Service Composition Based on Semantic Vocabulary
YuHui Ning, ShuXia Yu, YuYue Du and Wei Liu
College of Information Science and Engineering, Shandong University of Science and Technology,
579 Qianwangang Road, Huangdao, China
Keywords: Web Service, Service Composition, Semantic Vocabulary, Concept Replacement, Service Clustering.
Abstract: With the development of Web services, semantics is introduced into Web services, and the information
recognition ability can be enhanced with the semantic description of Web services. Thus, the technology of
service composition and binding can be improved. At present, the main method of service composition,
discovery, binding and replacement is similarity computing based on an ontology tree. However,
constructing the ontology tree and deploying the concepts in it are difficult, and the operability of service
composition with similarity computing cannot meet user requirements. To solve the above problem, a new
method of service composition is introduced based on semantic vocabulary. The constructing methods of
the semantic vocabulary, concept replacement, service clustering, and service composition are given.
Finally, the validity and correctness of proposed methods are illustrated by simulations and the comparative
analysis of simulation results.
1 INTRODUCTION
With the development of information technology,
Web services based on XML (Extensible Markup
Language) have developed rapidly (Sun and Jiang,
2008). Web services can be used distributed on a
cross-platform system. At present, the service
providers keep releasing many services into Internet.
Although there are many choices for users, the
efficiency and precision of service discovery are
decreased. In order to solve this problem, the
concepts of service groups (Liu et al. 2007), service
pools (Sheng et al. 2009), service clusters (Deng and
Du, 2013, Gao et al. 2006) and service communities
(Quan et al. 2009, Liu et al. 2009) have been
introduced. Their main idea is that the services with
the similar functions should be clustered before
binding a particular service to a composite service
(Hu and Du, 2013). Clustering services can
minimize the cardinal number for service discovery.
Thus, the service discovery can be sped up.
WSDL (Web services Definition Language) has
some limitations (Deng and Du, 2013). The data of
services described with WSDL cannot be recognized
completely by the computers. Thus, the ability of
service clustering is limited.
With the research of Web services, many
scholars introduce the semantics into Web services
(Xiong et al. 2010). The information recognition
ability can be enhanced with the semantic
description of Web services. Thus, the service
composition and binding would be improved. The
current main method for service composition,
discovery, binding and replacement is similarity
computing based on an ontology tree (Wu et al.
2005). However, constructing an ontology tree and
deploying the concepts in it remain a difficult task.
The operability of service composition based on
similarity computing (HAN et al. 2009, and Xie,
2011) is insufficient.
To solve the above problems, a method for
composing Web services based on semantic
vocabulary is introduced in this paper. The
constructing methods of semantic vocabulary,
concept replacement, service clustering, and service
composition are given. At last, the validity and
advantage of the proposed methods are illustrated by
simulations.
The rest of this article is organized as follows.
The overall design is introduced in Section 2. A
semantic vocabulary is constructed in Section 3. The
method of service clustering is given in Section 4.
The service composition is introduced in Section 5.
Simulations and comparative analysis are given in
Section 6. Concluding remarks are made in Section
7.
223
Ning Y., Yu S., Du Y. and Liu W.
Service Composition Based on Semantic Vocabulary.
DOI: 10.5220/0005426402230228
In Proceedings of the Fourth International Symposium on Business Modeling and Software Design (BMSD 2014), pages 223-228
ISBN: 978-989-758-032-1
Copyright
c
 2014 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

2 OVERALL DESIGN
The overall design of service composition is given in
this section. The advantage of the methods proposed
in this work is introduced.
The procedure to realize service composition is
shown as follows. Firstly, the semantic vocabulary is
constructed via expert judgment (Wu, 2007 and
Wang, 2008). The service concept class set and
instance set are defined in semantic vocabulary. For
example, a service concept class can include {Book
Ticket, Charge, Grade …}, The instance of the
concept “Book Ticket” can include {Boat Ticket,
Air Ticket …}. Second, the descriptive concepts in
Web services are replaced with the class concept
based on the semantic vocabulary. The services are
clustered based on the types of services. Finally,
service composition is oriented to user demands.
