The Design of a Robust Natural Language Interface to a Decision Aid.

Abstract

One of the main goals in the design of natural language interfaces to information processing systems is robust interaction, the capability on the part of the interface to cope with ill formed, though quite understandable, user input. This paper discusses three design principles: (a) the need for a conceptual level of representation of the user's inputs and the operations and data structures supported by the target system; (b) a bottom up mode of language analysis to permit at least a fragmented representation of problematic user input; and (c) methods of model-directed diagnosis, to allow the repair of those representations and the formation of reasonable target-system queries. These principles are illustrated by examining the design and implementation of DESI (Decision Support Interface), an experimental natural language interface to a commercial Decision Support System (DSS) for multidimensional numeric databases. The problems are examined which arise when one is allowed to use free English in place of, or to supplement, the set of commands provided by the DSS. These problems include ungrammatical, misspelled and fragmented input, word-sense disambiguation, ellipsis expansion, and referent selection. The diagnosis process involves recognizing the DSS command(s) intended by the user by using semantic information to type-check and fill in missing command parameters, asking appropriate questions where necessary. Keywords: Natural language, Artificial intelligence, Decision making. (jhd)

Document Details

Document Type
Technical Report
Publication Date
May 01, 1988
Accession Number
ADA198976

Entities

People

  • Mark A. Graves
  • Richard E. Cullingford

Organizations

  • Georgia Tech

Tags

Communities of Interest

  • Human Systems

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Databases
  • Decision Support Systems
  • Formal Languages
  • Information Processing
  • Information Systems
  • Language
  • Natural Languages

Readers

  • Artificial Intelligence
  • Computational Linguistics
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Information Retrieval
  • AI & ML - Machine Translation