A Semantic-based Approach for Ontology Module Extraction

Amir Souissi

, Walid Chainbi

and Khaled Ghedira

National School of Computer Science, SOIE, Manouba University, Manouba, Tunisia

Sousse National School of Engineers, SOIE, Sousse University, Sousse, Tunisia

Higher Institute of Management, SOIE, Tunis University, Tunis, Tunisia

Keywords: Ontologies, Modularization, Extraction, Semantics.

Abstract: Ontology modularization is crucial to support knowledge reuse on the ever increasing Semantic Web.

However, modularization methods that serve the reuse goal are often intended for humans to assist them in

building new ontologies, rather than for applications that need only a relevant part of an existing ontology.

Moreover, modules obtained are always subject to verification and maintenance by humans to validate the

semantic consistency of their contents. In this paper, we investigate how semantic comparisons may provide

a module relevant to a set of terms which are not part of the ontology. Our objective is to extract a module

which may be usable as a separate ontology. The user does not need to be familiar with the exact terms used

inside the ontology beforehand to extract from it a module for a specific application/knowledge sub domain.

1 INTRODUCTION

Ontologies have established themselves as a

powerful tool to enable knowledge sharing, and a

growing number of applications have benefited from

the use of ontologies as a means to achieve semantic

interoperability among heterogeneous, distributed

systems. Ontologies play a key role in one of the

newest areas of interest, the Semantic Web, as

confirmed by efforts such as OntoWeb, and OWL. A

widely quoted definition of an ontology was

proposed by Gruber who defines it as an explicit

specification of a conceptualization (Gruber, 1993).

An ontology specifies a vocabulary including the

key terms, their semantic interconnections, and some

rules of inference.

With the evolution of cooperative and distributed

systems, and the emergence of the Semantic Web,

ontologies have become an indispensable resource.

The number of ontologies available on the Web has

also increased due to the appearance of several tools

that assist users in creating their ontologies. This has

posed problems of understanding and reuse of those

resources already difficult to design. A solution was

then proposed by knowledge engineers namely

modularization. Spaccapietra indicates in

(Spaccapietra, 2005) that a module is a subset of a

whole that makes sense and can somehow exist

separated from the original ontology and not

necessarily supporting the same functionality as it.

He highlights five goals to modularization which are

scalability, complexity management,

understandability, personalization and reuse. He

considers that the understanding of what

modularization exactly means and what are the

advantages and the disadvantages which are

expected from modularization depend on these goals

assigned to modularization. Since ontology

construction is a labor intensive task and it is time

consuming, the modularization methods which serve

the purpose of reuse often focus at reusing

ontological modules for building new ontologies

(Cuenca Grau et al., 2007b; Doran et al., 2007). The

focus of this paper is on ontology modularization for

reuse. We aim to extract a part from an ontology in a

way such that it can be reused as an ontology instead

of the original one. Our objective is to allow

obtaining a module which covers a specific topic

from the ontology and to consider this module as a

new ontology modeling this topic. Since some

current ontologies are evolving to more expressivity

and complexity, we propose an approach which

targets ontologies without a clear internal structure

(more semantic relations and hierarchical staple

relations). Our approach intends to extract a module

relevant to a set of terms which may be different of

these employed inside the ontology. The idea is to

extract a module without being necessarily familiar

222

Souissi A., Chainbi W. and Ghedira K..

A Semantic-based Approach for Ontology Module Extraction.

DOI: 10.5220/0004544402220229

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2013), pages 222-229

ISBN: 978-989-8565-81-5

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

with the entities names inside the ontology.

This paper is organized as follows. Section 2

deals with the previously proposed techniques for

ontology modularization. Then in section 3, we

present our approach. A case study illustrating the

proposed approach is presented in section 4. In

section 5, we describe the usefulness of our

approach in an application domain namely

information retrieval. We conclude in section 6 with

a summary of the main points relevant to this study,

and we give directions for future work.

