Production system architectures are another way of representing knowledge. Proposed by A. Newell, production systems were originally presented as models of human reasoning. A set of production rules (each essentially a “pattern->action” pair) operate on a short-term memory buffer of relevant concept, although recent versions tend to have an unlimited memory of concepts. A basic control loop tries each rule in turn , executing the “action half” of the rule only if the “pattern harf” matches. Representing knowledge as pattern->action pairs has proved to be a very natural way of extracting and encoding rule-based knowledge in many applications and now production systems are widely used to construct special-purpose knowledge-based systems, socalled expert systems (Fixgure 3). Various abstract production systems , such as Emycin were devised to make this task easier.