Abstract: Highlights•As far as we know, this survey is the first work that attempts to present a comprehensive review of prompt-based sentiment analysis.•We detailly discuss the performance of LLMs (Large-scale language Models) for sentiment analysis, especially under few-shot conditions, further we also explore why LLMs are more challenging.•We summarize the latest prompt-based methods in various sentiment analysis tasks, and the influence of PMs’ biases, and analyze the potential gap in prompt engineering between sentiment analysis and other domains.•We discuss the future research directions to improve the performance of prompt learning methods in sentiment analysis.
Loading