Search strategy and the optimization of retrieval effectiveness

Future real-time information retrieval systems may be expected to utilize automatic text analysis procedures for the preparation of analyzed search requests, and user feedback information for the generation of a useful search strategy. The analysis procedures and the search strategies to be used will vary to some extent with the equipment used in the system, with the type of service to be furnished, and with the user population. If the user population is large, and service is to be rendered simultaneously to many users, then it is not possible to process each search request against an entire collection of stored items. Instead, a number of partial searches may be used to replace a single full search of the collection. In the present study, various partial search strategies are described, based partly on document and request groupings, and partly on user feedback information. The SMART system is used to evaluate these strategies, and to postulate an efficient, real-time, user-controlled search strategy.