Each phase of the system is described below in some detail. User interest and document indexing, i e., item tag preparation, may be done externally or by the system. In either case some machinable text, e.g., an abstract on IBM cards, must be provided. Each new document tag will be matched against the user tags for selective dissemination of a notice. Inquiries, received from users, will be matched against the tags to generate notices. In both cases the matches will be augmented by a small random sample of the nonmatches for evaluation purposes. Notices will have two parts, an item description, generated from the input text, and a response card. The latter will obtain a number of ratings regarding the item and allow the user to request, in the case of a document, a hard copy. Inventory control programs will determine the number of copies to print, purge item tags to slower access files and determine which extra document copies to purge from files, based on these user responses.
Unfortunately, ACM prohibits us from displaying non-influential references for this paper.
To see the full reference list, please visit http://dl.acm.org/citation.cfm?id=808516.