| Summary: | OWL ontologies are becoming increasingly important to the Semantic Web. Inconsistencies in OWL ontologies can easily occur, and it is a challenging task for ontology modellers, especially for non-expert ontology modellers, to resolve ~uch inconsistencies. Existing ontology debugging tools can identify the problematic axioms for unsatisfiable concepts. However, current systems provide limited support for selecting which problematic axioms to remove/modify to resolve the unsatisfiability. In the case of modification, the modellers are provided with limited information about the impact of changes on the ontology. Therefore, there is a growing ~eed for tools that can provide. guidance on (1) selecting the axioms to modify, and (2) how to minimise the impact of proposed modifications on the ontology. We firstly extend an existing tableau-tracing technique for debugging in order to pinpoint problematic parts 'ofaxioms('which are responsible for the unsatisfiability of a concept. This information allows us to rewrite faulty axioms by removing error-causing concepts, instead of removing axioms entirely. Secondly, we investigate the heuristics used by ontology engineers to resolve inconsistencies in ontologies. A number of empirical studies have been conducted to acquire a variety of such heuristics from ontology engineers. Some of the acquired heuristics are already incorporated in existing tools, however, we have also discovered a sizeable number of additional useful heuristics. Moreover, we investigate the usefulness of incorporating these heuristics in a software tool to guide non-expert ontology modellers to select appropriate axioms for modification. The result of our usability study shows that the heuristics are useful to help ontology users debug ontologies. Thirdly, in the case of modification, we study the usefulness of providing the modeller with detailed information about the impact ofchanges on ~he ontology. Specifically, we utilise the extended tableau-tracing technique to identify changes which are helpful in that they restore lost entailments (due to axiom removal), and those that are harmful in that they cause additional unsatisfiability. The work presented in this thesis constitutes an advance in debugging ontologies. The thesis combines a heuristic approach and formal methods for dealing with inconsistencies in ontologies. It contributes to knowledge in that it shows that the approach of acquiring human heuristics and encoding them in a tool is viable for ontology debugging. It also shows that the fine-grained approach to indicating lost entailments and recommending modifications is useful. The results of this thesis should be valuable to both the implementors of ontology management tools, and to ontology users, especially for the less experienced users of ontologies.
|
|---|
