CafeTran Espresso 2017 Harbinger - Release Candidate 4

Hello and Happy New Year!

 

CafeTran Espresso 2017 Harbinger aka Release Candidate 4 is available for download. The update can be performed via the Drag & Drop procedure as follows:


1. Run CafeTran.

2. Download 20170103_update.zip file from here (do not unzip or rename the file after downloading) and place it on your desktop.


Note: On Mac OSX, the Safari web browser may unzip the file automatically after downloading. To prevent this, just use the right-click menu and choose the "Download Linked File" command.

 

3. Drag and drop the downloaded file into the Project Dashboard and restart the program.


Alternatively, you can install the update via the Update button in the Help > About panel instead of dragging and dropping.


What's new:

  • The alternative connection to Yandex Translate MT service - see the option in Edit > Preferences > MT services tab. Required the API key from Yandex.com after the registration.
  • Experimental filter for xliff 2.0 files.
  • Automatic switch to preliminary memory matching with large translation memories.
  • Holding CTRL key in New Term/Fragment panels does not erase editor's contents.
  • Fixed the selection of words with German "ß" when changing the case.
  • Fixed a language direction issue when importing a glossary into the TM.

Igor


2 people like this

Yandex demo. The results are quite different from Google Translate and Bing Translator. Which doesn't say much per se.


image



Registration is easier than with MS or Google.


1 person likes this

Registration for Yandex translate was really easy. It is another bad MT provider for my language pair English>Turkish but why should I complain. Bad MT, more work for me. I wonder which CAT tools support Yandex currently. It is not supported by DVX3 and that is strange. I thought DVX3 was widely used among Russian translators. AFAIK memoQ and Wordfast Pro 4 do not support it as well.


Selcuk

>It is another bad MT provider for my language pair English>Turkish but why should I complain. Bad MT, more work for me.


These results from MT systems cannot be used directly. In my experience they often give nice suggestions for elegant / alternative wordings. I take the snippets that I like and leave the rest unconsidered.

Thanks for several nice improvements / feature additions.

>> Automatic switch to preliminary memory matching with large translation memories.

In Edit > Preferences > Memory, a new button has been added: "Switch to preliminary matching threshold." The unit of the value here is the number of TUs, I guess.

When I open a sufficiently large TM, however, preliminary matching does not occur automatically. Any tips?

 

Or the value may be the size of the TM (MB)? (default = 25000)

 

The threshold value is the number of loaded translation units. The Preliminary matching should start automatically after you choose a TMX file via the Dashboard. I have uploaded a new build where it also starts after opening a large TMX via the Memory menu. Then, it should kick off as you move to the next segment. The build also adds the standard keyboard shortcuts for basic formatting actions as requested in another thread.   

Great! Thanks!

 

Hi Igor,
I just installed the latest build and it's completely messed up the drop-down list of projects on the start page.
The project I was just working on is nowhere to be seen. The list now appears to show filenames rather than project names, some of which have de_DE appended.
Please fix!!
Jeremy

 

Oh: and with everything in lower case.

 

In connection with the automatic start of preliminary matching, I'd like to suggest that the default threshold be raised to a much higher level (300,000 or more) so that users not aware of this new feature will not be embarrassed with the automatic (for them, unintended) start of preliminary matching.

 

It is also a good idea to provide a "stop preliminary matching" button somewhere near the progress bar. Now, the only way to stop it seems to be exiting the program.


1 person likes this

Hi Jeremy,


Actually, it displayed project's .xlf file name instead. But you are right as it should display the project name given by the user. It's been fixed now. Thanks for noticing it!  

>I'd like to suggest that the default threshold be raised to a much higher level (300,000 or more) so that users not aware of this new feature will not be embarrassed with the automatic (for them, unintended) start of preliminary matching.


How about making depend on the total amount of RAM? 16 GB or 32 GB etc.

 I'd like to suggest that the default threshold be raised to a much higher level (300,000 or more)


The whole point of the automatic switching to the preliminary matching with large TMs is to make the search results available instantly. The TM size for the switch should be based on the average wait time for the result when the TM is set in the "Automatic" workflow. If it gets longer than 5 seconds, CT should make the switch. It is hardware dependent but I assume it may reach the 5-second limit somewhere in 25000 - 50000 segments range. The suggested 300000 segments size is way too large for the Automatic workflow.

Login to post a comment