Translators are at the core of everything we do at SDL Trados, so we want to celebrate their successes and tell you their stories. Learn how some of these translation professionals have grown their businesses and hear key advice to help accelerate your career.

Learn more: https://www.sdltrados.com/landing/we-are-sdl-trados.html

In the last Tool Box Journal, I featured an interview with Konstantin Savenkov, the founder and CEO of Intento, a service that determines which machine translation engine is the most appropriate and cost-effective for any given document. Intento is typically used in the form of plugins, including MS Office plugins, but for our (=translators') purposes also in a variety of translation environment tools.

Keep that in mind while we switch scenes for a second: I remember a conversation with David Canek, CEO of Memsource, a few years ago, about why he wasn't pursuing partnerships with other technology providers. Back then we talked about iLangL, a third-party service that allows tools to connect to content management systems. While iLangL had actually built an existing connector for Memsource, David told me then that he felt it was wiser to rely on one's own development capabilities if at all possible. (Click here to see an example of the CMS connectors that Memsource built on its own.)

While I did not ask David and Dalibor Frívaldský, Memsource's CTO, that question directly when I talked to them this morning about their own MT selector, Memsource Translate, I have a sense that the answer would be somewhere in the same ballpark. The solution they have developed is not so different from what Intento offers, only it's not a third-party service but directly integrated into Memsource.

Released last year, the service looks at texts in relation to the many machine translation engines Memsource provides access to. (Its list does need to be updated, though: Fair Trade Translation, for instance, has not been around for a while now.) It individually evaluates each language combination that is needed and then uses the MT engine most well-suited to support the translation of that project. Just like Memsource's two other AI-driven features -- the language-combination-specific recognition of non-translatables; and MTQE, or "machine translation quality estimation," which estimates the quality of an MT suggestion -- this is also driven by and used in combination with artificial intelligence technology. (This is one of the benefits of using a cloud-based tool through which you can learn from data across the network.)

I asked David and Dalibor whether there are plans to use this technology on a per-segment level, the same question I had asked Konstantin from Intento, and the answer was not immediately (the next feature, to be released in Q2, will be automatic domain recognition of texts in relation to assigning MT engines). I think the segment-specific recommendation is one of the more interesting "next things" when it comes to machine translation. With so many machine translation engines available, we've clearly reached a point where a technology like Intento or Memsource Translate could do the analysis they are already doing, decide which, say, three MT engines produce the best language-pair-specific results on a document basis, and then either show all three of those for translation purposes or do another more fine-tuned automatic evaluation of those three and just present the "best" MT suggestion.

A professional Machine Translation Tool for professional translators

Get your FREE trial and enjoy the benefits of instant project deliveries and high-profit margins. Post-edit all your jobs and make more money! 

www.goTranslation.com

I find translation environment tools interesting on a number of levels. Clearly, the most important level is how (and/or if) they help translators in their work. How much more productive can a translator become when using a particular tool, while maintaining or raising the quality of their work output? In that context I've always maintained that the cost of a tool, whether it's a monthly usage fee or the cost of a perpetual license, is really only relevant in relation to the gain in productivity. A free tool that doesn't make me more productive can end up "costing" me much more than one with a seemingly high price tag that radically increases my productivity.

Another area I've always found interesting is how the market for translation environment tools continues to develop. Many of us see that we have the same market leader today -- SDL Trados -- as we had 20+ years ago and assume that not much has happened. But that's not true. While the numbers do show that SDL has maintained a lead, I would argue that there is actually quite a bit of diversity in the market. (Of course, this doesn't mean that some tools or approaches have not failed. In fact, I was surprised about how many tools I had to take out of my latest edition of the Translator's Tool Box ebook.)

