TELKOM
NIKA
, Vol.12, No
.4, Dece
mbe
r
2014, pp. 10
17~102
2
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v12i4.434
1017
Re
cei
v
ed Au
gust 24, 20
14
; Revi
sed
No
vem
ber 1, 20
14; Accepted
No
vem
ber 1
7
,
2014
An Improved Entity Similarity Measur
e
ment Method
Gang lv
*
1,
2
, Chen
g
Zhen
g
2
, Sheng-bing Chen
3
1
Ke
y
of Intelli
g
ent Comp
utin
g and Si
gn
al Pro
c
essin
g
, Minist
r
y
of Educ
atio
n
,
Anhui Un
ivers
i
t
y
,
Hefei An
hui
23
003
9, Chi
n
a
2
Ke
y
La
borat
o
r
y
of Net
w
ork
a
nd Intell
ig
ent Information Proc
essin
g
, Hefei U
n
iversit
y
,
Hefei An
hui
23
060
1, Chi
n
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: lvgan
g@hfu
u
.
edu.cn
1
, chen
gzhe
ng@
ah
u.e
du.cn
2
, chens
b
@
hfuu.e
du.cn
3
A
b
st
r
a
ct
To facilitate t
he i
n
tegr
ation
of lear
ni
ng
resourc
e
s c
a
tegor
i
z
e
d
un
der d
i
fferent
ontol
og
y
repres
entati
o
n
s
, the tec
hni
q
ues
of
o
n
tol
o
gy
ma
ppi
ng
c
an
be
ap
pl
ied
.
T
houg
h
man
y
al
gorith
m
s
and
system
s have
been proposed for ontolog
y mapping, they
do not hav
e an
aut
om
atic weighting strategy on
class features
to automat
e the ontol
ogy
ma
pp
ing pr
oc
ess. A nov
el
meth
od of co
mp
utin
g the fe
ature
w
e
ights is pro
pose
d
. By feature se
ma
ntic
ana
lysis, the d
i
fferent entiti
e
s
simi
larity ca
lc
ulati
on
mo
del
an
d
w
e
ight calc
ul
ati
on
mo
del w
e
r
e
defin
ed. T
he r
e
sults sh
ow
that it makes the
ontol
ogy ma
pp
i
ng
pr
ocess mo
re
auto
m
atic w
h
il
e retain
ing s
a
ti
sfying accur
a
c
y
. Improve o
n
tolo
gy mapp
in
g effectiveness.
Ke
y
w
ords
: se
ma
ntics of features, ontol
ogy
ma
pp
ing, featu
r
e w
e
ight
1. Introduc
tion
Being on
e of the best in
st
rume
nt of kn
owle
dge p
r
e
s
entation an
d
the ba
sis of
semantic
w
e
b
tec
h
n
o
l
og
ie
s
,
on
to
lo
gy is
ma
in
ly de
sc
r
i
be
d
w
i
th RD
F
(R
es
our
c
e De
sc
r
i
p
t
io
n
Fr
a
m
ew
ork
)
and O
W
L
(
On
tology Web
L
angu
age
) rel
eased by W3
C be
side
s CY
CL, DO
GMA,
F-Logi
c a
nd
the
like
develop
e
d
an
d u
s
e
d
b
y
other
organ
ization
s
.
Currently, domain
ontolo
g
y ha
s bee
n a
pplied
in
many field
s
su
ch
as a
r
tificial i
n
telligen
ce,
so
ftwa
r
e
engin
eeri
ng, libra
ry
scie
nce
an
d sema
n
t
ic
web[1],[2]. The res
o
urc
e
s repres
ented by different
ontologies
in different fields
would
be
integrate
d
a
n
d
cla
s
sified vi
a ontolo
g
y m
appin
g
. As
th
e key fa
ctor o
f
ontology ma
pping, the
ent
ity
simila
rity me
asu
r
em
ent can be
gen
erally divided
i
n
to thre
e m
e
thod
s with
different b
a
ses,
namely, term
inology,
stru
cture
a
nd
se
mantics. Be
si
des, th
e p
r
o
c
ess of
mappi
ng
can
al
so
be
c
l
as
s
i
fied into three types
,
namely, man
ual, s
e
mi-aut
omatic
and automatic
[3],[4].
