MS-ONTO

Model and System for Supporting Ontology Evolution

Emile Tawamba

, Roger Nkambou

, Bernabé Batchakui

and Claude Tangha

ALOCO-LIRIMA, University of Yaoundé 1, Yaoundé, Cameroon

GDAC, UQAM Montréal, Canada

Keywords: Ontology Engineering, Ontology Evolution, Integrity Constraints, Semantic Web.

Abstract: Ontology is becoming the key knowledge capture structure in many domains. It plays a very important role

in the area of semantic web and is widely used in multiple fields including Intelligent Tutoring Systems

(ITS) and e-Learning. Ontologies are intensively used in domain knowledge modeling in specific areas

which can evolve. However, current tools used to implement ontologies fail to provide functions to ade-

quately ensure their evolution. To deal with this issue, we have developed an ontology evolution manage-

ment system named «MS-ONTO», founded on a formal description of evolution operators. MS-ONTO al-

lows the preservation of both the internal and external integrity constraints during the ontology evolution:

the external integrity meaning the preservation of its usage while internal integrity means its conformity to

the constraints (implicit or explicit) related to the ontology model itself. MS-ONTO should be integrated as

a plug-in in existing ontology editors such as NeOn Toolkit and Protégé.

1 INTRODUCTION

Nowadays, the use of ontologies in several research

areas cannot be over emphasized (Brewster, 2007).

The advent of semantic web and new approaches for

data web ease the access to or the sharing of web

resources. The main questions that come up remain

that of interoperability of the different systems of

information exchange between them. Even if some

languages such as XML and RDF play an important

role in the creation/description of resources, there is

yet no totally satisfactory solution for the terminolo-

gies and classification systems which should be used

in their indexing and scouting for necessary exploi-

tation and sharing (Psyché, 2007). Today, ontologies

greatly contribute in solving this issue by providing

a formal framework of conceptual modeling of dif-

ferent domains (Burcu, 2006).

Ontologies are proven effective in many areas

such as e-Learning, e-Commerce and many others.

Ontologies are mostly used in dynamic, distributed

and evolutive environments/systems. It is therefore

important to update ontologies in order for them to

reflect the changes (changes on the state, data, and

needs related to new functionalities) that affect the

systems life cycle.

Many other reasons can also cause a change in on-

tology. For instance, changes can occur when using

it in different contexts or when correcting errors in

conceptualization as well as changes in initial do-

main specification.

In order to enable ontologies to maintain their in-

terest as regards the applications for which they have

been constructed, evolution should be considered as

an integral part of the life cycle of ontologies design.

However, most of the tools used in ontology engi-

neering do not include an effective ontology evolu-

tion service. In fact, most of these systems do not

provide a clear framework for automatically manag-

ing the ontology evolution. The evolution process

mainly relies upon the human user. A good ontology

evolution process should prevent all abnormalities

that can be introduced regarding both the ontology

model itself (internal abnormalities) and the applica-

tion contexts in which the ontology is used (external

abnormalities). It is therefore important to provide a

system in which ontology evolution preserves both

internal and external integrity related to the initial

ontology. This paper presents an ontology evolution

management method based on a formal approach

using description logics and which enables the

preservation of internal as well as external integrity

(Noy, 2004) (Zablith, 2009).

319

Tawamba E., Nkambou R., Batchakui B. and Tangha C..

MS-ONTO - Model and System for Supporting Ontology Evolution.

DOI: 10.5220/0005085203190326

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2014), pages 319-326

ISBN: 978-989-758-049-9

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

The paper is organized as follows: the second sec-

tion presents some related works on the management

of ontology evolution. Our solution is then intro-

duced in the third section where a formal description

of the operators of changes is given with a general

presentation of our global approach towards ontolo-

gy evolution management. The fourth section is the

implementation and evaluation of this solution. The

paper ends with some concluding remarks.

2 RELATED WORK

This section defines some ontology related concepts

followed by existing ontology evolution manage-

ment techniques in order to bring out their limita-

tions.

2.1 Definitions

(Gruber, 1993) defined ontology as “an explicit

specification of a conceptualization”. Ontology is

also considered as the result of a complete formula-

tion and a rigorous conceptualization (hierarchical

organization of pertinent concepts, relationships

between these concepts, rules and axioms binding

them) (Amal, 2008).

Ontology evolution is the adaptation to changes

that are brought during its life-cycle and the propa-

gation of these changes at the level of the dependent

artifact, i.e. objects referred by the ontology as well

as ontologies and their related applications (Sure,

2004). There are various types of changes: changes

in the modelling domain (

Wohlgenannt, 2013), con-

ceptualization and specification changes (Djedidi,

2007), etc. Given that the domain is a part of the real

world, it is thus dynamic and evolves with time.

