Welsh accents

I saw a related thread here: Spanish accents.

Which with Spanish near voice collection it would be best to define a list of accents.

Welsh is already in voice collection but it doesn’t have any accents defined except a generic “UK” accent which isn’t a thing.

Is this something we need to rectify?

1 Like

Do you have a list of “official” accents for Welsh?

@josh_meyer correct me if I’m wrong but I assume an accent for Deep Speech is when the pronunciation is different, not when the pace and tone are different, right?

@nukeador – Welsh definitely needs better accents tags.

Someone from the community will have to confirm this (@dewi.jones @ynyr_llew), but my impression is that people agree on at least three accents: Northern Welsh, Southern Welsh, and Patagonian Welsh.

It’s hard to really define what an “accent” is, but pace / tone / pronunciation all distinguish accents. Pace and tone are actually very important for speech technologies.

1 Like

Asking because then we’ll need a very large list of accents for some languages. I did a first approach for Spanish here, but definitely we can go way more detailed.

Maybe defining accents should be a requirement before a language starts collecting data? Could we add this to Pontoon?

I don’t think pontoon is set up for that, since this don’t involve having a translation but a new set of lines with accents. But we should probably have a place where once a language has 5000 sentences, community can ask for activation providing with the accents list?

Josh is correct that Welsh has three major accents - north, south and Patagonian. Northern Welsh has an additional sound ‘u’ (sounded the same as ‘i’ in southern Welsh) and vowel length is another distinguishing factor. Patagonian Welsh shows a small Spanish influence in pronunciation. These aren’t showstoppers, though in our previous Paldaruo corpus we were able to ask more detailed questions about accents. I think if we could ask for activation with 5000 sentences we would go for an accents list for Welsh.