Research On Text Emotional Classification Based On Deep Learning

With the rapid development of the Internet and the popularity of smart phones,the network has become a platform for the participation of all people.Nowadays,a large amount of text data has been generated on the network.How to analyze and utilize these data is of great significance to help people make decisions in real life.As the basic task of natural language processing,text emotional classification has always been a research hotspot.Traditional text categorization methods mainly include dictionary-based method and machine learning method.Since the emergence of deep learning algorithm,the accuracy of text categorization has been greatly improved.At present,the common task of text emotional classification is emotional polarity analysis,that is,to judge the positive or negative emotions of the characters contained in a sentence.However,as the subjectivity and objectivity of sentences are not distinguished,it is likely to interfere with the classification results.For example,in film reviews,the sentence "Jack tries to maintain his good reputation" objectively describing the plot of the film,which contains positive adjectives such as "good",may mislead the classifier to classify it as positive emotion.But in fact,we can not get any positive or negative emotional information from it,and even mislead our judgment.If these objective statements can be filtered out in advance when classifying emotional polarity,the effect of emotional polarity classification will be improved.Therefore,how to classify the subjectivity and objectivity of the text and extract the subjectivity information from the text has become the focus of this paper.Based on the analysis of the task of text emotional polarity classification,the application of deep learning model in text emotional classification is deeply studied.The main research work of this paper is as follows:(1)It provides a new way to improve the task of text emotional polarity classification,filtering out the objective information before classifying.Three methods of text subjective information extraction based on deep learning model are proposed.(2)An improved subjective and objective text classification model is proposed.By analyzing and comparing the application principle of activation function in convolutional neural network,a subjective and objective text classification model based on convolutional neural network without activation function and long-term and shortterm memory network is designed.(3)The long-term and short-term memory networks and their variants are studied,and their application principles are put forward,which provide a basis for researchers to choose long-term and short-term memory networks in practice.(4)A subjective and objective text categorization model with attention mechanism is designed.In order to verify the validity of the model,this paper conducts a series of comparative experiments on the subjective and objective emotional data set of English movies.The experimental results are in line with expectations,and the results are analyzed and summarized.