Machine Translation

["Martin J. Dürst"]

Hello Michael,

On 2009/09/11 5:44, Michael Everson wrote:

> Any piece of text can be translated from one language to another.
> It can be translated well, or badly.
> It can be translated by a person, or by a machine.
> That is an authorship tag.
> There is no way of knowing whether a bad translation was written by a
> person or a machine.

I don't think this is true, in any case not in this generality. Anyway, 
if it were true, that might lend some support for the need for tagging 
(independent of how this might be done) the distinction between human 
and machine translators at the source.

> There is no way of knowing whether a good translation was written by a
> person or a machine.

I think this is definitely mostly not true. The chances that a good 
translation is done by a machine are very low. In case you know about 
good machine translations, maybe you could give us an example?

> This machine tag could not distinguish between an accurate translation
> or an inaccurate one. It has no a priori rationale but to be some sort
> of vague "caveat" and since a translation might as well be accurate as
> inaccurate, what good would such a caveat yield?

For research (including statistical) into how languages are actually 
used, it may in many cases make a lot of sense to exclude machine 
translations (even the good ones, in case they happen to exist).

Regards,   Martin.

-- 
#-# Martin J. Dürst, Professor, Aoyama Gakuin University
#-# http://www.sw.it.aoyama.ac.jp   mailto:duerst at it.aoyama.ac.jp