Han-Wen Nienhuys <[email protected]> writes:

> You don't need to change the website for this. Google Chrome already
> offers to translate pages that use languages outside of your regular
> set. Firefox does the same, here on Fedora 44 even with on-device
> translation.
>
> I assume users who can't read English would use automated translation
> for other parts of the web as well.
>
> We could consider dropping translations entirely, since we don't have
> the capacity to keep them up to date. They also complicate the build
> process by quite a bit.

We have several translations that are well-maintained and good.  The
problem I see with dropping good translations is that there is no
indications an AI would do better, and it would lead to a total waste of
global resources if everyone generated their own AI translation
on-the-fly.

When we find that AI produces reasonable translations, there may be a
point in attempting to do a whole-sale replacement of outdated
translations.

For example, I think that the German translation would be a reasonable
target because it is for one thing severely outdated, for another there
are a lot of German speakers on the lists who might be coaxed into
proofreading parts (possibly after stuff has already been checked in)
even if they are not up to participating in a manual translation.

There are other languages with bigger question marks, like Czech where
translations are only partial and outdated and it isn't clear there are
active users in contact with the LilyPond lists.

Then there are languages like Chinese which more likely than not have
significant user numbers, but they aren't in contact with the lists.

Where we are not in the situation to have reasonable amounts of contact
to people who could actually check translations, I don't think we have a
lot of options.

But if we want to venture into AI-based translations, obviously there is
no point in starting with those languages where the translations are in
a good state or where we don't have enough contacts to get a good
verification.

-- 
David Kastrup


Reply via email to