A global twitter sentiment analysis model for COVID-vaccination

The world faced a major health crisis during the COVID-19 pandemic for nearly two and a half years following its emergence in late 2019. Receiving prompt and accurate news played a crucial role in making various decisions to control the pandemic situation. In recent years, social media has demonstrated the effectiveness of disseminating information very rapidly. A spectrum of research is ongoing in the field of Twitter sentiment analysis. However, most of these studies consider only the local viewpoint to understand public sentiment. In this work, a fine-tuned transformer-based model is used to determine the sentiments of tweets locally. Additionally, various real data sources are augmented to enhance the understanding of sentiment polarity. Moreover, it is challenging to obtain a labeled dataset for analyzing sentiment from Twitter based on both global and local viewpoints. Hence, we propose three methods to achieve precise sentiment analysis by considering both global and local perspectives. The classification results obtained from these methods are used to train various traditional classification models. These results can serve as ground truth for further analysis. Our extensive analysis shows that more precise sentiments, based on the global perspective, can be extracted and are useful for training datasets in different machine learning techniques.