The Danger and Impossibility of Machine Translation

The Danger and Impossibility of Machine Translation

If you have ever played around with Google Translate (Yahoo翻訳、Excite翻訳、etc) you know that it is far from reliable. So I was very surprised when I heard, through a friend who attended the JAT presentation  “Does Translation Pay the Bills?”, that machine translations and native checkers are quickly replacing actual freelance translators. This is ridiculous. If you have ever tried to translate anything highly technical, you know that Google doesn’t cut it. But, I am sure they are not using Google. So the question is:

How far has machine translation come?

The answer, I have found, is not very far. Machine translation is light years away from where it was when the internet was first getting started and everyone was laughing at the kind of things Babelfish came up with. But, it is far from perfect. The general consensus in the market right now is that, even if you buy the most expensive software out there, the product will still be inferior to what a native speaker can produce. The reason is that language is just incredibly dynamic. Take the word 対応 for example. The dictionary has this as “correspond to”. However, it can function in an English sentence as “for”, “support”, “handle”, “tackle”, “address”, or (my favorite) NOTHING. There are times in translating when a word in the original simply has to be omitted. It happens with words like 対応 in Japanese and with pronouns in English. While it is perfectly acceptable in English to say “I put my homework in my bag”, it would be ludicrous to say   私は私の宿題を私のかばんに入れました. Just not necessary. Yet if you Google Translate the previous sentence, you come up with an odd variation. 「私は袋に宿題を置く」This shows a different problem: Lack of understanding of context. Any beginner student of Japanese would have been able to handle that sentence. Yet Google decided to use the more common translation of each word (袋 and 置く).  This is the heart of the problem. Translation software cannot assume it knows what you are talking about (as translators do). You have to spell it out. Oddly enough, if you take the ludicrous Japanese above and go J>E, you end up with “I was put in my bag with my homework.” Closer. Take out the 私は and you get: “I put in my bag my homework.” Now we have the second core issue: Word order. Google Translate has obviously been programmed with the correct STPOQV word order. However, it can’t always understand how that correlates to English sentences. It also doesn’t understand when someone breaks grammar rules. It can’t assume subjects correctly when they aren’t there. I shutter to think what it would to with languages that have a truly flexible word order. How can it know whether the beef ate the man or the man ate the beef?

I am not saying that machine translation will never be wide spread. As I learned, it already is. But one thing is true. It will never replace translators, particularly in one very important field:


You cannot sell machine translations of literature. (That is a bold statement but let me elaborate.) MT is enough to get by. You get the point of what you are reading and that is why it is common in legal and science translation. But people read novels for the experience. They want the story and the art of the language transferred into their own. Their are countless stories of beautifully translated phrases that became as famous in the target language as they were in the original because they were delicately re-crafted to evoke the same feeling as the original with different words. 「君の瞳に乾杯」has to be one of my favorites. “Here’s looking at you kid.” Same feeling, different words. That is what literature (in that case cinema) needs and that can never be done by a computer.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s