Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
15
,
No.
1
,
Febr
uary
20
25
, pp.
827
~
835
IS
S
N:
20
88
-
8708
, DO
I: 10
.11
591/ij
ece.v
15
i
1
.
pp
827
-
835
827
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om
An
i
nnovativ
e and e
ffici
ent app
roac
h for
searc
hing and
selecti
ng
web ser
vices o
perations
Sa
r
a Rekk
al
1
, Kahin
a Rek
k
al
2
1
Labo
ratory o
f
LR
I,
Co
m
p
u
ter
Scien
ce De
p
art
m
en
t,
Ba
d
ji M
o
k
h
tar
–
An
n
ab
a Univ
ersity
,
An
n
ab
a,
Alg
eria
2
Electr
ical
E
n
g
in
ee
ring
Depart
m
en
t,
S
alh
i Ahmed
–
Naa
m
a Univ
ersity
Cen
te
r,
N
aam
a,
Alg
eria
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Feb
15, 202
4
Re
vised
Sep 1
2,
2024
Accepte
d
Oct
1,
2024
The
ma
rk
et
ing
of
web
s
erv
ices
on
the
in
te
r
net
continues
t
o
in
cre
ase
,
result
ing
in
an
i
ncr
ea
sing
nu
mb
er
of
web
servi
c
es
and,
th
ere
for
e
,
oper
a
ti
ons
offe
ring
equ
ival
ent
func
ti
ona
li
t
i
es.
As
a
conse
qu
enc
e
,
find
ing
an
appr
opriate
web
service
(o
per
ation)
for
a
pa
rticular
ta
sk
has
bec
o
me
a
difficult
cha
l
le
nge
,
ta
k
in
g
a
lot
of
time
and
l
ea
d
ing
to
an
insuffi
cient
select
ion
of
rel
ev
ant
serv
ices.
Th
is
work
aim
s
to
propose
a
n
ew
appr
oa
ch
f
acili
t
at
ing
the
sea
rch
and
lo
ca
l
i
za
t
ion
of
r
el
ev
an
t
web
servi
ce
s (o
per
ations) i
n
a
n
ac
c
ept
ab
le
ti
me
whi
le
ensu
ring
the
to
ta
l
it
y
of
th
e
response
.
Thi
s
appr
o
ac
h
is
divi
d
ed
int
o
thr
ee c
ru
ci
a
l
phase
s. The
firs
t
step invol
v
es
c
oll
e
ct
ing
web
se
rvic
es
fro
m
var
ious
unive
rsa
l
desc
rip
ti
on,
di
scove
ry,
and
integra
t
ion
(UD
DI)
reg
istries
and
dif
fer
en
t
do
ma
ins
and
form
ing
spec
ialized
sub
-
reg
istri
es.
T
he
se
con
d
phase
invo
lve
s
t
he
ex
tracti
on
of
oper
ations
from
var
ious
servi
ce
s,
foll
owed
by
a
simi
l
arity
study
whos
e
goal
is
th
e
format
ion
of
cl
ust
ers
of
simi
la
r
oper
ations.
The
thi
rd
phase
pr
oce
ss
es
user
r
e
q
uests
by
id
entifying
the
desire
d
fe
at
ure
s
.
A
li
st
of
oper
at
i
ons
is
the
n
provi
ded
to
the
cl
i
ent
,
in
cl
uding
the
non
-
fun
ction
al
prop
erties,
fro
m
which
th
ey
se
le
c
t
th
e
on
e
that
best
m
ee
ts
the
ir
nee
ds
and
b
egi
n
to
invok
e it
.
Ke
yw
or
d
s
:
Cl
us
te
rs
Qu
al
it
y o
f
se
rvi
ces
Re
sp
onse
ti
me
Searc
hing fo
r web
servic
es
Sele
ct
ion
of we
b
se
rv
ic
es
Web ser
vices
operati
ons
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
BY
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
Sara
Re
kk
al
Lab
or
at
or
y of
LRI,
C
ompu
te
r
Science
De
part
ment,
Ba
dji M
okhtar
–
A
nnaba
Unive
rsity
23000 A
nnaba
,
A
lge
ria
Emai
l:
sara.
rekkal@
un
i
v
-
a
nna
ba.dz
1.
INTROD
U
CTION
The
c
omp
uter
sci
ence
fiel
d
ha
s
ev
olv
e
d
a
l
ot
in
rece
nt
ye
ars.
The
siz
e
a
nd
c
omplexit
y
of
s
of
t
war
e
are
co
ns
ta
ntly
grow
i
ng.
T
his
is
du
e
to
t
he
ne
w
an
d
numer
ou
s
nee
ds
that
are
m
or
e
a
nd
more
co
mp
le
x.
Wit
h
this
ev
olu
ti
on,
sever
al
a
pproaches
ha
ve
e
mer
ged
t
o
im
pr
ove
the
pr
oductivit
y
a
nd
eff
ic
ie
nc
y
of
s
of
t
war
e,
includi
ng
ser
vi
ce
-
ori
ente
d
arc
hitec
ture
(
S
OA
),
wh
ic
h has
be
en
acc
ompanie
d by
a
n
e
xpone
ntial
incr
ease i
n
the
numb
e
r of ser
vi
ces an
d
se
rv
ic
e pro
vid
e
rs.
In
t
his
move
m
ent,
the
sea
rc
h
for
the
best
s
erv
ic
es
repres
e
nts
a
real
c
halle
ng
e.
Fin
ding
the
rig
ht
serv
ic
e
to
i
nvoke
impli
es
ha
vi
ng
re
searc
he
d
and
disco
ve
red
a
set
of
pote
nt
ia
l
serv
ic
es.
It
is
then
a
quest
i
on
of
searchi
ng
an
d
sel
ect
ing
the
mo
st
releva
nt
to
in
voke.
S
earchi
ng
f
or
a
nd
sel
ect
ing
r
el
evan
t
web
s
erv
ic
es
pr
ese
nts se
ve
ra
l chall
en
ges rel
at
ed
to:
a.
Heter
og
e
neity
of
te
c
hnologie
s:
Var
io
us
te
c
hnologies
a
nd
protoc
ols
(
e
.g.
,
simple
ob
je
ct
acce
ss
prot
oc
ol
(S
O
AP)
,
re
pr
e
sentat
ion
al
sta
t
e
trans
fer
(RE
ST)
,
rem
ote
procedu
re
cal
l
(
gR
PC)
)
ca
n
be
us
e
d
to
c
onstr
uct
web
se
r
vi
ces.
Be
cause
of
the
se
te
chnolo
gie
s'
div
e
rsity,
the
search
a
nd
sel
ect
ion
of
ser
vi
ces
can
be
ta
il
or
e
d
to meet t
he un
i
qu
e
r
e
quireme
nts
of
a
n
a
ppli
cat
ion
that
is m
or
e
comple
x.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
827
-
835
828
b.
Ser
vice q
ualit
y:
It
mig
ht b
e
c
ha
ll
eng
in
g
to
as
sess
t
he
le
vel of
se
r
vices
th
at
are
of
fer
e
d.
C
ost
,
ef
fecti
ve
nes
s,
avail
abili
ty,
de
penda
bili
ty,
se
cur
it
y,
a
nd
othe
r
facto
rs
ca
n
al
l
aff
ect
how
good
a
se
rv
ic
e
is.
It
is
crit
ic
al
t
o
est
ablish selec
t
ion
proces
ses t
hat consi
der
t
he
se f
act
ors.
c.
Dynamic
e
nv
i
r
onments:
T
hes
e
set
ti
ng
s
al
lo
w
f
or
fr
e
quent
add
it
io
ns
,
del
et
ion
s,
a
nd
up
dates
of
ser
vic
es.
Eff
ect
ive
ma
na
geme
nt
a
nd
a
dap
ta
ti
on
to
c
hanges
in
the
s
e
dy
namics
is
cru
ci
al
f
or
ser
vice
researc
h
and
sel
ect
ion
.
d.
Secu
rity:
To
preve
nt
assa
ults
an
d
protect
da
ta
pr
i
vacy,
it
is
necessa
ry
to
ens
ur
e
the
se
cur
it
y
of
s
peci
fic
serv
ic
es
.
e.
Com
plexity
mana
geme
nt:
Searc
hing
a
nd
c
hoosi
ng
a
mong
th
e
ma
ny
op
ti
ons
offe
red
mi
gh
t
get
com
plica
te
d.
I
t
is
crit
ic
al
to
ha
ve
e
ff
ic
ie
nt
sy
ste
ms
i
n
place
f
or
filt
erin
g,
cat
eg
or
iz
in
g,
a
nd
prese
nting
serv
ic
es
so that
u
se
rs
ca
n
c
hoose
m
or
e ea
s
il
y.
f.