Influenced by fac
t
ors
such as
c
l
ass
i
fic
a
tion
sc
heme, repres
entation language, and
backg
rou
nd
kno
w
le
dge, t
he o
n
tology i
n
a
same fie
l
d may
app
e
a
r
quite
different. Th
erefo
r
e,
whe
n
stu
d
yin
g
the i
s
sue of
ontology m
a
pping, b
e
si
de
s the
re
sea
r
ches
on th
e cl
ass mat
c
hin
g
of
different entities, the features
(i.e. relations
)
between them al
so matters
. Generally, the s
y
s
t
em
of ontology m
apping possesses two st
rat
egies, nam
el
y, single strategy and
multi
-
strategy [5],[6].
Whe
n
multi
-
strategy is ad
opted, differe
nt sim
ila
rity
measurement
s shall be combine
d
into
a
singl
e o
ne
properly.
Du
rin
g
the
process, most
wei
ght
dist
ribution
o
f
re
sou
r
ces is mad
e
b
a
sed
on
the experi
e
n
c
e
s
or
expe
riments of th
e expert
s
no
wad
a
ys, whil
e this meth
o
d
rem
a
in
s time-
con
s
umi
ng a
nd un
stable
whe
n
used in
Web resource
s re
presente
d
by different ontologi
es[7],
[8].
Ontology m
a
pping i
s
a
ki
nd of p
r
o
c
e
s
s in
wh
i
c
h
the e
n
tity of the source
o
n
tology
(incl
udin
g
cla
ss a
nd features)
woul
d be mappe
d and
rep
r
e
s
ente
d
by a target ontology, and the
simila
rity me
asu
r
em
ent al
so in
clu
d
e
s
o
t
her relat
ed e
n
tities
o
w
ing to
ce
rtain rel
a
tional
featu
r
es
besi
d
e
s
the entity itself. A
con
c
e
p
t of “universa
lity” among cl
asse
s in ont
ology rep
r
e
s
entatio
n
is
prop
osed in t
h
is the
s
i
s
: if a feature p
o
ssesse
s
a
hi
gh
universality, the pa
rtition d
egre
e
of a
cla
s
s
woul
d be
com
e
low a
nd th
e simila
rity would thu
s
re
main indi
stin
guishable, n
a
m
ely, the larger a
feature’
s univ
e
rsality becomes, the sm
aller t
he
wei
ght will get. And the following
comes
the
detailed expl
anation
s
.
2. Semantics
Featur
es
Since the ont
ology po
ssesse
s many fea
t
ure ty
pes
su
ch a
s
tags, a
nnotation
s
, attributes,
relation
s (p
arent cla
ss an
d sub
c
la
ss)
and exampl
e
s
, the distinctive feature among entitie
s is
calle
d “uni
qu
ene
ss” [9],[10]. As a hypothesi
s
, if t
he
ontology of a feature is un
ique, mean
while
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 4, Dece
mb
er 201
4: 101
7 – 1022
1018
there i
s
a cl
a
ss
with same
feature
s
in a
nother
ontolo
g
y, then we consi
der th
e a
bove ontolo
g
i
e
s
equal
to e
a
ch
othe
r. Just
a
s
we
ca
n
easily di
sting
u
ish
hum
an
bein
g
s
wh
en th
ey we
re
put i
n
to
a
grou
p
of ani
mal by
notici
ng the
featu
r
e of
“t
hou
ght
”, si
nce they
are the
o
n
ly spe
c
ie
s who
possess the
ability of thinking. On
the
contrary, since they maintai
n
the same features, it is hard
to distingui
sh
them whe
n
in
a cro
w
d.
