Requirements validation is an important activity of requirements engineering. Large language models (LLMs) have been proposed to improve the efficiency of requirements validation activity. However, the LLMs validate generally without considering the stakeholders’ characteristics of the software development projects that will use the requirements. In this paper, we propose PReVAL, a novel persona-driven LLM-based validation approach that utilizes persona to reflect the characteristics of stakeholders in the prompting of LLMs. A comparative experiment demonstrates the benefits of PReVAL in terms of consistency in validating requirements based on the persona. However, introducing a persona also increases unnecessary aspects of the validation result.
Copyrights © 2025