Semantic Matching to Achieve Software Component Discovery

and Composition

Soﬁen Khemakhem

, Khalil Drira

2, 3

and Mohamed Jmaiel

University of Sfax, National School of Engineers, Laboratory ReDCAD, P.B.W. 3038 Sfax, Tunisia

CNRS, LAAS, 7 Avenue du Colonel Roche, F-31400 Toulouse, France

Universit

e de Toulouse, LAAS, F-31400 Toulouse, France

Keywords:

Ontology, Software Component, Discovery, Composition, Integration.

Abstract:

In CBSE, current approaches based on software component are in short of ﬂexibility because of lacking se-

mantic. In this paper, we introduce an extended semantic discovery of software components in which non

functional properties of components are considered. We also introduce an ontology supported automatic com-

ponent composition method to realize the integration of the adequate component composite in the current

project.

1 INTRODUCTION

Since the middle of 1990’s, object-oriented program-

ming becomes a key component in software engineer-

ing. Many reserach in software development focus

on software reuse, (Khemakhem et al., 2011) while

the software composition becomes an important com-

ponent in software engineering. To resolve the dif-

ﬁculty of manually building a composite component,

there have been many studies on the automated com-

position of software components (Seog-Chan et al.,

2006). Previous works specify the semantic descrip-

tion of a component using service speciﬁcation lan-

guages such as OWL-S (Chaari et al., 2013), WSMO

(de Bruijn et al., 2005), and SAWSDL (Akkiraju

et al., 2007) based on a domain ontology, which is

built by an ontological language such as OWL (Smith

et al., 2004). Likewise, a developer’s request is speci-

ﬁed in terms of inputs, outputs, preconditions and ef-

fects. From these speciﬁcations, they automatically

build composite components using various AI (Ar-

tiﬁcial Intelligence) techniques. Previous methods

construct composite components, which take the in-

puts entered by a developer and return the outputs re-

quested, by repeatedly discovering and chaining ap-

propriate components. When two components are

chained, the preceding component should satisfy the

precondition of the following component. Generally,

two components with identical inputs, outputs, pre-

conditions, and effects are regard as identical com-

ponents. Therefore, if a composite software compo-

nent satisﬁes the inputs, outputs, preconditions, and

effects requested by a developer, it is regarded as sat-

isfying the user requirement. However, when compo-

nents or user requirements do not have pre-conditions

and effects, the conventional methods may generate

composite components, which differ from a devel-

oper’s needs. In general, two components with the

same IOPE (input, output, pre-condition and effect)

are considered to be identical each other (Chaari and

Fakhfakh, 2012). However, non-functional properties

(Junghans and Agarwal, 2010), which provide certain

information about component constraint to users, may

not need any preconditions and have no effects af-

ter execution. In previous works query is based on

IOPE, cannot compose correctly if the non-functional

aspect of the disred composite component is not spec-

iﬁed. In several cases, the non-functional constraints

play a decisive role in the choice of the most power-

ful composite component. previous works have ex-

ponential time complexity in proportion to the num-

ber of available components, since they have to con-

sider every possible combination of available compo-

nents. In this paper we aim to deﬁne a uniﬁed and

a complete approach to ameliorate the reuse of soft-

ware components in the CBD. Our approach encom-

passes the functional and the non-functional aspect in

different stage: description, discovering, composition

and integration. To achieve our objective we develop

three ontologies. Two ontologies are used to discover

atomic and composite components which are the dis-

covery and the shared ontology. The third ontology

145

Khemakhem S., Drira K. and Jmaiel M..

Semantic Matching to Achieve Software Component Discovery and Composition.

DOI: 10.5220/0004369201450148

In Proceedings of the 9th International Conference on Web Information Systems and Technologies (WEBIST-2013), pages 145-148

ISBN: 978-989-8565-54-9

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

is the integration ontology which describes the com-

ponent’s internal structure to facilitate its integration

in the current work after its selection. To improve

the integration process we also use, Output-Matching-

Service and Input-Output-convertor are service types

used in matching parameters. This paper is organized

as follows: Section 2 presents the discovery and the

integration ontology. We will devote section 3 to de-

tail the shared ontology for composite component dis-

covery. We explain in section 4 how mapping is done

between the discovery ontology and the shared ontol-

ogy. In conclusion, we will suggest some openings

and prospects related to this study.

2 THE DISCOVERY AND THE

INTEGRATION ONTOLOGY

We describe the semantics of components to express

knowledge about functional and non-functional as-

pects of a component. This knowledge comprises:

• The structural aspects that specify the compo-

nent’s internal structure. The developer uses these

aspects to determine if interaction exists between

component operations and other components used

to build the current project.

