Scalable semantic retrieval through summarization and refinement

Query processing of OWL-DL ontologies is intractable in the worst case, but we present a novel technique that in practice allows for efficient querying of ontologies with large Aboxes in secondary storage. We focus on the processing of instance retrieval queries, i.e., queries that retrieve individuals in the Abox which are instances of a given concept C. Our technique uses summarization and refinement to reduce instance retrieval to a small relevant subset of the original Abox. We demonstrate the effectiveness of this technique in Aboxes with up to 7 million assertions. Our results are applicable to the very expressive description logic SHℒN, which corresponds to OWL-DL minus nominais and datatypes. Copyright © 2007, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.