Comparing pre-existing translations

Currently I am working on a very difficult and rather lengthy technical text that no less than three colleagues worked on before. The client is not satisfied with any of the manuscripts yet, and I see why. Still, each of the pre-existing translations contains lots of good research, especially terminology-wise. So ideally I would set myself up for the job in a way that I can compare each original string with all three pre-existing translations at the same time so I can pick and choose the best parts from each. But how?

3 Replies Latest Replies: 17 Mar 2019 2:10 PM by Alison Field
  • Hi 

    The way I would do this is as follows:

    If you have 3 SDLXLIFFs*, one from each of the translators:

    Create a new project and add all the files to it, plus a file which only has the untranslated source. If they have the same filename then change each according to the translator, either add their name or give them a number. You may favour one translator over the rest, in which case give him or her the number 1, etc.

    Create a new termbase and add it to the project.

    Highlight all the files in the Files view. If you've named them according to the order in which you want to look at them, they will be listed in that order and also open in that order.

    Right click on any part of the highlighting and select 'Open for Review' from the dropdown menu that appears. This will open the files together as one big SDLXLIFF. I suggest you have the untranslated file last. This is called a virtual merge and opens 

    Now pin the Advanced Display filter open so you can use it. Select Translated; All repetitions under File Attributes

    then add an initial letter to the Source Content filter as follows:

    The ^ is a 'regular expression' that means that 'a' is the first letter in the segment. Remember to tick the Regular Expression box.

    Then click 'Apply Filter' and only all segments that begin with 'a' will be displayed.

    You can then look at what is listed and narrow it down further by adding a second letter so that you only have a handful of identical source units visible at a time. That way you can compare without scrolling.

    You can then translate the empty Target segment in the best way, having read all the translators' input. Or if you prefer one translation to the others, you can hit Ctrl+Enter in that segment and it will autopropagate all the others.

    You need to have 'Auto-propagate from: First segment...' ticked under File>Options here:

    I would also  use the other settings I've highlighted.

    *If you don't have any SDLXLIFF's you probably have Word documents for example, then you need to first align the source document to the target document x 3, the method for which is described here:

    Aligning files

    It's too long a subject for me to cover in this post. 

    If you only have hard copy documents, you would need to first scan them into a word processing format.

    You could also scan to PDF then open the PDFs in Studio which would create an SDLXLIFF, from which you could create a Word document that you could then process via alignment, etc.

    Then you can go back to the first instructions at the top and continue to open all the SDLXLIFF's together as I described there. It isn't necessary to create a fourth untranslated SDLXLIFF, you could just decide to amend just one of the translation SDLXLIFF's to be your final document instead. The first or last and stick to it.

    I hope the above meets your needs at least somewhat!

    All the best,


  • Another alternative to Ali's excellent suggestion would be have a translation memory for each of the translators, using all 3 TMs at once in your project, which would penalize each of the multiple translations by 1%, showing you all 3 alternatives as you go into each segment, each with a 99% match value.


  • In reply to Nora Díaz:

    OOH yes  of course!

    I love the way there's always more than one workaround to achieve something that Studio wasn't actually created to do... ;-)


    Ali :D