• The functional aspects that identify the function-

alities of the component is expected to provide

through many features. These features include

methods that are used to adapt the behavior of

the component to his context. The adaptation is

made by specializing and customizing. The other

kinds of features are used by the application spe-

ciﬁc part of a component-based software. Gen-

erally this type of information is speciﬁed by the

component’s methods.

• The non functional aspect speciﬁes the compo-

nent constraints related to communication or com-

putation. The non functional aspect includes

features such as performance, availability, reli-

ability, security, adaptability and dependability.

We distinguish static and dynamic categories of

non functional features. Static features, such as

security-related constraints, do not change during

component execution. Dynamic features, such as

performance-related properties, depend on the de-

ployment environment.

All these features represent different and comple-

mentary views of a component. The feature set used

to describe a component, depends on the developer

action: discovery and integration. The discovery of a

component is made by sending a query to the repos-

itory manager. Once a set of components has been

selected, additional features are speciﬁed to select a

component before integration. For the discovery ac-

tion, the query includes functional and/or non func-

tional features. For integration action, the structural

features have to be speciﬁed.

The underlying approach for SEC++ is based on

the following ontologies( see Figure 1): (Khemakhem

et al., 2010)

• The discovery ontology that speciﬁes functional

and non functional features.

• The integration ontology that describes the prob-

lem solving method (PSMs) used to specify the

component’s structural features.

Figure 1: Discovery and integration ontologies.

3 THE SHARED ONTOLOGY

FOR COMPOSITE

COMPONENT DISCOVERY

This approach exploits the advantages of seman-

tic composition approaches, powered by ontologies

at both component discovery and integration levels.

Building on top of that, we introduce an ontology-

based semantic approach. First, the semantic com-

ponent speciﬁcation provides a mechanism to enrich

atomic components with more semantics than the syn-

tactical method. Second, mapping atomic compo-

nents and other relevant concepts into a centralized

shared ontology offers a knowledge repository for

software components (see Figure 2). The objective

of semantic enhancement is to support ontological

heuristics in order to enable automated and dynamic

component composition. When our enhanced search

engine SEC++ receives a query from a consumer, it

ﬁrst searches the discovery ontology. Our approach

enhances the discovery ontology with a shared on-

tology. This centralized ontology represents relevant

WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies

146

components and concepts in a speciﬁc domain, con-

structed by mapping and integrating individual dis-

covery ontologies for software components. Here,

the ontological heuristics serves as guidelines to re-

spond to a developer request. After using ontologi-

cal heuristics on the shared ontology, SEC++ gener-

ates a number of alternative solutions to component

composition. These alternatives are then evaluated

by a decision engine using a set of criteria speciﬁed

by the developer. Such criteria may include QoS-

based optimization of component composition, busi-

ness rules and strategies. A selected optimal compo-

sition scheme is then executed.

As for the integration ontology, we employ prob-

lem solving method to develop a local ontology for

component. In the integration ontology we try to di-

vide the component process into tasks. Tasks are ei-

ther solved directly (by means of primitive methods),

or are decomposed into subtasks (by means of de-

composition methods). We use the Uniﬁed Problem-

Solving Method Language (UPML) to describe the

components of PSMs (task, method and adapter).

Similarly, the component model subclass is especially

beneﬁcial for composition ( see Figure 1). The pro-

posed approach utilizes the component model class in

two ways. For base components, a component model

keeps information about composability, which speci-

ﬁes when the component can be used in a composite

component. For composite components, a component

model maintains alternative composite solutions in-

crementally for reuse. This semantic enrichment pro-

vides a self-learning capability of component compo-

sition.

Integration

ontology

= Shared ontology

Software components

Discovery

ontology

Meta data

Ontological heuristic

Figure 2: An ontology-supported system for component

composition.

Local integration ontologies are consolidated in a

server by ontology mapping and integration. As a re-

sult, all relevant concepts and components in a do-

main are in the shared ontology, local discovery ontol-

ogy for a component is mapped into the shared ontol-

ogy, appearing as a node in the ontology tree. How to

organize all components into the repository depends

on domains and application requirements. For exam-

ple, for calculating Matrix we can maintain semantic

relationships (e.g., hierarchical and sibling relation-

ships) between Matrix operations. The shared on-

tology also represents other application-speciﬁc con-

cepts for mapping and integrating components. The

mapping and integration not only unite component

descriptions and concepts but also add more seman-

tics. Moreover, the shared ontology enables ontolog-

ical heuristics, thus facilitating dynamic component

composition. For example, we can study composabil-

ity of components based on some generic concepts.

