Tool Box Logo

 A computer journal for translation professionals


Issue 18-5-287
(the two hundred eighty seventh edition)  
Contents
1. Making the Move
2. Pivot Languages
3. The Tech-Savvy Interpreter: Interpreting Tech and the Power of 2.0
4. Interview with Spence Green
5. This 'n' That
6. New Password for the Tool Box Archive
The Last Word on the Tool Box
Hyperbole

In the British edition of WIRED magazine, its digital editor had this to say about the presumed overpromise of Google's new AI-driven conversation tool Duplex:

"A year ago, at [Google's developer conference] I/O 2017, the world's press collapsed into a flurry of hyperbole at Google's Pixel Buds. The promise of instant, on-the-fly language translation would 'change the world.' A few months later, when Google released its world-changing product, the hype collapsed. 'Overpromised, underdelivered,' lamented one review. But the promise was never (really) there. The Pixel Buds could supposedly perform real-time translation between 40 languages. It was the Babel Fish made real. The reality was rather different. The Pixel Buds were a mess of a product born out of a (very good) product demonstration."

I read that and cried, "Yes! Finally a journalist who doesn't know much about translation per se but realizes how off-message the general media is about translation technology!" But then I thought, why does this only come a year after the fact? Why are we -- who understand translation -- not better equipped and/or willing to stand up and cry foul when these hyperbolic press releases are disseminated in the first place? Shouldn't we comment on social media, write letters to the editor, and publish articles with an opposing view?

The answer clearly is yes, especially because we are truly well-equipped with an expertise that uniquely qualifies us, even though we might not always trust ourselves to raise our voices. And because of that qualification, our voices are welcome and respected -- as long as they are smart, constructive, and respectful.

ADVERTISEMENT

Join the memoQ Referral Program

memoQ has always been well-known of its collaborative approach. Now we have taken one step ahead, and launched a new initiative, which holds a truly collaborative spirit: the memoQ Referral Program.

memoQ users -- including server and translator pro owners -- can now earn credits by referring memoQ translator pro to translators.

Sounds interesting, huh? Visit the memoQ website to learn more about the program, and join today!

1. Making the Move

Despite writing a fair amount about (translation) technology, my computer equipment -- at least my hardware -- is rather modest: one laptop of just barely above-average processing power and a number of surrounding paraphernalia. Said laptop had reached the proud laptop age of five by the end of last year and started to slowly fail. While the hard drive itself had been swapped just a year or so ago and was working fine, other features failed in predictable succession (camera dead, wireless chip dead, several USB ports faulty, battery unreliable, etc., etc.). So I decided to purchase a new machine last November to replace its ailing cousin.

Like I said, that was last November, and only within these past couple of weeks did I actually end up making the switch. I hate to spend hours upon hours of being unproductive while installing programs, transferring files, switching registrations, arguing with various customer supports because of those registrations (or other installation issues), fine-tuning Windows and all kinds of other programs, and so on and so forth. Because while my hardware is modest, I'm relatively well equipped when it comes to software. After all, this is the "Tool Box" Journal, and while you could argue that the computer is the tool box, very few expect the container to be the focus; most would agree that it's all about the tools inside.

So I decided to put together a list of small tools that -- to me -- were essential when deciding what to transfer/install and what not. I'm not talking about the expected, non-negotiable -- and typically larger -- tools that have to be on the computer of any self-respecting translator. These would include the following:

  • Office installation (and I would actually say MS Office at that -- it's affordable enough to not want to have to deal with compatibility issues that invariably arise when using tools like LibreOffice, OpenOffice, or Google Docs)
  • Translation environment tools (unless, of course, you are completely browser-based in that respect)
  • Quality assurance tools
  • Terminology retrieval tools
  • Dictionaries (there are still a few that are non-web-based)
  • Bookkeeping/financial tools
  • PDF conversion tool
  • A set of the most common web browsers (Firefox, Chrome, Opera)
  • Graphic program(s)
  • Desktop publishing programs (I guess that's the only one you could forego, but I work a lot in FrameMaker so this is a must-have for me)
  • Security software

But then there's so much more. To me it's the smaller tools that make a lot of difference. Here are the ones that I didn't even think twice about installing again. (Please note that this is not about the actual tool but more about the kind of tool. There are always equally fine or maybe even better tools that do the same with a different brand name.)

