ReXCL Tool Automates Requirements Engineering with Adaptive Fine-Tuning

Researchers from the arXiv community have introduced ReXCL, a novel tool that automates two critical phases of requirements engineering: extraction and classification. The Extraction module uses a combination of heuristics and predictive modeling to convert semi-structured raw requirement documents into a clean, predefined schema. The Classification module then employs adaptive fine-tuning of encoder-based models (like BERT variants) to assign appropriate class labels to each requirement. This streamlines the often manual and error-prone process of structuring requirements, which is a bottleneck in software development.

Performance evaluations presented at the Requirements Engineering 2025 conference demonstrate that ReXCL significantly boosts both efficiency and accuracy compared to traditional manual methods. The tool can export its output directly into external requirements engineering tools, making it practical for real-world use. By automating the schematization of requirements, ReXCL reduces human effort and minimizes inconsistencies, potentially accelerating the entire software development lifecycle from specifications to implementation.