Google Translator Toolkit

[chinopinyin]

Google has launched a translation toolkit http://translate.google.com/toolkit

The most promising aspect is what they call Translation Memories: when you are translating Google first searches in a database of human translations. As more people use the toolkit the database improves http://bhopu.com/2009/06/google-translator/

Are problems with machine translation going to be over in the near future?

Has any of you used it?

[tooironic]

Old news. Google Translator Toolkit is not so much a "machine translation" tool but a Computer Assisted Translation (CAT) software (although it obviously uses machine translation capabilities). Translation memories have been in use for a few decades. Currently the most popular ones are TRADOS , Deja Vu and Wordfast. I haven't seen Google Translator Toolkit catch on in the profession as of yet, but it will be interesting to see how it develops.

[HashiriKata]

Are problems with machine translation going to be over in the near future?

No, MT will get better in many well-defined areas but many other areas will still need the human touch. Things that needs to be translated are after all products by humans, and as long as humans are still humans and not machines, problems with MT will always be there.

[Lugubert]

MY translation memories by TRADOS are useful to me because they retain MY use of words for MY respective customers. TRADOS also lets me work in MS Word using my several hundreds of since many years ago (some remaining from identical WordPerfect keys from their 4.2 or something) established and mentally automated keyboard macros and similar shortcuts. I don't think Google can match that for still years to come. For example, I have to know which customer for medical analyses uses test strips and who only have test sticks, and what the difference is in Swedish. How would Google manage that distinction?

As more people use the toolkit the database improves

It won't improve too much, but will rather get immensely confusing because of the many errors introduced by pretend professionals and plain amateurs. I sometimes feel that it's more common to find a Council Directive EEC number still as ...EEC in Swedish on the Internet and in any printed material rather than the correct ...EEG. An "Appendix II" will almost always be found untranslated instead of as "tillägg 2" etc. etc. unto infinity.

To create really useful quality filtered stuff, Google should really pay me a humongous consultancy fee. But, sorry guys, I'm retired and too lazy.

For the general MT quality problem,

注文殺到

Google translation: [swedish] Chumun omedelbar oro (Chumun immediate worry)

How does zhùwén get interpreted as Chumun?

Babelfish: The annotated text kills (Who'd want to employ that company?)

The Chinese was cut from a translation company advertisement. What does it mean?

[chrix]

Lugubert, I'm not following where you got this 注文殺到 from, but in Japanese this would mean something like "(a lot of) orders are rushing in". Chumun is the Korean reading of 注文.

[Lugubert]

Like I wrote, it was displayed on an Internet advertisement for a translation agency. I saw no Japanese grammar particles, so I assumed it was Chinese. Shouldn't there for example have been a を following 注文 in Japanese? Either way, the Korean reading from Google looks funny.

[chrix]

no. 殺到 is intransitive, and also it is a set phrase so you don't necessarily need any particles here. Anyways, it's Japanese all right. If you google for it, you get a lot of hits on this or that product being in high demand...

As clever as I find the advertisement (without seeing it) I think the point is moot, if this phrase were to appear in a longer text it would be clear it was Japanese and hopefully Bablefish and google would get it right too. You could try this again, with the phrase embedded in a carrier sentence.

Here's an example, that brings up another point about the particles: in headlines and the like, you can also get rid of them, so consider the following article about Calendars made of French cheese being popular with women worldwide:

Headline: チーズのカレンダーに注文殺到、ファンの女性がアピール

from the text: このカレンダーには世界中から注文が殺到し、飛ぶように売れているという。

You could feed those two into the tool and see what happens (but as I said, there are also examples where 注文殺到 is used as a set phrase where it's not a headline)

Edited January 9, 2010 at 02:21 PM by chrix