UMT-NMT v1.0: Proof-of-Concept Study on Normative Documents in Metrology (EN→UK)

Unified Metrology-TRIZ Neural Machine Translation (UMT-NMT v1.0) is a domain-specific EN→UK MT pipeline for metrology with built-in structural transparency.It integrates placeholder/constant logic, constrained decoding, a deterministic post-editor, and a TRIZ-aware reranker that enforce structural integrity by design.On Corpus A (128 pairs; seed=2025), the system achieves PPR=1.0 and FTA=1.0.Lexical quality at this scale is moderate (BLEU≈30.09; chrF≈49.15).The bundle includes anonymized data, inference code, and structural/semantic graphs.Strict evaluation on an 8-sentence quantum metrology slice confirms preservation of invariants under conservative decoding.