TELKOM
NIKA
, Vol.14, No
.4, Dece
mbe
r
2016, pp. 16
17~162
8
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i4.4787
1617
Re
cei
v
ed Se
ptem
ber 13, 2016; Revi
se
d No
vem
ber
20, 2016; Accepted Decem
ber 1, 201
6
Heterogeneous Informati
on Knowledge Construction
Based on Ontology
Jianhou Ga
n
1
, Gang Xie
2
, Yongzhe
ng Yan
3
, Wanqu
an Liu*
4
1
Ke
y
L
abor
ator
y of Educ
atio
n
a
l Informatizati
on fo
r Natio
n
a
l
i
t
ies Ministr
y
of Educati
on,
Yunn
an N
o
rma
l Univers
i
t
y
, Ku
nmin
g, Chin
a
1,2
F
a
culty
of Metall
urgic
a
l a
n
d
Energ
y
Eng
i
n
eeri
ng,
Kunmi
n
g Univ
ersit
y
of Scienc
e an
d T
e
chn
o
lo
g
y
,
Yunn
an, Kum
m
ing, Ch
ina
3
School of Elec
trical Informati
on an
d Eng
i
ne
erin
g, No
rth Ch
ina U
n
ivers
i
t
y
of
T
e
chnol
og
y,
Beiji
ng, Chi
n
a
4
Departme
n
t of Computi
ng, C
u
rtin Un
iversit
y
, Perth, W
A
6102, Australi
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
:
w
.
li
u@curti
n
.edu.a
u
A
b
st
r
a
ct
Describ
in
g and
represe
n
ting
mu
lti-sourc
e
a
nd
heter
og
ene
ous know
le
dg
e
is a popul
ar researc
h
topic i
n
r
e
cent
years. After i
n
v
e
stigati
ng k
n
o
w
ledge
for
m
in
g
proc
ess b
a
se
d o
n
mu
lti-sour
ce h
e
terog
e
n
e
ous
infor
m
ati
on res
ources, w
e
pre
s
ent a new
ap
proac
h in
w
h
ic
h differe
nt infor
m
ati
on res
ourc
e
s are p
u
t into
a
mutu
al
RDF
(S) data
mo
de
l, a
nd se
mant
ic re
ason
ing
of RD
F
(
S) is cond
uc
ted. More
over,
a know
le
dg
e b
a
s
e
constructio
n
fr
amew
ork for
mu
lti-sourc
e
h
e
terog
e
n
eous
i
n
formatio
n
s
o
u
r
ce w
i
th co
mbi
natio
n
of Ontol
ogy
know
led
ge
mo
del is
put forw
ard, an
d an a
l
gorith
m
of kn
o
w
ledge
base c
ons
tructio
n
is
also pr
op
osed,
in
w
h
ich the c
o
re
issues
are k
n
ow
ledg
e i
n
clus
ion
an
d up
dat
i
ng. T
hen
the ti
me c
o
mp
lexity
of our a
l
g
o
rith
m i
s
ana
ly
z
e
d. F
i
n
a
l
l
y, in or
der to
solve th
e h
e
ter
oge
ne
ous, an
d
unev
en
hori
z
o
n
tal of g
e
o
g
rap
h
ical
distri
butio
n
of ethnic
mi
nor
ity infor
m
atio
n
resourc
e
s in Y
unn
an Pr
ov
inc
e
, w
e
use the
prop
osed
meth
od to co
nstruct a
do
ma
in kn
ow
ledg
e bas
e for
ethnic
mi
norit
y infor
m
atio
n r
e
sourc
e
s, and
use this
mod
e
l to eva
l
uat
e the
efficiency for the know
le
dg
e
inclus
ion
alg
o
r
ithm i
n
resp
o
ndi
ng ti
me a
n
d
ind
e
xin
g
res
pon
din
g
ti
me for
different dat
a resourc
e
s in o
u
r
experi
m
ents.
Ke
y
w
ords
: RD
F
(
S), Ontology, know
ledg
e ba
se constructio
n
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
With the u
b
i
quitou
s
utilization of mod
e
rn
n
e
two
r
k
techn
o
logy a
nd digital te
chnolo
g
y,
kno
w
le
dge b
a
se
d on data
analysi
s
fro
m
different re
sou
r
ces h
a
s
become an i
m
porta
nt factor fo
r
eco
nomi
c
de
velopment. Currently, most Internet
se
rvice is
ba
se
d on a
distri
buted d
a
taba
se
integratio
n
work,
and
the
rapid
develop
ment of th
e
I
n
ternet
re
sult
s in
ple
n
ty of inform
ation
and
kno
w
le
dge a
v
ailable in n
e
tworks. The
s
e informatio
n and kno
w
l
edge a
r
e u
s
ually incom
p
l
e
te
locate
d in se
parate
so
urces, a
s
well a
s
in
compl
e
x diso
rde
r
an
d
they can
not
dire
ctly rep
r
e
s
ent
the new
kno
w
led
ge. Thu
s
it is not e
a
sy to
be un
derstood
by comm
on u
s
e
r
s. As it is q
u
ite
comm
on to
confront
with t
he em
ergen
ce of mult
i
-
so
u
r
ce
s, h
e
terog
eneo
us and
mass info
rma
t
ion
resou
r
ces
i
n
t
he current Web servi
c
e
s
,
tradition
al kno
w
led
ge
b
a
si
s
co
nstructio
n
method ca
nn
ot
meet the rapi
d nee
ds
of u
s
ers, thu
s
it i
s
ne
ce
ssary t
o
develo
p
a
new
kn
owl
e
d
ge con
s
tru
c
tion
method in
order to
provid
e ra
pid n
e
w
knowl
edge
se
rvices. In thi
s
pape
r, we
ai
m to co
nst
r
uct the
kno
w
le
dge b
a
si
s in a different way ba
sed on the
stu
d
y of the unified kn
owl
edg
e rep
r
e
s
entat
ion
for multi
-
source
hete
r
oge
n
eou
s info
rma
t
ion resour
ce
s
with a
n
aim to p
r
ovide
a n
e
w mo
de
l o
f
kno
w
le
dge se
rvice.
The e
s
sen
c
e
of kno
w
ledg
e basi
s
con
s
truction i
s
to
represent knowl
edge
re
aso
nably
and effectivel
y, namely effective kn
owl
edge rep
r
e
s
e
n
tation [1-5].
