Text Sentiment Intensity Prediction Method Based On Valence-Arousal Space

Text sentiment intensity prediction refers to the use of computational linguistics to analyze,process and induce text affective information and deduce sentiment intensity.Compared to the categorical sentiment representation,continuous dimension sentiment representation,such as valence-arousal space,can quantify sentiment intensity by mapping text sentiment information to the coordinates of the low dimension space to provide more fine-grained sentiment information.The thesis focus on the text sentiment intensity prediction based on valence-arousal space.The text sentiment intensity prediction methods are mainly on the basis of sentiment lexicon,machine learning model and deep learning model.Due to the computation simplicity and the lack of word order information,the former two methods can not fully reflect the complex sentiment changes in the text.Although the method based on deep learning model can solve the problem to a certain extent,the existing methods can not make full use of the text sentiment information,which leads to the relatively large prediction error and low precision.In order to solve the above problems,this thesis proposes a text sentiment intensity prediction model based on LSTM?CNN(Long Short-Term Memory?Convolution Neural Network).Firstly,to extract the text context information better,the model uses LSTM to retain the important information and forget the non important information in the text.Secondly,CNN is used to extract local sentiment features from the output of LSTM to ensure the comprehensiveness of feature extraction.The experimental results show that the LSTM?CNN model has higher prediction precision than existing baseline models.However,the LSTM?CNN model can be improved in the context information extraction,word attention allocation and model generalization.Therefore,the three problems are optimized by introducing Bi LSTM(Bidirectional Long Short-Term Memory),attention mechanism and batch normalization,and a Bi LSTM?CNN model based on attention mechanism is proposed.The experimental results show that compared with the LSTM?CNN model,this optimization model has a certain improvement in the prediction precision of text sentiment intensity.In addition,the prediction performance of the model is significantly better than the existing baseline models.