Generating Requirements Elicitation Interview Scripts with Large Language Models
Binnur Görer, Fatma Başak Aydemir
Abstract
Requirements elicitation interviews are the most popular requirements elicitation technique and an integral part of requirements engineering education. Good and bad interview scripts provide students with examples of applying the theory. Constructing an interview script requires technical knowledge, practical experience, and creativity. As a result, only a few educational interview scripts are available to the community. This paper explores automatically generating interview scripts with large language models through prompt engineering. Our contribution is two-fold: First, we present a graph representation of interactive interview scripts. Second, we apply prompt engineering techniques to generate business domain descriptions, linear scripts, and conversation pieces focused on certain types of mistakes. Our findings indicate that large language models face challenges in handling interview conversation graphs. However, we can enhance the quality of the generated interview scripts by decomposing the task into smaller components and refining the prompts to provide more precise instructions.