There h
a
ve
been ple
n
ty of
studie
s
on th
e method
s of kno
w
le
dge re
pre
s
entatio
n. Xu and Ye [6] have studie
d
the merits a
n
d
disa
dvantag
e
s
of currently
use
d
kno
w
le
dge rep
r
e
s
en
tation method
s, and
clai
me
d that Ontolo
gy
has g
r
eat po
tential in kno
w
led
ge re
pre
s
entat
io
n [6]. Fensel, et al., [7] developed the On-t
o-
Knowle
dge
system, a
kn
owle
dge
ma
nagem
ent
sy
stem
based
on O
n
tology,
whi
c
h
aim
s
at
solving the li
mited inform
a
t
ion sha
r
ing
p
r
oble
m
ca
use
d
by tradition
al kno
w
le
dge
basi
s
via usi
ng
the key-wo
rd
matchin
g
m
e
thod to
sea
r
ch
kn
owl
edg
e. In additio
n
, resea
r
che
s
on
kn
owl
e
dge
basi
s
system
con
s
tru
c
tion
have attra
c
te
d much atten
t
ion [8-10], a
m
ong
whi
c
h
a rep
r
e
s
e
n
tative
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1617 – 162
8
1618
work i
s
KMSphere [8], a knowl
edge
ma
nagem
ent sy
stem devel
op
ed by Inst
itute of Comp
uting
Tech
nolo
g
y Chin
ese Aca
demy of Scie
nce
s
. Cheng
[8] stated tha
t
the kno
w
led
ge man
agem
en
t
system
ba
se
d on
Ontol
o
g
y
is mai
n
ly to
solve
two
issue
s
, the
co
nstru
c
tion
of
kno
w
le
dge
b
a
si
s
and the evolu
t
ion of knowl
edge ba
si
s. Re
sea
r
ch
e
r
s
have sh
own t
he advantag
e
s
of Ontology
on
kno
w
le
dge re
pre
s
entatio
n in [8-12]. In this pape
r,
we i
n
tend to con
s
truct the kn
o
w
led
ge ba
sis
by
adoptin
g O
n
tology
with a
n
aim to
improve the
re
st
ri
cted i
n
form
ation resea
r
chi
ng in
traditio
nal
kno
w
le
dge
manag
eme
n
t system. T
he traditio
n
a
l
kno
w
le
dge
base main
ly generate
the
kno
w
le
dge
b
a
se
thro
ugh
the sim
p
le
knowl
edge
di
scovery. T
he
traditional
kn
owle
dge
ba
se
pro
c
e
ssi
ng is
a data pro
c
e
s
sing. Kno
w
le
dge is sea
r
ch
ed in the kno
w
led
ge ba
se
according to the
need
s
of th
e u
s
e
r
s.
Fa
cing
num
ero
u
s
and
ju
m
b
led
network kn
owl
edg
e, the traditio
nal
kno
w
le
dge
b
a
se
p
r
o
c
e
ssi
ng i
s
only li
mited to th
e
data level,
a
nd
can
not
achieve info
rm
ation
interaction.
There a
r
e m
a
ssive i
n
form
a
t
ion re
so
urce
s in Inte
rnet,
in whi
c
h
the
kno
w
le
dge fo
rmation
pro
c
e
s
s is
asso
ciated
with
multiple d
a
ta
sou
r
ce
s; me
anwhile, the
system
s,
gr
a
mmar,
st
ru
ct
ure
and semanti
c
s of resou
r
ce
s are h
e
terog
eneo
us; hen
ce, these information re
so
u
r
ce
s a
r
e nam
ed
as m
u
lti-sou
r
ce h
e
teroge
n
eou
s info
rma
t
ion re
so
ur
ce
s.
I
n
o
r
de
r t
o
co
nst
r
u
c
t
t
he mult
i-
so
ur
ce
hetero
gen
eo
us informatio
n kno
w
led
ge
basi
s
, the firs
t is to solve the sem
antic
hetero
gen
eity of
these i
n
form
ation re
so
urces [13
-
18].
Aiming at
solving the di
fferences an
d inter-op
era
b
l
e
appli
c
ation
s
,
the middl
ewa
r
e
solutio
n
t
o
inte
g
r
ate
the h
e
tero
gen
eou
s multi
-
source
datab
a
s
e
system i
s
put
forwa
r
d in [1
9, 20]. Focu
sing
on the a
c
curacy an
d speed
requi
re
ments of fusi
on
node, the alg
o
rithm of mat
r
ix deco
m
po
sition is pro
p
o
s
ed [21
-
24].
This alg
o
rith
m not only can
acc
e
lerate the s
p
eed of data fus
i
on, but als
o
re
ach
high
accu
ra
cy wh
en
un
symmetrical
bl
ock
matrix happ
e
ned or
som
e
relation m
a
tri
x
lost.
The solution
to the integration of h
e
terog
ene
ou
s data inclu
d
e
s XML, SO
A, data
wareho
use, etc.. XML is a gene
ral meth
od for stru
ctu
r
ed data
rep
r
ese
n
tation, a
nd it allows t
he
appli
c
ation p
r
oce
dure to st
ore a
nd tra
n
smit the
data that ca
n be u
n
derstood
by o
t
her ap
plication
pro
c
ed
ure, and sep
a
rates
the format an
d content
of the data from
the pro
c
e
ssi
n
g
method
s. SOA
(se
r
vice-o
rie
n
t
ed Architect
u
re
) is a
co
mpone
nt mo
del. It links d
i
fferent functi
onal unit
s
of
the
appli
c
ation p
r
oce
dure to a
cont
ract
by the well-
d
e
fin
ed
interfa
c
e. Data wareho
use
te
ch
nolo
g
y,
as a solution
to
the
inte
grat
ion
of heterog
eneo
us
datab
ase
s
, n
o
t onl
y integrate
s
t
he d
a
ta lo
cat
e
d
in different
region
s throu
gh data extraction
and
transfe
r tool, the data o
n
d
i
fferent ope
ra
ting
system
s, different data
st
ructures tog
e
t
her with
ce
rta
i
n data
patterns, but also e
n
su
re
s the d
a
ta
c
o
ns
is
te
nc
y.
The ea
rly versio
n of Ont
o
logy lang
ua
ge is si
milar to XML, which could b
e
simply
cathete
r
ized
into XML.
Wit
h
the
develo
p
ment
of
Ont
o
logy la
ngu
a
ges,
mo
st of
whi
c
h
are
b
a
s
ed
on XML. RDF
is a data m
o
del u
s
ed to d
e
scrib
e
obj
ects and thei
r relation
ship. It provide
s
sim
p
le
langu
age de
scription by
using XML
[25-28]. RD
FS is a si
mplified Ont
o
logy descri
p
tion
langu
age, wh
ich is rega
rde
d
as the voca
bulary
de
scri
ption langu
ag
e of RDF [29
-
31]. Usi
ng th
e
pre
-
defin
ed term
s, RDFS enabl
es u
s
e
r
s to define
t
he cate
go
ry, prop
erty and
the relation
ship
betwe
en the
mentioned t
w
o obje
c
t
s
. As a descrip
tion langua
g
e
of Ontolog
y
, RDFS can
(a)
define th
e re
sou
r
ces an
d
their
categ
o
ri
es,
(b) defin
e
the p
r
op
ertie
s
a
nd
de
scri
be the
rel
a
tio
n
s
betwe
en
obj
ects,
and
(c) define
the
relation
shi
p
s amo
ng
different
catego
ries a
nd va
ri
ous
prop
ertie
s
[3
2, 33]. Now RDFS i
s
wid
e
ly used in
Semantic
Web
,
Intelligent search en
gine,
data
excha
nge in the sem
antic l
a
yer, automa
t
ic link and
referen
c
e of i
n
formatio
n, digital library a
n
d
so on.