In
te
r
opera
bili
ty:
T
her
e
ma
y
be
prob
le
ms
with
i
nter
op
e
r
abili
ty
w
he
n
web
se
rv
ic
es
are
est
a
blishe
d
by
var
i
ou
s
e
ntit
ie
s.
T
o
gu
a
ran
te
e
s
uccessful
in
te
gr
at
io
n
ac
r
oss
c
ho
se
n
se
rv
i
ces,
t
hese
disti
nctions
must
be
consi
der
e
d d
uri
ng
t
he
se
rv
ic
e
search
and
sel
ect
ion
process
.
g.
Ab
se
nce
of
re
pr
ese
ntati
on
of
non
-
f
unct
ion
a
l
prop
e
rtie
s:
Choosi
ng
ser
vices
that
best
s
at
isfy
us
e
r
nee
ds
i
s
pen
al
iz
e
d
in
th
e abse
nce
of
qual
it
y
of se
r
vice (
Qo
S
)
re
pr
es
entat
ion
s
[
1]
–
[
3]
.
Eff
ect
ive
s
olu
t
ion
s
m
us
t
be
de
velo
ped
to
a
ddress
the
se
c
ha
ll
eng
es
an
d
e
nab
le
ef
fici
ent
an
d
secu
re
us
e o
f
we
b
se
r
vices
in d
ist
rib
uted
a
ppli
cat
ion
s.
This p
a
pe
r
first
ad
dresses the
“co
mp
le
xity
ma
na
geme
nt”
iss
ue
by
pr
opos
in
g
a
n
in
novati
ve
m
et
hod
th
at
al
lo
ws
a
go
od
re
presentat
ion
of
web
ser
vice
op
erati
on
s
cl
assif
ie
d
by
their
f
unct
ion
a
li
ti
es
to
facil
it
a
te
the
searc
h
a
nd
sel
ect
ion
pr
ocess.
It
is
im
portant
t
o
cl
ari
fy
at
t
his
po
i
nt
that
a
cl
ie
nt
seeks
a
s
erv
ic
e
for
it
s
f
un
ct
io
nalit
ie
s,
so
the
re
is
a
n
i
nterest
in
ori
en
ti
ng
the
ser
vic
e
searc
h
to
wa
r
ds
the
op
e
rati
on
sear
ch.
O
nce
the
operati
on
is
f
ound,
t
he
ser
vice
that
offe
rs
it
i
s
disco
ver
e
d.
Seco
nd,
this
a
ppr
oach
so
lves
t
he
chal
le
ng
e
of
a
“
dynamic
en
vir
on
ment.
”
Simi
la
r
op
e
rati
ons
are
groupe
d
into
cl
us
te
rs
in
ord
er
to
rep
la
ce
a
fail
ing
operati
on
with
an
ot
her
s
imi
la
r
on
e
.
More
over,
t
his
method
e
nsure
s
the
up
dating
of
t
he
database
i
n
cas
e
of
delet
ion
or
m
od
ific
at
io
n
of
a
n
op
e
rati
on
by
t
he
ser
vice
pro
vid
er
.
F
ur
t
hermo
re,
th
e
qual
it
y
of
ser
vices
is
t
reated
dif
fer
e
nt
ly
com
pa
red
to
e
xisti
ng
w
ork.
The
sy
ste
m
offer
s
cu
stome
rs
the
co
rr
e
spo
nd
i
ng
op
e
rati
ons
as
well
as
their
qual
it
ie
s,
and
it
is
up
t
o
them
t
o
sel
ect
those
that
su
it
them
be
st.
Finall
y,
t
his
wor
k
ens
ur
es
sec
ur
it
y
by
e
nca
psula
ti
ng
both
the
cl
ie
nt
an
d
the
prov
i
der.
A
me
diator
syst
em
e
nsures
t
he
i
nvoc
at
ion
of
the
desi
red
op
e
rati
ons
a
nd
retu
r
ns
t
he
res
pons
es
to
the
c
li
ent.
T
he
rema
ind
e
r
of
this
do
cume
nt
is
struc
ture
d
as
fo
ll
ow
s:
s
ec
ti
on
2
disc
us
se
s
relat
ed
w
ork
,
s
ect
ion
3
intr
oduces
t
he
pro
po
s
ed
a
ppr
oac
h,
s
ect
io
n
4
ou
tl
ines
the expe
riment
al
r
esults,
and
s
ect
ion
5
c
oncl
ud
e
s the
p
a
per.
2.
RELATE
D
W
ORK
Web
ser
vices
remai
n
e
xtre
mely
c
omplex
te
chnol
og
ie
s,
ge
ner
at
in
g
va
rio
us
c
halle
nge
s.
I
n
this
con
te
xt,
nume
rous
init
ia
ti
ve
s
ha
ve
be
en
unde
rtake
n
to
gu
a
ra
ntee
bo
th
their
a
vaila
bili
ty
an
d
cu
s
tomer
sat
isfact
ion
.
H
uang
a
nd
Z
ha
o
[
4]
dev
ise
a
ser
vice
disc
ov
e
r
y
a
ppr
oac
h
us
i
ng
m
ulti
dimensi
onal
s
erv
ic
e
represe
ntati
on
s
,
empl
oyin
g
m
et
hods
li
ke
W
ord
2V
ec
,
em
be
dd
i
ngs
f
rom
la
ngua
ge
m
ode
ls
(EL
M
o)
,
a
nd
te
rm
fr
e
qu
e
nc
y
-
i
nv
e
rse
do
c
um
e
nt
fr
e
qu
e
nc
y.
Th
ey
c
on
st
ru
c
t
a
simi
la
rity
matr
ix
base
d
on
w
ord
fr
e
qu
e
nc
y,
sta
ti
c
con
te
xt,
a
nd
dyna
mic
c
on
te
xt
featu
res,
e
nh
a
ncin
g
acc
uracy
th
rou
gh
c
onvo
luti
on
,
poolin
g,
a
nd
op
ti
miza
t
ion
in
a
neural
matc
hi
ng
netw
ork.
C
and
i
date
ser
vic
es
are
rate
d
usi
ng
pr
e
dicte
d
s
cor
es
from
t
he
matc
hi
ng
net
work,
yieldin
g
ta
r
get
ser
vices
f
or
s
pecific
qu
e
ries
.
Che
n
a
nd
K
ua
ng
[
5]
i
ntr
oduce
a
novel
m
et
hod
f
or
we
b
serv
ic
e
retrieval,
f
ocus
ing
on
c
ompar
ing
t
he
simi
la
r
it
y
of
se
rv
ic
e
i
nterf
ace
s.
Thei
r
al
gorit
hm
a
ut
om
at
ic
al
ly
adj
us
ts
keyw
ord
w
ei
ghts
in
re
quest
s
to
imp
rove
re
s
ult
accurac
y,
a
ssign
i
ng
l
ow
e
r
weig
hts
to
ke
ywords
s
ha
red
acro
s
s
the ser
vice set
.
Ach
i
r
et
al.
[6]
present
a
ta
xonom
y
that
th
ey
us
e
t
o
cat
e
gorize
ser
vice
disc
over
y
me
thods
in
a
n
internet
of
thin
gs
(IoT)
set
ti
ng.
T
he
y
asses
s
these
te
ch
niqu
es
base
d
on
a
r
ang
e
of
f
act
ors
,
goin
g
i
nto
how
well
they
w
ork
in
va
rio
us
sit
uati
ons
a
nd
ci
rc
umst
ances
as
well
as
t
hei
r
be
nefi
ts
and
dr
a
wb
a
cks.
T
hey
al
so
poin
t
ou
t
diff
ic
ulti
es
and
offer
i
de
as
for
f
uture
avenues
for
th
is
fiel
d
of
stu
dy.
Zarei
a
nd
Gaedke
[7]
present
distrib
uted
se
r
vice
disco
ve
ry
(
DISCO
)
“
We
b
ser
vice
disco
very
chat
bo
t
,
”
a
ch
at
bot
-
dri
ve
n
a
ppr
oac
h
t
ha
t
le
ts
consu
mers
ch
oo
s
e
web
se
r
vices
with
ea
se.
DISCO
m
akes
t
he
i
nter
act
ion
simpler
an
d
rem
oves
th
e
requireme
nt
for
us
ers
to be te
chn
ic
al
ly
kn
owle
dg
ea
ble
by usi
ng con
ver
sat
i
on
al
i
nterfac
es.
In
order
t
o
fi
nd
on
li
ne
ser
vices
i
n
so
ci
a
l
re
posit
or
ie
s,
Liz
ar
ralde
et
al
.
[8]
s
ugge
st
ext
racti
ng
char
act
e
risti
cs
fr
om
ser
vice
de
scriptio
ns
us
i
ng
var
ia
ti
onal
autoe
ncode
rs.
