Pattern Ranking for Semi-automatic Ontology Construction

When developing semantic applications, the construction of ontologies is a crucial part. We are developing a semiautomatic ontology construction approach, OntoCase, relying on ontology patterns as additional resources. A crucial part of this approach is how to select the appropriate patterns based on the input representation extracted from a text corpus. In this paper, we suggest a pattern ranking and selection approach with the ability to partially bridge the gap between abstract patterns and specific terms, as well as being specifically tuned to the characteristics of ontology patterns. Compared to existing ontology ranking schemes our approach adds indirect matching of terms as well as relation matching. An initial experiment indicates that OntoCase ranking performs better, especially when ranking small and abstract patterns, than existing ranking approaches.