Research On Chinese Text Sentiment Classification Based On Deep Learning And Its Application In Public Opinion Analysis

With the explosive development of China's Internet,the huge online social groups produced a massive network public opinion,its sentiments are extremely easy to be spread and infected.And how to capture the emotional trend of the people through the complex public opinion information,to guide the positive public opinion spread,to ensure social stability and harmony,is a very important research topic.The Chinese text sentiment classification is one of the core part of public opinion analysis,which is paid close attention and studied by scholars.It is difficult to obtain a higher Chinese text sentiment classification accuracy rate,not only because Chinese text data has many characteristics such as semantic pluralism,grammatical particularity,implicit expression and so on,but because most of the current Chinese text sentiment classification methods belong to the shallow learning method and there are some shortcomings such as limited textual representation ability and reliance on manual extraction of sample features.Therefore,how to adapt to the characteristics of Chinese text and further enhance the performance of Chinese text sentiment classification is an urgent need to study in the field of online public opinion analysis.This paper has carried on the research of Chinese language sentiment classification based on deep learning and applied the research to the analysis of network public opinion.The related research contents are as follows:1.Proposed deep learning sentiment classification model based on merge theme features.Based on merge theme features,new deep learning sentiment classification model-TB_LSTM s and TCNN is proposed.these new model can obtain high quality text features.The experimental results show that the highest accuracy rate of the two models can reach 91.1% and 91.9% in the binary sentiment classification,which is about 2% higher than that LSTM,CNN,RAE.2.Deep learning sentiment classification based on enhanced feature extraction.In order to enhance the feature extract ability of the deep learning model,this paper megre high-level text feature from the two kinds of deep learning sentiment classification model(TB_LSTM and TCNN),So build the TB_LSTM + TCNN sentiment classification model.The experimental results show that the accuracy of binary sentiment classification of TB_LSTM + TCNN is higher than that of TB_LSTM sentiment classification model and TCNN sentiment classification model by 0.8%-1.6% under the same conditions.3.Intelligent modeling of Multi-dimensional public opinion analysis based on deep learning.Aiming at the comprehensive demand and accuracy requirement of Chinese network public opinion analysis,this paper constructs multi-dimensional public opinion intelligence analysis model based on deep learning.The model can use the deep learning sentiment classification model based on enhanced feature extraction to realize accurate Chinese text sentiment classification,And to achieve multi-dimensional emotional classification,multi-dimensional emotional trend analysis and prediction combined with the theme model and time series model.4.Empirical study of multi-dimensional public opinion analysis based on deep learning.In order to validate the validity of the multi-dimensional public opinion analysis model based on deep learning,this paper takes the "Wei Zexi event" as an empirical case,based on the multi-dimensional public opinion intelligence analysis model proposed in this paper,realizes the deep analysis of the "Wei Zexi event",realizes the analysis of the multi-dimensional opinion sentiment trend,the hot spot tracking and the emotional evolution,and proves the validity and practicability of the model by comparing with the expert 's conclusion..