Irish dialects or general question about language variant support

[AlanGreene]

Hi folks, I had some questions about support for dialects and other language variants while reviewing recent PRs for Irish language support.

TL;DR

Is there some formal policy or approach regarding support for dialects or language variants in this repo? Also, any details on technical implementation or process relevant to identifying the dialect / variant used as input so the same can be used for output?

Brief context

Irish has 3 main regional dialects, plus the "standard" (An Caighdeán Oifigiúil) Irish which is taught in schools, used in official documents, etc. There is a large amount of overlap between these of course, but there are significant differences in some parts of the language. There is a single ISO language code for Irish (639-1: ga, 639-2: gle), with no identifiers for the dialects.

Question

I'm curious if there's an overall policy / strategy in the project around support for dialects or other language variants, or if it's more ad-hoc. For example, I see Brazilian Portuguese (pt-BR), and Swiss German (de-CH), but a single French.

Is there a recommended / accepted approach already in place for similar situations in other languages? Should we just stick with "An Caighdeán" (standard Irish), or would it be acceptable to include phrases for the other dialects too?

My current best guess is that we could include support for the dialects in the input phrases, but for output would have to stick to the standard form as there'd be no way to specify which dialect to use to match the speaker even if reasonable TTS models were available for the dialects.

Related discussions

I couldn't find much discussion of the technical aspects of dialect / variant support. The closest I found were:

• Sunset of Canadian French: Sunset the fr-CA language #2435
  • this was more about lack of maintenance
• Swiss German Rename de-CH to gsw? #2804