Analyzing User Actions within a Web 2.0 Portal to Improve a Collaborative Filtering Recommendation System

The current Web manifests the problem of information overload, especially due to the success of the Web 2.0 paradigm, in which users provide new contents quickly. To help people find the most valuable information, many Web sites include a recommendation system based on a rating mechanism. However, such approach cannot be used when a rating mechanism is not present and, in addition, it does not take into account all the actions performed by the users. We propose an extension of the collaborative filtering approach to design a more effective recommendation system that overcomes those limitations.