Transformative Insights of Sentiment Analysis with Explainable Transformer Models

Sentiment analysis is an important component of natural language processing, giving significant insights into the attitudes, emotions, and opinions of individuals or groups, and influencing business strategies. Despite major advancements, the need for highly accurate, robust sentiment analysis models persists. This emphasizes the necessity for continuously exploring innovative approaches that can enhance the interpretability and efficiency of sentiment analysis. This study evaluates and compares the effectiveness of leading transformer models, including DistilBERT, ELECTRA, T5, ERNIE, and GPT-2, in accurately capturing and analyzing sentiments within the Amazon product reviews dataset. The results are benchmarked against current methodologies, emphasizing the Lexicon-Enhanced BERT model, acclaimed for its robust sentiment classification performance. We conducted a comparative analysis of performance metrics including accuracy, precision, recall, and the F1-score. GPT-2 resulted in an outstanding performance, achieving an accuracy rate of 94.5%. By enhancing model performance and interpretability. Our research represents a notable advancement in sentiment analysis.