MONO91 Other Potentially Related Research chapter Mary Elizabeth Stevens National Bureau of Standards "Te ....... uses, as quasi-descriptors, word-sets chosen from the Oxford English Dictionary (e.g. , any word falling between A-Ah) and relies on the subsequent correlation of terms to make sense of his seemingly bizarre choice." 1/ Lefkovitz is concerned with the so-called "automatic stratification" of a file in which both generic or associative relationships and exclusive partitioning is used to facilitate search. He claims: TI... The exclusive partitioning implies a separation of descriptors into groups such that no two descriptors in a group co-occur in any given document description of the file. This arrangement presents the dissociative properties of the file, or forbidden combinations. When coupled with a superimposed display of the `inclusive' or associative properties of the file a unique classification of the descriptors of this file results, which is based solely upon the association of the descriptors themselves within the document descriptions and not upon an arbitrary set of classes constructed by professional indexers." 2/ The purpose is to assist the searcher by warning him that if he chooses more than one descriptor from any one group as terms in his search request, there will be a null response from this particular file. However, the particular application considered involves a limited number of highly quantifiable or scalable "attribute-value" pairs, (for so the descriptors involved are defined), such as "Age-23", and "Hair-red". It is by no means obvious that comparable exclusive partitionings could be achieved for literature items or that the recomputations necessary as new items enter the file can be achieved on a practical basis. 6. OTHER POTENTIALLY RELATED RESEARCH In this section we shall consider certain areas of potentially related research that may prove applicable to the improvement of automatic indexing techniques. First is the area of thesaurus construction and use, which in turn is somewhat related to the develop- ment of statistical association techniques, especially for "indexing-at- time-of-search" and search renegotiations. Natural language text searching will also be briefly considered, together with related research in the general area of linguistic data processing. 6.1 Thesaurus Construction, Use, and Up-Dating The first area of potentially related research which promises improvements in automatic indexing procedures is that of thesaurus lookups by machine. There are several different possible definitions of the word "thesaurus" in the context of informa- tion storage, selection and retrieval systems. The first is that it is a prescriptive indexing aid, or authority list, serving the function of normalizing the indexing language, primarily by the use of a single word form for words occurring in various inflections, by the reduction of synonyms, and by the introduction of appropriate syndetic devices. The second definition relates to the intended function for the provocation and suggestion to the indexer or the searcher of additional terms and clues, and it follows the idea of word groupings related to concepts as in a traditional thesaurus like Roget's. The third 1i 2/ Cleverdon and Mills, 1963 [131], p. 8. Lefkovitz, 1963 [353], Preface, pp. VIII-IX. 1[OCRerr]4