A computer journal for translation professionals

Earlier this week, I wistfully tweeted ascreenshot of the Google news digest for the keyword "translation" a day or so after the Microsoftdemo of real-time interpreting through Skype was gobbled up by more than 300 mainstream media outlets.

The only other time translation or interpretation received that kind of media attention in the last couple of years was probably when the scandal around the "fake" sign interpreter at the Mandela memorial broke.

So what do we need to do to embrace these opportunities and point to the efforts and needs of "real" (I hesitate to write "human") translators and interpreters?

I'm not an interpreter, so I can't speak for them (if I were one, by the way, I think I'd be rather grateful for Laura Burian, Miguel Garcia, and Barry Olsen's video produced at TEDx Monterey). Actually, I'm not really sure that I can speak for translators, either -- but at least I have a better idea of what's happening on those fronts. And "fronts" may indeed be the right term because some of the rhetoric that we've been using against and about each other in social media and elsewhere resembles the language of war and destruction rather than understanding and appreciation.

I'm glad that there were a number of blog posts in the last few days (especially this one by Terena Bell) as well as the publication of the The Bright Side of Freelance Translation calling for a positive spirit in our industry. After talking with one of its co-editors, Andrew Morris, I understand that some have criticized The Bright Side for being too "happy-clappy" and too "Disney" -- and perhaps the website and some of the callouts in the book lend that impression (though I wouldn't have thought of those particular terms...). But if you actually read through the book, you'll gain a different impression -- you'll find stories of translators who love what they do, despite the occasional hiccup, and are not ashamed to talk about it.

It's easy to become too inward-focused and possibly too negative when we're working on our own. It's also easy to use technology that allows us to publish instantaneously and in an unreflected manner (I for one have been guilty of this). But we ought to be careful not to hurt each other, not to hurt the person who is using or offering a different business model or a different technology. I do think we need to be wise in what we support, and we need to be vocal about the things that are detrimental -- but we need to remember two things: my needs are not necessarily identical to those of my fellow translators, and we're ultimately more effective in our disagreements if we do it in a spirit of love.

And that is really the only way we can be successful in promoting ourselves and our profession because it will ultimately give us a position of strength.

Let me try to put this into practice: Anyone who has seen Microsoft's Skype demo can make up his or her own mind about it. I don't think it's necessary to add anything to the many voices of those who have already chimed in. There's just one thing that really frustrated me: Both Microsoft's new CEO Satya Nadella and Gurdeep Singh Pall, the head of Skype at MS, praised the new technology (or "the new combination of old technologies")  as "magic." And that is something that I think is really detrimental to virtually everyone. Leaving aside the fact that the results they demonstrated were far from "magical," everyone knows that technology is not magic. It's these kinds of misnomers that make a discussion on things like machine translation virtually impossible. After all, how can you even start to talk about magic?

(By the way, were you as confused as I was about their choice to use German for the demo since German is a notoriously difficult language for machine translation? As it turns out, the language recognition part of the process is more reliable in German than in languages like Spanish which is why they chose German.)

A couple of newsletters ago, I talked about the Multidimensional Quality Metrics initiative, or MQM, a "framework for describing and defining quality metrics used to assess the quality of translated texts and to identify specific issues in those texts." This initiative is part of the EU-sponsored QTLaunchPad project. (You can read the article in edition 234 in the archives.)

I also mentioned back then that there is a parallel framework being developed, the so-called Dynamic Quality Framework (DQF) under the umbrella of TAUS, the Translation and Automation User Society. Fellow translator and fellow Berliner (ex-Berliner in my case) Victor Dewsberry took it on himself to write a lengthy blog post on DQF awhile back. I would encourage you to read it because in my view it is written with exactly the spirit that I mention in the introduction to the newsletter. It's highly critical of DQF and its motives, but it communicates in a witty and yet positive spirit that -- if you read through the comments -- actually leads to some changes in the DQF program.

DQF started in 2011 as an effort by various large organizations (and some smaller ones). Just as Victor notes in his article, however, there are no translator groups represented. This really should not surprise us: TAUS plays an important role in the word of translation, but its main interest does not lie with the individual translator. Instead, it is mostly concerned with increasing automation, especially in regard to machine translation. I assume that even the folks at TAUS would agree with this assessment.

