Exploring Transformer-Based Model in Sentiment Analysis of Movie Review

The rapid development of technology has led to a significant increase in internet usage, with approximately 5.44 billion users worldwide. Now internet use has become an important part of everyday life. The internet is not only used to search for information but also as a platform for individuals to express their sentiments on various topics. Sentiment analysis is determining whether online comments are positive or negative. The movie industry in particular, benefits from sentiment analysis as audience reviews have a significant impact on viewership and financial success. Traditional models such as SVM, Naïve Bayes, and VADER have been widely used for sentiment analysis, but fail to handle complex language in movie reviews, which often leads to misclassification. However, transformer-based models have demonstrated superior capabilities in understanding context and word relationships, making them more effective for sentiment analysis. This research uses several of the most popular transformer-based models such as BERT, XLNet, and DistilBERT and compares them with traditional models such as SVM, Naïve Bayes, and VADER to identify the model that provides the best balance of accuracy, efficiency, and adaptability for sentiment analysis. The research results show that the XLNet outperforms other models with the highest scores, namely accuracy of 93.1%, recall of 91.0%, precision of 95.0%, and F1-score of 93.0%.