Supplementary Material: On Converting Natural Language Requirements into Semi-Formal Templates Using LLMs

Paper: On Converting Natural Language Requirements into Semi-Formal Templates Using LLMsConference: 2026 34th IEEE International Requirements Engineering Conference (RE) This supplementary material contains the data used in the evaluation of our LLM-based approach for converting natural language requirements into the MASTeR semi-formal template syntax. If you use this artifact in your research, please cite: @inproceedings{rosskothen_2026_convertReqsToMaster, title = {On Converting Natural Language Requirements into Semi-Formal Templates Using LLMs}, author = {Roßkothen, Julian and Fuchß, Dominik and Erdösi, Florian and Floruß, Maria and Keim, Jan and Hey, Tobias}, booktitle = {2026 34th IEEE International Requirements Engineering Conference (RE)}, year = {2026}, organization = {IEEE} } Contents File Description rephrased_reqs_data.csv LLM-rephrased requirements with all computed metric scores and survey ratings for each configuration survey_solution_data.csv Source requirements with human-rephrased ground truth (MASTeR) and survey ratings limesurvey_survey_export.lss LimeSurvey export file containing the complete survey structure and questions metrics.py Implementation of the MASTeR Template Metric and Embedding Mover's Distance metric used for evaluation prompts/ Jinja2 prompt templates and reference implementation of the reflective formalization loop Dataset Origin The source requirements and human-rephrased ground truth originate from Großer et al. [1]. References [1] Großer, K., et al. "A Comparative Study of Requirements Rephrasing into Semi-formal Syntax Templates." 2023. Dataset available at: https://doi.org/10.5281/zenodo.13343495