This th
esi
s
d
e
fines
2
1
,
c
c
Com
f
as two differe
nt se
mantic fo
rm
s
whi
c
h rep
r
e
s
ent two
feature
-
ba
se
d ontologi
es with di
ffere
nt feature types in relate
d sema
ntics similaritie
s
.
For
instan
ce, rel
a
ted se
manti
cs fo
r the string ty
pe “t
ag” a
nd “an
notation” mi
g
h
t be a set of
synonym
s
, while the relati
onal se
manti
cs of a re
late
d feature mig
h
t be a set of classes
whi
c
h
c
o
nnec
t via certain
relations
. If value(c
,f) is
defin
ed to
expre
s
s Feat
ure f’
s value
of Cla
ss
c, a
nd
sem
(
f,c) t
o
e
x
press
Featu
r
e f’s semanti
c
a
s
so
ciat
ed
value of
Cla
s
s
c, the fo
rm
ula for the va
lue
of Feature f, ontology c
1
a
nd c
2
and
2
1
,
c
c
Com
f
ca
n be define
d
as follo
ws:
2
1
2
1
2
1
,
,
,
,
,
c
f
sem
c
f
sem
c
f
sem
c
f
sem
c
c
Com
f
(1
)
Among whi
c
h
sem
(
f,c
1
) an
d sem(f,c
2
) a
r
e
re
spe
c
tively the syn
onyms of val
ue(c
1
,f) and
value(c
2
,f). Bes
i
des
,
the s
i
milarities
bet
wee
n
bin
d
ing
pro
perty f, c
1
and
c
2
ca
n al
so
be d
e
fined
a
s
follows
:
otherwise
f
c
value
f
c
value
if
c
c
Com
f
0
,
,
1
,
2
1
2
1
(2
)
More
over, a
s
for the value
of relational
featur
e
s
such
as “p
are
n
t cl
ass”, “sub
cla
s
s” an
d
“exampl
e
”, it can be
con
s
ide
r
ed a
s
a
colle
ction
of
ontologie
s
whi
c
h ori
g
ina
t
e from a ce
rtain
feature. And the relatio
nal feature
s
of c
1
and
c
2
could t
hus b
e
define
d
as the follo
ws:
f
c
value
f
c
value
f
c
value
f
c
value
c
c
com
f
,
,
,
,
,
2
1
2
1
2
1
(3)
The valu
e of
co
nceptual f
eature
2
1
,
c
c
Com
f
dra
w
n from
the a
bove
calculat
ion
can
be used to ca
lculate the va
lue of weig
ht of a feat
ure. If we define O
as the ontol
o
g
y, C for a se
t
of entities
whi
c
h
belo
n
g
s
to
O, F
for a
group
of feat
u
r
es of C,
which
in
clud
e “tag
s”, “an
notatio
ns”,
“pa
r
ent cl
ass”, “bind
s
”, “re
l
ations
”, “exa
mples” an
d the like, the
2
1
,
c
c
Com
f
definition of
a
feature is a
s
f
o
llows:
2
,
)
,
(
n
c
c
com
CM
C
c
c
j
i
f
f
j
i
(4)
Among
which
n rep
r
e
s
ent
s the amou
nt of cla
s
ses in
C; and
c
i
, c
j
a
r
e the
cla
s
se
s of C,
the weig
ht of Feature f coul
d thus be d
e
fined a
s
the follows:
f
f
CM
W
1
(5
)
3. Similarit
y
Measur
e
men
t
If the weight of the entity
feature
s
re
pre
s
ent
e
d
by the two ontologi
es is worke
d
out, the
simila
rity me
asu
r
em
ent of
different
cla
s
ses
ca
n
be
cal
c
ulate
d
b
y
integratin
g
variou
s feat
ure
weig
ht, for the simila
rity measure
m
ent
of clas
s an
d feature inte
r influen
ce
s ea
ch othe
r du
ri
ng
the pro
c
e
ss[1
1
],[12]. Since a class is d
e
scrib
ed by
a set of feature
s
, the sim
ilarity of feature
s
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
An Im
proved
Entity Sim
ilarity
Mea
s
urem
ent Method (Gang Iv)
1019
sho
u
ld
be t
a
ke
n into
a
c
cou
n
t when
doing
si
mil
a
rity measure
m
ent. When
ma
ke
ontol
ogy
mappin
g
, iterative algorith
m
would b
e
a
dopted in thi
s
thesis.
