Automated translation with human post-editing — does that sound tasty? If so, read on and comment below. Because a two-year-old survey of 228 individuals involved in global business by leading software translation solutions provider SDL revealed that — in Asia — 35% were already or else were planning to use automated translation with post-editing.
Translation for Business Global Survey survey, run jointly with the Association for Machine Translation in the Americas (AMTA) and the European Association for Machine Translation (EAMT), also showed that businesses in Asia are more concerned about quality of translation than their European counterparts.
The spell-check functionality of the earliest word processing programs was an early instance of computers helping us standardize our writing to increase value for a larger audience. However, given that translation is still 50% art 50% science, and business use any number of creative approaches to solving translation problems –from online and cloud-based to in-house and outsource — I wonder if that number is too low? I wonder — for example — if it accounts for use of Google Translate or Language Weaver by freelancers whose output belongs to enterprises who say they don’t or wouldn’t use MT (machine translation) / CAT (computer assisted translation).
The translation software business is about translating massive continuous work streams from one language to another — in real time, if possible. So if you’re wondering why you’d ever need anything faster, more convenient and — sometimes, anyway — accurate that Google Translate . . . well, maybe you wouldn’t. But that particular interface is presumably of no use to businesses who need software that learns both from what’s available on the Internet as well as the local intranet. while at the same time keeping all work product safe and confidential.
So what’s happened in the word of CAT and MT since the survey in late 2009? Well, the paid version of Google translation is now open for business with some predicting the free one will gradually become less useful. (That’s certainly what happened with Google’s free ET-phone-home service. I got unlimited free calls to the US (from Indonesia) for exactly one year and the quality and reliability was usually fine for business. Dried up on New Year’s Day 2011.)
Anyway, Google Translate power users (whose Trados or similar software connects directly to the Google translation servers ) have been scrambling for new solutions since that part of the functionality was officially “deprecated” in May 2011.
Speaking of SDL, however, the company purchased recently Language Weaver last year (from the CIA) for $US 42.5 million. Which means they don’t have to rely on Google technology to power the whole thing. And with everything else SDL has bought recently, this makes them an obvious leader in translation software.
Finally, here’s a year-0ld survey of professional translators (human ones) on the present and future of CAT. It takes more of a grass-roots approach and the upshot — I think — is this: how can you not use automated translation functionality when it’s already integrated in the industry-standard translation software like SDL Trados, Wordfast, Lingotek, Across, memoQ, Alchemy Publisher, MetaTexis, MultiTrans that so many translators already use. (Here’s a fairly comprehensive list of industry tools.)
(How this works in practice is Google makes suggestions for each of your sentences based on what it’s seen in the past. And you accept or reject, then edit.)
But just like the enterprises, translators often say they don’t use automation. Probably because computers still have such a hard time competing with humans and language use is one of the obvious arenas. So the pros may feel they’re less likely to confuse clients by saying, “No I don’t rely on automation.”
Depending on the language pair, industry context and type of document that may well be 100% true. Often enough Google — for example — can’t begin to make sense of the client’s text. Other times — happily enough — the situation is quite different.
So the question becomes, how do businesses and translated make use of the increasingly sophisticated options and programs that leverage CAT technology — not can computers actually translate human languages.