To
trai
n
the
a
ut
oen
c
od
e
r,
the
y
us
e
a
dataset
of
17,
113
ser
vices
f
rom
the
“
Pr
ogram
mableWe
b.
c
om
”
a
pp
li
c
at
ion
pro
gr
a
m
ming
inte
rf
ace
(
AP
I
)
.
Ngu
yen
et
a
l.
[
9]
prese
nt
a
s
ophisti
cat
ed
cl
ust
ering
-
based
t
echn
i
qu
e
f
or we
b
s
er
vice
s
ug
gestio
ns
with
the goal
of
ac
hie
ving
a
wide
r
ra
nge
of
res
ults.
T
hey
pro
vid
e
a
disti
nct
set
of
recommen
datio
ns
by
ta
king
i
nto
accoun
t
functi
onal
inte
rest,
div
e
rsity
trai
ts,
an
d
qu
a
li
ty
of
ser
vice
(QoS
)
pr
e
fer
e
nces.
B
y
us
e
of
we
b
se
rv
ic
e
gr
a
ph
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
An
i
nnov
ative
a
nd eff
ic
ie
nt
appro
ac
h
for
se
ar
c
hing
and
se
le
ct
ing
web
ser
vi
ces o
pera
ti
ons
(
Sa
r
a Rek
kal
)
829
const
ru
ct
io
n
a
nd
var
ie
d
-
widt
h
cl
ust
erin
g,
the
y
i
mpro
ve
the
qual
it
y
of
onli
ne
se
rv
ic
e
sug
gestion
s
by
impro
ving r
e
co
mmendat
io
n
ac
cur
ac
y.
Abram
owic
z
e
t
al.
[
10]
pr
e
se
nt
a
n
a
rch
it
ect
ur
e
f
or
filt
erin
g
a
nd
cl
ust
erin
g
we
b
se
rv
ic
es
that
ma
kes
us
e
of
us
e
r
an
d
ap
plica
ti
on
prof
il
es
t
hat
a
re
sp
eci
fied
us
in
g
we
b
ont
ology
la
nguag
e
for
s
erv
ic
es
(
O
WL
-
S)
.
I
n
order
t
o
imp
r
ov
e
data
re
fin
ement
an
d
sa
ve
execu
ti
on
ti
me,
filt
ers
co
mp
a
re
ser
vice
-
relat
ed
cl
us
te
r
s
us
in
g
cl
us
te
rin
g
a
nal
ys
is.
A
cl
us
te
ring
te
ch
nique
i
s
use
d
i
n
[11]
to
gro
up
heter
og
e
ne
ous
se
r
vices
a
nd
c
on
ce
ntrates
on
web
se
r
vic
e
disc
over
y
usi
ng
O
WL
-
S
in
co
njuncti
on
w
it
h
we
b
ser
vic
es
descr
i
ption
la
nguag
e
(WS
DL)
to
def
i
ne
se
rv
ic
e
semanti
cs.
A
m
e
d
i
a
t
o
r
b
u
i
l
t
o
n
a
g
l
o
b
a
l
o
n
t
o
l
o
g
y
d
r
a
w
n
f
r
o
m
l
o
c
a
l
o
n
t
o
l
o
g
i
e
s
o
b
t
a
i
n
e
d
b
y
h
o
s
p
i
t
a
l
i
n
f
o
r
m
a
t
i
o
n
s
y
s
t
e
m
s
(
H
I
S
)
i
s
u
s
e
d
i
n
[12]
t
o
p
r
o
v
i
d
e
a
s
e
m
a
n
t
i
c
i
n
t
e
r
o
p
e
r
a
b
i
l
i
t
y
a
r
c
h
i
t
e
c
t
u
r
e
.
T
o
h
a
n
d
l
e
p
o
s
s
i
b
l
e
s
e
m
a
n
t
i
c
c
o
n
f
l
i
c
t
s
,
w
e
b
s
e
r
v
i
c
e
s
a
r
e
t
a
g
g
e
d
w
i
t
h
l
o
c
a
l
o
n
t
o
l
o
g
i
e
s
f
r
o
m
a
p
p
l
i
c
a
t
i
o
n
s
.
T
h
e
m
e
d
i
a
t
o
r
h
e
l
p
s
c
o
m
p
o
s
e
w
e
b
s
e
r
v
i
c
e
s
f
o
r
c
o
m
p
l
i
c
a
t
e
d
q
u
e
r
i
e
s
,
b
u
t
i
t
d
o
e
s
n
o
t
g
o
i
n
t
o
g
r
e
a
t
d
e
p
t
h
a
b
o
u
t
h
o
w
t
o
i
m
p
l
e
m
e
n
t
i
t
.
Using
a
mac
hin
e
le
ar
ning
-
base
d
met
hodolo
gy
to
pre
dict
Q
oS
fea
tures
f
rom
s
ource
c
od
e
measu
reme
nts
and
assess
ser
vice
re
puta
ti
on
thr
ough
c
red
i
bili
ty
an
d
us
a
ge
histo
ry,
Ra
ngaraja
n
[
13]
present
a
novel
arc
hitec
ture
for
we
b
s
erv
ic
e
disco
ve
ry
a
nd
sel
ect
ion.
D
ynamic
Hilbert
cl
us
te
r
ing
,
a
te
c
hn
i
que
f
or
cl
assifyin
g
onl
ine
se
rv
ic
es
a
ccordin
g
t
o
c
onve
x
set
sim
il
arit
y
,
is
i
ntr
oduce
d
i
n
[
14]
.
B
y
cal
culat
ing
the
mathemat
ic
al
s
imi
la
rity of SO
AP
messa
ges, t
he
met
hod g
roup
s
the
m int
o ver
y
simi
la
r
clusters
.
M
oha
mme
d
et
al.
[
15]
pr
opose
a
dee
p
le
a
r
ning
fr
a
mew
ork
t
o
e
nhance
s
erv
ic
e
co
mpos
i
ti
on
qu
al
it
y
for
cl
oud
us
er
s,
with
a
f
ocus
on
locat
io
n
awar
e
ness
.
T
he
ir
str
uctu
re
in
it
ia
ll
y
reduces
data
dimens
i
ons
an
d
integrates
pa
rtic
le
swarm
opti
miza
ti
on
with
a
dee
p
le
ar
ning
lo
ng
s
hort
-
te
rm
me
mor
y
ne
twork
t
o
acc
urat
el
y
est
imat
e
Q
oS
values
,
c
onsid
erin
g
l
ocati
on.
T
he
fr
a
mew
ork
ai
ms
to
op
ti
mize
ser
vice
sel
ect
ion
to
r
edu
ce
consu
mer
c
os
t
s
an
d
dem
onstr
at
es
superi
or
predict
io
n
a
nd
c
omposi
ti
on
acc
ur
ac
y
c
ompa
re
d
to
existi
ng
m
od
el
s
on r
eal
dataset
s.
Song
a
nd
Co
[16]
pr
ese
nts
a
te
ch
nique
for
dev
el
op
i
ng
unifie
d
m
odel
in
g
la
ngua
ge
(UML
)
-
base
d
cl
oud
se
r
vices,
integ
rati
ng
cl
oud
a
nd
UML
modeli
ng
el
em
ents
int
o
a
hier
arch
ic
al
meta
model.
This
fa
ci
li
ta
te
s
the
creati
on
of
serv
ic
e
-
ori
ent
ed
cl
ou
d
m
od
e
li
ng
met
hodolog
ie
s
base
d
on
model
-
dr
i
ven
arch
it
ect
ure
(
M
D
A
)
and
model
-
vie
w
-
c
ontrolle
r
(
MVC
),
sim
plif
ying
the
de
sig
n
of
cl
oud
a
ppli
cat
ion
s
a
nd
enab
li
ng
hiera
rch
ic
al
reu
se
m
od
el
s
.
3.
PROP
OSE
D APP
ROAC
H
As
pr
e
viousl
y
sta
te
d,
this
res
earch
ai
ms
to
introd
uce
a
m
et
hodo
l
ogy
f
or
disc
ov
e
rin
g
a
nd
sel
ect
ing
web
ser
vices
op
e
rati
ons.
It
is
imp
or
ta
nt
to
note
that
cl
ie
nts
typ
ic
al
ly
seek
se
r
vices
ba
sed
on
their
functi
onal
it
ie
s.
T
her
e
fore,
it
is
ad
va
ntage
ous
to
fo
c
us
t
he
searc
h
on
fin
ding
sp
eci
fic
op
e
rati
ons.
Once
the
desire
d
operat
ion
is
ide
ntifi
ed,
the
co
rr
es
pondin
g
se
r
vice
is
ide
ntifie
d
t
oo.
