Shmuel Pfeifel

Shmuel Pfeifel said:

To may astonishment, not all segments were transferred to the first file, even if they were identical.

I would have to say that this is very unlikely that they were exactly the same.  Perhaps there are formatting tags that you are not displaying.. or something like that?  If you can extract a few sentences from your text file, and then copy the same segments fom the word file to a new file and reproduce then we can test this.

Will be happy to be proven wrong :-)

Hi Paul,

Obviously they were identical in terms of text. they were a conversion of the same file. Should not this at least appear as a very good partial match? also what would copy/pasting the same text to a word file achieve?

Shmuel Pfeifel

You said about the original:

Shmuel Pfeifel said:

I converted it to plain text

Then you said this about the new one:

Shmuel Pfeifel said:

Subsequently, the client sent me a word file of the same PDF file with minor conversion variations from my file.

A plain text file and a Word file are not the same thing.

I can only base my comments on the information you provide unless you also share some files to prove your point.  In this case you should provide the following:

1. sample plain text file
2. sample converted word file

You only need to provide this for the affected sentences, so the ones that don't match in your TM.

Shmuel Pfeifel said:

also what would copy/pasting the same text to a word file achieve?

It would give me a small sample to work with.  That's all... I don't want your entire file, just the offending parts.  If Studio really gets this wrong it should be simple for you to reproduce this.

Ok, just to be accurate, both files used for creating both projects are word files. One is the result of a conversion that I made using a ABBYY into plain text then saved to wordx. The other is a wordx file that the client created and provided to me. They both have slightly different formatting and tagging issues and hangups due to the limited capabilities of PDF file conversion software, but they are otherwise identical in terms of text.

However, this does not seem to be the issue. Even if they are different in terms of tags and cannot be populated during pre-translation, these segments are not so different as not appear as a TM match at all (I am tuned to 70%)

I can obviously send you the offending segments of both projects, but, do you expect a segment to contain a fatal error that would prevent it from being updated to the TM? without any error message to the user? that sounds extreme. How can you assure, then, that this does not happen right left and center with all documents?

Since I have them on my machine, wouldn't it be more effective to recreate the problem with those two projects on my machine? We can repeat the process back and forth between them and you can watch this happen.

Shmuel Pfeifel

Shmuel Pfeifel said:

Since I have them on my machine, wouldn't it be more effective to recreate the problem with those two projects on my machine? We can repeat the process back and forth between them and you can watch this happen.

No... I'd like to verify the issue being Studio or not first.  If I can't reproduce it or explain it then we can look at your machine.

Shmuel Pfeifel said:

do you expect a segment to contain a fatal error that would prevent it from being updated to the TM?

What fatal error?  You have not mentioned an error before?

I wonder what you are trying to find in that text that is not Studio's fault. Failure to detect a corrupted feature of a document that prevents Studio from processing it properly (saving it to the TM), as well as failure to record and save a segment in a project to the TM for any other reason, both require notification to the user. I did not receive any notice of any problem of any kind. One way or the other Studio should have done something that it did not do, either save the information properly and than retrieve it properly, or provide appropriate notices of failure to do so.

Regardless, I will prepare a small word file for you tomorrow and send it here so that you can check what you need.

Hi Paul,

As promised I am attaching a file for your review.

The file contains two examples of  source file excerpts of segments remaining blank in the second project after they were translated and populated to the main TM in the first project. Not even a partial match. Trados failed in this way for about 30% of the second file.

Please note that the parsing rules for preparing the two files are the same (same template), so the resulting segmentation is also the same.Paul sample.docx

Enjoy!

Shmuel Pfeifel

Thanks for the doument.  They are not exactly the same because one has line breaks in there and the other does not.  But despite this I still get a 98% match which seems fairly appropriate to me given the lack of line breaks:

Are you sure you have your TM results set to et matches below 100%?  How were you attempting to get the translations... by interactive work or a pre-translate batch task?  If the latter that's probably where you have gone wrong as the default is 100% for a pre-translate:

Now you begin to see the problem. My pre-translate is indeed set at 100%. My search translation value is set at 70%. I was expecting to get part of the matches from pre-translate, and I did, and part during interactive work, to account for the small difference. However, despite the 98% match, I did not get anything during interactive work. So nothing seems to be wrong with pre-translate as far as I understand. the problem is that I am not getting the rest of the TM matches below 100% during interactive work.

I don't see the problem at all!

I get this match whether I pretranslate or work interactively. If you have time right now we can take a quick look at your machine and try to get to the bottom of this?