ONTOLOGY-BASED DYNAMIC SERVICE COMPOSITION

USING SEMANTIC RELATEDNESS AND CATEGORIZATION

TECHNIQUES

Yacine Rezgui and Samia Nefti

Informatics Research Institute, University of Salford, M5 4WT Salford, UK

Keywords: Web services, Service composition, Ontology, Semantic Relatedness, Categorization.

Abstract: Organizations need to migrate their legacy systems to higher order applications capable of engaging in

automated modes of collaboration to support distributed business processes. This requires a change of focus

from intra-enterprise system integration through agreed data structures to inter-enterprise business process

integration through smart composition of web-serviced applications. The paper presents an approach aiming

at supporting ontology-based semantic composition of web-services to support distributed electronic busi-

ness processes. This new generation of composite services is semantically coordinated in a secure, scalable,

and resource-aware environment. Two services, at the heart of the service composition exercise are fea-

tured, namely: the semantic compatibility and categorization services.

1 INTRODUCTION

Service-oriented computing is becoming the promi-

nent paradigm for leveraging inter and intra enter-

prise information systems, creating opportunities for

smart organizations to provide value added services

and products. The benefits of web services include

the decoupling of service interfaces from implemen-

tation and platform considerations, the support for

dynamic service binding, and an increase in cross-

language and cross-platform interoperability (Ferris

and Farrell, 2003). This new form of computing

should move from its initial “Describe, Publish,

Interact” capability to support dynamic composition

of services into reinvented assemblies, in ways that

previously could not be predicted in advance (Heu-

vel and Maamar, 2003; Rezgui, 2007a).

However, composed web service applications are

not adaptive to change. If the requirements of the

application change or need extending, the service

composition needs to be re-specified from scratch. It

is currently not possible to define and implement a

web service composition once and use it in similar

designs with some variations in a later stage. A more

flexible approach allowing service re-use, extension,

specialization should be supported. Also, web ser-

vice composition methodologies have a focus on

syntactic integration and therefore do not support

automatic composition of web services. As high-

lighted in (Sycara et al., 2003), semantic integration

becomes crucial for web services as it allows them

to (a) represent and reason about the task that a web

service performs, (b) explicitly express and reason

about business relations and rules, (c) understand the

meaning of exchanged messages, (d) represent and

reason about preconditions that are required to use

the service and the effects of having invoked the

service, and (e) allow intelligent composition of web

services to achieve a more complex service.

The paper describes a web services infrastruc-

ture, developed within the EU funded e-Cognos

project (Rezgui and Meziane, 2005) and refined in

the follow-up EU funded FUNSIEC project (Barresi

et al., 2005), aimed at facilitating and supporting the

execution of distributed business processes imple-

mented through a coordinated composition and

invocation of web-enabled, service-oriented, Enter-

prise Information Systems (EIS). This service infra-

structure has been specified and implemented for the

Construction sector, but with a view to be general-

ized and used across industry. First, the paper pro-

vides a conceptualization of the proposed service

infrastructure. The latter establishes the middleware

foundation as well as the necessary components,

including services, enabling the support of distrib-

113

Rezgui Y. and Nefti S. (2007).

ONTOLOGY-BASED DYNAMIC SERVICE COMPOSITION USING SEMANTIC RELATEDNESS AND CATEGORIZATION TECHNIQUES.

In Proceedings of the Ninth International Conference on Enterprise Information Systems - SAIC, pages 113-120

DOI: 10.5220/0002369301130120

 SciTePress

uted business processes. The paper then describes

the different services developed to support intelli-

gent service composition. Two of these services are

then illustrated, namely: the semantic compatibility

and categorization services. Finally the paper dis-

cusses the limitations of the research as well as fu-

ture work.

2 SUPPORTING DISTRIBUTED

E-PROCESSES

E-processes are typically designed, developed, and

deployed by enterprises that want to compose inter-

nal capabilities with third-party capabilities, either

for internal use or to expose them as (complex,

value-added) e-services to customers. They can be

described as the smart aggregation of services that

are captured in the form of a Composite Service

complying with well-defined business rules. They

have the ability to be reused outside their scope and

become generic services. Composite services are

executed within shared workspace environments. A

shared workspace refers to an online web environ-

ment involving authorized actors, united for a busi-

ness or practice purpose. Access to data / informa-

tion / knowledge from within a shared workspace is

organized via dedicated elementary or composite

services, invoked by actors through their assigned

role(s). Shared workspaces can be defined at differ-

ent levels of granularity; these can range from sup-

porting collaboration within a complex construction

project, to nurturing a small community of practice.

