A heuristically ‘optimal’ knowledge base organization technique

SOT (self-organizing tree) is a general-purpose knowledge base organizing program which automatically computes an efficient discrimination scheme for accessing large numbers of facts, rules, etc., expressed in a slot-filler representational format. The discrimination scheme supports a variety of accessing modes, including exact, instance and template matching; and ‘best first’ and resumable search. This paper describes the organizing algorithm and the nature of the search process; and gives performance statistics for the system in several tasks.