Conceptualization can also change due to a new

observation or a restructuring of knowledge. Ontol-

ogy could be adapted to be reused in different tasks.

The conceptualization of ontology could be refined

through an interactive and incremental construction

process. Ontology enrichment tools are another way

by which an existing ontology can evolve as new

concepts as well as new relationships that are ex-

tracted and added (Booshehri, 2013). It is therefore

glaring that ontology must undergo changes and

therefore evolve (Cuenca, 2012).

Ensuring the evolution of ontology is a costly

operation which requires the services of a competent

expert in this domain (

Scott, 2013). However, the

significant role of ontologies makes it essential that

they should be kept up to date so as to reflect the

changes which affect the life cycle of the systems

and the applications for which they were conceived.

Many research teams and projects have worked

on ontology evolution and brought out interesting

findings. According to their methods, two broad

categories stand out: ontology versioning and man-

agement of changes (Kondylakis, 2011).

2.2 Ontology Evolution Based on the

Management of Versions

In this first group, the evolution is managed through

the creation and maintenance of different versions of

the same ontology. KAON, CONCORDIA and

SHOE are classic examples of tools and methods

that implement the view.

KAON (Gabel, 2004) offers a range of tools for

ontologies and the semantic web. Presently, it is one

of the rare ontology management systems which has

a function dedicated to the recording of changes.

During the evolution, KAON saves all the changes

carried out to move from one ontology version to

another in a folder as an ordered sequence of

RDF/XML declarations. The main drawback in

KAON approach is that, it deals only with elemen-

tary changes.

CONCORDIA (Shaban, 2010) defines a concep-

tual model for the management of changes of a med-

ical terminology. This model adds to each class a

unique identifier and these classes could only be

subsequently and logically withdrawn but not physi-

cally erased. As such for each class, the Concordia

model is capable of tracing all the parent-classes or

withdrawing children through its identifiers.

HEFLIN proposes SHOE (Heflin, 1999) as a

language for representing knowledge on the web. It

is a language based on FOL (First order logic).

SHOE is based on HTML, which offers primary

terms for the management of multiple versions while

allowing the association to each version of ontology,

a unique identifier and a code stating the compatibil-

ity with former versions. It brings out the relevance

of each revision/operation on data and requests.

As a whole, ontology evolution solutions of this

group do not really treat ontology evolution. They

rather provide a model for analyzing links between

ontology versions, but do not pay attention either to

the management of the dependent artifacts (referred

resources, related ontologies, programs in which the

evolved ontology is used) or to the impact of chang-

es on the ontology internal structure and/or seman-

tic. Moreover, the authors do not provide any func-

tional framework to integrate the totality of method-

ological elements that they propose.

KEOD2014-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

320

2.3 Evolution Based on Management of

Changes

The proponents of this approach think that it is nec-

essary to have a follow-up of changes: the request

for change has to be interpreted, analyzed and exe-

cuted under control. In fact, the system should be

able to notify a change that might lead to an incon-

sistent ontology, better still, it should at least help

the engineer in evolution operations if it cannot

completely automate them.

The ontology must remain consistent while

evolving. Ontology is consistent in relation to a

model if and only if it respects all the constraints of

the model. These constraints are either invariables or

user-defined.

The invariables constraints are those related to

the ontology structure; for instance, removing a root

concept should not be allowed without a correct

reorganization of the ontology. User-defined con-

straints are those defined by the user such as the

maximum number of instance related to a given

concept. A change maintains consistency only when

the resulting ontology is consistent. Hence, change is

defined as a function ch(args, prec, postc) with args

representing the arguments, prec representing all the

preconditions and postc representing all the post-

conditions. Some operations could lead to several

options of possible ontologies; it is therefore neces-

sary to choose the most consistent one that responds

to user’s needs. However, several iterations should

be needed before obtaining an acceptable resultant

ontology (Stojanovic, 2004).

This approach has many disadvantages; it con-

sists of a solution which has a high algorithm com-

plexity due to multiple possible iterations without

any specific switch-off conditions. The reason being

that at each stage, choices have to be made using

relevant heuristics. Another approach in this catego-

ry is that of Delia (Delia, 2008). It consists of a semi

automatic system where the user is involved in the

ontology evolution process. It proposes versioning

of changes which then serve to resolve eventual

semantic referencing (based on the ontology) fail-

ures. The system records the changes during the

evolution phase of the ontology and provides the

user with a list of possible semantic references that

are impacted. The management of changes is hereby

concentrated on the control of the consistency of

semantic referencing (meaning the usage of the

target ontology – external integrity).