2 RELATED WORK

Several modularization methods of ontologies have

been proposed in the literature. These methods are

based on two antagonistic approaches. The first is a

composition approach in order to obtain a modular

representation. The modularization can be perceived

from this perspective as a mechanism for assembling

some ontologies (modules) into a coherent network

that can be referred to as a single entity, modular

ontology. The result is a set of integrated or inter-

connected ontologies into a larger and more complex

network. The second approach is a decomposition of

a large ontology, which contains a large number of

concepts and relations into a set of smaller modules,

easy to understand and manage.

Decomposition methods proposed in literature,

belong mainly to two large families. Partitioning

methods are automatic and provide a set of modules

that can be disjoint or overlap. Examples include

partitioning methods that produce disjoint modules

(Cuenca Grau et al., 2007b; Stuckenschmidt and

Klein, 2004). Some others, like partition-based

methods (MacCartney et al., 2003), allow modules

to overlap. As for the extraction methods, they

involve the user in the extraction process and

provide a single fragment of the ontology. These two

categories of methods are generally based either on

logical criteria (Cuenca Grau et al., 2007a) or on

structural criteria (Ghiraldi et al., 2006; Noy and

Musen, 2009; Seidenberg, 2009). In both cases,

human intervention is necessary after the

modularization process to verify that the module is

covering a consistent knowledge area. We believe

that this is due to the fact that these methods neglect

the semantic aspect in the modularization process.

The methods based on structural criteria target

specific ontologies. This is the case, for example, of

the method of Seidenberg (Seidenberg, 2009) where

the ontology referred, is the GALEN ontology

(Rector and Rogers, 1999), which is characterized

by the strong presence of hierarchical relationships

between concepts. The results of modularization are

favourable only in the case of ontologies that have a

structure similar to that described at the outset.

The other methods based on criteria of

description logic, define the module formally by

setting the logical conditions in advance. Portions of

ontologies that satisfy these conditions are

considered as modules. Although these methods

consider a certain level of semantics, the modules

are usable only if humans validate the module, by

browsing it to estimate the concepts that are relevant

to its application. Cuenca Grau et al. propose in

(Cuenca Grau et al., 2007b; Cuenca Grau et al.,

2005) an algorithm to obtain partitions whose

elements are disjoint, starting with a formal

definition in order to characterize ontologies that are

susceptible to be decomposed safely. Indeed, if the

ontology does not have certain formal characteristics

defined by the algorithm, it cannot be modularized.

This is not ideal because it reduces the number of

ontologies ready to modularization. The work of

Cuenca Grau et al. is based on the notion of

conservative extensions (Ghiraldi et al., 2006; Lutz

et al., 2007). This means that essential inferences

about entities contained within an element of such a

partition should be preserved. Whilst conservative

extensions can theoretically be used to define an

ontology module, they cannot currently be used in

practice as deciding if an OWL-DL module is a

conservative extension is undecidable (Doran et al.,

2007; Lutz et al., 2007). In (Wandelt and Möller,

2012), the aim is to introduce modularization

techniques for ABoxes in order to obtain a set of

modules to release the main memory burden of DL

reasoning systems for semi-expressive ontologies.

They have proposed to transform an ABox to a

graph by mapping each individual in the Abox to a

node in the graph and then to decompose this graph

relying on connectedness-based graph partitioning

techniques. The algorithm gave a negative result for

SHOQ

DL (nominals problem, completeness

problem). In order to ameliorate results, an

intentional-based modularization by splitting role

assertions with ABox-splits is presented. This

method relies on internal paths of role assertions

between individuals. The method did not consider

the semantic relations expressed by these assertions

and the decomposition is completely depending on

the graph structure. Furthermore, there are no user

requirements considered during the splitting.

Both classes of methods mentioned above reduce

the reuse possibility. Indeed, these methods have

been dedicated for specific ontologies often

ASemantic-basedApproachforOntologyModuleExtraction

223

characterized by particular structures and properties.

In addition, human intervention, for checking the

semantic consistency of the concepts that make up

the module, is required before using the extracted

module by the final application.

In this paper, we propose an approach which take

into account the semantic aspect in the

modularization process (an application may use the

extracted module without the need for human

intervention in order to validate it). Moreover, the

extraction process does not begin from internal

properties of the ontology. When a software agent

would extract modules from a set of different

ontologies of different knowledge domains, it is not

necessary for it to know about the entities inside

these ontologies. As a starting point, a set of terms

relevant to some domain is entered. The module is

produced using a semantic matching between these

terms and the ontology concepts. The produced

module is intended to be considered as a separate

ontology relevant to these terms.

