Evaluation of criteria on information retrieval
Tsunenori Ishioka · Systems and Computers in Japan · 2004
Abstract We investigate van Rijsbergen'sF‐measure, the break‐even point, and 11‐point averaged precision, all of which can be translated into one‐dimensional scalar quantities from the precision and the recall. These investigations can be done by comparing to the tetrachoric (four‐fold) correlation coefficient and phi (four‐fold point) coefficient, which are often used as indices of statistical association in a 2 × 2 contingency table. The results show that when the fallout rate is less than 0.1, theF1measure has similar properties to the phi coefficient, the break‐even point is almost equivalent to the phi coefficient, and the 11‐point averaged precision should be a measure which is larger than the phi coefficient and smaller than a tetrachoric correlation coefficient. © 2004 Wiley Periodicals, Inc. Syst Comp Jpn, 35(6): 42–49, 2004; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/scj.10583