ated
th
e
f
o
llo
w
i
n
g
s
tep
s
ar
e
p
er
f
o
r
m
ed
.
Ste
ps
inv
o
lv
ed
in t
he
pro
po
s
ed
CL
UD
I
P
SO
ba
s
ed
m
uta
t
io
n
S
tep
1
:
A
ll t
h
e
d
i
m
en
s
io
n
s
o
f
t
h
e
v
elo
cit
y
v
ec
to
r
ar
e
n
o
r
m
al
ized
b
etw
ee
n
[
0
,
1
]
r
an
g
e.
S
tep
2
:
B
ased
o
n
[
1
8
]
th
e
r
an
d
o
m
n
u
m
b
er
i
s
ca
lcu
la
ted
S
tep
3
:
A
ll t
h
e
d
i
m
en
s
io
n
s
th
at
ar
e
ab
o
v
e
‘
r
’
ar
e
ch
o
s
en
i
n
t
h
e
p
o
s
iti
o
n
v
ec
to
r
an
d
u
p
d
ated
.
S
tep
4
:
Up
d
ated
C
lu
s
ter
is
p
r
o
v
id
ed
as th
e
r
es
u
lt.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
A
n
I
mp
r
o
ve
d
S
imila
r
ity
Ma
tch
in
g
B
a
s
ed
C
lu
s
teri
n
g
F
r
a
mewo
r
k
fo
r
S
h
o
r
t a
n
d
S
en
te
n
ce
…
(
M.
Jo
h
n
B
a
s
h
a
)
555
3.
RE
SU
L
T
S AN
D
AN
AL
Y
SI
S
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
p
r
o
p
o
s
ed
tex
t
b
ased
cl
u
s
ter
i
n
g
a
l
g
o
r
ith
m
i
s
co
m
p
ar
ed
w
ith
th
e
ex
is
t
in
g
Fu
zz
y
C
-
Me
a
n
s
(
F
C
M)
an
d
P
air
w
i
s
e
R
a
n
d
o
m
S
w
ap
(
P
R
S)
clu
s
ter
i
n
g
tech
n
iq
u
e
s
f
o
r
t
h
e
m
etr
ics,
s
u
c
h
as,
a.
Me
an
Sq
u
ar
e
E
r
r
o
r
(
MSE
)
b.
P
r
o
ce
s
s
in
g
T
i
m
e
c.
P
ea
k
Sig
n
al
No
is
e
R
atio
3
.
1
.
M
ea
n Squ
a
re
E
rr
o
r
(
M
SE
)
T
h
e
MSE
is
ca
lcu
lated
u
s
i
n
g
t
h
e
eq
u
atio
n
(
2
)
,
∑
∑
‖
‖
(
2
)
W
h
er
e,
K
in
d
icate
s
t
h
e
i
n
d
ica
to
r
f
u
n
ctio
n
.
T
h
e
D
d
en
o
te
s
t
h
e
n
u
m
b
er
o
f
o
b
j
ec
ts
an
d
th
e
E
d
en
o
tes
th
e
n
u
m
b
er
o
f
cl
u
s
ter
s
.
E
ac
h
o
b
j
ec
t
b
elo
n
g
s
to
t
h
e
cl
u
s
ter
w
it
h
t
h
e
m
i
n
i
m
u
m
E
u
clid
ea
n
d
is
tan
ce
to
t
h
e
ce
n
ter
ce
n
tr
o
id
.
F
ig
u
r
e
2
.
C
o
m
p
ar
is
o
n
o
f
t
h
e
MSE
f
o
r
th
e
p
r
o
p
o
s
ed
f
r
a
m
e
wo
r
k
w
it
h
t
h
e
F
C
M
an
d
P
R
S
Fig
u
r
e
2
s
h
o
w
s
th
e
co
m
p
ar
ati
v
e
an
al
y
s
is
o
f
p
r
o
p
o
s
ed
tex
t
-
b
ased
clu
s
ter
i
n
g
w
i
th
th
e
e
x
is
tin
g
F
C
M
an
d
P
R
S
tech
n
iq
u
es
r
eg
ar
d
i
n
g
th
e
MSE
v
a
lu
e
s
.
