Transformer-Based Sentiment Analysis in Social Networks: A PRISMA-Based Systematic Literature Review
TL;DR
A PRISMA based framework is presented to examine recent research published during 2021 to 2026 on transformer based sentiment analysis, indicating a strong shift toward fine-tuned pre-trained transformer models, such as BERT, RoBERTa, and domain-specific variants, along with increasing interest in large language model-based sentiment analysis.
Abstract
Sentiment analysis plays a significant role in understanding public opinion from the huge amount of data generated on review websites, social media platforms, and online forums. Recent advances in Natural Language Processing (NLP), particularly transformer-based models, have significantly enhanced sentiment classification by capturing contextual and semantic relationships from the data. This study presents a PRISMA based framework to examine recent research published during 2021 to 2026 on transformer based sentiment analysis. The outcomes indicate a strong shift toward fine-tuned pre-trained transformer models, such as BERT, RoBERTa, and domain-specific variants, along with increasing interest in large language model-based sentiment analysis. This study highlights the contributions of recent research, classifies the transformer architectures, contrast, reveals several critical challenges and future research directions, existing research gaps, trends, explainable, and scalable transformer-based sentiment analysis systems. This study provide up-to-date reference for researchers and practitioners seeking to develop scalable, accurate and effective sentiment analysis systems.