A content-based approach for detecting users' shift of interests

Adaptive Navigation Systems (ANS) are intended to support search in hypermedia environment. On the Web, ANS cope with the lack of user feedbacks with unintrusive methods to detect and understand the users' searching behaviours. This article tackles to the problem of detecting Web users' shifts of interests. Most existing methods focus on implicit interest indicators like the time spent on a page or the amount of page read. However, previous researches on these indicators have shown they do not convey enough clues about the users' interests. This paper presents a new unintrusive method to detect the shifts of focus using the content of the accessed pages. The proposed approach is based on the assumption that the issue of detecting the shifts of focus comes down to detecting paragraph boundaries in a very large document made up with the content of the accessed documents. The article also shows how an existing shift detection algorithm can also be considered as a text-segmentation approach. The two approaches are compared with an evaluation with real-users.