Automatic Identification of Information Searcher Intentions During an Information Seeking Session

There is a clear and increasing realization that a large proportion of information seekers using search engines and other information retrieval (IR) systems engage in “search sessions” or “information seeking episodes”, that comprise sequences of behaviors beyond that of a single query and the resulting Search Engine Result Page (SERP). Although there is a substantial literature on the possible goals or intentions that information seekers might engage in during such search sessions, to date there has been little or no substantive research on how to identify such intentions (or even to confirm that such intentions are indeed evident in information seeking). We propose a project to relate very low-level behaviors during information searching sessions to progressively higher-level behaviors and then to information seeking intentions. The results will be the association of sequences of behaviors such as eye fixations and mouse movements with specific intentions such as learning about a topic, evaluating relevance of retrieved items, reformulating queries in response to search results, and modification of search goals. This will lead to a model of the information seeking process, and a means both for evaluating IR system support for the various steps in this process, and for proposing, and testing methods for such support of the individual steps, and of the process as a whole.