How to handle really large TMs?

Question

This question came up after advising a colleague about how to convert hundreds of Transit projects to Trados: 
 What is the best way to handle really large TMs (say 500,000+ TUs)?

Lydia Simplicio · Accepted Answer

Hi Hans Lenting Not sure how to answer this as there are so many schools of thought at what works well for 1 may not be sustainable for another. Its OK to have 1/2 million Tu's but I would wonder how many TM's of that size do you have and even more importantly how many TM's of that size would be part of any given Project. You may find this of interest or to certainly be aware of: https://gateway.sdl.com/apex/communityknowledge?articleName=000005098 
 Generally when it comes to large TM's users wonder about performance followed by appropriate leverage/TM content reuse. Example you may have 1 segment that has 5 hits / possible candidates. Which one will you select? Users consider things like: 
 
 splitting large TM's into smaller targeted TM's (based on domain / customer) 
 working with more than 1 TM in a project but in hierarchy 
 sort ordering results based date most recent added 
 utilising field values so there is metadata that complements your hit results Going back to my scenario where may have 1 segment that has 5 hits / possible candidates, it could be Trados Studio helps you further target specific ones based on metadata data. Further explanation: Project you are working one is categorised as ABC In your TM you have 5 potential hits: 1 TU categorised as 1ABC 2 TU categorised as 1XYZ 3 TU categorised as 2XYZ 4 TU categorised as 2ABC 5 TU categorised as 3ABC 
 
 Due to project/TM settings and appropriate filtering you could reduce the 5 potential hits down to 3 based on ABC. Filtering and putting TM hit penalties in place helps ensure that if you do work with large TM's you are able to eliminate and target specific TU's for appropriate reuse. 
 Documentation around TM filters can be found here: https://docs.rws.com/813470/570411/trados-studio-2021-sr2/specifying-tm-custom-fields https://docs.rws.com/813470/332288/trados-studio-2021-sr2/creating-tm-filter-penalties 
 If you had a specific concern please update this post so we can address it 
 Have a good day 
 Lyds

Jerzy Czopik · Answer

I have several TMs with reasonable size of several hundred thousand segments and a few with over 1 million segments. Huge TMs are not really workable well, so what I do is creating project translation memories in each project. This way I can utilize these really large TMs in a sophisticated way. A project TM is small and thus fast and works even over the network, as the complete project including the project TM is located on NAS. The huge mother TM is stored locally on a SSD M2 HDD of my PC and still searchable for concordance. This method allows to combine huge TMs with reasonable working speed. 
 For the future I do not feed the biggest TMs anymore.

Trados Studio > 2. Trados Studio

How to handle really large TMs?

Top Replies