Hi Robert, > Hi, I think this is a tremendous work and I'm thinking if your resources > shouldn't be used for other things. Some time ago I posted a request about > Bayes filters etc. IMO it should be possible to use those 1200 taged > messages as the base to train a bayes filter and than have it categroize > the rest of the messages. What do you think? Rob
I think the tagging would be an excellent way of training a Bayes network. Whether it could tag as well as a human -- well you prove it, and we'll be begging to use your network to tag new threads. (Personally, as a DMOZ editor, I believe that "humans to it better". But I'm happy to be proved wrong) Feel free to use the ML archives and their tags as a training data set. You can download the ML archive (one message at a time!) using the Library Data Services Interface: http://www.rebol.org/cgi-bin/cgiwrap/rebol/documentation.r?script=lds-local.r The LDS request 'get-rml-message will return the current tags for the thread (field is 'topic-list -- not currently mentioned in the documentation) Sunanda. -- To unsubscribe from this list, just send an email to [EMAIL PROTECTED] with unsubscribe as the subject.
