Emotional Analysis And Research Of Microblog Text Based On Deep Learning

With the rapid development of the Web2.0 era,a large number of data have emerged on the Internet and many social platforms.It left a lot of people's information,including the emotions,thoughts,ideas and opinions of everyone.These data exist in an unstructured form.If they can be malicious,they extract emotional data,it will provide great value for the supervision of social public opinion,the development of merchandise sales,after-sales information feedback,information warning and so on.At present,Microblog is a very popular social platform.It is in the form of short text or multimedia communication.It has a great reference value for the study of social information.Firstly,this paper gives a brief overview of the basic theory of emotional analysis,and introduces the existing technology of micro-blog text emotional analysis.The emotional lexicon based micro-blog text affective analysis technology and the machine learning based micro-blog text affective analysis technology are two major categories.At present,the two technologies have developed relatively mature,and deep learning is a new field developed on the basis of machine learning.It has a remarkable achievement in processing image and audio recognition,and it also has outstanding performance in text emotional analysis.Secondly,after a preliminary understanding of deep learning.The basic idea of the convolution neural network is further introduced.Then the local connection and weight sharing features of the convolution neural network and the subsampling operation are described.In the end,a CNN-SVM microblog emotional analysis model is presented And relationship between the accuracy of the emotional analysis of the model and the number of iterations of convolution operation of the convolution neural network is studied by experiments.Finally,the comparison of the two experimental results is proved that the CNN-SVM analysis model presented Two groups of experiments were carried out by using the NLPIR public corpus and the actual data crawled by " microblog" as experimental data.Each group of experiments used three models,the microblog sentiment analysis model based on sentiment dictionary,the microblog text emotion analysis model based on convolutional neural network and the CNN-SVM emotion analysis models in this paper,to verify the experiment.Then select the evaluation indicators for evaluation.