Effective browsing technique based on behavioral collaborative filtering on social streams

In recent years, Social Networking Services (SNSs) are growing in popularity, and generating new articles moment by moment. However, when huge article streams are delivered from the SNS, it is not easy to browse them efficiently because users would sometimes skip valuable articles. In this paper, we propose a method to recommend an unread article in order to achieve efficient browsing. Our method estimates the preference of a user on a delivered article based on the browsing behavior of the user, and predicts the preference of each unread article based on the collaborative filtering approach. Our system estimates the value of each unread article for the target user based on the behaviors of users who might be highly similar to the target user's behavior of reading articles, and utilizes the estimation results for composing unread articles into a stream in an appropriate order to realize efficient browsing.