话题检测与跟踪任务中的话题跟踪研究  

Study on Topic Tracking in Topic Detection and Tracking Task

作　　者：张海川 李胜东 石俊涛 吴建伟 王元 ZHANG Haichuan;LI Shengdong;SHI Juntao;WU Jianwei;WANG Yuan(Department of Computer Engineering,Langfang Yanjing Vocational Technical College,Langfang Hebei 065200,China;Department of Mechanical and Electrical Engineering,Langfang Yanjing Vocational Technical College,Langfang Hebei 065200,China)

机构地区：[1]廊坊燕京职业技术学院计算机工程系,河北廊坊065200 [2]廊坊燕京职业技术学院机电工程系,河北廊坊065200

出　　处：《信息与电脑》2022年第11期151-153,共3页Information & Computer

基　　金：廊坊市科技支撑计划项目(项目编号:2021011024)。

摘　　要：话题检测与跟踪(Topic Detection and Tracking,TDT)任务是对互联网热门话题和敏感话题进行信息处理,受到了研究者的广泛关注。其中,它的子任务之一是话题跟踪任务,即跟踪热门话题和敏感话题。话题跟踪任务的关键技术是话题/报道表示模型和文本分类算法。因此,本文主要研究话题跟踪关键技术,分析关键技术的优缺点,并采用话题/报道表示模型表示话题和报道,同时利用文本分类算法判断报道与话题的相关性以跟踪同类话题,采用话题检测与跟踪评测方法评估话题跟踪结果,设计通用的话题跟踪系统。研究结果表明,该系统具有良好的应用前景。Topic Detection and Tracking(TDT) task is the information processing technology of hot topics and sensitive topics on the internet.One of its subtasks is topic tracking.Topic tracking task is used to track hot topics and sensitive topics,which has attracted extensive attention of researchers.It’s key technologies are topic/story representation model and text classification algorithm.We mainly study the key technologies of topic tracking,analyze the advantages and disadvantages of the key technologies,use topic/story representation models to represent topics and stories,adopts text classification algorithms to judge the correlation between stories and topics to track similar topics,uses TDT evaluation method to evaluate topic tracking results,and designs a general topic tracking system,which has a good application prospect.

关 键 词：话题跟踪 文本分类算法 话题检测与跟踪(TDT) 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]