PATTERNS IN ONTOLOGY ENGINEERING:

CLASSIFICATION OF ONTOLOGY PATTERNS

Eva Blomqvist

School of Engineering, J

onk

oping University

P.O. Box 1026, SE-551 11 J

onk

oping, Sweden

Kurt Sandkuhl

School of Engineering, J

onk

oping University

P.O. Box 1026, SE-551 11 J

onk

oping, Sweden

Keywords:

Ontology Engineering, Ontology Patterns, Pattern Classiﬁcation.

Abstract:

In Software Engineering, patterns are an accepted way to facilitate and support reuse. This paper focuses on

patterns in the ﬁeld of Ontology Engineering and proposes a classiﬁcation scheme for ontology patterns. The

scheme divides ontology patterns into ﬁve levels: Application Patterns, Architecture Patterns, Design Patterns,

Semantic Patterns, and Syntactic Patterns. Semantic and Syntactic Patterns are quite well-researched but the

higher levels of pattern abstraction are so far almost unexplored. To illustrate the possibilities of patterns on

these levels some examples are discussed, together with ideas of future work.

1 INTRODUCTION

Recent developments in the area of Ontology En-

gineering involve semi-automatic ontology construc-

tion to reduce time and effort of constructing an on-

tology. In particular for small-scale application con-

texts, reduction of effort and expert involvement is an

important requirement.

Ways of reducing this effort are to further facil-

itate semi-automatic construction of ontologies, but

also to introduce reuse in Ontology Engineering. Pat-

terns have proved to be a fruitful way to handle the

problem of reuse. In Software Engineering it is al-

ready the commonly accepted way to build software

and there are not only Software Design Patterns, but

also Data Model Patterns, generic Problem Solving

Methods, Object System Models, and many other ap-

proaches. Classiﬁcation of ontology patterns is a ﬁrst

step towards a structured use of patterns in Ontology

Engineering. The main contribution of our research,

and focus of this paper, is the classiﬁcation and cate-

gorisation of patterns for Ontology Engineering.

In section 2 the background of the research is pre-

sented. Further, in section 3, our approach for classi-

fying ontology patterns is described and some exam-

ples of the different levels are discussed. In the ﬁnal

part, section 4, our research is summarised and some

ideas for future work are presented.

2 ONTOLOGY PATTERNS

TODAY

Ontology is a popular term today, used in many ar-

eas and deﬁned in many different ways. In this paper

ontology is deﬁned as:

An ontology is a hierarchically structured set of

concepts describing a speciﬁc domain of knowledge

that can be used to create a knowledge base. An on-

tology contains concepts, a subsumption hierarchy,

arbitrary relations between concepts, and axioms. It

may also contain other constraints and functions.

Even using this deﬁnition, ontologies can be used

for many different purposes, and they can be con-

structed and structured in many different ways. One

of the most common ways to describe the level of

generality of an ontology is by using the structure in

(Guarino, 1998). The structure shows how a general

Top-level Ontology can be specialised into a Domain

Ontology or a Task Ontology, which in turn can be

combined into an Application Ontology (describing a

certain task specialised for a speciﬁc domain).

Another categorisation of ontologies can be ob-

tained by classifying them by their intended use (van

Heijst et al., 1997). There are three main levels, Ter-

minological Ontologies, Information Ontologies and

Knowledge Modelling Ontologies, where each level

adds further complexity to the ontology structure. For

413

Blomqvist E. and Sandkuhl K. (2005).

PATTERNS IN ONTOLOGY ENGINEERING: CLASSIFICATION OF ONTOLOGY PATTERNS.

In Proceedings of the Seventh International Conference on Enterprise Information Systems, pages 413-416

DOI: 10.5220/0002518804130416

 SciTePress

example, in a Terminological Ontology often a simple

taxonomy of terms is enough, while in a Knowledge

Modelling Ontology more advanced axioms and con-

straints are very often required.

The ontology community has not yet adopted the

pattern idea on a broader scale. There exist a few pat-

terns for ontologies, and all of them are specialised for

some speciﬁc kind of ontology (with respect to both

level of generality and usage area, as deﬁned above).

