Global Positioning System Reference
In-Depth Information
Determination of Relevant Links
At this stage, the problem is not the connection with the Data Base
Management System but the identifi cation and the choice of relevant
attributes among the database schema. They should satisfy the better
relevance criteria. The identifi cation is therefore equivalent to determine
the Relevant Links (RL). In that sense, we propose to classify the RL in
order to facilitate the search.
Several formalisms exist to handle a classifi cation, such as graph-based
or logic-based paradigms for example (Bedel et al. 2007). We propose to
use a lattice of concepts for the classifi cation of RL. Foerster et al. (2010)
and Messai et al. (2008) modify the lattice structure while analyzing end-
users' query. In contrast to this approach, we propose starting from the
data model of GetFeature or GetFeatureInfo data model to add or to delete
attributes in order to defi ne the data model associated with end-users'
queries. This approach is similar to a conventional viewed mechanism in
relational database.
Conventional applications in Information Systems have the knowledge
of the database schema (eventually with the formalism of views). In our
context, applications use a Web service that accesses to a distant database.
Users do not have an a-priori knowledge of the database schema nor the
database administrators have an a-priori knowledge of end-users' needs.
Therefore a view mechanism cannot be provided. Nevertheless, data base
administrators have a very good knowledge of the database schemas that
are offered to external queries.
We proposed the concept of Relevant Links that provides a semantic
relationship between database schemas and users. Starting with a user
query, we propose to offer a semi-automatic design of a result database
schema. The Data Base Administrator is no longer in charge of providing
static external schemas (i.e., views in relational databases) for the set of
queries. A middleware on the clients' computers is in charge of providing
a dynamic “view”. This operation is transparent to end-users. The “view”
is relevant to an end-user's point of interest and is designed according to
the operational complexity (s)he is willing to accept.
This section is organized into four parts. In the fi rst part, we present
the formal defi nition of lattice of concepts. In the second part, we present
the links of this lattice and the available RL. In the third part, we present
the extended formal defi nition of a RL. Then, in the fourth part, we present
the determination of RL.
Lattice Formalization
We propose a classifi cation built over a partial ordered structure named
lattice of concepts (Wille 1982; Nauer and Toussaint 2009). This structure
Search WWH ::




Custom Search