Abstract
We propose a method of information filtering for Web browsing based on a model of user's interests. We have already proposed Context-Sensitive Filtering (CSF) for browsing on CD-ROM encyclopedia, which follows up shifting user's interests in this case. We confirmed the effectiveness of CSF. However, it is not very effective for Web browsing. We assumed that one of the reasons is the difference of pages between the Web and CD-ROM encyclopedia. We describe improvements of modeling to estimate user's interests and filtering method by dividing Web pages. We confirmed the effectiveness of our filtering through an experiment to compare several filtering methods.