T
h
e
ef
f
ec
tiv
e
n
ess
o
f
an
y
p
r
o
to
co
l
p
r
o
p
o
s
ed
is
d
eter
m
i
n
ed
w
it
h
t
h
e
m
i
n
i
m
u
m
M
SE
v
al
u
es.
T
h
e
ex
is
tin
g
F
C
M
an
d
P
R
S
p
r
o
v
id
es
th
e
MSE
v
alu
e
s
o
f
1
9
1
an
d
1
4
0
f
o
r
s
in
g
le
cl
u
s
ter
.
T
h
e
y
p
r
o
v
id
e
1
7
an
d
1
6
f
o
r
1
5
clu
s
ter
s
.
B
u
t,
th
e
o
p
ti
m
al
cl
u
s
ter
i
n
g
-
b
a
s
ed
s
i
m
ilar
it
y
m
ea
s
u
r
e
m
e
n
t
i
n
p
r
o
p
o
s
ed
tex
t
-
b
ased
clu
s
ter
in
g
r
ed
u
ce
s
t
h
e
v
alu
e
s
to
1
1
0
an
d
1
5
f
o
r
s
in
g
le
an
d
1
5
clu
s
ter
s
r
esp
ec
tiv
el
y
.
T
h
e
co
m
p
ar
ati
v
e
an
al
y
s
is
b
et
w
ee
n
th
e
p
r
o
p
o
s
ed
T
B
C
w
it
h
t
h
e
ex
i
s
ti
n
g
P
SR
(
w
h
ich
p
r
o
v
id
es
m
i
n
i
m
u
m
v
al
u
e
s
)
s
tated
th
a
t
t
h
e
p
r
o
p
o
s
ed
T
B
C
ac
h
iev
ed
th
e
2
1
.
4
2
an
d
6
.
6
7
%
r
ed
u
ctio
n
in
MSE
co
m
p
ar
ed
to
P
SR
f
o
r
m
in
i
m
u
m
an
d
m
ax
i
m
u
m
cl
u
s
ter
s
r
esp
ec
ti
v
el
y
.
3
.
2
.
P
r
o
ce
s
s
ing
T
i
m
e
T
h
e
p
r
o
ce
s
s
in
g
t
i
m
e
o
f
t
h
e
p
r
o
p
o
s
ed
f
r
a
m
e
w
o
r
k
a
n
d
t
h
e
ex
is
tin
g
F
C
M
a
n
d
P
R
S
i
s
s
h
o
w
n
i
n
Fig
u
r
e
3
.
Fro
m
t
h
e
f
ig
u
r
e
it
’
s
o
b
v
io
u
s
t
h
at
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
p
r
o
d
u
ce
d
o
p
tim
al
P
SN
R
th
a
n
t
h
e
e
x
is
tin
g
clu
s
ter
i
n
g
tech
n
iq
u
e
s
.
Fig
u
r
e
3
s
h
o
w
s
th
e
co
m
p
ar
ati
v
e
an
al
y
s
is
o
f
p
r
o
p
o
s
ed
tex
t
-
b
ased
clu
s
ter
i
n
g
w
i
th
th
e
e
x
is
tin
g
F
C
M
an
d
P
R
S
tech
n
iq
u
es
r
eg
ar
d
in
g
th
e
p
r
o
ce
s
s
in
g
ti
m
e
v
al
u
es.
T
h
e
ef
f
ec
tiv
e
n
es
s
o
f
an
y
p
r
o
to
co
l
p
r
o
p
o
s
ed
is
d
eter
m
in
ed
w
it
h
th
e
m
in
i
m
u
m
p
r
o
ce
s
s
in
g
ti
m
e.
T
h
e
p
r
o
ce
s
s
in
g
ti
m
e
o
f
t
h
e
ex
i
s
ti
n
g
F
C
M
a
n
d
P
R
S
ar
e
7
3
an
d
1
7
s
ec
s
f
o
r
s
i
n
g
le
clu
s
ter
.
T
h
e
y
p
r
o
v
id
e
3
6
an
d
3
4
s
ec
s
f
o
r
1
5
clu
s
ter
s
.
B
u
t,
t
h
e
o
p
ti
m
a
l
clu
s
ter
in
g
-
b
ased
s
i
m
il
ar
it
y
m
ea
s
u
r
e
m
en
t
in
p
r
o
p
o
s
ed
tex
t
-
b
ased
clu
s
ter
in
g
r
e
d
u
ce
s
th
e
v
al
u
es
to
1
0
an
d
2
4
s
ec
s
f
o
r
s
i
n
g
le
an
d
1
5
clu
s
ter
s
r
esp
ec
ti
v
el
y
.