Definition
s: o
r
iginal
ontol
o
g
y
1
1
1
,
F
C
O
, target
on
tology
2
2
2
,
F
C
O
, c
1
an
d
c
2
f
o
r cla
ss
colle
ct
ion,
F
1
and F
2
for feature
colle
ction.
1
i
e
and
2
j
e
for entities
,
and the c
l
as
ses
and
feature
s
also
belong
s to their ontol
ogie
s
O
1
and O
2
. In order to benefit the expressio
n
of the
aforem
ention
ed al
gorith
m
, relate
d va
ria
b
les are d
e
fin
ed a
s
foll
ows:
2
1
,
j
i
k
e
e
Sim
for the entity,
and the si
mi
larity weight
for
1
j
e
and
2
i
e
would be
worked o
u
t after applying th
e iterative
algorith
m
for k times a
nd it would al
so b
e
rep
r
e
s
ente
d
by
2
1
,
j
i
k
e
e
ISim
.
2
,
,
,
1
2
2
1
2
1
i
j
k
j
i
k
j
i
k
e
e
Sim
e
e
Sim
e
e
ISim
(6
)
The entity
1
i
e
whi
c
h d
e
scri
bed by a
se
t of feature
colle
ction
ca
n be d
e
fined
as
1
1
2
1
1
1
,
,
)
(
i
l
i
i
i
e
f
e
f
e
f
e
F
, among wh
ich
1
F
f
t
,
l
t
,
1
. Another entity which
descri
bed
by
a se
t
of
feat
ure
colle
ction can
also b
e
defin
ed
as
2
2
2
2
1
2
,
,
)
(
j
m
j
j
j
e
g
e
g
e
g
e
F
, among which
2
F
g
t
,
.
,
1
m
t
After ap
plying the
iterative alg
o
rithm
2
1
2
1
,
max
,
m
i
k
m
j
i
k
e
e
ISim
e
e
ISim
for
k time
s, an
adju
s
ted
result
2
1
j
i
k
e
e
A
can b
e
wo
rked out via cal
c
ulatio
n. In order to calcula
t
e
2
1
1
,
j
i
k
e
e
Sim
, we define
s
m
k
k
k
g
A
g
A
g
A
VF
,
),
(
,
2
1
to adjus
t
2
j
e
F
to
1
i
e
F
. And related
Formul
a 7
is
as
follows
:
1
1
1
2
1
1
2
1
1
,
,
,
,
,
i
i
e
F
f
f
e
F
VF
f
k
j
i
k
k
k
f
j
i
k
w
f
A
e
value
f
e
value
SIM
f
A
f
Sim
w
e
e
Sim
Among which
the attribute value of SIM
k
is based on i
t
s type:
(1)
If X and Y are not in the sa
me type, then SIM
k
(X,Y)=0
(2)
If X and Y are in the sa
m
e
type su
ch
as “ch
a
ra
cte
r
type” or “nu
m
eri
c
type” a
nd X=Y, the
SIM
k
(X,Y)=1, otherwise:
Y
f
sem
X
f
sem
Y
f
sem
X
f
sem
Y
X
SIM
k
,
,
,
,
,
(3)
If X and
Y
are both entity sets, then:
Y
X
e
e
Sim
Y
X
SIM
X
e
k
Y
e
k
,
max
,
max
,
1
2
2
1
As is sho
w
n
in Fig
u
re
1
the ontol
og
y re
p
r
esentat
ion of
synon
yms, the formula of
simila
rity mea
s
ureme
n
