Web user search pattern analysis for modelling query topic changes

Web search engine logs are a good source of information for Web user modeling in which user session analysis is often incurred. However, studies on Web logs assume a user session to cover the complete time period of the data set. In the absence of any further information, we define a user session to be related to the user search topics. Viewing sessions in this way can help overcome problems due to varied approaches in session delimiters. The study in this paper is based on a large corpus of Excite search engine logs. Human expert analysis was performed to identify topic changes. The distribution of topic changes across users is presented. In this paper, we also describe an automatic session detection method on the same logs. For this, we use temporal information in grouping successive user search activities with respect to a user search topic. We then compare these results with human judgements and analyse the errors incurred. These results provide a comparison with other studies on Intranet Web search engine logs.