Implementing service-oriented shared workspace

environments involves three key generic “roles”

defined at the service infrastructure level:

Work Space Service Provider (WSSP):

This has

the responsibility of managing the entire service

infrastructure (e.g. servers, computer resources,

services, etc…) and allocating shared workspace

environments to potential construction clients to

host their projects. This involves hosting the core

infrastructure through provision of and access to

both core services and Third Party Services (TPS).

Core services refer to services necessary for the

basic operation and management of services, includ-

ing TPS that have the particularity of being provided

by third party entities, namely, third party service

providers. WSSPs, through the core services, have

the capability to host multiple projects and to make

available different services (both core and TPS) to

workspaces. WSSPs can also play the role of appli-

cation integrators, providing help and assistance to

migrate legacy applications and legacy systems to

Web services.

Third Party Service Providers (TPSP): These

represent various companies, including software

houses, interested in making their software applica-

tion(s) accessible through a service-based middle-

ware solution hosted and managed by the WSSP.

These companies have their services published

within the WSSP UDDI registry. Typically, these

services would be geared to serving a particular

purpose for the Virtual Enterprise (VE) to which

they are being made available. Examples of services

offered by TPSP include structural dimensioning

service, HVAC simulation service, procurement

service, and facility management service.

Figure 1: Conceptualization of Service Composition.

These services are offered by organizations that

procure the service implementation, supply the de-

scription of the service, and provide related techni-

cal and business support. These can also represent

service aggregators that consolidate multiple ser-

vices into a new, single service offering.

Work Space Clients (WSC):

The shared work-

space will involve stakeholders representing con-

struction companies collaborating within the context

of a project. This collaboration is supported and

enabled through the WSSP platform. While one

company, acting in the capacity of Workspace man-

ager, would configure and administer the Work-

space, others would make use of the core and TPS

services made available to the project.

The proposed conceptualization of service com-

position (illustrated in Figure 1) is based on BPEL.

Business processes interact, on a peer-to-peer basis,

Method

Service

1..*

belongs to

CompositeService

Workspace

WorkspaceRole

0..*

1..*

is applicable in

0..1

can be broken down into

0..1

comprises

AssignedServiceMethod

refers to

is assigned to1..*

is created within

is coordinated

0..*

0..1

1..*

comprises

0..*

BusinessProcess

1..*

0..*

ServiceInteraction

0..*

1..*

Activity

0..*

involves

is involved in

generates

0..1

Class

applies to

1..*

implements

Project

1..*

hosts

supports

ICEIS 2007 - International Conference on Enterprise Information Systems

114

with a set of services by invoking one or several of

the methods they support. The way messages are

exchanged between the business process and the

Service Methods is described through the concept of

Activity. Activities can then be combined into com-

plex algorithms through the BPEL concept of Struc-

tured Activities. The proposed model supports the

allocation of such services to Shared Workspace

Environments. Methods represent API calls, or func-

tionality, of such services.

3 ARCHITECTURE AND CORE

SERVICE DESCRIPTION

In order to support e-Processes a layered service

architecture that makes use of established work,

initiatives and standards in the web services domain

(including BPEL, WS-Security, WS-Coordination

and Transaction) has been specified. Each layer

represents the main building blocks enabling the

project workspace through the three identified roles,

namely Work Space Service Provider (WSSP),

Third-Party Service Providers (TPSP) and Work

Space clients (WSC).

Referring to Figure 2, boxes within the WSSP

services represent the essential core services neces-

sary for the setting-up, operation and coordination of

a project workspace. The service manager box pro-

vides access to the API functions necessary for all

aspects of invocation, registration and de-

registration of services from third-party service

providers, as well as their publication in the local

(WSSP maintained) UDDI registry. The Business

Process Specification Layer (BPSL) includes the

API functions that enable service composition in

order to implement a given business process. This is

based on the following core services concerned with

service coordination, transaction, and security:

• Security Service: This service builds and

implements the WS-Security specification.

