MONO91 Automatic Assignment Indexing Techniques chapter Mary Elizabeth Stevens National Bureau of Standards 4.8 Other Assignment Indexing Proposals A few additional automatic assignment indexing proposals are under development. Examples for which experimental data is not as yet generally available include, for example, work at EURATOM, sonie preliminary experiments at Chemical Abstracts Service, work at General Electric, Bethesda, the proposed "Multilindex" system of Information Systems, Inc. , investigations by Slamecka and Zunde, and a special purpose development project at Goodyear Aerospace. Meyer-Uhlenried and Lustig report for the EURATOM developments as follows: .... . Procedures are being developed which allow based upon given keyword lists first for abstracts: (a) to assign significant keywords and (b) based upon hierarchically organized keyword lists, to assign the documents in question to specific subject fields. "Experiments were made at first on narrow fields with so-called micro- thesauri, they showed encouraging result[OCRerr] when automatic and manual assign- ment were compared. Positive results depend of course on the quality of the abstracts and the significance of the words employed in them. It remains to see how far this favorable prognosis is confirmed by keyword collections of more complex contents." 1/ Friedman and Dyson (1961 [203]) have reported on manual experiments designed to relate words occurring in a sample of abstracts from a particular section of Chemical Abstracts to the title or heading for that section. Significant words in these abstracts were counted and the number of occurrences as well as the number of different abstracts in which they appeared were determined, with a rank order listing as a result. It appeared, from inspection, that it should be feasible to develop, for each CA section, a relatively small vocabulary of words that would be descriptive, and indicative of, the subject matter contained in it. They conclude: "In our opinion, the results were signifi- cant, the small vocabulary of words did select a large percentage of the abstracts in the section it was based on." 2/ A project at Information Systems Operations, General Electric, on possibilities for automatic indexing and abstracting of text has been reported in the November 1962 issue of Current Research and Development.3/The META project (Methods of Extracting Text Automatically) is said to be concerned with the use of statistical, linguistic, and semantic criteria for analysis and selection of significant words and significant sentences from text. Computer programs are being developed in modular fashion for the GE-225 computer. 1/ 2/ 3/ Meyer-Uhlenried and Lustig, 1963 [417], p.229. Friedman and Dyson, 1961 [203], p. 10. National Science Foundation's CR&D report, No. 11 [430], p.. 97. I OS