CRANV2 Main test results chapter 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. - 109 - Section 3 Precision Devices Two subsets of questions have been used to investigate precision [OCRerr]evices. Subset 4 consisted of 19 questions all having seven starting terms, while Subset 5 had 17 questions all having eleven starting terms. For reasons explained in Chapter 2, these tests made on precision devices required a different search rule (B) to what has been used so far. This rule required that the main term in a concept must be present ff the qualifying term was to be accepted, e.g. with the concept 'Hinged flap', '5[OCRerr]in[OCRerr]ed' would no[OCRerr] be accepted unless 'Flap' was also present. Therefore it is necessary first to present the results of searches made on these two subsets with search rule B, so that later comparison can be made. Each subset is tested with Index Language I.l.a and 1.6.a. There now follow the results of tests with (b) partitioning, (Figs. 4.310T -4.313T), (c)interfixing, {Figs. 4.320T -4.323T)and (a) partitioning plus interfixing {Figs. 4.330T - 4.333T), with both subsets of questions presented separately. For each precision device results are given when two recall languages are used, 1.1 Natural Language and 1.6. the aggregate of synonyms, quasi-synonyms and word forms. Because of the large clerical effort required to obtain these results, the precision ratios were not obtained at the lower levels of coordination. This, combined with the relatively small sets of questions used, limits the usefulness of any graphical presentation, for the results show little consistency. Therefore in the plots 4.340P, 4.341P, 4.342P and 4.343P, only a single generalised curve has been drawn for the four sets of data that are presented in each curve. The relative positions of the various symbols give an indication of the performance of the particular system. LIST OF FIGURES 4.300T 4. 301T 4. 302T 4.303T 4. 310T 4.311T Precision Index No. of Question Document Device Language Questions Subset Collection I.l.a 19 4 1400 (Search I.l.a 17 5 1400 Rule B) 1.6.a 19 4 1400 1.6.a 17 5 1400 PARTITIONING I.l.b 19 4 1400 " I.l.b 17 5 1400