Termbase in translation engine not seen in Studio

I'm experimenting with the use of translation engines plus (at the same time) cloud-based TMs and cloud-based termbases. It seems that when I don't add a cloud-based termbase as a stand-alone resource, Studio says "No open termbase" even if the translation engine contains one. 

What I was trying to check is how hits in a termbase in an engine are presented. It seems they are not presented at all, unfortunately. (I've already discovered that hits in a TM which is contained in a translation engine are simply indicated by "NTM", so there is no way to differentiate between such hits and the hits in an MT engine also contained in the translation engine -- which is a pity.)

What I would really like to see is

- hits in cloud-based TMs presented with the source (wich seems to be the case)

hits in cloud-based termbases presented with the source (wich also seems to be the case)

- hits in translation engines presented in a way that I can tell whether they come from a TE-contained TM or an MT engine. 

- Hits in a TE-contained termbase presented so that I can tell that this is the case.

Apart from these observations (the correctness of which I can't guarantee since it's easy to get lost during this type of expermimentation), I am impressed by the results from the SDL Generic MT engines. But I wonder why each language appears 111 times in the list of generic engines -- I would have thought that, since a language always appears as part of a pair, the number of occurrences would be even.

  • It seems that when I don't add a cloud-based termbase as a stand-alone resource, Studio says "No open termbase" even if the translation engine contains one. 

    Correct.  This is a work in progress.  Translation Engines are not properly supported in Studio yet so you have to manually add the cloud termbase.  If you open the file for translation in Studio from the Language Cloud interface then you'll find that the termbase is added automatically, so even there this needs to happen for Studio to be able to use it.

    What I was trying to check is how hits in a termbase in an engine are presented. It seems they are not presented at all, unfortunately. (I've already discovered that hits in a TM which is contained in a translation engine are simply indicated by "NTM", so there is no way to differentiate between such hits and the hits in an MT engine also contained in the translation engine -- which is a pity.)

    You're incorrect here Mats.  I did a small test using your post as the source and added a translation engine that contains the following:

    • TM, populated a little
    • DeepL (add-on in the cloud)
    • SDL MT (from the cloud)
    • Termbase, populated a little

    This is what I see:

    Screenshot of Trados Studio interface showing translation results with error message 'No open termbase' despite translation engine containing one.

    The numbers represent these:

    1. TM result using the same context as Studio
    2. DeepL result.  Unfortunately Studio doesn't see this as NMT which it should do.  This is a bug that needs fixing.
    3. SDL MT result.
    4. Terminology results
    5. Fully translated file populated in the online editor by me

    The other point to note here is that the translation details will tell you where thing came from:

    Close-up of translation detail tooltip indicating a context match from Translation Memory with provider MatsLinder.

    Close-up of translation detail tooltip indicating a draft status from Automated Translation with provider MatsLinder.

    Close-up of translation detail tooltip indicating a draft status from Neural Machine Translation with provider MatsLinder.

    This isn't very helpful as it doesn't tell you which MT engine was used or which translation memory.  It only tells you which translation engine provider, which I called MatsLinder.  I think this could (read as should) be improved by adding this information into the metadata of the TM in the translation results view as well as in these translation details ( )

    And it's further confused because the SDLXLIFF correctly contains this for DeepL where the result was inserted as NMT by Language Cloud in the Online Editor and the specific MT engine noted:

    Close-up of translation detail tooltip indicating a translated status from Interactive origin with Neural Machine Translation by DeepL provider.

    So no surprise you're confused.  We have a lot of inconsistency here that needs to be resolved between Language Cloud and Trados Studio.  Finally, hots from the termbase are not represented in the Translatin Results pane.  They are represented as shown above in 4.  You can see where the termbase came from here:

    Screenshot of Term Recognition pane in Trados Studio showing terms from Mats Linder Glossary with arrows pointing to matched terms.

    emoji


    Generated Image Alt-Text
    [edited by: Trados AI at 7:06 AM (GMT 0) on 5 Mar 2024]