UPPER-BOUNDS FOR RETRIEVAL PERFORMANCE AND THEIR USE MEASURING PERFORMANCE AND GENERATING OPTIMAL BOOLEAN QUERIES - CAN IT GET ANY BETTER THAN THIS

The best-case, random, and worst-case document rankings and retrieval performance may be determined using a method discussed here. Knowledge of the best case performance allows users and system designers to (a) determine how close to optimality their search is and (b) select quer ies and matching functions that will produce the best results. A metho d for deriving the optimal Boolean query for a given level of recall i s suggested, as is a method for determining the quality of a Boolean q uery. Measures are proposed that modify conventional text retrieval me asures such as precision, E, and average search length, so that the va lues for these measures are 1 when retrieval is optimal, 0 when retrie val is random, and -1 when worst-case. Tests using one of these measur es show that many retrievals are optimal. Consequences for retrieval r esearch are examined.