ASIS&T 2006 START Conference Manager    

Identification of User Sessions with Hierarchical Agglomerative Clustering

G. Craig Murray, Jimmy Lin, Abdur Chowdhury

ASIS&T Annual Meeting - 2006 (ASIS&T 2006)
Austin, Texas, November 3-9, 2006


We introduce a novel approach to identifying Web search user sessions based on the burstiness of users’ activity. Our method is user-centered rather than population-centered or system-centered, and can be deployed in situations where users choose to withhold personal content information. We adopt a hierarchical agglomerative clustering approach with a stopping criterion that is statistically motivated by users’ activities. An evaluation based on extracts from a popular search engine’s logs reveals that our algorithm achieves 98% accuracy in identifying session boundaries compared to human judgments.

START Conference Manager (V2.52.6)