Text Emotion Analysis Based On Neural Network

Emotion analysis is an important branch of text mining and natural language processing.With the development of the Internet,social networking and e-commerce,a great deal of text is produced every day.People often post their opinions on an incident on weibo or comment on products purchased on taobao.The emotional analysis of public opinions and comments with emotional tendency on weibo,taobao and other platforms can provide help for the decision-making of government departments and enterprises.Traditional emotion analysis models are built using machine learning methods.The traditional machine learning algorithm is unable to extract features,which leads to the low accuracy of emotion prediction.Aiming at the problem of low prediction accuracy of traditional machine learning algorithms,this paper uses neural network to build an emotion analysis model and combines the attention mechanism to improve the prediction accuracy of the model.The main research work and results are as follows:(1)Considering that convolutional neural network can effectively extract local features of text.The deep convolutional neural network model and piecewise convolutional neural network model were built respectively.Using different convolution structures to extract text structure features and word2 vec and attention mechanism are combined to improve the prediction ability of emotion analysis model.(2)Considering that the long-term and short-term memory network can effectively solve the problem of long-distance dependence of traditional circular neural network.Two-way long and short term memory network can effectively extract contextual information.To provide more emotional information for model prediction which this paper uses two-way short and long term memory network to build emotional analysis model and combines word2 vec and attention mechanism toimprove the prediction accuracy of the model.(3)Long and short-term memory networks can effectively extract contextual information,and convolutional neural networks can extract local text information.In this paper,combining the advantages of these two neural networks,an improved serial hybrid neural network model is proposed and a multi-level attention mechanism is merged.A longitudinal comparison experiment was conducted on the Chinese hotel reviews and English movie reviews data sets,and compared with some other existing models.The experimental results show that the improved model has strong learning ability,and can fully extract text features and predict Both the accuracy and the F value of the comprehensive performance index are better.