Reliable Requirement Engineering Using LLM for Tag Governance

This artifact accompanies our CASCON 2025 paper on Reliable Requirements Engineering. It provides the full implementation of our High-Level JSON (HLJ)–centric tag governance pipeline, including scripts, datasets, and evaluation harnesses. The artifact enables users to reproduce reported results—such as Precision, Recall, and F1-score across multiple models (GPT-4.1, Opus4, Meta-70B)—and to inspect per-tag evaluations, audit trails, and reasoning logs. By combining SBERT-based semantic similarity, hybrid clustering, and canonical tag mapping, the pipeline ensures results are not only reproducible but also fully auditable. Config-driven pipelines (v0, v1, v2) make experiments easy to replicate or extend. We expect this resource to support both researchers extending our methods and practitioners validating LLM outputs in requirements workflows.