I confessed earlier that CL/NLP is having a dull patch as a whole, but that is not true of machine translation, which is having a mini- renaissance, with method sometimes called statistical (following Mercer and Jelinek) and sometimes example-based (following Nagao). In fact there is no real difference between them and both rest entirely on corpus data provided by human translators, whose skill they attempt to learn, and with increasing success, as any user of internet free translators knows. There is no clear dividing line here at all between the parts of this large field, only, it seems, bad feelings. Yorick
On 19 Aug 2008, at 10:52, Lou Burnard wrote:
> Yorick says "what I have never seen able to see is what corpus
> linguistics (in the sense in which the phrase is owned by the main
> contributors to this debate) is FOR, except the production of better
> As far as I am aware one of the largest communities interested in
> consuming the fruits of corpus linguistics (whether you're talking
> about corpora or the methods attached to them) is that of people
> engaged in the humdrum but utterly mysterious business of language
> teaching and translation.
> Sadly, none of that community seems to have seen fit (yet) to
> contribute to the present discussion. But I think if they did they
> might suggest that corpus linguistics is very definitely "for" those
> wanting to ground their pedagogic practice in language as
> experienced, rather than language as theorized (which is of course
> experience too, but not quite the same order).