«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

26

期数:

2016年06期

页码:

25-30

栏目:

智能、算法、系统工程

出版日期:

2016-06-10

文章信息/Info

Title:

 Microblog Topic Detection Based on LDA Model and Multi-level Clustering

文章编号:

1673-629X（2016)06-0025-06

作者:

 刘红兵[1] ; 李文坤[2] ; 张仰森[2]

 1.太原科技大学 电子信息学院；2.北京信息科技大学 智能信息处理研究所

Author(s):

 LIU Hong-bing[1];  LI Wen-kun[2] ; ZHANG Yang-sen[2]

关键词:

 ; LDA模型; 话题检测; 改进的Single-Pass聚类; 层次聚类

Keywords:

 ; LDA model; topic detection; improved Single-Pass clustering; hierarchical clustering

分类号:

TP391

文献标志码:

A

摘要:

 随着微博这一新兴社交媒体的广泛应用，以微博为背景的相关研究不断涌现，其中基于微博的话题检测是当前研究的热点之一。结合微博文本的相关特点，文中提出了一种基于LDA模型和多层聚类的微博话题检测方法。首先，通过LDA模型对微博数据建模并提取特征；其次，利用改进的Single-Pass聚类和层次聚类对微博数据进行聚类，从而发现热点话题。通过在大规模微博数据上进行话题检测实验，通过LDA建模比通过TF-IDF进行特征选择和权重计算效果好；改进的Single-Pass聚类能够处理第一遍Single-Pass聚类未处理的微博，提高了初步聚类的精度，并且为下一步层次聚类减少了时间；多层聚类的聚类效果在准确率、召回率和F值三方面均比单一聚类算法的聚类效果好。显然，文中的话题检测方法是可行的，也是有效的。

Abstract:

 With the wide application of microblog,emerging social media,relevant research is being emerged on microblog. The topic de-tection based on microblog is one of the hotspots in current research. In combination with the relevant characteristics of microblog,a mi-croblog topic detection based on LDA model and hierarchical clustering is proposed. First,LDA model is applied for modeling and feature extraction to microblog data. Then,the improved Single-Pass clustering and hierarchical clustering is used on microblog data clustering and the hot topic is found. Experiment on large-scale corpus shows that it is more effective through the LDA model than by TF-IDF for feature selection and weight calculation;the improved Single-Pass clustering can deal with the untreated microblog by the first Single-Pass clustering,which can improve the accuracy of the initial clustering and reduce the time of hierarchical clustering;it is more effective through the hierarchical clustering than the single clustering in accuracy,recall and F -value. Clearly,it is feasible and effective by the LDA model and multi-level clustering to detect the microblog topic.

常用功能

更新日期/Last Update: 2016-09-19