Inferential Question Answering: Extending Converse.

Abstract

The paper describes a deductive component developed for use by an existing question-answering system. The emphasis is on deduction within a question-answering context rather than within a mathematical inference system. The inference component was designed to find, for a given input question, the relevant general premises needed for deduction from a very large number of possible premises, most of which are irrelevant to any particular problem at hand. The inference system first constructs preliminary, skeletal derivation proposals. The purpose of these proposals is to find possible deductions before any attempt is made to verify the proposals. Verification is thus delayed until overall proof plans have been established. Later phases of processing examine the variable flow within a proposal to detect possible collisions and search the fact file for compatible sets of values for instantiation. The use of semantic information to aid deduction is incorporated into the system. (Author)

Document Details

Document Type
Technical Report
Publication Date
Jan 31, 1973
Accession Number
AD0754801

Entities

People

  • Charles Kellogg
  • Larry Travis
  • Philip Klahr

Organizations

  • System Development Corporation

Tags

DTIC Thesaurus Topics

  • Collisions
  • Verification

Readers

  • Artificial Intelligence
  • Computational Linguistics

Technology Areas

  • AI & ML
  • AI & ML - Information Retrieval