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

There is a reason why I haven't written about memoQ for a while. memoQ is a very powerful and highly capable translation environment tool, but there have been a number of miscalculations in its product packaging, one in particular, that caused a bit of frustration among some users. It frustrated me because of Kilgray's unwillingness to discuss this openly with me.

The business model that Kilgray adopted a few years ago was this: After purchasing memoQ, you were strongly encouraged to purchase annual support contracts (at 20% of the list price of the memoQ version you own). These entitled you to support as well as upgrades to full releases (versus only relatively minor updates that were free to all). If you didn't purchase such a contract, you would have to pay up the missing period between the end of your last support contract and/or the one-year support of a newly purchased version to upgrade to the new version. Guess what? Not everybody purchases the support contracts -- particularly when version 7 just kept on coming up with a multitude of updates, some of which very well could have been complete upgrades. (Version 7.0 -- aka memoQ 2014 -- was released in June 2014, and for version 7.8 alone -- aka memoQ 2015-- four different versions were released from May 2015 on with a host of new features, including the Trados Studio plugin, find/replace in tags, customizable QA checks, weighted counts, introduction of MS Translator Hub, and so on and so forth.)

When version 8 came along at the end of February, however, some people -- who hadn't been "forced" into a support contract for the 2.5 years of version-7 bliss -- had to fork over quite a bit of money, and they felt that version 8 wasn't actually worth the upgrade price. Suddenly, the previously well-liked Kilgray team were the new villains on the block. Well, that's probably over-exaggerated, but the truth is that all of this didn't go over well, especially for freelance translators who were particularly poorly served with that version. I think Kilgray quickly realized their mistake, and they have tried to make versions 8.1 and the now released 8.2 more palatable to that user segment.

After not having written about versions 8.0 and 8.1, I decided now was a good time to speak about 8.2 with Gergely Vándor, memoQ's Product Designer Team Leader.

Version 8.1 included features like the TransPDF service (I reported on that in edition 264 of the Tool Box Journal), the round-trip transfer of Trados Studio track changes, and searches in previews.

Version 8.2 naturally comes with its own array of new features (more on that below), but during our conversation, Gergely made an interesting comment that got me thinking. He said that what users of translation technology (and in particular translators) want today is not so much new features but improvements of existing features.

I think he is right in one way: Translation environment tools, and in particular tools like memoQ,  Trados Studio, Wordfast Pro, Déjà Vu, and others, are chock-full of features, so much so that it's puzzling to choose from an array of options even for the most basic of processes. Do we want those features to work flawlessly and be continuously improved? There's no doubt about that. In fact, sometimes I wonder whether we would benefit from a pruning rather than adding of features -- or at least some dynamically driven feature display to only those users who want to use them. (AI, anyone?)

But (and I'm sure you expected this to come up at some point) that's only part of the story. We might be missing some technology thought leadership on the side of tool developers. And maybe that's not even their fault; maybe it's us who are not good at talking to them and sharing what we want.

When technology developers first introduced translation memory -- or when they introduced a combined processing of terminology data and termbase data, or when they introduced subsegment matching, or when QA feature sets were introduced, or when server- and cloud-based workflows were introduced -- they provided leadership and benefitted most of us in the process. (Of course, it's not true that "they" did all that in a vacuum without any push from their users, but it's also true that they still were the ones who came up with the mechanics and enabled us to use the processes that these features give access to.) I happen to think we're not nearly at the end of those kinds of breakthrough developments. As Lilt has shown in the last couple of years when they introduced adaptive and interactive machine translation, or when Déjà Vu introduced fuzzy match fixing with machine translation, there is a long way to go in how we can find ways to use machine translation data much more productively than we thought possible. New features for new features' (or the developers' glory's) sake are useless and undesirable. But we (or I should probably say: I) want developers to come up with ever better ways to seamlessly access external data that make me a better translator. That certainly includes (but is not limited to) machine translation data in its various forms.

