Trados 2021 > Filters for exporting translation memories

Dear community:

Why is it that, while we can create filters for searching in translation memories, save them, and use the same for batch delete, there is no way to use those saved filters for exporting? 

There is no option to import filters in export options:

Once and again I need to export tmxs fulfilling a mixed bunch of conditions (project number + a range of dates, etc.). But I cannot reuse these conditions from a file, I must write them each time by adding them one by one. This is quite frustrating, especially when the filters are complex and several conditions come together, or when there are many conditions. Is it ever going to be implemented? 

Thanks in advance,



    Indeed this is something that should be enhanced.  If you haven't already perhaps you can raise an idea for this and ensure it is on the radar of the products team.

    In the meantime here's an idea, created with the help of ChatGPT.  I'm sure it could be improved, but maybe it'll be helpful for you if you do this often.  First take a copy of your TM into a new folder so you don't risk losing anything.  Then ask ChatGPT something like this:

    Write the script that could delete all the rows in the translation_units table apart from those where the creation_user contains Anca and the creation_date is greater than 2011-03-01 and less than 2011-06-01. Then run the VACUUM statement at the end.

    The table information can be found by opening your TM in DB Browser for SQLite.  ChatGPT responded with this:

    FROM translation_units
    WHERE creation_user NOT LIKE '%Anca%'
    OR creation_date NOT BETWEEN '2011-03-01' AND '2011-06-01';


    I can put this into the "Execute SQL" tab like this:

    Execute the script and get a result like this:

    So it deleted 109044 TUs from my TM (it originally contained 109056 TUs).  I click on yes to remove all the spaces (might not even be necessary to run this VACUUM instruction... I just do it to remove the spaces left by the deletion) and then close DB Browser and open the TM in Studio.  It now only contains 12 TUs which is the result of running this query in Studio with a filter:

    Now I just export to TMX.

    I think once you have the script and get used to it then repeating this process by copying your TM, executing the script, opening the edited TM in Studio and exporting to TMX might be faster than creating all your filter rules over and over again.

    Definitely not ideal and I agree we should implement this idea in the product, but maybe this is an interesting solution for all kinds of complex queries that are even beyond the capabilities of the filters.  So things you always wanted to be able to filter on and were not able to?  If you have a play with ChatGPT or any other AI tools to help do come back and share your experience in this forum!

  • Dear Paul:

    Many thanks for your answer, and sorry, but I don't think is feasible :( 

    I find it impractical to have to export the memories, search for segments, clean them, etc., all that on the outside, for so many memories. We can't do that either because of time constraints.

    To give you an idea of the scale of the problem, on this particular occasion, I have to export with a filter involving 50 conditions (project numbers) over 4 memories that are hosted in Groupshare, each with 45 language combinations.

    The search filter, I got it. It would be soooooo easy to have that same filter to export my tmx...

    Nevertheless, I appreciate you taking the time to try and help, Paul :) - but again, I really look forward to the day when this essential option is implemented.

    Warm regards, 



    but I don't think is feasible :( 

    Understood.  Did you create the idea?  Worth doing that and sharing the link so others can vote for it.  Make sure you also add the information on the scale of the problem for you.

    I appreciate you taking the time to try and help

    No worries... to be honest I saw a good opportunity to encourage a little exploration of the AI stuff :-)

Reply Children