The advantages of the methods proposed in this
paper are described below. Firstly, from the
procedure to the realization of service composition,
the semantic vocabulary can be obtained from the
expert judgment. The service concept class and
concept instance can be obtained from semantic
vocabulary. The descriptive concepts of services can
be unified before service composition. The
information recognition ability can be enhanced with
the semantic description of services. Thus, the
service composition and binding can be improved.
Second, the types of services can be obtained from
the expert judgment before service clustering. The
overall design is shown in Figure 1.
Figure 1: The overall design of service composition.
3 CONSTRUCTION OF
SEMANTIC VOCABULARY
The service and service semantic vocabulary are
defined in this section. The constructing algorithm
for semantic vocabulary is given next.
Definition 1. Service=(No, Cons) is a Web
service, where
(1) No denotes the unique label of the service;
and
(2) Cons denotes the descriptive concepts of the
service.
Definition 2. Ctable=(Class, Instance, Relation,
Category) is the semantic vocabulary, where
(1) Class denotes the service concept class;
(2) Instance denotes the service concept instance;
(3) Relation denotes the relations between
concept classes and instances; and
(4) Category denotes the types of services.
Algorithm 1: Construction of semantic
vocabulary.
Input: Service set Tp={Wservice
1
, Wservice
2
, …,
Wservice
n
}.
Output: Semantic vocabulary Ctable=(Class,
Instance, Relation, Category).
Step 1: Create a new semantic vocabulary
Ctable, and
Ctable.Class=Ctable.Instance=Ctable.Relation =∅.
Create a concept variable K=∅. Create a one-
dimension array A[e], and e=+∝.
Step 2: Traverse a service set Tp={Wservice
1
,
Wservice
2
, …, Wservice
n
}.
2.1: Suppose that the current item of Tp is
Wservice
t
, traverse Wservice
t
.Cons.
2.1.1: Suppose that the current item of Cons is
Con
j
, put Con
j
to K.
Step 3: Traverse K.
3.1: Suppose that the current item of K is K
o
, a
result can be obtained via the expert judgment. If K
o
belongs to the concept class, copy K
o
to
Ctable.Class.
Step 4: Copy K to Ctable.Instance.
Step 5: Traverse Instance.
5.1: Suppose the current item is Instance
p
,
traverse Ctable.Class.
5.1.1: Suppose that the current item is Class
q
, a
result can be obtained via the expert judgment. If
Class
q
is the concept class of Instance
p
, put the
relation=<Class
q
, Instance
p
> to Ctable.Relation.
Step 6: Traverse Ctable.Class.
6.1: Suppose that the current item is Class
q
, and
there are already w types of services. A result can
be obtained via the expert judgment. If Class
q
Fourth International Symposium on Business Modeling and Software Design
224

belongs to the rth type of a service, and r<w, then,
put Class
q
to A[r]. Else, put Class
q
to A[w+1].
Step 7: If the number of types of services is z,
then put A[1], A[2], …, A[z] to Ctable.Category.
Step 8: Output semantic vocabulary Ctable.
Algorithm 1 presents a method for constructing
semantic vocabulary. Its workflow is shown in
Figure 2.
Figure 2: Workflows of constructing vocabulary.
4 SERVICE CLUSTERING
The method for service clustering is given in this
section. The algorithm of service clustering is given
as follows.
Algorithm 2: Service clustering
Input: Service set Tp={Wservice
1
, Wservice
2
, …,
Wservice
n
}, semantic vocabulary Ctable=(Class,
Instance, Relation, Category).
Output: Service cluster set Clusters.
Step 1: From Ctable, obtain the number of types
of services i. Create a one-dimension array A[i]={0}.
Create a service cluster set Clusters=∅.
