Architecture for Including Personalization in a Mobile GIS via Semantic Web Techniques - Geographical Information Systems: Trends and Technologies - page 227

Spatial RDF Data in Smartphone
This subsection proposes a framework to store and manage RDF data in
smartphones and to use spatial reasoning to create a system that performs
simple geographic operations over spatial RDF data.
In order to perform geographical queries over ontology data, first of
all we import the GeospatialWeb library into Android OS and we integrate
spatial predicates in the Androjena framework. Therefore, we create a
system that contains Androjena plus ARQoid and TDBoid components
and GeospatialWeb library. The system has been deployed into a prototype
system for an Android smartphone.
Feasibility tests
To test whether the architecture is feasible, we implemented three different
GeoSPARQL queries, the most common in a tourism application. The first
one selects all the points of interest inside a given area ( within function),
the second one selects the nearest points from the user position ( nearby
function), and the third one selects the distance between two points of
interest ( distance function).
Figure 7 shows that it is feasible to integrate a GIS with spatial
operations over RDF data. It shows the result of within query over the
ontology: the points of interest within the region of La Sagrada Familia (a
POI in Barcelona). The query returns all the POIs of LinkedGeoData from
this area, which are displayed over a map.
Effi ciency tests
The next step is to check whether such integration is effi cient enough and
scalable to be used in end-user applications. Therefore, we performed some
queries with the GeoSPARQL within function done over several geographic
ontologies, in order to evaluate the scalability of results. Additionally, we
also perform tests with simple SPARQL queries, without spatial reasoning,
in order to compare the results.
The smartphone used for the tests is an HTC Desire, which has Android
2.1 operative system, 1.0 GHz processor and 576 GB of RAM. Each test is
repeated five times, in order to have an average of results, because they
may be variable.
Results of SPARQL queries are displayed in Table 1 and results of
GeoSPARQL queries are displayed in Table 2. Note that ontologies in Table
1 and Table 2 have different number of triplets, since it is not possible to
use the same ontologies to query spatial and non-spatial data.