Annoying bug: entries from Trados-based TMs not recognized during pretranslation
Olaf Reibedanz Argentina Local time: 01:14 Member (2003) English to German + ...
Dec 26, 2011
Dear colleauges,
There is a really annoying bug that exists both in DVX 1 and in DVX 2 and that has been driving me crazy for years:
When the client has provided me with a TM (in .txt format) that he has exported from Trados, and when I import that TM into my Déjà Vu TM and do a pretranslation, many fuzzy matches (even very high ones with 99% concordance) are not recognized by Déjà Vu!
This is extremely annoying and has already cost me thousands of euros, because when I outsource those segments to other colleagues, I pay them the normal rate, whereas I only get a fuzzy rate from the end client.
To avoid this problem, what I ususally do is check all the segments one by one, by looking for high fuzzy matches in the AutoSearch window, making sure I do not outsource those segments but, unfortunately, I sometimes forget to do this when working under great time pressure.
Has anybody else encountered this problem and, if so, have you found a good workaround?
Thanks a lot for your help!
Kind regards,
Olaf
[Edited at 2011-12-26 16:01 GMT]
Subject:
Comment:
The contents of this post will automatically be included in the ticket generated. Please add any additional comments or explanation (optional)
Do you mean that they are not found as fuzzy matches, or that they are not found at all?
[Edited at 2011-12-26 17:09 GMT]
Subject:
Comment:
The contents of this post will automatically be included in the ticket generated. Please add any additional comments or explanation (optional)
RieM United States Local time: 00:14 Member (2006) English to Japanese + ...
Do you...
Dec 26, 2011
Hi,
I haven't had the same problem yet, and so maybe I'm doing something different...
Do you pre-process files within Trados and import them as ttx or sdlxliff？ Or do you just import raw files and apply pre-translate?
Do you have the same problem with the exact match? Or only with fuzzy matches?
How do you set up the Pr-translate window, especially under "Overwriting"?
Rie
Subject:
Comment:
The contents of this post will automatically be included in the ticket generated. Please add any additional comments or explanation (optional)
Olaf Reibedanz Argentina Local time: 01:14 Member (2003) English to German + ...
TOPIC STARTER
Further details
Dec 28, 2011
Dear Lorenzo and RieM,
Regarding your questions:
1) I don't do any pre-processing of the Trados TMs. I directly import the "raw" .txt files into DVX.
2) The Pretranslate settings are the same as usual. But I only have this problem with .txt files that come from Trados.
3) The problem only concerns fuzzy matches, but not exact matches. Some of the fuzzy matches are recognised in the AutoSearch window, but not inserted into the empty cells during pretranslate.
However, I must correct myself in one aspect: This problem only occurred in DVX 1, but not yet in DVX 2. In fact, in this particular case, I simply forgot to untick the option "Accept only exact matches". But I will go through my old projects (the ones I did with DVX 1) and try to find the projects where the problem occurred, and then see what happens when I pretranslate the same files with DVX 2. I will keep you posted.
[Edited at 2011-12-28 02:24 GMT]
Subject:
Comment:
The contents of this post will automatically be included in the ticket generated. Please add any additional comments or explanation (optional)
Grzegorz Gryc Poland Local time: 06:14 French to Polish + ...
DVX matching algorithm flaw.
Dec 30, 2011
Olaf Reibedanz wrote:
Regarding your questions:
1) I don't do any pre-processing of the Trados TMs. I directly import the "raw" .txt files into DVX. (...)
IMO it may not work.
In this caseyou should always consider a preprocessing in Trados.
The matching algorithms are different in Trados and in DVX and it's perfectly possible you may not get the expected matches.
E.g., if you consider an example like:
EN: I see the cat.
FR: Je vois le chat.
Trados (Workbench) will propose a 95% match for
EN: I see the cats.
while DVX will recognize only a 75% similitude.
The problem is DVX has a fundamental matching flaw and if a word differs only by one letter, it's considered as a 100% different word which is obviously false in most cases.
I.e. the fact the sentences like I see the cats. and I see the dogs. return both a 75% match for the example above is aberrant.
So, in order to make DVX more sensitive, you should lower the matching threshold (normally, I use something like 50-55%).
PS.
I did receive your e-mail, I'll contact you soon.
I simply got a 3 weeks long heavy infection, I'm slowly getting alive now...
Cheers
GG
Subject:
Comment:
The contents of this post will automatically be included in the ticket generated. Please add any additional comments or explanation (optional)
To report site rules violations or get help, contact a site moderator:
SDL Trados Studio 2011 is the latest market-leading translation memory software from SDL. Now with Track Changes, Bilingual Word Files support, new Display Filter, AutoSuggest and more great details.
DVX2 Professional is the most popular version of Déjà Vu X2 and with good reason. Fast and flexible, Déjà Vu X2 Professional combines Atril’s Intelligent Quality technology with an array of powerful, customisable productivity and quality assurance
The translation workplace 