Step 2: Traverse service set Tp.
2.1: Supposing that the current item is
Wservice
j
, traverse Wservice
j
.Cons.
2.1.1: Supposing that the current item is Con
k
,
traverse Ctable.Relation.
2.1.1.1: Suppose that the current item is
Relation
q
=<class, instance>. If instance is Con
k
,
then replace Con
k
with class in Relation
q
.
Step 3: Traverse Ctable.Category.
3.1: Supposing that the current item is
Category
r
, traverse the concepts in Category
r
.
3.1.1: Supposing that the current item is Con
e
,
traverse Tp.
3.1.1.1: Supposing that the current item is
Wservice
j
, traverse Wservice
j
.Cons.
3.1.1.1.1:Suppose that the current item is Con
w
.
If Con
w
==Con
e
, then put Wservice
j
to A[r].
Step 4: Put A[1], A[2], …, A[i] to clusters, and
output clusters.
From Algorithm 2, the descriptive concepts in
services are unified based on semantic vocabulary.
Services are clustered according to the types of
services. The workflow of service clustering is
shown in Figure 3.
Figure 3: Workflows of service clustering.
5 SERVICE COMPOSITION
The method for service composition is introduced in
this section. The user requirements are defined and
the workflow of service composition is described.
Four types of service composition are given. The
first one is the simple composition. It means that the
user requirements can be satisfied by a single
service. The second is sequence composition. It
means that the requirements can be satisfied by
some service executed consecutively. The third is
parallel composition. It means that the requirements
can be satisfied by some services executed in
parallel. The last is mixed composition. It is the type
mixed by sequence and parallel composition.
From the types of composition, the results of the
service composition binding to a user are a set of
services in an order. The user requirements are
introduced as follows.
Definition 3. Ur=(No, Cons, Relation) is a user
requirement, where
(1) No denotes the unique label of a user
requirement.
(2) Cons denotes the descriptive concepts of a
user requirement. The structure of Cons is Cons=(v,
w, …, k), where v, w, and k are descriptive concepts,
and denote the atomic demand.
Service Composition Based on Semantic Vocabulary
225

(3) Relation denotes the relations between
descriptive concepts in a user requirement. The
structure of Relation is Relation=(x&i&…&p)
||(q||e||…||n), where “||” denotes logic “or”, and “&”
denotes logic “and”.
The algorithm of service composition is shown
as follows.
Algorithm 3: Service composition
Input: User requirement Ur=(No, Cons,
Relation). Semantic vocabulary Ctable=(Class,
Instance, Relation, Category). Service cluster set
Acluster={cluster
1
, cluster
2
, …, cluster
i
}.
Output: The result of service composition.
Step 1: Traverse Ur.Cons.
1.1: Suppose that the current item is Cons
j
,
and traverse the item of it.
1.1.1: Suppose that the current item is Con
k
,
and traverse the item of Ctable.Relation.
1.1.1.1: Suppose that the current item is
Relation
q
=<class, instance>. If the instance is Con
k
,
replace Con
k
with class in Relation
q
.
Step 2: Traverse the atomic demands in Ur.Cons.
Suppose that the current item is Cons
i
. Traverse
Ctable.Category. Create one-dimension array
a[k]={0}.
2.1: Suppose that the current item is
Category
r
. Traverse Category
r
.
2.1.1: Suppose that the current item is Con
e
.
Traverse Cons
i
.
2.1.1.1: Suppose that the current item is Con
j
.
If Con
j
==Con
e
, a[r]++.
2.2: Suppose that the maximum value of a[k]
is a[t]. Traverse cluster
t
, and set a[k]={0}.
2.2.1: Suppose that the current item is
Wservice
t
. Traverse Wservice
t
.Cons.
2.2.1.1: Suppose that the current item is Con
w
.
Traverse Cons
i
.
2.2.1.1.1: Suppose that the current item is
Con
j
. If Con
j
==Con
w
, a[t]++.
