Translation is hard

Thursday 5 June 2003

I found a link in my referrer logs from Peter Gloor's blog. It's in German:

Kaum zu glauben, aber es gibt immer noch Leute, die genug Zeit haben Ihre persönliche Home-Page aufrecht zu erhalten und einen unwiderstehlichen Drang verspüren sich im Web irgendwie zu präsentieren. Nun, was ich gesehen habe als ich neulich zwei oder drei Mal über Ned Batchelder's blog gestolpert bin, war ein völlig neues Erlebnis, dem auch ich nicht widerstehen kann. Die Zeiten der langweiligen, statischen Home Pages sind vorbei.

I was, of course, curious to know what Peter had said, so I used Babelfish to translate it to English. This is what I got:

To hardly believe, but there is to feel still people, the enough time your personal homepage upright too received and an irresistible urge in the Web to present itself somehow. Now, which I had seen as I recently two or three times over Ned Batchelder's blog are gestolpert, was a completely new experience, which also I cannot resist. The times of the boring, static Home Pages are past.

I know automatic translation is hard, and we've all heard the stories of famously bad translations. One story about an early translation system claims that "the spirit is strong, but the flesh is weak" translated to Russian and back came out as "the vodka is good, but the meat is rotten", which is actually pretty good! But this passage is almost nonsense. I don't know any German, and I got almost as much meaning trying to read the German as I did reading the "English".

In fact, the main word I want to know from the German is "gestolpert", and Babelfish didn't translate it as anything, passing it straight through!

I was going to say that Microsoft Word's grammar checker wouldn't even accept the English as grammatically correct, but in fact, it only minded three things in the paragraph (aside from "gestolpert"!), so even it is falling down on the job.

I don't mean this to be a typical "ha ha, translation sucks!" comment. I knew before I hit the Translate button that the translation would be bad. But I'm still surprised each time anyway. Even after 30 or 40 years of research, computers remain really bad at things humans can do easily, and we humans are so good at them, that even when we know the computers are bad, we keep forgetting just how bad.


Martijn 9:06 AM on 5 Jun 2003

Gestolpert means literally: tripped over, here meaning 'came across'....
Anyway you can have very funny results if you translate a piece of text back and forth a few times ;-)

Jim 10:36 AM on 5 Jun 2003

Hee! I had a similar experience with a referer from a blog in Icelandic. The referer was pointing to my modest Movable Type Trickle extension, and described it as "Snilld!"

For all I knew, Snilld is Islensk for "Pounds sand with the force of Thor's Hammer."

With no Babelfish for that particular language, I just asked on one of the geek lists I'm subscribed to, and somebody knew. "Snilld!" is like American "Cool!" or British "Brilliant!"

Add a comment:

Ignore this:
not displayed and no spam.
Leave this empty:
not searched.
Name and either email or www are required.
Don't put anything here:
Leave this empty:
URLs auto-link and some tags are allowed: <a><b><i><p><br><pre>.