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

Dataset: Automatic Generation of Explainability Requirements and Software Explanations From User Reviews Authors:Martin Obaidi, Jannik Fischbach, Jakob Droste, Hannah Deters, Marc Herrmann, Jil Klünder, Steffen Krätzig, Hugo Villamizar, and Kurt Schneider This replication package accompanies the study on Automatic Generation of Explainability Requirements and Software Explanations from User Reviews. It includes all datasets and scripts necessary to reproduce the evaluations and analyses presented in the paper. The dataset consists of 58 user reviews with explainability needs, for which explainability requirements and explanations were manually and automatically generated. The package is divided into three key components: Evaluation of Explainability Requirements and Explanations Data from two studies where participants evaluated manually and ChatGPT-generated explainability requirements and explanations. Preference votes and reasoning categories (Clarity, Style, Tone, Correctness, Level of Detail, Relevance) assigned by study participants. Formulation of Explainability Requirements and Explanations Dataset from a workshop with four requirements engineers, who manually formulated explainability requirements and explanations based on user reviews. Tool for Automated Explainability Generation Source code of a UI-based tool that automatically generates summaries, explainability requirements, and explanations from user reviews via ChatGPT API calls. The tool allows manual refinement of generated outputs and supports requirements engineers in analyzing user feedback. A README file is provided, detailing the folder structure, study methodology, and instructions for reproducing the results. This package ensures transparency and enables further research on LLM-supported explainability requirement engineering. This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Grant No.: 470146331, project softXplain (2022–2025). Citation If you use this dataset, please cite the following publication: Obaidi, M., Fischbach, J., Droste, J., Deters, H., Herrmann, M., Klünder, J., Krätzig, S., Villamizar, H., Schneider, K.: Automatic Generation of Explainability Requirements and Software Explanations From User Reviews. 2025 IEEE 33rd International Requirements Engineering Conference Workshops (REW). License This dataset is provided under the Creative Commons Attribution 4.0 International License (CC BY 4.0). Contact For questions regarding the dataset, please contact the corresponding author as listed in the publication.数据集：从用户评论自动生成可解释性需求（Explainability Requirements）与软件解释文本（Software Explanations） 作者：Martin Obaidi、Jannik Fischbach、Jakob Droste、Hannah Deters、Marc Herrmann、Jil Klünder、Steffen Krätzig、Hugo Villamizar及Kurt Schneider 本复现包配套于《从用户评论自动生成可解释性需求与软件解释文本》的研究工作，内含复现论文中评估与分析所需的全部数据集与脚本。本数据集包含58条带有可解释性需求的用户评论，针对这些评论，我们已通过人工与自动方式生成了对应的可解释性需求与解释文本。 本复现包分为三大核心模块： 可解释性需求与解释文本的评估 包含两项研究的相关数据：研究参与者对人工生成与ChatGPT生成的可解释性需求及解释文本开展评估，并给出偏好投票与推理类别标签，类别涵盖清晰性（Clarity）、风格（Style）、语气（Tone）、正确性（Correctness）、详细程度（Level of Detail）与相关性（Relevance）。 可解释性需求与解释文本的构建 包含一场研讨会的数据集：该研讨会邀请了四位需求工程师，他们基于用户评论人工构建了可解释性需求与解释文本。 自动可解释性生成工具 包含一款基于用户界面（UI）的工具的源代码：该工具通过调用ChatGPT API，可从用户评论中自动生成摘要、可解释性需求与解释文本。工具支持对生成结果进行人工微调，并可辅助需求工程师开展用户反馈分析工作。 本复现包附带README文件，详细说明了文件夹结构、研究方法以及复现实验结果的操作指南。本包旨在提升研究透明度，并为依托大语言模型（LLM）的可解释性需求工程领域的后续研究提供支撑。 本研究受到德国研究基金会（Deutsche Forschungsgemeinschaft, DFG）资助，项目编号为470146331，项目名称为softXplain（2022–2025）。 引用说明 若使用本数据集，请引用如下文献： Obaidi, M.、Fischbach, J.、Droste, J.、Deters, H.、Herrmann, M.、Klünder, J.、Krätzig, S.、Villamizar, H.、Schneider, K.：《从用户评论自动生成可解释性需求与软件解释文本》，发表于2025年第33届IEEE国际需求工程会议研讨会（REW）。 许可协议 本数据集采用知识共享署名4.0国际许可协议（Creative Commons Attribution 4.0 International License, CC BY 4.0）进行分发。 联系方式 若对本数据集有任何疑问，请联系论文中列明的通讯作者。