After decades of innovation, we now have pocket-size computers that are more powerful than ones that used to take up whole rooms. But despite all the hardware advances we’ve made and the decades of research in computer natural language processing, Google translater still gives me this translation.
何しているの？ – What are you?
Hmm… nice try. The point I’m trying to make is that translation is and always will be an art and not a science. Especially with languages that are so different from each other such as English and Japanese. There will always be different interpretations and decisions to make on defining what the “best” translation is. It’s like trying to piece together the same lego set but with entirely different pieces. You can get some things to look similar or even almost identical but you’re gonna have to improvise on places where the pieces just don’t match.
However, there is an easy benchmark for determining how good a translation is for language acquisition: “how does the translation help you learn the language”? As adults learning a second language, it behooves us to learn new words by translating to our native language. You can save a lot of time by memorizing the word 「友達」 as “friend” rather than learning how a baby might from scratch. However, with longer sentences and more abstract concepts, translation can often be more of an hindrance than an aid depending on how you go about it. Ideally, translations should serve as a stepping stone to learn the core concepts with the aim of doing away with translation altogether. In most cases, this means going for the most literal translation.
For example, which of the following is a “better” translation?
食べてもいいですか？
1. May I eat it?
2. Even if eat, is good?
In the first translation, the translator (in this case, me) made a lot of decisions to try to craft what I thought was the most natural translation (“may” vs “can”, etc). If I was hired to translate a Japanese movie or text, #2 would be a terrible choice. But as a language learner, looking at translation #1 doesn’t help me understand anything about the core concepts that can explain other similarly structured sentences nor does it help me internalize the language for my own use.
食べなくてもいいですか？
Even if don’t eat, is it good?今から行っても間に合わないよ。
Even if go from now, will not make it in time.
Some concepts just don’t translate into English very well at all as you can see in the following sentences.
そのようです。
It is that manner/appearance.朝早く起きるようにしています。
Morning doing in manner/appearance of waking up early.やっと泳げるようになりました。
Finally became manner/appearance of able to swim.この映画は、もう見たような気がします。
As for this movie, feeling of manner/appearance of having already seen.
But it doesn’t matter how bad the translation is, as long as it gives you an idea of the intent of the original Japanese and helps you conceptualize and internalize the concepts. That’s really the “best” translation. That also means my translations are just as bad if they don’t work for you, so take translations as just a hint for you to figure out the meaning on your own. When you can “feel” the meaning without quite being able to express it in English, that’s when you know you’ve truly learned it.
Agree 100%, which made me skeptical of this, as cool as it looks: http://www.telegraph.co.uk/technology/news/6493869/NEC-unveils-Tele-Scouter-translation-glasses.html
When so much of a language is contextual, as it is in Japanese, it just seems to me nearly impossible to develop a device or program to fluidly and consistently translate it.
Not without some sort of sophisticated AI at least. At that point, we may as well kiss humanity goodbye. 🙂
I always wonder why translation hasn’t gone more web2.0 yet. If you see the “What am I” mistranslation you should be able to flag it, or even better, offer an alternative. Enough flags/alternatives should lead to a correction. But that takes a lot of resources, which brings us to another thing: free online translators aren’t exactly computation at its best. They have to handle thousands of connections a second, they’re going to have to make some sacrifices for the sake of efficiency…
In google translate, you can contribute a better translation. Don’t know what that does though. Google has tens of thousands of servers in huge and sophisticated data centers and can search billions of websites in under a second. I don’t think efficiency should be an issue.
A good translation program would not only have more-literal or less-literal translation capabilities, but be able to either sense the formality of the original text or designate the level of complexity and formality in the translation for things such as word choice and sentence construction. Any of the rules we use to parse and construct sentences can be used by a computer, it’s just a matter of spending enough time… lots and lots of time, to identify each specific word and phrase and its relation to others in different contexts… to make those rules. But it doesn’t sound like a terribly interesting project!
In the meantime, I think the more useful teaching material does try to use the literal translations, even if it may make some readers uncomfortable. The illusion of comprehension vs its actual reality.
> ねこであります。
> Dearimasu cat.
> Any of the rules we use to parse and construct
> sentences can be used by a computer, it’s just
> a matter of spending enough time…
Actually, a computer *can’t* use all the rules that humans use to understand language, because many of the rules rely on understanding the actual meaning of what is being said, and knowing what the words and phrases represent, in a way that computers can’t do. Computer scientists use the term “AI Complete” for this kind of problem: it is generally believed to only be really solvable by complete AI, that is to say, by making the computers as smart as people.
Tae, I agree that laying things out word-by-word can be useful for a language learner, but I would hesitate to call it “translation”. It’s really more like an interlinear. For a language learner this is useful, but the use case is altogether different from that of a real translation. A translation is intended to communicate, to someone who doesn’t know the original language, the same ideas that someone who does would get from the original. In the use case you’re talking about, the goal is different. Rather than conveying the same ideas as the original text but in the target language, you’re instead primarily conveying things about the way the language is used (grammatical patterns, nuances of the meanings of words, and so on). This is useful, but not in the same ways or for the same reasons as translation.
Of course, the ultimate translation problem is song lyrics, where you’re expected to preserve the original meter as well as the meaning and, if possible, the rhyming scheme as well The most impressive example I have ever seen is Hedge’s beautiful English translation of Ein’ feste Burg ist unser Gott, but translating between German and English is not nearly as challenging as, say, translating between Japanese and English.
But if I said interlinear, nobody would now what I was talking about (include myself). The art of translation for sake of language learning is to use a less literal translation for parts you are already familiar with (for example, は and が) so that you can focus on the the parts you’re trying to learn.
Translation can be a difficult thing. Dictionaries and computer translation devices have come an extremely long way in the past decade, but will never be able to become 100% accurate because of the contextual nature of language.
I feel you might get a laugh out of this one:
http://translationparty.com/
I guess the real test of an automated translator that intends to output something natural is whether it is adequately reflexive. Google translate is not, in oft hilarious ways. Of course, because Japanese relies on so many “intuitive” subjects and objects and even implied verbs it’ll never work. Well, short of dumping out a little piece of situational clipart of the mising subjects when translating E->J. Actually, that could be even more amusing… 😉
求めよ、さらば与えられん==>Ask and you shall receive Farewell (笑)
Your critique of automated translation is very in line with the type of thinking and language philosophy which I attempt to impart on my own blog. Without taking this too far, I believe there is a critical moment in the learning process when we challenge our inner existing knowledge (for example, one’s native language in our head) with expression, such as translation or interpretation, forcing us to reflect on and evaluate the real “essence” of things.
@Jonadab the Unsightly One: “Of course, the ultimate translation problem is song lyrics, where you’re expected to preserve the original meter as well as the meaning and, if possible, the rhyming scheme as well.”
I have been translating pop songs for a while, and recently organizing and expanding on my “interpretations” in English little by little at foreignlanguagemusic.wordpress.com/. While I do consider the meaning of the song as a whole, I try to avoid otherwise “creative” interpretations, and I pretty much ignore poetic devices like meter and rhyme. My goal is to make learning the original lyrics easier by offering a rough translation and grammar/vocabulary notes. As Dr. Alexander Arguelles mentions on his website regarding reading Great Books in the original text: ” if works are worth reading and rereading, then surely they are worth reading as they were written.”