The formal analysis of document retrieval systems

Abstract

AbstractThe output structure of text retrieval systems resembles formally a description of state that is well known in other fields. Such a (statistical) description allows the derivation of a unique and measurable performance criterion that, in turn, may be used in discussing various parameters of systems performance.The standard quantities “recall” and “precision” appear to be incomplete and dependent measures of system output. The method is illustrated by application to a small test data base and query sample.