Analyzing Customer Sentiments: A Comparative Evaluation of Large Language Models for Enhanced Business Intelligence

This paper presents a comparative evaluation of Large Language Models (LLMs) against established sentiment analysis tools, focusing on their efficiency in parsing complex customer reviews for sentiment and emotion. Our research reveals that LLMs are capable of detecting a wide range of emotions, offering granular insights that transcend traditional analysis capabilities. However, the study also highlights a notable discrepancy in reproducibility and reliability among smaller LLMs (SLMs) when compared to the VADER model, a stalwart in sentiment analysis. These findings elucidate the potential and limitations of employing SLMs for business intelligence purposes, particularly in automating sentiment analysis of customer feedback. We discuss the implications of these outcomes for businesses looking to leverage AI for customer insights, emphasizing the importance of model selection based on the specific requirements of reproducibility and depth of analysis.