Database Reference
In-Depth Information
explain why we have chosen OntoDB. Moreover,
we describe briefly the EXPRESS data modelling
language which is used to formally define our
model. The different models that are manipulated
in our approach are described in this language.
We detail afterward our approach of preferences
model and formally describe its three components
using EXPRESS. The integration of the preference
model into the OntoDB system is presented in this
section as well. Before concluding and outlining
some future directions of this research work, we
illustrate the approach of preferences model by an
implementation within the tourism domain.
to the type of used metric, two different ways
of expressing preferences have been proposed:
qualitative
and
quantitative
approaches.
Qualitative approaches (Chomicki, 2003;
Kießling, 2002) allow users to define (relative)
preferences between tuples. The preferences are
defined on the content and define a binary relation
between tuples (Chomicki, 2003). For example, if
we consider two tuples
t
1
and
t
2
, the expression
t
1
>
t
2
means that the user prefers the tuple
t
1
rather
than
t
2
. Kießling and Kostler follow a qualitative
approach as well, named
constructor
approach.
The preferences are expressed by a strict partial
order and are formally described by first order
logical formulas (Kießling, 2002). The defined
constructors are integrated within the Preference
SQL relational language (Kießling, and Kostler
2000). For instance, the constructor
Highest
(
c
)
is used to express that for 2 tuples
t
1
and
t
2
, we
prefer the tuple having the higher value for the
column
c
. This approach is referred as the BMO
(
Best Match Only
) query model and is identical
to the
winnow
operator defined by Chomicki
(Chomicki, 2003).
Quantitative approaches in the other hand
allow users to define scoring functions to com-
pute a numeric score or an absolute preference
for each tuple (Agrawal and Wimmers, 2000;
Koutrika and Ioannidis, 2004; Das et al., 2006).
The results are sorted according to this score.
In this context, Agrawal and Wimmers define
preferences by introducing a preferred value for
each column in the database's tables (Agrawal
and Wimmers, 2000). For instance, let us con-
sider the table Hotel defined as
Hotel
(
name
,
priceMin
,
priceMax
). The preference <
*,40,80
> indicates that preferred hotels are those having
room price between 40 and 80. This preference
is then used to compute a score between 0 and 1
for each hotel. Koutrika and Ioannidis introduce
the notion of
atomic preferences
by specifying a
set of pair <
condition
,
score
> where
condition
is a condition on the values of columns and
score
is the degree of interest between 0 and 1 of this
BACKGROUND
This section addresses the state of the art in two
directions followed by this article. On the one hand,
we overview the preference modelling approaches
in the areas of databases and of Semantic Web,
and on the second hand the database system that
embed ontologies and their instances. Our goal is
to study the existing approaches in order to propose
a preference model as generic as possible.
Related Work on Preferences Models
Handling preferences has been addressed in
various information systems research areas. The
related work described in this section is mainly
concentrated on two major subareas: Databases,
Semantic Web and data warehouse domains.
Preferences in Databases
Handling preferences in the database domain has
been addressed in many research work (Kießling
and Kostler, 2000; Kießling, 2002; Chomicki,
2003;Agrawal and Wimmers, 2000; Koutrika and
Ioannidis, 2004; Viappiani et al., 2006; Das et al.,
2006). Preferences in this context are defined on
the logical model level of the database, specifically
on the column values of the tables. According
Search WWH ::
Custom Search