Using provenance to debug changing ontologies

On the Semantic Web ontologies evolve and are managed in a distributed setting, e.g. in biomedical databases. Changes are contributed by multiple persons or organizations at various points in time. Often, changes differ by certainty or trustworthiness. When judging changes of automatically inferred knowledge and when debugging such evolving ontologies, the provenance of axioms (e.g. agent, trust degree and modification time) needs to be taken into account. Providing and reasoning with rich provenance data for expressive ontology languages, however, is a non-trivial task. In this paper we propose a formalization of provenance, which allows for the computation of provenance for inferences and inconsistencies. It allows us to answer questions such as ''When has this inconsistency been introduced and who is responsible for this change?'' as well as ''Can I trust this inference?''. We propose a black box algorithm for reasoning with provenance, which is based on general pinpointing, and an optimization, which enables the use of provenance for debugging in real time even for very large and expressive ontologies, such as used in biomedical portals.