Mining Method And System Research Of Sentiment Orientation For Network User

With the rapid development of mobile Internet and the continuous innovation of information technology and the popularization of smart devices,the Internet has covered a large area of people's learning,working,life,traveling,entertainment and other fields.People share their opinions on different platforms,express their attitudes towards hot events,comment on the quality of products,and express their feelings and emotions through online texts.Exploring the potential emotional tendency of users in the text has great commercial and social significance for product research,market analysis,and early warning of network public opinion.The methods of text sentiment analysis can be divided into two categories: emotional dictionary-based methods and machine learning-based methods.The analysis method based on sentiment dictionary relies on the sentiment dictionary,and the quality of the sentiment dictionary directly affects the classification effect of the model.Although this method is simple,it requires a huge manpower to maintain the dictionary,and the model is not portable.The traditional machine learning method requires the participation of domain experts when extracting text features.The text feature modeling in different fields is different,so the labor cost and portability issues are not properly solved.Aiming at the shortcomings of dictionary-based and traditional machine learning methods in text emotion mining,this paper combines convolutional neural network(CNN)and attention mechanism to design a model for network user sentiment classification(MWABCNN).Based on the theoretical model research,the network sentiment analysis prototype system is designed.The specific work of this paper is as follows:Firstly,based on the study of the word vector structure of text sentiment analysis,the existing word vector training model is compared and experimental research,and the training method of word vector based on word2 vec and GloVe model is designed.The pre-training experimental study of word vectors was carried out using Wikipedia and Weibo corpus.The results show that the pre-trained word vector performance is better than the randomly initialized word vector.Secondly,a new neural network model(MWABCNN)that combines multi-channel convolutional neural networks and attention mechanisms is designed.Compared with Native Bayes,support vector machine and recurrent neural networks classification algorithm,The results show the performance advantages of the MWABCNN model in network text sentiment analysis.Thirdly,the parameter optimization of the MWABCNN model is studied.According to the static and non-static of the input word vector,the size of the convolution kernel,the dropout probability,the size of the training corpus,etc.,multiple sets of different controlled trials were to set up to verify the impact of these parameters on the accuracy of the model classification and to tune the parameters.Fourthly,based on the theoretical model of text sentiment analysis,the prototype system of network text sentiment analysis based on deep learning(NUSAS)is designed and implemented.