An analytic measure predicting information retrieval system performance

摘要: Abstract The performance of information retrieval, hypertext linkage, and text filtering systems may be measured by using historical data or estimating Bayesian probabilistic artificial intelligence methods. measurement is necessary to evaluate document retrieval systems, electronic mail filters, office and, in general, from databases when the searcher has incomplete about characteristics records retrieved. We provide a method for precision quality without examining individual database documents. This requires knowledge only query expressed need set parameters constant all queries. concepts historic expected are examined. analytic measure used examine system relevance feedback increase accuracy parameter estimates. Use precision-document graph instead commonly precision-recall examined, several uses computer-human interface suggested, including its use as graphic aid assisting users deciding stop searching. An economic model stopping problem provided, conditions examined under which user should searching either not feedback.