T
o
ach
ie
ve
this
obje
ct
ive,
a
reorg
a
nizat
io
n
of
the
we
b
se
r
vices
s
pace
is
man
dato
ry.
Si
nce
it
is
diff
ic
ult,
eve
n
i
mpo
ssible,
to
m
od
i
fy
the
un
i
ver
sal
desc
riptio
n,
disc
overy,
a
nd
inte
grat
ion
(
U
DDI
)
re
gistries,
a
mediat
or
was
pro
po
se
d.
T
he
la
tt
er
performs
the
f
ollow
i
ng tasks
(which
w
il
l be
discusse
d i
n de
ta
il
):
−
Orga
nizing we
b
se
rv
ic
es
acco
rd
i
ng to
t
heir d
iffer
e
nt f
ie
l
ds
.
−
Fo
r
min
g
c
ommu
niti
es o
f
si
mil
ar operat
ions.
−
Handli
ng
us
er
requests.
3.1.
Or
ganizi
ng
web
ser
vices a
cc
ordin
g t
o their dif
fere
nt
fiel
ds
This
orga
nizat
ion
colle
ct
s
WSDLs
from
UDDIs
ac
ro
s
s
va
rio
us
fiel
ds
an
d
cat
e
goriz
es
them
int
o
sp
eci
al
iz
ed
s
ub
-
re
gisters.
Eac
h
WSDL
is
a
s
s
ign
e
d
to
a
s
pec
ific
domain
,
s
uc
h
as
weat
her,
math
operati
on
s,
or
fina
nce,
to
stre
amli
ne
the
res
earch
process.
This
st
ru
ct
ur
e
d
orga
nizat
io
n
s
impli
fies
acce
s
s
an
d
en
ha
nce
s
the
eff
ic
ie
nc
y of s
erv
ic
e
disco
ve
r
y,
a
s il
lustrate
d i
n
Fi
gure
1.
3.2.
F
orming
commu
nities
of simi
lar
op
e
rat
i
on
s
This ste
p
is
of
utmost
imp
or
ta
nce
for
the
foll
ow
i
ng
:
a.
Re
search:
If an
operati
on is
di
sco
ver
e
d,
all
si
mil
ar ones
a
re
disco
ver
e
d
t
oo.
Th
is
re
du
ces
s
earch
ti
me.
b.
Substi
tuti
on
:
O
ne op
e
rati
on is
rep
la
ce
d by an
oth
e
r
if t
he
la
tt
er f
ai
ls.
3.3. C
omm
un
itie
s’
f
orm
at
i
on ste
ps
A
c
om
m
unit
y
i
s
de
fine
d
as
a
set
of
simi
la
r
operati
ons.
T
wo
operati
ons
are
co
ns
ide
re
d
si
mil
ar
if
t
hey
sh
are
ide
ntica
l
inputs
a
nd
ou
t
pu
ts
.
Let
O
a
nd
O′
be
tw
o
op
erati
on
s
e
xtrac
te
d
f
rom
dif
fere
nt
we
b
ser
vic
es
S
1
and S
2.
The
simil
arit
y
bet
we
en
O
a
nd O′ is
cal
c
ulate
d
acc
ordi
ng to
t
he
Algorithm
1:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
827
-
835
830
Figure
1. O
rg
a
nizing we
b ser
vices
by f
ie
ld
Algorith
m
1.
P
rop
os
ed
alg
or
it
hm
(
1
,
2
)
//
(
)
: Measures the degree of similarity between two words.
If word1 or word2 is not in WordNet then
Score
←
Jaro
-
Winkler(word1,word2)
else
Score
←
Wu
-
Palmer(word1,word2)
return Score
(
1
,
2
)
//
(
)
:
Me
as
ur
es
th
e
de
gr
ee
of
si
mi
la
ri
ty
be
tw
ee
n
tw
o
id
en
ti
fi
er
s.
An
identifier
can
be
made
up
of
several
words.
Hungarian
-
Maximum
-
Matching
finds
a
matchin
g
of maximum scores and calculates their average.
for (int i = 0; i < Ident1.length; i++)
for (int j = 0; j < Ident2.length; j++)
Mat[i][j] = SimWord(Ident1[i], Ident2[j])
FinalScore =
Hungarian
-
Maximum
-
Matching (Mat[i][j]))
return (FinalScore)
(
1
,
2
)
//
(
)
: Measures the degree of similarity between two types.
Score
←
Min [Sim (T, T’), Sim (T’, T)]/2
return Score.
(
1
,
2
)
//
(
)
: Measures the degree of similarity between two parameters. Each
parameter has an identifier and a type.
ScoreSimIdent ← SimIdent (partList 1.ident,
partList 2.ident)
ScoreSimType ← SimTypes (partList 1.type, partList 2.type)
FinalSimScore ← (ScoreSimIdent+ ScoreSimType)/2
return FinalSimScore.
(
1
,
2
)
//
(
)
: Measures the degree of
similarity between two messages
(input/output). Each message has a list of parameters.
for (int i = 0; i < msg1.partList.length; i++)
for (int j = 0; j < msg2.partList.length; j++)
Mat[i][j] = SimParts (msg1.partList[i], msg2.partList[j])
FinalScore = Hungarian
-
Maximum
-
Matching (Mat[i][j]))
return FinalScore
(
,
’
)
//
(
)
: Measures the degree of similarity between two operations.
ScoreIdent ← SimIdent(O.Ident, O’.Ident’)
ScoreSimInpMsg ←
SimMessages(O.inputMessage, O’.inputMessage)
ScoreSimOutMsg ← SimMessages(O.OutputMessage, O’.OutputMessage)
FinalScore ← (ScoreIdent + (ScoreSimOutMsg+ ScoreSimInpMsg)/2)/2
return FinalScore.
Co
llecting WS
DL
s Files
1
Creating
specia
liz
ed sub
-
r
eg
isters
UDDI R
eg
istry
M
edia
tor Sy
stem
Specia
lized S
ub
-
r
eg
isters
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
An
i
nnov
ative
a
nd eff
ic
ie
nt
appro
ac
h
for
se
ar
c
hing
and
se
le
ct
ing
web
ser
vi
ces o
pera
ti
ons
(
Sa
r
a Rek
kal
)
831
Key P
oin
ts:
a.
(
)
empl
oys Wu
-
Palmer
simi
la
r
it
y
[
17]
w
he
n
bo
t
h
w
ords
exi
st
in Word
Net;
ot
herwise,
it
ut
il
iz
es
Jaro
-
Win
kler
s
imi
la
rity
[
18]
.
b.
The
H
unga
rian
ma
ximum
ma
tc
hin
g
al
gorit
hm
[
19]
dete
rm
ines
the
ma
xi
mu
m
sc
or
es
a
nd
c
o
m
pu
te
s
their
aver
a
ge
simi
la
r
it
y
sco
re.
c.
(
)
reli
es on a
sim
il
arit
y
ta
ble for
typ
e
s,
as
d
et
ai
le
d
in
source
s
[
20], [
21]
.
3.4.
As
signin
g
a func
tion t
o the
co
m
muni
ties
The
f
un
ct
i
on
is
the
ta
sk
pe
rformed
by
th
e
ope
rati
on.
The
na
me
of
the
ope
rati
on
is
usual
ly
meanin
gful
a
nd
re
pr
e
sents
it
s
ta
sk
.
T
her
e
f
or
e
,
it
is
us
ed
to
desc
ribe
the
functi
on
of
e
ach
co
mm
un
it
y.
T
he
functi
on
ma
y
ha
ve
se
ver
al
ke
ywords
(take
n
from
dif
fer
e
nt
op
e
rati
ons;
re
pe
at
ed
w
ords
ar
e
rem
oved
)
re
f
err
in
g
to it
,
as s
how
n i
n
Fig
ure
2.
3.5.
H
an
dli
n
g
user re
ques
ts
3.5.1.
Search
s
te
ps
As
kn
own,
us
e
r
queries
ca
n
be
simple
(see
king
one
feat
ure)
or
c
omplex
(try
i
ng
to
l
oc
at
e
mu
lt
iple
featur
e
s
i
n
on
e
quer
y).
Sim
ple
queries
po
se
no
prob
le
m
,
unli
ke
c
omplex
on
e
s.
To
re
me
dy
this
an
d
fac
il
it
at
e
the searc
h see
Figure
2, it
is re
comme
nde
d
t
o:
−
Simpli
f
y
the
c
omplex
re
qu
e
s
t
into
s
ub
-
requ
est
s,
al
lowi
ng
the
cl
ie
nt
to
de
te
rmin
e
t
he
num
ber
of
featu
res
they wis
h
to
se
arch f
o
r
at the
beg
i
nn
i
ng
;
−
Pr
ovi
de
a s
peci
fic num
ber o
f key
w
ords refe
r
rin
g
to
each
d
e
sired f
un
ct
io
n
;
−
Re
trie
ve
th
os
e
ke
ywo
rds
a
nd
matc
h
the
m
with
t
hose
li
ste
d
i
n
t
he
ta
ble,
re
su
lt
in
g
i
n
a
li
st
of
operati
ons
from w
hich
t
he
cli
ent can
selec
t t
he
one t
hat
best s
uits
their
needs.
