摘要
针对主题检测和追踪的第五个技术任务连接分析,提出了一种事件主题相似性分析技术·通过引入领域知识库,将基于词汇的分析技术提升到领域知识计算层面·当输入不同两个文档时,采用该分析技术进行识别文档内容所涉及到的事件主题是否一致·首先采用FIFA模型进行内容主题识别,然后采用LDM模型进行事件主题相似性计算分析·实验结果显示主题相似性计算正确率为64%,召回率为69%·
An event topic similarity computation technique for the fifth TDT techniquelinking detection was put forward. Domain knowledge base is used to upgrade topic similarity computation technique from lexical analysis to domain knowledge analysis. The linking detection technique is used to decide whether two documents (or stories) are on the same event topic. There are two steps in the analysis procedure which includes using FIFA modal to identification content topic, and using LDM modal to compute event topic similarity. Experiment results show that precision and recall percents of the model are 64% and 69%.
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2003年第11期1041-1044,共4页
Journal of Northeastern University(Natural Science)
基金
国家自然科学基金
微软亚洲研究院资助项目(60203019).
关键词
内容主题识别
事件主题分析
主题检测和追踪
领域知识
连接分析
content topic identification
event topic analysis
topic detection and tracking
domain knowledge
linking parsing