For speciﬁc ontology languages, patterns are be-

ing developed, e.g. the OWL-patterns of (W3C-

SWBPD, 2004). Examples of meta-structures for on-

tologies, which can be denoted patterns, are Seman-

tic Patterns for describing implementation indepen-

dent logical constructs (Stuckenschmidt and Euzenat,

2001)(Staab et al., 2001). These patterns can for ex-

ample describe notions like locally inverse relations

and composition of relations. Similar to Software De-

sign Patterns are the Ontology Design Patterns dis-

cussed in (Gangemi, 2004), and the Design Patterns

developed for ontologies in Molecular Biology (Re-

ich, 1999). These patterns are quite general and de-

scribe for example how to keep track of updates be-

tween different partitions of an ontology.

3 CLASSIFICATION OF

ONTOLOGY PATTERNS

In order to study the use of ontology patterns in a

broader perspective, there is need for a general clas-

siﬁcation of patterns. In this section such a classiﬁca-

tion is presented and discussed.

3.1 Classiﬁcation

Our classiﬁcation of ontology patterns uses the fol-

lowing levels (an illustration of the levels and their

interdependencies can be seen in Figure 1):

• Application Patterns - Purpose, scope, usage and

context of the implemented ontology or ontologies,

including interfaces and relations to other systems.

• Architecture Patterns - A description of how to

combine or arrange implemented Design Patterns

in order to fulﬁll the overall goal of the ontology.

• Design Patterns - A small collection of Semantic

Patterns that together create a common and generic

construct for ontology development.

• Semantic Patterns - Language independent descrip-

tion of a certain concept, relation or axiom. A

meta-description of a Syntactic Pattern.

• Syntactic Patterns - Language speciﬁc ways to

arrange representation symbols, to create a certain

concept, relation or axiom.

To really beneﬁt from ontology patterns they

should be studied, and used, at all these levels. As

noted in section 2, there are some ontology pat-

tern approaches present today, but these are mostly

connected to the lower levels, such as Syntactic

Patterns (W3C-SWBPD, 2004), Semantic Patterns

(Staab et al., 2001) and Design Patterns (Gangemi,

2004) (Reich, 1999).

Ontology Application Patterns aim to describe

generic ways to use the implemented ontologies, in

terms of purpose, context, interfaces etc. This idea

of abstracting the best ways to apply and use an on-

tology, or several ontologies, within some context or

application is an important issue. Here many mod-

els and common practices exist, but not formalised as

patterns. A simple generalisation of existing models

could result in initial Application Patterns for ontolo-

gies. The application level patterns will be very much

dependent on the ontology usage area, and may also

be domain dependent.

Ontology Architecture Patterns aim to describe a

generic way to design the overall structure of an ontol-

ogy, in order to fulﬁll the goal of the ontology in ques-

tion. So far, no Architecture Patterns for ontologies

have been presented. Important questions are whether

to divide the ontology into components or modules, or

to divide it into layers or use other construction prin-

ciples. On the architecture level there could be great

possibilities to draw from experiences in other ar-

eas. Architectures tend to be similar whether it comes

to Problem Solving Methods, software or ontologies.

Architectures also do not tend to be application or do-

main speciﬁc, but the patterns created certainly need

to be adapted to either manual or automatic use.

Ontology Design Patterns aim to describe a

generic recurring construct in ontologies. When im-

plemented, the construct will form a small piece of

the complete ontology (perhaps a component or mod-

ule). There exist patterns that are generic and domain

independent, for manual construction of domain on-

tologies. Such patterns can be very useful but there is

also a need for more speciﬁc patterns, e.g. domain

dependent patterns, both for manual and automatic

use. The main idea of patterns for automatic use is

to be speciﬁc enough, allowing them to be used auto-

matically or semi-automatically, and at the same time

generic enough to be useful in several ontologies of a

certain domain.

