新闻文本话题关联识别研究  

Research of the Story Link Detection in Topic Detection and Tracking

作　　者：张晓艳[1] 王挺[1] 陈火旺[1] 

机构地区：[1]国防科技大学计算机学院,湖南长沙410073

出　　处：《计算机工程与科学》2008年第4期131-134,138,共5页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60403050)

摘　　要：本文首先简单介绍了话题发现与追踪的发展史及主要任务,然后重点分析比较了现有话题关联识别模型的使用情况,并测试了文档中各层次知识特征对话题关联识别性能的影响。其测试结果对如何评判已有的识别系统或构建好的识别系统具有一定参考价值。As a key technology of topic detection and tracking, story link detection is always a issue from the beginning. There will be a great help if we can improve the link detection. At first we briefly introduce the history and primary tasks of topic detection and tracking. Secondly different kinds of link detection systems will be analyzed and compared. Thirdly we test the usefulness of different knowledge features. Finally, we provide the hypothesis of constructing a good link detection system.

关 键 词：话题关联识别 话题发现与追踪 知识特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]