On Saturday 25th Sept, we are completing planned maintenance work on the SDL/RWS Account platform. The outage is expected to last from Saturday 03:00 BST to Monday 27th Sept 11:00 BST. During this time you will not be able to access SDL accounts or activate / deactivate licences. Apologies for the inconvenience this causes.

Translation Memory Updates Option: Merge vs Overwrite

Hello. I have a question for the batch processing option when updating translation memory.

So far, I kept using the overwrite options. It is because I understand the overwrite option will add new translation units, overwrite the existing translation units if the same source segment but different target segment, and keep the existing translation units unchanged if the same source and target segments.

However, I found some translation units that don't exist on the translation memory.

Today, I tried to translate the words "Supported RAID". It is the words frequently appear on the document so far I worked. Of course there should be the translation unit on the translation memory. However, I realized it is not added on the translation memory. There is similar translation unit "Supported RAID Level" but not exactly the same unit "Supported RAID".

I used the overwrite import option for a long time but I tried with the add option today using the sdlxliff file from the previous project. Then, "Supported RAID" is finally added to the translation memory as a new translation unit today.

I don't understand why it happens.As my request, I don't want to increase the duplicated translation units. When only the target segment is updated, the existing translation unit with the same source segment should be updated.

Could anyone tell me the differences between merge/add and overwrite options, and what update/import option will be the best one to meet my request?

  • Hi

    Thank you for your description.

    I think understand what it is you are saying - allow for me to summarise.

    You are translating Supported RAID and initially it is not saving in your TM.
    Only when you run an update TM and did you find Supported RAID in your TM
    However this newly added entry, did not come from your current project but from a previous project / sdlxliff

    Let me break down my responses.

    1. Definitions of TM import options

    Merge translation units

    Replace with the latest translation only TUs that have matching context. All the other TUs with the same source segment remain unchanged. This is the default and recommended option for updating TMs.

    Add new translation units

    Add translations that have the statuses below as new TUs to the TM. This means that the existing TUs in the TM are not overwritten.

    Overwrite existing translation units

    Import the new TUs and overwrite the existing TUs in the TM.

    Source: SDL Documentation (rws.com)

    This means its its only Add new translation units that would respect if target translations differ. Meaning same source sentence localised differently.

    2.Task of updating TM

    You mention Overwrite existing translation units - which tells me you update your TM's at project / file level.

    This is different to updating your TM on the fly meaning as you localize each segment throughout your file / project.
    By extension I am not surprised you mention you are translating Supported RAID and initially it is not saving in your TM.

    Saving Translations on the fly is a choice that you can control.
    Please check File - Options - Editor - Automation as shown below.

    As its not ideal if you import other translated files in order to ensure your TM was fully updated.

    I hope some detail here is of help? 
    If you require further assistance please do let me know


  • Hi 

    Thank you for your reply.

    > 1. Definitions of TM import options

    "Support RAID" had appeared in the previous projects many times so it is not a new translation unit on the current project.

    I may have selected the overwrite option when "Support RAID" initially appeared, in such a case, will the initial translation units not be added to the TM?

    I understand "Add new translation units" will add any translation unit pairs if the source and target segments are different. If so, I'm wondering if many duplicated segments will exist on the TM. I understand many of the translation units that contain the same source segment will decrease the match percentage. I want to avoid decreasing the match percentage as much as possible, and want to only remain the latest translation unit to the same source segment.

    In this case, "Add new translation units" is an only option?

    > Task of updating TM

    Will the automation option you suggested be for the project TM? Or will it be for the main TM as well?

    Anyway, I have enabled the option already. However, most of project I work will be the package. The TM will be updated using the project packages translated by other agents. In this case, will the option work to update the TM, too?

  • Just to give you some peace of mind as you are mentioning both Project & Main TM's as well as packages.

    To clarify: When you open a project and you run the batch task "Update Main TM" - it will ALWAYS use the the translations from the project sdlxliff files.
    Even if the files cam from other agents as you mentioned.

    But yes regardless if you call the update of Project TM or Main TM - ensure that your update settings are fit for purpose.

    Hope this helps