DATA FILE SIZE AND ITS RELATION TO THE BAYESIAN EFFECTIVENESS OF AN INFORMATION RETRIEVAL SYSTEM
Interim rept. for 16 Feb-15 Apr 1965
DUNLAP AND ASSOCIATES EAST INC NORWALK CT
Pagination or Media Count:
A simple Bayesian measure of system effectiveness for information retrieval systems is proposed. The measure combines the recall and precision ratios of an information system with the utility structure of the system user. Using the measure, it is possible to show that effective systems are possible only under a very narrow set of conditions. In particular, it is shown that using present state-of-the-art indexing, it is not possible to have effective systems with file sizes much in excess of 100,000 documents.
- Computer Hardware