CRANV1P1 Indexing Procedures 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. -47 - by a mixture of the devices already described - by hierarchical linkage, by confound- Lugword forms, etc. "L'Unite" uses none of these, although the last-mentioned will in fact normally be a prominent accompaniment of its class definition, since different word forms will usually appear in the same alphabetical cluster. So it must be ad- mitted that it forms a discrete index device in its own right. The above eleven devices may be compared with the list of techniques for con- trolling index languages given by B. C. Vickery in his book 'On retrieval system theory' (Ref. 9 ) (see Chapter 1). Of those listed above, three devices, namely Bib- liographic coupling, Associative machine indexing, and L'Unite, were not mentioned by Vickery. The others include all the techniques given by Vickery, since a number of these were variants of the more broadly defined devices above. We have tried to distinguish the basic device itself, as a method of class definition, from the different ways in which it might be implemented in different index languages. The latter may be regarded as different amalgams of the various devices, with fur- ther differences resulting from the various methods of file organization. The different ways in which coordination is applied in preeoordinate and post- coordinate systems, have already been mentioned. The fundamental difference is that in the former a limited number of coordinations are made and the resultant com- pound headings are then filed in linear order, and rules observed (as to citation order, etc. } to allow determination of the exact position of any particular combination. This difference has repercussions for the other devices. Even in a largely single-entry pre- coordinate system, for example, 'weighting' of an elementary kind is implicit in the restrictions placed on the number of entries which can be recognized. Links are fundamental to a precoordinate system by 'partitioning', e.g., separate entries would be made for Copper-Hardness and for Titanium-Conductivity. Roles are indicated in a precoordinate system by citation order or by explicit syntactical devices; in a faceted index an index description such as Wings - High aspect ratio - Drag - Low angle of attack conveys by its citation order of Thing (Wing, etc. ) - Property (Aspect ratio, etc. ) - Process (Load, Drag, etc} - Condition (Aerodynamic parameters, Angle of attack, etc. ) that High specifies Aspect ratio and Low specifies Angle of attack. In an Alpha- betical Subject Catalogue, a similar function is served by such headings as Children in art, Pressure vessels - Heat transfer, Acceleration - Psychological effects. Synonyms are treated with varying strictness of interpretatio[OCRerr] in different systems. Often, it is a reflection of the degree of specificity sought, as when one system dis- tinguishes Potential flow from Irrotational flow and another confounds them. The problems of synonymity occurring at the level of multiple term descriptions raises a particular problem for post coordinate systems, since it implies a degree of pre- coordination at some point in the system; e.g. a Ground effect machine is a synonym for Air cushion vehicle, although there is no synonymity between the individual con- stituent terms. Confounding of word forms may be implemented at the indexing stage, via a con- trolled vocabulary (e. g. , Conducting see Conduction}, or by search rules (e. g., accept Conducting + Conduction + Conductor}. In a precoordinate system, where the dif- ferent forms are separated in different categories, confounding may be possible only at the search stage; e.g. , by consulting the A/Z index of a classified index and ob- serving the variant forms used.