Linguistically-Evaluated Integration of Semantic Web Data into Domain...

The present paper reports on an experiment aimed at optimizing Semantic Web data for the construction of domain-specific ontologies. For this purpose, an Ontology Engineering Tool (The Semantic Graph Tool, SGT) was developed[1]. SGT facilitates the integration of RDF-Data-structures into a domain ontology. Exemplary source for data integration is delivered by Freebase, a major source for the Google Knowledge Graph. As for the linguistic workflow employed, it allows for removing trivial and repetitive facts contained in Freebase as well as adding more detailed information necessary for subsequent domain ontology engineering. Regarding the ontology engineering workflow, basic NLP (natural language processing) strategies like tagging, chunking, keyword weighting, and co-occurrence identification is carried out. As a result, text snippets for building a domain ontology can be identified. The following step is to integrate those Freebase terms, which are similar to the respective concept nodes. Similarity calculation is done with the help of the Lesk algorithm [1]. The labelling method follows the guidelines described in [2].

[1] All programming was done during the internship of Alessandro Aichmann at Econob, Lakeside Park, Klagenfurt.

http://www.econob.com/index.php/en/company/about-us. The authors wish to express their gratitude.

[2] Günther Fliedl, Christian Kop, Jürgen Vöhringer (2010). Guideline based evaluation and verbalization of OWL class and property labels. Data & Knowledge Engineering 04/2010; 69:331-342.