Emotional Tendency Analysis Of Deep Text Based On SVM

With the continuous development of the Internet,people can express their opinions on various social platforms,and network texts are also increasing,such as users' comments on certain products and opinions on hot news,etc.These comments contain rich subjective feelings of users,and emotional analysis of these network texts has gained more and more attention and practical value in the field of natural language processing.One of the most important tasks of emotion analysis is emotion classification,which aims to divide the opinion text into different emotional polarities,such as positive and negative.In the face of a large number of unstructured texts,it takes a lot of resources to analyze them by manual annotation,while some traditional classical emotional analysis methods are not comprehensive enough to extract emotional information in the process of text analysis,and the classification effect is not ideal when facing short texts with sparse features and huge data.Therefore,this thesis proposes an emotional analysis network model based on support vector machines(SVM),which is composed of convolutional neural network(CNN)and bidirectional gated recursive unit(Bi GRU).The main research contents of this thesis are:(1)This thesis proposes an emotion analysis model combining multi-channel CNN and Bi GRU.This model can fully extract the feature information of different positions of sentences through three CNN channels.At the same time,combined with the advantage that Bi GRU model can link the text context information,it can enhance the model's ability to extract text emotion feature information and improve the model's performance of text emotion classification.(2)In order to improve the generalization ability and robustness of the model,this thesis proposes an emotion analysis model which combines neural network with SVM.In this model,multi-channel CNN and Bi GRU models are used as tools to extract SVM input vectors,and the output vectors of multi-channel CNN and Bi GRU models are used as inputs to SVM models,and the emotional polarity of texts is classified by SVM classifier.This model and other classic models are trained on Chinese and English data sets respectively.By comparing the classification accuracy of each model,it is verified that the improved model proposed in this thesis has higher accuracy and better classification performance for text emotional polarity.