Evaluation of a Personal Information Agent derived from the Context Modelling of Evolving Information Needs

This paper presents a pilot study of a novel context modelling method for capturing an evolving model of information needs during search tasks to assist users in satisfying their long-term needs. Short-term sessions are clustered using terms extracted from explicitly marked search results to create and update aspects of a user profile. The terms that most significantly define the aspects of a user profile are determined using the Ostensive Model and these aspects are used to create a Personal Information Agent to help users satisfy their long-term needs. A novel evaluation technique is introduced comprising four participants performing two simulated work situation tasks over three search sessions where user questionnaires and log file analysis measure the performance of the context modelling method.