It Takes Two: A Dual Stage Approach for Terminology-Aware Translation

---

Setup

pip install -r requirements.txt
pip install -e .

Set up API keys:

cp .env.example .env
# Add your API keys to .env

Quick Start

Generate synthetic training data:

python scripts/generate_synthetic_data.py \
  --input data/ende_dev.jsonl \
  --output data/synthetic/synthetic_ende.jsonl \
  --target-lang de --mode multi

Filter and train:

python scripts/filter_synthetic_data.py --input data/synthetic/synthetic_ende.jsonl --threshold 0.86
python trainer.py --data-files data/processed/filtered_*.jsonl

Results

Lang	Mode	BLEU	chrF2++	Proper SR	Random SR
DE	proper	48.06	70.74	0.98	0.73
ES	proper	58.51	76.08	0.99	0.78
RU	proper	35.80	63.57	0.98	0.72

Full results in paper.

Citation

@inproceedings{jaswal-2025-takes,
    title = "It Takes Two: A Dual Stage Approach for Terminology-Aware Translation",
    author = "Jaswal, Akshat",
    editor = "Haddow, Barry and Kocmi, Tom and Koehn, Philipp and Monz, Christof",
    booktitle = "Proceedings of the Tenth Conference on Machine Translation",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.wmt-1.112/",
    doi = "10.18653/v1/2025.wmt-1.112",
    pages = "1344--1350"
}