ISR10 Evaluation of Document Retrieval Systems chapter Joseph John Rocchio Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. 5-[OCRerr]1 log precision = ln (n0!) n 0 ln[OCRerr] [OCRerr]i) i=1 a[OCRerr]vantage of these in[OCRerr]ices lies in the fact that they are simpler and therefore easier to comp[OCRerr]te than the normalized indices. In addition, the ra[OCRerr]ik recall parameter requires no scale expansion to assume a useful range. [OCRerr]he disadvantage of [OCRerr]oth these measures is their dependence on n , the num[OCRerr]er of relevant documents. As n varies 0 0 from query to query this dependence makes it impossible to average the unnormalized indices over a set of search requests to produce a meaningful systern!s evaluation.