Methods for evaluating the number of relevant documents in a collection

Abstract

Two essentially practical methods which do not rely on the uncertainties of the statistical approach are proposed. The method of limiting recall has been used with a document collection of over one million items; it requires test searches to be carried out independently by several groups. The use of a very new and powerful retrieval aid named ZOOM is also proposed as a form of super-search for each test question, the results of which can be assumed to be close to the total number of relevant documents. Used together, these two methods would provide cross-verification.