Sentiment and Concern Classification on Metaverse Governance Responses Using Naïve Bayes and Support Vector Machine (SVM)

The rapid advancement of immersive technologies such as the metaverse has introduced new opportunities and challenges for digital governance. Understanding public perception of these technologies is essential for designing governance systems that are transparent, inclusive, and responsive to citizens’ needs. This study analyses public sentiment and concerns regarding the use of metaverse technology in governance by applying two machine learning algorithms: Naïve Bayes and SVM. The dataset, consisting of open-ended survey responses from participants in The Gambia, was pre-processed through tokenization, stopword removal, and TF-IDF vectorization before model implementation. The results indicate that both algorithms can classify sentiment into positive, neutral, and negative categories; however, SVM consistently outperforms Naïve Bayes across all evaluation metrics. The SVM model achieved an accuracy of 88.6 percent and an F1-score of 0.873, demonstrating superior capability in recognizing contextual and semantic nuances within short text responses. In contrast, Naïve Bayes tended to overclassify responses as neutral, reflecting its limitation in capturing word dependencies. These findings confirm that SVM is better suited for sentiment analysis involving complex linguistic expressions and context-dependent opinions. The study contributes to the growing body of research on artificial intelligence in public policy by demonstrating how machine learning can provide deeper insights into citizen perspectives on emerging digital technologies. Such analytical approaches can assist policymakers in identifying public expectations, addressing concerns, and fostering trust in metaverse-based governance systems.