Research On Two Kinds Of Revision Strategies Of Inconsistent Beliefs In Commonsense Reasoning

In the reasoning based on knowledge, knowledge used can't be always consistent because of incompletion,uncertainty,imprecision of knowledge and collision from different knowledge sources. So such knowledge can only be regarded as some beliefs other than as absolute truth. With the development of men's cognition, new belief can often be conflicted with the old ones. It is an important problem how to efficiently reason making use of inconsistent beliefs.At present two attitudes are mostly taken based on inconsistent beliefs. One is to revise the old beliefs in order to keep the consistency of the knowledge base, i.e. belief revision methods [1-10]. The other is to accept inconsistent beliefs and to reason in the precondition of tolerating inconsistency, i.e. unrevision methods [11-17].In belief revision system, when facing new information it is essential how to incorporate it into a consistent knowledge base and to keep the knowledge base consistent. When the new belief is consistent with the old knowledge base, the new belief can be directly added to the knowledge base. But when inconsistent, some beliefs in the old knowledge base often need be deleted in order to keep the revised knowledge base consistent. The deletion is not random. It should reserve more existing information in the old knowledge base and make belief revision conform to the principle of Minimal Change.Recently the problem of belief revision has received most attention in many fields such as artificial intelligent, philosophy and database. There are many different frameworks for belief revision but AGM [1] is the most notable one. Most others rely on the foundations of AGM. They define...