I don't know if you can do this from code in your provider, but Studio users can select in Studio's options to "Only show matches from the first translation source that returns matches".  The drawback is that the user will only see one translation memory.

It might be possible to check the "Origin" property of the TranslationUnit object in your search method for whether it is from translation memory, but I suspect that this will not have been set for blank segments until after your MT provider ends up being called.  However, if your users have already run a batch pretranslation, you could check for the origin and/or confirmation status and exit your search algorithm accordingly.

Hello again,

Although I admit that I might not know all the Trados Studio features and options, but I did a lot of debugging.

Firstly, the only thing that I see when the translation method is called is ConfirmationLevel, and thus I can simply not call the translation method, if there is already something in the translation result(a “draft” or “Translated” status), which is far from perfect. It means I can sometimes not show the translation in cases where they might be useful and the scenario where our translators complained the most was the case where they just translated the same string and they have to wait for some wrongly translated AT translation.

About Daniel Brockmanns suggestion. I even tried out SDL integrated AT solution to check weather I am doing something wrong. With neither mine nor the SDL solution did it actually pre translate anything (as the Idea that was implied that it would store an AT suggestion somewhere, in an object), and it actually translated only on clicking on the translation field. The only way that I found to actually translate the text was to use the option “Apply automated translation”, which adds the translation in translation segment. Our translators did not seem to like that the segments where automatically filed.

To sum up: What I seem to be able to do is say to the translators to use pre- translation with the configuration that it auto-fills the translation segments, and then set the plugin no to translate if there already is a draft. This seems sub-optimal and if there is a better way please let me know.

Hi Karlis,

Indeed that was my suggestion - use the option "Apply automated translation" in the pre-translate batch task. This will only be used if there is a no match situation, i.e. if there is no TM result. That seems to be the mainstream use case for machine translation and postediting - it will allow the user to start postediting right away when they open the file, rather than having to wait for results that would be the same as during pretranslation anyway.

Also, as another potenital tip - when you create a package after creating the project, you can remove the link to the MT system in that step (the option in the UI is in the last step of the project wizard). I would hope that this is also available from the project automation API.

I must admit, though, that I'm not an expert for the API details so it may be good if others can chime in as well - I'm rather commenting from a process/best practice perspective how to apply MT and TM best when you use the UI.

Thanks,

Daniel

Hi Karlis,

Indeed that was my suggestion - use the option "Apply automated translation" in the pre-translate batch task. This will only be used if there is a no match situation, i.e. if there is no TM result. That seems to be the mainstream use case for machine translation and postediting - it will allow the user to start postediting right away when they open the file, rather than having to wait for results that would be the same as during pretranslation anyway.

Also, as another potenital tip - when you create a package after creating the project, you can remove the link to the MT system in that step (the option in the UI is in the last step of the project wizard). I would hope that this is also available from the project automation API.

I must admit, though, that I'm not an expert for the API details so it may be good if others can chime in as well - I'm rather commenting from a process/best practice perspective how to apply MT and TM best when you use the UI.

Thanks,

Daniel