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. - 44 - The importance of the genus/species, or thing/kind relation needs little explan- ation. Since Aristotle, it has been the central relation considered in traditional logic. In information retrieval its use as an indexing device ('hierarchical linkage,) is that it provides an automatic relevance network based on the inclusion relation, e. g., if laminar boundary layer is a kind of boundary layer, then documents on the latter are likely to have some degree of rel4vance (possibly a high one) to questions on the former. It is also the basis of the notion of 'specificity', which is a fundamental parameter determining precision in index performance. It might be argued that the [OCRerr]:en-generic relations displayed in all library classifications also possess an 'h[OCRerr]clusion, quality so far as documentary classes go; e.g., a precoordinate class with the index description Delta wing - Drag could be said to be 'included' in the class Delta wings just as much as the true species Sweptback delta wing is included. But the inclusion relation of the first example is weaker, in tr..at it is a shared one; Drag on a delta wing is as much included in Wings-Drag as in Delta wings, whereas Sweptback delta wing is entirely included in the class Delta wing. That is to say, the non-generic relation is essentially one of conjunction ('coordination,) rather than complete inclusion. Certainly, no other single relation rivals it in ubiquity or importance for retrie val, although collectively the other relations contribute significantly to class defi- nition in indexing. In precoordinate classified indexes it is normal in question- programmir.g to move from the terms of a category or facet, each one constituting a hierarchy, to the particular term which gives rise to the facet; e.g., to move from Conductivity (of titanium) in the Properties facet of the class Titanium, to Titanium. The object of systematically arranging classes in such indexes is to provide a per- manent and constantly available mechanism for manipulating the classes in this way, and not only by movement within a strict genus/species hierarchy. Classes are ex- panded or contracted by moving also from one category to another. For example, in a question on compressor operation, the class examined may be expanded by moving from particular parts of the compressor (Blade, Shroud, etc. ) to particular processes (Stage interaction, Stage stall, etc.) or to particular characteristics of these (Inlet blade angle, Stall limit line, etc.) In fact, the term 'hierarchical, is frequently used as a synonym for the process of subordination which is the essence 0f a precoordinate system. In this view of hier- archical linkage a chain such as Delta wing - Sweptback - Transonic speed - Low angle of attack - Lift is regarded as reflecting a hierarchy every bit as much as a true genus] species chain like Metals-Titanium, or Aircraft-Heavier,than air-Monoplane-with Delta wing. It is possible to go even further and refer to the precoordination of an Alphabetical subject catalogue (which, of course, subordinates some terms to others in its subject headings, e.g. Aircraft-Design) as a hierarchical system (Ref. 21). Even if this last extreme view of the term 'hierarchy, is rejected, we are left with the fact that the identification of hierarchical linkage with the full range of rela- tions displayed by a highly organized library classification system does not give us a basic device which can be measured in the same way as can coordination, weighting, etc., but is a varied mixture of relations capable of defining classes. In traditional library classification it is well established that any relationship may be used to establish 'subclasses, of a given class. In the sense that library classification deals with classes of documents, and that documents on any aspect whatsoever of a subject x can be regarded as belonging to some subclass of 'Docu- ments on x', then Extrusion of plastics is just as much a subclass of Plastics as