Sentiment Analysis and Contrastive Experiments of Long News Texts
Weinong Niu, Lin Yuanbo Wu · 2019
In view of the current patterns and characteristics of information dissemination, this paper makes a sentiment analysis of network news. It aims to effectively control the spread of negative and false information, maintain social stability, and provide support for relevant departments to master the right of disseminate discourse. The "Belt and Road" related corpus obtained through web crawler is used as the experimental corpus. We use the dictionary method, machine learning method and deep learning method to analyze the preprocessed corpus respectively. Among them, machine learning algorithms include Logistic Regression, Naive Bayes and Support Vector Machine(SVM). The method of deep learning is Long Short-term Memory(LSTM) network. The experimental results show that the accuracy of the dictionary-based method is lower than the other two methods. The accuracy of the machine learning-based method depends on the selection of the feature vector dimension. In contrast, deep learning-based method have higher accuracy.