  • PureText is a tool that allows you to paste content from your clipboard (i.e., anything you have just copied) into unformatted pure text. Great if you copy something formatted from a website or email (or whatever) into any other environment that supports formatted text. In those cases, the formatting from the source invariably messes up your document. This tool prevents that. It's super lightweight, unobtrusive, and free. And, yes, MS Word also allows you to paste unformatted text, but it neither does that in a straightforward manner nor are we just talking about MS Word anyway.
  • While we're talking about copying and pasting, I really like ClipMate (or one of its many "competitors," such as the open-source Ditto). ClipMate essentially creates a database of all your copied content and allows you to do all kinds of things with it, from accessing something you copied a week ago, to editing the copied content, to merging hundreds or thousands of "clips" into one and pasting it into one large record, and so on and so forth. I don't have it running all the time (though it will do that if you don't disable it), but once I do need its functionality, I LOVE it. We'll see how relevant ClipMate will be once Windows 10's upcoming "Cloud Clipboard" is released in the fall, but my guess is I still won't uninstall my good and trusted Mate.
  • Another (two) program(s) to help with copying and pasting . . : I've written about AutoUnbreak a number of times before, but it's just so helpful! There are a good number of PDF conversion tools out there. None of them are flawless, but with a little bit of patience and a semi-clean PDF you'll get what you want. Sometimes you don't want the whole PDF converted, but you just need to copy a paragraph or two. AutoUnbreak allows you to automatically take out all the silly paragraph markers at the end of every line in the PDF file and paste whatever you have as actual flowing text.
    This is what's so great about these programs. They do only one tiny thing, but they do that very well. And if that's the thing that otherwise costs you a lot of time to fix, they can be awesome. A similar sort of tool, though run by a completely different logic, is
    ABBYY Screenshot Reader. This tool assumes that any text it detects is an image and then converts it via an optical character recognition process into text. This is really helpful when dealing with content in sources like Google Books or other ebook content where you can't actually copy text. You can purchase this as a standalone tool, but it also comes packaged in a number of other ABBYY products, including FineReader. (Allow me to digress a moment about the -- non-existing -- ABBYY customer support: I had a fully legal licensing code for a non-current version of FineReader which was no longer downloadable from the ABBYY website. I wrote to them three times asking for help and never even received an answer. Since this is the second time I've been burned by ABBYY support, I have to believe that not providing adequate support must be company policy. By the way: I was able to find the right version at FileHippo, which provides access to current and previous downloads of all kinds of software.)
  • I guess a screen-capture program like SnagIt (or about 450 other screen-capture programs like it) is clipboard-related, too. While Microsoft has done a little more for screen-capturing in Windows lately with the slightly enhanced Snipping Tool and the much better Snip (which frustratingly has just been discontinued), to me it's important to have a tool that I can use to quickly communicate what needs to be said. Often the best way is to do a screen-capture ("a picture is worth a . . .") and write something on top of it, or merge several of them, or crop them, or . . . That's what a tool like SnagIt does.
  • For some reason I don't use Replace Studio Pro (the successor of the aptly named Search & Replace) as much as I used to, but I still didn't hesitate to re-install it. It allows you to simultaneously search through as many text-based files (.txt, .xml, .html, .rtf, .xliff, .mif, etc.) as you want and -- if you know what you're doing (!) -- replace any text-based element. It works incredibly fast, integrates itself into your Windows/File Explorer context menu, and makes me happy every time I use it (particularly because it also offers a safety net by offering to back up all the files in which you replace content).
  • For the same kind of files (text-based files), it's really important to have a more capable text editor than the Windows-supplied Notepad. The easiest choice here is the free Notepad++, but I also like and use EmEditor, particularly because of how well it works with code pages (and it's not quite as crazily feature-heavy -- or quite as "++" -- as Notepad++).
  • While these text editors are all able to work with HTML files, I find it helpful to have an editor just for those purposes. In my case I like the unpronounceable (and, at this point, very ancient, but still very capable) Nvu for HTML 4 files and -- by the same maker -- BlueGriffon for HTML5 files.
  • The zipping utility that comes with Windows clearly is not sufficient. My tool of choice is the no-nonsense and endlessly powerful 7-Zip (well, not quite endless -- if you receive a lot of Mac-specific .sit and .sitx files, you might need StuffIt). The only thing you don't want to do with 7-Zip is to compress files into the 7-Zip-proprietary format .7z (because it's, you guessed it, proprietary).
  • And, lastly, every Windows 10 (or 8) computer that I've ever set up or helped someone with gets a copy of Classic Shell, which emulates the behavior of the old Windows 7 Start menu.

