Join the SDL Trados team for some fun this December with a festive calendar.

Each day some exciting content will be revealed -- what will it be?

A free resource? A special offer?  

Keep checking the calendar to find out!

One reason it takes me longer than might sometimes be immediately apparent to publish each edition of the Tool Box Journal is that I have a quite a few talks/interviews with people and companies that don't actually end up in the Journal. My goal with this publication is to present you -- translation professionals -- with information that's relevant, and quite a few things that at first seem to fall into that category do not end up making the cut. That was kind of the idea I had about Termsoup, a small Taiwanese company that produces a translation environment tool I recently stumbled over on LinkedIn (they had also exhibited at the ATA this year but I had missed them there).

After talking to the company founders Joanne Chou and Vincent Li, it turned out that this one certainly fell into the category of relevant.

Termsoup has been around since 2016, and its market so far has been Asia, with a predominant focus on Taiwan. That's important because of the kind of translators the tool is built for.

In the last three years, Taiwan has seen about 40,000 new book titles published annually, with about a quarter of those, 10,000 books a year, translated out of other languages into Chinese (source). Needless to say that's significant, and it's also significant that this is a market predominantly handled by the publishers themselves who contract with translators and typically do not ask those translators to use any kind of technology. (Compare this with translation agencies who generally do ask translators to use some kind of translation environment, in the case of Taiwan typically Trados Studio.)

Joanne was among those book translators, in her case non-fiction, looking for a solution to make her work easier. Since tools like Trados and memoQ felt like overkill, and MS Word was definitely not the ideal translation environment, she asked her partner Vincent to help her code a new tool. This collaboration led to Termsoup, a very minimalistic browser- and cloud-based translation environment tool. If you ask me it's a bit too minimalistic, but I also work in technical translation for different kinds of output.

The name of the tool reveals that the emphasis is not on translation memory or machine translation but on terminology. You can access third-party terminology and add terminology as you translate -- not automatically, mind you, but by highlighting a source term or phrase. I prefer that terminology be displayed automatically -- as in other translation environments -- and they promised to look into it. I also suggested a less mouse-heavy approach (anytime you have to move your fingers from the keyboard to your pointing device, you lose time and it's a break in your workflow), which also was well received.

So what then does the tool do? Termsoup...

• uses MateCat's filters to convert a very large range of file formats behind the scene into XLIFF, which is then presented in a bilingual format (either vertical or horizontal) and which will be re-exported back into the original file format once the translation is done.
• does not slavishly carry out segmentation on a sentence level (as you can see in the image below).
• automatically accesses third-party, language-specific dictionaries and other resources that display results if you select a word or a phrase (which can then be entered into your own preferred glossary).
• allows you to connect to an MT engine (right now Google Translate through Termsoup's own API or "application programming interface"), which is paid as part of the monthly fee for the tool.
• hosts its data on Amazon's AWS servers in Japan (and allows you to save backups to your Google Drive if you choose to).
• allows several users to work on a text simultaneously.

Here is what it looks like in any of the supported browsers (Chrome, Firefox, Safari):

You can see the sparse interface with its focus on the text and the term search on the right, activated by the highlighting of a term in the source column.

According to Joanne and Vince, the by far biggest user request is to relax segmentation even more so the text is shown strictly in paragraphs (but, if possible, still saved as traditional sentence-based translation units in the TM). The plan is to do just that in the near future.

Here's what I find so interesting about this tool: for not being on most translators' radar, Termsoup is a remarkably successful tool with its approximately 500 paid users who have translated about 300 books within its environment. It caters to a specific kind of user group -- translators of long, often narrative texts -- in a very different market from the traditional model of translation buyer -> agency -> translator with a relatively established technology infrastructure. Termsoup is widely taught in translation classes in Taiwan and Hong Kong (side-by-side with SDL Trados), often due to students requesting its inclusion in the curriculum.

