Ontology-Based Integration of Heterogeneous, Incomplete and Imprecise Data Dedicated to a Decision Support System for Food Safety - Data Warehousing Design and Advanced Engineering Applications - page 90

Figure 7. The SPARQL query associated with a view

his/her selection criteria. These preferences are
expressed by fuzzy sets as presented in Subsec-
tion “The fuzzy set theory”. Since fuzzy sets are
not supported in a standard SPARQL query, we
propose to “defuzzify” the MIEL++ query before
translating it into SPARQL. This defuzzification
is performed in two steps.
When the fuzzy value of a selection criterion
has a hierarchized symbolic definition domain, it
is represented by a fuzzy set defined on a subset
of its definition domain. Such a fuzzy set defines
degrees implicitly on the whole definition domain
of the selection attribute. In order to take those
implicit degrees into account, we propose to
perform a closure of the fuzzy set as defined in
Thomopoulos & al. (2006). Intuitively, the closure
propagates the degrees to more specific values of
the hierarchy. Then, for each selection criterion
represented by a fuzzy set, we can perform the
defuzzification of the fuzzy set which consists in
deleting the degrees in the case of a DFS and in
only keeping the interval which corresponds to
the support in the case of a CFS.

Example 5
Let us consider the following MIEL++ query Q
where FoodProduct, Contaminant, Contamina-
tionLevel are the projection attributes and where
FoodProduct and ContaminationLevel are the
selection attributes. Figure 8 presents (i) on the
left, the closure and the defuzzification of the
fuzzy value FoodProduct_Preferences={1.0/rice +
0.5/cereal} associated with the selection criterion
FoodProduct according to the type hierarchy of the
type FoodProduct of Figure 2 and (ii) on the right,
the defuzzification of the fuzzy value Conta_Pref-
erences={0.5, 0.75, 1.7, 1.8} associated with the
selection criterion ContaminationLevel.
The defuzzified MIEL++ query can now be
translated into a SPARQL query where the CON-
STRUCT clause allows the answers of the query
to be built according to the projection attributes
of the MIEL++ query and the SELECT clause
contains the selection criteria and the threshold
of the MIEL++ query. All the selection criteria
are represented into the FILTER clause of the
SELECT clause.