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. - 81 - (3} Term and species: If the basic class were Non circular cylinder and its synonyms (H75), this would be expanded by the addition of Cone cylinder + C.c. bodies + Ellip- tic c. + Elliptic c. of eccentricity ½.[OCRerr] + Hemispherical c. + C. with h. nose + Ogive c. model + Flat faced c. + C. without corners. (4) Term and species (selection}: a choice was made from (3) based on the context of the question asked. For example, in a question on the kinetic theory of gases, when programming the term Gases, only those kinds of gases which reflected in some way the problem of the question were selected - such as Ideal gas, Real gas, High temperature gas, Dissociating gas, Equilibrium gas. (5) Superordinate - i. e., adding to the basic class its immediate containing genus and as many more genera beyond that as appeared sensible; the number of steps included would rarely exceed three. To Non-circular cylinder (H75} would be added, for example, Cylinder + Body of revolution + 3-dimensional body. It should be noted that only the superordinate term was taken - not its species as well; the search is the equivalent of the traditional library search under 'more general' heads. (6) Generic (narrow}. - i.e. adding to the basic class its immediate containing class {genus} and all the other species in the same array (subfacet) as the basic class; e. g., to Non-circular cylinder would be added Cylinder and the rest of the array based on circularity of shape, but excluding those kinds of cylinder (Inconel cylinder, Flat faced cylinder, Long cylinder, etc. ) reflecting other principles of division {Material, Edge properties, Length, etc. ). Similarly, if the basic class were Super- sonic flow, this programme would add to it all other kinds of flow designated by speed, but excluding kinds of flow based on other principles, such as viscosity, compressi- bility, degree of turbulence, etc. (7) Coordinate (selection}: a choice was made from (6) of the most likely terms, but excluding the superordinate term. Since by definition the classes of an array are mutually exclusive this was never a very promising search and in fact was not often productive of any terms. But in those border line situations referred to above, where the concept of generic hierarchy can only be realized practically by accepting a less-than-precise category such as ,characteristics' or 'phenomena', the likelihood was greater; e.g., in a question on Air drag the coordinate class Atmospheric rota- tion was accepted. Another example is that of opposites, or near-opposites, like Laminar flow and Turbulent flow, where a document frequently refers to the one even when its primary subject is the other. (8) Generic (broad}: this added to (6) as many more superordinate terms as seemed reasonable, together with all their species - i.e., not just those restricted to the immediate array (subfacet) in which the basic term appeared. * For example, if the latter were Supersonic flow, this search would now bring in documents indexed by any kind of flow - Laminar and Turbulent, Conical and Parabolic, Equilibrium and Non-equilibrium, etc. This somewhat undiscriminating acceptance of the complete contents of a hierarchy is the equivalent of the 'generic search' as usually understood in machine searching. (9) Systematic Collateral (selection): this was a selection from (8} analogous to the selection from (6) which produced coordinate classes (7) - again excluding the super- ordinate terms themselves. This search was more productive than (7) since there is often a close connection between concepts from different arrays of the same genus. This fact underlies the correlation of properties and the principle of definition by