3 PROPOSED APPROACH

In general, the modularity of ontologies serves three

principal goals:

 The reuse of the fragments (modules) of ontologies

in the construction of new ones.

 The interoperability of the distributed systems

through the interpretation of the local semantics of

ontologies that constitute modules in the global

system.

 The extensibility for evolution and maintenance,

and the scalability for efficient reasoning by

localizing the inference in the module rather than

to reason on all the ontology.

In this paper, we propose an approach that serves

the purpose of reuse. However, reuse here is not

intended to assist developers in building new

ontologies, as is the case with the other methods of

modularization. In fact, we seek primarily to help

the user obtaining a relevant ontology module,

which captures a set of knowledge from a wider

existing ontology. Indeed, it would be interesting to

give the user methods and tools that offer an extract

from an ontology, which plays the role of ontology

in itself. Thus, reusing the module by integrating it

directly into an application, saves time to build a

dedicated ontology. Note here that the user may be

human or machine. It is rather the case of

applications that want to use these modules, which

interest us the most because we are looking to

propose a solution that makes use of modules as

ontologies, independently of human intervention.

The modularization approach we propose is part

of the decomposition approaches of monolithic

ontologies. It is an extraction method since

it aims to extract a relevant ontology module. The

aim of the approach is to provide the user with an

ontology module that covers a sub-domain of the

domain of the ontology. The method should allow

the user to express its needs by entering the concepts

which interest him. The result is a fragment

composed of concepts and relations that are relevant

to the module i.e., which have semantic relationship

with the concepts submitted by the user. We

consider a semantic relationship between two

concepts, as one of the four logic functions as

follows:

─ Identity Relation: it is a semantic relation between

two concepts that have the same syntax, the same

attributes and operations. Example: Identity

(Person, Person).

─ Synonymy Relation: it is a semantic relation

between two concepts that express the same

meaning. Example Synonymy (Person,

Individual).

─ Classification Is-a Relation: two concepts where

one is expressing a particular case of the other.

Example: Is-a (Student, Person).

─ Antonymy Relation: is used between two concepts

that have opposite meanings. Example Antonymy

(Registered, Unregistered)

For experimental reasons, we consider only these

four semantic relations. These semantic relations

exist in WordNet which is a large lexical database

for English language. It groups words together based

on their meanings and label the semantic relations

among words. We exploit these properties to

identify the semantic relations between the concepts.

For example, in an ontology that describes an e-

learning course, the user may be interested in

participants in that course. The method should

extract a module semantically rich on participants,

from the ontology of departure. For this purpose, we

verify if one of the semantic relationships described

above exists between the keywords entered by the

user and the concept of the ontology. The

comparison operation is only restricted to named

concepts.

We motivate our approach as follows:

 User Involvement: In the context of reuse, the user

should be satisfied with the result. If not satisfied,

he should be able to better communicate his needs

to be taken into account in the process of

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

224

modularization. Our approach involves the user

(human or computing system) in the process of

modularization, unlike the automatic

decomposition approaches (Cuenca Grau et al.,

2007b; Stuckenschmidt and Klein, 2004). This

allows him to express his needs regarding the

result he looks for. He begins by entering the

central concepts for the module he wants to

achieve. When he gets a result that does not satisfy

his needs, he starts the modularization by changing

the settings (e.g., concepts to include, concepts to

exclude …) to refine the result and get a different

module of the previous one.

 More Semantics: The extraction methods proposed

by Seidenberg, and Noy and Musen involve the

user (Noy and Musen, 2009; Seidenberg, 2009).

But these methods depend heavily on the structure

of the ontology. In addition, the concepts selected

by the user are part of the ontology to decompose.

In fact, their algorithm follows the links between

concepts to determine the portion to be extracted.

