Pre-translation does not work

Question

In a localisation project we were given reference material which I converted into a TM. We were first given only missing sentences to translate. After translation I created a project TM. 
 When creating a project with the complete file to localise, only a very small percentage (~0,1 %) was pre-translated. 
 I had checked TM contents previously (they are organised in such a way, that normally we should get only Context matches), and TMs are activated. 
 What could be wrong?

Paul · Answer

Hi Beate Vigliano Reiter , How did you create the TM? Could it be that there was a penalty being applied which reduced your matches to 99% and therefore missed the default 100% for pre-translation? Regards Paul

Paul · Answer

Dear Beate, 
 
It's going to be tough to give you a precise answer without seeing your files but I can see at least one potential problem with your workflow. If there are any tags in the resx file then converting the way you do could result in the tags becoming translatable text by the time they are in the Excel file, or not there at all. This would obviously result in a loss of leverage when running the resx files through Studio against your TM. 
 
Why didn't you just align the resx files so the correct filetypes were used for your reference TM? 
 
Regards 
 
Paul

Paul · Answer

Thanks for the file Beate... I can see why you won't use Alignment with files like these, very painful indeed! So this is what I did. 
 First I did use Alignment, and I ran an analysis on the file using the aligned TM and the alignment penalty set to zero. I get this: 
 
 I then did the same thing after using the Glossary Converter to convert your Excel file to TMX and then upgrading it. I get this: 
 
 Slightly different, but still not the result you wanted and the differences will be mostly accounted for by placeables. There are also things that should be tags in these files but I did not prepare them at all so everything is translatable text and probably a like for like with your workflow. 
 So, why are these not all 100%? Well, the first thing is easy... duplicate translation penalties. For example: 
 
 This would account for so many 99% matches throughout the file. Then you have things that are linked to segmentation like this: 
 
 So you only have a 69% match against the excel based TM because Studio has segmented the text in the resx in a different way to your Excel sheet. You can see this as #71 is separate to #72 and yet the first match in the TM includes the whole thing. If I use the aligned TM I see this: 
 
 You might not have exactly the same result as you used the bilingual Excel filtype to create your TM... although if I check it using the Bilingual Excel filetype I see this which is to be expected as it will use the cells for segmentation which is different to using the resx natively: 
 
 Anyway... I hope I gave you enough food for thought and you can see why you are not going to get the Context Matches all the way through the file? 
 Regards 
 Paul

Paul · Answer

Hi Beate, when you updated your bilingual Excel file into the TM which setting here did you use? 
 
 I'm showing the defaults... but would be useful to know which you used. 
 Regards 
 Paul

Paul · Answer

Hi again Beate Vigliano Reiter , 
 In the meantime I did this test again and this time followed exactly the route you are following to ensure we don't start discussing something irrelevant. Previously I just wanted to explain where some of the differences may lie but can see I confused things by not having any Context Matches in my example. The analysis results I have the way you are doing it are these: 
 
 So I just need to look at reasons for the less than CM matches. So... first of all, the correct translation for "Nature" is used. Clearly I used a bad example before: 
 
 If you want a thorough explanation of how CM matches work then perhaps review this article: 
 https://multifarious.filkin.com/2013/02/13/100percent/ 
 Now, we still have a duplicate translation issue for quite a bit because Studio is still unable to determine context for many of the translations. To see why you need to dig a little deeper. Take the source word "Code" for example. This appears in a lot of places... like this for example: 
 
 So the question is why is this a 99% when "Nature" above is a CM? Take a look at this one: 
 
 Exactly the same context (previous source and target result) but a different segment. It is therefore a duplicate. In fact this one word occurs 25 times in the document you provided and only two of them are CM. 
 The rest are almost certainly down to segmentation as you used the Excel file to update the TM as explained previously. I checked a few out and this did seem to be the case. 
 I hope this all makes a little more sense for you know and I'm sorry I didn't do this exactly the same way you did in the first place!! Interestingly I also looked using Passolo for the alignment with Daniel Brockmann . This tool aligned the resx files very rapidly and because they are software strings and it's ID based was entirely accurate. But the resultant TM actually provided worse leverage in Studio than any of the other three methods I've tried now. This is probably down to segmentation differences again and so if you did this the benefit would be in translating the files in Passolo too. 
 Regards 
 Paul

Paul · Answer

Unknown said: 
 In a localisation project we were given reference material which I converted into a TM. We were first given only missing sentences to translate. After translation I created a project TM. 
 When creating a project with the complete file to localise, only a very small percentage (~0,1 %) was pre-translated . 
 
 I really think I have confused you a lot in this thread Beate Vigliano Reiter so I came back to the original question and think I missed the point altogether!! I highlighted the relevant parts and think your problem may be related to something much simpler. 
 I think you probably created a Project TM which you used for the Bilingual Excel Project, and then you "think" you created a Project for the RESX files with the TM you updated. However, I think that perhaps you didn't update the Main TM in the Bilingual Excel project and as a result you have used an empty TM for the analysis of the RESX files. 
 Does this seem more likely to be the cause of your problem? 
 I think I got caught up in the enjoyment of comparing leverage results (sadly a guilty pleasure for me!!) 
 Regards 
 Paul

Paul · Answer

Hi Beate Vigliano Reiter , 
 Yes, this is possible. You need to prepare the file for translation using a Translation Memory that has paragraph segmentation. So what you could do is create a new blank TM based on this setting: 
 
 Then use this TM to prepare the projects for translation, then add your real TM for the work you do. This way you'll get the same segmentation in the RESX as your Excel file. 
 Good idea! 
 Regards 
 Paul

Trados Studio > 1. Trados Studio

Pre-translation does not work