Thursday, August 9, 2007

Machine Translation by Google

You may have noticed the little flags that now appear at the top of the right sidebar for this blog. Essentially, what this allows readers to do is to click on the flag of choice and get a sense of what the site says in the language represented by the flag.

I wouldn't call this "translation" per se, because, while this tool does give readers some idea of what the text means in another language, it does not have the necessary quality or accuracy that I like to associate with that term.

As a longtime user of CAT (computer-aided translation) tools, such as TRADOS and others that are common in the translation field, I have always been appreciative of the great advances in Translation Memory tools, but highly skeptical of automated/machine translation. With all of the automated translation tools I've ever seen, the translations are so nonsensical that it is not even worth using the tool in the first place, as the results end up communicating information that is wildly different from the source text.

For those who have never worked in the field of translation, here is a basic but important difference:

Computer-Aided Translation (CAT) = A HUMAN translator uses software programs to partially automate certain tasks. With many programs, the source text is divided up into Translation Units, which are stored along with their various translations, so that the next time they need to be translated, previous translations can be leveraged. CAT is extremely common these days, and many professional translators are familiar with such software tools.

Machine Translation (MT) = No human is involved, and the translation is performed entirely by a computer. Instead of a meaning-for-meaning conversion, this is frequently akin to a mere substitution of words in one language for words in another. Often, this results in unusable output. In many common phrases, such as, "Cat got your tongue?", MT programs typically render the information on a word-by-word basis into the other language. Obviously, that can make for some pretty entertaining material.

Most translators and interpreters have had at least one moment of side-splitting laughter at the expense of web-based MT tools. It's often quite entertaining to enter text into the "virtual translator" into the target language, then copy and paste the translation in again to do a back-translation, and sit in anticipation of the hilarious nonsense that is often produced.

When I first heard about Google Translation, I assumed it was just another of these programs. I had an opportunity to try it out the other day, when I came across some pages that mentioned the From Our Lips project in Chinese, but having zero proficiency in written Chinese, could not understand what the page said. When I used the tool to translate the information from Chinese into English, I could see that the site owner had essentially translated the call for submissions, and that the English version I was looking at was in fact a back-translation. While the translation by Google was far from perfect, I was able to get a general sense of what the site was saying. This also enabled me to thank the site owner.

So, I decided to add the tool to this blog temporarily, to give individuals from other languages a chance to try out the tool, just as I did with the site in Chinese, and especially seeing as how the project has already generated interest from interpreters in language pairs that do not include English, but who are still planning to submit stories in English. After all, the editing process will ensure that the stories are of a professional and publishable quality.

I may decide to remove it at some point in the near future, but for now, I thought it might be interesting for readers to test out on the site. I don't believe the tool in its current format is of much (if any) help to translators and interpreters in their professional work life. However, for the average layperson surfing the web who just wants to get a sense of what a site may say, it can be useful, as it was for me recently when reading information about this project in Chinese. Needless to say, though, it's still a far cry from a professional translation.


Nadine said...

Excellent! I plan to add this to my blog as well. I am a translator too, and while I despise machine translation, am glad to know about this for web navigation as you suggest.

Nataly said...

Glad to know it was helpful, Nadine, and thanks for your comment! :-)

Anonymous said...

Thanks for the explanations. I did not know the difference between "MT" and "CAT". As an interpreter, I never thought I needed to know, but I am glad I do now!

Mike said...

I know lots of translator friends who use LEC’s translation software ( I tried that and it is really incredible. The translation quality is better than Google. See the example below.

美国的经济很发达。(Meaning: The economy of the United States is prosperous.)

LEC: The economy of the United States is prosperous.

Google: The United States economy is a developed country.