WS-Security defines the core facilities for

protecting the integrity and confidentiality

of SOAP messages, and is specified in a

way that accommodates a wide range of se-

curity models (including identity-based se-

curity, access control lists, and capabilities-

based security) and encryption technolo-

gies.

• Coordination Service: This service builds

on WS-Coordination, which defines an ex-

tensible framework for coordinating activi-

ties using a coordinator and set of coordina-

tion protocols. This enables participants to

reach consistent agreement on the outcome

of distributed activities. The coordination

protocols that can be defined should ac-

commodate a wide variety of activities, in-

cluding protocols for simple short-lived op-

erations and protocols for complex long-

lived business activities. It provides consis-

tent control of the execution of the services

forming the composite service.

• Transaction Service: It is based on WS-

Transaction, which leverages WS-

Coordination by (a) extending the WS-

Coordination context to create a transaction

context, (b) augmenting the activation and

registration services with a number of addi-

tional services (Completion, Completion

WithAck, PhaseZero, 2PC, Outcome Noti-

fication, BusinessAgreement, and Busines-

sAgreementWithComplete), and (c) defin-

ing two particular coordination types:

Atomic Transactions (AT) and Business

Activity (BA).

Figure 2: Proposed Service-oriented Architecture.

The Third Party Service Provider Layer (TPSP)

represents all web-serviced applications that are

ready for invocation and use as part of a service

composition exercise in order to implement a busi-

ness process. As explained in section 2, any existing

EIS or legacy application has the potential to be

promoted to become a web service. The members of

a workspace (WSC) can use adapted software cli-

ents made available by the service provider (WSSP)

to collaborate and invoke services. They have also

the possibility to extend their existing portal (or EIS

Service Manager

API

RMI/SOAPRMI/SOAP

SOAPSOAP

RMI/SOAP RMI/SOAPSOAP SOAP

INTERNET

Service Discovery

Service Publication

Service Inspection

Security Communication

Coordination

Transaction

Business Process Specification Layer

Native APIs

Legacy

System 2

WSDL

SOAP SOAP

WSDL

Native APIs

Semantic

Representation

Service

Intelligent

Resource

Discovery

Service

Categorisation

Service

Multi-Modal

Interfaces

Web-Serviced

Enterprise

Information

Systems

Service

Infrastructure

Management/

Service

Composition

Semantic

Integration

Secure Service

Transactions

Coordination

Service Invocation

Semantic

Compatibility

Service

Ontology

Service

WSDL

Legacy

System 1

SOAP

Legacy

System 3

Workspace

Management

Service

INTERNET

WSDL

Legacy

System 1

SOAP

ONTOLOGY-BASED DYNAMIC SERVICE COMPOSITION USING SEMANTIC RELATEDNESS AND

CATEGORIZATION TECHNIQUES

115

application) by implementing relevant API functions

and/or providing transparent access to relevant core

services.

A number of dedicated services have been speci-

fied and developed to facilitate and support service

composition. The issue of web service composition

involves three fundamental aspects: (a) the identifi-

cation of the required functionality to implement the

desired business process; (b) the discovery of the

web services that perform the identified functional-

ity; and (c) the management of the interaction be-

tween those web services.

The service composition layer includes neces-

sary functionality for the consolidation of multiple

services into a single composite service and specify-

ing its business process behavior through the defini-

tion of activities and their protocols, including mes-

sage exchanges that take place between a process

and all its partners (represented through WSDL

descriptions of their web services). Activities are

used to describe ways in which messages are in-

voked on, and acted upon by, each process partner,

and can be combined into complex algorithms using

the BPEL structured activity definitions. These in-

clude the ability to define a sequence (ordered suc-

cession of steps), to define a loop (while), to have a

conditional branching (while), or to specify the

parallel execution of several steps (flow) as concep-

tualised in Figure 1. The services developed to sup-

port semantic composition include:

• Ontology Service: this provides the func-

tionality required to make the selected on-

tology available to the other services, which

may require it. This is achieved via the e-

Cognos Ontology Server (e-COSER) that is

devoted to handling all the ontology-related

issues within the system (Lima et al., 2005).

