Keyword Extraction Based on Implicit Feedback

To improve the results from search engines and make them more personalized for the user, we need to find out about the interests of a particular user. Many of the search personalization methods analyse documents visited by the user and from these documents infer the user’s interests. However, this approach is not accurate, because the user is rarely interested in the whole document; he might be interested in parts of the document only or the document does not have to interest him at all. In this article we analyse the user’s activity on a web site, called implicit feedback. This feedback is represented by the user’s behaviour in the Web browser: time spent browsing, mouse cursor movement, clicking, scrolling etc. Our method is then able to more accurately extract keywords from the documents by putting more weight on the keywords the user is more interested in and disregarding those which did not interest him. We extract keywords directly from the text and also by using traditional methods enhanced by the implicit feedback.