ISR11 Design Consideration for Time Shared Automatic Documentation Centers chapter M. E. Lesk Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. X-3 of collection selection and updating, occur only in real-life systems. Knowledge of these problems as they affect fully automatic documentation Systems are almost nil. They are analogous to problems that must be faced in large nonautomated documentation systems, but there exist also rrany differences. For example, in many nonautomated systems, the user [OCRerr]resents his query to an intermediary who translates it into some index language which is used by the system. In fully automatic systems however, this step is avoided, since the user queries the system directly using the natural language. The relationship of the user to the direct, fully automatic system, is thus clearly different and must be studied before the best user service techniques can be discovered. Such problems have not been studied before because an operating fully automatic documentation center is needed to study them. Ovcr the last few years, methods have however become available for automatic content analysis; these methods have been carefully studied on small collections, [OCRerr]Tith 1[OCRerr]nown performance capabilities, and can now be applied to larger collections in operational environments. The best reason for studying the design of fully automatic infor- mation retrieval centers at the present time is, however, the fact that small systems have finally demonstrated that some mechanized analysis techniques perform more effectively than many manual analysis techniques; it may therefore be expected that fully automatic systems would probably offer better service to real users than present-day manual or semi-manual systems. It may also be hoped that the construction of a real,operating, fully automatic system would dispel the general skepticism about the use- fulness of automatic information retrieval.