Abstract
Routing and filtering are terms used to describe the process of comparing text streams (such as newswire feeds) to the appropriate users. Although similar to the task of information retrieval, there are significant differences that impact on text representation, the retrieval model, and representing information needs as profiles. In this talk, I will present a model of routing that uses Bayesian networks and a system implementation based on this model. I will also compare approaches to learning profiles from examples of relevant and non-relevant texts. Examples from applications and the government-sponsored TREC evaluations will be used to describe problems and solutions.