CafeTran Espresso 2019 Forerunner

It is time to start rolling out the Forerunner (aka Preview) version of CafeTran Espresso 2019. This update and the subsequent builds this year will make up the CTE 2019 version to be released in December/January.

The update can be performed via Drag and Drop as follows:

1. Run CafeTran.

2. Download file from here and place it on your desktop. Do not unzip or rename the file after downloading.

Note:   On Mac OSX system, Safari web browser may unzip the file automatically after downloading it. Before you download the update file, uncheck the  following Safari option:

Safari > Preferences > General > Open "safe" files after downloading.

3. Drag and Drop the downloaded file anywhere in CafeTran's initial screen - the Dashboard.

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

Important! Please complete all your translation projects in your current CafeTran version before updating.

What's new in this update:

  • Autopilot. This may be a very handy feature during the Review or QA phase of translation. You can also use it when translating short, repetitive/similar phrases or numbers. It allows hands-free navigation with the set delay between the segments. The default delay of 2000 milliseconds (2 seconds) can be increased in Edit > Preferences > General tab. Activation/deactivation of the Autopilot is done via the Action menu. After that, you will notice the Autopilot button in the target segment editor. Click this button to turn on/off the Autopilot. Going to the next segment triggers off its functioning. To suspend it for a while (e.g for the correction of the current segment), press the Esc key or click the mouse on the target segment editor.
  • Automatic transfer of numbers being the full segments which end with the white space character.
  • Recognition of language variants in SDLTB termbases.
  • Support for subtitles (.itt and .srt file formats).
  • Added video resource preview for translation of subtitles.



3 people like this

You are right, it does. When an in-context review of the unknown spellings is needed, I'll take that route instead. Thanks!

> quickly filter target segments

Doesn't QA > Check spelling in targets segments do the job?

About Task > List words with unknown spelling, I like the easy way to add these to the user dictionary, but miss a way to easily copy / list in text mode and/or quickly filter target segments.

Reasoning for the last one (filtering target segments, which, to be fair, can be achieved by just providing  an easy copy text feature): Sometimes, I wish to review the unknown spelling in context before deciding what to do.

Maybe an Edit user's spelling dictionary button (along with one  to open unknown words list in text mode?) could make sense here.


About the 3-letter upper case rule, I don't have an opinion, for now at least.

Agree entirely with tre, etc. – having this non-configurable and on by default is a disaster waiting to happen.

"you sacrifice a basic feature for most users against a comfort feature for very few users" is exactly right. It's not just another odd design choice, it's utterly stupid.

> neural Machine Translation algorithms (e.g. Google Translate and DeepL), which are heuristic by their nature

Well, I do human translation. And maybe this is why a text pretranslated with fuzzies or MT alone does not make a good job.

> there is far less work spotting any 3-letter upper-case spelling mistakes, not to mention low probability of making them, than adding such words to user's spell checker dictionary

Indeed this is a good example for heuristics, but it is far from being reality for the daily routine.

> but I don't remember seeing a spelling error in 3-letter upper case words in real texts

Maybe this was before texts were translated with the latest CafeTran feature. As written above, you sacrifice a basic feature for most users against a comfort feature for very few users. What a pity.

I couldn't disagree more. The "precision" in fuzzy matches, auto-assembling or machine translation - the core of CAT tools functionality? It is just the opposite! It is approximation, and the very lack of precision is commonplace even in the latest neural Machine Translation algorithms (e.g. Google Translate and DeepL), which are heuristic by their nature. They are getting better but are far from being precise - and let it be so. :) 

Just a few words concerning heuristics.

If we take the correct article from Wikipedia here, you will se that the sky is not really blue. And perhaps we stumble upon the paragraph named Pitfalls. And if we follow the questions unter Trade-off, this very new feature would not have been introduced. Sorry, CafeTran is not a virus scanner. And it should not work with rules of thumb. A CAT tool needs precision and flexibility, not heuristics. This means of course also complexity (just multiply the important options in CafeTran to find out the number of possible configurations). When reading the article above, I cannot understand the latest release.

> ...I'm used to underlined non-translatables

Red-underlining is not much of an issue but false QA with dozens of segments to check may be so. In my opinion, there is far less work spotting any 3-letter upper-case spelling mistakes, not to mention low probability of making them, than adding such words to user's spell checker dictionary. We do make spelling mistakes, but I don't remember seeing a spelling error in 3-letter upper case words in real texts. An option might be the solution though.


Igor, could you please make it an option.

I don't think the benefits of such implementation are greater than disadvantages, either.

I'm used to underlined non-translatables but not to actual typos' being ignored.

Perhaps I'm wrong but I guess three-letter acronyms even with known spelling won't be included in the auto-completion anyway, because they are too short.



Not to forget the possibility to catch any wrong-spelled of these acronyms (some source texts – manuals – are less carefully edited than you might imagine, hey, they are only technicians).

I have to agree with Torsten here. I think that this tree-letter acronym workflow is a rather rare user case. With all due respect, of course.

In my own daily work, I often translate software interfaces that contain three-letter abbreviations. It would be unpleasant if these wouldn't be checked.

Personally, I don't see any harm in adding any abbreviations (or acronyms) to the custom dictionary. It's quite simple to edit this or to use different custom dictionaries for different types of text (one for software strings, another for novels, etc.).

Catching any untranslated words that are in the custom dictionary (e.g. UNO), can be done via:


Maybe my example was far-fetched and "stupid thing" was unfriendly, however there are the cases:

  • whole texts or segments in upper-case -> spellcheck does not work properly (typos such as "ADN" or "FRO" do not seem to have any importance)
  • texts where the acronyms that you do not want to add all acronyms to the userlist -> comfort
You did your choice, that's a pity.

1 person likes this

> OMG...

You make it way more dramatic than it really is. The heuristics of this algorithm is very practical, being useful in much more cases (e.g OMG) than doing any harm (e.g. your far-fetched examples to prove your point are in practice much rarer than the short upper case 3-letter acronyms). With this feature, I responded to the user's request  for such 3-letter upper case word handling by the spell-checker, and calling it "stupid thing" is a bit of disregard for other users preferences. And in this particular case,  I would say "most users' preferences", which is what the heuristics is all about.  By the way, a lot of algorithms are heuristics, the method often applied in the programming. From Wikipedia:

"A heuristic technique (/hjʊəˈrɪstɪk/; Ancient Greek: εὑρίσκω, "find" or "discover"), often called simply a heuristic,  is any approach to problem solving, learning, or discovery that employs  a practical method, not guaranteed to be optimal, perfect, logical, or  rational, but instead sufficient for reaching an immediate goal. Where  finding an optimal solution is impossible or impractical, heuristic  methods can be used to speed up the process of finding a satisfactory  solution. Heuristics can be mental shortcuts that ease the cognitive  load of making a decision. Examples that employ heuristics include using  a rule of thumb, an educated guess, an intuitive judgment, a guesstimate, stereotyping, profiling, or common sense."

Making options for each single user is definitively not practical. Perhaps let's wait what other users say as this feature can be made optional.


I'd agree to that, but my first thought was "well, he won't do such a stupid thing" ;-)

Hi Torsten, I don't install this build until the three-letter ignoration has been solved. Can test it on the MacBook Pro at home.

Login to post a comment