T
h
e
co
m
p
ar
ati
v
e
a
n
al
y
s
is
b
et
w
ee
n
th
e
p
r
o
p
o
s
ed
T
B
C
w
it
h
th
e
e
x
is
t
in
g
P
SR
(
w
h
ic
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
1
,
Feb
r
u
ar
y
2
0
1
7
:
55
1
–
558
556
p
r
o
v
id
es
m
in
i
m
u
m
v
al
u
es)
s
t
ated
th
at
th
e
p
r
o
p
o
s
ed
T
B
C
ac
h
iev
ed
th
e
4
1
.
1
7
an
d
2
9
.
4
1
%
r
e
d
u
ctio
n
in
p
r
o
ce
s
s
in
g
t
i
m
e
co
m
p
ar
ed
to
P
SR
f
o
r
m
i
n
i
m
u
m
a
n
d
m
a
x
i
m
u
m
cl
u
s
ter
s
r
esp
ec
ti
v
el
y
.
Fig
u
r
e
3
.
P
SNR
C
o
m
p
ar
is
o
n
Fo
r
th
e
P
r
o
p
o
s
ed
Me
th
o
d
an
d
E
x
is
ti
n
g
F
C
M
A
n
d
P
R
S M
et
h
o
d
3
.
3
.
P
e
a
k
Sig
na
l N
o
is
e
Ra
t
io
(
P
SNR)
T
h
e
P
SNR
is
th
e
r
atio
o
f
th
e
m
ax
i
m
u
m
p
o
s
s
ib
le
v
al
u
e
o
f
t
h
e
s
ig
n
al
an
d
t
h
e
p
o
w
er
o
f
d
is
to
r
tin
g
n
o
is
e
th
at
a
f
f
ec
t
s
t
h
e
q
u
alit
y
o
f
t
h
e
r
ep
r
esen
tatio
n
it i
s
ca
lcu
lated
b
y
th
e
f
o
llo
w
i
n
g
eq
u
a
tio
n
,
(
√
)
(
3
)
W
h
er
e,
d
en
o
tes
th
e
m
a
x
i
m
u
m
s
ig
n
al
v
alu
e
t
h
at
ex
is
ts
in
t
h
e
o
r
ig
in
al
d
ata.
Fig
u
r
e
4
s
h
o
w
s
th
e
P
SNR
co
m
p
ar
is
o
n
f
o
r
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
an
d
th
e
e
x
is
ti
n
g
FC
M
an
d
P
R
S
m
et
h
o
d
.
Fig
u
r
e
4
.
P
SNR
C
o
m
p
ar
is
o
n
f
o
r
t
h
e
P
r
o
p
o
s
ed
Me
th
o
d
a
n
d
E
x
is
t
in
g
F
C
M
a
n
d
P
R
S M
et
h
o
d
Fig
u
r
e
4
s
h
o
w
s
th
e
co
m
p
ar
ati
v
e
an
al
y
s
is
o
f
p
r
o
p
o
s
ed
tex
t
-
b
ased
clu
s
ter
i
n
g
w
i
th
th
e
e
x
is
tin
g
F
C
M
an
d
P
R
S
tech
n
iq
u
es
r
eg
ar
d
i
n
g
th
e
P
SN
R
v
alu
e
s
.
T
h
e
ef
f
e
ctiv
e
n
ess
o
f
a
n
y
p
r
o
to
co
l
p
r
o
p
o
s
ed
is
d
eter
m
i
n
ed
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
A
n
I
mp
r
o
ve
d
S
imila
r
ity
Ma
tch
in
g
B
a
s
ed
C
lu
s
teri
n
g
F
r
a
mewo
r
k
fo
r
S
h
o
r
t a
n
d
S
en
te
n
ce
…
(
M.
Jo
h
n
B
a
s
h
a
)
557
w
it
h
th
e
h
i
g
h
er
P
SNR
.
T
h
e
P
SNR
v
al
u
es
f
o
r
F
C
M
a
n
d
P
R
S
ar
e
2
5
.
3
an
d
2
6
.
