Project Summary

Our project aims to develop a sophisticated tool that leverages prompt engineering techniques to effectively elicit requirement revisions from ChatGPT, taking into account the capabilities of large language models (LLMs). The primary focus is on exploring the extent to which LLMs can refine, review, and improve requirements in terms of ambiguity, consistency, completeness, and verifiability. To achieve this, we will be formulating specific prompts that address the aforementioned questions. By interacting with ChatGPT, we will assess its ability to provide valuable insights and suggestions for refining the requirement set. The challenge lies in finding the most suitable approach to pose these questions to the model, ensuring that we obtain meaningful and actionable responses.

An important aspect of this tool’s development involves establishing a systematic way to parse and interpret the model’s answers. We aim to structure the output in a canonical format that is both readable and understandable by machines. By achieving this, we can transform the generated text into a structured representation that facilitates further analysis and processing.

Overall, our tool will facilitate the extraction of requirements from natural language text using ChatGPT. We recognize that large language models possess the potential to contribute significantly to the requirement engineering process. By utilizing prompt engineering techniques, we can guide and train ChatGPT to provide valuable input on requirement refinement.

The tool’s workflow will involve an iterative process of interacting with ChatGPT, refining prompts, and evaluating the model’s responses. This collaborative approach will allow us to progressively enhance the quality and accuracy of the requirement revisions. The successful development of this tool holds promising implications for the field of requirement engineering. By integrating LLMs into the process, we can tap into their vast language understanding capabilities to enhance requirement refinement. Furthermore, the tool will aid in bridging the gap between human language and machine-readable formats, enabling seamless interpretation and utilization of the revised requirements.

In conclusion, our project focuses on building a robust tool that harnesses prompt engineering to effectively engage ChatGPT in revising requirements. By extracting insights from natural language text and leveraging the capabilities of LLMs, we aim to optimize the requirement engineering process and improve the quality and accuracy of refined requirements.

Project Team

• Maged Elaasar, Ph.D. (PI, Modelware)
• Abdelwahab Hamou-Lhadj, Ph.D. (Co-I, University of Concordia)
• Mohammad Hamdaqa, Ph.D. (Co-I, Polytechnique Montreal)
• Kareem Elaasar (Research Intern, CalPoly Pomona)
• Brandon Hsu (Research Intern, UCLA)
• Seif Ashraf (Research Intern, AUC)

Sponsors