Knowledge-Based Extensible Natural Language Interface Technology Program
Abstract
This research project addressed the problem of developing knowledge-based natural language interface technology that is extensible via natural dialogue between user and computer system. Natural language understanding systems need to be extensible to accommodate changes in the target application system to which they interface as well as to accommodate new users. Typically, however, current systems cannot be extended as part of a normal dialogue session. Instead, extensions must be incorporated and compiled into the interface off line before the interface is loaded for use. This can be costly in terms of down-time and frustrating for the new user. The solution that was pursued in this project was to develop a natural language interface system, called Lydia, that is extensible via methods that are modeled after human behavior. Specifically, the following methods were implemented in the Lydia system a) learning by being told including the ability to understand natural language when it is used as its own meta-language to explain new concepts, relation, and rules and b) being able to infer the category and attributes of new words from their linguistic context when used in a natural language sentence. Keywords: Expert system; Speech recognition; Computer programs; Man computer interface.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 30, 1989
- Accession Number
- ADA216664
Entities
People
- Jeannette G. Neal
Organizations
- Calspan-University of Buffalo Research Center