CRANV1P1 Formation of Index Languages 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. - 77 - consisting often of a sole member; e. g., Stage efficiency is evidently a property and Stage performance a process - but they are treated as virtually synonymous; Cascade losses constitute a factor in efficiency or performance, but hardly a 'kind of efficiency'. Stage matching is another concept which lies on the borderline between processes and properties. It is possible to say, however, that all these rather subtly related notions are 'Stage characteristic[OCRerr] and in this way the facet structure is maintained without undue complexity. Other examples of a certain amount of violence being done to the strict nature of generic relations may be found, as at M4/33 (see Appendix 5.3) where complexly related terms are grouped as Atmosphere properties and characteristics, or at P7/26 Processes and properties of Vortices. Similar situ- ations inevitably arose in the single-term hierarchies in areas like Mechanics and Dynamics where the conditions of what Ranganathan has called 'canonical' classi- fication tend to hold. Another minor liberty, not demonstrated in the example above, was taken in the treatment of qualifying terms like Theory, Approximation, Experimental data, when these were found precoordinated as in Hypersonic flow approximation. It could be argued that these do not narrow the extension of the term they qualify and should therefore be disregarded - i.e., treated as synonymous with the term alone. Theo- retically, this is why they are usually placed (in the guise of 'Form divisions'} at the very beginning of the subdivisions of a term in conventional classification. In the search programmes, however, they were included in the 'Terms and species' sub-programme. This was later seen to be a mistake, but it is not thought that this was serious in view of the very few terms involved. Although the differently related facets follow and interrupt each other without clear signs of demarcation in the schedules, the different relations were strictly ob- served, of course, when the search programmes were compiled; i.e., when expanding a class by generic hierarchy, only those terms standing in a true generic relation to that class were counted; e.g., Compressor + Centrifugal c. + Axial flow c. + Dru.'[OCRerr]: construction + Disk construction + Jumo 004 + Single stage c. + ... would be gSve:[OCRerr] as the full generic expansion of a partict[OCRerr]lar kind of compressor (see the Generic - Broad search below}. Any terms not standing in a true generic relations (e. g. , Axial flow c. blade, Stage characteristics, Irrotational flow, etc. ) would be ignored. Multiple hierarchical relations The major weakness of the linear display of classes just described, in which a particular class (concept) is located in one place only (albeit a carefully chosen one) is that it fails to show the further generic relations a class may have. For example, Jet interference is subordinated to a category Jet characteristics, in which its class mates are Jet exit, Jet location, Jet energy, Jet structure, Jet emission, etc. It could equally well be subordinated to a category of Causes of interference with class mates like Wake interference, Forebody interference, Support system interference, Wave reflection interference, Wing-Body interference, etc. But in the schedule described these last terms are 'distributed relatives'. There are two traditional methods of handling this problem in a real-life classi- fied index: by multiple-entry as with UDC, where the number of entries for a given compound-class-description are multiplied, and filed according to a different classi- fication, so that Jet interference appears in a class Jet (divided into Jet.characteristics} and also in a class Interference (divided into Causes of interference). Or, by leaving these other connections to be indicated by an A/Z relative index, in which all the