| ADVERTISEMENT |
| Terminotix announces SynchroTerm 2018 which offers the following new features:
- Cleaner and more modern interface
- Duplicates detection feature at record creation
- SDLXLIFF and XLIFF files support
- 64-bit compatible
- Faster extraction engine
SynchroTerm is a terminology extraction software used by a number of international organizations, LSPs and freelancers alike to accelerate the process of creating term records by employing statistical algorithms to automatically identify equivalent terms. If you have any questions about Terminotix products, please write to info@terminotix.com. |
|
1. Resources
|
| A few months ago, a now-retired Russian translator contacted me to ask whether I could make use of his many Russian dictionaries. Since I'm heading the Resources committee for the ATA, the question was not completely random; still I really couldn't think of a practical way to make the right kind of person happy with that generous gift. Long story short, we finally figured out that it would be a dandy thing to have a dictionary exchange at the ATA annual conference! Not only will his dictionaries be laid out and offered as free takeaways for those who would like to add them to their existing collection (you can find them on a specifically designated table at the conference), but your unused dictionaries can be offered as well (maybe you've given up on a certain specialization, you have duplicates, or -- bless your heart -- you've retired). So, please (!), bring those dictionaries to the ATA conference in New Orleans at the end of next month. If you're not able to come (I'm specifically thinking of those blissful retired folks), give me a shout and we can see that your dictionaries get there without you. Someone mentioned that paper dictionaries are not really being used anymore. Maybe I've become an old geezer, but I beg to differ. Especially those highly specialized dictionaries that will never enter the digital world are still very much appreciated. |
|
ADVERTISEMENT
|
| This is the last chance to become a proud owner of legendary TO3000, AnyCount, and Projetex. Monthly pricing is around the corner, and owned licenses will soon not be available. Exclusive offer for Tool Box Journal readers from Advanced International Translations: translation3000.com/aitpn/557-55.html 10 copies of each product available at 50% off. Focus on translation, not administration. Get your copy now. |
|
2. Intento
|
|
Before this article let me say a couple of things. I may sometimes seem to talk about machine translation as a tool for professional translators as if we've all agreed that this is a given and as if it's already in all our tool boxes. I know that's not the case. Many professional translators still have very strong opinions about why machine translation is not an option to supplement the many other resources we use -- and I think that's great (although I suspect that many of those think of machine translation only in relation to post-editing, which I don't consider a productive process myself). I think, however, that we
are at a point with this technology where we don't have to be upset about each other's workflows, where we can recognize that there might be a number of ways to get to high-quality and productive deliverables for our clients. And that includes listening to and watching how others do their work. It's really in that spirit that I talk about machine translation as profitable for those who select it as one tool in their well-filled tool boxes.
I've only mentioned
Intento before briefly (as an upcoming plugin to
Smartcat -- see
here), but I've been following it very closely because I'm intrigued by what it's doing. I finally had a chance to get together again with its CEO Konstantin Savenkov.
Here is what
Intento is and what it does: It's a consolidation platform that provides ready-made API (application programming interface) access to virtually all cloud-based machine translation services through a single
Intento API connector. The user -- whether they connect through an off-the-shelf translation environment tool or through a custom product -- can then switch on and off any of the different machine translation services according to how they perform.
Intento charges a set fee for processing text through its system (about US$ 3 per million characters for high-volume users) plus whatever the machine translation engine charges -- unless the client wants to use their own API key for those engines (which they might prefer if there are separate confidentiality agreements with the MT provider or if they have customized engines). Also, if the client asks for it,
Intento can perform a quality assessment of the different MT engines for the client and a specific set of data and give recommendations on how well the different engines perform. (Interestingly, Konstantin assured me that while they do provide that service, they much prefer to have the client do that themselves and just sell "product.")
To make sense of all this, let's first have a look at the available machine translation engines:
- Alibaba Cloud Machine Translation
- Amazon Translate
- Baidu Translate API
- DeepL API
- Google Cloud AutoML Translation API (beta) and Google Cloud Translation API
- GTCOM Yeecloud Machine Translation API
- IBM Watson Language Translator Service v2 (SMT) and v3 (NMT)
- Microsoft Translator API v2.0 (SMT) and v3.0 (NMT)
- ModernMT Enterprise Edition
- PROMT Cloud API
- SAP HANA Cloud Translation Hub
- SDL Language Cloud Translation Toolkit
- SYSTRAN PNMT API and REST Translation API
- Tencent Cloud TMT API (beta)
- Yandex Translate API
- Youdao Cloud Translation API
Even at a quick glance you'll see that the service offering goes significantly beyond the typically available engines, including Amazon and the trainable Google AutoML, IBM, and Baidu neural engines. For those (and other) engines, you would be hard-pressed to find API stand-alone connectors in your translation environment tool, for instance.
So then, can you as a translator just connect to
Intento? Well, it's a little more complicated than that. If you are using
Smartcat you can, but unfortunately that's the only translation environment tool so far with an existing connector. (By the way,
Smartcat has just
raised $7 million in funding.)