• Semantic Representation Service: This se-

mantic representation service makes use of

the XML DOM API (Miniaoui et al., 2005),

which transforms XML documents into

trees. For compatibility detection between

two sevices, the XML string data structure

is more efficient than the tree structure be-

cause it preserves all paths within an XML

document and transforms the XML mining

problem to a string mining. The string map-

ping is performed in two stages. The First

stage consists in encoding all the edges (tag

names) of the XML DOM tree in a digital

format. Then, the pre-order pass of the tree

allows encoding the tree into a string. Zaki

(2002) proposes the encoding of the XML

tree into string. His algorithm inserts a char-

acter (-1) in the string to indicate the

movement in the tree. This latter will be

used in our semantic compatibility.

• Semantic compatibility service: this service

is used to inspect service definitions at run-

time and determine the level of compatibil-

ity between services to engage into a com-

position mode. This makes use of the eCog-

nos Construction ontology (Lima et al.,

2005) that takes a pivotal role in the process

by checking semantic relatedness between

concepts used by each partner service.

• Categorisation Service: This service is pri-

marily used to provide a context and crite-

ria-based categorisation of the information

held within the UDDI registry for effective

use and mining by potential business part-

ners. As things stand, this can be queried

using the native UDDI API. However, this

presents serious limitations and inefficien-

cies in that it relies on data and information

defined by business and service providers

and do not model the vagueness and the un-

certainty of stored information. Further-

more, our approach handles uncertainty in

the reasoning using fuzzy concept.

Figure 3: Intelligent resolution of semantic Compatibility

for web services.

• Intelligent Resource Discovery Service:

this service provides a standardised inter-

face to query and make use of the Intelli-

gent Categorised UDDI Registry generated

Web Serviced

Application

Method A2

Method A1

Method A3

Method A4

Method A5

Method A6

Web Serviced

Application

Method B1

Method B2

Method B3

Method B4

Method B5

WSDL:

XML-Based

Representation

WSDL:

XML-Based

Representation

API DOM

Semantic Representation Service

XML String

Structure

Semantic Compatibility

Service

Ontology

Service

Construction

Ontology

UDDI

Registery

Categorisation

Service

Intelligent

Categorised

UDDI

Registry

Find

Closest

Compatible

Service

Intelligent

Resource

Discovery

Service

(1)

(2)

(3)

(4)

(5)

ICEIS 2007 - International Conference on Enterprise Information Systems

116

through the categorisation service described

above.

The above-mentioned services are illustrated in

Figure 3 detailing the process of composing an e-

Process from elementary semantically compatible

services.

4 SEMANTIC COMPATIBILITY

SERVICE

This service is used to inspect service definitions at

run-time and determine the level of compatibility

between services to engage into a composition

mode. This makes use of the strings generated by

the semantic representation service and the eCognos

Construction ontology (Lima et al., 2005) that takes

a pivotal role in the process by checking semantic

relatedness between concepts used by each partner

service.

Moreover, services involved in inter-working

have to understand each other’s when they are sup-

posed to perform meaningful cooperative actions.

Service definitions are described in the XML-based

WSDL format.

Different mining techniques have been proposed

in the literature (Laurent et al., 2005). The main

difference between our semantic compatibility ap-

proach and the existing ones is that our method

handles uncertainty in the reasoning using fuzzy

concept (Miniaoui et al., 2005). Given two XML

string structures, the main idea used in our approach

is to measure the degree of overlap between two

strings by introducing the inclusion concept. This

means that an XML string is included within another

with a certain degree. A degree of inclusion is de-

fined based on notion of relationship between con-

cepts using ontology. This inclusion is measured by

using the semantic relatedness between labels (tags

name in the strings). Thus, the vector representation

of an XML structure provides sufficient statistics for

string mining and for computing the level of com-

patibility or expected overlap and inclusion between

XML strings.

Let us now consider that we are provided with

knowledge on the data (strings) given by the seman-

tic representation service. The idea is to find the

degree of compatibility between those two strings.

