Using ChatGPT-4 to facilitate sentiment analysis for narrative data

Abstract Sentiment analysis is essential for interpreting textual data, often categorising sentiments as positive, negative, or neutral. While human analysis can achieve this, large datasets require automated tools for efficiency. However, many sentiment analysis programmes provide only summary statistics, limiting deeper insights. Generative AI , such as ChatGPT-4, addresses this limitation by offering more detailed sentiment and thematic insights. Unlike traditional tools, ChatGPT-4 allows for in-depth analysis, though relying on it alone may not be advisable. This paper explores integrating ChatGPT-4 into sentiment analysis workflows for narrative data, emphasising the importance of carefully crafted prompts and validating AI findings with manual checks to ensure reliability and accuracy.