I
very strongly encouraged Konstantin to look into developing ready-made access plugins for tools like
Memsource,
memoQ, and
SDL Trados. Konstantin has already talked with the first two, but with the third, SDL, there has not been any interest among his customers. As to why that was, he surmised that his customer base is mainly very large translation companies who, according to him, often don't use
SDL Trados.
So,
Intento's focus is on large customers with large throughput (existing customers include Welocalize, One Hour Translation, and Janus Worldwide), but I hope I made my plea heard ("What are 20 large LSPs in comparison to 20,000 small ones or 300,000 professional translators?").
By the way, I had a chance to briefly touch base with Olga Beregovaya, VP of Language Services at Welocalize, about their use of
Intento. While at this point it accounts for only a relatively small percentage of their total use of machine translation (they just started working with each other), she couldn't speak highly enough of the high-quality service that Intento -- and in particular -- Konstantin provides. At this point, Welocalize uses
Intento for their "multi-engine strategy" that evaluates a mix of generic and their own customized engines through Intento. They also use Intento's evaluation service to get a quick overview of what mix of engines they should be using for any given customer, product, or language combination.
Of course, many smaller agencies are also looking into including MT in their product portfolio to distinguish themselves in a smart way. I can think of a number of good and potentially productive ways for smaller agencies to do this very seamlessly with
Intento.
And for translators? Here's what excites me: You have heard me say over and over again that only in the rarest of cases is machine translation useful in the context of post-editing (i.e., the "correction" of one suggestion from one engine). What is much more interesting for translators is the ability to access (or not access) the fragments of many machine translation engines at a time. This is exactly what
Intento can provide: simultaneous access to any number of suitable machine translation engines, and a configuration that can be changed at a whim with just a few clicks. Clearly it's important to know about the confidentiality settings of the different engines (remember, Google, Microsoft, DeepL, and SDL all assure us that they treat our data confidentially if we use their API to access -- which
Intento does), but once you know that in relation to the requirements of your clients, there is potentially a large data treasure out there.
It's possible that
Intento's team of 10 would have to be enlarged a bit to also focus on those new markets, as would their marketing strategy (presently consisting of reports, conferences, and relations with industry analysts), but to me it seems they would leave a very large chunk (in fact, the very biggest) of the potential market untapped if they do not include smaller LSPs and translators. (Plus, when I asked Konstantin whether they are profitable already he just rolled his eyes and said "We're a startup" -- so what difference do a few more employees make in that case, right?)
|
|
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! Would you recommend memoQ? Your review is highly appreciated. Click here! |
|
3. Similar to Similis
|
| The other day I talked to François Brown de Colstoun, the CEO of Lingua et Machina. Lingua et Machina used to develop and market the tool Similis. Here's what the Translator's Tool Box ebook says about Similis: "Similis differentiates itself from most other tools because it comes with a very high-level linguistic "knowledge" in seven EU languages (English, Dutch, German, Spanish, Italian, Portuguese, and French), which it derives from a powerful engine that was originally developed by Xerox for its XTS tool. This engine gives Similis the analytical power to apply linguistic rules to a number of processes, including alignment and automatic extraction of terms and phrases from translation memory content. In both of these processes Similis offers extremely high accuracy (of course, only in the language pairings mentioned above). For the actual translation process it offers two different environments: a hybrid Word/Similis environment for the translation of all files directly compatible with Word (Word, RTF, text files, etc.) and a separate environment for HTML and XML files. What makes the translation memory matches remarkable is the existence of "chunks," fragments of translation memory matches that the program was able to automatically extract from larger matches with the help of the XTS engine." Overall I wouldn't have recommended it as a main translation environment tool to anyone, but the terminology extraction engine was fabulous (in the languages it supported). And it didn't hurt that it had been a free download for several years. Well, Similis was taken off the market earlier this year, and Lingua et Machina gives the following reasons: 1. Similis is an older technology that is more and more difficult to keep alive. 2. Similis represents an obsolete way of working. Online platforms are much more efficient, particularly as far as updates are concerned. I kind of have to agree with that line of reasoning, even though it's too bad that Similis won't be offering that awesome term extraction feature anymore. Enter Libellex, a new product that is replacing Similis and addresses the concerns François mentions. Unlike Similis, Libellex is a completely web-based translation environment tool, though like Similis, it also has the incorporated XTS engine. And whatever languages are not covered for alignment and term extraction through the linguistics-based XTS engine are now supported through a -- likely weaker -- statistical model. Of course, Libellex is much more than "only" alignment and term extraction, but it seems that François has decided to use the complete tool at this point mostly in connection with his own service offering as a translation agency. Lingua et Machina works mostly for large French companies with complex XML-based documentation in content management systems (IBM Cognos, in particular), something in which he feels Libellex excels. (Just as side note: I think it's quite remarkable that more and more technology vendors only see their product in connection with services.) François gave me a quick demo of the tool, and I have to admit that I was not overwhelmed with what I saw -- it looked a little bit like browser-based tools