Understanding Online Public Sentiments: A Machine Learning-Based Analysis of English and Chinese Twitter Discourse during the 2019 Chinese National Day

As the Internet gradually penetrates people's daily lives, individual citizens are empowered to demonstrate and exchange opinions and sentiments at any time anywhere. Online communities are increasingly participating in the agenda-setting of public affairs and official policies. However, how to depict online public opinion and to what degree does it influence the real world are still unclear. This study addresses the above problems by analyzing Twitter discourse during the 2019 Chinese National Day with a machine learning-based approach. Over 300,000 English and Chinese tweets were collected between Sept 30 and Oct 3, and a hybrid method of support vector machine (SVM) and dictionary was applied to evaluate the sentiments of the collected tweets. This method avoids complex structures while yielding an average accuracy of over 96% in most classifiers used in the study. The results indicate alignment between the time of National Day celebration activities and the peak of sentiments revealed in both English and Chinese tweets, although the sentiments of the two languages tend to be in opposite directions. The sentiments of tweets also diverge from nation to nation, but are generally consistent with the country's official relations with China. The linguistic features of the tweets suggest different concerns for Twitter users who have different sentiments towards China. Future studies should prolong the collecting period and refine the algorithms used. Specific attention can also be paid to important countries pairs like China and the United States.