The DQF platform at this point consists of three parts:

• A knowledgebase that contains reports, models, and templates of other organizations using the framework
• A tool for "content profiling" that allows the user to describe the kinds of projects that are being translated with the goal of receiving a recommendation for how evaluation criteria should be used
• Three tools to assess and compare translated content

These three tools are responsible for the following tasks:

• Ranking the quality of different machine translation engines
• Testing the productivity of translators vs. post-editors
• Evaluating translations in regard to fluency, adequacy, and according to an error typology

In our context, the third tool is probably the most interesting, but let me first say a few things about the other two tools.

To the machine translation user, it makes sense to be able to use a tool to compare the quality of the output of various machine translation engines. For this purpose the user can upload a text of 250-1,000 sentences (fewer are possible but not recommended), have those translated by a number of machine translation engines of his choice (or he can connect his own customized engine and compare it with others), and then have (human) evaluators score the results on a per-segment basis -- either by selecting the best of the lot or by ranking them. The results are then presented in a report that should give a relatively good overview of which engine performs the best. According to Attila Görög from TAUS, ideally you would have four to five different evaluators work through the data to give you the best results.

It seems like a pretty straightforward system, but I'm not quite sure that it always presents the most realistic results. While it's easy to compare non-customized engines with this method, I wonder how reasonable (and realistic) the comparison of customized engines is, considering the enormous amount of effort you have to invest into building each of those for your particular purpose. Would you go through the customization of several engines and then make a decision on which one to use with this tool? I'm not sure that makes too much sense. Still, it's an interesting idea that takes the evaluation of machine translation away from the vague BLEU and other kinds of scores.

For the second tool, which is supposed to test the productivity of translators vs. machine translation post-editors, you also upload a text, half of which is machine-translated and the other half is presented as non-translated. It then measures how much time is spent on the translation of the respective segments vs. the post-editing. I found this tool really quite useless because it bears no resemblance to how a translator works -- without any terminology, TM data, or other productivity tools -- and it therefore gives a very inaccurate assessment of true productivity. I hope that not too many users of the framework will use this tool to make a decision on whether to use MT -- and I also hope that TAUS will add a disclaimer explaining the relative uselessness of this process. Plus, the same evaluation principle -- measuring time to see how fast one translates vs. post-edits -- is already available in a number of other tools, including certain versions of OmegaT, Trados Studio, and memoQ.

The last tool shares some similarities with the processes of MQM, the above-mentioned error typology system for human and machine translation. DQF offers a very similar error typology (the folks from both MQM and DQF told me that the different sets of typologies will eventually be harmonized since the main difference consists mostly of using different names for some of the items).

In addition, TAUS's DQF also offers an evaluation of "fluency" and "adequacy" -- both typologies that really are mostly, if not only, geared toward machine translation.

To use the last evaluation tool, you again upload a text, select the typologies you want to evaluate, and your evaluators then will be presented with the different sets of queries for each of the respective areas. The main error typology categories the evaluator can choose from include accuracy, language, terminology, style, and country standard. At this point, however, it's unfortunately not possible to add annotations to the selections the evaluator makes -- though, according to Attila, this will be implemented soon. My feeling is that this would be more heavily used within translation environment tools themselves. Both Wordbee and memoQ already offer the fully implemented typologies (including those for fluency and adequacy) with the ability to add comments, which is helpful -- and indeed of central importance.

I think translators would be well advised not to completely ignore efforts like this one by TAUS (or the other by QTLaunchPad). Though there are some weaknesses in the design of the actual tools I describe above, what better opportunity could we possibly have to demonstrate why translation by professional translators cannot be replaced with machine translation in the majority of cases? And that post-editing often does not save time over other more intelligent ways to use external data -- including machine translation data?

Next Sunday I'm going to be part of a workshop discussing the MQM model. I'm really looking forward to it, and I hope to suggest that finding better ways of measuring quality can only benefit professional translators. 

Simple. Stable. Swift. The new memoQ 2014 is released!

Project templates and workflow automation, image localization and importing terminology from other tools are just a few of the many features and concepts that memoQ 2014 brings.

Learn more about memoQ 2014 and give it a try by clicking right here.

Fluid is a tool by the Portuguese developer Joăo Albuquerque that allows you to analyze XLIFF (translation file exchange), TMX (translation memory exchange), or CSV files for terms or subsegments and replace those with their respective translations. You can do this in large batch operations or one by one, making sure that the context allows for the particular translation. The tool also enables processes such as consistency checks.

I tried it, but I was still a little puzzled about how this would fit into the workflow of a translator who is already using a translation environment tool (which you would do with formats like TMX and XLIFF). So I asked Joăo:

If a translator preprocesses a file with your tool, she would have a really hard time finding any perfect or fuzzy matches (or machine translation) because a part of the segment is already translated. And I'm not sure that I see an advantage in postprocessing.