Figure
2. Searc
h
ste
ps
3.5.2. Tex
t
si
mi
lariti
es
In
this
wor
k,
it
is
necessar
y
to
assess
the
si
mil
arit
y
betwe
en
tw
o
ba
gs
of
word
s:
the
fir
st
rep
re
sent
s
the
ke
ywords
entere
d
by
the
us
er
,
an
d
the
s
econd
re
prese
nt
s
the
keyw
ord
s
descr
i
bing
th
e
functi
on
pro
vid
e
d
by
each
co
mm
un
it
y.
T
her
e
ar
e
seve
ral
meth
od
s
f
o
r
cal
cula
ti
ng
this
meas
ure,
incl
udin
g
C
os
ine
,
Jacca
rd,
Dice,
Pears
on,
a
nd
Eucli
dea
n,
a
m
ong
oth
e
rs.
A
c
cordin
g
t
o
[22
],
[23
]
,
t
he
performa
nces
of
Cosine
simi
la
r
it
y,
the
Jacca
rd
coe
ff
i
ci
ent,
an
d
the
Pears
on
coe
ffi
ci
ent
are
ve
ry
cl
os
e
an
d
si
gnific
antly
bett
er
tha
n
th
os
e
of
t
he
Eucli
dea
n
distance.
Acc
ordi
ng
to
[
24]
,
the
Dice
i
nd
e
x
a
nd
the
Jacca
rd
ind
e
x
ha
ve
sim
il
ar
pe
rforman
ces.
I
n
this w
ork, Jacc
ard simi
la
rity
[
25]
was
c
hose
n; it
is simple
a
nd ef
fecti
ve, a
nd the
res
ults a
re s
at
isf
ying.
3.5.3. T
hresh
ol
d
d
etermi
n
ati
on
The
de
gr
ees
of
simi
la
rity
ret
urne
d
by
the
Ja
ccard
meas
ur
e
ra
ng
e
bet
wee
n
[0,
1]
.
Acc
ordi
ng
to
our
ob
s
er
vations, t
her
e
are five
subinter
vals:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
827
-
835
832
−
[0,
0.2]
:
No si
mil
arit
y.
−
[0.2,
0.5]
: Li
tt
le
simil
arit
y.
−
[0.5,
0.7]
:
Av
e
rag
e
simil
arit
y.
−
[0.7,
0.9]
:
Very simil
ar.
−
[0.9,
1.0]
: Simi
la
r.
Th
us
,
the simi
la
rity
th
res
ho
l
d
sta
rts
f
rom 0
.
7,
mea
ning
t
hat
two bag
s
are
c
onside
red
simi
la
r
if
an
d
only
if
they
hav
e
a
sc
or
e
great
er
tha
n
or
equ
al
t
o
0.7
.
Con
se
quently
,
the
co
nce
rn
e
d
functi
on
or
c
ommu
nity
(a
nd
al
l
it
s
op
e
rati
ons)
w
il
l be
r
et
urne
d.
3.5.4. T
he
sel
e
ction
pr
ocess
The
la
st
ste
p
r
et
urns
a
li
st
of
operati
ons
f
rom
w
hic
h
the
c
li
ent
sel
ect
s
one
operati
on
over
an
oth
e
r
base
d
on
l
y
on
it
s
non
-
f
un
ct
io
nal
pr
op
e
rtie
s.
To
il
lustrate
t
hi
s
idea,
c
onsid
er
the
f
ollo
wing
e
xam
ple:
a
c
li
ent
search
e
s
f
or
functi
on
1.
The
e
ntire
cl
us
te
r
(O1,
O2,
O
3)
is
r
et
urned,
an
d
th
ei
r
non
-
f
unct
io
nal
crit
eria
are
al
so
disp
la
yed
as s
how
n
i
n
Fi
gure
3
. T
he
cli
ent th
en
sel
ect
s th
e
operati
on t
hat best
su
it
s their
nee
ds
.
Figure
3. Sele
c
ti
on
resu
lt
s
4.
RESU
LT
S
AND
DI
SCUS
S
ION
4.1.
Ex
peri
menta
l
e
nv
ir
onme
n
t
The
e
xp
e
rime
ntati
on
was
c
arr
ie
d
out
usi
ng
a
set
of
r
eal
web
s
er
vices
source
d
from
dif
fer
e
nt
domains
,
incl
udin
g
mes
sagi
ng,
weathe
r,
a
nd
re
ser
vatio
n
s
ys
te
ms.
Th
ese
serv
ic
es
we
re
caref
ully
sel
ec
te
d
to
ens
ur
e
a
di
verse
re
pr
ese
ntati
on
of
us
e
cas
es,
co
ve
rin
g
a
wide
sp
ect
rum
of
pr
act
ic
al
ap
plica
ti
on
s.
Thi
s
div
e
rsity
was
cru
ci
al
f
or
th
oro
ughl
y
eval
ua
ti
ng
the
r
obust
ness
of
our
appr
oach
a
cr
oss
va
rio
us
sce
nar
i
os
.
Con
se
quently
, i
t al
lowed
us t
o assess
how
ef
f
ic
ie
ntly our me
thod
c
ou
l
d handle real
-
w
or
l
d com
plexiti
es.
4.2.
A
ppli
ca
tion
d
evel
op
me
nt
The
a
pp
li
cat
io
n
us
ed
f
or
this
researc
h
was
dev
el
op
e
d
us
i
ng
NetB
eans,
w
hich
is
a
r
obust
integrate
d
dev
el
opment
e
nv
i
ronme
nt
(IDE)
ba
sed
on
J
ava.
The
pri
ma
ry
obje
ct
ive
of
this
ap
plica
ti
on
is
t
o
strea
mli
ne
a
nd
op
ti
mize
the
proces
s
of
sea
r
chin
g
f
or
rele
van
t
op
e
rati
ons.
T
o
ac
hieve
this,
t
he
ap
plica
ti
on
gro
ups
simi
la
r
op
e
rati
ons
int
o
cl
us
te
rs
,
org
anizi
ng
t
hem
base
d
on
c
ommo
n
featu
res.
This
cl
us
te
ri
ng
ste
p
sim
plifie
s
the
su
bse
que
nt
sea
rch
by
na
r
ro
wi
ng
dow
n
t
he
opti
ons
an
d
ma
king
it
easi
er
t
o
locat
e
the
de
sired
functi
ona
li
ti
e
s
eff
ic
ie
ntly
.
4.3.
Me
thod
ol
og
y
f
or
re
orga
nizi
ng
web ser
vices sp
ace
4.3.1.
Org
an
iz
ing
w
eb ser
vices by
do
m
ain
The
fir
st
ste
p
in
our
ap
pro
ach
is
to
orga
nize
we
b
ser
vi
ces
WSDL
accor
ding
to
th
ei
r
sp
eci
fi
c
app
li
cat
io
n
do
mains,
s
uc
h
a
s
we
at
her,
mess
agin
g,
a
nd
rese
rv
at
io
ns
.
T
his
init
ia
l
cat
egoriz
at
ion
hel
ps
str
uctu
re
the
avail
able
s
erv
ic
es
a
nd
fa
ci
li
ta
te
s
their
su
bse
que
nt
ma
nag
e
ment.
B
y
gro
up
i
ng
we
b
serv
ic
es
i
nto
di
sti
nct
domains
, we la
y
the
fo
undatio
n for m
ore ta
r
ge
te
d
an
d
e
ff
ic
i
ent searc
hes
.
4.3.2. F
orm
ati
on
of
c
ommuni
ties of simi
lar
o
per
ati
on
s
On
ce
the
we
b
ser
vices
are
orga
nized
by
domain,
we
procee
d
with
a
more
detai
le
d
a
nalysis
t
o
identif
y
an
d
gro
up
simi
la
r
operati
ons
within
each
domain
.
These
“
c
ommu
niti
es
”
of
operati
ons
ar
e
f
ormed
base
d
on
f
unct
ion
al
simi
la
riti
es,
mea
ning
op
erati
on
s
t
hat
pe
rf
orm
co
mp
a
r
able
ta
sk
s.
