Testing DeepL Translator

Hello everyone,

I’ve played a bit with DeepL Translator. You can find the translator available here freely: https://www.deepl.com/translator

You can also use my testing account (credentials in bidwarden) especially if you don’t want to be limited by batch of 5000 characters.

The quality of the translation is not bad at all. Main advantages over Google translate:

  1. It isn’t Google :muscle:
  2. You can upload entire documents (but only Microsoft licensed at the moment).
  3. They claim the documents and data are deleted when you are finished

However, this is not a tool we can let translate things on its own. It really needs a review. For example, in French sometimes the word “hub” was translated as “Moyeu” :eyes:. In both French and German, it didn’t know how to translate the pronoun “you”, alternating between “vous/Sie” the polite way and “Du/tu” the familiar way.
More generally it will also need reviews because we need it to fit the correct terms we have on the platform. For example, order cycles are translated “cycle de commande” in French but we are currently using “Cycle de vente” (Sales cycles).

If you pay a plan https://www.deepl.com/pro.html#team you can access the API. So I guess this is what you had in mind @MyriamBoure when you were thinking about this tool? To maybe connect it to Gitbook and automate translation?

With absolutely no reviews, this idea would be great in order to onboard more countries into using OFN. Today you need to speak English to understand what OFN does. With an automated tool, we could have our guides translated in languages we don’t have within the community yet. It wouldn’t be perfect but it would let the word out.

Today however, DeepL only translate to a limited set of languages yet:

  • English
  • French
  • German
  • Spanish
  • Portuguese
  • Italian
  • Dutch
  • Polish
  • Russian

To conclude, my personal opinion is that I don’t see us working on a automated tool for this now. However we can use this tool manually to be more quick into updating our guides in several languages.
The free plan can work for now. It would be fair to pay for it, but I think I would prefer seeing people updating guides be paid first.

But I would love to have other feedback on this :slight_smile:

@Anna especially if you already though of a Spanish version after the Catalan one.
@luisramos0 if you have some time to play with the Portuguese translator it could be interesting as well.
@Theodore for German and Dutch

Pinging also @louise (put I can’t find her) @lin_d_hop @Kirsten for English, because it goes also the other way around.

1 Like

Sounds like this will be a great tool though it will require development etc. In summary I agree with your conclusion @Rachel that it doesn’t make sense to automate now.

Personally for UK I would pay for a tool that provided an automatic translation for any user guide pages that France creates if it was saving us money. Looking at the payment plans they have:

  • Individual plans 5.99 euro for 5 translations /month or 19.99 euro for 20 translations

  • Team plans that reduce the price when you have 3 users or more.

The plans are thus very flexible and scale-able. The big cost is the initial integration such that a new page provided results in an auto-API call. Then we gain financially if the automation saves us enough time that it pays for itself.

From some back-of-envelope calcs I don’t really think that we are at that point yet.

The actual gains in productivity will come from having a similar structure of the user guides between countries. This will make it easier for a user guide maintainer to easily make updates using any translation system that they choose.

Similarly it would be great if we can create an automated push notification of a diff when a file is changed. This will help other people maintaining the user guide to act with ease.

Thus I would propose as next steps:

  1. Unify the guides for anyone interested. France has put work into this so makes sense to follow their lead. Everyone can agree that the old format is now outdated and needs updating.

  2. Create a system such that a push notification (even just to Slack) notifies when a change is made by anyone to their guide. This might be easy enough with gitbooks.

  3. If people are spending more than 2-3 hours per month updating their guide from another’s translation it will probably be worth exploring automation. The automation will not remove all time requirements so perhaps even waiting for a bigger time commitment to consider automation makes sense.

Ping @Kirsten @MyriamBoure @Rachel @Anna @luisramos0 @Theodore

Thank you @Rachel for your investigation on this… I get your point and also agree with your conclusion for now. So this option of having an automated translation is not relevant for now. But I think we could still evolve toward having only one version we maintain and update, and then local versions we ONLY translate basically, especially as we evolve from instance guides to language guides. And we can then use easily DeepL for instance but still check and correct automatic translation.

I will go back to the other discussion to comment more generally on this and your point @lin_d_hop so let’s keep the conversation there : User Guide as a Living Document - how to keep it current/rich/relevant? and there User guide harmonizations : what could be our new common skeleton?

For now we just work in catalan, but it could be great to try it for spanish! thanks!