Multiclass Sentiment Analysis of Political Viewpoints Using XGBoost and BERT

A recent paper published on arXiv (2608.11049) investigates multiclass sentiment analysis concerning political opinions expressed on social media. This study, categorized as a cross-type submission, tackles the difficulty of automatically distinguishing various sentiment classes related to political figures and issues. The researchers developed and assessed two machine-learning techniques utilizing XGBoost and BERT, training and evaluating them on a labeled dataset of political posts from social media using conventional classification metrics. The findings revealed that the XGBoost model attained an F1-score of 0.2835, while the BERT model's results were not entirely detailed in the abstract. This research underscores the increasing significance of sentiment analysis in comprehending political discussions, capitalizing on the extensive political content produced on social media.