The difference between our approach and the exist-

ing mining approaches is that in our reasoning the

inclusion detection is not only performed on same

tag names. We also consider the fact that two differ-

ent labels could be semantically related even if they

don’t have the same name. This relatedness can be

detected by using semantic relatedness measure

based on the Construction ontology (Lima et al.,

2005), which allows us then to measure the inclu-

sion between strings structure. The semantic relat-

edness used is Hirst–St-Onge (Hirst and StOnge,

1998) measure. The idea behind this measure is that

two lexicalized concepts are semantically close if

their synsets are connected by a path that is not too

long and that “does not change direction too often”.

The strength of the relationship is given by:

kdlengthpathCccrel

−

),(

; where d is

the number of changes of direction in the path using

the ontology, and C and k are constants; if no such

path exists,

),(

ccrel

is zero and the labels are

deemed unrelated.

For instance, let’s take three cases where two strings

S and X could be included within another with dif-

ferent degrees.

Case 1:

String S: ABCDEF-1G-1H

String X: ABYDEF-1G-1H

The string X is similar to S if

1),( =YCrel

The string X is highly included in S if Semantic re-

latedness is low

Case 2:

String S: AB-1C-1D-1KLMNY

String X: AC-1B-1D

The string X is fully included in S if the

),( BCrel

is height (B is synonym of C)

The string X is partially included (with a certain

degree) in

0),(

BCrel

Case 3:

String S: ABC-1D-1F

String X: AE-1D

The string X is fully included in S if the

),( CBrel

is high and is semantically related to E.

In the string, each label will be represented by

fuzzy membership degree describing the ancestor-

descendant relationship measured by the semantic

relatedness value between given labels.

5 CATEGORISATION SERVICE

The Categorisation service provides a context and

multi-criteria based clustering of the information

held within the UDDI registry for effective exploita-

tion by potential business partners. As argued

ONTOLOGY-BASED DYNAMIC SERVICE COMPOSITION USING SEMANTIC RELATEDNESS AND

CATEGORIZATION TECHNIQUES

117

above, the native UDDI API presents limitations due

to the reliance on service providers for the descrip-

tion of their services. This does not model the uncer-

tainty and the vagueness of stored information.

Categorisation techniques employed to date, includ-

ing hard clustering methods, have shown some limi-

tations. These can be overcome with fuzzy cluster-

ing technique where service descriptions are attrib-

uted to several clusters simultaneously and thus,

useful relationships between businesses and service

categories may be uncovered, which would other-

wise be neglected by hard clustering methods. Fur-

thermore, services may belong to several clusters at

the same time. This is in line with the Fuzzy c-

means clustering technique as it adopts a non-binary

approach by assessing the membership of a service

to a cluster. Each service is assigned a vector of

membership degree over all categories or disci-

plines. Hereafter, a service is assimilated to a text

document.

To be able to cluster these text documents one needs

to map them to numerical feature vectors. After

applying the traditional pre-processing methods

(stopwords elimination and stemming using the

Porter Stemmer (Porter, 1980), each document is

represented by a vector model. Under this represen-

tation model, documents are mapped into vectors

usually normalized in a high dimension concept

space. The axes of such a concept space are different

dictionary terms and its coordinates are term weight-

ing values corresponding to dictionary terms.

In its simplest form, each weighting term correspond

to the term-frequency (TF) value (Zhao and Karypis,

2002). The resulting document vector will be de-

fined by :

()

mtf

tftftfd ...,..........,.........,

= ;

where tfi is the frequency of the i th term in the

document.

Thus, this vector representation of documents

provides sufficient statistics for computing the dis-

tribution of terms within each document. Than, each

distribution is approximated by a Gaussian. Thus,

this new vector representation will allow us to find

expected overlap between two vectors in terms of

corresponding distributions.

In order to perform the fuzzy clustering of these

distributions, one needs to find a new distance

measure which can measure effectively the distance

between two distributions.

In this service we adapted our Modified algorithm

proposed in (Nefti and Oussalah, 2004) to measure

the distance between the distributions.

