Supporting Exploratory Search Tasks with Interactive User Modeling

摘要

This paper presents the design and study of interactive usermodeling to support exploratory search tasks. Contrary totraditional interactions, such as query based search, querysuggestions, or relevance feedback, interactive user modelingallows a user to perceive the state of the user model atall times and provide feedback that directly rewards or penalizesit. The technique allows the user to continuouslytune the system’s belief about the user’s evolving informationneeds. We demonstrate that such functionality isuseful in exploratory search where users need to get accustomedto a body of literature in a domain. We conductedtwo experiments where scientists carried out exploratorysearch tasks with our implementation of an interactive usermodeling and retrieval system (SciNet) and two baselines:SciNet from which interactive user modeling was excludedand a real-world baseline (Google Scholar). The resultsshow that interactive user modeling can help users to moreeffectively find relevant, novel and diverse informationwithout compromises in task execution time.