7
d
B
f
o
r
s
in
g
le
cl
u
s
ter
.
T
h
e
y
p
r
o
v
id
e
3
5
.
7
an
d
3
6
.
1
d
B
f
o
r
1
5
clu
s
ter
s
.
B
u
t,
th
e
o
p
tim
al
clu
s
ter
i
n
g
-
b
ased
s
i
m
ilar
i
t
y
m
ea
s
u
r
e
m
e
n
t
i
n
p
r
o
p
o
s
ed
tex
t
-
b
ased
cl
u
s
ter
i
n
g
in
cr
ea
s
e
s
t
h
e
v
alu
e
s
to
2
8
.
1
an
d
3
7
.
5
d
B
f
o
r
s
in
g
l
e
an
d
1
5
clu
s
ter
s
r
esp
ec
tiv
el
y
.
T
h
e
co
m
p
ar
ati
v
e
an
al
y
s
is
b
et
w
ee
n
th
e
p
r
o
p
o
s
ed
T
B
C
w
it
h
t
h
e
ex
i
s
ti
n
g
P
SR
(
w
h
ich
p
r
o
v
id
es
m
ax
i
m
u
m
v
alu
e
s
)
s
tated
t
h
at
t
h
e
p
r
o
p
o
s
ed
T
B
C
ac
h
iev
ed
th
e
4
.
9
8
an
d
3
.
7
3
%
i
m
p
r
o
v
e
m
e
n
t
i
n
P
SNR
v
al
u
e
s
co
m
p
ar
ed
to
P
SR
f
o
r
m
i
n
i
m
u
m
an
d
m
ax
i
m
u
m
clu
s
ter
s
r
esp
ec
tiv
el
y
.
4.
CO
NCLU
SI
O
N
T
ex
t
clu
s
ter
i
n
g
is
t
h
e
p
r
o
ce
s
s
o
f
g
r
o
u
p
i
n
g
th
e
lar
g
e
a
m
o
u
n
t
o
f
in
f
o
r
m
at
io
n
i
n
to
m
ea
n
i
n
g
f
u
l
clu
s
ter
s
.
E
x
is
ti
n
g
FC
M
an
d
P
R
S
cl
u
s
t
er
in
g
tech
n
iq
u
es
ar
e
u
s
ed
f
o
r
clu
s
ter
in
g
th
e
tex
ts
in
th
e
d
o
cu
m
e
n
t.
B
u
t,
t
h
es
e
m
et
h
o
d
s
d
o
n
o
t
p
r
o
d
u
ce
an
o
p
ti
m
al
p
r
o
ce
s
s
i
n
g
t
i
m
e,
p
ea
k
s
ig
n
al
n
o
is
e
r
atio
an
d
m
ea
n
s
q
u
ar
e
er
r
o
r
v
alu
es,
h
en
ce
i
n
t
h
is
p
ap
er
an
ef
f
icie
n
t
tex
t
b
ased
clu
s
ter
in
g
f
r
a
m
e
wo
r
k
is
p
r
o
p
o
s
ed
to
clu
s
ter
th
e
t
ex
t
d
o
cu
m
e
n
t
s
th
a
t
co
n
tain
s
b
o
th
t
h
e
s
e
n
ten
ce
s
an
d
s
h
o
r
t
tex
t
s
.
I
n
itial
l
y
,
t
h
e
d
ataset
is
p
r
ep
r
o
ce
s
s
ed
to
r
em
o
v
e
th
e
n
o
i
s
e,
th
e
n
t
h
e
s
i
m
ilar
it
y
b
et
w
ee
n
t
h
e
w
o
r
d
s
is
ca
lc
u
lated
u
s
in
g
th
e
co
s
i
n
e
s
i
m
ilar
it
y
.
B
ased
o
n
t
h
e
co
m
p
u
ted
s
i
m
ilar
it
y
,
t
h
e
v
ec
to
r
d
ata
is
g
e
n
er
ated
.
T
h
e
v
ec
to
r
d
ata
is
t
h
en
clu
s
ter
ed
u
s
in
g
t
h
e
C
L
UI
DI
P
SO
tec
h
n
iq
u
e.
T
o
o
p
tim
ize
t
h
e
clu
s
t
er
s
,
m
u
ta
tio
n
p
r
o
ess
is
d
e
p
lo
y
ed
.
