Six-Year Multilingual Translation Dataset Released with Fully Automated WMT Pipeline

A team open-sourced a six-year, high-quality multilingual dataset and models, and won best systems on some language pairs in the WMT Creole shared task using a fully automated pipeline, while noting mismatched orthography in archival data as a key remaining challenge.

2026-11-06 ~ 2026-11-06 · 3 related posts

• Orthography mismatches remain the hard part of mining archival data for low-resource MT — anas_ant · 2026-11-06
• 6-Year-Open-Source Multilingual Dataset Ships, Author Says Only Today's Models Make It Feasible — anas_ant · 2026-11-06
• Zero-human pipeline produces best system in WMT Creole MT shared task — anas_ant · 2026-11-06