Harvesting online ontologies for ontology evolution

Ontologies need to evolve to keep their domain representation adequate. However, the process of identifying new domain changes, and applying them to the ontology is tedious and time-consuming. Our hypothesis is that online ontologies can provide background knowledge to decrease user efforts during ontology evolution, by integrating new domain concepts through automated relation discovery and relevance assessment techniques, while resulting in ontologies of similar qualities to when the ontology engineers' knowledge is solely used. We propose, implement and evaluate solutions that exploit the conceptual connections and structure of online ontologies to first, automatically suggest new additions to the ontology in the form of concepts derived from domain data, and their corresponding connections to existing elements in the ontology; and second, to automatically evaluate the proposed changes in terms of relevance with respect to the ontology under evolution, by relying on a novel pattern-based technique for relevance assessment. We also present in this thesis various experiments to test the feasibility of each proposed approach separately, in addition to an overall evaluation that validates our hypothesis that user time during evolution is indeed decreased through the use of online ontologies, with comparable results to a fully manual ontology evolution