Chinese Text Sentiment Analysis Based On Converged Neural Networks

Sentiment Analysis allows computers to understand human language emotions correctly and effectively by establishing effective analytical methods.Because China's economy is developing rapidly,Chinese has become very popular.Chinese sentiment analysis has become a high-profile scientific issue.The paper uses converged neural network to analyze the sentiment analysis of Chinese comment texts.Firstly,it applied the fractal convolutional neural network model(Fractalnet)to Chinese sentiment analysis.It obtained a sentiment analysis model based on fractal convolutional neural network.The experiment adjusts the width of that model in Weibo comments and hotel reviews.The experiment result shows that when the column width is 3,the fractal convolutional neural network model has an accuracy and F1 value exceeds the model that the width is 2 or 4.Secondly,it merged the Fractalnet with the TextCnn model.It designed a Chinese sentiment analysis model based on Text_Fractalnet.The experiment compared Text_Fractalnet with TextCnn,Fractalnet and vDcnn model.The experiment result shows that the Text_Fractalnet model has an accuracy and F1 value exceeds the other models.Finally,it introduced a parameter in the Text_Fractalnet model fusion layer.This parameter is named the fusion rate.It adjusts the eigenvector weight of the TextCnn and Fractalent branches.The experiment compares the Text_Fractalnet model which introduces the fusion rate with the one which does not introduce the fusion rate.The experiment result shows that when the fusion rate is in multiple values,the former has the accuracy and F1 values exceed the latter.