One aspect of the market that especially interests me is how different companies try to monetize their products. When I look back 15 years or so, almost everyone monetized their tools by selling perpetual licenses and releasing paid upgrades every year or two. Some were a little slower in releasing those paid upgrades than others, but virtually everyone did it. And then there was a phase -- I think it was ushered in by Star's Transit product, but it was followed by many of their competitors -- where free licenses were given out to translators for versions of their tool that really were meant to work only on projects created in large server-based editions owned by translation buyers or agencies. The hope was that translators would so enjoy working with the tool that they would later happily buy a paid version. That model didn't work so well (partly because most translators were looking for fewer different tools on their computers rather than more).

Then came the age of SaaS (Software-as-a-Service), which typically meant tools were "rented" with monthly or annual payments. Since most of those tools were server-based, updates happened automatically with no additional cost or hassle for their users. With some exceptions, that model worked well for agencies and translation buyers (and still does), but not particularly well for freelance translators. I assume this is because of a (false) sense of security of "owning" a software license rather than using it as a paid service. It's "false," of course, because the concept of continuous "ownership" - which, by the terms of the contract, is not real ownership to start with -- only works until you update your operating system; the software support is suspended (Windows 7, anyone?); or your work environment -- clients, file types, etc. -- simply requires you to upgrade.

So a good number of companies have looked for completely new business models to sell their products. One of the more interesting ones has been Smartcat.

I was actually aware of Smartcat long before it was released. I recently met again with Boston-based CEO Ivan Smolnikov to have him update me on how things are going and to give me some insights on a recent shift in their strategy.

Let's have a few data points on Smartcat first.

Smartcat is a cloud-based platform that attempts to address virtually all aspects of the language service ecosystem. From the point of view of the translator (and really anyone else), everything has to be performed in a browser. Once your files are uploaded to a server where they are processed, they are presented to you in the typical tabular (source on the left, target on the right) translation interface.

The supported file formats include a very large range of word processing, software development, subtitling, content management system, and desktop publishing formats (including Trados package formats). Even more interesting, the list includes image formats, including graphic files and image-based PDFs. For these formats, Smartcat uses the ABBYY-owned FineReader OCR engine and does a fine job and in a large number of languages -- if the fonts used on the graphics and PDFs are regular fonts.

As far as translators are concerned, the OCR services are already a small part of their business solutions, because this is the only thing you pay for -- well, plus the use of machine translation services. Here, unlike with other tools, you don't pay a fee to any of the MT providers directly. Instead, you pay a fee to Smartcat (US$20 for 500 pages at 250 words, but that pricing structure is likely to change). Accordingly, if translators don't want to use OCR or MT services, they can use the tool completely for free.

And actually this is also true if you are a translation agency or a translation buyer. How does that work? Well, the business model behind this is what Smartcat calls "Connected Translation," a solution that covers all aspects of the translation business, including vendor management (a large marketplace offers access to thousands of freelance translators and agencies), workflow management, and highly automated billing services. It's the latter that are fee-based if you use Smartcat to send payments out (and have it automatically figure out different rates, currencies, payment schedules, and payment types -- all otherwise time-intensive jobs).

Only very recently, Smartcat has now also introduced some paid corporate accounts that include individual customer service, including customized solutions for non-supported content management systems. I asked Ivan who would pay for that if it otherwise could also have been had for free -- and he told me that corporate customers who primarily rely on Smartcat asked them to introduce pricing so they felt they could rely on the service that essentially runs their business.

(Also, a variety of back-office features are available only in the paid editions. To review those, follow the link above.)

I also asked whether Smartcat is enough of a business and workflow solution that medium-sized or larger agencies would select to use it exclusively as their translation management system. While they certainly could, he said, many still use legacy tools like XTRF or Plunet alongside Smartcat -- something he hopes will diminish once Smartcat begins to decouple the translation management system from their translation system in the next few months. (As every agency owner and project manager knows, the problem is that there is no way a single translation system is enough to satisfy all customers, and it's problematic if the management system is closely and exclusively tied to one translation system.)

Translation buyers, on the other hand, who naturally use only one translation system do tend to use Smartcat for all the other translation-management-related tasks as well.

New parts of their strategy include the possibility of purchasing premium services as we discussed earlier, as well as the fact that the marketplace, an essential component of their business strategy, now also includes agencies (previously only freelancers could offer their services). This is an attempt to cater especially to translation buyers who often prefer to select one vendor rather than having to select multiple translators.