Thi
s
gro
up
i
ng
has
a
du
al
pur
po
se:
it
sim
plifie
s
the
sea
r
ch
by
al
lo
wing
simi
la
r
operati
on
s
to b
e q
uic
kl
y
f
ound
w
he
n
a
releva
nt
oper
at
ion
is i
den
ti
fie
d,
a
nd it
f
aci
li
ta
te
s the su
bs
ti
tuti
on
of operati
ons
when an
alt
ernat
ive is n
ee
de
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
An
i
nnov
ative
a
nd eff
ic
ie
nt
appro
ac
h
for
se
ar
c
hing
and
se
le
ct
ing
web
ser
vi
ces o
pera
ti
ons
(
Sa
r
a Rek
kal
)
833
4.
4.
Se
arch
p
rocess
The
a
pp
li
cat
io
n
ma
na
ges
use
r
queries
base
d
on
the
desire
d
functi
onal
it
ie
s.
W
he
n
a
use
r
sp
eci
fies
the
op
e
rati
ons
the
y
need,
the
sea
rch
is
la
unche
d
across
the
pre
viously
f
ormed
com
munit
ie
s
of
operati
ons
.
Tha
nk
s
to
this
cl
us
te
ri
ng
orga
nizat
io
n,
no
t
on
l
y
is
the
exact
ope
r
at
ion
quic
kl
y
f
ound,
bu
t
al
l
simi
la
r
op
e
rati
ons
are
al
so
propose
d,
offer
i
ng the
use
r
a c
ompre
he
ns
ive
set o
f rel
evan
t
opti
ons
a
s sho
wn in
Fig
ur
e
4
.
Figure
4. Searc
h
to
ol
for web
serv
ic
es
operat
ion
s
4.5. V
alida
tio
n and
analy
sis
Since
it
is
dif
ficult
,
if
no
t
imp
os
sible,
to
fin
d
simi
la
r
to
ols
for
c
ompa
rison,
we
have
us
e
d
tw
o
measu
res,
prec
isi
on
a
nd
rec
al
l,
as
s
how
n
in
Figure
5,
to
va
li
date
the
re
su
l
ts,
pa
rtic
ularly
the
rele
van
ce
of
the
sel
ect
ion
be
for
e
a
nd
a
fter
the
re
orga
nizat
ion
of
the
web
se
r
vices
sp
ace
.
T
hese
meas
ur
e
s
al
low
us
t
o
qu
antify
the
accu
racy
of
our
resu
lt
s,
pro
vid
i
ng
a
cl
ea
rer
picture
of
how
well
our
method
ide
ntif
ie
s
relevan
t
se
rv
ic
es.
Additi
on
al
l
y,
we
c
onduct
ed
f
ur
t
her
anal
ysi
s
to
eval
uate
the
imp
r
ov
e
ment
i
n
s
earc
h
ti
me,
as
s
how
n
i
n
Figure
6, w
hic
h
is c
ru
ci
al
for unde
rstan
ding
the ove
rall
eff
i
ci
ency
of
our
a
ppr
oach.
We
note
a
dis
cern
i
ble
impro
veme
nt
in
the
per
ti
ne
nt
sel
ec
ti
on
of
we
b
se
rv
ic
e
op
e
rati
ons
afte
r
the
rear
rangeme
nt
of
t
he
we
b
se
rv
ic
es
s
pace.
This
imp
r
ov
e
ment
is
re
pr
e
s
ented
in
a
hi
gher
acc
uracy
measu
re,
wh
ic
h
i
nd
ic
at
e
s
the
r
el
eva
nce
of
the
ope
rati
on
s
retu
rn
e
d,
a
nd
a
hi
gher
rec
al
l
measu
re,
w
hich
in
dicat
es
t
hat
al
l
releva
nt
opera
ti
on
s
wer
e
re
tur
ned
beca
use
of
the
cl
ust
ers
that
were
est
ablishe
d.
A
dd
it
i
onal
ly,
this
rear
rangeme
nt
has
le
d
to
a
sign
ific
a
nt
dec
rease
i
n
sea
rc
h
ti
me.
It
is
mu
c
h
more
e
f
fici
ent
to
sea
r
ch
i
n
a
well
-
str
uctur
e
d
env
i
ronme
nt,
wh
e
re
act
io
ns
are
gro
upe
d
in
to
cl
us
te
rs
a
nd
cl
assifi
ed
by
domain,
tha
n
in
an
unorga
nized
one.
T
he
ac
qu
ired
fin
dings
dem
onstrat
e
the
in
disputabl
e
eff
ic
ac
y
of
our
m
et
hodo
logy,
emp
hasizi
ng
it
s
capaci
ty
t
o
preci
sel
y
pinpo
i
nt
pe
rtinent
act
ion
s
.
O
ur
meth
od
's
ef
ficacy
ha
s
bee
n
vali
dated
b
y
meti
culo
us
.
Figure
5. Pr
eci
sion an
d recal
l
befor
e
and
a
fte
r
the
r
e
orga
nizat
ion
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
827
-
835
834
Figure
6
.
Resp
on
s
e
t
ime
5.
CONCL
US
I
O
N
Ov
e
r
the
la
st
decad
e
,
an
e
xpone
ntial
incre
ase
in
the
numb
e
r
of
se
r
vices
a
vaila
ble
on
t
he
we
b
has
been
obser
ve
d.
This
sit
uatio
n
necessit
at
es
th
e
creati
on
of
a
n
ef
fici
ent
so
l
ut
ion
f
or
t
he
sea
rch
a
nd
sel
ect
ion
of
web
ser
vices.
I
n
this
w
ork,
a
simple
yet
hi
ghly
e
ff
ect
i
ve
a
ppr
oach
was
pro
posed
,
e
nsur
ing
t
he
c
omple
te
ness
of
the
res
pons
e
as
well
as
the
releva
nce
of
th
e
sel
ect
ion
within
a
n
acce
pta
ble
ti
me
f
rame
.
It
in
volves
di
vi
din
g
the
cl
ie
nt's
r
eq
uest
i
nto
se
veral
su
b
-
f
un
ct
io
ns
and
sea
rc
hing
f
or
the
opera
ti
on
s
that
pe
rform
these
f
unc
ti
on
s.
The
cl
ie
nt
the
n
sel
ect
s
the
one(s)
t
hat
best
meet
their
nee
ds
ba
sed
on
no
n
-
functi
onal
pr
op
e
rtie
s
an
d
be
gin
s
t
o
us
e it
.
REFERE
NCE
S
[1]
Z.
Gu
i,
J.
Cao
,
X.
Liu, X
.
Ch
en
g
,
an
d
H.
Wu
,
“Glo
b
al
-
s
cale res
o
u
rce
su
rve
y
and
perfor
m
an
ce
m
o
n
ito
ring
of pu
b
lic OGC
web
m
ap
serv
ices,”
ISPRS
I
n
tern
a
tio
n
a
l Jou
rnal o
f Geo
-
Info
rma
tio
n
,
v
o
l.
5
,
n
o
.
6
,
p
p
.
1
–
24,
Ju
n
.
2
0
1
6
,
d
o
i: 10
.33
9
0
/ijg
i5
0
6
0
0
8
8
.
[2]
K.
Hu
,
Z.
Gu
i,
X.
Ch
en
g
,
H.
W
u
,
an
d
S.
C.
McClu
re
,
“
The
co
n
cept
an
d
te
ch
n
o
lo
g
ies
o
f
q
u
al
ity
o
f
g
eo
g
raph
ic
in
form
atio
n
serv
ice
:
im
p
rov
in
g
u
ser
ex
p
erience
o
f
GIS
ervices
in
a
d
istr
ib
u
ted
co
m
p
u
tin
g
en
v
iron
m
en
t,”
IS
PR
S
Inter
n
a
tio
n
a
l
Jo
u
rn
a
l
o
f
Geo
-
Info
rma
tio
n
,
v
o
l.
8
,
n
o
.
3
,
p
p
.
1
–
2
6
,
Mar
.
20
1
9
,
d
o
i: 10.
3
3
9
0
/ijg
i
8
0
3
0
1
1
8
.
[3]
Y.
Ch
en
g
,
W
.
G
e,
an
d
L
.
Xu
,
“
Qu
ality
o
f
g
eo
g
raph
ical
in
form
ati
o
n
serv
ices
ev
alu
atio
n
b
ased
o
n
o
rder
-
rel
atio
n
,”
in
Co
mmu
n
ica
tio
n
s i
n
Co
mp
u
ter a
n
d
I
n
f
o
rma
tio
n
Scien
ce
,
2
0
1
8
,
p
p
.
6
7
9
–
6
8
8
.
[4]
Z.
Hu
an
g
an
d
W
.
Zhao
,
“A
se
m
an
tic
m
at
ch
in
g
ap
p
roach
ad
d
ressin
g
m
u
ltid
im
en
sio
n
al
repres
en
tatio
n
s
for
web
s
ervice
d
isco
v
ery,”
S
S
RN E
lectro
n
ic J
o
u
rn
a
l
,
v
o
l.
