From: Rick DeNatale <rick.denatale@...>
Date: 2006-08-31T08:18:00+09:00
Subject: Re: [ANN] How to spy on the Japanese Rubists

On 8/30/06, Hal Fulton <hal9000@hypermetrics.com> wrote:
> Paul Robinson wrote:

> > Why is machine translation so bad anyway? It's only changing one set  of
> > vocabulary and grammar for an equivalent set - it would seem that  the
> > biggest problem with most tools out there is ridiculously small  foreign
> > language dictionaries and tiny grammar rule sets. Very odd,  and
> > probably quite easy to fix.
>
> Not to be unkind, but that's very naive. I can understand people
> thinking that way... but only until they've studied the field in
> detail, or better yet, tried it themselves.
>
> So try it. If you fail, you'll have learned. If you succeed, you'll
> be rich and famous.

There'a a famous story about an early attempt on machine translation.

Sometime in the 1960s the US Air Force funded a project for
Russian/English translation.

They decided that good test cases could be obtained by taking famous
quotations, translate them to Russian and then back again.

Two of the tests were:

"Out of sight, out of mind."
and
"The spirit is willing, but the flesh is weak"

and they came back respectively as:
"Invisible idiot"
and
"The vodka is strong, but the meat is rotten."

T'aint just a matter of grammar and vocabulary.

I've been trying to program my meatware to do NL translation for 40 or
more years with limited success.

-- 
Rick DeNatale

My blog on Ruby
http://talklikeaduck.denhaven2.com/