News event tracking using an improved hybrid of KNN and SVM

Abstract

News event tracking is the task of associating incoming stories with events known to the system. A tracking system's goal is to automatically assign event labels to the subsequent news stories. The paper presents an improved fusion algorithm for news event tracking based on the combination of the KNN and SVM. The improved KNN utilizes density function to select some cluster centers from negative examples and the improved SVM uses sigmoid function to map the SVM outputs into probabilities. The problem of effective density radius selection is discussed, and the performance differences between the event tracking method proposed in this paper and other methods are compared. The experimental results with the real-world data sets indicate the proposed method is feasible and advanced. © 2009 Springer-Verlag Berlin Heidelberg.