Abstract: A method and an apparatus for constraint-oriented inference provides improved knowledge base formation efficiency and an ability to maintain consistency among slots. Constraints indicating relationships among slot values of frames are stored collectively, pointers which make the constraints accessible by relevant slots are attached automatically, and the constraints are simplified by inserting slot values determined by processes of inference.

Abstract: According to this invention, an attribute of an object and a relation of the object with another object is expressed by a paired set of an attribute name and its attribute value. A knowledge base of the frame type is used. In the knowledge base, the attribute of objects and the relation among different objects are expressed in terms of the attribute name/attribute value set, and framed and stored for each objects. The knowledge base is weighted depending on the strength of the attribute of the object and the strength of the relation between different objects. An inference processing unit accesses the knowledge base, and extracts the strong attribute or the strong relation whose strength is presented by the weight, and obtains the interference result. Knowledge can be accessed in a natural way, based on the strength of the relation between objects and the strength of the attributs of the objects, thus enabling a flexible and high-speed infernece processing.