Back to memoQ 8.2 and its new features. You can read about all of them, grouped for the different kinds of user groups, right here, but here are the ones I found particularly interesting:

• Machine translation has two new ways of application: You can do a "concordance search" for a machine translation suggestion (i.e., highlight a part of a segment and retrieve a suggestion from associated MT engines). Even more useful, you can also use machine translation to fix fuzzy matches ("Patch fuzzy matches"). The latter feature can potentially be a very useful productivity boost because the data retrieved from the MT engine is only the "incorrect" part in the fuzzy segment, which is then automatically inserted. Naturally, the less morphologically rich your target language, the more useful this feature is.
• Also related to MT, users of the server edition can now batch-assign MT configurations to translators. The benefit is that translators don't have to configure (and potentially pay for) this themselves when working on server-based projects, and project managers have a lot less explaining to do. (And by the way, the default setting for setting up server-based projects is still that the translator can configure any additional resources to his or her liking.)
• "X-Translate" allows the translation from a previous version of a file to be used to pretranslate the next version of that file without actually going through the translation memory. This is helpful because it allows for a great deal of reliability of the translated (probably better: transferred) segments. While this feature has been around, it is now possible to also do this with documents in which the content has been shuffled around a bit.
• The "cascading-filters"-feature -- i.e., the ability to define and process nested formats within other formats -- has also been there for a number of years (I told Gergely that I thought this is one of the strongest features of memoQ), but it's now also possible to apply regular expressions to the cascading filters and thereby fine-tune the processing. I asked Gergely for an estimate of how many users actually use the rather geeky and complicated regex language, and he said maybe two or three percent, but these are still important.
• "Rigi" is now integrated into memoQ (as it is in SDL Passolo and Trados Studio). I actually met with Rigi's CEO Henk Boxma to learn more about the tool, but we decided that I would write about it in more detail at a later point. Just this much now: Rigi is a developer-centric tool that is able to capture data within its appropriate context in dynamic web applications and WPF .NET apps. The translator can then translate that in her regular environment while showing a dynamic view of the content in a reference window. Very clever -- especially from the vantage point of translators who are often frustrated with lack of context in the localization of modern software applications.

Naturally, the list of new things -- and improvements to existing features -- is much longer, but overall it seems to me that memoQ should win back some of the translation community's good will that it lost earlier this year. 

30% Off on crossMarket Premium Subscriptions!

The online network brings together translation service providers and buyers. Join now and save 30 percent for the first year of your premium subscription. Log in to crossMarket, click "Upgrade," and enter the discount code TOOLBOX30 during the payment process. Valid from October 5 to October 19.

www.crossmarket.net

I have previously written about the machine translation efforts of the World Intellectual Property Organization (WIPO). I've been impressed with what they do, not primarily because of the quality of their translations but because of their very inquisitive and creative approach to MT. The feature that I talked about in the past was a beta version of their statistical machine translation engine, which not only presented a translation of a given text but also dozens of variants of any fragment of that text that would show up when you highlighted said fragments.

A version of this has now been released as an official product, but it is now based on neural machine translation. You can choose between the different variants by simply highlighting the respective part in the target field. There is still a bug in the system's capitalization rules for the variants, and unfortunately the rest of the segment doesn't change when one subsegment is changed (as it would in DeepL), but once again the idea is really cool. It points to a process that might turn out to be very promising for translators who could choose from many variants of fragments while they translate. In reality, however, this might have to wait awhile since there is no API at this point, so it's not possible to integrate this right into a translation environment tool.

There is no clear language on the ownership of the data once it's contributed or edited on the website. I asked Bruno Pouliquen of the WIPO team, and this was his response:

To make it short: we are not using the post-editions at all. This is only a facility offered for users, not a way to gather "crowd sourcing contributions."

In the last Tool Box Journal, I mentioned Trados Studio as the only tool that offers easy access to Google Translate's neural and statistical machine translation engines, something that can be really useful for some language combinations and depending on how you use MT (if at all).

The latest beta version of OmegaT (4.1.2) also offers super-easy access (in the last official version of OmegaT, things were rather difficult to set up). In the beta version, all you need to do is select Options> Preferences> Machine Translation> Google Translate> Configure and then select "Premium Edition" if you want to use the neural engine, or unselect that option if you prefer the statistical engine.

Wordfast Pro 5 (see elsewhere in this Tool Box Journal) also seems to offer that option, but it apparently does not work (at least not in my attempts: I received only neural responses, no matter what I checked).

In my conversation with Kilgray for the memoQ article in this edition, there was simply no awareness on Kilgray's side that this would be a helpful feature and -- since it's very easy to implement -- should be supported. This is a good example of where we as translators and technology users need to have a strong voice and vigorously ask for that feature.

So, tool vendors, PLEASE implement a switch between the neural and statistical machine translation engines in your Google Translate interface!

The My SDL Trados app, at your finger tips

My SDL Trados is a new, free companion app that allows you to stay in touch with the latest SDL Trados news, education and professional development resources on your mobile phone or tablet.

