在线新闻主题检测系统的设计与应用  

Design and Application of an On-line News Topic Detection System

作　　者：万小军[1] 杨建武[1] 

机构地区：[1]北京大学计算机科学技术研究所,北京100871

出　　处：《华南理工大学学报（自然科学版）》2004年第z1期42-46,共5页Journal of South China University of Technology(Natural Science Edition)

摘　　要：利用主题检测技术可以从海量新闻信息中实时检测到主题信息,从而将新闻信息按照主题组织并加以利用.文中通过改进加窗策略,采用自适应倒排文档频率,设计了一个中文新闻主题检测系统并进行了实验.结果表明了该系统的有效性.该系统在新华网数据中心的成功应用进一步表明系统达到了实用需求.Topic detection technique can be used to detect news topics from a great amount of news stream, and help us organize and utilize news information according to their topics. In this paper, a Chinese news topic detection system is designed and tested by using the improved time window strategy and adopting the self-adaptive inverse document frequency. It is then indicated that the proposed system is effective. The application of the system in the data center of XinHua Net further illustrates that the system meet the application demands well.

关 键 词：主题检测 增量式聚类算法 加窗策略 自适应倒排文档频率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]