On Sat, Oct 18, 2003 at 10:43:04AM +0000, Jonathan Matthews wrote: > Anthony DeRobertis had the gall to say: > > OK, we get a fair number of these. So do some other people. None of the > > claimants ever seem to respond when asked about the details. From > > googling, here are some other references: > [snip] > > > Personally, I sort of suspect address collection or other scam. I > > suspect this because of all the messages we've gotten, and all I can > > find on the web, the are quite similar in ways you would not expect, > > such as putting the commas in "50,000,000". No one wrote "50000000" or > > "50.000.000". I'd expect that if these were messages generated by > > confused lusers, we'd see more variation in them. > > [...] > > I'm always a little dubious about telling bogofilter that they're spam, > as they include valid nouns which might easily come up on lists. Ideas, > anyone?
That is the neat thing about bogofilter (and other bayesian classification methods): if a word in the spam message really does appear more often in non-spam messages, then that word will not contribute to marking future messages as spam. -- gram
signature.asc
Description: Digital signature
