A searcher can continue to step down the chain until the retrieved document set is reduced to a small number. But during the earlier steps of a search, this set is large. In principle, it can be ranked according to some criterion, and the top-ranked items displayed first. NCSU offers rankings according to publication date, title A–Z, author A–Z, call number, popularity (based on circulation data?), and “relevance” (the criteria used to make relevance assessments are not stated).
We encounter a problem in an online faceted documentary classification. Suppose, in a library OPAC, that the search is stopped at a high-level term such as Computing, because the aim of the search is to find a general book on the subject. The retrievable set will be all documents to which the term “computing” has been assigned—a large set comprising documents on Computing and all its subdivisions. We need a way of ranking documents from general to special, so that the most general books on computing will be displayed first. At NCSU, sorting the set by call number (based on the Library of Congress shelf classification) might provide an answer. But within the faceted classification alone, it is not easy to visualise a solution.