基于OTSCM模型的主题情感在线追踪

发布时间：2018-06-17 19:55

本文选题：OLDA模型 + 主题情感　；参考：《现代情报》2017年12期

【摘要】：网络舆论主题情感在线分析对舆情研判与管理起着十分重要的作用,当前的主题情感模型存在着主题与情感建模关系不紧密,情感挖掘偏斜等问题,容易造成舆情误判。文本在OLDA(On-Line Latent Dirichlet Allocation,OLDA)模型的基础上引入情感参数,并提出情感遗传思想,建立基于情感遗传的在线主题情感混合模型OTSCM(On-Line Topic and Sentiment Combining Model)。该模型把t-1时间片内的主题情感分布作为t时间片内主题情感分布的先验,通过构造主题情感演化矩阵,生成t时间片内文档—主题、主题—特征词以及主题—情感词3个分布,最后使用交叉熵方法计算t时间片内主题分布与t-1之前主题分布的相似度,得出t时间片内主题情感演化结果。本文在5个数据集上对OTSCM进行了验证,并与其它流行算法进行了对比,实验表明,文本方法在主题情感在线识别方面达到了良好的效果。
[Abstract]:Online analysis of online public opinion theme emotion plays a very important role in the research and management of public opinion. The current thematic emotional model has some problems such as the relationship between theme and emotional modeling is not close and emotional mining is skewed, which can easily lead to misjudgment of public opinion. Based on the OLDA-On-Line Latent Dirichlet allocation Model (OLDAA), this paper introduces affective parameters and puts forward the idea of affective heredity, and establishes an online thematic affective hybrid model based on affective heredity, OTSCM On-Line On-Line topic and observation Model (OTSCM). In this model, the theme emotion distribution in t-1 time slice is regarded as a priori of the theme emotion distribution in t time slice. By constructing the theme emotion evolution matrix, the document and theme in t time slice are generated. Finally, the cross-entropy method is used to calculate the similarity between the theme distribution in t time slice and the theme distribution before t-1, and the result of theme emotion evolution in t time slice is obtained. In this paper, OTSCM is validated on five datasets, and compared with other popular algorithms. The experiments show that the text method achieves a good effect in the online recognition of subject emotion.
【作者单位】：蚌埠医学院卫生管理系;中国科学技术大学综合国力信息监测中心;
【基金】：国家自然科学基金项目“面向路径隐私保护的移动群智感知数据收集研究”(项目编号:61672038) 安徽省高校人文科学重点项目“基于对比挖掘的医疗卫生网络舆情的发现、跟踪及倾向性分析”(项目编号:sk2015A405) 安徽省高校自然科学重点项目“基于统计模型检验和风险分析的婴儿培养箱安全性评价关键技术研究”(项目编号:KJ2017A223)
【分类号】：TP391.1

【相似文献】