基于组合神经网络的舆情短文本表示模型降维方法研究

Research on Dimension Reduction Method of Public Opinion Short Message Text Representation Model Based on Combinatorial Neural Network

随着网络舆情对整个社会的舆论影响力越来越大,为了有效的防控恶性事件,实时监控网络舆情也越来越尤为重要.本文针对舆情短文本信息的特点,提出基于组合神经网络自动聚类方法来构建舆情短文本的表示模型,通过特征词语义相似度构建词簇.实验结果表明:在海量文本分析和舆情发现的过程中,该方法可以有效提高舆情短文本构建表示模型的精度,提升舆情发现效率.

With the rapid development of the Internet and the popularization of social media,the influence of cyberspace public opinion on the whole society grows tremendously.In order to effectively prevent and control the malignant events,real-time monitoring of cyberspace public opinion is becoming more and more important.Based on the characteristics of public opinion short message texts,this paper proposes an automatic clustering method based on the combinatorial neural network to construct the expression model of public opinion short message texts by constructing word clusters according to the semantic similarity of feature words.The testing results show that in the process of massive text analysis and public opinion discovery,by reducing the dimension of the text representation model,this method can effectively improve the accuracy of public opinion short text construction model and greatly increase the efficiency of public opinion discovery while guaranteeing real-time public opinion acquisition.