OCR+VLM pipeline mines grammar books to boost translation for 4,000+ low-resource languages
anas_ant · x · 2026-11-06
A new paper from Chutong Meng and Barron Associates substantially improves translation for extremely low-resource languages by extracting parallel data from linguistic documentation materials. Key points: thousands of languages have grammar books and archival data versus only a few hundred with web presence; an OCR+VLM pipeline mines these materials, and even a few hundred/thousand sentence pairs prove highly useful. On Qwen, 512 mined examples lifted average chrF++ from 9.36 to 30.76 for languages without NLLB support (e.g., ddo 0.17→24.24, mnc 3.92→27.65), and from 20.76 to 29.66 with NLLB support. The upshot: under-resourced languages no longer need to rely solely on the Bible for training and evaluation.
More from Research
- Kording Lab amplifies thread: objective functions as concise descriptions of what neural nets learn — KordingLab · 2026-11-06
- LLM peer review is here to stay, researchers say — but 'excuses to reject' era must end — examachine · 2026-11-06
- UChicago names 31 researchers to 2026 Rising Stars in Data Science, with focus on world models — cindy_x_wu · 2026-11-06
- Surge researcher praises Tinker: parallel RL experiments without GPU management — echen · 2026-11-06
- Nemotron 3.5 Super + Surge RL matches GPT-5.6 Terra across 12 benchmarks — echen · 2026-11-06
- BMVA symposium on computer vision interpretability opens CFP, London, Jan 2027 — HildeKuehne · 2026-11-06