There were discussions in one of the forums about machine translation and about the funny stuff it comes up with. I think I already threw in my two cents by saying that MT has its place in workflows that involve human translation. It is actually no accident, because MT tools are actually ancestors of our CAT tools (or our Translation Environment Tools). Although there are some people rightfully critical about the use of CAT tools, it appears that the advantages often outstrip the problems. Don't get me wrong - I am all for a critical position towards any translation memory application (in fact - towards any computer application). This critical distance is in particular useful for translators as our task is quite complex. But (and of course there is a "but"...) don't dismiss MT to easily. There is a lot of stuff in a translation process that can be automated - and some of it even has to do with languages. Think, for example, of terminology recognition tools. They've come a long way since some computer linguists came up with concordancers and word lists. Today we do have bilingual concordancer, which are able to harvest terminology from bilingual files. And any translation memory file is by definition a bilingual file.
I remember that a few years ago I saw a demonstration of a prototype of Trados' Multiterm. It was in Germany on a TEKOM exhibition. I was rather skeptical of the product for two reasons: on one hand the demo worked on a Greek to Finnish file (so there was no way for anybody to check the results) and secondly it costs 2500 Euro - so even if a Finnish to Greek translator came along, s/he wouldn't be able to afford it. :wink:

Well - today those applications do work and they are enormously useful and save technologists a lot of money.

But there's more to MT.. but not for now.