1

Segmentation, pretranslation and populating strings

Federica Sozzi

March 18, 2021 16:01

None

Follow

Good afternoon,

What are the best practices for automatically populate strings with pre-translated text in a job?

We often reuse paragraphs, which were already translated, adding some changes (sentences or phrases) and we would like to have the identical parts populated automatically.

E.g.: I uploaded a file simply applying minor changes, like punctuation, to an already translated one. According to the word count, 90% of the text is pretranslated, but in the editor the majority of the strings are empty and not populated. 

How does the segmentation affect the matches?

Thank you

• Facebook
• Twitter
• LinkedIn

1 comment

• 0
  
  Project Managers April 09, 2021 08:32

  Hi Federica,

   

  Since changes were made to the text, the similarity with the translation memory might have dropped below the pre-translation threshold. I would suggest using the "Translate from memory" batch action and lowering the similarity rate:

  

   

  If you are referring to the fact that you expect the segments to be populated automatically, without having to look and copy from the translation finder, this option in the translation editor settings will help:

  

   

  I don't know if the breakdown of the word count analysis is a concern here.

  - if you expect a higher pre-translation rate, the threshold for pre-translation should be higher. Those small differences will be shown as fuzzy pretranslations.

  - Otherwise, if you do not want to have fuzzies, the option in the editor settings will be of help when working on the file. You will need to revise those carefully, especially if there are numbers or changed elements difficult to spot at first sight.
  
  

  When uploading new files to an existing project, the threshold approach will work well. When a new memory has been added to the project in the meantime, the option "Translate from memory" is needed as a "general baseline check".
  
  

  However, when working in parallel on two files in the same project, the option in the Translation Finder seems more appropriate, bringing the best matches to your view. And if a proposal is wrong, you can always correct it on the spot and move on.

   

  Best,
  Andrea