Google demonstrates phone that translates text

One reason why machine translation doesn't work on Japanese is all the missing information. Sentences commonly have no subject. Most nouns have no plural. Tenses are fuzzy. European languages, including English, need number, gender, tense and identifiable subjects and verbs to produce grammatical sentences. Since such information is not provided explicitly in Japanese, the translator has to supply it by analyzing the context, reading between the lines, or just guessing. People still do those things infinitely better than machines.

Ranger -

Mmm, they aren't called 'Words Dictionary' or 'Pictures Dictionary' either....

Folk get confused because kanji has the same form in the singular and the plural in Japanese. Usually when foreign words are used in English, they keep their grammar rules unless the speaker doesn't know what they are or the word is used often enough to be assimilated into English thus taking on English rules of how to form the plural. Just because the plural form is the same as the singular form, it doesn't mean the noun is uncountable.

If you're still confused, let's take the good old English word 'sheep'. In both singular and plural, it's 'sheep'. But you can still count the little blighters.

paulinusa -

I think it depends to some extent what kind of publication you're reading, and what you're reading about. The written word has the advantage (usually) over the spoken word in that you have the kanji to help with the meaning in words like ame/candy (飴) and ame/rain(雨). But then if it's a report about the stuff falling from the sky it's more than likely not candy, just like with hashi if you're eating your dinner with it it's probably not a bridge, so I don't think the kanji really make all that much difference. If they did people would be writing everything down instead of talking. (I know people do sometimes write kanji in the air in a conversation).

I don't know what it is with machine translation. Other languages seem to be turned into something that is at least reasonably understandable, but for some reason Japanese seems to blow a fuse so that more often than not total jibberish is produced.

I can spreak Nihongo o.k. but can't read at all, so I know that the spoken word sometimes needs more context and emphasis for understanding. An obvious example: ame(candy)/ame(rain). My previous question about Kanji being specific was a bit clumsy, so allow me to ask in another way. Let's say you pick up a Japanese newspaper and you're a native level reader. Is what you read (a) open to interpretation and in need of more context and emphasis like the spoken word, (b) more precise but still needs some context and emphasis, or (c) very precise. Or is the key word here interpretation, not translation, and computer software still can't process the written word like the human brain?

OK Cleo, then why aren't all those dictionaries called "Kanjis Dictionary"?

Please don't make translation software too "perfect"... Some of us still have to earn a living...

kanji is a noncountable noun, I believe. Same with sake (the drink). Just being a dweeb here.

Nah, not the same at all. You can have a jugful of sake, but not a jugful of kanji. Or kanjis. If you can point your finger and enumerate, it's countable.

酒 = one kanji 漢字 = two kanji (or kanjis)

Though if you have enough jugfuls of sake it gets a lot harder to count the kanji (or kanjis).

Hey, I can be a dweeb too! ;-)

This would be great if only international roaming data fees weren't so high... Cloud computing means the necessity of a fast data connection - something that is extremely expensive when overseas, when this software is most needed.

kanji is a noncountable noun, I believe. Same with sake (the drink). Just being a dweeb here.

quick, buy kyodo news a thousand of them

Well, I doubt that there is an official dictionary that manage slang. Some times Im forced to go to Urban dictionary.com when dealing with english, because Dictionary.com is not eough for deal with informal english. And dont get me started with words that just dont have an english equivalent word or phrases where the meaning is lost without a cultural context, like modern jokes or old sayings. But the real problem with translation programs is grammar. Most programs get stuck because there is no space betwen words in japanese and guess where one word start and end need a strong "context recognition" program. I think that for japanese the big progress can be get in every ketai a kanji recognition program. That can make dictionary searches waaay faster, in place of crawling with find a kanji by radicals when we are reading the newespaper. A translation tool for japanese, is still far away on the future. But a free portable dictionary for kanjis can help.

Great! How does the phone translate an intelligent question like "Ima, doko?" into English?

A specific character means a specific thing. Please correct me if I'm wrong.

You're wrong. :-)

For example, a character like 通 (the first character in the word meaning 'interpret) can mean (or rather, be used in words that mean) to come and go, commute; to pass through without stopping; to be an expert or connoisseur; currency; ventilation; trade & industry; sum total; a telephone call; and a bowel movement. In real life it's usually quite clear what's meant, but machines and software still have a long, long way to go before they can read context.

like some strange form of garbled poetry

Yes!

I'm not complaining though, it's the lack of half-decent machine translation that keeps the wolf from the door. Building the Tower of Babel was, in retrospect, a Good Thing.

Good points. Thanks

paulinusa, some kanji might have a general meaning for it, but some might have different nuances. You have to keep in mind that when it comes to translation it's not just about each individual word (or kanji). Grammar comes into play, too, not to mention expressions and context that software might not be able to pick up and translate into natural English (or other language).

And it's hard to figure out why kana or kanji would be so difficult to translate at this point. We're not talking about handwriting. A specific character means a specific thing. Please correct me if I'm wrong.

Translation tools are not reliable if what you want to translate goes beyond a simple phrase or sentence. Anything more complex and it's literally "lost in translation". My students submit speeches that have obviously been translated through a website such as Google or Yahoo and I can't understand what they are trying to say. Reading their own limited English, as flawed as it may be, is a lot easier than trying to read the garbled mess that comes out of these translation tools.

Two years ago, a guy came to my house to tell me about the nearby construction. He handed me a map with the Japanese text translated into English by Google's server. It was almost completely incomprehensible but the funniest thing I'd read in a long time! Hope they've improved since then.

I agree. Translations of Japanese websites into English on Google (or other sites for that matter) has always been problematic. What usually appears is like some strange form of garbled poetry.

No machine translator works, perhaps english to spanish or english to french but english to Japanese never works, it gets about 20% of it correct for simple sentences.

It translated “Fruhlingssalat mit Wildkrautern” as “Spring salad with wild herbs.”

google servers have no better job to do?

hahaahaha, it is made for Americans who do not undersand foreign languages

Great, let's see if it work for more complicated things like Chinese and Japanese Kanji.

Thanks goolge for translating the my text and.