t of
“Book”, an
ent
ity in t
he
so
urce
ontolo
g
y a
nd the
on
e i
n
target
ontolo
g
y
is
as
follows
:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 4, Dece
mb
er 201
4: 101
7 – 1022
1020
class
er
k
k
k
k
erclass
r
k
W
r
A
Book
value
r
Book
value
SIM
r
A
r
Sim
W
Book
Book
Sim
_
sup
1
02
01
1
sup
02
01
1
,
,
,
,
class
sub
label
class
er
W
W
W
OnSubClass
Similarity
OnLabel
Similarity
ss
OnSuperCla
Similarity
_
_
sup
Among which
Similarity On Supercla
ss,
Simila
rity On Label an
d Similarity On Subcl
a
ss
are
the co
rre
s
po
ndin
g
sim
ilarity
of
feat
ure
s
(Sim
k
)
and th
e fe
ature
weight
(W) for featu
r
es,
sup
e
r-cl
ass a
nd su
b-cla
ss
by runni
n
g
the simila
rity measure
m
ent.
Figure 1. The
Ontology Re
pre
s
entatio
n
Figure 2. Com
pari
s
on o
n
Experi
m
ental
of a Sample
Resul
t
s
Duri
ng
the
proce
s
s of
ro
un
d-robin,
if
ne
are
s
t a
d
ju
stment fun
c
tion
A
k+1
and the
simila
rity
function Sim
k+
1
are in
the
same
value
with A
k
an
d
Sim
k
, then e
nd the
circul
ation an
d iteration.
The adju
s
tme
n
t algorithm i
s
as follo
ws:
PROCEDURE:
Ontolog
y
Map
p
in
g
INPUT
:
Ontology
O
1
,O
2
OUT
P
U
T
: Alignnme
n
t A
BEGIN
W
1
=
C
omputeW
eig
h
t(O
1
)
W
2
=
C
omputeW
eig
h
t(O
2
)
A
0
=
C
omputeInitia
l
Ali
gnme
n
t(O
1
, O
2
)
Sim
0
=
C
ompute
I
nitialS
i
mil
a
rit
y
(O
1
, O
2
,A
0
)
k = 1
WHIL
E k
≠
-1
F
O
R e
i
in O
1
F
O
R e
j
in O
2
PU
T
(
Sim
k
, ComputerS
i
milarit
y
(e
i
, e
j
, A
k-1
))
END_FOR
END_FOR
A
k
=
GetAlignment(Sim
k
)
IF Sim
k
≒
Si
m
k-1
AND A
k
≒
A
k-1
TH
E
N
k
= -1
ELSE
k =
k +1
END_IF
END_WHI
L
E
OU
T
P
U
T
(
A
)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
pre
cision
rate
recall
rate
F
‐
measure
Lily
MapPSO
TaxoMap
AFW
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
An Im
proved
Entity Sim
ilarity
Mea
s
urem
ent Method (Gang Iv)
1021
END_BEGIN
END_PROC
E
DURE
As i
s
sho
w
n
i
n
the
above
algorith
m
, if t
he
re
spe
c
tive entity num
be
r for O
n
tology
O
1
a
nd
O
2
is n and m
,
the time complexity of th
e very algorit
hm woul
d be
O (n
×m).