T
h
e
m
u
ta
tio
n
p
r
o
ce
s
s
is
r
ep
ea
ted
till
a
n
o
p
ti
m
al
cl
u
s
ter
is
o
b
tain
ed
.
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
tex
t
b
ased
clu
s
ter
i
n
g
f
r
a
m
e
w
o
r
k
is
co
m
p
ar
ed
w
i
th
t
h
e
ex
is
t
i
n
g
F
C
M
an
d
P
R
S
clu
s
ter
i
n
g
m
e
th
o
d
s
.
W
h
e
n
co
m
p
ar
ed
to
th
e
e
x
i
s
t
i
n
g
m
et
h
o
d
s
,
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
r
ed
u
ce
d
th
e
p
r
o
ce
s
s
in
g
ti
m
e
an
d
M
SN
v
al
u
es
an
d
i
n
c
r
ea
s
ed
th
e
P
SN
R
v
al
u
e.
T
h
u
s
o
u
r
tex
t
b
ased
cl
u
s
ter
i
n
g
f
r
a
m
e
w
o
r
k
is
p
r
o
v
ed
to
b
e
b
etter
th
an
th
e
e
x
is
t
in
g
cl
u
s
ter
in
g
FC
M
a
n
d
P
R
S
m
et
h
o
d
s
.
RE
F
E
R
E
NC
E
S
[1
]
A
.
G
ra
n
a
d
o
s,
K.
K
o
ro
u
tch
e
v
,
a
n
d
F
.
d
e
B
o
rja
Ro
d
rig
u
e
z
,
"
Disc
o
v
e
rin
g
Da
ta
S
e
t
Na
tu
re
t
h
ro
u
g
h
A
lg
o
rit
h
m
ic
Clu
ste
r
in
g
Ba
se
d
o
n
S
tri
n
g
Co
m
p
re
ss
io
n
"
,
IEE
E
T
r
a
n
s
a
c
ti
o
n
s
o
n
Kn
o
wled
g
e
a
n
d
Da
t
a
E
n
g
i
n
e
e
rin
g
,
v
o
l
.
2
7
,
p
p
.
699
-
7
1
1
,
2
0
1
5
.
[2
]
S
.
J.
L
e
e
a
n
d
J.Y.
Jia
n
g
,
"
M
u
lt
il
a
b
e
l
T
e
x
t
Ca
teg
o
riza
ti
o
n
Ba
se
d
o
n
F
u
z
z
y
R
e
lev
a
n
c
e
Clu
ste
rin
g
"
,
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
F
u
zz
y
S
y
ste
ms
,
v
o
l.
2
2
,
p
p
.
1
4
5
7
-
1
4
7
1
,
2
0
1
4
.
[3
]
T
.
Wei,
Y.
L
u
,
H.
Ch
a
n
g
,
Q.
Zh
o
u
,
a
n
d
X
.
Ba
o
,
"
A
s
e
m
a
n
ti
c
a
p
p
r
o
a
c
h
f
o
r
tex
t
c
lu
ste
rin
g
u
sin
g
W
o
rd
Ne
t
a
n
d
lex
ica
l
c
h
a
in
s"
,
Exp
e
rt
S
y
ste
ms
wit
h
A
p
p
l
ica
ti
o
n
s,
v
o
l.
4
2
,
p
p
.
2
2
6
4
-
2
2
7
5
,
2
0
1
5
.
[4
]
T
.
P
e
n
g
a
n
d
L
.
L
iu
,
"
A
n
o
v
e
l
in
c
re
m
e
n
tal
c
o
n
c
e
p
tu
a
l
h
iera
rc
h
ica
l
tex
t
c
lu
ste
rin
g
m
e
th
o
d
u
si
n
g
CF
u
-
tree
"
,
Ap
p
li
e
d
S
o
ft
Co
mp
u
ti
n
g
,
v
o
l.
2
7
,
p
p
.
2
6
9
-
2
7
8
,
2
0
1
5
.
[5
]
M
.
Yu
a
n
a
n
d
Y.
S
h
i,
"
T
e
x
t
Clu
st
e
rin
g
Ba
se
d
o
n
a
Div
id
e
a
n
d
M
e
rg
e
S
trate
g
y
"
,
Pro
c
e
d
ia
Co
m
p
u
te
r
S
c
ien
c
e
,
v
o
l.