In this
paper, we firs
t
study the i
n
formation
proce
s
sing ba
sed on
m
u
lti-so
urce
hetero
gen
eo
us resource
s, by which
we de
scri
b
e
RDFS fo
r the informatio
n
,
and tran
sfo
r
ms
diverse d
a
tab
a
se
s into
one
mutual
RDF
S
model; the
n
ba
sed
on th
e Ontolo
gy knowl
edge
mo
del,
we p
r
e
s
ent t
he kn
owl
edg
e basi
s
con
s
truction
pro
c
edure for m
u
lti-sou
r
ce an
d heteroge
ne
ous
informatio
n reso
urce
s, an
d also
develo
p
an alg
o
rith
m for kn
owl
e
dge ba
si
s co
nstru
c
tion; fin
a
lly,
we evalu
a
te
the appli
c
a
b
ility of proposed alg
o
rit
h
m
throu
gh a
spe
c
ific knowl
edge ba
sis
c
o
ns
tr
uc
tio
n
.
The rest of the pap
er i
s
orga
nized a
s
fo
llows. Se
ction 2 gives
RDFS
de
scri
ption for
multi-source
hetero
gen
eo
us info
rmatio
n re
so
u
r
ces;
Section 3
con
s
tru
c
t
s
knowl
edge
ba
si
s
frame
w
ork fo
r multi-sou
r
ce hetero
gen
e
ous info
rmat
i
on re
sou
r
ce
s
;
S
e
ct
ion 4 puts forward the
algorith
m
for kno
w
le
dge
basi
s
con
s
truction
wi
th
multi-source heteroge
n
eou
s inform
ation
resou
r
ces; Section 5 repo
rts the experi
m
ental
re
sult
s; and Se
ction 6 con
c
lu
de
s this pa
per.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Heteroge
neo
us Inform
atio
n Knowle
dge
Con
s
tru
c
tion
Based o
n
Ont
o
log
y
(Ji
anh
o
u
Gan)
1619
2. RDFS Des
c
ription for
Multi-sou
r
ce
Hete
roge
ne
ous Informa
t
ion Resou
r
c
e
s
2.1. Kno
w
l
e
dge Form
a
tion Proce
ss for Mul
t
i-sou
r
ce Hetero
gene
ou
s Informati
on
Reso
urce
s
In this pa
pe
r, we
are
o
n
ly aiming
a
t
spe
c
ific a
p
p
licatio
n sph
e
re
and
solv
ing the
“inform
a
tion
silo” i
s
sue i
n
sha
r
ing
a
nd co
mmuni
cating of mu
lti-sou
r
ce an
d heterogen
eou
s
informatio
n reso
urce
s. In terms of the
kno
w
le
dge
structu
r
e, mo
st
of
currently use
d
kn
owl
e
dge
sou
r
ces a
r
e
hard
-
a
nalyze
d
un
stru
ctu
r
e
d
text, semi
-structu
red
web
pag
e info
rm
ation in
co
mp
lex
diso
rde
r
, or widely used
con
n
e
c
ted d
a
taba
se
s.
In this case, rese
archin
g o
n
the kno
w
le
dge
formation process for
multi
-
source
and
heterogeneous i
n
form
ation reso
urces will
be
benefi
c
ial if
we
can
devel
op an
efficien
t method to
el
iminate
the
semantic hete
r
ogen
eity amo
ng vari
ou
s da
ta
sources;
on t
he other
hand, it will be also
helpf
ul to transform
the un
solved new information
resou
r
ces int
o
a mutual kn
owle
dge de
scription. Fo
cu
sing on differe
nt data sou
r
ces, putting da
ta
into mutual
RDFS mo
del
a
nd de
scri
bing
data
with th
e RDFS a
r
e
a critical
ste
p
for
kno
w
le
d
ge
basi
s
con
s
tru
c
tion.
Figure 1 sh
ows the kno
w
led
ge form
ation
pro
c
e
s
s for multi-source hete
r
o
gene
ou
s
informatio
n reso
urce
s an
d it can be
divided into
two ste
p
s. Fi
rstly, populati
ng differe
nt data
sou
r
ces to
an
RDF m
odel,
and
se
co
ndly
,
forming
the
kno
w
le
dge
m
odel th
rou
gh
RDFS
seman
t
ic
reasoni
ng process. Next, we will addre
ss these two i
s
sues
separatel
y
.
Figure 1. Knowled
ge form
a
t
ion pro
c
e
ss
of mult
i-so
urce hetero
gen
e
ous info
rmati
on re
sou
r
ce
s
2.2. RDFS Descriptio
n
a
nd Semantic
Reaso
ning
2.2.1 RDFS Des
c
ription
RDFS i
s
a
co
llective wo
rd f
o
r RDF
(Resource
De
scri
ption Frame
w
ork) a
nd its e
x
tension
RDF S
c
h
e
m
a
, whi
c
h is an asse
rtio
n langu
age
usin
g sta
nda
rd vocabul
ary to repre
s
ent
comm
and
s.
The
basi
c
co
nce
p
tion i
s
t
o
u
s
e
sim
p
le
statem
ents to repres
ent
r
e
so
ur
ce
s.
E
a
ch
statement co
ntains
three parts,
n
a
mely
Subject, Pre
d
icate,
an
d
Ob
je
c
t
. In orde
r to adopt RDFS
for b
e
tter
de
scribi
ng
multi-sou
r
ce
heterogen
eou
s i
n
formatio
n, kno
w
led
ge i
s
un
derstood
to
b
e
a
s
a combin
atio
n of a
seri
es of
resource
s. RDFS u
s
e
s
prope
rty an
d p
r
ope
rty value to
de
scribe
resou
r
ces. O
ne RDFS de
scriptio
n is defi
ned a
s
followi
ng:
Defini
tion 1:
Statement::=
<
s
ubjec
t,predic
a
te,objec
t
>
Subject
is u
s
ed to descri
b
e RDFS resource,
predi
cate
is for some specific f
a
ctor or
c
h
ar
ac
te
r
i
s
t
ic o
f
su
b
j
ec
t, or
its
r
e
l
a
tion
ship to
other p
r
ope
rtie
s, an
d
obj
ect
i
s
th
e prope
rty va
lue
whi
c
h could a
l
so be a
subj
ect
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1617 – 162
8
1620
As to spe
c
ific sph
e
re, RDFS off
e
rs
kn
owl
e
d
ge de
scripti
on for mult
i-so
urce
hetero
gen
eo
us info
rmatio
n. Based
on
Definition 1
,
one ha
s th
e followin
g
steps for
RDFS
descri
p
tion
s:
1.
Establishing
a commo
n v
o
ca
bula
r
y o
r
set of
re
so
urce
s fo
r a
spe
c
ific
sp
here,
among
which
the co
mmon
vocab
u
lary o
r
set of resource
s
shoul
d b
e
ea
sy to un
derstand
and
can
be u
s
e
d
con
s
i
s
tently for de
scriptio
n
;
2.
Usi
ng the RDF Sch
e
ma
langu
age to establi
s
h the
commo
n vocab
u
lary in some sp
ecif
i
c
appli
c
ation a
r
eas;
3.
For the
ne
w
RDFS vo
ca
b
u
lary, one
sh
ould d
e
scri
be
the cla
s
s, property an
d re
sou
r
ces
as
a
whol
e, so a
s
to provide a g
ood ba
si
s for modelin
g a specifi
c
sp
here;
4.
