Peña, Botp
9/29/2004 6:28:00 AM
Hi,
From experience (as admin of email), spam is very difficult to combat
technically because you have to read the message before declaring it spam.
The problem is exacerbated since not all users treat spam equally in meaning
and in text (some are of different language). By and large, the most
effective defense we have here is using the humble bayesian algorithm/filter
in combination w a spam admin and helpful/disciplined users.
Thus for a wiki situation: if a user submits, he will have to pass thru a
bayesian filter before his page is finally posted. If his post is flagged as
spam, it will still be posted but with a readonly notice below that it was
flagged as spam and _his changes_ will be deleted after a scheduled periodic
scan.
Maybe we can create a bayesian learner that learns fr a set of text inputs.
consider this quiz#3 (Hi JamesEGII :). As corollary, we then create a
bayesian filter module that accepts text and spits out a bayes result in
percentage basing on what it learned on #3. Consider this quiz #4.
kind regards -botp