2
1
0
,
2
0
2
2
,
d
o
i: 10
.2
1
3
9
/
ss
rn.4
0
7
6
7
0
9
.
[5]
K.
Ch
en
an
d
C.
Ku
an
g
,
“Web
ser
v
ice
d
isco
v
ery
b
ased
o
n
m
ax
im
u
m
weig
h
ted
b
ip
artite
g
raph
s,”
Co
mp
u
ter
Co
mmu
n
ica
tio
n
s
,
v
o
l.
1
7
1
,
p
p
.
5
4
–
6
0
,
Ap
r.
20
2
1
,
d
o
i: 10
.10
1
6
/j.comcom
.20
2
1
.01
.0
3
1
.
[6]
M.
Ach
ir
,
A.
Ab
d
elli,
L.
Mok
d
ad
,
an
d
J.
Ben
o
th
m
an
,
“
Service
d
isco
v
ery
an
d
selectio
n
in
I
o
T:
a
su
rvey
an
d
a
t
ax
o
n
o
m
y
,”
Jo
u
rn
a
l
o
f Netwo
rk a
n
d
Co
mp
u
ter App
lica
tio
n
s
,
v
o
l.
2
0
0
,
p
p
.
1
–
4
4
,
Ap
r.
20
2
2
,
d
o
i: 10
.10
1
6
/j.jnca.2
0
2
1
.10
3
3
3
1
.
[7]
B.
Zarei
an
d
M.
G
aedk
e,
“Disco
:
we
b
serv
ice
d
isco
v
ery
ch
atb
o
t,”
IA
DIS
Inter
n
a
tio
n
a
l
Jo
u
r
n
a
l
o
n
WW
W/In
ter
n
et
,
v
o
l
.
1
8
,
n
o
.
2
,
p
p
.
1
6
–
2
8
,
Dec.
2
0
2
0
,
d
o
i: 10
.3396
5
/i
jwi_
2
0
2
0
1
8
2
0
2
.
[8]
I.
L
izar
ralde,
C
.
Mateos
,
A
.
Zun
in
o
,
T.
A.
Majchr
z
ak
,
an
d
T.
-
M.
Gr
ø
n
li,
“Disco
v
ering
web
serv
ices
in
so
cial
web
serv
ic
e
repo
sito
ries
u
sin
g
d
eep
v
ariation
al
au
to
en
co
d
ers,”
Informa
tio
n
Pro
c
ess
in
g
&
Ma
n
a
g
ement
,
v
o
l.
5
7
,
n
o
.
4
,
p
p
.
1
–
1
9
,
Ju
l.
2
0
2
0
,
d
o
i: 10
.1016
/j.ipm.20
2
0
.1
0
2
2
3
1
.
[9]
H.
H.
Cu
o
n
g
Ng
u
y
en
,
B.
T.
Kh
iet,
V.
L.
Ng
u
y
en
,
an
d
T.
T.
Ng
u
y
en
,
“An
eff
ectiv
e
m
eth
o
d
for
clu
sterin
g
-
b
ased
web
serv
ice
recom
m
en
d
atio
n
,”
Inter
n
a
tio
n
a
l
Jo
u
rn
a
l
o
f
E
lectrica
l
a
n
d
Co
mp
u
ter
Eng
in
eerin
g
,
v
o
l.
1
2
,
n
o
.
2
,
p
p
.
1
5
7
1
–
1
5
7
8
,
Ap
r.
2
0
2
2
,
d
o
i: 10
.1159
1
/ijec
e.v1
2
i2
.pp
1
5
7
1
-
1
5
7
8
.
[10
]
W
.
Ab
ra
m
o
wicz
,
K.
H
an
iewicz
,
M.
Kacz
m
a
rek,
an
d
D.
Zys
k
o
wsk
i,
“A
rchitectu
re
for
we
b
serv
ices
filter
in
g
an
d
clu
sterin
g
,”
in
S
econ
d
Int
er
n
a
tio
n
a
l
Co
n
fere
n
ce
o
n
Inter
n
et
a
n
d
Web
Ap
p
lica
tio
n
s
a
n
d
S
ervices
(I
CI
W’0
7
)
,
May
2
0
0
7
,
d
o
i: 10
.1109
/ICIW.2
0
0
7
.19
.
[11
]
R.
Nay
ak
an
d
B.
Lee
,
“Web
se
rvice
d
isco
v
ery
with
ad
d
itio
n
al
sem
an
tics
an
d
clu
sterin
g
,”
in
IE
EE
/W
IC/ACM
Inter
n
a
tio
n
a
l
Co
n
feren
ce on
Web
Intellig
en
ce (
WI
’
0
7
)
,
No
v
.
2
0
0
7
,
p
p
.
5
5
5
–
5
5
8
,
d
o
i: 10
.
1
1
0
9
/W
I.
2
0
0
7
.82
.
[12
]
L.
S
.
Taps
o
b
a,
Y
.
Tr
ao
re,
an
d
S.
Malo,
“Towards
a
n
architectu
re
for
th
e
in
terop
erabilit
y
o
f
h
o
sp
ital
in
form
atio
n
sy
ste
m
s
i
n
Bu
rkin
a Faso
,”
S
tu
d
ies in
Health
Tec
h
n
o
lo
g
y an
d
I
n
fo
r
ma
tics
,
v
o
l.
2
7
2
,
p
p
.
1
5
9
–
1
6
2
,
Ju
n
.
2
0
2
0
,
d
o
i: 10
.3
2
3
3
/
SHTI
2
0
0
5
1
8
.
[13
]
S.
Ran
g
araja
n
,
“
Qo
S
-
b
ased
web
s
ervice
d
isco
v
ery
an
d
selectio
n
u
sin
g
m
achi
n
e
lea
rnin
g
,”
ICS
T
Tra
n
sa
ctio
n
s
o
n
S
ca
la
b
le
Info
rma
tio
n
Sys
te
ms
,
v
o
l.
5
,
n
o
.
1
7
,
p
p
.
1
–
8
,
May 2
0
1
8
,
d
o
i: 10
.41
0
8
/
eai.
2
9
-
5
-
2
0
1
8
.15
4
8
0
9
.
[14
]
N.
A.
Al
-
Mus
awi
an
d
D.
Al
-
Sh
a
m
m
ary,
“
Dy
n
am
ic
Hilb
ert
clu
sterin
g
b
ased
o
n
co
n
v
ex
set
for
web
serv
ices
ag
g
regatio
n
,”
Inter
n
a
tio
n
a
l
Jo
u
rn
a
l
o
f
Electri
c
a
l
a
n
d
Co
mp
u
ter
Eng
in
eerin
g
,
v
o
l.
1
3
,
n
o
.
6
,
p
p
.
6
6
5
4
–
6
6
6
2
,
Dec
.
2
0
2
3
,
d
o
i:
1
0
.11
5
9
1
/ijece.v1
3
i6
.pp
6
6
5
4
-
6
6
6
2
.
[15
]
A.
M
.
Moh
am
m
ed
,
S.
S.
A.
H
ay
tam
y
,
an
d
F.
A
.
O
m
ar
a,
“Locat
io
n
-
awar
e
d
eep
learnin
g
-
b
a
sed
fr
a
m
ewo
rk
for
o
p
tim
izin
g
clo
u
d
co
n
su
m
er
q
u
ality
o
f
serv
ice
-
b
ased
s
ervice
co
m
p
o
sitio
n
,”
Inter
n
a
tio
n
a
l
Jo
u
rn
a
l
o
f
Electric
a
l
a
n
d
Co
mp
u
ter
Eng
in
eerin
g
,
v
o
l.
1
3
,
n
o
.
1
,
p
p
.
6
3
8
–
6
5
0
,
Feb
.
2
0
2
3
,
d
o
i: 10
.11
5
9
1
/ijece.v1
3
i
1
.pp
6
3
8
-
650.
[16
]
C.
-
Y.
So
n
g
an
d
E.
-
S.
Ch
o
,
“A
se
rvice
-
o
riented
clo
u
d
m
o
d
elin
g
m
eth
o
d
an
d
p
rocess
,”
Inte
rn
a
tio
n
a
l
Jo
u
rnal
o
f
Electrica
l
a
n
d
Co
mp
u
ter Eng
in
ee
rin
g
,
v
o
l.
1
0
,
n
o
.
1
,
p
p
.
9
6
2
–
9
7
7
,
Feb
.
2
0
2
0
,
d
o
i
: 10
.115
9
1
/ijece.v1
0
i1
.pp9
6
2
-
977.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
An
i
nnov
ative
a
nd eff
ic
ie
nt
appro
ac
h
for
se
ar
c
hing
and
se
le
ct
ing
web
ser
vi
ces o
pera
ti
ons
(
Sa
r
a Rek
kal
)
835
[17
]
M.
