Lexicographic Repair Under Querying Prioritised DL-Lite Knowledge Bases

Abstract

This article discusses the issue of inconsistency in responses from various DL-Lite knowledge bases. This inconsistency problem is at the origin of several sources of assertions with different levels of reliability. The various solutions proposed in the literature that have to do with retrieving an exhaustive and coherent list of responses are not satisfactory from the point of view of reliability and performance. The solution that we present to solve this problem is articulated around two phases: the first phase consists of interrogating the different knowledge bases to retrieve all of the possible answers, which may be inconsistent and/or contradictory, and the second phase consists in repairing these inconsistencies and/or contradictions. To do this, we propose an approach based on three algorithms that we developed in this framework: a first algorithm for non-defeat repair, a second algorithm for lexicographic repair and a third algorithm for non-defeat repair based on lexicography of possible inconsistent responses. The experimental study carried out on the different data collections, as well as the analysis of the results obtained, confirm the performance of our approach as well as its efficiency in regards to productivity and complexity in terms of execution time.