Search path visualization and session performance evaluation with log files

Abstract

This paper discusses new strategies for the performance evaluation of user search behavior. For the Log Analysis for Digital Societies (LADS) task of LogCLEF 2009 are proposed three different levels of user performance: success, failure and strong failure. The goal is to compare and measure session performance on a qualitative as well as on a quantitative level. The results obtained with both methods are in good agreement. Primarily they show that it is possible to investigate user performance from interpreting the user interactions recorded in log files. Both the qualitative and quantitative results give rise to a refinement of our operational definition of performance. © 2010 Springer-Verlag Berlin Heidelberg.