Download here »

You may not have realized it, but a race is on in the tech sector, and it appears that Google has just broken away from the pack. The race? To produce the first set of actually usable earphones or earbuds that can translate (not interpret, more on this distinction in a minute) conversations across languages in "real time."

We can thank Gene Rodenberry's original TV series Star Trek and Douglas Adams' The Hitchhiker's Guide to the Galaxy for popularizing the idea of a gizmo -- the universal translator -- or a leech-like creature (eww...) that feeds off brainwaves -- the Babelfish -- being used to cross the language barrier. The two were no more than storytelling conventions that made it possible for any alien species to have its own rich and complex language and culture but still be able to communicate with Captain Kirk and other characters quickly and efficiently -- in the language of the viewer, of course -- so as not to bore them with too many pesky subtitles or some other bilingual character providing a consecutive "translation."

The whole concept of a machine doing our job has been scoffed at for years by professional interpreters, and with good reason. Yet, there is something terribly attractive about being able to put something in your ear and simply listen in your language to what someone is saying to you in another. In essence, this is what conference interpreters, with the help of technology, have been doing since the Nuremberg Trials. The difference now is that the technology not only delivers the interpretation but is beginning to be able to produce it as well, circumventing the human interpreter altogether. With the onslaught of these in-ear devices, we interpreters now face the same questions that our translator colleagues have been grappling with since the advent of machine translation.

I am not saying that these devices are here to replace us. They are not, nor can they...yet (never say never). This is where I would like to make a clear distinction. The tech world that is producing them always refers to these gadgets as "translators" rather than "interpreters." I say we let them. Not one of these new devices is capable of truly interpreting what a human being says. They are, in effect, speech-to-speech machine translators with speech recognition and speech synthesis tacked on either end. No human is involved in evaluating the social, cultural or other contextual elements that can impact understanding. Thus, interpreting the original intended meaning is left to the human listener.

Expanding Access? Maybe

When I speak to translators and interpreters about technology, I often divide technology's effects into two categories -- replacement and expansion. That is, technology can either replace us or expand our opportunities. In the case of these new mobile speech-to-speech translation devices, I see them expanding rudimentary communication across languages, the kind that professional interpreters likely wouldn't be hired to interpret for in the first place.   

It is difficult to cut through all the hype around these in-ear translators since the companies' promotional materials are all effusively positive, using phrases like "the world's first translating earphones" or "it all happens simultaneously without interruption." The tech and general media only compound this self-praise from the companies with headlines like "Google's Pixel Buds translation will change the world." Interestingly, many of the promotional videos for these products have an English-speaking-boy-meets-non-English-speaking-girl plot that can only be resolved with speech-to-speech translation technology, so the companies are definitely putting their marketing savvy to work.

The reality is much less superlative and problem free. In real life, this kind of technology-assisted communication is nowhere near as seamless or glamorous. Additionally, it is still unclear just how or if end users will employ this technology in useful or meaningful ways, but the possibility is there. The anecdotes of first responders using Google Translate in emergencies and other scenarios when finding a qualified interpreter is logistically impossible or financially impractical continue to appear in the media. Time will tell if these technologies will be up to the task.

The Main Contenders

This incipient speech-to-speech in-ear gadget segment of the tech sector has become quite crowded as of late. Here are a few of the contenders. It is important to note that none of these devices is currently available off the shelf. You can preorder with delivery starting as early as mid-October for some, mid-December for others.

