CRANV1P1 General Considerations chapter Cyril Cleverdon Jack Mills 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. CHAPTER 1 General Considerations The original Aslib-Cranfield investigation on the efficiency of indexing systems (refs. 1, 2 and 3) did not, by itself, produce firm answers to what is one of the basic problems in information retrieval, namely the decision as to which index language should be used. Certainly it did not, as some people had anticipated, demonstrate that one system was 'better' than another, either generally, or in any given situation. The positive contributions of Cranfield I can be grouped into four areas: 1. It swept away a number of popular misconceptions concerning indexing and index languages that were extant in 1957 when the project commenced. Every index language had its passionate adherents and opponents. The modernists against the traditionalists, those arguing for natural language against controlled vocabulary, those preferring alphabetical as opposed to classified arrangement, all could find both comfort and dismay in the results of Cranfield I. It was shown to be not true that postcoordinate indexing was vastly superior to precoordinate indexing, it was not necessary to put 120 entries into a card catalogue to retrieve a document covering five concepts, yet on the other hand it was not true that a postcoordinate system (at that time usually associated with the Uniterm system) necessarily need have weaknesses due to lack of term control; the chain index did not provide a satisfactory means of entry into a single order facet classified catalogue nor, on the other hand, did engineers find any particular difficulty in using the long numerical notations of the Universal Decimal Classification. Such were only some of the view- points which had been endlessly argued without any experimental evidence to justify either side. 2. With the test of the index of metallurgical literature of Western Reserve University, it was shown that an evaluation could be made of an operational system with comparatively little effort and by using only a small sample of the collection. Since that time improvements have been made in the methodology, and experience has shown in what respects improvements are still necessary, but the general methods first tried in 1962 have been successfully used in a number of different applications (e.g. Refs. 5 and 6). 3. It stimulated a considerable amount of discussion (see, tor instance, the bibliography in ref. 4) which has helped to clarify the problems of information retrieval, and created an interest in the methodology of evaluation. 4. It provided sufficient data to enable provisional statements to be made covering a number of aspects of information retrieval systems. It was in the new hypotheses which could be formulated that the earlier project is of main importance in regard to the present work. Swanson (Bef. 4), in the most exhaustive and scholarly review of Cranfield I that has been made, has listed the following points which appeared to him as being significant.