基于主题提取的海量微博情感分析

发布时间：2018-02-24 07:00

本文关键词： 海量微博聚类主题提取情感分类　出处：《南京大学学报(自然科学)》2017年03期 　论文类型：期刊论文

【摘要】：从海量微博数据中分析公众对某一社会事件的情感倾向具有重要研究意义,而海量微博文本稀疏规模庞大,导致传统方法处理这一任务时面临诸多挑战.提出一种基于主题聚类的海量微博情感分析方法.首先基于高质量微博数据挖掘频繁项集,设定语义相关阈值,筛选重要频繁项集进行谱聚类,得到主题关键词.基于主题关键词对海量微博数据依据语义相关度归类,最后结合情感词典对每类中的微博检索主题关键词前后修饰距离内情感词及否定词,结合表情符号计算微博情感值.在百万规模中文微博上进行实验,证明该方法能准确按主题归类且能有效在该主题上进行情感分类.
[Abstract]:It is of great significance to analyze the emotional tendency of the public on a certain social event from the massive Weibo data. As a result, traditional methods face many challenges when dealing with this task. This paper proposes a massive Weibo affective analysis method based on topic clustering. Firstly, based on the frequent itemset mining of high-quality Weibo data, the semantic correlation threshold is set. The important frequent itemsets are selected for spectral clustering, and the topic keywords are obtained. Based on the topic keywords, the massive Weibo data are classified according to the semantic relevance. Finally, the emotion words and negative words in the distance before and after the subject keywords are retrieved by Weibo in each category are combined with the emotion dictionary, and then the emotional value of Weibo is calculated by using emoji. The experiment is carried out on the million-scale Chinese Weibo. It is proved that this method can accurately classify the subject and can effectively classify the emotion on the topic.
【作者单位】：山东管理学院信息工程学院;南京大学计算机科学与技术系;
【基金】：国家自然科学基金青年项目(71301086) 山东省电子政务项目(2150511) 山东省科技厅星火计划(2013XH17003) 教育厅科技计划(J14LN62)
【分类号】：TP391.1

【相似文献】