Approximate information retrieval based on multielement bounds

摘要

Ontology-based information systems can offer precision and effective information retrieval but suffer from the problem of ontology heterogeneity. Approximate information filtering approaches use approximations of concepts to solve this problem. Current methods only consider one-to-one concept subsumption relations, so cannot yield the closest approximations of concepts. No effective algorithm has been previously reported to find the closest approximations of concepts. This paper defines multielement least upper bounds and multielement greatest lower bounds that contain disjunctions or conjunctions of concepts. The approximations based on them are proved to be the closest possible approximations of concepts. Multielement bounds may contain much redundancy: this increases the expression complexity of the approximations of concepts. We also define simplified multielement bounds to remove redundancy, and provide effective algorithms to find them.