An appealing example consists of clustering a set of

Gaussian distributions

)(xG

, defined by their

mean

and covariance matrix

Σ ( ni ..1= , n is

the number of distributions)

[]

))()(

exp()(

xxxG

μμ

−Σ−−=

−

(1)

However, the distance structure in the product space

of Gaussian parameters need not correspond one to

one to the inclusion structure between Gaussians,

which has influence on the location of the cluster

prototypes. Therefore, we constrained the fuzzy

clustering algorithm additionally, in order to detect

the inclusion of document in the cluster prototype.

In (Nefti and Oussalah, 2004), we refocused on

standard FCM algorithm when dealing with prob-

abilistic distance structure which incorporates a

component of variance-covariance document (see

for more details Nefti and Oussalah, 2004),. We

showed that the use of such distance leads to optimal

solutions in terms of prototype canters and member-

ship function matrix U in a reasonable computa-

tional time.

The proposed modified FCM algorithm for Gaus-

sians is based on the probabilistic distance measure.

Consider a set of distributions, represented by Gaus-

sians in the form of (1). This set may be interpreted

as a set of points

),(

, ni ..1= , in m-

dimensional functional space with metrics, defined

by (2),

is a column mean vector

()

and

is a diagonal covariance matrix

()

Let

designate the BHATTACHARYYA dis-

tance from datum to cluster center

)*,*(

)*(

jiij

μμ

ΣΣ

μμΔ

−

⎥

⎦

⎤

⎢

⎣

⎡

−=

−1

[]

)*(

ΣΣ

(2)

Then, by analogy with standard FCM method, the

proposed algorithm optimises objective function J,

which may be expressed in the following form

⎟

⎠

⎞

⎜

⎝

⎛

−+Δ=

∑∑∑∑

====

1111

ijij

uuJ

(3)

ICEIS 2007 - International Conference on Enterprise Information Systems

118

where

are Lagrangian multipliers. The necessary

conditions for optimality of the objective function J

may be found by setting its partial derivatives to

zero; namely:

=+Δ=

∂

−

iijij

λα

, (4)

=−=

∂

∑

, (5)

()

=−

⎥

⎦

⎤

⎢

⎣

⎡

−=

∂

−

∑

μμ

ΣΣ

,(6)

[]

()

[

]

,Ee**u

jiji

⎭

⎬

⎫

⎩

⎨

⎧

⎟

⎠

⎞

⎜

⎝

⎛

−+−=

∂

∑∑

−

μμΣΣ

(7)

where

e and

E are column vectors

()

matrices with the only non-zero components

{

}

and

{}

E , respectively. Equations (4) and (5) may

be used to derive

∑

−

⎟

⎠

⎞

⎜

⎝

⎛

(8)

From (9) we may infer, that

The detail of this calculus, based on the derivative of

(8-9) with respect to its parameters and the proof of

the existence of uniqueness of the solution is omit-

ted here (Nefti and Oussalah, 2004).

Table 1: Proposed modified fuzzy clustering algorithm.

Step

Fix c,

nc <≤2 ; fix

∞

<≤

1 ;

fix

; initialise

)0(

U by using standard

FCM algorithm

Step

Calculate the c fuzzy cluster prototypes

),(

jjj

v Σ

solving (7), (9) numerically

Step

Calculate

)1(

U using (11) and

),(

jjj

obtained at Step 2

Step

≤−

)0()1(

UU STOP,

otherwise

)1()0(

: UU = and return to Step 2

Finally, the proposed modified fuzzy clustering

algorithm can be given as described in Table 1.

6 CONCLUSIONS

The paper argues that the ability to integrate dispa-

rate, heterogeneous Enterprise Information Systems

to implement a distributed business process is facili-

tated by the loosely coupled nature of Web services.

These provide the ability to combine and aggregate

services into sophisticated, higher order, composite

services. Moreover, the proposed service-oriented

computing paradigm migrates the traditional stand-

alone-hosting model to a networked one, by allow-

ing web services to dynamically discover and hook-

up to web services offered by different providers.

The approach promotes the creation of domain-

specific vertical libraries of services that are modu-

lar, well documented, implementation-independent,

and interoperable.

