Partitioning large ontologies based on their structures

Abstract

With awareness of ontology capabilities in processing semantic web information, the number of ontologies have been increasing over the past decade. However, there are still some difficulties in working with ontologies having large sizes (that is having considerable amount of concepts and relationships) resulting from high time and space complexity of the processing involved. To overcome these problems, some researchers tend to use clustering and fragmentation techniques to partition the ontologies into meaningful parts called sub-ontology. Such partitioning can be used to process sub-ontologies locally and then combine those processing results to gain final results. In these manners, the technique chosen for the partitioning is an effective factor in the quality of the final results. In this paper we have proposed an efficient new structure-based method for partitioning an ontology to the meaningful clusters. Although, this method can act completely automated, it also enables the user to determine the number of final clusters in each level of granularity. The time-complexity of this method is of where n is number of concepts in the ontology. Key words: Ontology partitioning, sub-ontology, closeness, cluster similarity.