Topical relevance relationships. II. An exploratory study and preliminary typology

The first part of this two‐part series argues that the assumption of topic matching between user needs and texts topically relevant to those needs is often erroneous. This second part reports an empirical investigation of the question, “What relationship types actually account for topical relevance?” In order to avoid the bias of topic‐matching search strategies, user needs are back‐generated from a randomly selected subset of the subject headings employed in a user‐oriented topical concordance. The corresponding relevant texts are those indicated in the concordance under the subject heading. The study compares the topics of the user needs with the topics of the relevant texts to determine the relationships between them. This examination reveals that topical relevance relationships include a large variety of relationships, only some of which are matching relationships. Others are examples of paradigmatic relationships or syntagmatic relationships. Indeed, there appear to be no constraints on the kinds of relationships that can function as topical relevance relationships. They are distinguishable from other types of relationships only on functional grounds. © 1995 John Wiley & Sons, Inc.