WikiGaze: Gaze-based Personalized Summarization of Wikipedia Reading Session

Abstract

Wikipedia is an open-content encyclopedia that receives billions of page views per month. It has been observed that in a single reading session, Wikipedia users visit multiple articles. To reduce the problems of overload and loss of information, there has been a growing interest in the research community to develop new approaches to present the only necessary information to the users. Automatically generation of personalized summaries is a proven remedy for the information overload problem. In this paper, we propose a technique to generate personalized summaries for Wikipedia articles by analyzing the reading patterns of users. To perform reading pattern analysis, we track eye gaze during the article reading session. Eye gaze analysis helps in identifying the attention distribution of a reader over an article. We extend the proposed approach to generate a summary for multiple articles visited during a user's Wikipedia reading session. We capture a dataset representing the reading pattern of Wikipedia users. We make this dataset publicly available for research community1.