Can you now see why I always dread switching to a new computer?

ADVERTISEMENT
SDL Trados Studio 2019 is coming soon -- Exclusive Preview Webinar

The user experience for everyone, from beginners through to experts, has been put at the heart of SDL Trados Studio 2019.

Discover how industry insights, trends and customer feedback inspired the upcoming new features in Studio 2019.

Sign up for this free webinar >> 

2. Pivot Languages

The disconnect that exists between stating strong opinions about the quality of machine translation and not actually revealing the language combination has been talked about a lot -- and for good reason. Even in the shiny world of neural machine translation, the quality differences between language combinations remain significant.

In this same context, one thing that hasn't been discussed much among translation professionals is the impact that "pivot language" translation can have. While we don't really know how widespread this practice is among vendors like Google or Microsoft, we know for certain that in "small" languages this is the norm rather than the exception. What does this mean? Rather than translating directly from language A to language B, a pivot language is used -- typically one with the most material in relation to both languages A and B -- and the translation goes from A to P(ivot) to B. This clearly brings a lot of advantages, particularly access to any kind of language combination within the number of languages covered by one engine, if there is at least one common language. But there are also a lot of disadvantages (remember the whisper game "Telephone"?), and it behooves us to at least be aware of them.

English-to-Finnish translator Risto Siikarla recently sent me this example that he stumbled upon. The Finnish phrase "anoppi kylässä" means something like "mother-in-law is visiting," but if you have Google Translate translate it into English, something entirely different and rather rude comes out on the other end. We've all seen pranks like this before (I assume that's what this is) where a number of users are able to manipulate a translation engine by overwhelming it with a concerted effort of "voting" for one translation. If, however, a pivot language is used -- which in the case of Finnish seems to be the case for at least all European languages -- then the prank is ever more powerful and a "little" disconcerting to anyone using this for productive purposes. To see what I mean, try changing the target language into any other European language, or, since this likely will be "fixed" once it comes to Google's attention, you can also see it in this graphic. (Viewing the graphic may enable you to appreciate why I didn't actually use it in the newsletter.)

Again, there is no problem with pivot languages in general, and if one agrees that there is some value in freely available machine translation tools for overall communication purposes, then one needs to be welcoming to "tricks" like this as well. After all, "relay interpreting" is also used, even at institutions like the EU or the UN. But when instances like this reveal how widespread this is -- at least for "smaller" languages like Finnish -- professionals who use these services need to be aware. In fact, it would be very helpful to have access to lists provided by the vendors that show which language combinations are translated in what manner.

ADVERTISEMENT

Your style, your wording, your content. Utilize translations quickly and reliably throughout your company.  

Corporate machine translation powered by STAR MT

Watch this short video for more information on STAR MT functionality and usage.

www.star-group.net

3. The Tech-Savvy Interpreter: Interpreting Tech and the Power of 2.0 (Column by Barry Slaughter Olsen)

May is a tremendously busy time for interpreters and educators, and for me it's a double whammy -- a busy interpreting season and the end of the academic year, with its qualifying and exit exams. So, my bandwidth is limited when it comes to tinkering with new technologies or researching emerging trends in the interpreting space.

That said, I keep interpreting technology providers in my sights and I am seeing a trend that indicates that the space is beginning to mature. Let me explain.

When I first started observing interpreting technologies seriously in 2010, startups were bursting on the scene with great fanfare in the hope that their creation (e.g. a VRI platform or an interpreting management system or the like) would be bombarded with eager clients ready to use the service. (Anyone remember Verbalizeit, Capiche or Babelverse?) When this didn't happen in a few months, they would either "pivot," an entrepreneurial code word for radically changing your business model or your targeted customer base in an attempt to save the business, or they would simply quietly fade away as their startup capital, their enthusiasm, or both, ran out. These early attempts at interpreting technology never made it to the next stage: version "2.0."

Fast forward to 2018 and a lot has changed. We now have a fast-growing interpreting technology segment with multiple offerings, including remote interpreting platforms, glossary management and term extraction programs, streaming audio delivery platforms that replace traditional FM and infrared systems, and a growing number of interpreting management systems. Several of these companies have now been around for three or four years, others even longer than that. They aren't seeing exponential growth, but they are staying the course and experiencing a steady organic increase in their customer base. The fact that they haven't flamed out in the first 12 months of their existence is a step in the right direction.

In today's software-driven world, interfaces get old quickly and new features can often be added without a huge investment in resources. Users provide direct feedback on the functions they wish a platform or service could provide, and companies respond. This is how technology evolves and improves. In the last six months multiple companies have released new versions of their platforms that include everything from improved echo cancellation (for sound quality) to entire makeovers of user and interpreter interfaces. Here are just a handful of recent examples:

Boostlingo 2.4 -- Boostlingo is a unified interpreting management platform that enables OPI, VRI, and on site interpreting. This upgrade launches on the same day as this publication, so I haven't seen it yet. According to the company announcement, Boostlingo's latest version will include the ability to schedule OPI and VRI calls in advance, a way to send written communications between clients and interpreters, and a totally revamped notification module. There are many more new features which I hope to evaluate in the new future. The long list of upgrades and tweaks has surely come as a result of clients' requests, which means Boostlingo has built a client base that is using the platform regularly and which is providing meaningful feedback.

InterpretBank 5 -- This is an app for PC, Mac and iOS designed for interpreters to create glossaries and manage and lookup terminology. I'm a newcomer to InterpretBank but I definitely like what I see. Since I started to use the software at the beginning of the year, there have been multiple upgrades and improvements. My simultaneous interpretation students worked with the software for the first time this year and have been impressed with its capabilities as well. The fact that the software is already on version 5 proves it has staying power.

KUDO 2.0 -- KUDO is a cloud-based web conferencing and live events platform with simultaneous interpretation. Earlier this month, KUDO launched an entirely redesigned user interface that incorporates a web conferencing feature with remote simultaneous interpretation. KUDO's VP of Communications Ewandro Magalhaes told me the driving force behind this new interface was to respond to client requests for multilingual virtual meetings. I attended the launch event from my office in DC and was duly impressed by the experience. You can read more about the event here. Yes, it is a promo piece, but it is informative nonetheless. KUDO has only been on the market for just over a year and has already pushed out a comprehensive redesign based on market demand.

Linguali Event -- When this "no equipment" simultaneous interpretation solution launched in 2016, it focused on replacing traditional simultaneous interpretation equipment for smaller bilingual meetings with Wi-Fi connected devices (Check out my video of Linguali IS from July 2017 to understand the original concept). The folks at Linguali quickly realized that there was a much bigger market out there for large events with simultaneous interpretation in multiple languages that could use smart devices as receivers. Their response to market forces in October 2017 was Linguali Event -- a more robust Wi-Fi-based interpreting delivery system that can provide simultaneous interpretation for up to 7,500 participants at once. It is a classic example of a company responding to market needs.

ZipDX Multilingual* -- Of all the solutions mentioned here, ZipDX has been around the longest. Its remote simultaneous interpreting solution for conference calls has gone through multiple iterations and its virtual interpreter console is now on version 4. There hasn't been a radical change to the platform or a major overhaul since it was first released, but look "under the hood" and you'll see constant tweaks to improve situational awareness (icons to indicate if someone is connected via mobile phone, land line or VoIP, or signals to show when an interpreter has Internet connection problems), improvements in audio processing (echo cancellation and automatic gain control), or improved chat functions for interpreters.

