Jonas Hahnfeld via Discussions on LilyPond development
<[email protected]> writes:

> On Tue, 2026-05-12 at 21:17 +0800, Kunpeng He via Discussions on
> LilyPond development wrote:

>> - As an amateur music lover, I cannot guarantee the overall quality
>> of the translation. Are there any quality standards we should meet
>> before merging? Or would a complete version be acceptable, with
>> polishing to follow later?
>
> Can you please clarify to what extent you used the AI to translate the
> documentation? Using LLMs to translate (parts of) phrases that were
> later checked by a human is fine, I guess. Such machine translators
> exist since a long time (Google Translate etc.) and in the end could
> be considered "better dictionaries".

Machine translations actually have become amazingly better in the last
few years.

> On the other hand, just running an LLM on the entire documentation
> without careful checking is different in my opinion: What do we gain
> by integrating such translation? Users could already translate the
> English documentation on-the-fly and an "official" translation would
> look like it was vetted by a human.

The environment gains by not everybody running their own translators.

> You are writing "few days", which is very little for careful checks.
> This leads me to believe that it's the latter, and in that case I
> personally would be against accepting it.

I do expect that good LLMs would easily do 98% of the job these days.
The results still need to get vetted by a human and compared with the
actual behavior of LilyPond.  That's the threshold for entering actual
commits.

-- 
David Kastrup


Reply via email to