Adopting
RQ
L (RDF
Que
r
y Lang
uag
e)
to que
ry
fo
r o
ne o
r
more
RDF
or Sch
e
m
a
mo
de, a
nd
returning to the co
rrespon
ding vari
a
b
le
to bind the list [34, 35].
In the RDFS
step
s me
ntio
ned a
bove, it
usu
a
lly
uses
several
con
s
t
r
uctio
n
e
n
tities, which
inclu
de: (1
) The main
entity is
rdfs
:res
ourc
e
,
and two subcl
a
sse
s
are
rdfs:
c
las
s
and
rdfs:p
rop
e
rty
;
(2) Th
e co
re ch
ar
act
e
ri
st
ic
s incl
ude
s r
df
s:
su
bCl
a
s
s
Of
,
rdfs:
s
ub Prop
erty Of
,
rdfs
:type
; (3
) The
co
re
con
s
trai
nts are
rdfs
:range
,
rdf
s
:do
m
ain
,
refs:constraintPro
p
e
rty
,
rdfs:
c
on
strai
n
tResource
.
In the foll
owi
ng p
a
rt
of thi
s
p
ape
r,
we
give
the
rel
e
vant definitio
n
s
of
RDFS
Descriptio
n
and Semanti
c
Re
asoning,
then a new
method for di
ffe
rent inform
ation re
sou
r
ces bein
g
put into
a mutual RDFS data mod
e
l is presente
d
. Next, a
kn
owle
dge b
a
si
s const
r
u
c
tio
n
frame
w
o
r
k i
s
put
forwa
r
d with
combi
nation
of Ontology know
l
edge m
o
del, and an a
l
gorithm of knowl
edge b
a
s
is
con
s
tru
c
tion i
s
develo
ped.
2.2.2. RDF
(S
) Semantic
Reasoning
In the
kno
w
ledge
form
ation p
r
o
c
e
s
s for
multi-source
hete
r
o
gene
ou
s inf
o
rmatio
n
resou
r
ces,
RDFS d
e
scrip
tion is me
re
ly use
d
to
stand
ardi
ze
the RDF/XM
L seri
alizatio
n
statement
s. Ho
wever, onl
y the formal repre
s
e
n
ta
tion
of gramma
r can b
e
con
d
u
c
ted for ma
ch
in
e
impleme
n
tations,
whi
c
h
cannot avoi
d the ambi
gui
ty of RDF lan
guag
e comp
rehen
sio
n
. In the
prog
ram
m
ing
impleme
n
tation, judgm
en
t on the tr
u
e
of RDF
(S) statem
ent
is requi
red
to
impleme
n
t RDFS autom
a
t
ed rea
s
o
n
in
g. Therefor
e
,
this pap
er offers
sev
e
ral im
porta
nt
con
c
e
p
tion
s, as the theo
ret
i
cal ba
si
s of RDFS
rea
s
on
ing; in the me
antime, owin
g to the fact that
RDFS
as on
tological de
scriptio
n la
ng
uage, the
s
e
theoreti
c
al
d
e
velopme
n
ts are
not o
n
l
y
the
inevitable evi
den
ce for a
u
t
omat
ed re
asoning, but al
so suitable f
o
r the Ontol
o
gy rea
s
onin
g
in
kno
w
le
dge b
a
si
s co
nstruct
i
on.
Defini
tion 2
:
Brief explan
ation
o
n
RDF
graph.
A
s
fo
r the
V in
RDF
gra
p
h
de
noted
a
s
I=<IR,IP,IEXT,IS,IL>, whe
r
e, IR d
enote
s
the n
o
n
-
em
pty resource
s set a
nd i
s
called the
dom
ain
of explanatio
n I,
IP
is the prop
erty re
so
urces
set, IEXT
is the resource
s se
t whose prope
rty is
mappe
d into
resou
r
ces-resource
s
colle
ction, that is IEXT:
2
IR
I
R
IP
, IS is a
mappin
g
fro
m
URI
into resourc
e
s
or
property, that is IS:
UR
I
I
R
I
P
, IL re
pre
s
e
n
ts a m
appi
ng f
r
om th
e type
d
literals of V in
to reso
urce
s, that is IL:typed literals
→
IR.
From
Definiti
on 2, RDF semantic l
ang
uage
sho
u
ld
firstly map V in the RDF g
r
aph i
n
to
the element
s in domain IR,
then the elements a
r
e m
appe
d as the
binary relatio
n
in domain IR.
Thus,
unde
r t
he sim
p
le int
e
rp
retation
of a given
RDF
grap
h, it can
be cl
early d
e
fined by the
RDF
grap
h assig
n
m
ent whi
c
h is as a true val
ue judgm
ent method of RDF graph.
Defini
tion 3:
Assig
n
ing RDF
g
r
aph.
O
ne RDF graph will give the following assi
gning
rule
s:
(1) If E is a non-type a
r
gu
ment, then
()
IE
E
;
(2) If E is a type arg
u
me
nt, then
()
(
)
IE
I
L
E
;
(3) If E is URI
, then
()
()
IE
I
S
E
;
(4) If E is
a no
n-e
m
pty triple
no
de
,,
sp
o
, then
I(E)=True, a
nd
only when,
s,
,
VP
Vo
V
,
()
IP
I
P
and
()
,
(
)
(
(
)
)
Is
Io
I
E
X
T
I
P
(5) If
E
is an RDF g
r
a
ph, then
()
IE
F
a
l
s
e
, and only when
E
co
ntains
certai
n triple nod
e
E*
whic
h mak
e
s
()
IE
F
a
l
s
e
, otherwi
se
()
IE
T
u
r
e
.
In RDF
sem
antic rea
s
oni
ng, RDF gra
ph assi
g
n
ing
is a judg
me
nt method o
n
definin
g
true valu
e. O
n
the b
a
si
s
of this, RDF
se
mantic
sp
ecifi
c
ation
is i
n
tro
duced to
de
rive a con
c
ept
for
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Heteroge
neo
us Inform
atio
n Knowle
dge
Con
s
tru
c
tion
Based o
n
Ont
o
log
y
(Ji
anh
o
u
Gan)
1621
con
d
u
c
ting t
he true
state
m
ent ove
r
other true
statements,
con
s
eq
uently it
will le
ad to
RDF
implicatio
n co
nce
p
t and RDFS implicatio
n con
c
e
p
t.
Defini
tion 4
:
Sim
p
le im
plication
rul
e
s.
Given
S
a
n
d
E are
both
RDF
g
r
aph, i
f
every
simple expl
a
nation satisfi
e
s that
S
can
meet
E
, then it is called
S
simple impli
c
ating
E
; in other
words, if each model of RDF graph
S
is
E
’s model, then
S
simply implicates
E
.
In pra
c
tice, d
e
fining impli
c
ation rule
s is
mainly use
d
to achi
eve ma
chin
e rea
s
o
n
i
ng. The
basi
c
con
c
ep
tion is that if an RDF
gra
ph co
ntain
s
some fo
rm
s of triple node
,,
sp
o
, after
applying impl
ication rule
s, the triple no
de ca
n be a
dded into the
RDF g
r
ap
h. RDF
sema
n
t
ics
spe
c
ification
[21] mainly d
e
fines thre
e i
m
plicatio
n
rul
e
s, n
a
mely t
he
simp
le im
plication rules,
RDF im
plication rule
s an
d RDFS impli
c
a
t
ion rule
s.