To which Joăo answered the following:

I see the role of this as a "hermaphrodite" in the industry workflow...

Your statement is true! But I use both this and the most (compulsorily $700's) used CAT tools, and I can't imagine not using this! Maybe I should work harder on my presentation and focus on the most productive functions.

For example, I'm currently working on a product with a very particular terminology, and I know I wouldn't be able to recycle it with 100% precision with those other CATs! I find that they serve the big translation companies and MT factories too much, always giving an overly heavy workload to translators somewhere in the line production....

There you have it. I think his tool has some potential -- it might not be quite where it needs to be, but it could be worth taking a look. 

Success can be measured in many different ways. There's prestige, there's money, there's satisfaction -- you know the list. But the ultimate measure of success is when your adversaries adopt your ways. As the saying goes, imitation is the sincerest form of flattery. What better recognition can there be that you're doing something right?

A couple of weeks ago fame, Trend Micro, the security software company, reported this:

VOBFUS malware is known for its polymorphic abilities, which allow for easy generation of new variants. We recently came across one variant that replaces these abilities for one never seen in VOBFUS malware before-the ability to "speak" several languages.

WE MADE IT! Our message has penetrated to the bad guys. Now even they know that well-internationalized and localized software performs better. Can there be any better recommendation of our communication skills than that?

Introducing the SDL Customer Experience Desk 

The new SDL Customer Experience Desk has launched to help answer any questions you may have around joining the SDL community or about our leading translation software, SDL Trados Studio 2014.

"Whether you are completely new to translation software, or looking to switch from another tool, we are here to help. We're happy to provide the information you are looking for, so please get in touch." 

Steven Whale, Manager - Customer Experience Desk 

Chat with us live » 

"Individual translator" is how Kilgray (for the time being) designates what you and I know otherwise as a "freelance translator." They came to that conclusion after a survey they took among participants during the last -- and I'm sure very raucous -- memoQfest. I think, though, that I'll stick with "freelance" or "professional translators."

Be that as it may, the new version of memoQ has plenty to offer for all different groups of users, even though it might not be apparent at first glance because the interface looks deceptively similar.

When I talked to Kilgray's Gábor Ugray about this latest version, he stressed that unlike previous versions, this one carries more of Kilgray CEO István Lengyel's imprint than that of the development team -- and this might just show in some of the business-oriented developments we can see in it.

Gábor did not seem quite so clear about the overall theme of this particular release (they love themes at Kilgray, with the two previous ones being "quality" and "refactoring"), but he said that "simplification" would best express what this was about. When he said that my heart made a BIG jump! Really? Had they finally listened to me? Had they finally truly simplified things? Imagine my disappointment when I discovered that there really was no visible change in the somewhat overcrowded menus and options. Still, Gábor has a point. Sort of.

Here are the bigger changes in a super high-level overview:

• A stronger emphasis on automation, project management, and a tie-in with the web-based Language Terminal
• A stronger utilization of third-party terminology
• Very perceivable steps toward processing acceleration
• An emerging emphasis on multimedia

Let's start with the first of these points. Kilgray has been touting templates for project creation for several weeks now, and I wasn't particularly awed by the idea. It felt more like something they should have done some time ago. Still, the way they've implemented this is pretty clever. Now you can start a project on the basis of a previous project (that's the idea of a template), but Kilgray's idea of templates also involves the concept of project automation. If you select Tools> Resource Console> Project Templates and then select one of the available templates, you have the option to select Automated Actions. This in turn will allow you to define a whole chain of events that are dependent on each other. This means you can define that when a file is imported, an analysis is created, the files are pretranslated, and so on and so forth. This is more likely very handy for LSPs and for the senior project managers working there, but I can see that they will really like it.

For the "individual translator" it should be a welcome option that you can now right-click on any memoQ-supported file in Windows/File Explorer, select Send to and memoQ. memoQ will then analyze the language of the file (and will even find out if there is more than one language -- Gábor called that the "world's most hidden feature"), make that the source language, and also decide on the target language based on your keyboard layout and other indicators.

The integration of projects with the online Language Terminal also continues to become stronger and stronger. To the point, in fact, that LT is now becoming more like a translation management system where details about projects, vendors, and clients are stored and communication is enabled between all of these groups. I've reported before that Kilgray holds great hopes for Language Terminal -- at one point they called it one of "the best things that can happen to the language industry" with the vision of it becoming the industry-wide hub of translation management. This requires collaboration with other technology vendors, however, and so far only Plunet and XTRF are working on some kind of integration. Wanna guess whether SDL is doing that as well?

