CafeTran

Hello all, 

I am trying to check the wonders of pre-translation. 

I see in the wiki that the pre-translation mode is recommended for TMs under 30 000 segments. Do you still recommend using this feature even for shorter TMs?

Is there any workflow I should implement when working in retranslation (besides adding new segments to the Project TM for automatic recognition?)

Please share your views. Very much appreciated. 

Thanks!

I don't mind trying to explain Pretranslation, but I think it's a waste of time. Don't use Pretranslation. Total Recall made it redundant, and is way faster and smarter. Unless I miss something.

H.

Hans, or CT users in general: since there is the option to create a .tmx file for terms, or Termbase, then my question is: how this TM would behave in comparison to a glossary, where I am storing basically terms? Is there any hidden perk I should be aware of when using termbases? 

Thank you again! 

Thank you sooooooooooooooooo much Hans for your help. It is highly appreciated!!!!:-))))) 

Igor took over, see the KB.

H.

TMX: More control, better fine-tuning.

H.

Hi EC,

The CafeTran Experts seem to be too busy to answer your question, so I'll give it a try. A try, because it isn't the easiest of subjects. Intervention of Igor - and a KB entry - are appreciated.

In the "normal mode," CafeTran matches your source text with your resources when you go from segment to segment. That's fine, keep it that way. Unless:

You'll have to wait for CT to finish searching before you can start entering your translation. This usually happens when your resource(s) are large. It depends on your hardware how big they'll have to be to cause that delay. Your computer/laptop gets hot when working in the normal mode (no further comments, I don't use a laptop)

You should then use pretranslate (or "Total Recall", see below). There are two varieties:

On-the-go. Select Pretranslate from this segment in the menu under Translation. Give it a head-start for a minute or two, and start/continue working Wait till the pretranslation process has finished, save the result as a TMX file

In both cases, the results will be in CT’s cache (correct me if I’m wrong, Igor), and may/will interfere with your current job, e.g. if you used a term in segment n that’s different from the results of the pretranslation, you’ll hang, because CT will pick the term from the pretranslation. So I’d advise against On-the-go under all circumstances. The second option is much better. Save the pretranslation, and restart CT.

However, you can now import large resources in the database, and let CT extract (whole) segments and words from it. The resulting TM is much smaller than the imported one, there’s no problem with the cache, so there’s a good chance your worries are over. See the KB on his site under Solutions.

H

PS: Entering text in this site is lousy.

EC: besides adding new segments to the Project TM for automatic recognition?

If I'm right about the cache, that won't help you a bit. Not for Auto-Assemble. However, checking the Term Consistency box for the project TM will at least point to any mistakes during QA. I wrote about my approach for TM settings here. As you may expect, it's against the prevailing opinions here.

H.

Ipse: Don't use Pretranslation. Total Recall made it redundant, and is way faster and smarter. Unless I miss something.

I missed something. The TM, resulting from TR, can be too large to be processed the normal way.

H.

Thank you, Wooden for your reply. 

I am reading your article (very interesting and didactic, thank you as well for that) and you say that if my TM  for segments is getting really big that I should try the manual workflow. What does it imply? I read in the Wiki >>cManual: for TMs of only chosen segments. ????

We work with a general TM that we grow with results and terms coming from other projects in order to use them in future projects. Terminology may get to a ceiling some day, but segments vary ( technical item description -what we translate- vary per client). Which is the workflow recommended then? In you article you say optional workflow integration: Pre-translation, but I see Keep all duplicates, low priority, manual...and I get lost here>>  keep all duplicates>> does it imply all the versions for a translation are kept? Manual>> do I have to search manually for coincidences??

So activating the options translation memory and Termbase at the same time causes conflict??

Thank you again for your help.

I forgot to ask> I assume you recommend to exclude the Big Mama TM from consistency check as we are keeping all duplicates, so same source segments, different translations...Am I right? 

EC: ...you say that if my TM for segments is getting really big that I should try the manual workflow. What does it imply?

Manual workflow: Perform the concordance search yourself. In other words, enter a word or a phrase in the search box, and hit Return/Enter

Automatic workflow: CT performs the concordance search if you go to the next segment. This can take long if the TM(s) get too big. In that case you should choose between Manual workflow or Total Recall. I think you can safely say the option to "pretranslate" is outdated because of Total Recall (Igor?)

I read in the Wiki

The Wiki is - like my outdated blog - a personal view. One I often disagree with.

...and I get lost here>> keep all duplicates>> does it imply all the versions for a translation are kept?

Yes. I think that's OK for the Big Mama (general TM). Not for your ProjectTM.

So activating the options translation memory and Termbase at the same time causes conflict?

I don't fully understand this. But if you use Auto-Assembly (AA), they can conflict, which is why you'll have to set priorities.

I assume you recommend to exclude the Big Mama TM from consistency check as we are keeping all duplicates, so same source segments, different translations...Am I right? 

Right. For consistency check, I think you should only enable it in the ProjectTM, and any client TMs or termbases.

H. 

Thanks again for your reply. Going back to your answers: 

I think I am missing one of the two options you suggest for pre-translation. If you have the chance please to clarify, that would be great. Thanks in advance. 

You should then use pretranslate (or "Total Recall", see below). There are two varieties:

On-the-go. Select Pretranslate from this segment in the menu under Translation. Give it a head-start for a minute or two, and start/continue working Wait till the pretranslation process has finished, save the result as a TMX file.  (which is the second?)

In both cases, the results will be in CT’s cache

Also, in case anybody can help: have you encountered an error message when performing a pre-translation? I get it when I pre-translate against two memories...