of five or so years ago, but then there is that alignment and term extraction... Fortunately, he plans to offer those exact features as standalone tools in the near future and add some web-crawling functionality to it (so you can harvest project-specific bilingual data from the web). I'll be sure to let you know what those tools look like when we come to that point. In fact, just the day before I finished working on this Tool Box Journal, François contacted me again to announce the release of ComPair (not an easy word to type since Microsoft Word autocorrects it to "compare" every time you type it with an English language setting), a tool that allows you to align translated documents (and save them as a TMX file) or compare two different versions of a document with the changes highlighted side by side (and save that as an HTML file). The documents can even be in different file formats, including scanned documents (which are processed with optical character recognition in a surprisingly large number of languages via the open source Tesseract product). The cost of the tool is $5 per language a month (after a month's worth of free testing). I had a chance to run a few tests with the tool. The alignment feature worked adequately but was complicated by no true editing capabilities and no pre-configured way of joining and splitting segments. What I liked was the ability to alternatively show or hide tags and the fact that for the TMX export all tags were omitted. What I did not like was that there is only one selectable language variant per language (so only en-GB for English, for instance). Still, I can see potential for making this into a more powerful tool, especially with the above-mentioned web-crawling functionality, but I think there is a bit more work that needs to be invested. I really liked the compare feature, though. It works on the same range of files, including heavily formatted documents or PDF files. And the greatest thing is that the compared output can be exported to HTML with color coding to show all the differences (and then shared with someone who does not have access to the tool). |
| 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 300 dictionaries in 26 languages. WordFinder Unlimited was voted the winner of the Process Innovation Challenge at LocWorld in Warsaw 2018. Click here to find out 10 things you need to know about WordFinder Unlimited |
|
4. This 'n' That
|
| In the last edition I mentioned an OCR program for a number of Indic languages. Now another for all Devanagari-based languages has shown up: Sanskrit CR (and the -- anonymous -- person behind this wants us to cleverly pronounce it "Sanskrit Seer"). The application is really just a gateway to either the Google or the above-mentioned open-source Tesseract OCR engines, either of which you can select in the menu after you upload your image file. Spanish machine translation company tauyou (aka the company with the coolest website font) has been acquired by TransPerfect/translations.com. Ironically enough, it seems that the possibility of being acquired has -- so far -- been the best way for machine translation vendors to make money (this excludes companies like Microsoft or Google who have presumably figured out how to make money in their very multi-channeled ways). I asked Diego Bartolome, formerly the CEO of tauyou and now the Director of Artificial Intelligence at TransPerfect/translations.com, how reasonable it is that the tauyou services of customizing machine translation engines will be used by translators or smaller LSPs and/or competitors, considering that they would commit client-specific data to TransPerfect/translations.com to customize machine translation engines. Here is what Diego said: "At translations.com we have NDAs in place with our clients, that guarantee that their data remains confidential. That's also the case with machine translation, which indeed is part of the translations.com portfolio. We can offer the MT customization services to any company, and we are mostly oriented to our top verticals, i.e., Life Sciences, Travel & Hospitality, Legal, Financial, etc. Responding to your question, it would be harder for competitors to use our MT than for smaller companies. I have seen in the past many medium-sized LSPs using SDL's MT, but direct SDL competitors might not use their MT." Judging from other translation tools that TransPerfect/translations.com uses, including Wordfast and Alchemy Catalyst, this might very well be so. And speaking of SDL, here is something that I've wanted to write about for more than a year now, and I've decided to delay it one more time only to visit it again later this year: SDL's endeavors to use morphological recognition in terminology processing in the SDL Language Cloud Terminology tool. This tool has been in beta pretty much exactly since I've been planning to write about it, and it's expected to be moved out of beta in Q4 of this year (when it will also announce pricing, etc.). I'll write more about it then. If you have no idea what I'm talking about, take a look at this introductory video of the product, which also mentions the "linguistic fuzzy search." You will find a cloud-based product that is a lot less heavy and complex than MultiTerm and, since it's cloud-based, is completely sharable among collaborators. SDL's development team has put significant effort into the linguistic match feature (which distinguishes itself from the fuzzy match search that looks only for the similarity of three identical successive characters by looking at the underlying morphology of languages that use morphology), for which they used a third-party product (Elasticsearch) that they modified. More in a couple of months. |
| ADVERTISEMENT |
| SDL Trados Studio 2019: New Feature Spotlight Webinars Do you want to see the new SDL Trados Studio 2019 features in action? Join us as we deep-dive into some of the exciting new features in Studio 2019, which provide you with a brand new user experience - increasing your productivity and translation quality.
- How to use Tell Me technology in Studio 2019 and the new on-demand tutorials, tips and tricks. Register here >>
- How to use the brand new Project Wizard in Studio 2019 and perform mid-project updates. Register here >>
|
5. 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 endofsummeralready.
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. Here are two reader who have added the icon to their websites: www.swissglobal.ch/jobs/ www.capstan.be 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 |
|
