Research On Emotion Classification Of Network Short Texts Based On Deep Neural Network

With the advent of the information age,the development of Internet technology is becoming more and more mature,all kinds of network media emerge at the historic moment,from the beginning of tencent QQ chat,to the later douban film,sina weibo.People can express their ideas and opinions anytime and anywhere,and this convenient and rapid way of exchanging information brings about an increasing amount of data.These data contain people's views and opinions on things or events.The use of natural language processing technology to analyze these data and find the emotional tendency contained in them has an important impact on our practical applications such as public opinion monitoring,commodity marketing and financial analysis.Text emotion classification,also known as text orientation analysis,is one of the research hotspots in the field of natural language processing in recent years,attracting the attention of many researchers,and many practical and effective classification algorithms have been proposed.Based on the deep neural network model,the emotion classification algorithm has gradually become one of the mainstream methods to solve the text emotion classification problem due to its excellent feature extraction ability and the advantages of hardware computational support.Based on the deep neural network,this paper conducts text emotion classification research on network short texts.Firstly,it obtains the film review of douban movie with the help of network crawler,and then preprocesses and emotion labeling the crawled data according to the corresponding criteria and standards,so as to obtain the text emotion classification data set.In order to make full use of the text emotional information,word2 vec tool is used to vectomize the part of speech features and lexical features respectively,and then the vector splicing is carried out and used as the input of the convolutional neural network.On this basis,considering that the traditional maximum pooling method is likely to lose feature information,a KMCNN model is proposed to replace the maximum pooling with k-max pooling.Experiments show that KMCNN based text emotion classification has better classification performance than other models.Considering that the learning of text sequence features needs to be combined with context,a kmcnn-gru model is proposed on the basis of KMCNN,and the locally important features extracted by KMCNN and the sequence features extracted by GRU are fused in the feature extraction layer to enhance the ability of the model to capture text emotions in the way of feature fusion.The experimental results show that the text emotion classification model based on KMCNN-GRU can learn more semantic features and has better classification accuracy than other classification models.