*Full disclosure: I helped design the ZipDX Multilingual remote simultaneous interpretation platform, so the argument can be made that I am not an unbiased reviewer.

These are only a few of the many interpreting technology offerings out there. It is not a comprehensive list of updated platforms. Whenever I "name names," I run the risk of missing someone or something. So, if you know of any other interpreting technology that has undergone a recent upgrade be sure to send me a quick email and tell me about it.

The upshot of all these upgrades is clear. The market is maturing and solution providers are learning how to respond to client needs. We are past the early stage of outside companies chasing an elusive slice of the fast-growing interpreting pie by building a "killer translation app" that could supposedly crowdsource bilinguals all over the world in the hopes of making a quick buck. Now, we have multiple companies built from the ground up by players from inside the field and highly knowledgeable about the market. And that, my friends, is a good thing.

Do you have a question about a specific technology? Or would you like to learn more about a specific interpreting platform, interpreter console or supporting technology? Send us an email at inquiry@interpretamerica.com.

ADVERTISEMENT

With GT4T, the ultimate online reference tool, you can use Google Translate, DeepL and many others without ever having to leave your work environment.

GT4T works in all programs. Get ready to be amazed.

Learn more and download at gt4t.net.

4. Interview with Spence Green

Following is an interview with Spence Green, the CEO of Lilt. Once you read it you'll understand why I felt it was a good idea to interview him.

Nice to touch base with you again, Spence. We've known each other since before you actually launched a product, and I've written about Lilt a number of times in different outlets. What has always fascinated me about you and your company is that you are an industry outsider, but from my perspective you've seen this as an opportunity rather than a disadvantage. Is that correct? And if so, could you explain that?

Spence: I'm not sure if it's an advantage or a disadvantage. It certainly means that we entered this project with a different set of assumptions. For example, we believed that MT could radically accelerate translation based on research that we conducted in 2012. Actually, the first paper to show improvements in human translation throughput was published in 2009. Surprisingly, this remained a contested perspective until the release of neural MT in 2016, but in fact neural MT simply optimized an approach that we already knew worked well.

It's been a disadvantage in that we didn't understand the complexity of the translation supply chain, and the relationship dynamics in that supply chain. The logistical and business model problems that need to be solved are surely greater than the technology problems on which we originally focused.

What have you learned about the translation industry since you launched Lilt 2½ years ago? Are there areas where you could teach us something about "us" (whoever that might be) that you can see more clearly than we see ourselves?

Spence: There seems to be a widespread belief that translation is noble work, and improves the human condition. That's good and true and not the case for many other industries. I know plenty of people for whom work is not a calling. That rarely seems to be the case in translation.

In my experience, what is most obviously missing from this industry is a culture of operational efficiency. Early in my life, and later at Lilt, I managed software engineers. Engineers seem compulsive in their need to write fewer lines of code, to optimize their work processes and tools, and to produce more faster. Everyone's looking for an edge, and this is good, because more software gets written, and more products get created. Recently I've started managing sales teams, and the culture is very much the same. Sales reps use tools and technology to generate and handle more deal flow, increase close rates, and shorten sales cycles. This is good for commerce.

In contrast, many in the translation industry seem to assume that the faster you translate, the lower quality work you produce. That just isn't generally true for human performance tasks. For example, my co-founder John is a much better programmer than I am. Not only does he produce more code every day than I do, but also his code is better than mine. This is true in many other human endeavors be it programming or translation or sales or running.

If the industry really cares about making information universally accessible so that more people can participate in the knowledge economy, then people should be championing operational efficiency. That means significant automation of PM work, use of MT on every segment, and so on.

Lilt seems to have gone through a number of developments -- not in a feature-based sense, but in how you|ve positioned Lilt. Is that true? Where do you see Lilt now and in the next couple of years? Would you say that Lilt evolved the way it did because you had outside investors?

Spence: Yes, we've gone through at least four different business model pivots as we've tried to figure out this industry. There are submerged rocks, some of which we've hit, and others which we've avoided. It seems that we've finally figured out a viable path forward, and we'll be announcing more about that in the coming months.

Our investors have been enormously helpful while thinking about e.g., the go-to-market strategy, our pricing, and the evolution of our product.

