Multi-language cards with bitext

I noticed that this discussion was not a feedback post, and therefore didn’t have the ability to “vote” on it. Here I’d like to propose an implementation.

In short, the problem is to provide knowledge in multiple languages. For example:

• In Quebec, Canada, legislation enforces availability of documents in French. This is a problem that we must solve.
• Our team may work with documents in their native language, but as we onboard more people who may not speak that language, we need a way to provide answers that will also be maintained, and invalidated.
• Updates to one language must invalidate other languages until someone revised and translated them.
• Translation requires bitext (seeing both texts aligned, side by side) otherwise it’s very, very difficult to do without using external tooling.
• Missing translations shouldn’t mark the card as not verified, but out of date or missing translations must be unverified.

I wrote a proposition that, I think, would complement this request very well, but I’ll try and avoid doing both and only focus on the language part here: 

1. Define what languages you want to support. I would suggest to do this per collection.
   1. I think there should be a primary language, and a list of secondary languages per collection too.
2. In cards, select a language that the SME (or the verification team) will consider the primary language. This card will behave like cards to right now, i.e. they don’t “care” about other languages.
3. Allow showing other languages
   1. Either switch to another language, or;
   2. Show both at the same time
4. Each language can have one of three statuses
   1. Not translated
   2. Translated (translated after the main article was verified)
   3. Translation outdated (a modification to the primary language was made, so the translation is out of date)
5. Edit mode would also allowing showing the primary AND one of the secondary languages

There’s room for improvement but hopefully this can trigger some discussions internally on how companies outside of the US could benefit from Guru without the headaches of localization.

As a bonus, it would be very nice to see an integration with DeepL for any non-translated card (while clearly being identified as “machine translation”), so we can just look at automatic translations, make a few adjustments and verify them.