[L10n] Integrate translation platform

[Korb]

1. Weblate (used by F-Droid too, open-source, self-hostable, our recommendation),
2. Crowdin or
3. Transifex (both proprietary but free for FLOSS projects).

https://f-droid.org/2024/08/15/twif.html#community-news

and

4. Lokalise (“We offer Lokalise for free to open source projects”)

Don't forget about metadata files and Instructions for translators (an example of a good instruction).

[Cwpute]

Yes ! i'd be very glad to help transjate the game into french if a crowdsourcing tool is eventually chosen :)
My first choice would go to Weblate.

[Sesu8642]

I created #114 for adding support for multiple languages. It should be implemented first.

Since context needs to be provided anyway, I wonder if AI tools can provide similar quality translations instead or at least provide a base to be improved upon. The risk of AI outputting some copyrighted text should be minimal in this use case.

[Cwpute]

The text of Feudal is pretty simple and wouldn't be a problem for an LLM to translate no.
For better results, i suggest @Korb and i provide translations for the languages we wish to translate Feudal in, populate the Glossary, which will certainly improve further translations in other languages.

How does that sound ?

[Korb]

I wonder if AI tools can provide similar quality translations

Here I explained why AI translation of apps is a bad idea. It will significantly complicate the process of getting the translation up to par.

@Cwpute, the use of the glossary is non-negotiable. It's mandatory. Initially, work on the terminology dictionary can be made publicly available, and editing access rights can be reduced only after long-term translators with a clear vision of the big picture have been identified for each language and have proven themselves worthy of promotion to proofreader.

[Cwpute]

Oh i don't think we need too much restrictions or hierarchy on who should be able to edit the glossary or translations, or even if we need one individual responsible for each translation branch ! For a bigger project with lots of dialog or text i would definitely agree, but here it would be way overkill.

I understand and share your concerns regarding LLM-translated apps and my preferences always goes towards fully human-translated work. Although i can certainly understand how in some situations a badly translated app is better than an untranslated one that someone, as a foreign language speaker, wouldn't understand at all.

Anyway, whatever is decided upon, i'll happily provide a french translation for it !

[Sesu8642]

I understand @Korb's concerns about LLM translations. Some of the problems mentioned might have been caused by poor execution. I guess I will do my own experiments. Of course, I won't turn down a volunteers translation in favor of the inferior LLM generated one.
I need to evaluate how difficult it is to integrate the crowd sourced translation services and if it's worth it. Contributions to the translations could also be submitted as pull requests or by sending the property files. This does limit the number of people who are willing to contribute, of course. But first let's focus on #114 first.