On the whole, the fore-mentioned approaches do

not solve the problem of evolution on all its dimen-

sions. Some are concentrates on structural aspects

(internal integrity) while others focus on the external

implications (external integrity). In addition, they

lack a formal framework which entirely integrates

the ontology evolution engineering.

3 THE PROPOSED APPROACH

FOR MANAGING THE

EVOLUTION OF ONTOLOGY

Despite the fact that several works have been carried

out on the evolution of ontologies, the tools which

should make it possible to ensure a guided evolution

which guarantees the consistency of final ontology

are still to come. The solution that we propose com-

prises four points: we start by taking into account

the semantics behind the various operations of

changes which can affect ontology during its life

cycle. Secondly, we set up some integrity constraints

which must be respected during any execution of an

operation or of a set of operations of change. Third-

ly, we control the execution of operations according

to some used case we proposed, and we end by giv-

ing a detailed report of the execution of the changes

with the new version of ontology.

3.1 Formal Description of Operators of

Change

These operators of change are formally defined

using Description Logic (DL). Evolution operations

are usually done on ontology elements including T-

Box axioms (concepts, roles, restrictions, attributes

etc.) as well as A-Box axioms (instances).

There are two groups of operators: Basic DL op-

erators such as negation, generalization, spe-

cialization, etc., and the other operators which we

defined using DL primitives (Stojanovic, 2004)

(Baader, 2007) (Gagnon, 2007). These operators

include adding, removing, merging or grouping

ontology elements (concept, role, restriction, etc.)

Each operator is semantically defined as illustrated

in the following for 3 instances of operators. Our

contribution here is that, actions related to each

operator are explicitly represented in the semantic.

The following examples illustrate some of these

operators and their formal semantic descriptions. Let

us consider (O, C, I, intscon,



) where O is an

ontology, C is a set of concepts, I is a set of instanc-

es, instcon is a function that relates a concept to the

set of its instances and 



is a relation called con-

cept hierarchy, we have:

 AddInstance (i,C):

MS-ONTO-ModelandSystemforSupportingOntologyEvolution

321

Add an instance to a concept 

Pré-conditions: ∈







∧

∃ ∈  ∖







, ∉









Post-conditions:  ∈ 







Actions :

1.  (ABox containing a statement C)

2. ( an instance different from )

3.  (Add an instance )

 AddConcept C: Add a concept C

Pré-conditions:



C, root



∉H



∗

∧∃D∈

Post-conditions:



C, root



∈H



∨



C, root



∈



∗

Actions:

1.  (TBox)

2. ⊑  is subsume by)

3. ⊑

⊺

 RemovConcept C: Remove a concept

Pré-condition: ∈∖







∧∃D





∈





,





∈



∧







,





∈





Post-condition:



,





∉



Actions:

1.  ( TBox containing concepts C



,…,C



C and D)

2. (∃1in) C



⊑C→C



⊑D

3. C⊑D

(C is subsume by D)

4. C⊑ (Remove C)

3.2 Constraints of Semantic Integrity

We also defined a set of constraints of semantic

integrity (CSI) which must be taken into account

throughout the evolution process. Each single opera-

tion of change should lead to the checking of the

CSI before the system

can approve or reject the

change. In all cases, a report must be drawn up.

An example of CSI could consist in setting a

maximal DL expressivity that should be preserved

after the evolution process. In this way, it should be

possible to preserve the logic of the initial ontology

if needed. In this case, taking into account the ex-

pressivity includes three main steps:

 Identifying the structures allowed for each sub-

language in a «database of structures of sub-

languages». Several level of language could be

obtained: For example OWL-LITE, OWL-DL,

OWL-FULL (Zghal, 2007) if we refer to OWL1.

 Creating an inspector of structure, a program that

is able to verify the ontology structures and to

categorize them, in order to determine the ontol-

ogy expressivity.

 Ensuring that the degree of expressivity of the

ontology is maintained.

The expressiveness management model de-

scribed above is presented in figure1. The initial

ontology undergoes an audit which determines its

expressiveness. Each operation of change which is

applied to the ontology must absolutely respect the

expressiveness. After the ontology expressiveness is

verified, a report is produced (Figure 1).

Figure 1: Model of management of the expressivity.

3.3 Execution of the Operations of

Change

The execution of an operation of change or a range

of change goes beyond the control of the expressivi-

ty of ontology. In fact, in addition to the CSI, each

operation has a set of pre-conditions and of post-

conditions to be verified before and after execution.

