Database Reference
In-Depth Information
RDF data warehouse. Then, the answer to this
SPARL query must be “refuzzified” in order to
be able to measure how it satisfies the selection
criteria.
To measure the satisfaction of a selection
criterium, we have to consider the two semantics
-imprecision and similarity- associated with fuzzy
values of the XML/RDF data warehouse. On
the one hand, two classical measures (Dubois &
Prade, 1988) have been proposed to compare a
fuzzy set representing preferences to a fuzzy set
having a semantic of imprecision: a possibility
degree of matching denoted ∏ and a necessity
degree of matching denoted N. On the other
hand, we propose to use the adequation degree
as proposed in (Baziz & al., 2006) to compare a
fuzzy set representing preferences to a fuzzy set
having a semantic of similarity. The comparison
results of fuzzy sets having the same semantic
(similarity or imprecision) are aggregated using
the min operator (which is classically used to
interpret the conjunction).
Therefore, an answer is a set of tuples com-
posed of the pertinence score
ps
associated with
the relation, three comparison scores associated
with the selection criteria in the data warehouse:
a global adequation score
ad
associated with the
comparison results having a semantic of similarity
and two global matching scores
∏
and
N
associ-
ated with the comparison results having a semantic
of imprecision, and, the values associated with
each projection attribute. Based on those scores,
we propose to define a total order on the answers
which gives greater importance to the most perti-
nent answers compared with the ontology. Thus,
the answers are successively sorted according to
firstly ps, then
ad
and thirdly a total order defined
on
∏
and
N
where
N
is considered as of greater
importance than
∏
.
in Figure 4 is given below:Result = {ps= 0.75,
ad=0.66, N= 0.0,
∏
= 1.0, FoodProduct=(0.66/
Rice + 0.5/Rice Flour), ContaminationLev-
el=[1.65,1.65,1.95,1.95]}.
The Ontology Alignment
As already said, the CARAT system is composed
of contamination data indexed by the CONTA
ontology and consumption data indexed by the
CONSO ontology. Both types of data concern
food products: the contamination data are mea-
sures of chemical contamination for food products
and the consumption data are about household
purchases of food products. Therefore, the deci-
sion support system of the CARAT system needs
correspondences to be found between food prod-
ucts of the CONTA ontology and food products
of the CONSO in order to estimate the exposure
of a given population of consumers to chemical
contaminants.
Since the CONSO ontology is updated every
year by the company which provides the TNS
WORLD PANEL data and, on the contrary, the
CONTA ontology remains stable, the CONSO
ontology is considered as the source ontology in
the alignment process and the CONTA ontology
as the target ontology. A simple mapping between
food product names of the CONSO ontology and
food product names of the CONTA ontology is
not efficient because only a little set of names
have words in common. Therefore we have used
an additional knowledge to map food products:
the food product description available in both
ontologies. For this purpose, the content of the
CONTA ontology presented in Subsection “The
structure of the CONTA ontology” is extended
with an international food description vocabulary
called Langual (Ireland & Moller, 2000). Langual
is composed of predefined characteristics and
of predefined associated values partially order
by the subsumption relation. Figure 10 gives an
excerpt of the extended version of the CONTA
ontology expressed in RDFS: the symbolic type
Example 7
The answer to the query of Example 6 com-
pared with the first row of the table presented in
Table 1Table 1 A Web data table and annotated
Search WWH ::
Custom Search