Terminology-aware machine translation
TaMTAS translates life-science texts with their terminology intact, then adapts them for the people who need to understand them.
The challenge
The project develops a terminology-aware, document-level translation and augmentation system for life sciences. It combines Large Reasoning Models, quality estimation, automatic post-editing and audience adaptation across English, Spanish, Catalan, Estonian and Irish.
How it works
The technical workflow follows the interdependence between work packages described in the project proposal.
Corpus & terminology
Multilingual corpora and enriched terminology databases establish the terms the system must preserve.
WP2 · TBXTools · parallel and comparable corpora
Reason & verify
An LRM translates at document level. Quality estimation and automatic post-editing identify terminology errors and feed corrections back.
WP3 + WP4 · LRM · QE · APE · DPO
Adapt
The translated text can be simplified and supported with summaries, glossaries and explanations for different audiences.
WP5 · text augmentation · audience adaptation
Ambiguity, made visible
This conceptual demonstration shows why scientific translation needs document context and controlled terminology.
Real-world validation
Stakeholders contribute authentic documents and expert or public feedback to evaluate accuracy, clarity, fluency and usefulness.
Genomic research, scientific outreach and the enrichment of high-level scientific terminology in Catalan.
Medical research and guideline development, where consistent terminology is essential to prevent misunderstanding.
Accessible bilingual scientific and healthcare communication for patients, clinicians, researchers and the public.
