The MADLAD-400 open source translation model supports many languages. After enabling support for languages not supported by other services (T354666), this ticket proposes to enable support for languages which are only supported by one external service to reduce dependencies on them.
We want to enable the selected languages in the MinT test instance (not for Content/Section translation yet).
Based on the analysis in T343340, these are the languages selected:
Supported only by Google Translate
- Corsican (co)
- Divehi (dv)
- Western Frisian (fy)
- Hawaiian (haw)
Gan Chinese (gan)Not supported in MADLAD-400
Supported only by Yandex:
- Chuvash (cv)
- Eastern Mari (mhr) chm is used in MADLAD: "Unfortunately, we use the macro code chm for Meadow Mari (instead of the correct mhr), and mrj for Hill Mari"
- Western Mari (mrj)
- Yakut (sah)
- Udmurt (udm)