Ontology-Based Database Approach for Handling Preferences - Data Warehousing Design and Advanced Engineering Applications - page 254

ogy is used to define preferences, the approaches
are static and not enough flexible and generic to
handle different preference models and are hardly
adaptable in other contexts. This is because these
approaches handle mostly preferences at ontol-
ogy's instances level. For the storage issue, any of
the approaches provide a real storage possibility
of preferences model. Some of them hard code
the models in the application (e.g. (Siberski et
al, 2006)). None of the previously mentioned
approaches offer simultaneously the possibility
to define preferences at the semantic level, allow
their persistence and provide a dedicated language
for querying with preferences facilities.
To overcome these drawbacks, we propose
an approach of preferences management by
introducing a data model which is abstracted
from the concerned database logical model.
Indeed, our approach is based on an Ontology
Based Database (OBDB) system which offers a
persistence mechanism allowing the storage of
both preferences and the actual data in the same
infrastructure. The OBDB approach offers also
the possibility to manage the ontology which de-
scribes the content of the database. Our approach
acts at the semantic level by dynamically linking
the preference model to the ontology describing
the meaning of the semantic data.

the instances descriptions. However, this approach
raises serious performance issues when queries
require many self-joins over this table.
Double schema approach. In this approach,
ontology descriptions and instance data are stored
separately in two different schemas (Alexaki et
al., 2001; Broekstra et al., 2002; Ma et al., 2004).
The schema for ontology descriptions depends on
the ontology used model (e.g., RDFS (Lassila,
1999), OWL (McGuinness et al., 2004), PLIB
(IS010303.02, 1994). It is composed of a set
of tables used to store each ontology modelling
primitive like classes, properties and subsumption
relationships. For the instances data, different
schemas have been proposed; i) a vertical table
to store instance data as triples (Ma et al., 2004;
Broekstra et al., 2002); ii) a binary representation
where each class is represented by an unary table
and each property by a binary table (Alexaki et
al., 2001; Broekstra et al., 2002; Pan and Heflin,
2003; Abadi et al., 2007); iii) finally, a table per
class representations (also known as class-based
representations) has been proposed where a table
having a column for each property associated
with value for at least one instance of a class is
associated to each class (Dehainsala et al., 2007;
Park et al., 2007). These three basic approaches
have also some small variants. For more detail,
the user may refer to (Theoharis et al., 2005).
Separating representation of ontology descriptions
and instance data leads to better performance and
query response time. However, this approach as-
sumes a fixed ontology model.
The OntoDB approach . OntoDB (Dehainsala
et al., 2007; Pierra et al., 2005) proposes to add
another schema to the previous approach. This
schema, called meta-schema, records the ontol-
ogy model into a reflexive meta-model. For the
ontology schema, the meta-schema plays the same
role as the one played by the system catalog in
traditional databases. Indeed, the meta schema
supports: (1) a generic access to the ontology,
(2) the evolution of the used ontology model,
and (3) the storage of different ontology models

Ontology Based Database (OBDB)

Over the last years, many OBDB architectures
have been proposed in the literature. They can be
classified in 3 categories according to the schemas
they handle.
Single triple-based table approach. In this
category, information is represented in a single
schema composed of a unique triple table (sub-
ject, predicate, object) (Harris and Gibbins, 2003;
Chong et al., 2005; Petrini and Risch, 2007;
Alexaki et al., 2001; Broekstra et al., 2002; Ma
et al., 2004; Dehainsala et al., 2007; Pierra et al.,
2005). This table, called vertical table (Agrawal
et al., 2001) can be used both for the ontology and