Research On Emotion Classification Of Public Opinion Based On Convolutional Neural Network And BiLSTM

With the rapid development of network information technology in recent years,the influence of Internet on people's life has penetrated into all aspects.People are usually used to express their views on the goods,things and people in the world through words,and the Internet is a value medium in this era.People communicate with each other through words on the Internet to convey their feelings.With the help of this kind of extensive communication method,efficient and convenient information and value transmission are realized.Therefore,mining the emotional relationship between characters and their transmission under the Internet environment is not only helpful to promote the research and development of NLP,but also has practical value in people's actual life.In short,the handling and analysis of online public opinion events by Chinese institutions,on the one hand,conveys the government's emotional tendency towards similar events,on the other hand,plays a good guiding role for the same type of events in the future.This paper mainly includes the following contents:(1)based on the consideration of how to improve the efficiency of text feature extraction,this paper extends and improves the traditional model.In this paper,considering that the external information of Chinese in reality has a certain impact on the emotion of the text,and it can improve the accuracy of the emotion classification of the text,so the theme of the text is extracted through related technologies,and the two emotion classification models based on the theme fusion are constructed.Then the model is proved to be effective by experiments,which shows the feasibility of feature fusion.(2)in view of the differences of the current Internet texts and the defects of some noise categories,based on the above idea that feature fusion can enhance the effect of emotion classification,this paper proposes a model integrating the traditional two deep learning models and a more robust text emotion classification,and then tries to integrate the attention mechanism based on this,so that the model proposed in this paper can be better extracted Output local information and context information of the text.Finally,several groups of comparative experiments are designed to prove that the classification performance of the model is improved,and the optimized model is effective.