Wednesday, January 17, 2018

Google Translate (谷歌翻译)

One of the biggest fears for people like me, whose job involves using my language abilities and who has invested a lot of time and effort into learning languages, is the advent of a universal translator or similar artificial intelligence device, that's able to totally wipe out all need for any translation work or humans with language abilities. For years (after the Iphone was created, Internet replaced newspapers, Netflix replaced movie stores, etc.)  language translation has been the next technological advancement in the making, may experts warning that it is coming sooner or later. It's definitely something I get concerned about sometimes, but don't lose too much sleep over.

Google Translate, at least for Chinese and Japanese, is getting better, a step up from back in the day when their subjects and objects were arranged into sentences that made no grammatical sense, and even my sister, who's not that well versed in Chinese, was making fun of it. When I had some renewed misgivings about my language abilities today I plugged in some English sentences and had a mild heart attack at first due to the preciseness of the translations that it spat out, at least for rather simple sentences. They were grammatically correct and pretty broad in its vocabulary list, even including some common idioms they could decipher and give the right definition, like "it was raining cats and dogs" wasn't literally translated to cats and dogs but as "it rained heavily." Smart work, Google. The kicker, though, and what gave me a little bit of hope for my mere human language abilities, is more nuanced words that have more than one definition and needs context to be interpreted. For example, plug in these sentences to Google Translate:


1.) I have had it up to here with you - (Google misses the "I'm angry!" nuance)

2.) I might turn on him later - (Google misses the "betray" nuance of this sentence due to all the words being common words and "turn on" usually used like turning on a television).

3.) I had the hots for him. - (Google misses the "I really like him!" nuance of hots.)

4.) I tore him a new one. (Google misses the "severely criticized" portion of this).


Google can program in a bunch of unique idioms in all languages (although there are still some that it misses, like "waiting for Godot" or 犬猿の仲, meaning fierce enemies), but it still can't detect the phrases or expressions that use ordinary words that just have a different meaning, like a robot Google has to go through all the variations and possibilities but ultimately choose the most likely meaning of "turn on" like "turn on a TV" because that gets the most hits, while a human would able to read the whole sentence and know immediately that it contains a secondary meaning. Same in Chinese and Japanese, except so much more Kanji characters to confuse a machine (don't get me even started on trying to read a handwritten Chinese paragraph with scriggly writing and shorthand!) 
Google Translate, though, admittedly is great at typing in a single word or expression and getting different translations for that one word. Although that might also be unhelpful because which one of the 3 translations it spits out for "fire" is the right one? Too many variations sometimes can bog down the whole exercise.
Google Translate still has a way to go. But do other translators have better success? Apparently, Google Pixel phones now have a universal translation device where one can talk into the phone and the phone spits out the translation, but it has trouble picking up what the speaker says due to the various intonations (Chinese and its 4 intonations!) and accents people have. Which, by the way, is a problem that humans have to, deciphering what the other people are saying, even if you know the language, sometimes I don't even understand another people speaking English if it's accented.


Ultimately, I do think one day there will be an "Instant Translator" that can translate exactly what one says immediately to facilitate conversations. But perfecting it can take years, and translating a language does not capture the nuance of a language quite as well as what native speakers can do. Each language is its own different world, and there's a beauty to each language that can't be just turned into a mechanism and made practical by a robot, although it can get pretty close. For now, I'll just rest on my laurels that my language skills are still pretty marketable........until they aren't. When there comes a day when I see this instant translation device being used on the street and among all my friends, I'll have to admit it's over. (And really, at that point the human race might have to admit defeat to robots and AI because they've deciphered when of the most basic things that humans possess). But until then, I'll still keep fighting! 파이팅 (doesn't translate correctly on Google Translate)

* Apparently this type of thing has a name, based on the cult movie Hitchhiker's Guide to the Universe, called "Babelfish."

Fantasize on,

Robert Yan

No comments: