A computer newsletter for translation professionals

Someone asked me last week: Do you still always come up with enough material for your newsletter? The answer is I don't have to: the material comes to me. If there was enough to cover in translation technology 219 editions ago, there is certainly more than enough today. Case in point: I had planned to write an article about the Canadian tool Logiterm and the British tool XTM, but neither article made it - the bulk of content for this edition forced them into a holding pattern for the next edition.

Is translation a waste of paper?, an article last week on the ForeignExchange Translation blog, made me smile when I read it. I don't do any medical translation -- though I have translated many instruction sheets and similar items that I know have never been read -- but when I looked back yesterday and read the comments I felt proud to be a translator. Man, we are a great group of folks. Mind you, I don't agree with every comment, but I LOVE the truthfulness that speaks through the passion infusing most of the comments.

Oh, and here's another (maybe not too comforting) thought: that same content that you and I translate, some of which may never be read by its intended audience? It will be read by the large web crawlers that collect multilingual materials to train their machine translation engines. So it wasn't all for naught. Or was it?

Microsoft has just released its latest edition of Office. You're excused if you missed that memo. Unlike previous releases, this all happened with remarkably little fanfare. To make a hash of my metaphors, Microsoft might feel it has bigger fish to fry, and a new version of Office might be assumed to be easy sledding. Maybe it's also that even Microsoft got confused about its terrible naming strategy of the new product and didn't really know how to promote it.

Here is the difference between Office 2013 and Office 365 in a nutshell: When you buy into the Software-as-a-Service plan "Office 365" (at a $100-a-year price tag), you can download five copies of Office 2013 for five different computers with online storage. If you buy Office 2013, you receive the typical license but you get only one install and will inevitably end up paying a lot more because you have to update it at some point as well.

Now that Adobe has successfully employed the SaaS model for some time, Microsoft has also realized that it's nice to have an ongoing revenue stream, particularly because users tended to upgrade only every two or three versions anyway. So it's really "encouraging" you to go the 365 route. (Still confused? Read this article and you'll get it.)

More important than all of this, of course, is to ask whether an upgrade makes sense to start with. In short, the answer is no, though there are some interesting features (and I'm now talking from the translator's perspective). I can't think of many reasons why you shouldn't go with the new version if you have to purchase one with a new computer anyway (or you're entitled to a free upgrade because you recently purchased a computer with Office 2010 -- which was the case for me).

Well, there actually is one reason why some should abstain. While files of the new version of Office are compatible with most translation environment tools, they are presently not compatible with Trados Studio. I'm sure that this will be remedied fairly soon, but in the meantime you'll have to down-save your files if you use Office 2013 and Trados Studio is your TEnT of choice.

The two things that you'll notice immediately when you open any of the Office programs is the different look and feel: everything seems flatter and more web-like and most certainly more touch-friendly, much in line with Windows 8. The other thing that you'll notice when you start to type is that the cursor moves differently -- it sort of hovers along rather than jumping from character to character. It looks kinda cute, but it gets annoying when you type really fast because it can't keep up with you. Unfortunately, there is no setting within Word that allows you to change that behavior, though of course there are some "hacks" to get rid of it.

Let's quickly cover some of the features that are relevant for translators: The biggest and most important is that grammar and spellcheckers are now available for all languages (that are supported in the first place) for free. While it is still possible to purchase language packs (these come with translated help and user interface -- but you can also download the translated versions for free with the Office 365 offer), grammar and spellcheckers are yours to download and install at your whimsy. That's great (and rather embarrassing for Microsoft that this is available only now), but it's made even better by the fact that you are prompted automatically whether you want to install a new language once it's detected in your text (if you find the reminders annoying, you can disable them under File> Options> Language).

Other things that are helpful for the translation-editing-proofreading process are a "Reply to Comments" feature that allows you to respond within a comment with the automatic addition of your name and the time of your response.

I also liked a new "Simple Markup" feature. Here you can show the location of markups without showing all the markups in the text detail, which, as we all know, all too often makes a document virtually unreadable. A redesigned "Read Mode," which now emulates ebook reading, could have been nice for editing a Word document but I found this really disappointing and much harder to use than the normal edit mode. What is nice, on the other hand, is a new feature that asks you whether you want to jump back to the place you had last viewed when opening a longish document -- very helpful for editing purposes.

There is also a much heralded automatic PDF conversion feature. It works like most of you PDF-jaded translators would expect: nicely with simple PDFs, not so nicely for complex files. Still, it might give you better results than what your TEnT offers, so it might be worth a try. 

Why do people work with memoQ

"I'm sitting here at my desk doing some rather tedious translation work. Well, I wouldn't say that it is a delightful pastime on a Saturday afternoon - but memoQ makes it really bearable. It is such a pleasure to work with this translation environment."
Carola Striebel -- freelance translator

To learn more, attend free educational webinars and download the full-featured trial version from http://kilgray.com.

Shortly before Language Weaver, the statistical machine translation company, was acquired by SDL in 2010, it had started to offer a cloud-based solution for machine translation. SDL has branded this in the last couple of years as BeGlobal, marketed it as a separate product to large companies, and integrated the baseline product (i.e., the product that is not specifically trained for a certain purpose) into Trados Studio.

This coming month, SDL is going to release a trainable version of BeGlobal for LSPs. While this might not sound too breathtaking, I think it's an important step, and that's why I sat down with some folks from SDL to have a closer look at the product.

But before we go on and you work yourself into an increasing sense of rage about MT being covered once again, make sure you read this discussion that I recently had with Charlotte Brasler for the ATA Chronicle. It's certainly nothing definitive about machine translation, but it does represent an attempt to shine light on the various ways one can look at the topic.

Done?

Good.

If an LSP wants to use customized machine translation today, it has a number of options. There are the traditional rules-based engines, such as SYSTRAN and PROMT, which can be trained both with data (primarily terminology data) and by fine-tuning and tweaking the linguistic rules of the target language. Then there are many MT programs that are based on the open-source Moses engine. These include some commercial offerings like Asia Online as well as many "homemade" options, and they are primarily trained with translation memory data because these are statistical machine translation engines. And lately, Microsoft has been in the running with the Microsoft Translator Hub, which allows you to train a statistical MT engine with your own data (which in return you'll need to donate to Microsoft).

SDL's offering now attempts to lower the entry threshold for LSPs to add machine translation to their offering without having to invest in the expertise that some of the other offerings require (in particular the Moses solutions and the training-intensive rules-based engines). BeGlobal is available in a fairly large number of language combinations, and any of these language combinations can be further trained with TMX (translation memory exchange) files with a minimum of 200,000 words and 15,000 translation units. These files need to be pre-processed to eliminate non-linguistic entities such as tags. Once that is done they can be imported into the engine where they are processed. After 24 to 36 hours you'll get a notification on how the trained engine compares to the baseline engine (it uses an internal evaluation process that applies the industry-standard BLEU score to the baseline vs. the trained data).

If the score is no better than the baseline, you can retrain the engine by scrubbing your data some more or adding some other or more data -- something that you can also continuously select to do as your translated data grows.

As expected, the user interface within Trados Studio makes it easy to connect to the trained cloud-based engine sitting on SDL's servers. The same logic that is applied to any machine translation within Trados Studio (or any other TEnT for that matter) is applied to this automatic translation: If a translation memory match of a certain customizable quality level is found, use that; otherwise, use the machine translation. Within Studio's interface it is marked as AT (automatic translation) so the translator knows that she will have to apply a different kind of editing to these segments than the TM matches.

The cost of the system is $1,500 per year for three million translated words and something between $18,000 and $19,000 for 100 million words. The training is free.

These are the mechanics of the system along with some additional thoughts:

• Confidentiality: I don't have a problem with the fact that the system is cloud-based per se, but there is an option to "donate" your data for the good of mankind SDL's machine translation engine. I find it preposterous to even offer that. It's one thing that Microsoft or Google are doing that -- but their business is not in translation. SDL's is. I also talked a bit with the SDL folks about the progress they have made on the baseline engines for the various language combinations. While the majority of data comes from public web resources, they have also used data they have translated for their clients for which it was not "expressly prohibited that it could be used for such purposes." Kind of a touchy area, if you ask me -- and, by the way, potentially one that you could find yourself in if you use clients' data to train the engine.
• Language combinations: Don't assume that the level of quality you might achieve in one language combination will automatically be applicable to another language combination. If you really do want to use machine translation for certain kinds of text, you will have to run tests for each of your language combinations with different engines, and it is almost a given that you will find that different engines perform better with some languages (and some languages may not be suited at all).
• Product captivity: If you are certain that you are staying within SDL's suite of products for a long time to come, this might be a good choice. If you want to leave your options open, you will want to choose an MT technology that can be used within more than just one tool (to my knowledge, Lingotek is the only other tool that has a built-in interface for Language Weaver/BeGlobal aside from SDL products).
• Integration into Studio's resources: As I mentioned above, it is very easy to integrate the engine into the workflow, but I found it disappointing that other resources, and I'm thinking primarily of termbase data, are not automatically considered. While they are displayed during the translation process, they are not automatically inserted if there is a discrepancy between the term used in the machine-translated segment and the termbase's term. I know that this is a weakness of most other tools as well and needs a considerable amount of development to make possible, but until it happens machine translation is essentially just a parallel process rather than a deeply embedded one that truly aids the translator -- but again this is a problem that is not exclusive to SDL's products.

Multilingual computing has always faced the challenge of different input methods and keyboard layouts for different languages (and even competing methods and layouts within one language -- think QWERTY vs. the Dvorak layout for English). This will continue to be a challenge with new solutions cropping up here and there. For Windows users, Microsoft has been remarkably good at offering built-in language keyboards for more than 120 languages and the ability to extend or compose your own language keyboard with the Microsoft Keyboard Layout Creator.

With much computing now happening on mobile platforms, there is an increasing need for other input methods that are not bound to a single operating system. Google has recently multiplied its offering of language keyboards by adding no fewer than three additional Chinese input methods, among others. Some of these can be downloaded and integrated into various operating systems (including Windows); others are available only within different Google tools, in particular Google Translate, which adds an input icon in the source language field that provides access to various keyboards or input methods.

Some of you use the US-International keyboard on Windows computers when dealing with multiple Western languages, which gives you access to a good number of special characters that are not part of the English alphabet.

Moshe Devere just pointed me to another keyboard that offers most, if not all, of the additional US-International keyboard characters plus characters with breve (˘), inverted breve ( ̑ ), dot below and above (·), and the macron (¯) as diacritical marks. While interviewing folks for the research on Found in Translation, we encountered some of the struggles of languages with those diacritics, in particular when it comes to languages like Māori or Yorùbá (see pages 219 and 215).

You can download versions of the Alt-Latin keyboard right here. It will easily integrate itself into your system (both Windows and Macintosh) and you can install it with the normal keyboard installation routine (this explanatory article is helpful though slightly outdated, and you can find updated information in my Tool Box ebook) as one of the available keyboards. To enter characters with diacritics such as å, ā, ą, ạ, ȧ, or ă, you will need to simultaneously press the right Alt key and the key that corresponds to the grey diacritics marked below, release both keys, and then press the key for the primary character.

Surveys by Common Sense Advisory and ProZ.com have all identified SDL Trados technology as the most widely used translation memory software in the industry.

We're also proud to announce that another 4300 translators of the EU will be joining us.

Why don't you join us too?

Sign-up to our Translators Group Buy and save up to 35% off SDL Trados Studio 2011 Freelance. You will also get FREE access to SDL QuickStart where experts can help you make the most of SDL Trados Studio.

BUY NOW - Hurry, there's only limited units available.

When I was home in Germany last week, I was happy to discover that my mom had finally bought her first computer -- grandkids, here comes Oma on Skype! Naturally and rather unfortunately, it was a Windows 8 computer, which felt rather impenetrable to her with her faint memories of the Windows NT and XP computers back at work many years ago. I tried to explain, putting together a list of keyboard shortcuts and trying to clear the app jungle, but it still didn't really click for her. So I ended up pulling the last arrow out of my quiver: I installed Classic Shell -- and that did it.

Classic Shell is an incredibly popular freeware program with a total of four million downloads that was originally developed for Windows Vista and Windows 7 users who were unhappy with those new interfaces, especially when it came to the altered Start menu. The program restored the old and familiar interface along with all its functionality. I'm not sure why this was important -- I've always preferred the Vista/7 interface over its predecessors -- but that's how people are: the familiar feels safer.

Now, with Windows 8, I'm the Luddite: I can understand why users don't want to start out in the Disney-like app interface. I want to have the computer boot right to the desktop, a desktop with a Start button that gives me access to all the programs and utilities I had there before.

This is exactly what Classic Shell does.

It's a relatively small download, it's free, and the installation should work seamlessly. There are admittedly a LOT of options you can choose from after you install it -- and some of them I find hard to understand -- but you don't even need to bother about those. Just leave everything in the default settings, and you will (almost) think you're back in an earlier version of Windows. (If you find later that there's a certain behavior you don't like, you can still fine-tune the settings, which is much easier when you actually know what you're looking for.) If for some reason you end up in app mode again (this might happen while you're adjusting some settings), just press WinKey+T or WinKey+D and you're right back in Desktop Mode.

And breathe a sigh of relief. Life can go on as normal . . .

Giveaway! 

Quick now, you language geeks: What is "Annotators Instill" an anagram for? The third and thirteenth correct answers will win a signed copy of Found in Translation, the book that Mental Floss calls "a spirited tour of the world of translation" and that your colleagues have been raving about. 

Just reply with your anagram answer to this newsletter.  

One of the unfortunate side effects of MS Word-based translation environment tools has always been the dependence on changes in MS Word, resulting in hard games of catch-up for some tools. (As a side note: Wordfast Classic is already compatible with Word 2013.) Reader Dino Azzano brought up this perfectly legitimate question: Does a similar issue exist for browser-based TEnTs, particularly because new versions of browsers (especially Chrome and Firefox) are being released at an extremely rapid pace?

I asked the makers of MemSource, Wordbee, and XTM how they handle the many changes in browsers and the many kinds of browsers.

Here is what they had to say:

It sounds like Dino's question was right on the money and this indeed provides a healthy challenge, albeit one that most of the developers master.

As a subscriber to the Premium version of this newsletter you have access to an archive of Premium newsletters going back to May 2008.

You can access the archive right here. This month the user name is toolbox and the password is oregonrain.

New user names and passwords will be announced in future newsletters.

If you would like to promote this newsletter by placing a link on your website, I will in turn mention your website in a future edition of the Tool Box newsletter. Just paste the code you find here into the HTML code of your webpage, and the little icon that is displayed on that page with a link to my website will be displayed.

Here is a website that mentioned the Tool Box last month:

french-to-english.net/machine-translation-blog  

If you are subscribed to this newsletter with more than one email address, it would be great if you could unsubscribe redundant addresses through the links Constant Contact offers below. 

© 2013 International Writers' Group