Hi Yasemin Ak 

Thank you for reaching out! 

This can happen if you have the Cloud resources and/or the Language Weaver provider activated. See this articles that explain a bit what happens and what you can do:

https://gateway.sdl.com/apex/communityknowledge?articleName=000018765

https://gateway.sdl.com/apex/communityknowledge?articleName=000008379

Basically the machine translation provider, Cloud Resources or LW Plugin, is used by Trados Studio like a Translation Memory, searching it at every segment. If you perform also pre-translate and then go segment by segment, you can end up with extra usage. 

You can monitor your usage in Language Weaver as shown in this article: https://gateway.sdl.com/apex/communityknowledge?articleName=000013218

My suggestion is to try on a smaller document, to see how what you in do in Trados Studio affects your LW machine translation usage. 

Hope this helps, 

Radu

Hi Radu, 

Yes, I use Pretranslate (to use 'change all' option for certain terms) and then I go segment by segment. In this case, Weaver must be using its resources two times for each segment, thereby doubling the used character count. I have not considered it before. Thanks for clarifying the issue.

Kind regards, 

Yasemin 

I find it really helpful to learn about this problem. So, we should not use Pretranslate function to avoid double counting, should we? Sorry for asking this, how should we use Machine translation efficiently? (For now, I have not started using it yet due to some problems with installation but just want to exclude extra usage in the future). Thank you.

Why don´t you just deactivate Language Weaver resource upon pretranslation???

You configure the project settings as: TM as a priority source (if you have any, of course). Then you add LW (in my case, it´s DeepL) as a source of machine translation, which will be applied only in case nothing is found in your memory with the match% you define.

To define a TM as a main source, use arrows, so that your TM (or TMs) appear FIRST in the list of resources:

Now go to Batch tasks-Pretranslation and don´t forget to mark the box: Apply machine translation as in p.3 below. If you don´t, only TM results will appear after pretranslation:

It´s in Spanish, but you can deduce:

Step 1: define match % of fuzzies that will be taken form your TM(s). Important: TM-based translations will not (or at least, MUST NOT, subtract from the 6 mln.characters you have).

You can even go further in case of documents with lots of numbers, alphanumeric data and other non-translatables: before running pretranslation, filter them, just copy them to the target segments and change their status to Translated.

Then be careful to check option 1 "Maintain the existing translation", so that Trados would skip them during pretranslation.

Step 2: check or uncheck the confirmation boxes (depends on how much you trust your TM, of course); I personally prefer to go through CM and 100% and when they are not confirmed, the program does not skip them (you also have filters for those, so it´s up to each translator),

Step 3: Don´t forget to check this option. If not, MT will not be applied, but MAY count the characters from your 6 mln."account", since LW will be activated in your project settings, so be careful with that.

When you are done and have reopened the project in the editor window, go to the Project settings again and remove it from the list of the language resources. 

Hope, this helps. 

Best regards!