MONO91 Automatic Assignment Indexing Techniques chapter Mary Elizabeth Stevens National Bureau of Standards Swanson and his TRW associates have further proposed extensions of the prespecified unique clue-word technique. For example, it is suggested that machine processes of comparing words of titles, subtitles and chapter headings to lists of possible subject heading can be extended in sophistication by machine lookups of synonym groups and of characteristic subject-word associations. 1/ Frequency weightings may be taken into account, and similar measures of association and subj[OCRerr][OCRerr]t-indicativeness may be developed for phrases as well as for individual words. - In general, however, the apparent success of this clue-word technique in tests to date should be considered in the light of the special character of the items, their extreme brevity, and the high probability that the fact-word incidence involved in news reporting is not typical of less popular and less factually oriented materials 3/ Continuing work along similar lines has been carried forward at Ramo-Wooldridge in the `1Word Correlation and Automatic Indexing Program!! sponsored by the Council on Library Resources (1959 [490] and [491]). Here, the objectives are to develop and apply clue-word techniques to material that is much more representative of the scientific and technical literature. The thesaurus[OCRerr]groups, now called `1indexonym" groups, are made up of words and phrases selected by extensive human analysis as being significantly "useful- for- retrieval[OCRerr]purposes". New items would be processed in a word and phrase lookup operation, with each word or phrase being initially assigned the identifier number codes of all groups to which it belongs. However, unless a particular group 5 number is repeated several times within the space of a few paragraphs, it is not used as the basis for the actual assignment of an index tag. Provision would be made for calling human attention to items having a numbe of words that are not deleted by processing against a "useless-for-retrieval purposes" list, but that are not found in any of "accepted" groups. It is suggested that in this way it should be possible to "ascribe measures of automatically recognizable `newness' to technical articles!!. 4/ 4.2 Maron's Automatic Indexing Experiments By April of 1959, the reports of work at Thompson Ramo-Wooldridge on automatic indexing and related problems submitted for the Current Research and Development in Scientific Documentation series included reference to Maron and a "probabilistic model for the assignment of index tags", as well as to Swanson's continuing projects. 5/ 1/ 2/ 3/ 4/ 5/ Swanson, 1962 [584], p. 469. Swanson, 1963[580], pp. 1-2. See also Mooers, 1963 [424]. Thompson Ramo Wooldridge, 1959 [491], p. 2A. National Science Foundation's CR&D report No. 5 [430], p.34. 93