And it's the selection process -- of translators and possibly agencies -- where much of the current development effort (of an impressive 70 or so developers) is being focused: using computer learning (if you want to be fancy, you can call it "AI") to match the right vendors with any request. This kind of matching goes beyond just availability or pricing but also looks at (the quality of) already translated materials and matches that with newly available materials. As I've said a number of times before, AI is finding its place in many shapes and forms within our processes (and project managers will be a more endangered species than translators -- see here).

If you are a translator (or an agency for that matter) who is still looking for customers, it might not hurt to register in Smartcat's directory.

GT4T is an intuitive MT add-on to all your Windows/Mac apps. It is now also a file translator that supports a great number of file formats. Get your Office documents or CAT files pre-translated without having to upload to a server.

GT4T

Michel Xynos from Greek service provider Intertranslations has developed and just released a machine translation service called goTranslation that he brands as "Professional Machine Translation" services. It's an interesting product in that it uses a whole lot of third-party technology and applies them in a number of different constellations for different projects.

Here's the process:

• Upload a number of file types (the various Office file types and non-image-based PDF files).
• Receive a price quote (5-euro minimum and then .1 euro per word).
• Upon payment and language combination selection, these files are converted (in the case of PDF files), translated, and formatted according to the original.
• Receive a notification that you can download the fully translated files, accompanied by bilingual files in table format that also contain the data of said files (the bilingual files come with minimal formatting -- underline, italics, bold).

I ran a test with a short but heavily formatted Word document and a corresponding PDF file with fairly impressive results. The formatting in the translated Word files was perfectly preserved and, more importantly, the formatting did not throw off the machine translation. The results from the PDF file (which is output as a Word document) were not quite as good (formatting was overall preserved but not as cleanly, and in one case the machine-translated flow was thrown off) but still useful.

I asked Michael about the kind of technologies he uses, and though he didn't want to disclose the actual product names, he did say that it's a potpourri of various MT engines and PDF converters that are used for different situations. In the case of my little test, I wouldn't be surprised if the "culprit" were DeepL because it returned the same results when I double-checked, but that's just a guess. And it's really beside the point because there are obviously other technologies involved, including the one that created the bilingual file.

So what's the usefulness for professional translators? I could see some coming up with creative solutions to use that bilingual file for something, even including turning it into a TM after some post-editing. But again, I think that we (me included) are still at the beginning of our journeys to find out how to reasonably integrate machine translation into our workflows.

I do find it super interesting that a little cottage industry has sprung up to specialize in buying from the large generic engines, reselling MT services to translators and end-users (in the case of GoTranslation, there will eventually be an option to select your own MT engine if you want to use that), and then adding some value-added service. (Another value-added service that Intertranslation will also offer at some point will be post-editing services on these documents, but that will be less interesting to translators, I assume.)

Oh, and as far as privacy, GoTranslation says it fully complies with the GDPR regulation. I assume this means they don't use engines like Amazon because, unlike the ones from Google, Microsoft, and DeepL, it does not guarantee that it won't use your data for their own purposes, even if you use their paid service.

PDF Translation for Professionals

Translating PDFs is easier and quicker with TransPDF.

• Integrated into Memsource, memoQ, ONTRAM
• Compatible with all CAT tools.
• Fast log-in for Proz members.

Try it FREE for your next PDF project.

5 Translation Technology Trends to Watch Out for in 2020

memoQ's Trend Report is a curious look into the most important developments the memoQ team believes will influence the landscape of translation technology and related fields in the coming year.

This year's report presents five trends that have managed to break out of the silos of the translation industry and establish themselves as prominent themes worldwide. Beyond audiovisual translation and hybrid translation workflows, these articles also explore how the global game localization market is expanding, how the translation industry in the Asia-Pacific region is expected to grow and what the most important security trends in the localization industry will be.

Learn more and read memoQ's 2020 Trend Report by clicking here.