CONCLUSION

In this work, we present and evaluate a framework that can take advantage of a topic taxonomy as part of a filtering language. Fuzzy aggregation of the estimated topic probabilities proved to exhibit superior performance than Boolean aggregation. OWA aggregation operators improved fuzzy aggregation in an inversely proportional manner to classification accuracy. Future work includes the study of the proposed framework, including automatically learned OWA aggregation as well as further deployment of the framework on the Web.