Start a new topic

About the ProjectTM - how to edit it

I have been using CafeTrans for a long time, but I am just starting to try to get the hang of translation memories. I made a total recall table and am storing my project memory in total recall. However, my problem is that the TM is not recognizing translations correctly, for example, it memorizes the translation of the word "eylem" as "the cervical" where it should be "labor". It is doing this for many words and phrases. I am struggling to find a way to edit these entries and correct them. I am translating from Turkish to English. Can the language pair be the cause of this, or is there something I might be doing wrong?

Any suggestions? 

And thank you in advance. I have been reading through the forums and I can see this is a very helpful community. 

Hi Lifeindark,

First of all, please update CafeTran to the latest 2016 Ichiro version. There have been improvements to the "phrase guessing" algorithm. Secondly, take a look at the "Subsegment hits threshold" value in Edit > Preferences > Memory tab. In the latest Ichiro version (and late 2015 versions), the default value is 3 so please increase that number from 2  (the previous default value) to 3. If it doesn't help, increase this value even more to 4 or 5 for your source language.


If you really need to edit a TM, there are two ways:

  • On the go: Just click the blue number to the right of the segment in the tabbed pane you want to edit:

Make your changes in the pop-up window:

  • The whole TM: Close the current project. Drop the TMX file on the Dashboard, and select Edit TMX file. You can then edit the TM like any other project, and you can use a number of tasks:


I take it the TM you mention is the resulting TM from the "Recall Memory" action in the database.

CafeTran tries to guess the meaning of single words from that TM. The more hits there are for a word, and the higher the "Subsegment to virtual threshold" (in Edit | Preferences (Options in Windows?) | Memory), the more accurate the guess will be. But it's still guesswork.

A better way to get results, is to add as many words/phrases/fragments to a Memory for Fragments. For more on this, see and


Login to post a comment