Thread: Translating AutoCAD files with MemSource

Hi All! A few tips on how to associate TranslateCAD and MemSource.

As you may know, TranslateCAD is a useful tool that exports "editable" content from AutoCAD files, making it ready to process the translation with the help of a CAT Tool. My experience so far had been with SDL/Trados, but I found some hurdles to jump when it came to use MemSource.

• Use "trans1" TXT file: Same as for SDL. No need to prepare the file any other way.
• Disable MT: Do not use Google as it disrupts the token encoding (##000001##) by adding extra spaces.
• Pretranslate non-translatable and leverage matches from the TM without blocking full/ICE match (101%) matching segments.
• Fix added characters from target: Not sure why, MemSource added a couple of symbols to the first token that need to be deleted from the target file.
• Save TXT using "unicode" encoding: Open TXT and save it as "unicode" file (instead of an ANSI one).

Why do I advocate not blocking full matches? AutoCAD callouts for drawing may need twisting of word sequence in the target language. You need to see things in context. Sometimes you may even get to the point of realizing there is a mistranslation when looking over/proofreading the PDF in the target language.

Hope it helps!

Hi Gentle, how you doing?
Thank you for the detailed explanation.
You say that MemSource added a couple of symbols to the first token that need to be deleted from the target file... How you noticed that? you tried first doing a pre translation? to make sure you got all the text translated?

A few days ago, I tried exporting text from an autocad file and then I did a few changes on the txt file (as if it were the translation) and when I tried to import text back to the autocad file I couldnīt see the text I had "translated".
I thought that that text might have been in a hidden layer? Any idea?

Originally Posted by analaura

A few days ago, I tried exporting text from an autocad file and then I did a few changes on the txt file (as if it were the translation) and when I tried to import text back to the autocad file I couldnīt see the text I had "translated".
I thought that that text might have been in a hidden layer? Any idea?[/COLOR]

I faced a somewhat similar situation when trying to replace text straight into the TXT file and reimporting it back with the help of TranslateCAD. I only tried a few lines, but it would not show as translated in the AutoCAD file. However, I decided to put it through SDL and pre-translate it with MT, and it suddenly worked. However, I still don't know why!!! But, my advice would be to try to translate everything and with the help of a CAT Tool. If you'd like to use MT with MemSource, you face the issue described above, for which I've found a workaround. But, that will go in my next chapter!

Thank you Gentle.
Yes, I would go ahead and do a trial with the Memsource pre translate function, and Iīll see what happens. I donīt know why either that text didnīt get translated...

Originally Posted by analaura

Thank you Gentle.
Yes, I would go ahead and do a trial with the Memsource pre translate function, and Iīll see what happens. I donīt know why either that text didnīt get translated...

So, here goes my next tip...

Don't forget that MT (automated translation) will corrupt the encoding! The workaround I've found unfortunately adds a few steps, as follows:

1) Prepare a project with MT "disabled.
2) Load it with the "trans1" TXT files TranslateCAD suggests as your input for the translation.
3) Pre-translate files by simply pre-translating non-translatables (ie., clicking on Pre-translate, then on the Pre-translate where empty, and chosing only Pre-translate non-translatables from the Machine Translation section of the pop-up window. Also, check "Set segment status to confirmed for non-translatables".
4) Edit the project and enable MT (Google, for instance).
5) Pre-translate files again, this time you may use a combination of TM (whichever fuzzy level you prefer) and MT (below that certain level, empty segments will be populated with Google).

This way, Google will find the encoding filled out with the correct code (will not have the chance to alter it) and you will be able to use TM+MT.

Remember that AutoCADs are always tricky, as segmentation is arbitrary (similar to images callouts) so that you may need to totally translate segments by turning them upside down.

Hi Gentle,
Thank your for the tip...
Now, what if instead of pre-translating the non-translatables I go for the option "add copy source where empty"? will that "save" the right encoding? and then translating w/o MT. Would this be another option?

Originally Posted by analaura

Now, what if instead of pre-translating the non-translatables I go for the option "add copy source where empty"? will that "save" the right encoding? and then translating w/o MT. Would this be another option?

Unfortunately, that will also populate segments that need to be translated in the end. In other words, will give you no chance to later enable MT and pre-translate segments that need to be translated (as Google will find text already in the target segment).

However, that will certainly work if you plan to manually open the file and proceed to do it manually segment by segment (and, of course, you may also have MT enabled when doing so). But, this means you will find a pre-populated target segment with text in the source language, you will clear this content (by hitting Ctl+D) and decide on how you want to replace it by the target language (either by using what Google/MT suggests and pre-editing it, or simply translating that from scratch -100% human effort!-).