I think we (maybe I should say: I) tend to have a really good idea of how things should be and how the industry works, but this is one of those cases where once again we (I) discover how diverse the world of translation is and how much room there is for great variances in the technological solutions needed for it.

Joanne and Vincent are planning to market Termsoup beyond the Asian market (which, by the way, virtually excludes China due to payment problems between China and Taiwan) and will add customer service resources in the time zones that would then be needed.

Memsource 2019: Our Top Feature Picks

Read our round-up of the most important improvements made to Memsource this year; from Memsource Translate, our AI-powered feature which automatically selects the best MT engine for your content, to the launch of Memsource Editor for Mobile, the first-ever comprehensive mobile CAT tool.

Read more

We are about to say goodbye to the second decade of the 21^st century. Ten years ago, interpreting was on the verge of another significant technology-induced change, similar to the seismic shift from consecutive to simultaneous ushered in by the Nuremberg Trials. This latest shift has been brought about by staggering advances in mobile communication technology.

From Telepresence Studios to Web Conferencing

Cisco introduced its first Telepresence videoconferencing equipment in 2006 at a price point and with tech requirements that made it an option only for the largest of companies with very deep pockets. But its HD cameras, life-size video screens and high-fidelity audio made it seem like you were actually in the same room with people thousands of miles away. These initial systems (Cisco wasn't the only belle at the ball, of course) all ran on proprietary codecs and could only connect with endpoints using technology from the same provider. Initially, Cisco tech couldn't talk to Tandberg tech, which couldn't talk to Polycom tech, which couldn't talk to . . . you get the picture.  So, they were all building closed, proprietary systems.

Early on in this decade, there was a push toward "standards-based" systems, which made it possible (but not always easy) for equipment from different providers to connect to the same videoconference. Then, a few short years later, Web-RTC made it possible for in-browser audio, video, and data communications. The advent of Web-RTC was a catalyst for the development of remote interpreting platforms largely because it brought down the barrier to entry into the videoconferencing and web conferencing space, allowing tech entrepreneurs to create interpreting-friendly platforms at a fraction of what it would have cost just a few years earlier. More importantly, this and other technologies "virtualized" videoconferencing, turning it into a digital service that could be delivered to just about any endpoint (e.g. a PC, a smartphone, a tablet, or a big videoconferencing room). In other words, you no longer had to buy expensive, purpose-specific hardware to participate in a videoconference.

Remote Interpreting Becomes Economically and Technically Feasible

Many of the first movers in the remote interpreting space have either closed up shop after burning through their initial investment capital or were acquired for undisclosed amounts by larger companies. (Does anyone remember Babelverse, Verbalizeit, or Capiche?) But in their wake, serious startups have emerged that have learned from the mistakes of others and seen significant year-over-year growth in their business. As an example that big companies are beginning to pay attention to multilingual, in October of this year, we saw the first mainstream videoconferencing platform release the beta version of a remote simultaneous interpretation feature.   

Still, remote interpreting hasn't been the juggernaut that many feared. It hasn't displaced much traditional face-to-face work in the grand scheme of things. It's just an additional service offering in a growing, diversifying and increasingly competitive market.  

Professional Associations Start to Catch Up

How have professional associations approached new technologies over the last decade? Well, finally, after years of flat-out rejection or indifference, in late 2017 and early 2018, the logjam finally cleared, and many professional associations began to address remote interpreting as a method of service provision that is not going to disappear. Efforts are underway to develop international standards and best practices. Meetings and workshops are being offered to teach practitioners what they need and how to provide remote interpreting. And interpreters are finally engaging with the technology providers to provide feedback to inform the product development process.

To be sure, there are still sharp disagreements about working conditions -- e.g. should interpreters only work from studios (RSI) or call centers (OPI and VRI) or is working from home acceptable? Fair and appropriate remuneration for shorter virtual engagements is still a moving target. Downward pressure on interpreting fees is a concern, but it is a general problem and not exclusive to remote interpreting.