2.3: If the maximum value of a[k] is a[m],
then replace Cons
i
in Ur.Relation with Wservice
m
.
Step 3: Output Ur.Relation.
The workflow of service composition is shown
in Figure 4.
Figure 4: Workflow of service composition.
6 EXPERIMENTS
To show the advantages of the proposed methods,
the experiment of service composition is given based
on semantic vocabulary in this section.
(1) Construction of services
From Definition 1, 1000 services are defined in
Sheet 1 of Microsoft Excel 2007. A part of services
in Sheet 1 is shown in Figure 5.
Figure 5: A part of services in Sheet 1.
From Figure 5, each line of Sheet 1 denotes a
service. The unique label of a service is defined in
the first column of Sheet 1, and the descriptive
concepts of the service are defined in the second
column. For example, from Figure 5, Service
1
=(No,
Cons), where Service
1
.No=1, and
Service
1
.Cons={Book Ticket, Air Ticket}.
(2) Construction of semantic vocabulary
From Definition 2 and Algorithm 1, the semantic
vocabulary can be obtained, and created in Sheet 2
of Microsoft Excel 2007. A part of semantic
vocabulary in Sheet 2 is shown in Figure 6.
Fourth International Symposium on Business Modeling and Software Design
226

Figure 6: A part of semantic vocabulary in Sheet 2.
From Figure 6, the service concept classes are
given in the first column of Sheet 2, and the service
concept instances are shown in the second column.
The relations between concept classes and instances
are given in the third column, and the types of
services are described in the fourth column. For
example, from Figure 6, the first type of service is
presented by {Sells Ticket, Air Ticket}.
(3) Concept replacement
From Algorithm 2, the descriptive concepts of
the services defined in Sheet 1 can be unified
according to the service concept classes defined in
Sheet 2. To improve the efficiency, the procedure
for concept replacement can be programmed by
Excel VBA (Visual Basic for Application) (Hu,
2014). It is convenient to extract and edit the data in
Excel Sheets by VBA. The workflow of a VBA
program for realizing service clustering is shown in
Figure 7.
Figure 7: The workflow of a VBA program.
Sheet 1 has been changed after concept
replacement. It is shown in Figure 8.
.
Figure 8: Sheet 1 of Excel after concept replacement.
From Figures 7 and 8, the concept “Book Ticket”
in service
1
is replaced by “Sells Ticket”, service
4
is
also changed, and so on.
(4) Service clustering
From Algorithm 2, fifty service clusters are
obtained by using VBA programs, and created in
Sheet 3. A part of service clusters in Sheet 3 are
shown in Figure 9.
Figure 9: A part of service clusters in Sheet 3.
In Figure 9, the names of service clusters are
given in the first column of Sheet 3, and the services
in clusters are given in the second column. For
example, from Figure 9, cluster
1
={service1,
service
2
, service
58
,…}.
(5) Service composition oriented to user
requirements
From Definition 3, fifty user demands are
defined in Sheet 4. A part of user demands are
shown in Figure 10.
Figure 10: A part of user demands in Sheet 4.
From Algorithm 3, a service composition can be
obtained, and created in Sheet 5. A part of service
composition is shown in Figure 11.
Figure 11: A part of service composition.
(6) Comparison
From Figure 11, the user demands can be
satisfied by the service composition based on the
methods proposed in this paper. The result of service
Service Composition Based on Semantic Vocabulary
227

clustering without semantic concept replacement is
shown in Figure 12.
Figure 12: Clusters without semantic concept replacement.
From Figures 11 and 12, obviously, the service
clustering ability in Figure 12 is decreased. For
example, the descriptive concepts “Book Ticket” in
service
1
are not a concept class. The computer
cannot recognize them from the service clustering.
Thus, service
1
does not appear in cluster
1
of Figure
12. Moreover, the service composition ability can be
decreased.
