A bi-level concise modular structured knowledge representation scheme

Abstract

Abstract For many years, major drawbacks that have been present in knowledge-based systems are their lack of structure-based knowledge access, high consumption of storage space, long response time, and incapability of tracing items related to each other, unlike a data processing system manipulating a data constructed according to the doubly linked list data structure. This article presents a bi-level knowledge representation scheme that addresses these problems along with several others. This scheme is based on a new knowledge access method in which the expert system communicates with the knowledge base through sub-knowledge sources rather than knowledge sources (facts and rules), i.e., through distinguished objects and relations. This method is used to design the scheme through building the knowledge base using a set of individual objects and relations. This individuality is shown to make it possible to construct highly efficient indices for these objects and relations. It is shown that although the physical structure knowledge achieves high performance through implementing a concise and structured version of the knowledge in its low level, it is organized to provide efficient service to all the tasks carried out by the accompanying knowledge based system. Furthermore, although the low-level knowledge is highly abstracted, it is easily browsed in its full text mode, just like many existing knowledge bases. One of the important issues that this scheme address is the optiional incorporation of certainty degrees that is used by an appropriate reasoning strategy. It is also shown that fuzziness manipulation can be carried out or halted without having to rewrite the knowledge stored physically.