ON-LINE TRANSLATION OF NATURAL LANGUAGE QUESTIONS INTO ARTIFICIAL LANGUAGE QUERIES.
Abstract
The report describes a computer program that demonstrates one approach to the problem of relating questions, posed in a limited subset of English, to a set of facts stored in an on-line data base. The program, called CONVERSE, is designed to use an existing data management system, to provide answers to questions. Where possible, CONVERSE translates an English question into one or more file-searching procedures. If complete translation is not possible, the program provides a user with information that may help him in defining new terms or rephrasing his question into acceptable English terms. CONVERSE accepts generic or browsing types of questions that ask for information about the data base as well as questions of a more specific nature. The translation procedure is based on a generative model of syntax and semantics that is comprehensive enough to automatically resolve some forms of syntactic and semantic ambiguity. A definition of 'fact' is introduced to help specify semantic associations between questions and data values. A dictionary and a series of syntactic, semantic interpretation, and query construction rules constitute a file of interpretive data that is used by the program in effecting question-to-query translation. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 28, 1967
- Accession Number
- AD0654595
Entities
People
- Charles H. Kellogg
Organizations
- System Development Corporation