DeepL Provider: support glossaries

callumbwhyte commented 7 months ago

We have a client that uses some domain-specific language in their content that does not translate well.

DeepL has a Glossary feature making it possible to provide an alternative list of translations to the engine.

You must pass the glossary_id field in the translation request (see docs), but Translation Manager does not offer this as an option (or provide a way to add programatically) 😔

It would be super awesome if we could optionally set a glossary ID in the DeepL connector settings at the very least, or a way to append this programatically... Even better would be the ability to set a different ID per language (as you may have different glossaries per source + target language)... Even even better would be for Translation Manager to manage the contents (+creation) of the Glossary itself as it needs to be created by the API – but I know I'm getting carried away there!

(PS the link to the DeepL docs on the Providers web page says "Google Docs" 😉)

KevinJump commented 7 months ago

Yep - been looking at this, does a couple of things, but wondering about the interface

Added IDeepLConfigurationProvider interface, that lets you define the options just before anything is sent to translation.

this has a default implimentation :

    public TextTranslateOptions GetOptions(string sourceLanguage, string targetLanguage, bool IsHtml)
        => new TextTranslateOptions
        {
            TagHandling = IsHtml ? "html" : string.Empty,
        };

but it hopefully going to have enough info that if you wanted to build something to do per language glossaries your could ?

I was going to also add some glossary options to the connector, but the per language pair thing probiblty makes it a bit more than a quick fix (we will need to have something where you pick both languages and then pick the glossary id, not hard, but not super quick).

I need to set aside a little bit of time for that, not least because i will need to go purchase the pro version to make sure i get all the options right!

at the moment, would you be ok with something that let you set glossaries programmatically ?

KevinJump commented 7 months ago

OK,

So we have a beta : (this is v12.x - but for deeplL v12 works on v10 - v13 of Umbraco)

https://www.nuget.org/packages/Jumoo.TranslationManager.Connector.DeepL/12.5.0-beta02

this has two features to help (i think).

it has the option to 'auto map' the glossaries, so it will attempt to find a glossary that matches the source and target languages being used, and if it finds one it will add it to the options.
it allows you to impliment a IDeepLConfigurationProvider interface (and register it in a composer) to replace how the options are defined for a job. (There is a base class DeepLConfigurationBase that you can use that does most of the default config, so you can choose to just do the bit that matters.


public class CustomDeepLConfigProvider : DeepLConfigurationBase, IDeepLConfigurationProvider
{
    public CustomDeepLConfigProvider(
        TranslationConfigService configService,
        Lazy<IDeepLTranslationService> deepLService) : base(configService, deepLService)
    {
    }

    public override TextTranslateOptions GetOptions(string sourceLanguage, string targetLanguage, bool IsHtml)
    {
        var options = base.GetOptions(sourceLanguage, targetLanguage, IsHtml);

        // custom things here ?
        options.GlossaryId = "my-id";

        return options;
    }
}

but maybe the mapping means you don't need this ?

also - added some more of the options, to the config