Using Ontological Contexts to Assess the Relevance of Statements in Ontology Evolution

Abstract

Ontology evolution tools often propose new ontological changes in the form of statements. While different methods exist to check the quality of such statements to be added to the ontology (e.g., in terms of consistency and impact), their relevance is usually left to the user to assess. Relevance in this context is a notion of how well the statement fits in the target ontology. We present an approach to automatically assess such relevance. It is acknowledged in cognitive science and other research areas that a piece of information flowing between two entities is relevant if there is an agreement on the context used between the entities. In our approach, we derive the context of a statement from online ontologies in which it is used, and study how this context matches with the target ontology. We identify relevance patterns that give an indication of relevance when the statement context and the target ontology fulfill specific conditions. We validate our approach through an experiment in three different domains, and show how our pattern-based technique outperforms a naive overlap-based approach.