基于深度学习和OCC情感规则的网络舆情情感识别研究  

Sentiment Analysis of Network Public Opinion Based on Deep Learning and OCC

作　　者：吴鹏[1,2,3] 刘恒旺 沈思 Wu Peng Liu Hengwang Shen Si(School of Economics and Management, Nanjing University of Science ＆ Technology, Nanjing 210094 Hubei Collaborative Innovation Center for Early Warning and Emergency Response Technology, Wuhan 430070 Jiangsu Collaborative Innovation Center of Social Safety Science and Technology, Nanjing 210094)

机构地区：[1]南京理工大学经济管理学院,南京210094 [2]安全预警与应急联动技术湖北省协同创新中心,武汉430070 [3]江苏省社会公共安全科技协同创新中心,南京210094

出　　处：《情报学报》2017年第9期972-980,共9页Journal of the China Society for Scientific and Technical Information

基　　金：国家自然科学基金"突发事件网民负面情感的模型检测研究"(71774084);"突发事件网络舆情演变过程中的人群仿真研究"(71273132);"基于时间感知模型的学术主题检索与演化挖掘研究"(71503124);国家社会科学基金"基于社会网络分析的网络舆情主题发现研究"(15BTQ063);安全预警与应急联动技术协同创新中心"面向突发事件网络舆情演变的群体行为建模与仿真研究"(JD20150401)

摘　　要：为解决网络舆情情感倾向性分析中语义理解不足和仅关注情感词典的现状,本文基于OCC模型认知情感角度建立情感规则,对网络舆情中突发事件的微博文本进行情感分类标注作为训练集,并对深度学习中卷积神经网络模型进行训练得到网络舆情情感识别模型。通过对比实验证明OCC情感规则标注使数据集情感分类更加精确,卷积神经网络的识别效果显著优于传统的机器学习方式(SVM),情感识别模型情感最高可达到90.98%的准确率。From the perspective of cognition, this study uses the OCC model to establish emotion rules to solve the problem of a lack of semantic understanding. This model is based solely on an emotional dictionary during the process of analyz- ing the affective tendency of Internet users. The affective tendencies of micro-blog texts on public opinion regarding an emergency are classified and labeled as the training set, and are used to train a neural model of deep learning to obtain the emotion recognition model of the public opinion on the network. The results indicate that the training set of emotion classifications labeled using the OCC model is more accurate, and that the distinguishing effect of the con- volutional neural network is significantly better than the traditional machine learning method （SVM）, achieving 90.98% accuracy.

关 键 词：网络舆情 OCC模型 深度学习 词向量 卷积神经网络 

分 类 号：G353.1[文化科学—情报学]