Geoscience Reference
In-Depth Information
Circle (Chesneau, 2006) and Derain Algorithm (Derain, 1905), Domingues
et al. had implemented ontology of colors for this purpose.
(James et al. 2010) had proposed directions for the application of ontology
matching techniques to solve different interoperability issues in the area of
image annotation and retrieval so we can replace 'image' with 'icon or tex-
ture' for example and test the feasibility of their framework. In the context
of semantic image annotation, ImageNet and LSCOM are two examples of
multimedia ontologies where the concepts are the nodes of the WordNet
ontology and the instances are the images, or the visual attributes in our
case, in the associated databases labeled by these concepts.
We can deduce that ontologies are convenient to represent visual knowl-
edge or map legend but we should bridge the semantic gap problem be-
tween the semantic level and the visual level representations. This can be
solved by 1) matching ontologies at the semantic level with ontologies at
the visual level, and 2) matching multiple visual ontologies in order to ex-
tract a common visual model for linguistic descriptions of images or icons.
Besides, we can apply “variable selection techniques” in machine learning
that can serve to rank the input variables (for example, the different icons
for same POI service) by their importance for the output visualization, ac-
cording to user's evaluation criteria, his context/profile and other semiol-
ogy constraints. Belief weights could be applied within OWL file (Belie-
fOWL) for each symbol to ensure ranking as well.
We decided then to develop a new type of geographic ontology framework
to build and match semantic and visual aspects of the providers' legends
towards a domain reference one for LBS. Because we are dealing with on-
tologies, we will use the tags 'properties' of OWL (Web Ontology Lan-
guage) standard to include the visual attributes of each POI concept such
as his icon, color, texture, font, number, etc. instead of normal XML tags
as per SE. Other interpretation was to extend OWL with a new tag called
Symbol in order to code the visual aspects for each concept as detailed be-
low. This later is implemented in the platform.
By respecting the proposed paradigm: “the visual ontology of concepts”,
each provider shall have its own local ontology that should be populated
via a Graphical user interface (GUI) implemented for this purpose. This is
done by inserting manually their semantic and visual service attributes
based on their legends, semantic concepts for the name of POIs and aside
the visual aspects of the correspondent symbol.
