Analyzing the Interplay Between Social Media Sentiment and Traditional Public Opinion in Politics

In the present day, due to the growing number of Web 2.0 tools, users are producing vast quantities of data in a massive and constantly changing manner. Opinion mining or Sentiment analysis (SA) is an informative technique to help obtain useful information from users’ data automatically. Over the years, a number of SA challenges have been solved using deep learning techniques which has led to the achievement of state-of-art performances. Thus, it is necessary to solicit help to help the researchers to learn the present progress and outstanding issues to be solved quickly. This paper discusses the multidimensional nature of public opinion in South Korea through analyzing the entire Korean tweets from January 1, 2023 to December 31, 2023. Using Twitter (X) API, we collected over 5 million tweets with emphasis on phrases related to South Korean president and significant events in the country. Through the application of set filters, the authors were able to arrive at a dataset of approximately 4 million tweets. These developed tweets were then taken through rigorous preprocessing in order to make them ready for SA. In the current work, the BERT model was chosen to act as the major focus of the study. This model was particularly trained on a given tagged dataset of Korean text for the purpose of classifying the text’s sentiment. The findings of the present study reveal complex patterns of the processes that define the nature of online sentiment, thus revealing that several forces work in concert to form public opinion. Despite certain concordances between the results of the online SA and the offline public opinion polls on presidential job performance approval, the relationship is rather weak.