Microsoft CEO Satya Nadella

When I first heard of Microsoft’s real-time voice translation app for Skype and CEO, Satya Nadella’s bold prediction to “imagine in the very near future technology allowing humans to bridge geographic and language boundaries,” I was moved by such a world-changing utterance. I was also curious, and a little skeptical: the translation tools already available, to be frank, have never worked very well, not even in text format. A real-time voice recognition service that is actually efficient seemed to me to embody the idiomatic expression, ‘don’t run before you can walk’.

In view of accents, dialects, colloquialisms, lexical nuances, constantly evolving idioms, could real-time voice translation between a Glaswegian football player and a French oenologist discussing the vicissitudes of their employment bridge the language barrier? Or, as so often happens with Google Inc.’s translation tool, Google Translate, the translations become an amusing comedy of errors.

Low expectations based on experience

Having worked as a writer and editor in non-English speaking countries for the past 15 years, Google Translate was a most welcome technology for me when it first arrived. I naively thought would be an expedient for my job at the time as editor of an English magazine based in Thailand. The pleasant surprise lasted about a day. It didn’t just have “limitations”, it was pretty much disastrous when translating Thai to English. Similar to an online dictionary, it was capable, at times, of translating very familiar nouns. There’s one word for ‘love’ in Thai (rak), and Google translated this word correctly. But request a translation of the the idiomatic expression, ‘I am in love’, and the result is nonsensical. So was new this technology any kind of revelation, or was it just a dictionary with delusions of grandeur?

When I was a judge of the Junior Dublin Literary Awards for Thailand, I noticed that some young students, with obviously poor English-speaking acumen, would often hopelessly revert to Google Translate to complete their essays. In 2012 a few students translated the same text, which went like this:

“…their faces are no different from worm dead rotten carcasses, duties under the mask, it will be better to aim at removing the removal of flesh and blood that surges at the foreground mask. I myself would like to fool around with something dirty, filthy beyond description…”

I’m not sure where they got that from, nor if they understood the gravity of their crude statement. In a cultural context, such a statement would likely get them expelled from school.

Lost in translation

Languages, which share a related genealogy such as French and English, may give translation tools some credibility, but then try translating just one sentence of a wordsmith like the notoriously difficult-to-translate French author Louis Ferdinand Celine, who uses all sorts of strange French parlance, and the translation will be an ocean away from its intended meaning. I know this because I once tried it while writing a story, and had a French academic friend explain to me that it was a laughable translation. In his own words I should only use Google Translate for words like cheese…but certainly not cheesed-off.

Seriously Impressive?

The demo of a German to English Skype conversation was called “seriously impressive” by one tech writer. And in the words of Peter Lee, Microsoft Research corporate vice president, “truly magical.” But are they talking in terms of advancement of flawed translation tools, or actual workability, accoutablity?

This is partly how the magic works: Neural network technology first picks up parts of speech, the sounds of vowels and consonants, and much like how the human brain functions through interconnected neurons, the neural network learns and evolves, it becomes more sophisticated, until, perhaps, it can understand a drunk Irishman speaking animatedly, and idiomatically, to his Thai bride who speaks in a strong colloquial northern Thai dialect. That would be magical. It’s more likely that the new translation tool will be effective when interlocutors are speaking related languages in a kind of Tarzan and Jane repartee simplicity.

Even if the tool is used at something like a business conference, in which well-spoken ladies and gentlemen speak slowly to each other, what happens as soon as one idiom in Finnish is miscommunicated to a Mandarin-speaking Chinese person? There is a great risk in putting your trust in translation technology. Improved voice recognition technology might be able to better understand our phonemes and morphemes, but how will it ever interpret the infinite complexities and disorienting nuances of meaning?

It will be interesting to see just how magical the new Microsoft Skype Translator is. Testers of the product have yet to post any reviews. To not be too harsh, translation tools do have value, and can be useful at times, but it’s likely we should not put too much trust in their efficiency, especially when important matters are being discussed, or typed. They don’t translate as a bi-lingual human does, they make very basic approximations. Translation tools have a long way to go; they are what they are…an expression itself that is untranslatable to many languages.

Photo credit: Daniel Zana via photo pin cc