Theoretical Considerations in Information Retrieval Systems Information storage and retrieval systems are composed of three major components: (a) identification of information and tagging it for effective retrieval, (b) searching strategy, how to enter the file to circumvent the scanning of nonrelevant material, and (c) file organization to make access to information efficient. For identification of information the paper suggests that a metalanguage (recently discussed in a paper by Goffman, Verhoeff and Belzer) associated with an object language be used. For searching strategy, a linear model for an evaluation function of relevancy is developed which rewards the system for retrieving relevant documents and not retrieving the nonrelevant, and penalizes the system for the escaped relevant documents and false drops. The inadequacies of a linear model are indicated. Two approaches to file organization are discussed. One is self-organization of the file based on its history and past performance, and the second is a self-generating subset of the file with a high probability of being relevant. CACM July, 1964 Belzer, J. Goffman, W. CA640715 JB March 9, 1978 7:55 PM 1032 5 1032 1032 5 1032 1032 5 1032