IRS13 Document Length 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. V-[OCRerr]7 Co-ordination of Two Terms TOTAL RELEVANT NON-RELEVANT [OCRerr] RECALL* PRECISION[OCRerr] INPUT AVERAGE DOCUMENTS DOCUMENTS DOCUMENTS RATIO RATIO TEXT LENGTH RETRIEVED RETRIEVED RETRIEVED Titl[OCRerr] 7 623 111 512 56.1% 17.&7o Indexing 1 1[OCRerr] 1,239 138 1,101 69.7% ii.i% Indexing 2 22 2,029 162 1,867 81.8% 8.0% Indexing 3 33 2,381 166 2,221 83e8% 7.0% Abstract 60 2,820 i66 2,65[OCRerr] 83.8% 5.9% L_ _- Co-ordination of Four Terms IN?UT AVERAGE TOTAL RELEVANT NON-RELEVANT RECALL* FRECISI0[OCRerr] DOCUMENTS DOCUMENTS DOCUMENTS TEXT LENGTH RETRIEVED RETRIEVED RETRIEVED RATIO RATIO Title 7 147 29 18 114.6% 61.7% Indexing 1 114 1140 60 80 30.3% 142.9% Indexing 2 22 295 87 208 143.9% 29.5% Indexing 3 33 1412 97 315 149.0% 23.5% Abstract 60 5214 92 1432 146.5% 17.6% * These ratios are computed using the aggregates (Ilmicro!? evaluation). Cranfield Project results comparing titles, abstracts, and indexing at three levels of exhaustivity, using search term co-ordination of two and four terms, Word Form Language 13a. Figure 33.