How to Reuse a Faceted Classification and Put
It on the Semantic Web
Bene Rodriguez-Castro, Hugh Glaser, and Leslie Carr
School of Electronics and Computer Science, University of Southampton,
Southampton SO17 1BJ, UK
{b.rodriguez,hg,lac}@ecs.soton.ac.uk
http://www.ecs.soton.ac.uk
Abstract. There are ontology domain concepts that can be represented
according to multiple alternative classification criteria. Current ontology
modeling guidelines do not explicitly consider this aspect in the rep-
resentation of such concepts. To assist with this issue, we examined a
domain-specific simplified model for facet analysis used in Library Sci-
ence. This model produces a Faceted Classification Scheme (FCS) which
accounts for the multiple alternative classification criteria of the domain
concept under scrutiny. A comparative analysis between a FCS and the
Normalisation Ontology Design Pattern (ODP) indicates the existence
of key similarities between the elements in the generic structure of both
knowledge representation models. As a result, a mapping is identified
that allows to transform a FCS into an OWL DL ontology applying the
Normalisation ODP. Our contribution is illustrated with an existing FCS
example in the domain of “Dishwashing Detergent” that benefits from
the outcome of this study.
Keywords: facet analysis, faceted classification, normalisation, ontol-
ogy design pattern, ontology modeling.
1 Introduction
Ontologies remain as one of the key components needed for the realization of
the Semantic Web vision. They bring with them a broad range of development
activities that can be grouped into what it is referred to as Ontology Engineer-
ing. Ontology Engineering for the Semantic Web is a very active research area
and has experienced remarkable advancements in recent years, although it is
still relatively new compared to other engineering practices within Computer
Science or other fields. A constant ongoing effort in Ontology Engineering deals
with harnessing the field with sound development methodologies analogous to
those successfully employed in Software Engineering for decades. One of the ob-
jectives of these methodologies is to address areas of the ontology development
process vulnerable to ad-hoc practices that could potentially lead to unexpected
or undesirable results in ontology artifacts.
This paper describes a specific, very recurrentmodeling scenario in ontology de-
velopment, subject to such vulnerability. The scenario consists of domain-specific
P.F. Patel-Schneider et al. (Eds.): ISWC 2010, Part I, LNCS 6496, pp. 663–678, 2010.
c© Springer-Verlag Berlin Heidelberg 2010

664 B. Rodriguez-Castro, H. Glaser, and L. Carr
concepts that can be represented according to multiple alternative classification
criteria. To the best of our knowledge, guidelines for the conceptualization and
representation of domain-specific concepts prone to be described based on multi-
ple (potentially alternative) classification criteria, has not been explicitly consid-
ered in the context of ontology modeling for the Semantic Web.
General examples of domain-specific concepts that exhibit the characteris-
tics described abound, going from a “bibliographic reference”, (which could be
classified according to several criteria such as “subject”, “author”, “publication
venue”, etc.); to a “toy” (which could be classified based on “suitable age”,
“brand”, “subject type”, etc.). The list of examples can go on. We have seen
in our own experience that lack of specific design guidelines leaves ample room
for conceptual errors when trying to develop a simple domain-specific ontology
model for such concepts. For example, common mistakes when trying to rep-
resent these concepts and their classification criteria are to use subsumption
relations between classes when in fact a part-of relationwouldbeinorder,orto
use subsumption to model relationships that are outside OWL DL expressivity
altogether.
Other examples of domain-specific concepts that can fit into the modeling
scenario described are particularly interesting because they are used in well-
known ontology development literature using OWL. They include: “Wine” [1],
“Person” (in the context of family history relations) [2], or “Pizza” [3]. However,
in none of them, they refer explicitly to the various classification criteria of the
domain concept that are considered implicitly, nor attempt to represent these
criteria explicitly in the respective ontology models developed.
To assist with these issues, we aim to put forward an initial set of basic
design guidelines to mitigate the opportunity for ad-hoc modeling decisions in
the development of ontologies for the problem scenario described. To obtain
the conceptual model of a domain-specific concept and its multiple classifica-
tion criteria we examined a simplified model for facet analysis in the field of
Library and Information Science [4]. The outcome of this facet analysis is a
Faceted Classification Scheme (FCS) for the domain concept in question where
in most cases a facet wouldcorrespondtoaclassification criterion.Toobtainan
ontology representation of the FCS, we examined the Normalisation Ontology
Design Pattern (ODP) [5] [6] [7]. A comparative analysis between a FCS and
the Normalisation ODP revealed the existence of key similarities between both
knowledge representation paradigms. The similarities allowed us to identify a
series of mappings to transform a FCS into an OWL ontology applying the Nor-
malisation pattern. Moreover, the ontology model obtained through this process
contains a valid OWL DL representation of the classification criteria involved in
the characterization of the domain concept.
To illustrate our contribution, we used throughout the document an existing
FCS example in the domain of “Dishwashing Detergent” [8]. In fact, there are
aspects of the work presented in this paper that could be viewed as a follow-up
to [8] in the context of the Semantic Web and we attempted to acknowledge that
in our title.