MONO91 Other Potentially Related Research chapter Mary Elizabeth Stevens National Bureau of Standards 1/ th[OCRerr] southwestern states.- A joint American Bar Foundation--IBM research program has been established to explore both text searching without prior indexing and automatic in- dexing techniques (Eldridge and Dennis, 1962 [183], 1963 [182]). In the Horty-Pittsburgh System, approximately 6, 000,000 words of text have been converted via Flexowriter to magnetic tape. An exclusion dictionary of 100 words is used to eliminate the most common words and a word-concordance is prepared, resulting in word-occurrence location indicia by position in sentence, paragraph and section of the statute. In searching, the user has available to him the alphabetized list of approximately 17,000 different words and it is up to him to think of the words and synonyms most likely to occur in statute sections likely to be the ones he seeks. Several search logics are available. One provides that at least one of a group of alternate words must appear; another requires that at least one from two or more groups must appear in the same sentence. Intra-sentence distance criteria are also utilized: "If the phrase `born out of wedlock' is sought, the operator... requires that the word `wedlock' appear in the same sentence, no more than three words after `born'." Obviously, for the same question the searcher would also have to specify synony- mous words and phrases- -"illegitimate children", "illegitimate births", "unwed mothers", "unmarried mothers", "illegitimacy", "bastardy", and so on. The reported success of the system is apparently due in large part to the ingenuity of the searchers in specifying the expressions and synonyms most likely to be used. Hughes comments as follows: "It should be noted that this system will be most efficient only whe'n the users are thoroughly familiar with the linguistic style of the source material and search is made on words known to occur in the appropriate statutes" 3/ 6.5 Other Examples of Related Research in Linguistic Data Processing Since, as Garvin has emphasized, "All areas of linguistic information processing are concerned with the treatment of the content, rather than merely the form, of docu- ments composed in a natural language," [OCRerr] much of the research in linguistic data processing is potentially applicable to both the development and the improvement of automatic indexing techniques. Thus developments in automatic content analysis, in psycholinguistics, in question[OCRerr]answering systems, may eventually find application to mechanized indexing systems. 1/ 2/ Horty, 1962 [278], pp. 59-60. 3/ Eldridge and Dennis, 1964 [182], p. Hughes, 1962 [284], p. IV-6 to IV-8. 90; Wilson, 1962 [645]. 136