Query Cluster: A Method For Web Search Behavior

There have been intensive research on user web search behavior since the late 1990s. Previous researchers collected data from search engines and analyzed explicit data (queries) to understand the characteristics of the user's search process, while other researchers analyzed the data of recruited subjects under experimental settings to understand the behavioral patterns in web usage. Although these researches provided an understanding of what users are searching for and how they are searching, both approaches did not provide rich user contexts that capture the reason why users are motivated to search, how long users' tasks (session) last, and other factors affecting user's search behavior. In this paper, we propose 'Clustered Query' as the unit of analysis in web search behavior studies. We found that users make their own Clustered-queries that yield better overview on their web search pattern, yet detailed individual web traces intact. The methodology consists of three phases and Log Catcher, Query Cluster, Monitoring tool, and Retrospective Interview technique are used in each phase. At the end of this paper, we also illustrate the process of the pilot and main study where the methodology is modified and validated.