Some new tests of relevance theory in information science

A central idea in Dan Sperber and Deirdre Wilson's relevance theory is that an individual's sense of the relevance of an input varies directly with the cognitive effects, and inversely with the processing effort, of the input in a context. I argue that this idea has an objective analog in information science-the tf*idf (term frequency, inverse document frequency) formula used to weight indexing terms in document retrieval. Here, tf*idf is used to weight terms from five bibliometric distributions in the context of the seed terms that generated them. The distributions include the descriptors co-assigned with a descriptor, the descriptors and identifiers assigned to an author, two examples of cited authors and their co-citees, and the books and journals cited with a famous book, The Structure of Scientific Revolutions. In each case, the highest-ranked terms are contrasted with lowest-ranked terms. In two cases, pennant diagrams, a new way of displaying bibliometric data, augment the tabular results. Clear qualitative differences between the sets of terms are intuitively well-explained by relevance theory.