Research And Application Of Emotion Classification Of Micro-blog On Deep Learning

With the rapid development of the Internet,people are active in many different social platforms and leave a lot of useful information,especially sina weibo platform.More and more people tend to write down their thoughts and express their inner feelings on the platform.These text messages often contain the user's comment on an event,the merchant's promotion of a product,and so on.Therefore,the massive text data generated on the microblog platform has certain mining value.Traditional text sentiment analysis methods mainly include sentiment lexicograpty-based methods and emoticons based methods,but these methods need to spend a lot of time when extracting features from text data.Using machine learning and deep learning to deal with emotion classification tasks can overcome the shortcomings of the above traditional methods.This paper mainly integrates deep learning technology with attention mechanism model to improve and optimize the existing sentiment analysis model and method of micro-blog text.In this paper,the Bi-directional Long Short-Term Memory(BiLSTM)model is introduced to construct the BiLSTM-ATT model,and the affective information in the word vector is automatically learned.Softmax layer is used to classify the affective output of the model.On the basis of the research of BiLSTM and Attention Mechanism model,an improved word vector is proposed to combine the self-attention Mechanism(SA)model with BILSTM,which is abbreviated as WD-BiLSTM-SA model.Aiming at the problem that the influence of word frequency information on word vector information is ignored in the traditional word vector,this paper improves the word frequency information based on the Word2 vec model,and adds word frequency information on the basis of the original word vector,so as to strengthen the initial information of word vector.In addition,compared with the attentional mechanism,the self-attentional mechanism relies less on external information and is better at dealing with internal information relations.Experimental results show that the improved model achieves better classification performance compared with the traditional model.Based on the research of Convolutional Neural Network(CNN)and BiLSTM model,a model combining the Convolutional Neural Network and BiLSTM and introducing the attention mechanism is proposed,which is abbreviated as CNN-BiLSTM-ATT model.On CNN the convolution operation can effectively obtain the advantage of the text in different information,and BiLSTM can effectively forecast the meaning of the text statements,so the combination,introducing attention mechanism construction of CNN-BiLSTM-ATT model,technology in the use of weibo creeper crawled take weibo corpus set on contrast experiment,the experimental results show that compared with separate CNN and BiLSTM separate model,the optimized model has better emotion classification effect.In view of the fact that text sentiment classification is not widely used,this paper also designs and implements a text sentiment classification system based on Django framework,which can be applied to different sentiment classification scene tasks.The back end mainly uses the improved WD-BiLSTM-SA model to carry out sentiment classification of text,while the front end is mainly implemented by using HTML,CSS,JQuery and other technologies.