Emotion-Based Chinese News Classification And Recommendation Research

In recent years,the rapid development of the Internet era has made personalized recommendation technology widely used.At the same time,the advancement of artificial intelligence technology has led to the innovation of robot technology.The new generation of robots has more humanized emotional computing capabilities.Starting from the idea of serving the empty nesters,a news recommendation system project that integrates user emotions is proposed.This paper mainly studies two sub-tasks in the project,emotion-based news classification task and user-sense-based news recommendation model construction task,which can be used as a technical foundation for the whole project.The main work of this paper is summarized as follows:(1)Firstly,the research background and significance of this dissertation are introduced,and the research status of text sentiment analysis and personalized news recommendation are summarized.Then from the perspective of text sentiment classification,the related techniques used in the machine learning-based classification method are introduced,including data preprocessing,text representation,text feature extraction,and common machine learning methods.Two basic artificial neural network models are also introduced.Then introduced two more mature recommendation algorithms,and briefly introduced the recommended methods,evaluation indicators and experimental methods commonly used in the recommendation system.(2)Secondly,for the emotion-based news classification problem,according to the characteristics of news and the comparison of common text sentiment classification methods,the multi-input channel convolutional neural network model is proposed based on convolutional neural network to solve emotional classification of news headlines problem.The input layer of the MIC-CNN model has three input channels,and the three channels are weighted and combined in the convolution pooling layer,and the emotion classification is finally completed by the softmax function.The comparison experiment shows that the method is feasible and the classification effect is improved compared with other methods.(3)Finally,for the news recommendation problem based on user emotions,mainly analyzes how to obtain news data with emotional information,how to obtain emotional information of users and how to integrate emotions to make interest recommendation three difficult points.After research,this dissertation uses the previous news sentiment classification method to obtain news with emotional information,uses existing techniques such as speech emotion recognition,facial expression detection and othermethods can obtain emotions to a certain extent;By analyzing the characteristics of people's emotions,five kinds of emotional recommendation strategies are given to make interest recommendation.Then the framework of the news recommendation system that integrates the user's emotions is given.Finally,the feasibility and effectiveness of the emotion-based news recommendation system scheme are verified by the experimental method of user study.