Research On Optimization Of Description Logics Reasoning

The network has changed ours life deeply. It enriches peopele's communication methods, increases approches of gaining knowledge, speeds up the generation of information, and affects ours leisure time. It is the most influential techniqy since the last half centry. But the web pages, which described in html, are aimed to people, but not to our computers, so they are difficult to be handled by softwares automatically. With the increment of network stations and web pages, people find more and more difficult to find what they want.Tim Berners-Lee, who is the inventor of WWW, proposed the semantic web as the next generation of WWW. By adding mete data to the web pages, refining the structuct of the web pages, semantic web aims to make the pages more acceptable to computers. Description Logics is the logic base of semantic web, it can used to duduce the implicated knowledge of the web. So using Description Logics, the knowledge on web pages can be handled by computer more easily and accurately.But it is well known that the ALC, which is a Decription Logic language and has a normal expressive power, is pspace-complete. So it is tough to applicate Description Logics to the environment of WWW. We must find way to optimize the reasoning of Descripton Logics.The attribution of a concept is defined by all the individuals belong to it. The individuals, which are produced by the reasoning process, is helpful to optimize the new reasoning process. But this method is invalid to complex concepts, because they may have a few individuals which can't represent their attribution.If all the individuals of a concept belong to another concept, we can induce that the former is a subconcept of the latter. This surely is not accurate, but it is really an approximation method. When deducting the satisfiability of a complex concept, it may take much of time, even be time out. So if we can forecast the complexity of reasoning, we may approximate the most time consuming one.For Desction Logic SHOIQ, it is known that query in it is an open problem. An appximate method is proposed to deal with this. When people give a query, they give different key words different prefence. So concepts are set with prefences according to the users. Prefence base query graph travel is proposed to utilize the prefence.Partitioning can decompose a big ontology into small size ones; Summarization can cut big ABox down to a much smaller one. These two technologies are potential to sove the big ontology reasoning. So the approximation reasoning based these two methods are also discussed.Based on the preference reasoning by manipulating logical chain, how to represent ranked knowledge base in Web Ontology Language (OWL) is proposed. Using OWL, the preference representation and reasoning are combined naturally. Furthermore, conditional preference can be easily added, and so the expressiveness is enhanced.Experiment for reasoning optimization based on individual, approximation reasoning, preference are given to show the efficiency.