Representing the Search Session Process

Information search is a cognitive process by which users work to satisfy an information need or solve a problem. Information system performance could be improved if it had a grounded belief about the user and their goal. To accomplish this, a system needs to recognize and predict user states and aspects of their task, including search intentions. High fidelity representation of user mental states during search would best capture the user situation and presumably allow for better calculation of actions to take to help the user achieve their task goal more quickly and/or with higher quality outcomes.;While there is research to infer user search context based on their activity with system, such as page use and query submissions, there has been little work to model user search session actions from a cognitive perspective. The dissertation addresses the development of a framework and methodology to represent some aspects of user information processing states in the information search process by exploiting eye movement patterns in the reading information acquisition process. This work is situated within a user-centered and cognitive model of the entire information seeking session.;Information processing states were learned for participants in a user study where tasks were varied by complexity and other factors likely to influence search strategy and tactics. The learned information processing states were examined as patterns of user search actions to investigate whether task differences could be distinguished. In principle, such a model could help to ground system inferences and predictions about the current search state of the user and some aspects of their task goal, i.e. what they are trying to accomplish. The results show tasks can be distinguished using information processing state patterns.;The results are discussed and potential for application in personalization and making simulated users considered. Limitations and directions for research are identified.