基于层叠模型的话题检测方法研究  

Topic Detection Research Based on Cascade Model

作　　者：谢林燕[1] 张素香[2] 戚银城[1] 

机构地区：[1]华北电力大学电子与通信工程系,河北保定071003 [2]国网信息通信有限公司,北京100053

出　　处：《郑州大学学报（理学版）》2012年第2期43-47,共5页Journal of Zhengzhou University:Natural Science Edition

摘　　要：针对相似话题难以区分的问题,提出了基于层叠模型的话题检测方法.该方法以Single-Pass聚类策略为基础,将新闻实体信息运用到话题检测中,改进时间相似度和地点相似度的计算方法,在底层利用文本内容相似度完成话题检测的任务,在高层结合时间相似度和地点相似度完成话题检测的任务.实验结果表明,该方法的性能优于传统的文本相似度算法.A novel approach was proposed based on the cascade model for the topic detection to effective- ly distinguish similar topics. Based on Single-Pass clustering algorithm, the entity information was used in topic detection, two similarity methods were proposed, including time-based similarity and location-based similarity. Moreover, at the bottom level of the system, content-based similarity was used to realize topic detection task. At the top level of the system, three similarity results were effectively combined to realize topic detection task. Experimental results showed that the performance of this novel algorithm was superi- or to the traditional text similarity algorithm.

关 键 词：话题检测 相似话题 向量空间模型 层叠模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]