Therefore, the validity and advantages of the
proposed method are illustrated.
7 CONCLUSIONS
To improve service composition, a new method of
service composition is proposed based on semantic
vocabulary in this paper. A semantic vocabulary is
constructed and some algorithms are presented such
as service concept replacement, service clustering,
and service composition. Finally, the validity and
advantages of the proposed methods are illustrated
by simulations and comparative analysis.
Further work will be the data processing
technology of the services oriented to big data.
Moreover, the platform construction of service
composition will be considered based on cloud
computing.
ACKNOWLEDGEMENTS
This work is supported by the National Basic
Research Program of China under grant
2010CB328101; the National Natural Science
Foundation of China under grants 61170078 and
61173042; the Doctoral Program of Higher
Education of the Specialized Research Fund of
China under grant 20113718110004; Basic Research
Program of Qingdao City of China under grant No.
13-1-4-116-jch; and the SDUST Research Fund of
China under grant 2011KYTD102.
REFERENCES
Deng, S.Y., and Du, Y.Y., 2013. Web service composition
approach based on service cluster and Qos. Journal of
Computer Applications. 33(8), 2167-2170.
Deng, S.Y., and Du, Y.Y., 2013. Web service composition
approach based on service cluster and Qos. Journal of
Computer Applications. 33(8), 2167-2170.
Gao, Y., Na, J., Zhang, B., Yang, L., 2006. 3-layer Web
services organization model for dynamic service
composition. Mini-Micro Systems, 27(10), 1879-1882.
Hu, Q., Du, Y.Y., 2013. Service architecture and service
discovery oriented to service clusters. Journal of
Computer Applications. 33(8): 2163-2166.
Hu, X., 2014. Automatic transmission of delivery
information based on GPRS. Journal of Neijiang
Normal University, 18(9): 107-109.
HAN, S., WANG, H.Y., CUI, L.Z., 2009. A user
experience oriented service discovery method with
clustering technology [C]// Second Int. Symposium on
Computational Intelligence and Design, China, Chang
Sha: IEEE Computer Society, 64-67.
Liu, S.L., Liu, Y.X., Zhang, F., Tang, G.F., 2007. A
dynamic Web services selection algorithm with Qos
global optimal in Web services composition. Journal
of Software, 18(3), 646-656.
Liu, X.Z., Huang, G. and Mei, H., 2009. Discovering
homogeneous Web service community in the user-
centric Web environment. IEEE Transactions on
Service Computing, 2(2), 167-181.
Quan, Z.S., Boualem, B., Zakaria, M., Anne, H.H.N.,
2009. Configurable composition and adaptive
provisioning of Web services. IEEE Transactions on
Services Computing, 2(1), 34-49.
Sun, P., Jiang, C.J., 2008. Using service clustering to
facilitate process-oriented semantic Web service
discovery, Chinese Journal of Computers, 31(8),
1340-1353.
Sheng, Q.Z, Benataliah, B., Maamar, Z., 2009.
Configurable composition and adaptive provisioning
of Web services. IEEE Transactions on Service
Computing, 2(1), 34-49.
Wu, J., Wu, Z.H., Li, Y., Deng, S.G., 2005. Web service
discovery based on ontology and similarity of words.
Chinese Journal of Computers, 28(4), 595-602.
Wu, J.Y., 2007. Requirements and configuration
management of project. Microcomputer & its
Applications, S1(2):112-114.
Wang, X.P., Wang, F.J., Wang, W.W., 2008. The
certification research of enterprise quality
management system. Management Review, 10(8): 131-
134.
Xiong, P C, Fan, Y., Zhou, M.C., 2010. A Petri net
approach to analysis and composition of web services.
IEEE Transactions on Systems, Manand Cybernetics,
Part A, 40(2): 376–387
Xie, L.L., 2011. Ontology-based semantic web services
cluster and discovery. Tianjin University, China.
Fourth International Symposium on Business Modeling and Software Design
228