CRANV2 Citation indexing and bibliographic coupling 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. - 243 - CHAPTER 7 CITATION INDEXING AND BIBLIOGRAPHIC COUPLING It is true [OCRerr]hat librarians do an almost religious job of storing information; it is 9]aced on record, but without evaluation, and much of it is not worth its rental of space. Each oper- ational unit tends to use its own special language, and translators are very few. In warfare, the questioning of returning forces is regarded as a highly skilled speciality. Would social action (and perhaps social research) not gain much if an analogous speciality could be created to assess and consolidate relevant information? Such an organization could not confine its attention to information retrieval, no matter }low efficient such a retrieval process might be. Storage and retrieval systems do not represent evaluation and consolidation of information . L.T. Wilkins: Social Deviance Page 111 In Chapter 7 of Vol. I, an account was given of the compilation of a citation index, and the subsequent preparation of bibliographic coupling groups, with a view to an evaluation being made of this index. As stated in this earlier account, it is a matter for some argument as to how this type of index can be tested in an experimental situation. In carrying out a test in an operational cnvironment, there would be no difficulty beyond the effort required, but although several different ways of presenting the results of this test have been considered, there does not appear to be any procedure which can be considered entirely satisfactory. For this reason two sets of figurcs are being given; the first method probably results in a performance which is better (in comparison with the result[OCRerr] obtained with conventional systems) than should be the case, but it has the major advantage that it does not involve any manipulation of the test results and therefore permits direct comparison to be made between different subsets of questions, different relevance decisions etc. If an evaluation of a citation index can be carried out in an experimental environ- ment, then the second method of presenting results is probably nearer the real performance of the system, and is used for comparison with the results of the conventional index languages. As described in some detail in Vol. I (page 110 and Fig. 7.5) the score sheets for each question gave the results with coupling strength from 1 to 7. The basic scoring at the seven coupling levels for the 42 aerodyn- amics questions with the 1400 document collection is shown in Fig. 7.1T and the results for this set of questions are presented in Fig. 7.2T. Fig. 7.3T presents the results for the 42 questions dealing with structures, while Fig. 7.4T gives the results for the 35 questions having 7 starting terms. A comparative plot of these three question sets is given in Fig, 7.5P. All the results so far shown are obtained with documents of relevance grades 1-4; Figs. 7.6T, 7.7T and 7.8T show the results for documents of relevance grades 1-3, relevance grades 1-2 and relevance grade 1, with the 42 aerodynamics questions, Fig. 7.9P plots the results of the four grades of relevance.