Decomposition-based Reasoning for Large Knowledge Bases in Description Logics

Reasoning in a Knowledge Base(KB) is one of the most important applications of Description Logic (DL) reasoners. The execution time and storage space requirements are both significant factors that directly influence the performance of a reasoning algorithm. In this paper, we investigate a new technique for optimizing DL reasoning with the purpose to minimize two factors above as much as possible. This technique is applied to speed-up TBox and ABox reasoning, especially for large TBoxes. The incorporation of this technique with previous optimization techniques in current DL systems can effectively solve intractable inferences. Our technique is called "ontology decomposing" in which decomposition of one ontology to many sub-ontologies is implemented such that it still preserves the semantic and inference services of original ontology. We are concerned about how to reason effectively with multiple KBs and how to improve the efficiency of reasoning over component ontologies.