Towards measurement of structural complexity for ontologies

Abstract

Ontologies represent knowledge of a particular domain and form an elementary unit for inference techniques on the semantic web. It is important to measure the complexity of ontologies in order for users to better understand, maintain, reuse and integrate them. Existing measures for the complexity of an ontology either handle complexity at schema and instance levels or define subjective parameters to measure ontology complexity. To this end, we present a semi-automated framework to measure the structural complexity at different abstraction levels of an ontology. Moreover, our metrics leverage the information provided by ontology editors, which can be useful to the ontology designer for selection of the ontology with optimum complexity among alternative ontologies. Our framework is inspired by the concept of component-based software complexity metrics and its metrics have been validated against Briand's and Weyuker's benchmarks. We have also performed comparative analysis on public ontologies.