TMX from Wordfast = import failed

Question

Dear All, I have already imported a good number of TMX derived from my Wordfast without any problems whatever. All of a sudden, with the most important TM ever, Trados won't even display any error message; Trados behaves, as if it did not recognise the TM at all. When I try to open the TMX file instead of the regular window (i.e. the four-step Upgrade Translation Memory), there is just a split second image of the window that disappears and no recognition results. 
 The size of the tmx file is 100 MB. I am puzzled. Would you get me out of this rabbit hole? Regards, 
 Maciej

Paul · Accepted Answer

Maciej Śpiewak 
 The problem seems to be illegal chars in eleven of the TUs in the TMX... for example: 
 
 I removed the offending TUs completely and it all imported correctly. Studio carries out some validation on the file prior to importing, and will fail with errors in the XML (as above), and will remove TUs that are duplicates, missing source or target etc. So the final result is this 
 
 I sent you a link to where you can download the repaired files. 
 Paul

Jerzy Czopik · Answer

Use any editor able to validate XML, for example Notepad++. Or Okapi Olifant - it should also detect such problems. There are also other dedicated TMX editors, just ask uncle Google :) 
 BTW, the illegal characters are usually the same in the complete TM, so using Notepad++ is a good idea, as you can use decent search & replace function there.

Paul · Answer

Maciej Śpiewak said: I have a TMX file not an XML file... 
 TMX is XML... look at the header: 
 
 Maciej Śpiewak said: And on top of this, I first heard of Sublime from you today. I have no literacy, no skill and getting disheartened slowly 
 Basically the problem is that Wordfast is exporting invalid XML so you have some segments in your TMX that you either need to remove, or fix. These are the invalid chars in XML: 
 
 In your file I removed the TUs instead of replacing the & with &amp; because they didn't look as though they should be there anyway and the TUs I removed were actually same source same target, missing source and only target, or just gibberish (not sentences at all)... so removing them was faster. To do that you need to understand the structure of a TMX. It's pretty simple, each TU has this structure.. I marked each one with a red square: 
 
 You can remove a complete TU (everything in the square) without damaging the file. So my process was this: 
 
 Identify which TUs had the invalid segments in them one at a time 
 Decide whether it was worth attempting to replace the invalid char with the correct entity, or some other character, or not 
 Replace the char, or delete the entire TU (what's in the red box) 
 
 That's it. Sometimes you can do smart replacements and perhaps the one from Jerzy with Notepad++ is clever enough to let you replace only the ampersands that should not be there, I don't know. But that's the approach I took. 
 I imagine others have their own way too... Evzen Polenka will almost certainly have a smart way of tackling these issues. 
 But one thing is for sure... if you want to be able to handle things like this you need to read up a little on why it's necessary (google invalid XML), read up on how XML is structured ( https://www.w3schools.com/xml/ ), and learn about TMX ( https://en.wikipedia.org/wiki/Translation_Memory_eXchange ) 
 Paul

Paul · Answer

Anna Brancaleon 
 Zip the files first Anna.

Trados Studio > 1. Trados Studio

TMX from Wordfast = import failed

Top Replies