Sentiment Analysis of Movie Reviews with Transformer Models: A Deep Learning Approach to Text Classification

Sentiment analysis (SA) is a method for making inferences from textual data. These inferences are used in many fields, such a business, social media monitoring, and healthcare. This study presents a comparative analysis of different Transformer models and other hybrid configurations to classify sentiment in a movie review dataset. The models compared are BERT, DistilBERT, RoBERTa, and hybrid architectures such as BERT-BiLSTM-XGBoost, BERT-DistilBERT-TF-IDF ensembled with XGBoost, DistilBERT-XGBoost, and BERT with XGBoost. Among these, the RoBERTa model outperformed the rest of them, attaining an accuracy of 96. These result show how well transformer models and hybrid models perform.