A system for UNIX command retrieval using a multilayered neural network

Technical Report · Mon Apr 01 04:00:00 EST 1991

OSTI ID:5739015

Preston, D R

This project developed a system that allows a user to access information on UNIX commands by entering a natural language description of the function the user wishes to have performed. The user is not constrained by having to phrase the query in Boolean notation, and the system does not use strict and-ing and or-ing to retrieve the information. This was done in two parts: first, the UNIX man pages were indexed using statistically-based automatic indexing techniques. This process resulted in a list of the terms contained in a man page document that are most useful for recalling the command described in that man page. Second, a multilayer neural network was defined that associated the extracted terms with their commands. The multilayer neural network allows a retrieval paradigm that falls between and-ing the query terms and or-ing the query terms, and allows the user to control the broadness of the results. 12 refs.

• Research Organization: Lawrence Livermore National Lab., CA (USA)
• Sponsoring Organization: DOE; USDOE, Washington, DC (USA)
• DOE Contract Number: W-7405-ENG-48
• OSTI ID: 5739015
• Report Number(s): UCRL-LR-107023; ON: DE91012105
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200 -- Mathematics & Computers
990300* -- Information Handling
COMPUTER CODES
EXECUTIVE CODES
INFORMATION RETRIEVAL
NEURAL NETWORKS