TranSenA: A Transformer-based Framework for Sentiment Analysis of Restaurant Reviews

Sentiment analysis (SA) of restaurant reviews is pivotal in understanding customer experiences and preferences, allowing businesses to enhance service quality and customer satisfaction. This study presents an automatic framework that leverages transformer-based techniques for SA of restaurant reviews in Bengali. To underscore the importance of this research, a new Bengali Restaurant Review Dataset (BRRD) has been meticulously developed, featuring 11,004 labeled restaurant reviews categorized into positive, negative, and neutral. The SA tasks were performed using various ML, DL, and transformer-based models, including LR, RF, DT, SVM, AdaBoost, CNN, BiLSTM, CNN + BiLSTM, XLM-R, mBERT, and BanglaBERT. Experimental evaluations reveal that BanglaBERT achieved the highest weighted f<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf>-score of 0.90, outperforming the other models used in this study.