A MULTILEVEL APPROACH TO INTELLIGENT INFORMATION FILTERING - MODEL, SYSTEM, AND EVALUATION

In information-filtering environments, uncertainties associated with c hanging interests of the user and the dynamic document stream must be handled efficiently. In this article, a filtering model is proposed th at decomposes the overall task into subsystem functionalities and high lights the need for multiple adaptation techniques to cope with uncert ainties. A filtering system, SIFTER, has been implemented based on the model, using established techniques in information retrieval and arti ficial intelligence. These techniques include document representation by a vector-space model, document classification by unsupervised learn ing, and user modeling by reinforcement learning. The system can filte r information based on content and a user's specific interests. The us er's interests are automatically learned with only limited user interv ention in the form of optional relevance feedback for documents. We al so describe experimental studies conducted with SIFTER to filter compu ter and information science documents collected from the Internet and commercial database services. The experimental results demonstrate tha t the system performs very well in filtering documents in a realistic problem setting.