Ontology Generation from Web Tables: A 1+1+N Approach

Abstract

Creating plenty of ontologies is a crucial task for relizing the Semantic Web vision. Techniques of learning ontology from text are far from mature because of the limitations of natual language understanding. However, discovering concepts and relations from tables is much easier. In this paper, we proposed a novel approach to generate ontologies from web tables. Our apporach can be described as a 1+1+N model: one core ontology which contains the core concepts of a given domain and serves as the global ontology, one standard table which serves as a reference during the construction of other ontologies, N new tables from which local ontologies are generated. Type spotters, instance-based schema matching, and some heuristics rules are employed to generate ontologies. Experiment results show that our appoach is feasible and quite effective.