IRS13 Search Matching Functions chapter E. M. Keen Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. 111-38 EVALUATION L'[OCRerr]UMBE[OCRerr] A[OCRerr]v.D PERCENTAGE* MEASURE USED OF INDIVIDUAL REQUESTS COLL[OCRerr]CTI0N [OCRerr]UT AND DICTIONARY TO DETERMINE NUNERIC LOGICAL BOTH MERIT SUPERIOR SUPERIOR EQUAL IRE-3 Abstract, Stem Normed Recall 26 76e5% 8 23.5% 0 Normed Precision 26 76.5% 8 23.5% 0 34 Abstract, Thesaurus Normed Recall 26 78.8% 7 22.2% 1 Requests -3 Normed Precision 30 88.2% 2* 11.8% 0 CRAN-1 Abstract, Stem Normed Recall 28 77.8% 8 22.2% 6 Normed Precision 26 70.3% 11 29.7% 5 42 Abstract, Thesaurus Normed Recall 19 51.4% 18 48.6% 5 Requests -3 Normed Precision 21 53.8% 18 46.2% 3 Text, Suffix I[OCRerr]I Normed Recall 14 4o.o% 21 60.0% 0 Normed Precision 17 48.6% 18 51.4% 0 Text, Stem Normed Recall 18 52.9% 16 47.1% 1 Normed Precision 16 45.7% 19 54.3% 0 ADI Text, Thesaurus-l Normed Recall 22 62.9% 13 37.1% 0 Normed Precision 23 65.7% 12 34.3% 0 35 Requests Abstract, Suffix tSt Normed Recall 17 51.5% 16 48.5% 2 Normed Precision 19 57.6% 14 42.4% 2 Abstract, Stem Normed Recall 17 51.5% 16 48.5% 2 Normed Precision 18 51.4% 17 48.6% 0 Abstract, Thesaurus Normed Recall 16 2*8.5% 17 51.5% 2 -1 Normed Precision 18 54.5% 15 45.5% 2 * Percentages do not include cases where both options h[OCRerr]ve equal merit. Comparisons of individual request merit giving the numbers of requests favoring numeric and logical with percentages on 10 options from three collections, according to merit assigned by normalized recall and precision. Figure 23.