Evaluation measures for interactive information retrieval

摘要: Abstract Several criteria and measures have been proposed used in evaluating interactive IR performance. There is no agreement about what a successful performance or which are the best existing evaluation measure(s). This study aims to identify measure(s) for Twenty of were selected natural environment, involving 40 real end-users from an academic setting with information problems, interacting six professional intermediaries searching large operational systems. These responsible costs their own searches. showed that value search results as whole single measure among selected. Precision, one most important traditional effectiveness, not significantly correlated success. Users appear be more concerned absolute recall than precision. The also identified two basic factors future can account much higher proportion total variance alone can. Seventeen new success categories suggested investigation.