The paper proposes an approach aiming at sup-

porting ontology-based semantic composition of

web-services to support electronic business proc-

esses. This has been implemented to provide the

middleware foundation as well as the necessary

components, including services, enabling the sup-

port of distributed business processes. Two services,

at the heart of the service composition exercise, have

been featured, namely: the semantic compatibility

and categorisation services.

However, while the testing and validation work

(not included in the paper due to length limitations),

was highly encouraging, the approach requires large

scale trials in order to better apprehend the business,

technical as well as process implications. Also,

given the emerging nature of the web service model,

web compliant services are yet to be developed to

form a critical mass of services that would encour-

age businesses to migrate to this new mode of col-

laboration (Rezgui, 2007b). From a technical point

of a view, the algorithms presented in the paper need

adapting to accommodate very high dimensional

data with acceptable processing times. These issues

are currently under investigation and will be re-

ported in future publications.

[] []

)9(

−

⎟

⎠

⎞

⎜

⎝

⎛

−

Σ+Σ

⎟

⎠

⎞

⎜

⎝

⎛

−

Σ+Σ=

∑∑

jiij

ONTOLOGY-BASED DYNAMIC SERVICE COMPOSITION USING SEMANTIC RELATEDNESS AND

CATEGORIZATION TECHNIQUES

119

ACKNOWLEDGEMENTS

The authors would like to acknowledge the support

of the European Commission under the eContent

FUNSIEC project

REFERENCES

Barresi, S., Rezgui, Y., Lima C., Meziane, F., 2005. Ar-

chitecture to Support Semantic Resources Interopera-

bility, ACM 14th Conference on Information and

Knowedge Management (CIKM), Bremen.

Ferris, C and Farrell, J., 2003. What are web services,

Communications of the ACM, Vol.46, No6.

Hirst, G and St-Onge, D., 1998. Lexical chains as repre-

sentations of context for the detection and correction

of malapropisms, In Fellbaum, pp. 305–332.

Laurent, A., Poncelet, P. and Teisseire, M., 2005. Towards

a Fuzzy Approach for Mining XML Mapping Sche-

mas. Fuzzy Logic and the Semantic Web Workshop,

Marseille.

Lima, C., El-Diraby, T. and Stephens, J., 2005. Ontology-

based optimisation of knowledge management in e-

Construction, ITcon Vol. 10, pg. 305-327.

Miniaoui, S., Wentland, and Forte, S., 2005. XML mining:

from trees to strings, in proc. of ICICIS 2005, Cairo,

Egypt March.

Nefti, S., Oussalah, M., 2004. Probabilistic-Fuzzy Cluster-

ing Algorithm, IEEE International Conference on

Systems, Man and Cybernetics, The Netherlands, Oc-

tober.

Porter, M., 1980. An algorithm for suffix stripping»,

Program, Vol. 14, no. 3, p.130-137.

Rezgui, Y. and Meziane, F., 2005. A Web Services Im-

plementation of a User Centred Knowledge Manage-

ment Platform, International Journal of Intelligent In-

formation Technologies, Vol.1, No 4.

Rezgui, Y., 2007a. Exploring Virtual Teamworking Effec-

tiveness in the Construction Sector, Interacting with

Computers, Vol. 19, no 1, pp. 96-112.

Rezgui, Y., 2007b. Role-based Service-Oriented imple-

mentation of a Virtual Enterprise: A Case Study in the

Construction Sector, Computers in Industry, Vol. 58,

no 1, pp. 74-86.

Sycara, K., Paolucci, M., Ankolekar, A. and Srinivasan,

N., 2003. Automated Discovery, interaction and com-

position of web services, journal of web semantics, 1

,27-46.

Van Den Heuvel, W.J. and Maamar, Z., 2003. Moving

Towards a Framework to Compose Intelligent Web

Services, Communications of the ACM, Vol.46,

No10.

Zaki. J., 2002. Efficiently Mining Frequent Trees in a

Forest, 8th ACM SIGKDD International Conference

on Knowledge Discovery and Data Mining.

Zhao, Y. and Karypis, G:, 2002. Criterion Functions for

Document Clustering: Experiments and Analysis,

http://www.cs.umn.edu/˜karypis

ICEIS 2007 - International Conference on Enterprise Information Systems

120