Defini
tion 5:
RDF im
plication rule.
Given
S
and
E
are two
RDF
graph
s, if one gra
p
h
whi
c
h ap
plie
s the simple i
m
plicatio
n rul
e
of
S
or RDF implication rule
simply im
plicates
E
, then it
is call
ed
S
RDF- impli
c
atin
g
E
.
Defini
tion 6:
RDFS im
plication rule.
Gi
ven
S
and
E
are two
RDS
graph
s, if one grap
h
whi
c
h ap
plie
d simple im
p
lication rule
of
S
, RDF implicatio
n rul
e
, and RDF
S
implication
rule
simply impli
c
ates
E
, then it is calle
d
S
RDFS-im
plicating
E
.
In other wo
rd
s,
S
RDFS-i
m
plicating
E
if and only if
RDFS
clo
s
ure of
S
simply
implicat
es
E
, equivalent
to the
co
nce
p
t of form
al
gramm
a
r RDFS ded
uctio
n
.
It guarantee
s the
validity
of
the RDFS lo
gical
rea
s
oni
ng for sema
ntic co
nc
ept, and provide
s
a theo
retical foundatio
n
on
kno
w
le
dge re
pre
s
entatio
n reasonin
g
.
3. Kno
w
l
e
d
g
e
Basis
Cons
tru
c
tio
n
Frame
w
o
r
k for Mu
lti-sour
ce
Heterog
e
neo
u
s
Information Reso
urce
s
3.1. Ontolog
y
Kno
w
le
dg
e Basis Mo
d
e
l
The differen
c
e b
e
twe
e
n
new
kn
owl
edge
ba
sis
frame
w
ork
a
nd tra
d
itiona
l DBMS
(Data
b
a
s
e M
anag
ement S
y
stem) i
s
that
traditional
DBMS is una
bl
e to rep
r
e
s
e
n
t
and de
al wit
h
rule
-ba
s
e
d
knowl
edge,
bu
t new
kn
owle
dge b
a
si
s h
a
s
a
uniform symbol an
d st
ructu
r
al
mod
e
l,
whi
c
h i
s
a
ra
tional
colle
cti
on of
de
scrip
t
ion kno
w
led
ge a
nd
proce
dural
kno
w
le
dge i
n
a
spe
c
ific
sph
e
re;
on t
he othe
r h
a
n
d
, Ontology,
as a
kn
owl
e
dge
rep
r
e
s
en
tation metho
d
, can
effecti
v
ely
rep
r
e
s
ent the
con
c
e
p
ts of
stru
cture an
d the rela
tio
n
s
betwe
en con
c
ept
s. So it can better
achi
eve
the "shared
con
c
e
p
tuali
z
a
t
ion" [36-41]. In
ce
rtai
n
appli
c
ation
s
,
we
combi
n
e the
two
i
n
a
rea
s
on
able
way, and then
provide
a ne
w kno
w
led
g
e
basi
s
m
odel,
intendin
g
to
provide
re
qui
red
data a
nd
normative inst
ru
ction
s
for the
kn
owl
edge
structu
r
e, a
n
d
build
a the
o
r
etical
ba
si
s f
o
r
building a
kn
owle
dge ba
si
s frame
w
o
r
k.
Defini
tion 7
:
Ontolog
y
rep
r
ese
n
tation
of kno
w
le
dge b
a
si
s
m
odel.
<Knowle
dge M
odel
>::
=<Dom
a
in
Kn
owle
dge
> <Reason
Kno
w
l
edge
> <Ta
s
k Knowle
dge
>
, in
whi
c
h Dom
a
in
Knowledge
rep
r
e
s
ent
s field knowl
edg
e used for
a
detailed d
e
scriptio
n of a
particula
r field of kno
w
le
d
g
e
type; Reason
Knowled
ge repre
s
e
n
ts re
aso
n
ing
an
d or metho
dolo
g
y kno
w
ledg
e, describi
ng
the
rea
s
oni
ng m
e
thod
s or st
eps of ge
neral kno
w
led
g
e
in spe
c
ific
area
s, su
ch
as matching,
a
gene
rato
r, an inferen
c
e en
gine, and oth
e
r ba
sic
con
s
truction
s; Ta
sk Knowl
edg
e rep
r
e
s
ent
s the
task
kno
w
le
d
ge, whi
c
h de
scribe
s the ta
rget kn
ow
l
e
d
ge of the syst
em to be ach
i
eved in stag
es,
inclu
d
ing the
sub-ta
sks of
decom
po
sition in the re
aso
n
ing p
r
o
c
ess and targ
et kno
w
ledg
e
in
rea
s
oni
ng.
Definition 7 can rep
r
e
s
ent
the three
-
level kno
w
led
g
e
system of "facts-con
cept
-rule", but
the Ontolo
gy kno
w
led
ge
model i
s
onl
y capa
ble of
better ge
ne
ralizi
ng a
nd
abstractin
g
the
kno
w
le
dge re
pre
s
entatio
n.
The cu
rrent kno
w
le
dge b
a
si
s
fra
m
e
w
o
r
k ca
nnot we
ll
rep
r
e
s
ent
t
h
e
update
proce
ss, th
us
Defi
nition 7
is u
s
ed a
s
a
sup
p
l
e
menta
r
y. Ne
xt, the Ontolo
gy man
agem
ent
is appli
ed in the co
nstructi
on of the pro
posed kno
w
le
dge ba
si
s fra
m
ewo
r
k.
3.2. Kno
w
l
e
d
g
e Basis F
r
a
m
e
w
o
r
k
Based o
n
the above an
alysis, we offer a
multi-source hete
r
o
gene
ou
s informati
o
n
resou
r
ces
kn
owle
dge ba
si
s frame
w
o
r
k,
as sho
w
n in
Figure 2.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1617 – 162
8
1622
Figure 2. Multi-so
urce h
e
terogen
eou
s inf
o
rm
atio
n re
so
urces
kno
w
le
dge ba
se fra
m
ewo
r
k
3.3. Diffe
ren
t
Data Sou
r
ce
s into RDF model
It is require
d
to have different data source
s,
thus convertin
g
different data
source
s
(SQL, XML,
RDF
) into
DOi throu
gh a
wra
ppe
r
i
. The convertin
g
method for
different type
s of
data va
rie
s
. In ap
plication,
wrap
per 1
is the
tra
d
itio
nal
wra
ppe
r
(mainly fo
cu
sing o
n
XML
or
traditional
da
ta whi
c
h
can
be
structu
r
al
ly descri
bed
),
whi
c
h
on
e
can u
s
e
the V
e
locity [42] f
o
r
conve
r
ting.
Wrappe
r 2 i
s
a relation
al dat
ab
ase
packa
ge
wrapp
er, we can use
D2
RQ
,
SquirrelRDF,
Virtuo
so
an
d othe
r tool
s to c
onvert i
n
to RDF
gra
ph, an
d the
n
acce
ss thro
ugh
SPARQL. Wrapper 3 i
s
an as
sociated data package
wrapper
(mai
nly for the
RDF data model in
sema
ntic web
)
, which use Pubby
[43] (a
ssoci
a
ted dat
a front end
) to map URI in supp
orting
We
b
bro
w
ser for t
he pa
ckagin
g
process. Th
e wra
ppe
r
i in the pro
c
e
s
s is ba
se
d o
n
the RDF (S)
descri
p
tion a
nd sem
antics reasonin
g
.
