A Novel Transformer Based Deep Learning Approach of Sentiment Analysis for Movie Reviews

Language and organ signs allow humans to convey their feelings. When a large number of people express their emotions about a particular topic, it indicates a point of col-lective assessment, in short people's sentiment. Due to the vast production, sharing, and transition of data, opinions, and product reviews online and in media, sentiment analysis is essential. It evaluates the information and categorizes it as positive or negative. Numerous machine learning and deep learning algorithms can be deployed to interpret the sentiment behind any opinion or text. The sequential processing inherent in sequence models often leads to prolonged execution times. However, the Transformer model parallelized processing reduces calculation time. To that aim, our research suggests a comprehensive deep learning approach that mitigates the drawbacks associated with sequential models while combining the favorable aspects of both Transformer and sequence models. This architecture uses the IMDB dataset of 50K movie reviews to analyze the reviewers' expressions and the proposed model has been experimented with three advanced transformer-based deep-learning approaches such as BERT, RoBERTa, and DistilBERT. Of these, the Robustly optimized BERT approach (RoBERTa) fares better than the other two, obtaining an accuracy of 95.02 %, which is higher than 94.12% of BERT and 92.82% of DistilBERT approaches as well as the state-of-the-art models. This study highlights the importance of sentiment analysis as a method for understanding the emotions and attitudes conveyed in movie reviews. The high score also forecasts the efficacy of a film by taking into account the mean emotion of all the critiques.