Presented at: Third International Workshop On Ontology Matching (OM2008)
by Stefan Dietze, John Domingue
The widespread use of ontologies raises the need to resolve heterogeneities betweendistinct conceptualisations in order to support interoperability. The aim of ontology mapping is,to establish formal relations between a set of knowledge entities which represent the same or asimilar meaning in distinct ontologies. Whereas the symbolic approach of established SWrepresentation standards – based on first-order logic and syllogistic reasoning – does notimplicitly represent similarity relationships, the ontology mapping task strongly relies onidentifying semantic similarities. However, while concept representations across distinctontologies hardly equal another, manually or even semi-automatically identifying similarityrelationships is costly. Conceptual Spaces (CS) enable the representation of concepts as vectorspaces which implicitly carry similarity information. But CS provide neither an implicitrepresentational mechanism nor a means to represent arbitrary relations between concepts orinstances. In order to overcome these issues, we propose a hybrid knowledge representationapproach which extends first-order logic ontologies with a conceptual grounding through a setof CS-based representations. Consequently, semantic similarity between instances –represented as members in CS – is indicated by means of distance metrics. Hence, automaticsimilarity-detection between instances across distinct ontologies is supported in order tofacilitate ontology mapping.
Resource URI on the dog food server: http://data.semanticweb.org/workshop/om/2008/paper/main/28
Explore this resource elsewhere: