Detecting Search Sessions Using Document Metadata and Implicit Feedback

It has been shown that search personalization can greatly benefit from exploiting user’s short-term context – user’s immediate need and intent. However, this requires that the search engine must be able to divide user’s activity into segments, where each segment captures user’s single goal and focus. Several different approaches to search session segmentation exist, each considering different features of the queries, but it may be helpful to also consider user’s implicit feedback on the search results clicked in response to the query. We propose a method for segmenting queries into search sessions which is based on document metadata and incorporates implicit feedback. Our approach also considers multitasking, where user shifts her current interest, but afterwards proceeds with the original task. We evaluated our approach on manually segmented query log and compared the results of our approach with results from other methods and showed that using implicit feedback can improve the performance of the segmentation task.