Webpage Javascript Translation Project

Have a look and please take the code you want. Suggestions would be great.

http://www.geocities.com/redragon3696/jtet.htm

Posted: Mon Apr 17, 2006 4:55 am Post subject: Post Rating: 0

When I click translate, I get an error. In other words, it doesn't work.

Posted: Mon Apr 17, 2006 4:57 am Post subject: Post Rating: 0

Is the program supposed to be working?

Posted: Tue Apr 18, 2006 2:38 am Post subject: Post Rating: 0

i'm sorry, but what does this have to do with fansubbing? scratch

should this be in the tech or general forums?

Posted: Tue Apr 18, 2006 3:32 pm Post subject: Post Rating: 0

Posted: Tue Apr 18, 2006 3:35 pm Post subject: Post Rating: 0

Posted: Tue Apr 18, 2006 6:12 pm Post subject: Post Rating: 0

I tried it on Firefox and IE, with Japanese characters and with Romaji. When I click "Translate the text area?" Nothing happens.

Posted: Tue Apr 18, 2006 6:46 pm Post subject: Post Rating: 0

I tried it just now. Works for me! I think if you were to type in just one word, there's a chance that the word won't be in the database. If the word can't be translated, then nothing happens. For example I entered 武蔵 (Musashi) and it didn't show anything. But when I entered 新選組 (Shinsengumi, where "shinsen" means new compilation, and "gumi" means group), it spit out "Gumi Group" which to me it didn't help any other than saying it is a group of gummies??? LOL! Lots more work to be done! For one, there should be some sort of response to kana characters that were not in the database, instead of just ignoring them.

Interesting project. Based on my experience and instincts, I don't think you'll be able to just simply parse words and phrases, match them in a database, and then spit it out. That's why systems like Babelfish are horrible at this.

If you think about it, most Japanese "speed" courses teach you to memorize phrases. Think about your brain being the database, and you've entered hundreds of Japanese words and phrases into it. And then in the real world, someone comes along and says something to you that just doesn't match, such as using a certain dialect or shortcut. You can't process the new information because you've never applied a grammar to the database. By the description that rent1999 gave, that's basically what is going on in his program. Taking Japanese words and "matching" them up with English words is very easy. Designing two grammars (one for Japanese and one for English) using computer algorithms and then interfacing the two grammars with some form of conduit is the challenging part.

Also, a lot of these databases attempt to translate EVERYTHING - especially proper nouns like names of people and places. To me, proper nouns should remain untouched and instead just translate to romaji. The challenge is to tweak the grammar into figuring out what kana are parts of proper nouns. Early on, Babelfish failed at this. But recently, they've improved a great deal.

In college, I took two years of linguistics, focusing mostly on language grammar structures. I did this when I was designing my own computer language (for compiling to machine language, but later on use it for artificial intelligence), and wanted to design the grammar of the language so that parsing it was efficient (we didn't have gigahertz processors in my day.)

--- groink