MONO91 Other Potentially Related Research chapter Mary Elizabeth Stevens National Bureau of Standards 1/ under the general heading of the association map technique, - although passing reference has been made to some of Doyle's suggestions and findings elsewhere in this report. Beginning in 1958 (Doyle, 1959 [168]) information retrieval projects at the System Development Corporation have had, among other objectives, that of developing ways to use computers in the processing and interpretation of natural language text. By February of 1959, a computer program was already in operation that could search fragments of about 100 words of keypunched text, match input words against a pre-established clue word selection list (i. e. , an inclusion dictionary) and substitute a short encoded form to be used for subsequent search. Processing of keypunched abstracts using this program in- volved computer time at the rate of four abstracts per second. Other features of this text compiler, and of subsequent text processing programs developed at SDC, enable the making of frequency counts and other statistical measures. Such features are then used for the investigation of, for example, word-word, word- document, and word-subject associations, looking toward the determination of answers to such questions as: "Do subject words have distribution characteristics within a library that a computer program can detect?" Doyle's investigations of word co-occurrences have included hypotheses and tests of various probabilistic measures in terms of observed frequenci[OCRerr][OCRerr], in terms of "boingi" words (so-called because of the mental sound effect they elicit), - in terms of adjacent word pairs and affinities between particular nouns and particular adjectives, !` and in terms of distinctions between frequency (the total number of times a word appears in a give nlibrary corpus) and prevalence (the total number of items in which a particular word appears). Si He has also stressed distinctions between adjacent words and high corre- lations for words that are not closely positioned together in text, as follows: 1/ Compare Doyle himself, 196Z, [163], p.383: "Swanson and others have offered thesauri of synonyms and related terms... (to assist in indexing or search processes)... An association map is, in a sense, an extension of this solution; it is a gigantic, automatically derived thesaurus. Confronted by such a map, the searcher has a much better `association network' than the one existing in his mind, because it corresponds to words actually found in the library, and, therefore, words which are best suited to retrieve information from that library." See also Wyllys, 196z [651], p. 16: "L. B. Doyle (1961) has invented a fascinating search tool which seems to us to belong at a level intermediate between automatic indexes and auto- matic abstracts; i.e., a possible search method might be to have the computer scan automatic indexes and compare the index terms therein with the request, then obtain the possibly pertinent documents and display their association map for the user to examine..." ZI 3' 4/ 5' Doyle, 1959 [168], p.6. Doyle, 1959 [165], p. 5. Doyle, 1961 [169], p. lZ; 1959 [165], p. 16. Doyle, 1962[163], p. 380. 1Z3