Representing Social Reality in OWL 2?
Rinke Hoekstra1,2
1 Department of Computer Science, Faculty of Exact Sciences, VU University
Amsterdam
hoekstra@few.vu.nl
2 Leibniz Center for Law, Faculty of Law, University of Amsterdam
hoekstra@uva.nl
Abstract This paper introduces a design pattern that allows for the
OWL 2 DL representation of concepts central to social reality: roles.
The work presented here is motivated by experiences in the development
of the LKIF Core ontology of basic legal concepts [10,8]. This paper
applies modelling steps identified in earlier work for the representation
of transactions [9] to the domain of roles. This is done by building on
Searle’s theory of social reality [14]. We use the new features of OWL 2
to approximate a reified relation, and show how the approach of [9] can
be reused to define a pattern for capturing roles, intentional concepts
and n-ary relations [13].
1 Introduction
This paper introduces a design pattern that allows for the OWL 2 DL repres-
entation of relational roles in social reality, and n-ary relations in general. We
apply the pattern to two use cases: the representation of n-ary relations by the
SWBP of the W3C3 [13] and a description of social reality [14].
The work presented here is motivated by experiences in the development of
the LKIF Core ontology of basic legal concepts [10,8]. To be able to adequately
capture the contents of legal norms [17], this ontology had to be equipped with
a large number of concepts for describing social reality: roles, beliefs, desires,
obligations, permissions, intentions etc. The problem of representing roles such
as ‘student’ has been discussed at length in the literature [15,12]: they can be
seen both as a class (‘Jane is a Student’) and as a relation (e.g. the student
relation between Jane and her university). The same duality holds for intentional
categories such as beliefs and desires. While a class conveys a stronger ontological
commitment than the relation, the relation is often more convenient and succinct
for practical use.
The representational construct in existing Semantic Web languages that
comes closest to capturing this duality is reification. Reification allows one to
address the relation between two or more resources (e.g. an RDF triple) as a
? This paper is a significantly revised version of a section in [8]
3 SWBP: Semantic Web Best Practices and Deployment group. See http://www.w3.
org/2001/sw/BestPractices.

primary entity (a resource). It is a sensitive subject, and is often associated
with messy modelling and has complicated semantics, especially when different
levels of reification are not stratified. On the other hand, reification is becom-
ing increasingly important as a candidate approach to attaching metadata to
individual triples, e.g. for the purposes of provenance tracking.
Unfortunately, the built-in reification capabilities of RDF are not sufficient
to our purposes. Namely, the existence of a reification (an RDF statement) does
not entail the existence of the corresponding relation [6]. Furthermore, OWL 2
DL [2] is not expressive enough to represent reified relations either, even though
these form a core part of use cases in social reality.
Current Semantic Web languages fall short in another, related respect: they
do not provide means to express n-ary relations. This issue was addressed by
the SWBP in the identification of a (famous) design pattern that represents the
n-ary relation as a class with multiple properties relating it to the relata of the
n-ary relation [13]. Unavoidably, this pattern has an effect similar to that of
RDF reification: the original relation between the relata is lost.
This paper applies modelling steps identified in earlier work for the repres-
entation of transactions [9] to the domain of roles. This is done by building on
Searle’s theory of social reality [14]. We use the complex role inclusion axioms of
OWL 2 to approximate a reified relation. We show how the new features of OWL
2 can be used to define a pattern for capturing intentional concepts, speech acts,
and the n-ary relations examples of the SWBP document [13].
Section 2 describes the problems of representing social roles and n-ary re-
lations in more detail. section 3 applies the steps of [9] to construct the design
pattern, and subsection 3.1 applies this pattern to the use cases described above.
2 The Problem
The LKIF Core ontology distinguishes three levels: physical reality, an inten-
tional level and a legal level. Each consecutive level builds on and expands the
level below it with more descriptive power. The distinction between these levels
is inspired by Valente’s functional ontology, where legal knowledge is considered
to be an abstraction of commonsense knowledge [16].
Categories described at the intentional level and above, are social constructs
that can be attributed to, or imposed on brute facts [14]. Brute facts are phe-
nomena of which the existence does not depend on human agreement. This is
similar to the way in which intentional and functional notions are generalised
over physical phenomena in the design and intentional stance of Dennett [1].
According to Searle, institutional facts are constructed by means of constitutive
and regulative rules. These are rules of the form:
X counts as Y in context C
Examples of constitutive and regulative rules are, respectively:
– Bills issued by the Bureau of Engraving and Printing (X) count as money
(Y) in the United States (C).