A continuous application of all levels of the pattern

classiﬁcation above would require deﬁned patterns for

all different kinds of ontologies (both level of general-

ity and usage area as deﬁned in section 2). Yet another

aspect to be taken into account is whether the patterns

will be applied manually or automatically. In Ontol-

ogy Engineering today, the approaches tend more and

more toward semi-automatic construction, even more

so than in Software Engineering. Automatically ap-

ICEIS 2005 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

414

plied patterns will probably play a very important role

in the future. Thus, when exploring and developing a

full-covering system of patterns for ontologies, it is

important to consider both the ontology categorisa-

tion, the pattern categorisation and the notion of man-

ual or automatic pattern usage.

Application Level

Architecture Level

Design Level

Semantic Level

Syntactic Level

Figure 1: Illustration of the classiﬁcation levels as a

continuous increase of granularity.

3.2 Discussion of Selected Examples

Our research mainly has a focus on ontology patterns

for automatic use, and speciﬁcally for Information

Ontologies on the domain or application level within

enterprises.

One can imagine two different approaches when

addressing this problem. The ﬁrst one is to take all ex-

isting ontologies and derive patterns from them. The

other one is to develop criteria of ”good design” and

construct patterns that reﬂect these principles. Unfor-

tunately, the second approach is not practically feasi-

ble since it is almost impossible to derive criteria of

how to design ontologies for all situations. The ﬁrst

approach is the common way, in for example Soft-

ware Engineering, to ”discover” patterns using exist-

ing structures. Unfortunately, enterprise ontologies

are so far quite scarce and often sparsely documented,

which makes it hard to extract patterns this way.

One ”middle course” is to draw from knowledge

already accumulated in other areas. Many patterns in

Computer Science also describe some kind of knowl-

edge, although they might have a different purpose.

Our approach is to study those patterns in order to de-

velop ontology patterns. So far the focus has been

to develop an overall picture of the ontology area, by

developing the classiﬁcation structure discussed pre-

viously, and to put existing approaches into this cat-

egorisation. In the following sections ﬁrst ideas for

populating the higher pattern levels are presented.

3.2.1 Ontology Application Patterns

No Application Patterns have so far been formalised

for ontologies, as noted earlier, but there exist many

models of ontology usage. The models typically show

how different ontologies interact and depend on each

other in a speciﬁc system. As a starting point, these

types of models (for example like in (Levashova et al.,

2003)) could be abstracted into Ontology Application

Patterns.

3.2.2 Ontology Architecture Patterns

At the next level, some Software Architecture Patterns

might be quite easily adopted by the ontology com-

munity. There are several approaches to modulari-

sation of ontologies (Stuckenschmidt, 2003) (Rector,

2003), which in combination with a Software Archi-

tecture Pattern, for example similar patterns to those

developed by Shaw (Shaw, 1996), could become an

architecture for ontologies. An example could be

the Implicit Invocation Architecture Pattern (Shaw,

1996), which instead of objects or processes could be

used with ontology modules.

PATTERNS IN ONTOLOGY ENGINEERING: CLASSIFICATION OF ONTOLOGY PATTERNS

415

3.2.3 Ontology Design Patterns

The third level in focus is the Ontology Design Pat-

terns. Here patterns similar to Software Design Pat-

terns have already been suggested, so the interesting

areas to explore involve more domain and application

speciﬁc patterns for semi-automatic use. For testing

the suitability of different patterns in Computer Sci-

ence to act as templates when constructing Ontology

Design Patterns, some example ontologies were cre-

ated. These were evaluated against real-world appli-

cations in order to decide which patterns to focus on.

Some patterns considered interesting for constructing

Ontology Design Patterns are Data Model Patterns

(Hay, 1996) (Silverston, 2001), Object System Mod-

els (Sutcliffe, 2002) and Analysis Patterns (Fowler,

1997).