As this decade ends, remote interpreting has not turned out to be the boogeyman many feared it would be. The technology is sound. Working conditions are generally good to very good, and the amount of remote interpreting work continues to grow steadily year over year. However, I feel the biggest concern we still need to address is the sense of isolation that many interpreters report feeling as they increase the amount of remote interpreting they do. Interpreters are, as a general rule, social animals, and "being there" has been a big reason why many have come to and stayed in the profession. Remote interpreting definitely changes that.

All in all, I think the interpreting profession is finally coming to terms with the effects of mobile technologies on how we provide our services, but it took us the better part of a decade to do so.

The Coming Decade: Artificial Intelligence and Interpreting

As we prepare to enter the third decade of the 21^st century, there is another technological advancement that is knocking at our professional door -- speech-to-speech translation in the form of apps or devices that can listen to speech in one language and then produce a translation in written or spoken form with little lag time. While we were collectively fretting over the advancement of mobile technologies that made remote interpreting possible, breakthroughs in artificial intelligence (AI) made speech-to-speech translation (what many are calling machine interpreting) a reality, albeit a far-from-perfect one.

Notable advances in speech recognition, neural machine translation and speech synthesis, now coupled with mobile broadband data access, are poised to fill the gaps where human-powered interpreting is neither feasible nor cost effective. The good news is that AI-powered tools will also help interpreters improve their performance and productivity. Computer assisted interpreting (CAI) will only grow in importance in the coming decade.

High-quality human interpreting in high-stakes interactions, in other words just about every situation in which interpreters are called to work today, will continue to be the standard of quality required by clients and the profession. I'm not concerned about being replaced by a machine anytime soon, but I am concerned about the way people will perceive what I do as an interpreter as speech-to-speech translation grows. So, there are alarm bells that need sounding. The growth and expansion of AI-powered language solutions will continue to capture the attention of mass media the world over. And this, in turn, will affect the way the general public perceives the language professions, with many believing that AI can do what we do just as well or better than we do.

And that, colleagues, is the real threat we will face as professionals over the next ten years.

In the coming decade, not only do we need to continue to be tech-savvy but also much more adept at public relations or coming up with adequate resources so PR professionals can do it for us. Otherwise, we run the risk of looking like disgruntled has-beens left behind in the wake of technological progress.

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.

New Book: Interpreters vs Machines: Can Interpreters Survive in an AI-Dominated World, Jonathan Downie (Routledge, 2019)

Interpreters vs Machines asks whether advances in automatic speech translation will lead to the replacement of human interpreters and how human interpreting should now be marketed, sold and delivered.

Due for release December 22^nd, available for preorder from all bookshops and from Routledge.

I've been a friend of Tuomas Kostiainen's tutorials on Trados's features for a long time. He gave a super-thorough class on the latest edition of SDL Trados Studio's verifications features at the ATA this year. You may appreciate his helpful handout as a download from his website right here.

Here is a new feature in some MS Office 365 editions (apparently not every 365 user gets access to it; Microsoft seems to be justifying this by adding a "[Preview]" behind the name of the feature). It's called "Rewrite," and it allows you to right-click on a word or phrase, select Rewrite (Preview), and then be shown other ways of saying what you meant to say:

This clever quote by Shanta Gokhale should not be rewritten, and I also know there are many other third-party tools that do this and much more (and better). Still it's a nice-to-have tool that's at your disposal. (Oh, and dear multi-language folks: This is available only in English at this point.)

In the past I've talked a lot about bringing journalists and other public figures into a discussion about translation and interpreting when they publish misconceptions about our professions. Very subjectively speaking, it seems that these "opportunities" are getting rarer (which would be great, if true!) -- but that doesn't mean they don't happen. A few weeks ago George Gene Gustines, a senior editor at the New York Times, wrote an article about an upcoming newly translated and Americanized version of the beloved Asterix (Astérix) comic book series. Super interesting. Unfortunately, he failed to mention the name of the translator. The following dialog on Twitter ensued:

Isn't that so cool? Wouldn't you love for all journalists to have that kind of self-confidence and greatness to acknowledge an oversight and learn from it? I'm fairly certain this will be the last time George fails to name a translator in a story about translation. 