3.4. Process
of Da
tab
ase
Cons
tru
c
tio
n
One can mat
c
h DOi and
CO by match
i
ng device int
o
the alignm
ent rule Ri (Ri is the
gene
rating
al
ignment
rul
e
matching
de
vice).
Giv
en
Ri, throug
h i
n
terme
d
iate f
ile i, inte
ra
ctive
data so
urce
s can b
e
esta
blish
ed bet
ween data
(X
ML, SQL, RDF) and
CO.
Whe
n
the dat
a can
be de
scribe
d according to the Ontolo
gy and then
co
n
v
erted into RDF, we
can
convert the da
ta
into the
kn
owledge
ba
si
s,
and
provid
e a
n
interf
a
c
e
to
outsi
de
acce
ss, i
n
we
ca
n
use
RDF q
u
e
ry
langu
age, S
PARQL, fo
r kn
owl
edg
e
retri
e
val p
r
ocessin
g
. T
he
whol
e
pro
c
e
s
s is
an
impleme
n
tation of ontolog
y knowl
edg
e theoretical m
o
del.
3.5. Ontolog
y
Manageme
nt
Ontology ma
nagem
ent is
mainly focu
si
ng on
kno
w
le
dge up
dating
process, whi
c
h a
c
ts
on the entire
cycle of kn
owle
dge ba
si
s co
nstruc
tio
n
, includin
g
ontology merging, ontolog
y
decompo
sitio
n
, and
onto
l
ogy evolutio
n. Je
na [
4
4
]
is
conve
n
i
ent tool fo
r merging
a
nd
decompo
sin
g
the ontolo
g
y of RDF
data
model. Th
e
ev
olution of
Ontology req
u
ire
s
u
s
to d
e
fine
s
o
me
r
e
le
vant r
u
le
s
,
as
s
h
o
w
n
in
F
i
gu
re 3
.
Thro
ugh m
a
tchin
g
, one
ca
n esta
blish th
e relatio
n
ship
betwe
en the
old
t
O
and th
e
ne
w
t+
n
O
, and the
n
ge
nerate
the ali
gnment
rule
i
R
. Given
i
R
, throu
gh ge
ne
rators, conversio
n
mode
l
can b
e
gen
erated, and the
n
it is convert
ed into
t+n
I
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Heteroge
neo
us Inform
atio
n Knowle
dge
Con
s
tru
c
tion
Based o
n
Ont
o
log
y
(Ji
anh
o
u
Gan)
1623
Figure 3. Ontology evolutio
n pro
c
e
s
s
4.
Constr
uction Algorith
m
for He
tero
geneou
s Information Kn
o
w
l
e
dge
Bas
e
Based o
n
the kno
w
led
g
e
basi
s
frame
w
ork di
scussed previo
usly
, we can de
velop a
multi-source
hetero
gen
eo
us i
n
form
atio
n resour
ce
s
kn
owl
edg
e basi
s
co
nst
r
uction
alg
o
rit
h
m,
whi
c
h is d
e
scribed in the fo
llowing
step
s.
First we anal
yze different
data so
urce
s, and return t
o
the triple node re
quired
by the
RDF.
We the
n
use interfa
c
es a
nd meth
o
d
s p
r
ovide
d
b
y
Jena fo
r further p
r
o
c
e
ssi
ng. Firstly, one
cre
a
tes th
e RDF mo
del u
s
i
ng Mod
e
lFa
c
t
o
ry, and the
n
usin
g the
r
e
a
d
( )
func
tion to read the
RDF
data, finally applying the iterator StmtIte
rato
r to return to the triple node collection.
For a pa
rticular field of
kno
w
led
ge,
one ca
n impleme
n
t the kno
w
le
dge
inclu
s
ion
algorith
m
and
kno
w
led
ge u
pdating al
gori
t
hm as de
scri
bed bel
ow.
Algorithm 1
:
Kno
w
l
e
d
g
e Inclusion
Al
gorithm
S::=<subje
c
t, predi
cate, o
b
ject
>|[<subj
ect, predi
cate
, object
>
] let that P is defi
ned a
s
a
triple nod
e in S. Data type is the dou
ble
colle
cti
on of the key value
of Map (key, value). The
key
mode of P (key, value) is (predi
cate, {
s
u
b
je
ct, obje
c
t}), in which the
key is the ID;
Input:
kno
w
l
edge
in
clud
e
d
Outpu
t:
KB
Step 1: If
SK
B
, th
en stop; othe
rwise go to Step 2;
Step 2: Let
.
subC
la
Tf
Ss
s
O
, and
CO
be t
he ontology knowl
edge b
a
s
e, if
TC
O
, then
output error a
nd stop; othe
rw
ise extract o
ne triple no
de
P
from
S
, tur
n
to Step 3;
Step 3: If
."
"
P
p
r
e
d
i
ca
te
ID
, then give key value of
ID
to
.
P
obje
c
t
and then mat
c
h to the
prop
erty colle
ction
PS
of
KB
; if
.
Pp
r
e
d
i
c
a
t
e
P
S
acco
rdi
ng to th
e in
h
e
rited
rel
a
tion
of p
r
op
erty
in RDF de
scription, incl
ud
e
p
r
edi
ca
te
; if
."
"
Pp
r
e
d
i
c
a
t
e
P
S
, and l
e
t
{}
PS
PS
pr
edi
c
a
te
then turn to Step 4;
Step 4: Let
{}
SS
P
, if
S
then
stop,
otherwise ext
r
act
one
tripl
e
nod
e
P
from
S
,
turn to Step 2.
Algorithm1
g
i
ves the
kno
w
led
ge i
n
cl
u
s
ion
p
r
ocess
for multi-so
urce heterog
eneo
us
informatio
n reso
urce
s kn
o
w
led
ge ba
sis con
s
tru
c
tion,
Next we pre
s
ent the Kno
w
led
ge upd
ating
in Algorithm 2
.
Algorithm 2
:
Kno
w
l
e
d
g
e Upda
ting
Al
gorithm
Input:
Key, ID and tripl
e
n
ode ne
ede
d to be upd
ated;
PS::={ID}(<[<s
ubjec
t, pr
e
d
i
c
ate, obje
c
t>]
>
;
Outpu
t
:
KB
Step 1: If
ey
KK
B
I
D
, it implies that the kno
w
led
ge ba
si
s doe
s not inclu
de the
kno
w
le
dge a
bout ID, and
stop; othe
rwi
s
e, ma
ke
{}
PS
PS
ID
,extract o
ne triple
node
P
, turn to
Step 2;
Step 2: Give
n
.
Pp
r
e
d
i
c
a
t
e
, if
..
P
p
r
e
di
c
a
t
e
KB
p
r
ed
i
c
at
e
, it implies that th
e ontol
ogy
kno
w
le
dge
d
e
scrib
ed i
s
n
o
t in the
kno
w
led
ge b
a
si
s,
then
sea
r
ch
P
, judge
whet
her
.
