Information Filtering in Changing Domains

The task of information ltering is to classify documents from a stream into either relevant or irrelevant according to a particular user interest with the objective to reduce information load. When using an information lter in an environment that is changing as time proceeds, methods for adapting the lter should be considered in order to retain the desired accuracy in classiication. We favor a methodology that attempts to detect changes and adapts the information lter only if inevitable in order to minimize the amount of user feedback for providing new training data. Nevertheless, detecting changes may require costly user feedback as well. This paper explores a method for detecting changes without user feedback and brieey discusses strategies for adapting information lters. Empirical results with two simulated change scenarios based on real-world text data show that our adaptive information lters perform well in changing domains even without user feedback.