CRANV1P1 Additional Tests 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. - 107 - on both the indexing of these documents and also their abstracts. Subsequently the decision was taken to extend this work so as to cover the whole of the Cranfield collection. Citation indexing It was in 1961 that the first major grant was given for a citation index (ref. 32), and the following year we were asked for our views on how a citation index could be evaluated. Citation indexing is basically a method of forming classes of documents which are all related through a common reference to a base document. There will, of course, be many occasions when the class consists of only a single entry; however, in the more numerous cases where the class consists of two or more documents, then citation indexing can be considered equivalent to bibliographic coupling at a strength of one. Bibliographic Coupling has been developed by Dr. M. Kessler at the Massachusetts Institute of Technology (ref. 20), and in relation to citation indexing, can be considered as a precision device, since it progressively narrows the class of documents as the demand for common references increases in number. Citation indexing and bibliographic coupling could therefore be tested in the same way as any other device; it was, however, necessary to prepare an index for this purpose. The first stage was to prepare xerox copies of the citations in the 1400 documents in the test collection; against each citation was put the code number for the citing documents, after which each citation was cut up so as to appear on a separate slip of paper. This resulted in some 20,000 slips of many various sizes, which had to be sorted into author alphabetical order. This being done, the slips were pasted onto sheets of paper; where two or more slips related to the same cited document, only one example was pasted in; the references to the additional citing documents were entered alongside. This can be seen in Fig. 7.1, which covers a series of references to papers by H.J. Allen, in particular a paper written with A.J. Eggers entitled 'A study of the motion and aerodynamic heating of missiles entering the earth's atmosphere at high supersonic speeds. ' (NACA TN 4047), This is shown to have been cited by thirteen papers in the test collection. This procedure resulted in a normal citation index; to obtain the index for bibliographic coupling required three further stages. First, each cited reference having two or more citations was givena code number, the paper by Allen and Eggers beingA25, and a separate card was prepared for each cited reference. On this reference card was written the code for the cited document (i.e. A25) and then, in numerical order the codes for the citing documents. Fig. 7.2 illustrates the reference card prepared in connection with the paper by Allen and Eggers shown in Fig. 7.1. The reference cards were sorted into numerical order depending on the lowest number on each card. Since these numbers represented the codes for the citing documents, they ranged from 1001-2400. Each card was then taken in turn, and the information from all reference cards having the same starting number was transferred to a master card. As an example, the master card shown in Fig. 7.3 illustrates the position with regard to document 1067, this number being poated in the top left hand corner. In the column headings are entered the code numbers for the documents which have been cited by document 1067, this information being obtained from the reference cards such as Fig. 7.2, and these being, in this particular case, A25, W32, A23, E23, F90, and 024. In the first column of the master card are entered the document numbers of all other citing papers, this information being again obtained from the reference cards. A tick is put against each number in the column under the appropriate heading