Pp
r
e
d
i
c
a
t
e
is
one
su
b-type
of existin
g
reso
urce
s; if
yes, in
stantia
te
P
; if no, let
..
KB
p
r
edicate
KB
pr
ed
i
c
a
t
e
{.
}
Pp
r
e
d
i
c
a
t
e
then insta
n
tiate it. If
..
P
p
r
e
di
c
a
te
K
B
pr
e
d
i
c
a
t
e
, instantiate directl
y
, and then turn
to Step 3;
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1617 – 162
8
1624
Step 3: Let
{}
PS
PS
P
, if
PS
, then st
op; otherwi
se
extract on
e
triple no
de
P
from
PS
, then turn to Step 2.
From a
bove
mentione
d two algorith
m
s,
the al
gorith
m
s mainly traverse thro
u
gh triple
node. By an
alyzing the p
r
ope
rty of triple nod
e,
on
e can a
c
hi
eve expan
sion
or modifi
catio
n
of
kno
w
le
dge
b
a
si
s. When
the
sea
r
ch of
triple n
ode
e
nds, th
ese al
gorithm
s
also
stop.
For lim
ited
numbe
r of tri
p
le no
de, the
step
s of the
s
e al
gor
ith
m
s are
also limi
t
ed, then the
com
putation
a
l
time of algorit
hm is
()
On
where
n
is the numb
e
r of the triple
node
s.
By using the RDF
query
l
anguage SPARQL one
c
an
retrieve knowl
edge from thi
s
model. First,
input SPARQL as
a st
ring,
then pars
e the st
ring to generat
e
an abstract syntax,
and then define query rul
e
s by al
gebra operators
provided by
SPARQL, finally calculate the
results in RDF grap
hs.
The kno
w
led
ge ba
sis
co
n
s
tru
c
tion al
go
rithm for multi
-
so
urce h
e
terogen
eou
s inf
o
rmatio
n
resou
r
ces in
clude
s the abo
ve algorithm
1 and the alg
o
rithm 2.
5.
A Practic
al Example
Yunnan i
s
a
n
area
with
ri
ch mi
no
rity re
sou
r
ces.
Cu
rrently there
is
not a
com
p
re
hen
sive
unde
rsta
ndin
g
on different
material cult
ure am
ong 2
6
ethnic min
o
rities in Yun
nan. The maj
o
r
rea
s
on i
s
that the nature of
these
national reso
urces i
s
het
erog
ene
ou
s, imbalan
ce, and
diso
rde
r
e
d
i
n
geo
graphi
cal
distri
buti
on. By
usi
n
g knowl
edg
e
ba
sis
co
nstruction
met
hod
prop
osed in t
h
is pa
pe
r, on
e can
co
nst
r
uct t
he multi-sou
r
ce hete
r
ogen
eou
s mi
nority inform
a
t
ion
resou
r
ces
kn
owle
dge ba
si
s, the step
s a
r
e sh
own as f
o
llows.
We
can
use
Protégé [4
5] to build the
Ontology
kn
owle
dge
ba
sis in eth
n
ic
minorit
y
domain, an
d
form the RDF graph be
tween do
mai
n
ontology a
nd appli
c
atio
n as sh
own
in
Figure 4. The
uppe
r pa
rt of Figure
4 is t
he ontol
o
g
y section of
RDF gra
ph, de
scribi
ng
con
c
e
p
ts
like nation, id
eology, cu
sto
m
s, etc., and
relation
s like
religion p
r
op
erty, languag
e prop
erty, etc.
The lower
se
ction of this figure
rep
r
e
s
e
n
ts the
re
al minority Wa
descri
bed by
Ontology an
d
its
related info
rm
ation. Two
se
ction
s
are di
stinguished through the
RDFS.
Figure 4. RDF Relation b
e
t
ween d
o
mai
n
ontology an
d appli
c
ation
example
By using
int
e
rfaces an
d
method
s p
r
o
v
ided
by
Je
n
a
, one
can
return t
he tri
p
le no
de
need
ed for
RDF de
scri
ption, and then
search th
roug
h triple
s in lo
ops.
Rea
s
oni
ng rul
e
s
ba
se
d o
n
the
RDFS are
given
i
n
T
abl
e
1. Next we con
s
tru
c
t th
e
kno
w
le
dge
b
a
si
s b
a
sed
o
n
the
kn
owl
e
dge
inclu
d
ing alg
o
r
ithm.
The
experi
m
ent ha
s
used
a hi
gh
-pe
r
fo
rman
ce
serv
er
fo
r data proce
s
sing. Th
e
sy
stem
environ
ment
is java: j
d
k1.5.0. The
max
i
mum h
eap
thre
shol
d i
s
2
56M. As we
aim at
differe
nt
data so
urce
s, ethnic inform
ation re
sou
r
ce dat
a set is
acq
u
ire
d
as
shown in Tabl
e 2.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Heteroge
neo
us Inform
atio
n Knowle
dge
Con
s
tru
c
tion
Based o
n
Ont
o
log
y
(Ji
anh
o
u
Gan)
1625
We
have
re
corde
d
the
respondi
ng time
of execut
ing i
n
clu
d
ing
algo
rithm of
ea
ch
data,
a
s
sho
w
n in
Fig
u
re 5. F
r
om
Figure 5, we
kno
w
t
hat the
respon
ding t
i
me of incl
udi
ng algo
rithm
of
different data
sou
r
ces i
s
diff
erent, the
re
spondi
ng
time
of Web
pag
e
is better t
han
other two dat
a
sou
r
ces in th
e same th
ree
tuple numb
e
r.
By applying SPARQL query language
with the
proposed knowledge
updating algorithm,
we ca
n se
arch the triple no
de, and finally return
the relevant sea
r
ch results on
RDF g
r
ap
hs.
We
aim to
se
arch an
d
return
the
“n
ation
”
a
nd
“culture”
as de
scri
bed i
n
RDF
data
sou
r
ce
on
“http://ethnic.
ynnu.edu.
cn/
e
thnic”. Thus we can co
nstruct the following
knowledge basis model:
PREFIX rdf: <http://www.w3.or
g/1999/02/22-rdf-s
y
ntax-ns
#
>
PREFIX nation: <http://ethnic.
ynnu.edu.
cn/ont
ol
ogy/nation#>
SELECT ?X ?Y
FROM<http://ethnic
.
