An Improved Framework for Online Adaptive Information Filtering

Adaptive information filtering is an emerging filtering technology that can learn the user interest/topic automatically during the filtering process and adjust its output accordingly. It provides a better performance and broader applicability than the traditional filtering technology, therefore is useful in Internet for managing sensitive information and presenting personalized content to Web user. In this paper we propose a new framework for online adaptive filtering, in which two different scoring/weighting and feedback mechanisms are implemented. Based on them, an incremental profile training method is introduced for locating user interest accurately, and a profile self-learning algorithm is also developed for adjusting user focus in test filtering. The experiments in the Reuters online news show our system performs better than the exist systems in the profile training and overall filtering results.