Our overall model of ontology evolution is pre-

sented in Figure 2.

The list of changes is a set of operations recorded

after evolution of the ontology or a set of operations

which one must apply to make the ontology evolve.

CSI is a set of constraints of semantic integrity

which must be verified to guarantee the consistency

of the ontology along its evolution. The list of the

operation of changes is drawn up by the operators of

DL on the one hand, and the operations which we

defined on the other hand as mentioned in the previ-

ous sections. The processor is a parser which carries

out the operations of change while respecting con-

straints. A report is produced at the end of the execu-

tion.

Initial Ontology

Process

Expressiveness

Ontology

Change

operato

Report

Check

KEOD2014-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

322

4 CURRENT IMPLEMENTATION

AND RESULTS

4.1 Implementation

MS-ONTO is implemented and current use-case

helps to control the ontology evolution giving as

inputs, a list of changes and the initial ontology on

which these changes are applied. The constraints of

semantics integrity (CSI) guaranteeing the consist-

ence of integrity all along the evolution and their

procedure of verification are also known to the sys-

tem.

A parser executes each change operation while

respecting predefined constraints. A report is pro-

duced at the end of the execution.

The operation class modelizes all the operations

which can be applied to the ontology. The diary of

changes contains the operations of change which the

ontology will undergo. The list of changes (changes

log) can be provided directly to the system or edited

by an Expert-user using an interface (Figure 4).

The

initial ontology which may eventually evolve is

contained in an OWL file.

The operations contained in the log are then applied

to the ontology. The integrity constraints are verified

and a detailed report is produced (Figure 5).

The overall processes and information sharing in

our system are depicted in Figure 3. Processing_01

helps the expert to edit or choose the change opera-

tors that will be applied to the ontology. Pro-

cessing_02 checks if each operator preserves the

expressiveness of the ontology. Processing_03 exe-

cutes the operations under the specified CSI. At the

end of all these processes, a report is produced con-

taining explanations about operation execution fail-

ures if any.

4.2 Validation

A very initial test of our system has been done using

a simple ontology on which we applied 6 change

operations including 2 that had to produce erroneous

results, which were successfully detected by the

system. However, a thorough validation is still to

come using more complex ontologies and meaning-

ful records of change operations.

Figure 3: Sequence diagram.

MS-ONTO-ModelandSystemforSupportingOntologyEvolution

323

Figure 4: The interface that allows the user to build up its journal of change.

Figure 5: Viewing the log of changes / Initial Ontology / Execution of operations of change.

KEOD2014-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

324

Figure 2: Overall model of MS-ONTO.

5 CONCLUSION AND FUTURE

WORKS

We have described a new system which makes it

possible to follow and control the evolution of on-

tologies. Our solution is founded on the formal

description of the semantic associated to each opera-

tion of change based on the Description Logics.

Each operation of change contains a set of pre-

conditions and a set of post-conditions to fulfill in

order to guarantee the consistency of the ontology

during the evolution. Furthermore, a set of con-

straints of semantic integrity was defined to better

ensure the consistency and the integrity of the final

ontology.

We intend to add to these constraints, a set of

metric of quality in order to measure the quality of

the final ontology. Also, many works remain to cope

with the external validation of the evolution which is

usage-context dependant. The key issue here is to

look for a possible generic core that can ease the

specification of context-dependant CSI related to the

ontology use in a specific domain. A first step could

consist in reconsidering Delia’s work on semantic

reference of learning resources using our own sys-

tem.

This work is still in its initial stage but we have

provided a framework where both internal and ex-

ternal validation of the evolution process can be

managed. This is a contribution as such integrated

solution does not exist. We intend to implement SM-

ONTO as a plug-in in order to ease its integration

with classic ontology engineering tools such as Pro-

tégé. We should also validate the system and collect

some data for its improvement. Finally, the current

implementation is used as a validation tool to ana-

lyze a list of changes operated on a given ontology

and provide feedbacks on that. Our next step here

will be to use the system in other use-cases where it

could act as an active coach during a live ontology

evolution (even building) process.

REFERENCES

Amal, Z., Nkambou R., 2008. Building Domain Ontolo-

gies from Text for Educational Purposes. IEEE Trans-

lations on Learning Technologies, vol. 1, N° 1, p.49-

62.

Baader, F., Calvanese, D., McGuinness, D., Nardi, D., and

Patel-Schneider, P., 2007. The Description Logic

Handbook: Theory, Implementation, and Applications.

Cambridge University Press, 622 pages.

