Analysis Internet Public Opinion Opinion Based On The Evolution Model Of Topic Events

With the rapid development of Internet in China,network public opinion analysis becomes more and more important for maintaining social and economic development.Network public opinion analysis technology mainly includes three important parts: network data collection,data preprocessing and public opinion analysis.The topic event evolution model extracts the change of topic with time in the data set by analyzing the network public opinion data of time series.The rapid development of deep learning technology has brought new development to emotion analysis based on topic model.As a document topic generation model,topic model can complement the characteristics of deep learning sequence modeling.The rapid development of microblog,e-commerce and various forums has brought great convenience to the public to express their views.With the increase of user size and data volume over time,it brings new challenges to the analysis of network public opinion.Therefore,this paper takes the topic event evolution model as the research object,combined with the development of deep learning technology,carries out the research on topic evolution and emotion analysis of topic event evolution model.The experiment shows that deep learning technology can bring new promotion and development to the research of topic evolution and emotion analysis in improving the accuracy of data classification.The main contents of this paper are as follows:(1)This paper focuses on the acquisition of network data and the preprocessing technology in natural language processing,focusing on the existing crawler framework,web page data analysis technology and data duplication methods,word segmentation,deactivation and word coding methods,feature extraction methods,similarity measurement methods,etc.Through the research of the network data crawler technology to ensure the data source of public opinion analysis,the research of the preprocessing technology in natural language processing found the shortcomings of the traditional methods.(2)Based on the existing topic evolution algorithm,a topic evolution algorithm based on topic emotion polarity is established.By constructing the word embedding model of corpus,improving the classification method of topic emotion polarity,calculating topic emotion factors,and introducing topic emotion factors into the process of topic evolution.Experimental results show that the accuracy of the proposed method is higher than that of the traditional method,and the confusion of the topic evolution algorithm based on this method is better than that of the previous method.(3)This paper studies the emotion classification method which integrates LDA(latent Dirichlet allocation)and Self-Attention mechanism.By simplifying the topic event evolution model to LDA model,the emotion analysis is carried out.According to the whole document topic generation process of LDA model,each comment data and topic information are spliced and input into the word embedding model for training,so that the cosine similarity of the same topic information in the word vector space before the SelfAttention mechanism classification is smaller.Experiments on the published short text emotion classification dataset show that the emotion classification method proposed in this paper is superior to the current mainstream method based on the combination of LSTM(Long Short-Term Memory)and attention mechanism,and the model complexity is lower.