Colinked descriptors: Improving vocabulary selection for end‐user searching

Abstract

This article introduces a new concept and technique for information retrieval called colinked descriptors. Borrowed from an analogous idea in bibliometrics—cocited references—colinked descriptors provide a theory and method for identifying search terms that, by hypothesis, will be superior to those entered initially by a searcher. The theory suggests a means of moving automatically from two or more initial search terms, to other terms that should be superior in retrieval performance to the two original terms. A research project designed to test this colinked descriptor hypothesis is reported. The results suggest that the approach is effective, although methodological problems in testing the idea are reported. Algorithms to generate colinked descriptors can be incorporated easily into system interfaces, front-end or pre-search systems, or help software, in any database that employs a thesaurus. The potential use of colinked descriptors is a strong argument for building richer and more complex thesauri that reflect as many legitimate links among descriptors as possible. © 1996 John Wiley & Sons, Inc.