Ontology similarity assessment based on lexical and structural model features extraction

This work discusses a method for extracting quantitative measures from terminology and instance base components of semantic models. The method introduces a multi-criteria analysis while comparing ontology models and assessing semantic similarity. The article presents theoretical overview of the method and tool implementing evaluation schemes supplemented with practical examples. The capabilities of the method can be used for semantic pattern recognition within knowledge bases, which can be utilised by analytical tools especially in the security domain (criminal threat, financial fraud detection, etc.). The specificity of security applications requires methods dedicated for analysis of hidden, indirect, comprehensive and versatile data. Structural analysis method and its implementation in form of ETOSE plugin for Protégé OWL environment delivers process-based approach for evaluating instance bases. The mechanisms has been designed to operate as a data flow interceptor, collecting the data and transforming them into instances expressed in a specific domain ontology (set of ontology modules). Presented quantitative approach has been applied in terrorist threat assessment, financial fraud identification tasks where certain templates of behaviour and associations can be described. The method and tool utilize structural and lexicon comparison of compared ontologies in order to deliver multicriteria evaluation of concepts, relationships and indirectly implemented axioms. Key-Words: ETOSE, semantic structural analysis, semantic similarity, ontology, quantitative analysis, similarity measurement, OWL2, Protege