Is there a way to provide additional context to help translation

[jtoppine]

Is there a way to provide additional context to assist the AI in choosing the right meanings of words where by themselves the words can be ambiquous?

Say I have Finnish document I want to translate to english. It has a title, and various content fields. The finnish title is Lähde, which depending on context may translate to wildly different meanings source, spring (of water), or go away. At the moment, by itself argos translates the title lähde to source, although it the correct meaning would be go away.

Content field might contain more text that opens up the context to what we are talking about. Let's say
lähde pois mun luota (go away from me). Indeed argos can translate this correctly, and infer the meaning of lähde in this sentence to be go away, and not source.

Wondering if it would be possible to translate only the title string lähde but give the content string lähde pois mun luota as a hint of the context.

I could just concatenate the strings to translate everything in one go so that argos is aware of the whole context, but I would not be able to read back just the title field that I'm actually interested in. Furthermore, not sure but it appears as if argos splits input sentences and translates them separately, as ambiquous word meanings in translations appear to be more correctly guessed if I remove all punctuation and translate the whole word soup in one go.

In an ideal world, imagining an API like this for example:

argostranslate.translate.translate(text, from_code, to_code, hints)

Where hints content would be text somehow considered to infer context ("are we talking about relationships, geology, or programming languages"), but not included in actual translation result.

But I would also be interested in ugly hacks or workarounds that might result in more accurate contextual translations.