Searching Natural Language Text by Computer

Publisher Website
Google Scholar
Abstract
A fundamental approach to automatic indexing and retrieval of library-stored information through investigating machine search of natural language text is described, and the results of preliminary experimental studies based on that approach are presented. A limited and manageable model of a library (together with search questions) constituted the object of the investigation; the effectiveness with which responsive information could be recovered was measured. The small scale of the model permitted direct examination of the entire collection as a basis for establishing practical measures of "relevance" or "responsiveness" to questions. In terms of these measures, the effectiveness of all information search techniques tested on the model was found to be rather low. Test search by computer was, however, significantly better than a conventional, nonmechanized subject-index method. Thus, even though machines may never enjoy more than partial success in library indexing, a small suspicion might justifiably be entertained that people are even less promising.

This publication has 0 references indexed in Scilit: