Published: Monday, May 15, 2000

Stump the geeks

Q: I am writing a report for school. We have to think of a problem and find a solution. My report is on language barriers. As you know, there are hundreds of different languages spoken in the United States alone.

My solution to the problem is a Universal Translator (like on "Star Trek"). Is it possible to make a translator that could take any language, translate it and send it out in the native language of the person using it? For instance, if I speak Spanish and the person I was talking to spoke French, could a device be made to take in Spanish, translate and send out French?

A: As a teacher, I am hesitant to do your homework for you. But as a geek, I am flattered that you asked and can't keep my virtual mouth shut.

I suspect that you are testing me to see if I still subscribe to Wired (yes, I was an early subscriber back when Wired was still k3w1 [cool], and yes, I am the kind of geek whose attention deficit disorder keeps him in the state of "meaning" to cancel now that Wired is tired, but who never gets around to it.). If you look at the May 2000 issue of Wired or visit the Wired Web site for the same issue http://www.wired.com/wired/archive/8.05/ you will find the best collection of easily accessible articles on Machine Translation (MT).

The short news on the subject is that MT doesn't work but that MT is getting better at not working. MT, my droogs inform me, will always be doubleplusungood cuz folks mess their language up a bit right regularly. If you were to say "Pound a pound of Polish polish," your translator might light up like a Christmas tree or maybe blow its top. You begin to see the problems, I hope.

Idioms, slang, context-sensitive usage and the very slippery nature of language are the deadly enemies of MT. Speaking the language into a MT device doesn't crack these problems. Spoken language adds other interesting problems to the mix.

Still, some translation is better in many cases than no translation at all. Word and phrase translators do a fairly good job of getting you to the train on time, to dinner and into the WC (or restroom). Companies like Franklin and Lingo make cute little helpers that will get you through some sticky places when traveling. Several companies make dictionaries and simple translators for the PalmPilot see http://www.palmgear.com/ and search for "translation" and the languages that you are interested in.

CPen, http://www.cpen.com/ works by dragging the pen tip over words as if you were highlighting them, but it would not likely work on speech. (You might get some interesting intercultural interactions by trying to drag a CPen over the lips of an Italian policeman, though.)

A long-time Net favorite is http://babelfish.altavista.com , which can give some very funny results. Purely for scientific purposes, I'll reproduce a sample babelfish session here.

First, a stanza of that great American song, "John Henry":

Well John Henry said to the captain,

I'm gonna take a little trip downtown

Get me a 30-pound hammer with that 9-foot handle

I'll beat your steam drill down, Lord God

I'll beat your steam drill down

Using babelfish, I translate that stanza into French:

Henry bon de John a indiqu au capitaine,

je suis prise de gonna que un petit centre ville de voyage

m'obtiennent un marteau de trente livres avec ce traitement de neuf pieds

je battrai vos exercices de vapeur vers le bas, seigneur God

je battrai vos exercices de vapeur vers le bas

Now let's go back to English:

Good Henry of John indicated to the captain,

I am taken of gonna that a small centre town of voyage

obtain to me a hammer of 30 books with this processing of 9 feet

I will beat your vapor exercises to the bottom, lord God

I will beat your vapor exercises to the bottom

A cheap shot to be sure, but one that points up the problems inherent in MT while providing a little amusement.

What you might try instead is a mix of human and machine translators. Like the United Nations, you could have near instant translation via a service available through your wireless phone. I think of such a business as Psychic Friends Network with a purpose. I demand a note of credit if you use this for your class. I demand stock if you set up a business to exploit this fine idea.

Until then, let me refer you to my favorite translator -- the Lamerizer, which mixes upper and lower case, substituting numbers for certain letters and new varieties of phonics. (Serious computer folks make fun of the "elite" kids who understand this argot by calling them "lamers.") David Barberi has made life simpler for would-be lamers with his Lamerizer at http://www.metalab.unc.edu/dbarberi/

The Lamerizer renders the John Henry stanza as:

WE|L j0hn H3nry sa1d to th3 CAptAin,

1'M gonna Tak3 A |1tT|3 tr1p Downtown

g3t m3 a ThiRty pounD hamM3r w1th thAT nIne Foot HAndl3

i'l| b3At Ur st3Am DrIll Down, |0rd gOd

1'|L b3at Ur stEaM DRIL| doWN

(Note that the same word doesn't translate the same way each time it occurs. d3AL With 1t, U dUpez 0f th3 ON3 wor|D gov3rnM3nt!)

Paul Jones director MetaLab, UNC-Chapel Hill E-mail questions to stumpthegeeks@nando.com. Please include your name and phone number.