The Role of Object Primitives in Natural Language Processing.

摘要

Natural language processing techniques rely in part on the use of functional knowledge about physical objects, and associative memory structure. In this paper, a representational system called Object Primitives is presented as an extension to the system of Conceptual Dependency for the purpose of representing physical objects and providing an organizing structure for associative memory. A computer program, OPUS, is described which applies this representational system to the problem of analyzing natural language sentences dealing with objects. Inferences derived from Object Primitive descriptions are made during the conceptual analysis by a system of demons, providing a framework for an integrated understanding system. (Author)