Sh
en
o
y
,
“A
n
ew
si
m
ila
rity
m
ea
su
re
for
tax
o
n
o
m
y
b
ased
o
n
ed
g
e
co
u
n
tin
g
,”
Inter
n
a
tio
n
a
l
jo
u
rnal
o
f
Web
&
S
ema
n
tic
Tech
n
o
lo
g
y
,
v
o
l.
3
,
n
o
.
4
,
p
p
.
2
3
–
3
0
,
Oct.
2
0
1
2
,
d
o
i: 10
.51
2
1
/ijwest.2
0
1
2
.3
4
0
3
.
[18
]
O.
Ro
zin
ek
an
d
J.
Mar
eš
,
“Fast
an
d
p
recise
co
n
v
o
lu
ti
o
n
al
Jaro
an
d
Jaro
-
W
in
k
ler
simila
rit
y
,”
in
2
0
2
4
3
5
th
Co
n
feren
ce
o
f
Op
en
Inn
o
va
tio
n
s Asso
ci
a
tio
n
(
FR
UCT)
,
A
p
r.
20
2
4
,
p
p
.
6
0
4
–
6
1
3
,
d
o
i: 10
.2
3
9
1
9
/FRUC
T61
8
7
0
.2024
.10
5
1
6
3
6
0
.
[19
]
Y.
Zeng
,
X.
W
u
,
a
n
d
J.
Cao
,
“Res
ear
ch
an
d
im
p
le
m
en
ta
tio
n
o
f
Hu
n
g
arian
m
eth
o
d
b
ased
o
n
t
h
e
stru
ctu
re
in
d
ex
redu
ctio
n
for
DAE
sy
stems,”
Jo
u
rn
a
l
o
f
Algo
rith
ms
&
Co
mp
u
ta
tio
n
a
l
T
echn
o
lo
g
y
,
v
o
l.
8
,
n
o
.
2
,
p
p
.
2
1
9
–
2
3
1
,
Jun.
2
0
1
4
,
d
o
i:
1
0
.1260
/1
7
4
8
-
3
0
1
8
.8.2
.2
1
9
.
[20
]
E.
St
rou
lia
an
d
Y.
W
an
g
,
“Structu
r
al
an
d
se
m
an
tic
m
atch
in
g
for
ass
e
ss
in
g
web
-
serv
ice
simila
rity,”
Inter
n
a
tio
n
a
l
Jo
u
rnal
o
f
Co
o
p
era
tive I
n
fo
r
ma
tio
n
Sys
tems
,
v
o
l.
1
4
,
n
o
.
4
,
p
p
.
4
0
7
–
4
3
7
,
Dec.
2
0
0
5
,
d
o
i: 10
.11
4
2
/S0
2
1
8
8
4
3
0
0
5
0
0
1
2
1
3
.
[21
]
P.
Pl
eb
an
i
an
d
B.
Pernici,
“URB
E:
web
serv
ice
r
etrie
v
al
b
ased
o
n
s
im
il
arity
ev
alu
atio
n
,”
I
EE
E
Tra
n
sa
ctio
n
s
o
n
Kn
o
wled
g
e
a
n
d
Da
ta
E
n
g
in
eerin
g
,
vo
l.
2
1
,
n
o
.
1
1
,
p
p
.
1
6
2
9
–
1
6
4
2
,
No
v
.
20
0
9
,
d
o
i: 10
.1
1
0
9
/TKDE
.20
0
9
.35
.
[22
]
A.
Hu
an
g
,
“Si
m
il
arity
m
easu
res
fo
r
tex
t
d
o
cu
m
en
t
clu
sterin
g
,”
in
6
th
New
Z
ea
la
n
d
Co
mp
u
ter
S
cien
ce
Resear
c
h
S
tu
d
en
t
Co
n
feren
ce (
NZC
S
RSC
2
0
0
8
)
,
Ap
r.
20
0
8
,
p
p
.
4
9
–
5
6
.
[23
]
T.
Z.
B
ah
arav,
G.
M.
Kam
ath
,
D
.
N
.
Tse,
an
d
I.
Sh
o
m
o
ron
y
,
“Sp
ectra
l
J
accar
d
si
m
ila
rity:
a
n
ew
ap
p
roach
to
estima
tin
g
p
airwi
se
seq
u
en
ce a
lig
n
m
en
ts,”
Patter
n
s
,
v
o
l.
1
,
n
o
.
6
,
p
p
.
1
–
1
0
,
Sep
.
2
0
2
0
,
d
o
i: 10
.
1
0
1
6
/j.patter.20
2
0
.
1
0
0
0
8
1
.
[24
]
E.
Neg
re
,
“Co
m
p
araison
d
e
tex
tes:
q
u
elq
u
es
ap
p
roch
es...
,”
h
al
-
00874280,
20
13.
[
O
nl
i
n
e
]
.
A
va
i
l
a
bl
e
:
ht
t
ps:
/
/
ha
l
.
sc
i
e
nc
e
/
ha
l
-
0
0
8
7
4
2
8
0
v
1
[25
]
İ.
Kab
asak
al
an
d
H.
So
y
u
er,
“A
Ja
ccar
d
simil
arity
-
b
a
sed
m
o
d
el
to
m
at
ch
stak
e
h
o
ld
ers
fo
r
co
llab
o
ration
in
an
in
d
u
stry
-
d
riven
p
o
rtal,
”
Th
e
7
th
Inter
n
a
tio
n
a
l
Mana
g
ement
Info
rma
tio
n
S
ystems
Co
n
feren
ce
,
v
o
l.
7
4
,
n
o
.
1
,
p
p
.
1
–
9
,
Ma
r.
2
0
2
1
,
d
o
i:
1
0
.33
9
0
/p
roceed
in
g
s2
0
2
1
0
7
4
0
1
5
.
BIOGR
AP
HI
ES OF
A
UTH
ORS
Sara
Rek
k
a
l
born
in
Bec
h
ar,
Alger
ia,
th
e
author
obta
in
ed
her
ma
ster's
degr
ee
i
n
com
put
er
sci
ence
from
th
e
Univ
ersit
y
of
Be
cha
r
in
2015
and
completed
he
r
doc
to
rat
e
in
2019
at
the
Univ
ersit
y
of
Ahmed
B
en
Bel
la
Or
an
1
.
She
is
an
assistan
t
prof
essor
a
t
Ba
dji
Mokhtar
Anna
ba
Univer
s
it
y
in
Alg
eria.
Her
prim
a
ry
ar
e
as
of
expe
r
ti
se
i
ncl
ude
web
serv
ic
es,
mul
t
i
-
cri
t
eri
a
decision
support,
e
mbedded
sys
te
ms
e
ngine
er
ing
and
par
alle
li
sm,
m
et
ah
eur
isti
c
al
gorit
h
ms,
and
nat
ur
al
la
nguag
e
proc
essing.
F
or
inqui
ri
es,
she
ca
n
be
contac
t
ed
a
t
e
ma
i
l:
sara
.
r
ekka
l@uni
v
-
anna
ba
.
dz
.
Kahina
Rek
k
al
born
in
Be
cha
r,
Alg
eria,
th
e
aut
hor
ea
rn
ed
her
Dipl.
In
f.
-
Ing
.
degr
ee
from
th
e
Univer
sity
of
B
ec
har
in
2009,
a
ma
st
er's
degr
ee
in
2012
,
and
a
doct
ora
te
i
n
2018,
a
ll
fro
m
th
e
sam
e
univ
ersity.
Curre
n
tl
y
serv
ing
as
an
assista
nt
profe
ss
or
a
t
S
al
hi
Ahm
e
d
Naa
ma
Univ
ersi
ty
C
ent
er
in
Al
ger
ia,
her
profe
ss
iona
l
fo
cus
en
com
passes
cha
n
nel
codi
ng
,
digi
tal
signa
l
proc
essing,
opt
im
izati
on
of
t
rel
li
s
-
code
d
mo
dula
ti
on
sche
m
es,
gen
et
i
c
al
gorit
h
ms,
c
o
mm
unicati
o
n
o
ver
multipath
and
fad
ing
channel
s,
orthogon
al
fre
qu
enc
y
-
divi
sion
mul
ti
p
l
exi
ng
(OF
DM
),
cha
nn
el
equ
al
i
ze
rs,
ren
ew
abl
e
ene
rgi
es,
and
A
rab
ic
na
tura
l
la
nguag
e
pro
ces
sing.
For
fur
the
r
communicati
on
,
she
c
an
be
con
ta
c
te
d
at
em
a
il
:
rek
kal
@
cuni
v
-
n
aa
m
a
.
d
z.
Evaluation Warning : The document was created with Spire.PDF for Python.