5
5
,
p
p
.
8
2
5
-
8
3
2
,
2
0
1
5
.
[6
]
K.
K.
Bh
a
rti
a
n
d
P
.
K.
S
in
g
h
,
"
A
t
h
re
e
-
sta
g
e
u
n
su
p
e
rv
ise
d
d
im
e
n
sio
n
re
d
u
c
ti
o
n
m
e
th
o
d
f
o
r
tex
t
c
lu
ste
rin
g
"
,
J
o
u
rn
a
l
o
f
Co
m
p
u
t
a
ti
o
n
a
l
S
c
ien
c
e
,
v
o
l.
5
,
p
p
.
1
5
6
-
1
6
9
,
2
0
1
4
.
[7
]
W
.
S
o
n
g
,
J.Z
.
L
ian
g
,
a
n
d
S
.
C.
P
a
rk
,
"
F
u
z
z
y
c
o
n
tro
l
G
A
w
it
h
a
n
o
v
e
l
h
y
b
rid
se
m
a
n
ti
c
si
m
il
a
ri
t
y
s
trate
g
y
f
o
r
tex
t
c
lu
ste
rin
g
"
,
In
fo
rm
a
ti
o
n
S
c
ie
n
c
e
s,
v
o
l.
2
7
3
,
p
p
.
1
5
6
-
1
7
0
,
2
0
1
4
.
[8
]
L
.
G
o
n
g
,
J.
Zen
g
,
a
n
d
S
.
Zh
a
n
g
,
"
T
e
x
t
stre
a
m
c
lu
ste
rin
g
a
lg
o
rit
h
m
b
a
se
d
o
n
a
d
a
p
ti
v
e
f
e
a
tu
re
se
l
e
c
ti
o
n
"
,
Exp
e
rt
S
y
ste
ms
wit
h
Ap
p
li
c
a
ti
o
n
s,
v
o
l
.
3
8
,
p
p
.
1
3
9
3
-
1
3
9
9
,
2
0
1
1
.
[9
]
M
.
Ya
o
,
D.
P
i,
a
n
d
X
.
Co
n
g
,
"
Ch
in
e
se
tex
t
c
lu
ste
rin
g
a
lg
o
rit
h
m
b
a
se
d
k
-
m
e
a
n
s"
,
Ph
y
sic
s
Pro
c
e
d
ia
,
v
o
l.
3
3
,
p
p
.
301
-
3
0
7
,
2
0
1
2
.
[1
0
]
Y.
S
.
L
in
,
J.
Y.
Jia
n
g
,
a
n
d
S
.
J.
L
e
e
,
"
A
si
m
il
a
rit
y
m
e
a
su
re
f
o
r
te
x
t
c
las
sif
ic
a
ti
o
n
a
n
d
c
lu
ste
rin
g
"
,
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
K
n
o
wle
d
g
e
a
n
d
Da
t
a
E
n
g
i
n
e
e
rin
g
,
v
o
l.
2
6
,
p
p
.
1
5
7
5
-
1
5
9
0
,
2
0
1
4
.
[1
1
]
G
.
L
iu
,
Y.
W
a
n
g
,
T
.
Zh
a
o
,
a
n
d
D.
L
i,
"
Re
s
e
a
rc
h
o
n
th
e
p
a
ra
ll
e
l
tex
t
c
lu
ste
rin
g
a
lg
o
rit
h
m
b
a
se
d
o
n
th
e
se
m
a
n
ti
c
t
re
e
"
,
in
6
th
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Co
mp
u
ter
S
c
ien
c
e
s
a
n
d
Co
n
v
e
rg
e
n
c
e
In
fo
rm
a
ti
o
n
T
e
c
h
n
o
lo
g
y
(
ICCIT
)
,
2
0
1
1
,
p
p
.
4
0
0
-
4
0
3
.
[1
2
]
C.
L
i,
Y.
T
a
n
,
a
n
d
J.
Ko
n
g
,
"
A
n
M
a
h
a
lan
o
b
is
d
istan
c
e
s
b
a
se
d
tex
t
c
lu
ste
rin
g
a
lg
o
rit
h
m
"
,
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
A
u
t
o
ma
ti
c
C
o
n
t
r
o
l
a
n
d
Arti
fi
c
i
a
l
In
telli
g
e
n
c
e
(
ACA
I
2
0
1
2
),
p
p
.
