Google Translate is not there yet.

In an article “Google’s Computing Power Refines Translation Tool” in today’s NYTimes  I read this:

“In a meeting at Google in 2004, the discussion turned to an e-mail message the company had received from a fan in South Korea. Sergey Brin, a Google founder, ran the message through an automatic translation service that the company had licensed.

“The message [in Korean] said Google was a favorite search engine, but the result [in machine translated English] read: ‘The sliced raw fish shoes it wishes. Google green onion thing!’”

Well, I said to myself, I should check this out, and see just what progress in machine translation Google has made. I’ve been an admirer of the Google search engine from the very beginning, and have always thought that these people, Sergey and Larry, and all the others up there at Mountain View, could do whatever they set their minds (now money, loads of it) to.

So in order to check out this report of their latest achievement I went to their Google Translate web site with an excerpt in Russian from the Chekhov story, In the Ravine, and requested a translation.

No problem. In seconds, actually less than than seconds, I got a complete translation of the passage, some 132 words in the original Russian becoming some 146 English words in translation.

Right away I looked to see if there were any of this sort of thing, “sliced raw fish shoes or green onions.” I really wanted from Google, the company I so admired, that there not be any. But almost immediately I read this (sentence?):

“When passers-by asked what this village, they said, — This is the same thing where the clerk at the funeral of all eggs eaten.”

And there they were, not fish shoes but eggs eaten.

Not too different from the translation from the Korean. I wondered if I should send the Google “translation” of Chekhov’s story to Sergey Brin, reminding him of the “sliced raw fish shoes” of 2004 which had led him to the push towards better translations. And after all, Sergey was Russian, Russian was his first language, and he would be able to quickly grasp the  machine’s failure to translate the Chekhov passage.

And if he were to see this what would he then say about the progress of Google’s machine translation services?

OK, this was a literary text, and as we all know literature is notoriously hard to translate, usually demanding long hours even from someone who knows both languages well.

My conclusion? So far Anton Chekhov’s language, unlike perhaps the language of math and science, maybe even the language of the social sciences, (not to mention the play of the game of chess as we saw in Big Blue’s victory over Kasparov), has resisted being translated by a machine. So far in regard to literature texts in general, taken from this one example of Chekhov, the score is Man 1, Machine 0.

Will it ever happen that the Machine wins? That, say, original poetry coming out of the machine in the new language, will accurately convey the author’s thought and/or image, much as it was in the original?

I would say probably not, unless somehow we can eliminate all individual differences, and that is not happening now, and will probably never happen. In fact, our individual differences are only growing as we find out more about ourselves.

Finally, a few explanatory notes on the Chekhov passage in question:

Here is the original Russian:

Когда прохожие спрашивали, какое это село, то им говорили:
— Это то самое, где дьячок на похоронах всю икру съел.

Here is Constance Garnett’s translation:

“When visitors asked what village this was, they were told:
‘That’s the village where the deacon ate all the caviar at the funeral.’”

And again, here is the Google machine translation:

“When passers-by asked what this village, they said:
—This is the same thing where the clerk at the funeral of all eggs eaten.”

What do you think? Has the machine translator done anything right in conveying the Russian into the English? Probably not, — or, if anything, very little. Probably only when the Russian word and its position in the sentence corresponded exactly to just one English word and sentence position, such as Когда, when, and где, where.

Otherwise, when there were a number of possible translations, or at least word meanings, the wrong ones were chosen, — passer-bys instead of visitors, clerk instead of deacon, eggs instead of caviar. And I need not even mention the errors of grammar and syntax, all leading inexorably (?) to the nonsense sentence in the result.

So Sergey, we wish you well, but perhaps you ought to show a little humility in regard to the ‘progress” of Google Translate. You’re not there yet, and there is some question, at least as far as the translation of literature is concerned, that you ever will be.


Explore posts in the same categories: Idle Thoughts

Comment: