Thesis Title:
Modular Ontologies for Spatial Information
Summary:
This thesis investigates the development
and application of spatial ontologies.
It also analyzes methodologies and techniques necessary to model
spatial ontologies and to apply them in di erent application scenarios.
First, we therefore analyze the variety
and diversity of spatial information.
For instance, it has not been analyzed in detail, which types and conceptualizations
are available to model spatial information. In particular, we
distinguish di erent types of spatial information by specifying their thematically
di erent perspectives on space. Based on these perspectives, spatial
information that is characterized by quantitative, qualitative, abstract,
domain-speci c, or multimodal aspects is distinguished. These di erent
perspective guide the development of respective computational models, i.e.,
ontology modules for spatial information. Ontologies are logic-based speci -
cations that categorize and classify semantic types and relations. To model
the spatial ontologies, we technically use OWL and Protege as well as CASL
and HeTS.
Second, we investigate how the developed
ontologies can be compared and
combined, technically as well as content-wise. Here, ontological modularity
is essential to provide combination methods for the di erent spatial ontologies.
We discuss and apply primarily the methods extension/re nement,
matching, and connection to combine ontology module that comply with
di erent spatial perspectives. The use of these combination methods between
ontologies are determined by means of their technical practicability
and thematic adequacy. The combination of spatial ontology modules is
then applied in di erent application scenarios.
Third, we integrate uncertain information
with the developed ontology modules.
As spatial applications have to cope with partially vague, uncertain,
or ambiguous types of information, we provide a method to take these aspects
into account in the ontological modeling of spatial information. No
standards are available today that support uncertain types of information
together with logically strict ontological formalizations. Hence, we also
examine technical and thematic approaches to combine uncertainties and
ontologies and to use them in particular application scenarios.
Finally, the developed spatial ontology
modules, their combinations, and
their uncertainty components are applied and evaluated within three applications:
in the area of architectural design and assisted living systems,
spatial ontology modules are applied to provide automated assistance and
to analyze design requirements; in the area of visual recognition, spatial ontology
modules are used to support object and scene classi cations; in the
area of natural language processing, spatial ontology modules are utilized
to interpret natural language in terms of qualitative spatial representations.
In conclusion, this thesis results in spatial ontology modules that are structured
by means of their thematically distinct perspectives on space. Ontological
modularity and uncertainties guide and support combinations and
extensions of the developed spatial ontologies. The modules' applicability
and adequacy are evaluated individually as well as within di erent spatial
application scenarios.