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. 5-27 dividing by the maximum possible error. [OCRerr]rom the previous ar[OCRerr]uinent this error is: n 0 max precision error in [OCRerr]N - (i-i) - in (n0!) i=1 in N! ln(n !) (N - n )! - o 0 - ln(N0[OCRerr] (the binomial coefficient). The normalized index of precision error is therefore: n 0 ln[OCRerr][OCRerr](i) - ln(n0!) i=1 ln(nNo) Pen (5.22) Again, since this is an index of precision error, it is desirable to reverse it. Therefore: p =1- n ln[OCRerr](i) - ln(n0!) i=1 ln(nN \) 0 (5.23) is the desired normalized index 0£ precision performance.. Since both these indic/es re£lect over-all system performance a value of 1 for either implies a value of 1 for the other, in contrast to the conventional recall and precision evaluation measures. The