4. Experimental resul
t
s
and An
aly
s
e
s
The test data of thi
s
thesis is OAEI
2009 Corpus (http://oaei.o
ntologymatching.org/
),
and th
e eval
uation of th
e
perfo
rman
ce
stand
ard
re
m
a
in to b
e
p
r
e
c
isi
on
rate, recall
rate
an
d F-
measure [13]. The re
spe
c
ti
ve definitions
of
the calcula
t
ion formula a
r
e as follo
ws:
extracted
standard
extracted
p
standard
stanard
extracted
r
r
p
pr
F
2
The test d
a
ta incl
ude
s 3
3
identified
classe
s,
24 re
lations, 4
4
at
tributes, 5
6
e
x
amples
and 20 exa
m
ples
with no
attribute. The
exper
ime
n
t has al
so
co
mpared the
prop
osed me
thod
(AFW)
with
L
ily, MapPSO, and
TaxoM
a
p, and
a
s
i
s
displ
a
yed i
n
Figure2
belo
w
, o
w
ing
to t
he
adoptio
n of
automatic fe
ature
wei
ght
cal
c
ulat
io
n, the mat
c
hi
ng efficie
n
cy
and th
e th
ree
perfo
rman
ce
stand
ard
s
ha
ve been imp
r
oved sig
n
ifica
n
tly.
5. Conclusio
n
By emphasi
z
i
ng the import
ance to represe
n
t feature
s
via the method of weig
h
t
and
analyzi
ng the
sem
antics
o
f
feature
s
, thi
s
the
s
i
s
ha
s
desi
gne
d the
com
puting
model
of enti
t
y
weig
ht and
calcul
ated the
simila
rity we
ight amon
g variou
s
relatio
n
s. Due to the ado
ption
of
iteration m
e
th
od an
d auto
m
atic featu
r
e
weig
ht ca
l
c
ul
ation, the Ont
o
logy-m
appin
g
efficien
cy h
a
s
been imp
r
ov
ed in relate
d
experime
n
ts. Besides, it
also po
sse
s
se
s better
chara
c
te
risti
c
s in
pre
c
isi
on
rate
, re
call
rate
a
nd F
-
me
asure when
co
mpa
r
in
g w
i
th o
t
he
r
s
y
s
t
e
m
s
.
Pr
io
r
i
ties
w
o
uld
be given
on t
he studies of
improvin
g the robustness and adj
ustable
capability of the algorithm
in
the near futu
re.
Ackn
o
w
l
e
dg
ement
Proje
c
t wa
s
suppo
rted by t
he Nature S
c
i
ence Fou
nda
tion of AnHui
(201
3SQ
R
L
074Z
D,
1408
085M
F1
35).Key Con
s
tructive
Disci
p
line of
He
fei
University, No. 2014x
k08,
Traini
ng O
b
je
ct
for Acad
emic
Lead
er of Hef
e
i University, No. 201
4dtr0
8
Referen
ces
[1]
Z
hou She
ng-c
hen, Qu W
en-
ting, Shi Yin
g
-
z
i, Sh
i Xu
n-zh
i
,
Sun Yun-che
n
. Overvie
w
o
n
Sentime
n
t
Anal
ys
is of Chi
nese Micr
obl
og
gin
g
.
Co
mput
e
r
Applic
ations
and Softw
are.
201
3; 30(3): 16
1-16
4.
[2]
Gang Lv, Cheng Zheng.
A novel frame
w
ork for concept detect
ion on lar
g
e scale video database and
feature po
ol.
Artificial Inte
l
lig
e
n
ce Rev
i
ew
. 2013; 40(
4): 391
-403
[3]
AH. Doan, J. Madh
avan, P. Domin
gos, A. Hale
v
y
. Ontol
o
g
y
m
a
tchin
g
: A machin
e le
arn
i
ng a
ppr
oach
.
Han
dbo
ok on
Ontolog
i
es
in Information System
s. 20
03: 39
7–4
16.
[4]
Xi
on
g F
a
n
g
, H
uan
g Ho
ng-
bin
,
Huan
g Yu-c
h
eng. An
Appr
o
a
ch of Inform
a
t
ion Sem
antic
Clusteri
n
g
Based o
n
Sem
antic Simil
a
rit
y
.