In our approach, the user may enter concepts that

can be internal or external to an ontology. It is a

new aspect in the operation of modularization that

other methods have not explored. Indeed, these

methods work with the concepts that make up the

ontology and do not address the case where the

user provides concepts that are not elements of this

ontology. The essential for us is that the module

should capture the meaning of concepts by looking

for concepts that are in strong semantic relation

with those of the ontology. Thus, the main

contribution of our approach is that the module is

determined on the basis of the semantic

relationship that can exist between internal and

external concepts.

 Low Coupling and High Cohesion: reuse and

extensibility are the goals sought in the operation

of an ontology modularization. Nevertheless,

achieving these two objectives requires that the

modules are loosely coupled and highly cohesive.

The coupling means the modules dependency.

Loose coupling means a weak relationship

between modules allowing flexibility for updating

and maintenance. So, each module can be

modified by limiting the impact of change to the

rest of the ontology. Cohesion measures the

dependence of the components of modules. In

other words, concepts, relations, and individuals

are strongly linked to each other within the same

module. So, cohesion denotes the degree of

relatedness of elements within the same module

(D’Aquin et al., 2009). If we consider two

concepts are strongly-related if there is a semantic

relationship between them. We can use the

semantic relationship, as a mean to identify how

strongly-related are the concepts, and consequently

if they are parts of the same module. So we can

reach high cohesion, in a portion of ontology,

based on the notion of semantic relationship.

Before beginning to describe the approach

currently being investigated, we propose the

following definition of a module: an extracted

ontological module is the relevant part of an

ontology to a set of terms which are not necessarily

the exact terms used inside the ontology. It is

intended to cover a sub-area of knowledge for which

a module needs to be extracted.

This definition implies that ontology module is a

single extract and it can be reused as a full

independent ontology. The user (human or machine)

may be not familiar with the content of the ontology.

If he needs to extend the module with new concepts

and relations then the module should be viewed as

an ontology itself. The quality of the module

depends on the relevance of the knowledge captured

by the module relative to the user query.

Our approach is based on two basic steps:

- 1

step: Identifying concepts that have a

semantic relationship with external terms.

- 2

step: composition of the module based on

the concepts identified in Step 1. All concepts that

appear in the definition of the concepts identified are

considered part of the module. The module is

composed from the union of all retrieved axioms.

The algorithm identifies a module by doing

comparisons between a term entered by the user and

a concept of the ontology. WordNet is traversed to

extract synonyms, antonyms or hyponyms

depending on the user choice. In case the term is

identical to a concept name, we consider it as part of

the module. If there is a semantic relationship

between them, the concept of the ontology is moved

to the module. In addition, in case the extracted

concept has an equivalent definition with another

concept in the ontology, all the definition is

extracted. So, all the concepts within the module

constitute a subset of concepts definitions that are

extracted from the original ontology. In case there is

not a semantic relationship between the compared

concepts, then the ontology concept is not extracted.

The algorithm continues so until all concepts in the

ontology are compared with the user concept.

At the beginning of the algorithm, the user may

choose the concept by entering its name. So, the

extraction procedure is automatic but it takes into

account the user requirements. In this paper, we

present the approach and show its feasibility at a

ASemantic-basedApproachforOntologyModuleExtraction

225

practical level as we show in the next section. The

aim of the paper is not to test our method on real

well known ontologies. It aims rather at proving the

contribution of the modularization based on a

semantic matching for some kinds of ontologies (i.e.

expressive ontologies not taxonomies). Thus, we

present theoretically, in the following paragraph,

some evaluation criteria which we can apply on this

approach.

As we are only interested in one module,

evaluation criteria dedicated to sets of

interconnected modules resulting from partitioning

techniques – redundancy, connectedness, and inter-

module distance–are not relevant in our technique.

However, since our method aims to produce a

relevant module to a set of terms in order to use the

module as an ontology, we can use evaluation

criteria for determining the quality of the ontology to

evaluate the quality of the onotology module. These

criteria are mainly the module cohesion, the richness

of the representation and the domain coverage.

 Module cohesion denotes the degree of relatedness

of elements within the module. Cohesion metrics

are based on the structure of the ontology: the

number of root classes, the number of leaf classes,

the maximum depth of the hierarchy.

 Richness of the representation denotes the amount

of conceptual information retained in the module.

