Ontology-Based Database Approach for Handling Preferences - Data Warehousing Design and Advanced Engineering Applications - page 250

INTRODUCTION Within most existing information systems,

even if the notion of preference has been inte-
grated in various application domains, it is not
explicitly modelled. They are often hard encoded
and disseminated throughout the applications that
exploit these information systems. Therefore,
they can not be shared and must be defined and
updated for each application. This is a burden for
users and yields to another layer of heterogeneous
modelling.
To overcome these drawbacks, we propose
an ontology based approach of preference model
which relies on an ontology based database sys-
tem (OBDB), namely OntoDB (Dehainsala et al.,
2007; Pierra et al., 2005). The proposed model is
formally described using the EXPRESS model-
ling language and the approach takes benefits
from the OntoDB system which offers a flexible
mechanism for storing together an ontology, its
model and its instances. Nowadays, ontologies are
well accepted as formal knowledge organisation
systems which describe the explicit semantics of
entities manipulated in a given domain (Gruber,
2003). Domain ontologies are used to provide
definitions and specifications of these manipulated
entities. These entities are defined following the
ontology model.
Our contribution in this article is twofold.
Firstly, we propose a shared and generic model to
represent user preferences. Then, we describe how
preferences model can be attached to an ontology
and manipulated on the meta-model level. The
sharable preferences model has been formally
defined using the EXPRESS modelling language
in order to make its definition ambiguity free.
Indeed, EXPRESS is equipped with a powerful
constraints language allowing defining precisely
the semantics of the defined model.
The rest of the chapter is subdivided as follows.
The next section gives an overview of the prefer-
ences handling in the database, Semantic Web and
Data Warehouse areas. In addition, to make the
chapter self-explanatory, we give an overview of
the different ODBD existing approaches and we

The rapid growth and the wide adoption of internet
technology make available a huge amount of data
managed by various information systems. When
searching over these disseminated data, users are
often submerged by the numerous returned results
in response to their requests. These results must
often be sorted and filtered in order to identify the
relevant information. Despite the fact that the “one
size fits all” approach has shown its limitation in
many applications particularly in the e-commerce
domain, our targeted application domain, most
information systems do not take into account the
variety of users' need and preferences.
Capturing and exploiting user's preferences
have been proposed as a solution to this problem
in many domains including database systems
(Kießling and Kostler, 2000; Kießling, 2002;
Chomicki, 2003; Agrawal and Wimmers, 2000;
Koutrika and Ioannidis, 2004; Viappiani et al.,
2006; Das et al., 2006), Data Warehouse (Bel-
latreche et al., 2005), the Semantic Web (Siberski
et al., 2006; Gursk ý et al., 2008; Toninelli et al.,
2008 ), Information Retrieval (Daoud et al., 2007)
and Human Computer Interaction (Cherniack et al.,
2003). Although preferences are defined using an
ontology in some approaches, most of the previ-
ously cited work, and particularly in the Database
domain, the preferences and their model are defined
according to the logical model underlying the tar-
geted system. The use of the preferences requires
having knowledge of this logical model.
Preferences express the sense of wishes and
preference based search is a popular approach for
helping consumers to find relevant items. Users
would like to find the best matches between their
wishes and the reality. Modelling preferences is
difficult because human preferences are complex,
multiple, heterogeneous, changing, and even
contradictory. Moreover, they are complex to
evaluate and according to the user's goals and
his/her current task, they should be evaluated in
the context they have been expressed.