A Comparative Analysis of Pretrained Models for Sentiment Analysis on Restaurant Customer Reviews (CAPM-SARCR)

Sentiment analysis plays a crucial role in understanding customer opinions and attitudes towards various products and services. In this study, proposed work is based on comparative analysis of five pretrained models, namely BERT Tokenizer, RoBERTa, MBERT, DeBERTa, and XLNet, for sentiment analysis on restaurant customer reviews. The dataset used for evaluation contains customer reviews labeled as positive or negative, indicating the sentiment associated with the restaurant experience. The main objective is to identify the most effective pretrained model for sentiment classification in the context of restaurant customer reviews. The models are evaluated based on testing accuracy and loss to determine their performance on the unseen test dataset. Among these models, RoBERTa and DeBERTa emerged as the most promising ones, achieving remarkable testing accuracies. These models demonstrated their exceptional capability to effectively capture and comprehend sentiment patterns in the feedback provided by restaurant customers. The outcomes of this study can provide useful insights into the use of pretrained models for sentiment analysis in the restaurant sector, as well as contribute to the advancement of natural language processing.