SARA: an experiment in natural language processing with event-type knowledge representation

Abstract

Recent development in natural language processing have shown that the inclusion of world knowledge in a processing system will inhance the understanding capabilities of the computer. This paper describes a system, called SARA, that implements some of the recent techniques used to organize the computers view of the world. An ability to understand events is demonstrated by generating a paraphrase of a short input story. The systems limitations are discussed as a function of the unsolved problems in designing a universal scheme for an adaquate knowledge representation.