• Lingmo Translate One2One earpieces do not have to be paired with a mobile phone app to function, making them the only autonomous speech-to-speech translator on the list. Lingmo is powered by IBM Watson artificial intelligence. It can translate any combination of nine different languages. Unlike the other gadgets on this list, Lingmo only does speech-to-speech translation. You can't listen to your favorite music or talk on your mobile phone. I should have a pair of the One2One headsets in my mailbox later this month for testing and will be writing a review for The Tech-Savvy Interpreter soon. Price $279.00.
• Google Pixel Buds made a big splash this week when they were unveiled at the Google launch event. Of course, they only work with Google Translate and the new Google Pixel 2 smartphone and on-demand translation is only one of their features. One big advantage of Pixel Buds over other in-ear translators is that they do not require a complete stranger to stick something of yours in their ear to converse with you. You hear the translation into your language in your earbuds, and your conversation partner hears the translation into his or her language through the front speakers of the Pixel 2 smartphone. This earbud and smartphone combination can translate 40 different languages. Price $160.00.
• Waverly Labs Pilot Translating Earpiece -- This in-ear translator started as a campaign on IndieGoGo and has garnered just under $5 million without producing a single device for sale. That pretty well proves there is pent-up demand for this technology. The question now becomes, can they make it work? Delivery of beta-test versions is going on now with main production and delivery starting in December. The Pilot Translating Earpiece must be paired with the Pilot App (currently available in the iOS App Store and Google Play). I've tested the app without the earpiece. "Seamless" or "easy" are not words I would use to describe my experience. The basic edition provides speech-to-speech translation between English, French, Italian, Portuguese and Spanish. You'll have to pay more for languages like Chinese, German, Greek, Korean, Russian or Turkish. Price $249.00.
• WT2 Real-time Translating Earphone -- The result of another more modest crowdfunding campaign on Kickstarter, this speech-to-speech translator also pairs to an app on your smartphone using Bluetooth technology. Like the Pixel Buds, the WT2 also has a mode that allows you to use your smartphone to converse with others rather than requiring that they put the other in-ear translator in their ear. The form factor is futuristic and the promotional video slick, but this initial review from Linus Tech Tips brings a dose of tongue-in-cheek reality to this segment, showing that things are nowhere near as simple as the promo videos would lead us to believe. No retail price available yet.
• The Braghi Dash Pro -- These earbuds are designed first and foremost as wireless earbuds, not as an in-ear translator. Think of it this way, enjoy these wireless earbuds to listen to your tunes, oh, and by the way, you can use them for speech-to-speech translation too when paired with the iTranslate (iOS and Android) or Converse (iOS only) apps. The translation feature seems a bit like an afterthought on these earbuds. Retails for $329.00

Why Should You Care?

It is easy for professional interpreters to be dismissive of these gadgets. They are nowhere near capable of providing the high-level interpreting we do as professionals in a multitude of settings. But we need to be -- at the very minimum -- aware of these speech-to-speech translation platforms and understand the premise behind them so we can intelligently explain the difference between what they do and what we do. Otherwise, we run the risk of appearing uninformed about a development that is directly related to what we do every day.

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.  

DS-Interpretation, Inc.

Specializing in Conference Interpreting since 1972. New cloud and app-based technologies including Linguali, cutting-edge equipment & sound booths and professional interpreters world-wide. We believe in relationship-based customer service that focuses on meeting the needs of interpreters. As Company Founder, Bill Wood, said years ago: "Interpreters will not be replaced by technology, they will be replaced by interpreters who use technology."  Find out why interpreters love working with us.

www.ds-interpretation.com 

Before writing this review of the latest version of Wordfast Pro (5), I re-read what I wrote almost exactly two years ago about Wordfast Pro 4. I encourage you to read it. Wordfast was unhappy with me for appearing too critical but I thought (and still think) that I pretty much "nailed" it. Wordfast 4 was not a release that distinguished itself with a lot of new features (except a really new and different interface); instead, it provided overall enhancements of existing features or introduced features that the competition had long ago implemented.

I said this was somewhat surprising because other Wordfast products, in particular Wordfast Classic and Wordfast Anywhere, were real trailblazers.

Now with version 5 of Wordfast Pro, it's sort of the same: relatively few completely new features and lots of enhancements. But while my assessment might have felt a little critical two years ago, I tend to be more positive now. Why? Because Wordfast 5 feels surprisingly mature.

The interface still has a really different feel -- but a now slightly altered, comfortable, and well-organized feel. memoQ's Gergely's comments -- that translators just want existing features to be improved rather than new ones introduced -- rings very true in this program (though I only partially agree with that view). And maybe it's just not quite right to look at Wordfast Pro as a standalone tool only; instead, we likely need to view it in the context of its complete offering, including the Classic and Anywhere programs, which all are deeply connected. The cloud-based (and still free) Anywhere forms the axis that connects everything else by providing access to translation memory and terminology data for its Classic and Pro siblings as well. (Oddly enough, this approach of using the complete Wordfast structure is used relatively sparingly, according to Wordfast's John Di Rico, my source for this report.)

As far as notable improvements to users of Wordfast Pro 3 or 4, there is an increased flexibility with project structures and file locations; the target preview is now rendered in HTML (so MS Office does not have to be installed on the same machine to make it work); and there are more reliable progress indicators; better processing of memoQ and Trados Studio packages (not yet including terminology, but that's in the pipeline); and speed increases (though I found the program's response speed to displaying dialogs with a great number of options rather slow).

