作　　者： ; ; ; ;

机构地区： 西安交通大学电子与信息工程学院

出　　处： 《西安交通大学学报》 2012年第10期24-28,共5页

摘　　要： 针对基于关键词元的话题内事件检测算法运行效率不高、不适合进行大规模文本话题检测的问题,提出了一种高效的关键词元聚类算法.该算法在进行词元簇选择时,为簇间相似度分配权值,并借鉴正态分布函数评估词元簇的个数,提高词元簇的选择精度,从而减少所需的词元聚类次数.实验结果表明,将改进的方法应用到舆情监控的话题检测中,能在不影响检测精度的前提下有效地提高算法的运行效率. An improved term-committee-based event identification algorithm is presented to meet the requirements of efficiency and accuracy in public opinion monitor system,where the original event identification algorithm can not be applied due to its lower efficiency.While the similarity between the clusters is calculated,the weight is taken into consideration simultaneously.Referencing the examples from normal curve,an evaluation algorithm is proposed to help choosing cluster with a proper term number,thus the improved algorithm only needs clustering once.The experiments indicate the operating efficiency for the required accuracy.

关 键 词： 话题检测 关键词元 舆情监控

领　　域： [自动化与计算机技术] [自动化与计算机技术]