MONO91 Appendix B: Progress and Prospects in Mechanized Indexing appendix Mary Elizabeth Stevens National Bureau of Standards The term mechanized indexing can be interpreted in two different ways: as involving the use of machines to produce indexes once the index entries have been pre-determined manually, or as involving the use of machines to select the index entries as well as to prepare the indexes. The first interpretation, that of machine compilation of indexes is perhaps best represented by the progressively more sophisticated mechanization used for the production of Index Medicus from manual "shingling", through sequential card camera operations, to the computer-based system using a high-speed phototypesetter, the Photon GRACE 1, 2/. As noted elsewhere in this report, machine capabilities have made practical the prepara- tion of citation indexes. In general, however, machine-compiled indexes work with the results of human intellectual efforts as applied in the subject content analysis of documents. We also find machines used to provide aids to the indexer. Two different tools may be employed to improve the quality of indexing. There are prescriptive aids in the sense of limiting and rigorously defining- the scope of index terms to be used, and there are suggestive aids in the sense of provoking ideas about additional terms that might be used. The first type may involve a mechanized authority list or thesaurus used to normalize proposed index term entries, as has been demonstrated by Schultz 3/ and Schultz and Shepherd 4/ from 1960 onward. The potential value of this technique is indicated by further investigations of Schultz et al 5/ in which it was found that index terms proposed by authors agreed more with terms employed by more than one member of a typical user group than did terms available in the document titles. Another example of developments in the use of a mechanized thesaurus is the system at Lockheed Missiles and Space Division, Palo Alto 6/. This type of tool is used to check proposed indexing terms against the terms of the system vocabulary, to prescribe choices between synonyms and different levels of spec- ificity, and to supply syndectic devices such as see also" references. Computer manipulations of thesauri can also be used to diversify search questions and to provide useful groupings of terms previously used in the system. The mechanized thesaurus can thus serve as the second type of aid by suggesting to the human indexer additional terms he might use. In effect, such a thesaurus provides a display of prior term-term, document- term and document-document associations observed in a particular collection, such as was demonstrated in the form of special purpose equipment in Taube's 11EDIAC11 7/ and the 11ACORN" devices at A. D. Little 8/. The associational thesaurus can also be used to aid in the resolution of ambiguities of natural language and to provide for updating in the light of changing terminologies or changes in the subject scope of a collection. What are the prospects for automatic updating and revision of a mechanized thesaurus? Luhn 9/ has suggested that a record of the num- ber of times words and groups are looked up would be 11an indispensable part of the system for making periodic adjustments based on the usage of words or notions as mechanically established. Another suggestion for the development of mechanized aids in human indexing proce- dures has been made by Markus 10/. This is to "explore the possibility of applying programmed teaching to indexing, with or without machines. Machine-compiled indexes rest upon the efficacy of human indexing and there is increasing reason to doubt that this will be 11good enough'1 for the future. It appears that there is a growing consensus with respect to inadequacies of present scope and coverage of indexing services. Cheydleur Ll/ emphasizes that: "The cost of manual classification and abstracting of all the articles in the world's hundred-thousand technical periodicals would be fantastic. The practicality of carrying it out in a coordinated and timely way by 223