Machine Translation

[Michael Everson]

On 11 Sep 2009, at 11:13, Martin J. Dürst wrote:

> Good dictionaries usually contain several entries for each word in  
> the source language; what you do in that case is lookup, not  
> translation.

Rendering of one language into another is translation, whether it is a  
word or phrase or sentence.

> Short phrases could indeed be the place where machine translation  
> has its best chance currently: memory is cheap, phrases can  
> eliminate ambiguity, and if they are short, they are suited to  
> lookup without much processing.

Yes, currently. And the technology will grow and grow over the next  
few hundred years until Star Fleet has its Universal Translator....

>> I have several on my iPhone, and Mac OS
>> ships a Systran widget with its OS X Dashboard. This is machine
>> translation.
>
> Mostly just lookup, isn't it?

No. You can have it process entire sentences.

The red sun was high, the blue low.
-->
Die rote Sonne war hoch, das niedrige Blau.

What do you have in your pocket?
-->
Was haben Sie in Ihrer Tasche?

I assume the first one ought to have read "die niedrige blau", but the  
sentence is fairly literary. (It's science fiction.)

There's nothing wrong with the second sentence however.

I still don't see how our subtags, which describe language varieties,  
ought to be dragged into this arena. Compare
<lang=de>Was haben Sie in Ihrer Tasche?</lang>
and
<lang=de-machine>Was haben Sie in Ihrer Tasche?</lang>

There is no linguistic difference to be tagged. There was a process  
involved in getting to the second of the two sentences, but that is  
something else again.

Michael Everson * http://www.evertype.com/