Context-Aware Behavior Analysis For Web Search

Web search engine is one of the most important applications on the Internet.It can help its users to search and access efficiently a large amount of Web information within a finite time and has become an indispensable tool for humans to acquire information and knowledge.With the development of research on Web search engines,the scope of the research has extended from merely focusing on retrieving results that are relevant to user's query to the whole interaction process between users and search systems.By analyzing and understanding user's search behavior,we can provide guidance for the design of search engines,evaluate the performance of search engines in user's perspective,and use user behavior logs to improve the ranking function and user experience of Web search engines.However,with the development of the Internet and the growth of search engine users,the information resources,search engine users,search environments,and search tasks are becoming more diverse,heterogeneous,and complex,a variety of search context factors may affect user's search behavior.For example,first,as the search tasks are more complex,the user often needs to submit multiple queries in a search session to complete them.The session-level context may affect the user's follow-up query reformulations and her interactions with search engine result pages.Second,with the emergence of Mobile Web,more an more users start to use mobile search engines.User's behavior patterns in mobile search will be,in many aspects,different from those in desktop search.Third,with the growth of search engine users,the background of users are increasingly more diverse.Long-term context factors,such as user profiles and users' search history may also influence their search behavior.Existing works on user behavior analysis for Web search engines have not sufficiently considered and analyzed the effect of search context on user's search behavior.To address this problem,we propose to conduct context-aware behavior analysis for Web search in this dissertation.The dissertation consists of four four related but separate works:(1)context-aware evaluation of Web search engines;(2)mobile-environment-aware measure for result usefulness;(3)constructing mobile-environment-aware click models;(4)user-context-aware analysis of complex search.These studies utilized two complementary research approaches,user study and large-scale log analysis,to systematically analyze how three search context factors,session context,search environment,and user profile,affect user's search behavior.The experiment results extended and deepened our understanding of user's search behavior.Based on the understanding of how search context affects user's search behavior,we further develop a user-centric evaluation framework and to improve the ranking function of Web search engines.