Scalable description logic reasoning architectures for the semantic web

点击翻译按钮获取中文摘要

The semantic web extends the WWW from a collection of data and documents, that are designed for human consumption and are often hard to find and use, into a collection of knowledge that is understandable and very convenient to use by humans as well as computer programs. The knowledge available in the semantic web depends on wide-spread information (e.g., collections of web pages) and on the compilation of information into (often distributed) formal ontologies or knowledge bases. The Web Ontology Language (OWL) was approved in 2004 as a W3C recommendation. It is a response to the semantic web initiative and has become a de-facto standard in knowledge representation for many scientific communities as well as companies and governments. OWL is partially based on a family of knowledge representation formalisms called description logics. The appeal of description logics is the immediate availability of highly optimized reasoners offering a wide range of automatic inference services relevant for supporting the semantic web. Description logic reasoners are usually based on so-called tableau procedures requiring highly sophisticated optimization techniques due to their inherent runtime complexity. The first objective of this proposal is to achieve better reasoning scalability and to support persistency. This goal is due to recent developments in the semantic web community where (simple description logic) reasoning has been applied to very large ontologies consisting of up to millions of assertions. The challenge is to provide persistency and ensure scalability even if complex reasoning is required. This motivates the need to design and evaluate tableau optimization techniques which support scalable reasoning on persistent and virtually unlimited ontologies. The second objective is to develop novel optimization techniques for combinations of description logic constructors such as number restrictions, nominals, and inverse roles because known techniques are insufficient.