Computer En
g
i
ne
erin
g an
d Scienc
e
. 2012;
34(1
1
): 175-
17
9.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 4, Dece
mb
er 201
4: 101
7 – 1022
1022
[5]
Jian
g Me
n-ji
n,
Z
hou Y
a
-qi
an,
Hua
ng
Xua
n
-ji
ng. S
y
n
o
n
y
m
o
us Entit
y
E
x
pa
nsio
n Bas
e
d
In
formation
De-
dup
licati
o
n
. Jo
urna
l of Chin
es
e Information P
r
ocessi
ng
. 20
1
2
; 26(1): 42-5
0
.
[6]
Cui
Xi
ao-Ju
n, Xia
o
Ho
ng-
yu, Ding L
i
-
x
in.
Distance-B
a
s
ed Ada
p
tive
Recor
d
Matchi
ng for W
e
b
Databases
.
Jo
urna
l of W
uhan
University (Na
t
ural Scie
nce E
d
itio
n)
. 201
2; 12(1): 1-9
[7]
Z
hao Ha
i-
xia,
Li Da
o-sh
en, L
I
U Yong, et al.
Rese
arch o
n
e
n
tit
y
e
x
tracti
on
method
of Dee
p
W
eb dat
a
integr
ation.
C
o
mp
uter Eng
i
n
e
e
rin
g
and A
ppl
i
c
ations
. 20
12;
48(3
6
): 160-
16
3.
[8]
Erlin E, R
ahmi
a
ti R, Ri
o U. T
w
o
T
e
xt Cl
assi
fiers in On
li
ne
Discussi
on: Su
pport Vect
or M
a
chi
ne v
s
Back-Prop
agat
ion Neur
al Ne
t
w
ork.
T
E
LKO
M
NIKA T
e
leco
mmu
n
icati
on
Co
mp
uting
Ele
c
tronics a
n
d
Contro
l.
201
4; 12(1): 18
9-2
0
0
.
[9]
Qi
Yu-do
ng,
Y
an Xi
ao-
bin, Xi
e Xia
o
-fan
g.
C
once
p
tua
l
mod
e
ls simi
lar
i
t
y
c
o
mputati
on
ba
sed o
n
LIS
A
theor
y.
Co
mp
u
t
er Engin
eeri
n
g
and App
lic
atio
ns
. 2012; 48(
3
)
: 40-42.
[10]
Don
g
De
ng-h
u
i
,
Xia
o
Gan
g
, Z
han
g Yu
an-mi
n
g
, Lu Ji
a-
w
e
i, Xu J
un. An SO
A Refere
nce M
ode
l Base
d
on Multi-gr
an
ul
arit
y
Serv
ice Li
brar
y
and Its Applic
atio
n.
Co
mp
uter App
lica
t
ions an
d Softw
are
. 2012;
29(1
0
): 152-
15
5.
[11]
Sun
M
i
ng, Lu
Chu
n
-she
ng, Xu Xi
u-
xin
g
.
A W
eb
Entit
y
Inf
o
rmation
E
x
tra
c
tion M
e
thod
B
a
sed
on
SVM
and Ad
aBo
o
st.
Computer Ap
p
licatio
ns an
d S
o
ftw
are.
2013; 30(4): 10
1-1
0
6
.
[12]
Albar
da A, Supa
ngk
at S
H,
Kuspri
ya
nto K, et al. Informatio
n
Intercha
nge L
a
yer base
d
o
n
Classification of In
formation U
s
e (IU).
T
E
LKOMNIKA T
e
lec
o
mmunic
a
tio
n
Co
mp
uting E
l
e
c
tronics a
n
d
Contro
l.
201
4; 12(2): 48
5-4
9
2
[13]
Y.s R. Jean-M
a
r
y
, EP. Shir
o
nosh
i
ta, et al.
Ontology
matching
w
i
th
s
e
m
antic
verification.
Journ
a
l of
W
eb Semantic
s.
2009; 21(
4): 121-
135.
Evaluation Warning : The document was created with Spire.PDF for Python.