Semantic Learning and Procedural Learning

The paper is concerned with the relationship between the problem of automatically generating, debugging and improving problem solving programs, and the problem of machine representation of knowledge. The design of a computer semantic memory is given and particular attention is devoted to the processes far storing and retrieving information. The problem solving system and its memory system are centered around the notion of production rules. The proposed memory shows the feasibility of such an approach by providing, on one hand. an efficient way to represent and access a large knowledge base. and. on the other hand. by relieving the learning processes from some of the burden of maintaining and modifying the production rule system according to the acquired experience.