Retrieval performance and information theory

摘要

This paper challenges the meaningfulness of precision and recall values as a measure of performance of a retrieval system. Instead, it advocates the use of a normalised form of Shannon's functions (entropy and mutual information). Shannon's four axioms are replaced by an equivalent set of five axioms which are more readily shown to be pertinent to document retrieval.The applicability of these axioms and the conceptual and operational advantages of Shannon's functions are the central points of the work.The applicability of the results to any automatic classification is also outlined.