Hi,
I think translation is fine, if you use sentence-by-sentence
translation. The problem with AI chatbots is that they reformat the
text, add slop to it and so on. Yo may use AI to translate, but they
make clear in your prompting that you want a trnalation that does not
change sentence structure, adding filler text, reformatting with emojis
and bullet points,...
In general if you don't understand the text after translation as a
contributor, because you don't know the language, then you have to add
your original text in your own language as a reference. If you
understand the text and verifid that it is not longer than your original
written text, feel free to post it - if you can identify yourself with
it. I sometimes also let AI cleanup and rewrite English texts written by
me, but I make sure, that the typical "UweSays" style stays alive!!!
Otherwise I can't publish it.... Maybe Mark Miller produces an UweSays
AI that can post texts in my style!
Uwe
Am 30.08.2026 um 13:55 schrieb Michael Sokolov:
I read the linked policy from Astral and liked it too. My only concern
was the very strong admonition about using LLMs to generate text made
me think it was not OK to use it for translation, but then at then
end, they say that is OK. I might just wordsmith it a little to work
in the exception sooner, but in the end the policy seems to get it
right - LLM as a tool in service of humans, communication remains
human-to-human (possibly mediated with translation tools)
On Sat, Aug 29, 2026 at 8:34 AM Robert Muir <[email protected]> wrote:
On Sat, Aug 29, 2026 at 4:48 AM Dawid Weiss <[email protected]> wrote:
This looks great, I think we should do a similar thing. Also, we should add
some global instructions for LLMs
to guide people's efforts and ensure they run all the checks prior to
contributing anything (some of this
is under help/ but a top-level AGENTS.md would help here).
Agreed. I also think maybe we should re-evaluate the help/ files
becoming markdown format? Main reason would be to take advantage of
markdown tooling. Our markdown linter will find broken links, the
formatter will keep them tidy, its easier for developers to navigate
between the files, etc. Would make the maintenance of these files
easier.
If nobody is opposed to the idea, maybe we can open a PR with 3 or
lines of text similar to one of those examples? Maybe we could improve
issue/PR templates to link to it. It might improve the signal-to-noise
ratio.
+1.
PR: https://github.com/apache/lucene/pull/16583
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
--
Uwe Schindler
Achterdiek 19, D-28357 Bremen
https://www.thetaphi.de
eMail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
