On 8/11/05, Luke Dicken <[EMAIL PROTECTED]> wrote:
> I doubt that any AI program is this far advanced, to be honest. It'd
> probably generate as much rubbish as good. The Fleisch-Kincaid
> algorithym just assesses text (not that well), it doesn't alter it.
The algorithms for doing this should be reasonably straightforward fwiw,
but you would need a complete buttload of valid training data in order
to make it work. If anyone is interested in doing this, a good starting
place is probably Knight's paper "A Statistical Machne Translation
Workbook". Using the methods outlined there as a basis and a reasonable
heuristic, you ought to be able to translate mechanically from Highbrow
english to Lowbrow english relatively accurately - but as I say, you
will need a LOT of test cases to make it work.
Luke
-
Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html . Unofficial list archive: http://www.mail-archive.com/[email protected]/
--
You can't build a reputation based on what you are going to do.
