On the Foundation of Evaluation.

In this article the derivation of a general retrieval effectiveness measure is investigated. It is shown that the measure cannot be justified on the basis of formal considerations associated with measurement theory. The implications of the mathematical condition which defines the relative importance a user assigns to precision and recall are also investigated. Inconsistencies in the definition of the associated weight are resolved, and it is shown that the influence of the weight on the effectiveness measure can produce counterintuitive results. An alternative effectiveness measure provided by the MZ-metric is recommended. [ABSTRACT FROM AUTHOR]