Global Positioning System Reference
In-Depth Information
The main drawback we found is the ineffi ciency of SPARQL and
GeoSPARQL queries. Therefore, in order to improve query time, we rewrite
the conceptual model of the ontology to a relational database schema. In
consequence, the instances of the ontology are inserted into the database
as relational data. Then, we could use SQL to perform queries, which is far
more effi cient than SPARQL queries.
To create the database from the ontology, we created the ontologies in
OWL format with triplets that represent the knowledge of each domain.
Then, we imported the ontology model into a relational database schema.
We tried to minimize the loss of semantics with a translation of the ontology
model to a relational database: each class of the ontology becomes a table in
the relational database. Therefore, the instances of the ontology are imported
into the tables as relational data.
In this way, OWL reasoners cannot be used in the system because
we implemented the ontology using a relational database representation.
Therefore, the inference capabilities of reasoners cannot be used in the
prototype. However, the system still has a clear and unambiguous defi nition
of the touristic and geographic information and supports inference rules
that allow taking profi t of such information. In particular, the system has
an ontology that stores the past trips of users (see Fig. 6) and some rules
use this information to adjust the personalization values of each user
according to her/his history, like tuning the average visit time of a point of
interest according to the visit time of similar points of interest in the user
past visits. Thus, even though moving the ontology to a relational model
avoids using inference engines, the semantics of the concepts are kept and
some basic inferences are still possible. In particular, most of the integrity
constraints and derivation rules of the ontology have been kept after a direct
translation of the RDF ontologies to the relational model, minimizing the
loss of semantics. Note that the fact of not using reasoners is not only due
to limitations of OWL/RDF frameworks in smartphones, but also due to
technical limitations that make the use of these reasoners impracticable for
real life smartphone applications.
Furthermore, it is important to take into account that personalization
is one of the requirements to achieve. Therefore, we must fi lter data and
suggest the best information for each different user. In order to do so, we
design a personalization algorithm over the database.
Personalization Algorithm
Section “Spatial RDF Data in Smartphone” showed that it is ineffi cient to
work with RDF data in a smartphone locally and proposed to work with
ontology data exported to a relational database. Thus, to get personalization
we designed a personalization algorithm with different rules that selects
Search WWH ::
Custom Search