CRANV2 Appendix 3A appendix Cyril Cleverdon Michael Keen Cranfield An investigation supported by a grant to Aslib by the National Science Foundation. Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. - 265 - APPENDIX 3A TABI.,ES OF GENERALITY NUMBER, AND FALLOUT, RECALL AND PRECISION RATIOS These tables provide a method of conversion between the recall, precision and fallout ratios and the generality number. The equations on which the tables are based, are given on pages 41 - 42. The method of using the tables is as follows. Assume three separate search results where in the first case the generality number is 20 in the second case it is 5, and in the third case it is 25. At 60% recall ratio, the precision ratio is 15% in the first case, 25% in the second case, and 60% in the third case. It is desired to equate the performance in regard to recall and precision. The correction is done by converting the precision ratios to what they would be at a generality number of 2. Reference to the tables for a general- ity number of 5 (page 270) shows that for a recall ratio of 60% and a precision ratio of 25%, the fallout ratio is 0.905%. Reference to the tables for a generality number of 2 (page 267) shows that for a recall ratio of 60% and the fallout ratio of 0.905%, the precision ratio would be approximately 11%. For the third case, where the generality number was 25, reference to the appropriate table on page 281 gives a fallout ratio of 1.026% for a recall ratio of 60% and a precision ratio of 60%. From page 267 it is found that this fallout ratio of 1. 026% at 60% recall occurs with a precision ratio of approximately 10%. The original and corrected ratios are then as follows: io Generality Number 2 5 25 Recall Ratio Precision Ratio Fallout Corrected Ratio Precision Ratio 15% 11% 10% , I 0.681% 0.905% 1. 026% 60% 60% 60% 15% 25% 60% This shows that the first case represents the best performance. [( ! i l h :i' i :t!