I was particularly excited about the third-party terminology integration into this new version -- and very disappointed when it didn't work in the beta version that I worked with, but I'm hoping those early bugs are now ironed out. What third-party terminology? Well, first of all, it is now possible to import Transit and Trados packages, including the terminology databases they might contain. It had been a real pet-peeve of mine that this had not been possible -- especially for Transit projects that often tend to be very terminology-heavy. Now it's apparently possible, though it didn't work when I tried it out...

The other terminology twist that I'm excited about is the integration with TaaS, or Terminology as a Service, the web-based terminology tools that I reported on a few issues ago (issue 229 in the archives). If you missed it, it's a tool that allows you to very intelligently extract normalized terminology from your project, pretranslate it with a number of web-based resources, and then correct that pretranslation and amend it. You can then either download it or -- in the case of memoQ -- leave it right there on the server and use it from there. This is great for sharing terminology with other translators you work with (or even translators you don't work with). What I found unfortunate in memoQ's implementation is that it uses memoQ's own term extraction module, which (depending on the language) is usually not as powerful as that of TaaS -- so there are plenty of ways for the folks at Kilgray to improve this feature.

Another new feature is the processing of embedded files. If you for instance have a PowerPoint file with an embedded Excel file, the text from both files is extracted and offered for translation. This goes just one level deep (it would not extract a Word file within the Excel file that is extracted within the PowerPoint file), but for all intents and purposes that should be enough. It's a feature that some other tools offer as well, and for those tool vendors who don't yet offer it -- it's high time to do that now.

memoQ also now extracts embedded graphic files (if you enable that under Import with Options). Of course, it can't offer you the extraction of text -- that would require an optical-character recognition engine -- but it allows you to transcribe the images (you'll have to highlight the images in the Translations view, select Details, and then Transcribe Image -- it took me forever to find that) and then transcribe them. It's a nice new feature but not earth-shattering. It might be more earth-shattering as the first step toward a greater attention to multimedia files. I think that you should be able to see a subtitling translation feature for videos with real-time preview in one of the next memoQ releases (like only Transit offers now) and that will be more interesting -- and more and more relevant now that videos and therefore video translation have become so very ubiquitous.

The speed of the processes in memoQ is also notably faster. This is one of those claims that you can almost always find on the what's-new list for a new version of any tool -- but this is different. It really is faster. Startup time is at least twice as fast, and the export of files now uses both processors in your computer (if you have dual processors), so that also goes ruckzuck (as we Germans like to say).

Oh, yes, and since we always like to see all the different articles in a Tool Box Journal tie in with each other, how about this: The TAUS DQF results can now be exported into a nice-looking report if you use that particular linguistic quality assessment within memoQ.

How's that for quality?

Across Language Server v6 

The most important new features at a glance:

• Across Dashboard: Relevant project information at a glance.
• crossWeb Review Mode: Easy involvement of international subsidiaries in the review process.
• Across Data Cube: Easy collection of indicators.
• Project Management Cockpit: Optimized work process for project managers.

Get an exclusive preview of the Across Language Server v6 even before its release: www.across.net/en/newversion  

I was going to write a longer article on this for this issue of the Tool Box Journal, but space (and time) are running out, so I will focus more on that in the next issue (also in the next issue: the new version of Across).

Maybe I can also use some of your input (if you have any) when I'm able to devote more time to this.

In a nutshell: Both the machine translation community and we have been focusing on post-editing as essentially the only way to use machine translation. And since post-editing really goes counter to what our translation process is like, it's no wonder that many of us feel so strongly opposed to working that way (of course, there are other reasons besides that as well).

I have recently worked in a couple of large projects where I tried a different way of using machine translation -- in the form of auto-write suggestion. Both Déjŕ Vu X3 and Trados Studio (with this or this plugin) allow for this. Especially when I had several machine translation engines offering me suggestions right on the point of my cursor and responding to something that I had already typed, I noticed that I a) was able to work faster and b) was not unduly influenced by the suggestions. Of particular interest to me, I was able to use all that in combination with voice recognition.

More in the next journal. 

As a subscriber to the Premium version of this journal you have access to an archive of Premium journals going back to 2007.

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

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

If you would like to promote this journal by placing a link on your website, I will in turn mention your website in a future edition of the Tool Box Journal. 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.

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

Should you be  interested in reprinting one of the articles in this journal for promotional purposes, please contact me for information about pricing.

© 2014 International Writers' Group