Investigating Large Language Models in Sentiment Analysis: Techniques, Trends, and Challenges

This paper provides an overview of a literature review for large language models (LLM) that provides a description of each step in the social sentiment analysis pipeline from data collection to actionable insights. This literature review also reveals the value of transformer-based models such as BERT and GPT to assist with the efficient processing of natural language data and performing more complex sentiment analysis types. Within the framework of this literature review, various components will also be discussed including the classification of sentiment, detection of emotion, and extraction of topics to assist in gaining more insight into how LLMs can assist with deeper understanding of user behaviour, emotion, and opinion. Within the literature review, there has also been an identification of trends and patterns that have emerged as a result of existing research that indicate public mood and perception across different domains. Major challenges such as ethical concerns, model biases or errors, and transparency and accountability are also presented in this literature review. Finally, conclusions and recommendations are provided for improving the performance of LLM sentiment analysis and reducing bias, as well providing future directions for further research regarding LLM-inspired sentiment analysis.