CDW: A Text Clustering Model for Diverse Versions Discovery
The development of information technology brings numerous online news and events to our daily life. One big problem of such information explosion is, many times there are diverse descriptions for one incident which make people confused. Although previous researches have provided various algorithms to detect and track events, few of them focus on uncovering the diversified versions of an event. In this paper, we propose a novel algorithm which is capable of discovering different versions of one event according to the news reports. We map documents to the topic layer to get the information of each topic Then we extract the highly-differentiated words of each topic to cluster the documents. Compared with previous work, the accuracy of our algorithm is much higher. Experiments conducted on two data sets show that our algorithm is effective and outperforms various related algorithms, including classical methods such as K-means and LDA.
Rong Xiao Liang Kong Yan Zhang Min Wang
Department of Machine Intelligence, Peking University, Beijing 100871, China Key Laboratory on Machi Ucap Corporation, DongGuan, Guangdong, China
国际会议
上海
英文
1161-1165
2011-07-26（万方平台首次上网日期，不代表论文的发表时间）