Dataset: Automatic Generation of Explainability Requirements and Software Explanations From User Reviews

We conducted two studies in which requirements engineers generated explainability requirements and explanations based on user reviews. These artifacts were created manually by the engineers in a workshop setting and then validated through a structured evaluation process. The evaluation involved two sets of participants, who rated the quality of the generated requirements and explanations. Participants selected their preferences without knowing whether the artifact was generated by ChatGPT or by the engineers, allowing for unbiased feedback. The dataset includes these generated requirements and explanations, participant ratings, and their reasoning, categorized into six key factors: Clarity, Style, Tone, Correctness, Level of Detail, and Relevance.Additionally, we have included the code for a tool that can automatically generate summaries, requirements, and explanations based on a list of reviews with explanation needs. This tool utilizes an API call to ChatGPT to create the texts, and it allows for manual adjustments to the generated content for further refinement.