The richness of semantic information in a module

depends on the richness of the mother ontology.

Richness metrics such as - the average number of

subclass relations per class- or the -average

number of domain relations per class- can be used.

 Domain coverage is the criterion which determines

how well the module fits the representational

requirements of the application that request it. To

determine the domain coverage, we need a suitable

representation of the domain that should be

covered by the module. Comparing a corpus of

documents with the module is a technique for

determining how well the ontological module

represents the content of the documents.

Another evaluation criterion which can be

considered is the performance measuring. It is

important to consider it, particularly when using a

modularization technique for the purpose of an

application.

We present in the next section some of the

screen shots of our developed system which was

tested under an ontology that describes an e-learning

course.

4 CASE STUDY

We provide an example of extracting a module from

an ontology to illustrate our approach. The ontology

expressed in description logic corresponds to the

following axioms:

a1 correction ≡ page ⊓ ∃associated.exercise

a2 exercise ≡ page ⊓ ∃associated.course

a3 course ≡ page ⊓ ∃caracterized. session

a4 session ≡ ∃caracterize.course ⊓

∃composed.module ⊓ ∃associated.test

a5 module ≡ ∃associated.Tutor ⊓

∃associated.registered ⊓ ∃compose. session

a6 test ≡ ∃associated. session ⊓

∃corrected.tutor ⊓ ∃performed.registred

a7 tutor ≡ ∃associated.module ⊓ ∃correct.test

⊓ teacher

a8 registered ≡ ∃associated.module ⊓

∃perform.test ⊓ student

a9 person ≡ teacher ⊔ student

Suppose the user wants to extract an ontology

module relevant to persons which participate in an e-

learning course. He may enter the term “person”,

which is the name of an internal concept. He may

also enter the terms “coach” or “unregistered”.

These words are syntactically different from

ontology concepts, but they belong to the same

context for the user, that is to say people which

participate in an e-learning course.

Result of the 1

Step:

If one refers to the semantic relationships

defined above, we find that there is an identity

relation for the concept person. An antonymy

between the concepts registered and unregistered.

And a synonymy between coach and tutor.

Result of the 2

Step:

The definitions that we found for these concepts,

in the Tbox of the ontology are:

a7 tutor ≡ ∃associated.module ⊓ ∃correct.test

⊓ teacher

a8 registered ≡ ∃associated.module ⊓

∃perform.test ⊓ student

a9 person ≡ teacher ⊔ student

Note that the concepts that have not a semantic

relationship with the original concepts (“teacher”

and “student”) chosen by the user appear in the

definition of the found concepts. Concepts like

“Module” and “Test” are considered as part of the

module because they are considered as part of the

definition of the concepts founded.

Figure 1. is a screen shot of our developed

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

226

system. By clicking on the button Display, all

extracted named concepts are listed above and the

module is created in a separate owl file (Figure. 2).

Figure 1: Screen shot of the modularization approach.

Figure 2: The obtained module in a separate owl file.

5 USEFULNESS

OF OUR APPROACH

Our approach is based on identifying the semantic

relations between terms of concepts. It can have

applications in many natural language processing

tasks, such as Information Extraction and

Information Retrieval. We discuss in this section the

usefulness of our approach in the domain of

Information Retrieval (IR).

An IR system allows users to look for

information in a collection of documents (or other

information sources) through queries usually

formatted as a set of keywords (Baeza-Rates and

Ribeiro-Neto, 1999). There are three main steps for

the process of IR: The indexing process, the query

processing and the matching between the query

Figure 3: Information retrieval processes.

terms and the documents. These processes are

visualized in Figure 3 (Goker and Davies, 2009).

The goal of the indexing process is to represent the

content of the documents in order to be used in

further searches.

There are two types of indexing: bag-of-words

indexing and semantic indexing. For the first type,

the indexing terms are extracted from the documents

content itself. It includes two steps: searching the

terms and weighting them. The second type aims to

rely on ontologies to represent documents. From this

point of view, the descriptors (indexing terms) are

chosen directly from the ontology rather than the

documents. So, documents are indexed by concepts

which reflect their meanings, rather than frequently

ambiguous words (Aussenac-Gilles and Mothe,

2004).

