MONO91 Other Potentially Related Research chapter Mary Elizabeth Stevens National Bureau of Standards matching program, applied to logical similarities 0£ texts related as by having various assigned descriptors or citations in common, might provide a basis for generating document surrogates by representing each text in a related group of texts with the words or sentences these texts have in common. 1/ In the case of man-machine interaction during search, it is suggested that the user should indicate the names of selected documentary items which are of particular interest, then: The machine forms an `hypothesis1 about the subset of articles likely to be of interest. It does this by examining all recorded statements common to the ones selected but not to the rejected ones. The weight of different attributes and degree of interest is taken into account. The machine may display this hypo- thesis or another random sample of titles consistent with it, or both." A1 6.6 Machine Assistance in Translations of Subject Content Indications to Special Search and Retrieval Language There are, also, in the areas of directly and indirectly related research, certain programs of research, development, and experimentation which include investigations of possibilities for using machines to assist in the "translation" of textual languages into special intermediate or "documentary" languages. Doyle's use of the inclusion list principle to extract specified content-indicative words and to encode them in his "bigram" index was an early but relatively trivial example. 3/ The work of Williams and her associates, at Itek and elsewhere, 4/ has involved the objectives of determining which of the subject-revealing implications of titles, abstracts and, if necessary, full text, are susceptible to machine detection and manipulation such that the implied as well as the explicit assertions made in a document may be incorporated in a formalized language for retrieval. While Williams, Barnes, Cardin and Levy, and others, have so far approached such tasks primarily from the standpoint of human analytic judgments, Coyaud (1963 [143]) has discussed at least preliminary work looking toward the automation of the analysis of natural language texts for purposes of encoding and organization of the terms and relationships to be used in the "documentation language" known as "SYNTOL" (Syntagmatic Organization of Language), this work has used a corpus based on biblio- graphic abstracts from the Bulletin signale'tique of the Centre National de la Recherche Scientifique, Psychophysiolog[OCRerr] section, for the period 1958-1960. NotwithstandIng such difficulties as determining rules for proper subdivisions of text, reduction of synonyms, resolution of lexical and syntactic ambiguities, and the fact that some words are always, 1/ ZI 3/ 4/ Kochen et al, 196Z [329], p. 2. Kochen et al, 1962 [OCRerr]28], p.7. Doyle, 1959 [168]. See also p. [OCRerr]23 of this report. See, for example, T. M. Williams, R. F. Barnes, 3r., 3. W. Kuipers, various references. 140