All file formats, all languages, all target groups, better quality, shorter time-to-market -- one standard solution.

Translation and Localization powered by STAR Transit

Watch the short video for more information on Transit functionality and usage.

To me it's not so much the "2020" that looks futuristic; it's when someone refers back to "2005" or "2012" as something long in the past and part of history that I get spooked. Wasn't 1984 supposed to be science fiction? (And speaking of science fiction: There have been some really great articles on science fiction and translation this past week, starting with this one about Ken Liu's Chinese-into-English translation of Liu Cixin's work.)

So, in that -- slightly uncomfortable -- spirit: Let's talk about "2019" as (almost) part of history, taking a look at what has happened in language and translation and how that might set things up to happen in 2020.

One appellation that made 2019 stand out was its designation as the International Year of Indigenous Languages (IYIL). There's always a choice of whether to be critical of what did or did not happen or try to look at it from a positive point of view. I think we had an unprecedented number of articles, radio features, and (I assume) television programs about indigenous languages, their relevance, and why it's important to support efforts to ensure their survival. While it's possible that not every person knew that it was the International Year of Indigenous Languages, even a semi-literate person would've been hard-pressed not to have encountered the plight of indigenous languages in some way or the other. I think that's a success. Also, rumor has it that the International Year of Indigenous Languages might be converted into an International Decade of Indigenous Languages. I would welcome that and would hope it would continue to be used as a way to educate the general public and to find ways to support indigenous communities in their efforts to preserve their languages.

I would also be delighted to have indigenous languages as an ongoing topic because of the Translation Insights & Perspectives project I've worked on for the past three years and have written about in this newsletter (and had a chance to speak about at some conferences in 2019). If you did not have a chance to read the original article or hear me out, you can do that here (since its publication, the tool has gained much more data) or you can visit the tool at tips.translation.bible.

Machine translation has been an ongoing topic in the world of translation, of course, but I think this year in particular we reached the point where we know two things: 1) we now have a good idea of what the latest developments of machine translation can and cannot do, and 2) we therefore know we don't have to worry about our jobs.

Whether we are post-editing machine translation, using machine translation as one of many resources, or are not using machine translation at all, each workflow provides viable opportunities to earn livelihoods as translation professional for the foreseeable future. I think it's exactly this fact -- that it comes down to our choice which of these three options we choose to use when we translate (assuming the languages we work with are supported by some kind of reasonable machine translation in the first place) -- that makes this a positive item on our checklist of what we have achieved this past year. Granted, you might have to look for new clients if all your clients want is for you to post-edit machine translation and that is something you prefer not to do (or vice versa), but that ends up being your choice. While it might require some effort on your part to look for new clients, it certainly should be well-invested effort if you feel strongly one way or the other about how you want to work with that technology.

We also know that the next wave of machine translation technology will come at some point. When it comes there will be further quality improvements, but my feeling is that we might have to wait a bit before we have a major jump like we had going from statistical machine translation to neural machine translation at the end of 2016. Why?

This is why:

The Abstraction Reasoning Corpus (ARC), introduced by AI scientist François Chollet,
tests intelligence systems with few training examples. (Source: Arxiv.org)

I borrowed this image from this article that talks about the limitations of the current kind of "narrow" artificial intelligence we are dealing with right now. Look at the image. Can you guess what the next square would have to look like? Unless you are having a real brain freeze right now, you'll have no problem deducing the appearance of the next square. With the current form of artificial intelligence we're dealing with, no computer would be able to tell you the answer based on the very small sample of those few squares. This is where our intelligence differs from machines. We don't need large samples to make logical conclusions, unlike machines. This is why we're good at language, and in extension with translation, especially when it doesn't fit certain patterns. Machines are limited in those cases.

