Lauro Colasanti 

Lauro Colasanti said:

The translations I receive seem not to take account of the terms defined in my glossary. Why?

Because there isn't a feature in Trados Studio that will do this out of the box.

Lauro Colasanti said:

I have problem in filling the requested data,

It's probably worth you reviewing this site:

https://cloud.google.com/translate/automl/docs

You need to set up the Google Translate API first, and if you want to use dictionaries to influence your MT results you'll need to full understand how it work and not just set up the basic Google Translate.

Lauro Colasanti said:

I'm new on Trados Studio 2022.

Do you specifically need to use Google Translate as your MT provider or are you just looking for any machine translation?

Hi Paul, thank you for your answer.

I'm trying to set up a Google Translation project as suggested in the link you send me.

1. I chose AutoML Translation
2. I created  a new set of Data: I uploaded  a .tsv file with my Glossary, but there are some problems:
   1.  I sent more than 600 couples but got only around 550 because
      1. Error: There are 67 examples failed with error: The sentences contains invalid UTF-8. 
      2. Warning: There are 14 examples skipped with warning: Duplicated examples
   2. I didn't diveded my couples in 3 files for training, evalution, confermations, but send them all in one file
   3. I cannot start training for my couples are less than 1000.

I would ask you:

1. Can I ignore the errors ?
2. Did I do well in sending everything in one file or should I divede the couples in 3 files (I would not know which criteria to follow)?
3. There is no more the difference between glossary (Termbase) and Translation Memory that I learned in Trados Studio?
   1. From my old translation I had created to files of couples:
      1. short sentences in Glossary (TermBase)
      2. long sentences in Translation Memory
   2. Should I put again all together to reach the 1000 limit?
4. What about the others requested data for MT Enhanced Google Advanced?

I was using Google Translation just because is more popular and because I have a free trial account for 3 months. If you have others suggestions please do it.

Thanks again, Lauro

Hi Paul, thank you for your answer.

I'm trying to set up a Google Translation project as suggested in the link you send me.

1. I chose AutoML Translation
2. I created  a new set of Data: I uploaded  a .tsv file with my Glossary, but there are some problems:
   1.  I sent more than 600 couples but got only around 550 because
      1. Error: There are 67 examples failed with error: The sentences contains invalid UTF-8. 
      2. Warning: There are 14 examples skipped with warning: Duplicated examples
   2. I didn't diveded my couples in 3 files for training, evalution, confermations, but send them all in one file
   3. I cannot start training for my couples are less than 1000.

I would ask you:

1. Can I ignore the errors ?
2. Did I do well in sending everything in one file or should I divede the couples in 3 files (I would not know which criteria to follow)?
3. There is no more the difference between glossary (Termbase) and Translation Memory that I learned in Trados Studio?
   1. From my old translation I had created to files of couples:
      1. short sentences in Glossary (TermBase)
      2. long sentences in Translation Memory
   2. Should I put again all together to reach the 1000 limit?
4. What about the others requested data for MT Enhanced Google Advanced?

I was using Google Translation just because is more popular and because I have a free trial account for 3 months. If you have others suggestions please do it.

Thanks again, Lauro

The file I uploade is in TSV

Lauro Colasanti 

Lauro Colasanti said:

Can I ignore the errors ?

I would suggest you are wasting your time unless you use the recommended number of TUs (Translation Units = couples) and do not ignore the errors.

Lauro Colasanti said:

Did I do well in sending everything in one file or should I divede the couples in 3 files (I would not know which criteria to follow)?

I have no idea, I don't really understand what you're doing.

In general I think you should be aware of the costs here.  Training an engine will run you into hundreds of euros and you also pay for the results when translating.  Are you aware of this?

I'd also note that we are currently in the process of separating Google and Microsoft from MT Enhanced and creating two separate plugins for these providers, using a better UI and UX.  It's a little way off before we're done, but if you want to make use of Google anyway I'd recommend you hold off until we have released improved plugins.

Lauro Colasanti said:

I was using Google Translation just because is more popular and because I have a free trial account for 3 months. If you have others suggestions please do it.

Well, first of all you already have free machine translation using Language Weaver out of the box in Trados Studio.  Have you tried that?  You'll find it under Cloud-based resources...

Then select Machine Translation from the dropdown:

That will get you free machine translation, but not the ability to use dictionaries.  There are paid versions if you're interested.

Alternatively, depending on your language pairs it may be worth looking at ModernMT as this is adaptable MT so it learns from your corrections as you work:

https://www.modernmt.com/translators/

The plugin is here:

https://appstore.rws.com/plugin/99/

There are more on the appstore ad I'd recommend you take some time to review them as some may provide exactly what you want : https://appstore.rws.com/?q=&category=6

Hi Paul, thanks again for your advice.

I put my Termbase and Transaltion Memory together and added few new phrases, so I reached the 1000 limit. I then uploaded the new TSV file. Now I'm training the dataset.

When uplaoding the data I was offered the choice to use 3 different files (for training, test and validation) or only one and the splitting is made automatically by Google. I choose 1 file.

I don't know exactly for the cost. Now I have a free trial for 3 months. Later I will see.

I'm using a free trial version of Trados Studio 2022 so I cannot use neither MultiTerm, neither Language Weaver, neither Cloud-based Resources...

I will try the other LT Machines.

Lauro