Ontology-Based Discovering of Geographic Databases Content - Advances in Cartography and GIScience - page 317

Specifications are the most detailed available source of knowledge about
geographic databases content. They describe the meaning of each element
of a database schema or, i.e. the semantics of a schema. However, as they
are mainly written for human readers, these specifications are only
available in natural language. So, they are not directly tractable. Providing
a formal representation of these specifications would enable to take
advantage of the knowledge they contain for a better comprehension of
databases contents and for data integration purposes (Mustière et al. 2003).
2.1.2 The topographic domain ontology
Our approach requires a shared taxonomy of geographic concepts
described in the studied databases (or several and aligned taxonomies,
however this issue is out of the scope of this paper for the sake of
simplicity). For our study, we use a bilingual (French/English) taxonomy
built at the COGIT laboratory from a semi-automated analysis of
geographic terms encountered in several data specifications (Abadie and
Mustière 2008). This taxonomy of the topographic domain contains more
than 760 concepts. In the future, we aim at using a richer ontology, in
terms of number of concepts and in terms of description of those concepts.
The production of this ontology is one of the goals of the undergoing
GéOnto project (Mustière et al. 2009). The enrichment of the original
taxonomy is made thanks to the analysis of two types of textual
documents: technical specifications of geographic databases and
travelogues. It is based on automated language processing techniques,
ontology alignment and also on external knowledge like dictionaries and
gazetteers of place names.
2.1.3 Specification ontologies
Our approach also relies on some 'specification ontologies', which are
local ontologies that formalize the content of the specifications of each
considered geographic database. The methodology used to build these
ontologies is described in (Abadie et al. 2010). This methodology
stipulates that each geographic database should be associated with a 'local
specification ontology ( LSO )', which describes its content. The ontology
first contains a transcription of the database schema into OWL formalism.
Then schema classes are annotated with concepts from the global