Functional requirements for very large knowledge bases

A large knowledge based system can be considered equivalent to a large data base system; but, instead of the more usual algorithmic rules, most of the rules are expressed in the form of specifications. Such a form of expression has the considerable advantage that theorem provers can be used to detect inconsistencies between the rules. The proposed general architecture can be summarized in a diagram: (Diagram) In this diagram, the APPLICATION involves physical variables such as time, part, distance, etc. The MODEL of the application defines a relationship between any combination of the corresponding mathematical variables. This is because the model contains general rules which relate these variables. These rules come either •from the user, who himself has induced general rules about the data, or •from the system's own GENERALISER, an adaptive pattern recogniser which induces rules from the data.