A Review on Sentiment Analysis Using Transformers and Ensemble methods
Y. Rama Devi, Aasish Bharthepudi, Adarsh Govindarajula · 2025
Sentiment Analysis is the field of study that examines people’s opinions, sentiments, valuations, appraisals, and attitudes and emotions from written language. It is one of the most active research areas in natural language processing and is also widely studied in data mining, web mining, and text mining. Sentiment analysis has become almost an essential task for companies to keep track of their brand or product success on the social media market. Sentiment analysis can help one gain insights about how people feel about things, i.e., their opinions. This paper presents a comprehensive review of sentiment analysis using transformer architectures and Large Language Models (LLMs), and their ensembles with traditional machine learning models for various sentiment analysis tasks. The performances of sentiment analysis tasks can be significantly improved by transformers and LLM-based approaches. The models that are covered in paper, including SAE, LSTM, BERT, RoBERTa, MPNet, GRU, XLNet and others are trained on different publicly available benchmark datasets used are IMDB Reviews and Twitter Airline. Reviews and so on to measure the performance of surveyed models, the metrics primarily considered are accuracy and $\mathbf{F 1}$-score. The main objective of this survey is to provide an overall understanding on transformer-based approaches for sentiment analysis, including advantages and Problems. Future scopes are depicted at the end. Index Terms—Sentiment Analysis, Transformers, SAE, LSTM, BERT, RoBERTa, MPNet, GRU, XLNet.