Brewster, C., and O’Hara, K. 2007. Knowledge represen-

tation with ontologies: Present challenges-Future pos-

sibilities. ScienceDirect Int. J. Human-Computer Stud-

ies 65, p.563–568.

Booshehri, M., Malekpour, A., Luksch, P., Zamanifar, K.,

and Shariatmadari, S. 2013. Ontology Enrichment by

Extracting Hidden Assertional Knowledge from Text.

International Journal of Computer Science and Infor-

mation Security, vol. 11, No.5, p. 64-72.

Burcu, Y., 2006. Ontology Evolution and Versioning: The

state of art. Vienna University of Technologies, As-

gaard-TR-2006-3, 28 pages.

Cuenca, G., Jimenez-Ruiz, B., Kharlamov, E., Zheleznya-

kov, E., and Dmitry, 2012. Ontology Evolution under

Semantic Constraints. Proceedings of the 30th Inter-

national Conference on Principles of Knowledge Rep-

resentation and Reasoning, p.137-147.

Delia, R., 2008. Gestion de l’évolution d’une ontologie :

méthodes et outils pour un référencement sémantique

évolutif fondé sur une analyse des changements entre

versions de l’ontologie. Thèse de Doctorat, Université

du Québec à Montréal, p.68-180.

Djedidi, R., Aufaure, M., and Abboute, H., 2007. Evolu-

tion d’ontologie: Validation des changements basée

sur l’évaluation. Journées Francophones sur les Onto-

logies (JFO), p. 55-73.

Gabel, T., Sure, Y., and Voelker, J., 2004. KAON-

ontology management infrastructure. SEKT informal

deliverable, 58 pages.

Gagnon, M., 2007. Logique descriptive et OWL. Ecole

Polytechnique de Montréal, 55 pages.

Initial Ontology

CSI

Changes

records



Semantic

Parser



All

change



Report

Final Ontology

MS-ONTO-ModelandSystemforSupportingOntologyEvolution

325

Gruber, T., 1993. Toward Principles for the Design of

Ontologies Used for Knowledge Sharing. In Interna-

tional Journal Human-Computer Studies 43, p.907-

928.

Heflin, J., Hendler, J., and Luke, S., 1999. A knowledge

Representation Language for Internet Applications.

Technical Report, University of Maryland at College

Park, 30 pages.

Kondylakis, H., Plexousakis, D., 2011. Exelixis: Evolv-

ing ontology-based data integration system. Proceed-

ing of the international conference on Management of

data, p. 1283-1286, ACM.

Noy, N., Klein, M., 2004. Ontology Evolution: Not the

same as Schéma Evolution. In Knowledge and Infor-

mations Systems, vol.6 Issue 4, p.428-440.

Psyché, V., 2007. Rôle des ontologies en ingénierie des

EIAD: Cas d’un système d’assistance au design péda-

gogique. Thèse de Doctorat, Université du Québec à

Montréal, p.13.

Scott, B., Pete, B., Mark, B., David, K., and Debra, S.

2013. PRONTOE - A Case Study for Developing On-

tologies for Operations. In: 5th International Confer-

ence on Knowledge Engineering and Ontology Devel-

opment (KEOD). SciTePress, Vilamoura, Portugal, 9

pages.

Shaban-Nejad, A., 2010. A frame work for analyzing

changes in health care lexicons and nomenclatures.

PHD Thesis, Concordia University, Montreal, Que-

bec, Canada, p.86.

Stojanovic, L., 2004. Methods and Tools for Ontology

Evolution. Thèse de Doctorat, Karlsruhe, 249 pages

Sure, Y., Tempich, C., 2004. State of art in ontology engi-

neering methodologies. SEKT informal deliverable

7.1.2, Institut AIFB, University of Karlsruhe.

Wohlgenannt, G., Belk, S., and Schett, M., 2013. A Proto-

type for Automating Ontology Learning and Ontology

Evolution. In: 5th International Conference on

Knowledge Engineering and Ontology Development

(KEOD). SciTePress, Vilamoura, Portugal. p. 407-

412.

Zablith, F., 2009. Evolva : A Comprehensive Approach to

Ontology Evolution. Proceedings of the 6th European

Semantic Web Conference, LNCS 5554, eds. L. Aroyo

et al., Springer-Verlag, Berlin, Heidelberg, p. 944-

948.

Zghal, S., Yahia, B., Nguifo E., and Slimani. Y., 2007.

SODA : une approche structurelle pour l’alignement

d’ontologies OWL-DL. Actes 1ères journées franco-

phones sur les ontologies (JFO), p.1-20.

KEOD2014-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

326