• Available within Memsource and memoQ
• Compatible with all CAT tools.
• Fast log-in for Proz members.

As far as how we're using machine translation, 2019 was good in the sense that we've talked more about creative uses of machine translation suggestions beyond just post editing, and I think we'll see more of that in 2020. But we will also see more ways to help us automatically evaluate the quality of different machine translation suggestions, maybe even on a per segment basis. Memsource has been doing some work in this area as well as Intento, which is embedded or embeddable in Smartcat, YiCat, SDL Trados, and memoQ, and has an API connector to use in any other kind of environment.

I have been puzzled to see that we seem to have headed in the wrong direction as far as separating technology and services. This was a big topic in 2005 (See what I'm saying? Reads like yesterday!) when SDL bought Trados and we discussed the problematic messiness of having a service and technology provider be the same. What we have now seen is that companies like Smartling and Lilt, both of which prided themselves on being technology providers only, were pushed by their investors to combine their technology with a service offering as well. In the case of Smartling this doesn't really matter so much, since the tool is really not accessible to most language providers as a standalone tool to start with. But it does matter in the case of Lilt, which is accessible as a technology only if you use it in providing services to clients via Lilt (and in that case you are free to use it for other clients as well). To me the lack of ready availability of this powerful technology is a sad and regrettable loss, but likely one that we have to blame ourselves for, since it's us who didn't support Lilt in its previous attempts to market to us directly.

I hope 2020 doesn't see more examples of this.

On a positive note, I do think 2020 will continue to see translation paid increasingly by the hour rather than the character, the word, the line, or the page. When I wrote about this topic the last (of many) time(s) about two years ago, it still seemed a bit like futuristic wishful thinking, but at this point it's neither futuristic nor wishful since companies have already started moving that way. We'll all be the better for it (provided we're smart enough to negotiate reasonable rates).

Here's something else that's good: one session at the ATA conference this year made me especially happy. It was a session about subtitling tools that can be used for translating, well, subtitles. It wasn't so much the tools' powerful features that made me glad; it was what they didn't have, especially in comparison to the ready availability in the translation environments so many of us use day-in and day-out. We've come such a long way in harnessing powerful ways to provide high-quality translation on so many levels within our technology. No need to copy and paste text into MS Word to spellcheck it or use external tools for word counts or consistency checks. These are features we use every day without even thinking about it (only realizing how far we've come when we look at other fields in a more nascent stage). And as far as subtitling? Fortunately a number of translation environments now support a number of subtitling formats quite well.

So I don't think we're going to see any breakthroughs when it comes to translation environment tools in this coming year, though we might see more specialized tools, along the lines of Termsoup that I write about elsewhere in this Journal. We'll also see a more pronounced use of cloud-based technology, whether in a hybrid form with a desktop tool and a cloud-based server or a browser-based tool that works on browser-based content. (By the way, it's remarkable how much later this is coming to the world of translation in comparison to other industries.)

The most important and at the same time oldest "development" of 2019 may be specialization. I know your eyes are glazing over right now, but since apparently I've been added to a database for translation agencies, I receive a lot of résumés from translators. It always makes me feel a little guilty when I delete the résumés without replying (though a little research would have shown that my company is not looking for translators), but every once in a while I look at several and am often gobsmacked to read about translation services offered without any recognizable specialization. It might've been fine to offer such general services in the 1990s but certainly not today. If anything, we can expect a hyper-specialization in the 2020s. So, if you're not ready for that yet, it's time to start.

Translator's Black Friday returns! Exclusively for the smartest among translators -- Tool Box Journal readers! Win a discount from 25% to 90% on AnyCount, TO3000, or Projetex. Spin the Wheel of Fortune and try your luck at https://translation3000.com/aitpn/557-74.html.

You've been a hardworking translator for the whole year, and now it's time to get your reward. Focus on translation, not administration. The wheel stops on December 15 -- get it or regret it!