You have repeatedly talked about why you think post-editing of machine translation is not the best approach to working with MT. Could you explain that (one more time)? Can you give us two additional reasons why you think Lilt is a compelling choice for a translator?

Spence: First, I want to reiterate that post-editing is a workflow, and machine translation is a technology. "Post-editing" and "machine translation" are not synonyms. What translators have done with fuzzy matches for decades is precisely post-editing, the difference being only the provenance of the matches.

In general, people like active work. Ask ten journalists if they prefer to write their own articles, or edit the articles of their colleagues. You'll find that people much prefer to write their own articles. Ask ten software engineers if they prefer to fix machine-generated code, or write their own code. You'll find that people much prefer to write their own code.

Post-editing, i.e., the fixing of MT output, is a passive task. I find that the people who promote it typically haven't spent much time doing it. The situation is rather better with neural MT, and many translators now use MT as a reference in their translation environments, but it's still the case that the fixing of errors is tedious work.

The good news is that we just need to think a little bit harder about the workflow. That's where interactive MT (i.e., Lilt) comes in. We see interactive, predictive interfaces all around us now: when we execute a web search, when we type a text message, and in the near future, when we type emails (see Google's new SmartCompose feature). The simplistic "auto suggestion" features in current translation environments, which simply select sub-segments from static MT outputs, are not true predictive typing. The MT system can do much more if given a partial translation to complete. That's what Lilt does.

Lilt is also adaptive, meaning that it learns as it's used. We expect our TMs to be updated instantly. Why don't we have the same expectations for MT?

Two weeks ago we released our second generation NMT system. I've been working on MT for ten years now, and the technology has finally become what we knew it could be. Last week, I was staffing a conference booth and giving demos for Arabic> English. I got so wrapped up in translating with the new system that I quit talking to people. That never happened to me before.

Empirical and anecdotal arguments can be made against post-editing. But the best argument against it is an interactive neural MT system. Everyone should try it, if only to see what's now possible.  

ADVERTISEMENT

Imagine all the world's best dictionaries at your fingertips!

For a fixed monthly fee, on all your devices, integrated in your daily work applications!

WordFinder Unlimited -- One service, 5 applications and more than 260 dictionaries.

Click here to find out 10 things you need to know about WordFinder Unlimited 

5. This 'n' That
I was tickled to read MultiLingual's Katie Botkin's review of Translation Matters in the latest issue of MultiLingual, which delightfully lines up with other flattering reviews such as this or this. And just for those who missed out on previous announcements: You can purchase Translation Matters on Amazon as a Kindle book or a printed book , but you can also buy it as a PDF right here. In fact, if you would like to support the Tool Box Journal for a year (either by adding to your existing Premium subscription or by subscribing for the first time) as well as have a copy of the PDF edition, you can do just that by purchasing the PDF , and I will add the subscription on top of that for the rest of the month of May.
 
While I was never able to benefit much from itranslate4.eu in my own translation work, I thought its concept was really smart. It was a consortium of European MT providers (MorphoLogic, SYSTRAN, Linguatec, pwn.pl, SkyCode, Amebis, Sunda Systems Oy, PROMT, and Trident MT) that allowed you to access one or -- if applicable in your language combination -- several machine translations at a time. It was founded when statistical machine translation had taken over the reins at the likes of Google and Microsoft, apparently in the hope that the combined power of the involved companies could make a difference. Though it was available through its API in a relatively wide number of translation environment tools, most of you must have felt like I did: that the translations, and even the fragments, were not quite good enough to be useful. So, despite its somewhat clever name (itranslate4.eu: "I translate for you"), it had to close shop.
 
I mentioned the 2018 meeting of the European Association for Machine Translation (EAMT) in Alacant/Alicante, Spain, that is to be held on 28-30 May 2018, and I also said that, for the first time ever, there will be a translator's track. What I did not say is there is a survey that translators are asked to fill out in preparation for the track. You can find it right here. Thanks for being good sports!
6. New Password for the Tool Box Archive
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 lowtide.
New user names and passwords will be announced in future journals.
The Last Word on the Tool Box Journal

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.

© 2018 International Writers' Group