As a simple example, when composing component

C2 that calculates the determinant of a real matrix by

receiving the output parameters of a component C1

that calculates the sum of two matrix which have a

natural type. At ﬁrst glance, these two components

cannot be composed. However, the relationship be-

tween real and natural is revealed in the type ontol-

ogy: natural is included in real.

4 MAPPING THE DISCOVERY

ONTOLOGY INTO

INTEGRATION ONTOLOGY

In this section, we present a prototype system

for mapping a discovery ontology of mathemati-

cal service into integration ontology, which is de-

veloped based on the proposed ontology-supported

software component integration. Our prototype

system employs W3C-recommended standards (i.e.,

RDF+OWL) for semantic description and ontologi-

cal engineering. The software utilized for this task is

Prot

e 3.4.4. At the discovery ontology description

level, the prototype system translates component de-

scriptions and then adds more semantics in the com-

ponent model class. The composability property of

the component model class can have values denoting

possible ways for component composition.

Taking the Linear system resolution component as

an example, its composability contains a list of possi-

ble parameter ﬂows (from inputs to outputs), each of

which can be a part of an alternative path in a compos-

ite component (Complex Matrix −→ Complex Ma-

trix), (Reel Matrix −→ Real Matrix).

Another way to exploit composability is ﬁrst to at-

tach composability to other properties with concrete

meanings, then associate composability with compo-

sition rules.

All individual discovery ontologies are mapped

together, appearing as nodes or subclasses in the

shared ontology. For example, the LSR component

is a subclass of the LS class (see Figure 3). The pro-

totype can extract some metadata from individual dis-

covery ontologies and map into the shared one. After

SemanticMatchingtoAchieveSoftwareComponentDiscoveryandComposition

147

Figure 3: mapping the discovery ontology individual in-

stances into shared ontology.

organizing those base Matrix Operation services into

a shared knowledge repository, the prototype adds

other concepts relevant to Matrix operations, either

domain-speciﬁc or generic, such as Type.

5 CONCLUSIONS

In this paper, we presented the latest version of

SEC: SEC++, which cover the whole set of on-

tologies. It is a persistent component for discov-

ering atomic and composite components. It deliv-

ers services and helps the developer to locate ap-

propriate components for integration into the cur-

rent work. To ensure the successfulness of this pro-

cess, we developed the discovery, the integration and

the shared ontologies. The ontological heuristics

in used to enable automated and dynamic compo-

nent composition. To improve the integration process

we have used Output-Matching-Service and Input-

Output-convertor service types used in matching pa-

rameters. In the future work, we plan to develop a

fuzzy search system that aims to discover component

which is most relevant to the user according to his/her

perception.

REFERENCES

Akkiraju, R., Bournez, C., Domingue, J., Farrell, J., Ferrari,

L., and Henocque, L. (2007). Semantic Annotations

for WSDL and XML Schema, W3C recommenda-

tion edition.

Chaari, T. and Fakhfakh, K. (2012). Semantic model-

ing and reasoning at runtime for autonomous sys-

tems engineering. In 9th International Conference on

Ubiquitous Intelligence and Computing & Autonomic

Trusted Computing(UIC/ATC).

Chaari, T., Tazi, S., Jmaiel, M., and Drira, K. (2013).

ODACE SLA: Ontology driven approach for auto-

matic establishment of service level agreements. IJS-

SOE, 1(3).

de Bruijn, J., Bussler, C., Domingue, J., Fensel, D., Hepp,

M., Keller, U., and Kife, M. (2005). Web Service Mod-

eling Ontology (WSMO), W3C recommendation edi-

tion.

Junghans, M. and Agarwal, S. (2010). Web service dis-

covery based on uniﬁed view on functional and non-

functional properties. In Proceedings of the 4th

IEEE International Conference on Semantic Comput-

ing (ICSC 2010). IEEE Computer Society.

Khemakhem, S., Drira, K., and Jmaiel, M. (2010). An in-

tegration ontology for components composition. In

International Journal of Web Portals, 2(3):35–42.

Khemakhem, S., Drira, K., and Jmaiel, M. (2011). Modern

Software Engineering Concepts and Practices: Ad-

vanced Approaches, chapter Description, classiﬁca-

tion and discovery approaches for software compo-

nents: a comparative study. IGI publisher.

Seog-Chan, Lee, D., and Kumara, S. R. T. (2006). A com-

parative illustration of AI planning-based web ser-

vices composition. SIGecom Exch., 5(5):1–10.

Smith, M. K., Welty, C., and McGuinness, D. L. (2004).

OWL Web Ontology Language Guide, W3C recom-

mendation edition.

WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies

148