>On the UT news, there is a link,
http://www.newscientist.com/article.ns?id=dn7054, about a new approach to automatic translations.
>
>The idea (that the software learns by comparing thousands of previous translations) seems an interesting new approach, but for me, in this case, "seeing is believing"; i.e., I doubt that it will produce a high-quality result.
>
interesting. i think it could work if its done correctly (i didn't read the whole article). if it learns from previous translation it only makes sense if the translator is told if it was correct or wrong. inteligent programing has been a pet-project of mine. i truely believe that its possible if the software is beeing tought the right stuff.
this in connection with current technowledgy - that should get pretty close. ofcourse a human is always gonna be better - in any area - just slower.
>IMNSHO, by the time automatic translators can be made that are really intelligent enough (if ever), they will no longer be needed. (It is my hope that mankind will learn a common language in schools all over the World. This language must not necessarily replace the languages spoken in each country.)