OpenAI: providing a second trusted translation to give more context

[TruffeCendree]

Describe the problem

The primary language of our app is French, but we have internally the skills for producing a quality translation in English.

For other langages, we intent to rely on OpenAI automation and fix issues with users feedbacks. Of course, some words or sentences may be ambiguous. To improve the quality of the translation, we would like to provide more context to the AI.

Describe the solution you would like

For our case, we were thinking about sending both the French and English translation to produce a third one. This may allow OpenAI to better understand the context when translating ambiguous words.

Describe alternatives you have considered

We will try to implement ourself a POC by updating openai.py.

Of course, a built-in implementation would benefit to more users.

Screenshots

No response

Additional context

No response

[nijel]

There are always more ways to provide a context to LLM. I'm curious if this significantly improves translation quality. I was also considering including some existing translations, but I couldn't see noticeable improvement while experimenting with this.

[github-actions]

This issue has been put aside. It is currently unclear if it will ever be implemented as it seems to cover too narrow of a use case or doesn't seem to fit into Weblate.

Please try to clarify the use case or consider proposing something more generic to make it useful to more users.

[TruffeCendree]

Thanks for considering this feature. We will probably experiment it and come back with some feedback.

It is a bit early, but I wish you a happy new year.