Finding explanations of inconsistency in multi-context systems (Q460632)

The article discusses inconsistency in a nonmonotonic multi-context system (MCS), which is defined as a ``flexible, abstract formalism to interlink heterogeneous knowledge sources for information exchange''. Such systems are becoming common in the context of the Semantic Web and developing methods for their consistency maintenance is increasingly important. An underlying assumption in an MCS is that contexts, each defined by its own knowledge base with an underlying logic, are autonomous sources that cannot be modified from the outside. The interaction between contexts is carried out by bridge rules that might be nonmonotonic and/or circular, thus causing potential inconsistencies in an MCS. The article presents two approaches to detecting and explaining such inconsistencies. The first, consistency-based (or diagnostic) approach, characterizes inconsistencies in terms of ``modified sets of bridge rules that admit equilibria'', where an inconsistency is defined as ``the lack of an equilibrium''. To recover the consistency of an MCS, pairs of sets of bridge rules are selected so that the rules in the first set are deactivated and added to the second set in unconditional form. The second, entailment-based (or explanation) approach, identifies a pair of sets of bridge rules whose presence or absence entail an inconsistency. Both approaches are illustrated by many examples and characterized in terms of their computational complexities. Prototype implementations by means of so-called ``HEX-programs'' are given. A comprehensive discussion of related work is presented as an appendix to the article. Another appendix contains all proofs and formal results.