Semantic indexing consists of two steps. The

first step consists of identifying the ontology

concepts or instances in the documents. The second

step consists of weighting the concepts for every

document according to the conceptual structure

which they are derived (Hele and Tanel-Lauri,

2001).

Combining the usability of keyword-based

interfaces with the power of semantic technologies is

one of the most challenging areas in semantic

searching. To use an ontology in an IR system, it

needs to choose it first. As much as ontologies in

different domains are now accessible, reusing them

could be a solution for ontology integration in IR

systems. In this case, ontologies are generally

chosen only based on the knowledge domain they

address (Baziz et al., 2005). Once the ontology

chosen, the knowledge it represents can be used

when indexing documents. Thus, the choice of the

Information

Feedback

Query

Documents

Indexe

Matchin

Indexing

Query

formulation

Retrieved

documents

ASemantic-basedApproachforOntologyModuleExtraction

227

ontologies which will be used for indexing is a

primordial step. Our ontology modularization

approach would be useful in this context.

In many studies, the choice of the domain

ontology which will serve to represent the corpus is

dependent of the task domain itself (Vallet et al.,

2005). Thus, the reusability of the ontology for

another task or another domain is not insured.

In fields, like medicine, ontologies have

especially great size, and contain many knowledge

domains. A collection of medical documents could

be represented by the ontology. We can have a

corpus which talks about a specific disease and

another corpus which talks about treatment of this

disease. As a result of a classic semantic indexing,

the two corpuses are indexed by a single ontology.

At the end, we obtain many concepts which are

shared to represent the two corpuses. This can affect

the relevance of the document retrieved later. In this

case, our modularization approach would be useful.

In fact, in this case, we aim to extract two modules,

from this ontology, which are semantically related to

the two corpuses. Every module is a representation

space of its correspondent corpus. When a query

concerning the disease is formulated, only the

documents which are indexed semantically by the

disease module are retrieved.

6 CONCLUSIONS AND FUTURE

WORK

In this paper, we proposed a method to extract

modules from ontologies based on semantic

relations identification. We considered four semantic

relations which are: Identity, Synonymy,

Classification and Antonymy. We considered that

two concepts are relevant for the module if there

exists a semantic relation between them. We have

used Wordnet to identify the semantic relation

between the external concept (the user request) and

the internal one (the ontology concept). The result of

the extraction is a module composed from these

concepts and their definitions.

We show that the use of semantic relations

makes the method less dependent to the structure of

the ontology to modularize. It is effectively intended

to high expressive and more complex ontologies

rather than ontology structures based on

subsumption relations. The user is involved in the

modularization process but he is not supposed

knowing the components of the ontology. His needs

are expressed as a list of relevant concepts for his

purpose. Hence, the method is automatic but takes

into account the user requirements. The user here

could be a human or an application program. In fact,

the main goal of this approach is to allow programs

to extract useful modules from available ontologies

on the Web. In this way, our goal meets the

objective of the semantic Web which is to allow data

to be shared, understood and reused across

applications.

In future work, we envision to evaluate the

usefulness of our approach. For this purpose, we

have to determine the possible evaluation criteria,

including application-dependent criteria, which can

be used to determine the quality of a module. We

intend to develop an IR system for medical Web

documents using ontology modules to index the

documents. The efficiency of the approach would be

discussed in the context of experiments that aim to

measure the relevance of the retrieved documents.

REFERENCES

Cuenca Grau, B., Horrocks, I., Kazakov, Y., Sattler, U.,

2007a. A Logical Framework for Modularity of

Ontologies. In International Joint Conference on

Artificial Intelligence, pages 298-303.

Cuenca Grau, B., Horrocks, I., Kazakov, Y., Sattler, U.,

2007b. Just the Right Amount: Extracting Modules

from Ontologies. In International World Wide Web

conference.

Cuenca Grau, B., Parsia, B., Sirin, E., Kalyanpur, A.,

2005. Modularizing owl ontologies. In Proceedings of

the KCAP Workshop on Ontology Management.