I did notice that there is only an English interface and help system. This seems odd for such a feature-rich program, and even I with my relatively good command of English have to read some of the options twice to really understand the nuances. Other similarly feature-rich programs such as Trados Studio, memoQ, and Déjà Vu have long translated their user interface and help systems (into eight, eight, and six languages respectively). Just because we're all translators doesn't mean everyone has a good command of English. This is also surprising because Wordfast is the workhorse used by mega-big and super-global LSP TransPerfect/translations.com, which at least partially owns the Wordfast company. (I know, I know: TransPerfect has embarrassed us all with its recent ownership-related antics, but many of us will have worked for them in the past or are still working for them, and if so, likely through Wordfast.)

I asked John about other large or large-ish translation companies that use Wordfast. There really aren't that many because, unlike many of its competitors, Wordfast has not built large-scale translation management into its program or lacks the connectors to management systems like Plunet or XTRF. The result: Much like the other Wordfast siblings, Wordfast Pro is really a tool for freelancers or small companies (again, with the notable exception of TransPerfect, which has its own management platform). And that is also reflected in the overall lower price than tools like memoQ or Trados. (Wordfast Pro has a licensing scheme according to which you have to renew the license every three years for 50% of the list price.)

One interesting new feature in Wordfast Pro is the flagging of often correctly spelled but wrong words (think pubic vs. public) or false friends. This is done according to manually created bilingual tables that do these checks for (at the present time) six languages (French, Spanish, Russian, Italian, Korean, and Portuguese) coming from English. I really like this idea, though the manual creation limits this feature to a very small number of languages. I could certainly imagine an AI-like approach to the feature, comparable to what Lilt does with morphological rules in the languages it supports (see the 269th edition of the Tool Box Journal). Either way, this might be a real pioneer for features that will show up in other tools as well.

And since we are talking about Wordfast, I also checked in with Yves Champollion about Wordfast Classic, the Word-based and Word-file-only translation environment tool (a descriptor that is becoming ever more true since Wordfast Classic's PowerPoint extension's recent discontinuation in favor of Wordfast Pro).

Rather than talking about how powerful Wordfast Classic is (this is the tool you should give serious consideration to if you exclusively translate Word documents), here is some info on its future.

In the latest version of the Translator's Tool Box, my computer primer for translators, I still say this about the Mac compatibility of Wordfast:

One clear advantage of Wordfast Classic is that it runs on any platform supported by MS Word, including Macintosh. It does not run on Word 2008 for Mac, but it does work again on Word 2011 and above for Mac.

True that, but I had not looked at Word 2016 for Mac. Here is the full story on that: The Visual Basic for Applications (VBA) module on which Wordfast Classic depends is once again next to non-existent in Word 2016 for Mac. Even though Microsoft has announced a fix, according to Yves, "Word 2011 is still the version of choice for WFC on a Mac."

What about the Windows version? Here is Yves again:

Let's put it this way: WFC will be alive as long as Microsoft Word supports VBA, and when that support ends, as long as translators are willing to run VBA-enabled versions of Microsoft Word.

All this is true only for locally installed versions of Word, so Wordfast is not designed to run in cloud-based instances of Word.

Wordfast Classic used to be updated every few weeks if not days, and while this is not happening anymore, it's not really for lack of development but more a decision to release more substantial upgrades. (And I'd be really grateful if the next upgrade wouldn't completely redo my Quick Access Toolbar in Word every time I open Wordfast Classic ...)

There is also a new, web-based aligner, which, much like XTM and LF Aligner, is based on data rather than just segmentation rules. I like that approach as I said elsewhere, but I'm just not so sure about this confidentiality clause:

The aligned material (all posted source and target material, and the alignment, or the TM, that is returned) are deleted every day. We do not disclose your material in any way. Our reputation is at stake. Note that our aligner learns from the submitted material by drawing associations between words or expressions, for the sole purpose of improving future alignments.

Essentially, this is the same argument used by all the online machine translation vendors. If you decide that their approach does not work for your data, I can't see that this would.

10 conferences you should not miss until the end of the year

As the conference reason becomes more active, the memoQ team is hitting the road again to meet their current and future users.

Meet our team, schedule a meeting, and let us help you make your translation workflow more effective.

We are looking forward to meeting you all over the world!