4
6
5
-
4
6
8
,
2
0
1
2
.
[1
3
]
S
.
H.
Ng
u
y
e
n
,
W
.
S
w
ieb
o
d
a
,
a
n
d
H.S
.
Ng
u
y
e
n
,
"
On
se
m
a
n
ti
c
e
v
a
lu
a
ti
o
n
o
f
tex
t
c
lu
ste
rin
g
a
lg
o
rit
h
m
s
"
,
in
IE
E
E
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
Gr
a
n
u
l
a
r Co
m
p
u
ti
n
g
(
Gr
C)
,
2
0
1
4
,
p
p
.
2
2
4
-
2
2
9
.
[1
4
]
M
.T
.
Ga
o
a
n
d
B.
J
.
W
a
n
g
,
"
T
e
x
t
c
lu
ste
rin
g
e
n
se
m
b
le
b
a
se
d
o
n
g
e
n
e
ti
c
a
lg
o
rit
h
m
s
"
,
in
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
S
y
ste
ms
a
n
d
In
f
o
rm
a
ti
c
s (
ICS
AI)
,
2
0
1
2
,
p
p
.
2
3
2
9
-
2
3
3
2
.
[1
5
]
K.
S
h
i
a
n
d
L
.
L
i,
"
Hig
h
p
e
r
f
o
rm
a
n
c
e
g
e
n
e
ti
c
a
lg
o
rit
h
m
b
a
se
d
t
e
x
t
c
lu
ste
rin
g
u
sin
g
p
a
rts
o
f
sp
e
e
c
h
a
n
d
o
u
tl
ier
e
li
m
in
a
ti
o
n
"
,
Ap
p
li
e
d
I
n
telli
g
e
n
c
e
,
v
o
l.
3
8
,
p
p
.
5
1
1
-
5
1
9
,
2
0
1
3
.
[1
6
]
R.
K
.
V
e
n
k
a
tes
h
a
n
d
N.I.
E.
M
.
I
n
d
ia,
"
L
e
g
a
l
Do
c
u
m
e
n
ts
Clu
ste
rin
g
a
n
d
S
u
m
m
a
riz
a
ti
o
n
u
sin
g
Hie
r
a
rc
h
ica
l
L
a
ten
t
Dirich
let
A
ll
o
c
a
ti
o
n
"
,
IAE
S
I
n
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
ia
l
I
n
te
ll
ig
e
n
c
e
(
IJ
-
AI)
,
v
o
l.
2
,
p
p
.
2
7
-
3
5
,
2
0
1
3
.
[1
7
]
S
.
Ba
n
o
,
K.L
.
Un
iv
e
rsity
,
K.R.
Ra
o
,
a
n
d
E.
S
r
i
P
ra
k
a
sh
Co
ll
e
g
e
o
f
,
"
P
a
rti
a
l
C
o
n
tex
t
S
im
il
a
rit
y
o
f
Ge
n
e
/P
r
o
tein
s
in
L
e
u
k
e
m
ia
Us
in
g
Co
n
tex
t
Ra
n
k
Ba
se
d
Hie
ra
rc
h
ica
l
Clu
ste
rin
g
A
lg
o
rit
h
m
"
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
Co
mp
u
ter
E
n
g
in
e
e
rin
g
(
IJ
ECE
),
v
o
l.
5
,
p
p
.
4
8
3
-
4
9
0
,
2
0
1
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
1
,
Feb
r
u
ar
y
2
0
1
7
:
55
1
–
558
558
[1
8
]
X
.
Hu
,
R.
C.
E
b
e
rh
a
rt,
a
n
d
Y.
S
h
i,
"
S
w
a
r
m
in
telli
g
e
n
c
e
f
o
r
p
e
r
m
u
tatio
n
o
p
ti
m
iza
ti
o
n
:
a
c
a
se
stu
d
y
o
f
n
-
q
u
e
e
n
s
p
ro
b
lem
"
,
in
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
0
3
IEE
E
S
w
a
rm
In
telli
g
e
n
c
e
S
y
mp
o
siu
m,
2
0
0
3
.
S
IS
'0
3
.
,
2
0
0
3
,
p
p
.
2
4
3
-
2
4
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.