3.2.4 Remaining Issues

So far our focus has been on Ontology Design Pat-

terns in practice, and currently an evaluation is being

conducted to see how useful the ontologies suggested

above can be, and how they might be generalised into

ontology patterns. The evaluation has shown that one

problem is to transfer constraints in the existing pat-

terns into axioms of the ontologies in a structured

way. Another problem is the differences in general-

ity of the existing patterns. Also, the underlying as-

sumptions that the patterns are based on might differ

between the ontology ﬁeld and the ﬁeld where the pat-

tern originated.

4 SUMMARY AND FUTURE

WORK

The main contribution of this research, so far, is the

classiﬁcation of ontology patterns into ﬁve different

levels. When using these levels it is also impor-

tant to consider the kind of ontology to be built, its

intended usage, and especially how it will be con-

structed. The notion of semi-automatic construction

poses new and different requirements on possible pat-

terns, they need to be more speciﬁc, appropriately for-

malised and maybe even domain dependent.

Our work has so far focused on testing patterns

from other Computer Science areas in order to eval-

uate their usefulness when adapted to the ontology

area. When considering future work, it has been

shown that many areas of the ontology pattern ﬁeld

require more research efforts, in order to reach a full-

covering system of patterns. Our research will con-

tinue to focus on the patterns intended for automatic

use since they are the natural next step in further fa-

cilitating semi-automatic ontology construction.

REFERENCES

Fowler, M. (1997). Analysis Patterns - Reusable Object

Models. Addison-Wesley.

Gangemi, A. (2004). Some design patterns for do-

main ontology building and analysis. Available

at: http://www.loa-cnr.it/Tutorials/OntologyDesign

Patterns.zip, downloaded 2004-10-04.

Guarino, N. (1998). Formal Ontology and Information Sys-

tems. In Proceedings of FOIS’98, pages 3–15.

Hay, D. C. (1996). Data Model Patterns - Conventions of

Thought. Dorset House Publishing.

Levashova, T. V., Pashkin, M. P., Shilov, N. G., and

Smirnov, A. V. (2003). Ontology Management, II.

Journal of Computer and Systems Sciences Interna-

tional, 42(5):744–756.

Rector, A. (2003). Modularisation of Domain Ontologies

Implemented in Description Logics and related for-

malisms including OWL. In Proceedings of the in-

ternational conference on Knowledge capture, pages

121–128, Sanibel island. ACM Press.

Reich, J. R. (1999). Ontological Design Patterns for the

Integration of Molecular Biological Information. In

Proceedings of the German Conference on Bioinfor-

matics GCB’99, pages 156–166.

Shaw, M. (1996). Some Patterns for Software Architec-

tures. In Vlissides, J. M., coplien, J. O., and Kerth,

N. L., editors, Pattern Languages of Program Design,

volume 2, pages 255–269. Addison-Wesley.

Silverston, L. (2001). The Data Model Resource Book, Re-

vised Edition, volume 1. John Wiley & Sons.

Staab, S., Erdmann, M., and Maedche, A. (2001). Engineer-

ing Ontologies using Semantic Patterns. In O’Leary,

D. and Preece, A., editors, Proceedings of the IJCAI-

01 Workshop on E-business & The Intelligent Web,

Seattle.

Stuckenschmidt, H. (2003). Modularization of On-

tologies. WonderWeb Deliverable D21, available

at: http://wonderweb.semanticweb.org/deliverables

/D21.shtml.

Stuckenschmidt, H. and Euzenat, J. (2001). Ontology

Language Integration: A Constructive Approach. In

Proceedings of the Workshop on Application of De-

scription Logics at the Joint German and Austrian

Conference on AI, CEUR-Workshop Proceedings, vol-

ume 44.

Sutcliffe, A. (2002). The Domain Theory - Patterns for

Knowledge and Software Reuse. Lawrence Erlbaum

Associates.

van Heijst, G., Schreiber, A. T., and Wielinga, B. J. (1997).

Using explicit ontologies for KBS development. Inter-

national Journal of Human-Computer Studies, 46(2-

3):183–292.

W3C-SWBPD (2004). Semantic Web Best Practices

and Deployment Working Group. Available at:

http://www.w3.org/2001/sw/BestPractices/.

ICEIS 2005 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

416