The Classification of Semantic Relations Based on Primitive Properties

Abstract

Although much research has been done on the classification of objects (for example, taxonomic hierarchies in biology), there has been little research to date on the classification of semantic relations among objects. This paper describes a method that uses the definitional properties of semantic relations to classify the relations in a manner similar to that used for the classification of nonrelation objects. The classification schema is being implemented in CYC, a frame-based, knowledge representation system under development at the Microelectronics and Computer Technology Corporation, Austin, Texas [Lenat and Gulla, 1990]. CYC researchers are attempting to encode into a computer common-sense knowledge about the real world, along with mechanisms for reasoning about that knowledge. CYC currently contains more than 5,100 relations (called slots), which are classified loosely according to several different schemes. This paper presents a unifonn classification scheme that is being used to reorganize these slots. The limitations of the scheme are also discussed.