Doran, P., Tamma, V., Lannone, L., 2007. Ontology

Module Extraction for Ontology Reuse : An Ontology

Engineering Perspective. In Conference on

Information and Knowledge Management CIKM’07,

Lisbona, Portugal.

Ghiraldi, S., Lutz, C., Wolter, F., 2006. Did I damage my

ontology? A case for conservative extensions in

description logics. In P. Doherty, J. Mylopoulos, and

C. Welty, editors, Proceedings of the Tenth

International Conference on Principles of Knowledge

Representation and Reasoning.

Lutz, C., Walther, D., Wolter, F., 2007. Conservative

extensions in expressive description logics. In

International Joint Conference on Artificial

Intelligence, pages 453-458.

Noy, N. F., Musen, M. A., 2009. Traversing Ontologies to

Extract Views. In Heiner Stuckenschmidt, Christine

Parent, Stefano Spaccapietra (Eds.), Modular

Ontologies concepts, Theories and Techniques for

Knowledge Modularization, Springer-Verlag Berlin,

Heidelberg, LNCS 5445, pages 245-260.

Rector, A., Rogers, J, 1999. Ontological Issues in using a

Description Logic to Represent Medical Concepts:

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

228

Experience from GALEN. In IMIA WG6 Workshop.

Seidenberg, J., 2009. Web Ontology Segmentation:

Extraction, Transformation, Evaluation. In Heiner

Stuckenschmidt, Christine Parent, Stefano

Spaccapietra (Eds.), Modular Ontologies concepts,

Theories and Techniques for Knowledge

Modularization, Springer-Verlag Berlin, Heidelberg,

LNCS 5445, pages 211-243.

Stuckenschmidt, H., Klein, M., 2004. Strutured-based

Partitioning of large concept hierarchies. In

Proceedings of the 3rd International Semantic Web

Conference.

Spaccapietra, S., 2005. Report on Modularization of

Ontologies. Knowledge Web Deliverable 2.1.3.1.

MacCartney, B., McIlraith, S., Amir, E., Uribe, T.E.,

2003. Practical Partition-Based Theorem Proving for

Large Knowledge Bases. In Proc. of the International

Joint Conference on Artiﬁcial Intelligence IJCAI.

d’Aquin, M., Schlicht, A., Stuckenschmidt, H., Sabou, M.,

2009. Criteria for ontology modularization techniques.

In Proceedings of Modular Ontologies, pages 67-89.

Gruber, T. R., 1993. A Translation Approach to Portable

Ontology Specifications. Knowledge Acquisition, Vol.

5, 199-220, November 2.

Baeza-Yates, R. and Ribeiro-Neto, B., 1999. Modern

Information Retrieval. Addison wesley Longman.

Goker, A., Davies, J., 2009. Information Retrieval Models.

In J.Information Retrieval : Searching in the 21st

century. Jhon wiley and sons, Ltd., ISBN-13: 978-

0470027622.

Aussenac-Gilles N., Mothe J., 2004. Ontologies as

Background Knowledge to Explore Document

Collections. Proceedings of Computer-Assisted

Information Retrieval (Recherche d'Information et ses

Applications) - RIAO 2004, 7th International

Conference, University of Avignon, France, pages

129-142.

Hele-Mai H., and Tanel-Lauri L., 2001. A survey of

Concept-based Information Retrieval Tools on the

Web. Proceedings of the 5th East-European

Conference, ADBI, vol 2, pages 29-41. Vilnius,

Lithuania.

Baziz M., Boughanem M., Aussenac-Gilles N., Chrisment

C., 2005. Semantic Cores for Representing Documents

in IR. Proceedings of the 20th ACM Symposium on

Applied Computing, ACM Press ISBN :1-58113-964-

0, pages 1020-1026.

Vallet D., Fernandez M., Castells P., 2005. An Ontology-

Based Information Retrieval Model. Proceedings of

the 2

European Semantic Web Conference, pages

445-470.

Wandelt S., Möller R., 2012. Towards ABox

Modularization of semi-expressive Description

Logics. Journal of Applied Ontology, V.7, pages 133-

167, DOI 10.3233/AO-2012-0105, IOS Press.

ASemantic-basedApproachforOntologyModuleExtraction

229