Event detection based on on-line news clustering

Abstract

In this dissertation, we develop and implementation a news event detection system by using an improved Single-pass incremental clustering algorithm. The objective of our work is to judge whether a current document is talking about the same event as the previous documents. Based on the traditional algorithm, its real-time and dynamic natures are guaranteed, and the improved algorithm solves the problem that the original algorithm is greatly affected by the input sequence. In addition, the new algorithm also improves the accuracy of topic detection. The improved Single-pass algorithm processes the text data by groups and calculates the similarity by average-link instead of maximum value. The experiment part verified that the improved Single-pass algorithm has great performance on Event Detection, with high accuracy and efficiency.