Meeting the Interlocking Needs of LF-Computation Deindexing and Inference: An Organic Approach to General NLU

Abstract

We argue that a "theory bottleneck" encountered in the 70's and early 80's in attempts to build comprehensive NLU systems led to a fragmentation of NLU research, which still persists. To some extent, this fragmentation represents an appropriate response to the variety and subtlety of remaining problems; but at this point, it also represents a loss of nerve: NLU is an organic phenomenon, and enough has been learned about the vexing problems of the 80's to try to integrate these insights and build more comprehensive theories and extensible implementations. On that premise, we have been building such a comprehensive framework . Its centerpiece is Episodic Logic (EL), a highly expressive knowledge representation well-adapted to the interpretive and inferential needs of general NL U. The logic has been successfully implemented in the EPILOG system, which has been applied to several domains proving EL's practicality: reasoning about small excerpts from the Little Red Riding Hood story, as a natural language interface for the TRAINS planning system at the Uni versi ty of Rochester, and for a message processing application at the Boeing Co.