ynnu.edu.c
n
/ethnic
/
ethnic>
WHE
R
E{?Y n
a
tion:cultu
r
e
?X}
Table 1. RDF
(
S) re
asonin
g
rules
Rule Name
Kno
w
n Condition
Reasoning Results
rdfs1
<s,p,o>(if o is lite
r
al)
_n:rdf:t
ype r
d
fs:Literal
rdfs2
(p rdfs:domain
x)
&<s,p,o>
s rdf:t
y
pe
x
rdfs3
(p rdfs:ra
nge
x)&
<
s,p,o>
o rdf:t
ype
x
rdfs4 <s,p,o>
s||o
rdf:t
ype
rdfs:
R
esource
rdfs5
(p rdfs:subProp
e
r
t
y
Of q)&
(
qrd
f
s:s
ubProper
t
y
Of r
)
p rdfs:subPrope
rt
yOf
r
rdfs6
p rdf:t
ype rdfs:Cl
a
ss
p rdfs:subPrope
rt
yOf
p
rdfs7
<s,p,o>&(p rdfs:subProper
t
y
Of q
)
<s,p,o>
rdfs8 s
rdf:t
y
pe
rdfs:Cl
a
ss
s
rdfs:subClassOf rdfs:Resource
rdfs9
(s rdf:t
ype x)&(x rdfs:subClassOf
y)
s rdf:t
y
pe
y
rdfs10
s rdf:t
y
pe
rdfs:Cl
a
ss
s rdfs:subClassof s
rdfs11
(x rdfs:subClassOf y)&(y
rdfs
:sub
ClassOf z)
x rdfs:subClassOf z
Table 2. De
scriptive data
of experime
n
ts
Data Source
File Number
Example Numb
e
r
Triple Node N
u
m
ber
Database Size (
M
B)
Relation Databas
e
2
134
2367
1.5
XML 34
767
4783
2.38
Web Page
67
1023
8921
3.7
Being differe
nt from the traditional al
go
ri
thm, usin
g the propo
se
d kno
w
le
dge in
cludi
n
g
algorith
m
and
the kno
w
led
ge upd
ating
algorith
m
is b
e
tter in the following
kn
owl
edge infe
ren
c
e:
(1) T
he tradit
i
onal alg
o
rith
m can o
n
ly be on stri
ng
matchin
g
for
the relation
al
databa
se
s a
nd
XML, whi
c
h
are t
w
o
data
so
urce
s. Bu
t, using
kn
o
w
led
ge in
clu
d
ing al
go
rith
m and
knowl
edge
updatin
g alg
o
rithm, the knowl
edge u
n
i
t
with sema
n
t
ics will be i
n
tegrate
d
, an
d it is easy to
unde
rsta
nd
the
sem
antic
relation
bet
ween th
e
kn
o
w
led
ge
units to the
ma
ch
ine. (2) Fo
r
RDF
data so
urce
s,
the tradition
al algorith
m
can only us
e the no
rmal form of RDF to
dire
ctly link the
kno
w
le
dge,
and e
s
tabli
s
h the rel
e
va
nce
amon
g
a
ll types of file storage
d
a
ta, and it lacks
sema
ntic info
rmation. B
u
t, Using
the
knowl
edge
in
clusio
n
sub
-
al
gorithm
an
d
the kno
w
led
g
e
updatin
g su
b
-
algo
rithm to extend kn
owl
edge o
n
the bas
i
s
of the tradition
al RDF triples, that
is
RDF
=
{Reso
u
rce, Attribut
e, Re
so
urce
Type, Attrib
ute Type},
a
nd the
ch
ara
c
teri
stics of t
h
e
resou
r
ces a
n
d
the prop
ert
y
is retained,
then one
can
discover n
e
w
kno
w
le
dge.
By applying the kn
owle
dg
e updating al
gorithm,
we can se
arch th
e triple nod
e, and the
respon
ding ti
me of differe
n
t
data so
urce
s is
sh
own
a
s
Figure 6. Fro
m
Figure
6, we kn
ow th
at the
respon
ding
time of
upd
ating al
gorith
m
of differe
nt d
a
ta sou
r
ces i
s
diffe
rent, th
e respondi
ng
time
of We
b pa
ge
is
slightly b
e
tter than
ot
her t
w
o
data
so
urce
s in t
he
same
qu
e
r
y. In the
sin
g
le
pseudo-di
s
tri
buted environment, we
can select
the Query SPARQL query in Figure 6 to
analy
z
e
the re
sp
on
se
time of RDFS
infere
nce mo
del u
s
ing
the
traditional
alg
o
rithm
and
o
u
r al
gorith
m
on
RDF d
a
ta so
urces, via co
mbining
with the su
b-al
go
rithm kno
w
le
dg
e updatin
g.
The re
sult
s are sho
w
n
in Figure 7. Th
rou
gh
the com
para
t
ive analysis of the
experim
ental result
in
Figu
re 7,
we can
dra
w
t
w
o
co
n
c
lu
sion
s: Fi
rst, with the
in
crea
sing
of th
e
numbe
r of th
e triple
s data
sou
r
ce, the
respon
se
tim
e
of the RDF
S
inferen
c
e
wa
s si
gnifica
ntly
increa
sed; S
e
co
nd, the
RDFS infe
re
nce
re
spo
n
se time to use the improved su
b-algo
ri
thm
kno
w
le
dge u
pdating i
s
less than the tra
d
itional algo
ri
thm.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1617 – 162
8
1626
In view of the above exp
e
rime
ntal an
alysis
, the
kn
owle
dge b
a
si
s co
nst
r
u
c
tio
n
method
prop
osed in t
h
is p
ape
r is
e
ffective. It is better
than t
h
e previo
us
un
improve
d
kno
w
led
ge b
a
si
s
in
terms
of RDFS inferen
c
e
time. Namel
y
, by firstly parsi
ng the
different data
source
s an
d the
n
executin
g the sub-algo
rith
m of knowle
d
ge incl
u
s
io
n and the su
b-
algorith
m
Knowle
dge up
d
a
ting
to build the knowl
edge b
a
s
is i
s
pro
m
isi
ng.
Figure 5. Re
spondi
ng time of includi
ng
algorith
m
of different data d
ource
s
Figure 6. Re
spondi
ng time of updating
algorith
m
of different data source
s
Figure 7. Re
spondi
ng time of RDF infe
re
nce b
e
twe
en
the traditional
algorithm a
n
d
our
algorith
m
6. Conclusio
n
In kno
w
led
g
e
engine
eri
ng
field, dome
s
tic an
d forei
g
n
resea
r
ch i
n
st
itutions an
d schol
ars
have made
much effo
rt on kn
owle
dg
e basi
s
co
ns
truction, but
there is n
o
t a comp
aratively
c
o
mplete
k
n
owledge
s
y
s
t
em. In this paper, th
rou
gh
studying
the
kno
w
le
dge
fo
rmation
proce
ss
of multi-sou
r
ce
heteroge
neou
s info
rmation
reso
urces,
we
u
s
ed
the
RDFS de
scripti
on for
sema
ntic
rea
s
oni
ng, an
d
analyzed the
Ontology
kn
owle
dge mo
del
represent
ation;
The
n
, we
provide
d
a f
r
ame
w
o
r
k wi
th kno
w
le
dg
e incl
usi
on
and u
pdatin
g
algo
rithms
for multi-so
u
r
ce
hetero
gen
eo
us i
n
form
ation
kno
w
led
g
e
; Finally
we dem
on
stra
te the effe
ctiveness of
the
algorith
m
thro
ugh a practi
cal appli
c
ation.
Ackn
o
w
l
e
dg
ements
The re
se
arch
is su
ppo
rted
by The
re
sea
r
ch
i
s
suppo
rt
ed by the Nat
i
onal Nature
Scien
c
e
Fund Proj
ect
(612
620
71),
Key Project o
f
Applied
Basi
c
Re
sea
r
ch
Program of Yunna
n Provin
ce
(201
6FA02
4
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.