Representing Complex Physical Objects in Memory

Wasserman, Kenneth; Lebowitz, Michael

Researchers an artificial intelligence have proposed and implemented several representation systems for use in computer programs that "understand" natural language input. Noticeably lacking from these systems is a robust and concise method of representing complex physical objects. This paper describes a language independent scheme for encoding real-world objects in a manner that captures elements of how people describe them. Two major groups of objects are distinguished: ''unitary'' objects that are described by a single "shape-descriptor"; and "composite" objects that are represented by a frame-based system that focuses on the physical relations that exist among obiects. The heart of this scheme is a primitive-based framework that classifies physical relations into three fundamental categories with five possible properties. Our current work on RESEARCHER, a program that employs this scheme while reading patent abstracts, is also discussed.



More About This Work

Academic Units
Computer Science
Department of Computer Science, Columbia University
Columbia University Computer Science Technical Reports, CUCS-037-82
Published Here
October 26, 2011