Combining inductive and deductive reasoning

Work in artificial intelligence has shown that rule induction is a useful aid to knowledge acquisition but that its general applicability is limited. In particular, induced rules can be difficult to understand and hard to modify. This paper describes a computer program to generate knowledge bases from examples in a form which can be interpreted either as a set of rules or as an inference network. The rules are readily understood so that the structure can be modified by the expert into a form that the program will tune again to fit the examples. This new tool for analysing data/knowledge draws together two very different techniques, inductive and deductive, used in building expert systems.