A Relevance-Directed Algorithm for Finding Justifications of DL Entailments

Qiu Ji,Guilin Qi,Peter Haase
DOI: https://doi.org/10.1007/978-3-642-10871-6_21
2009-01-01
Abstract:Finding the justifications of an entailment, i.e. minimal sets of axioms responsible for the entailment, is an important problem in ontology engineering and thus has become a key reasoning task for Description Logic-based ontologies. Although practical techniques to find all possible justifications exist, efficiency is still a problem. Furthermore, in the worst case the number of justifications for a subsumption entailment is exponential in the size of the ontology. Therefore, it is not always desirable to compute all justifications. In this paper, we propose a novel black-box algorithm that iteratively constructs a set of justifications of an entailment using a relevance-based selection function. Each justification returned by our algorithm is attached with a weight denoting its relevance degree w.r.t. the entailment. Finally, we implement the algorithm and present evaluation results over real-life ontologies that show the benefits of the selection function.
What problem does this paper attempt to address?