Interesting article from Adot about how it’s possible to over-train Thunderbird’s spam filter, by repeating too much of the same spam (presumably so that new patterns are overwhelmed by the number of old ones).
Spam, spam, spam and Thunderbird
Interesting article from Adot about how it’s possible to over-train Thunderbird’s spam filter, by repeating too much of the same spam (presumably so that new patterns are overwhelmed by the…
I’ve been using Apple’s mail application, and I wonder if the over-training thing applies to it too. The web page you referenced doesn’t say anything about other apps being susceptible to the problem, and there are some that suggest it’s a bug, but I wonder if it isn’t an intrinsic feature of bayesian spam filtering.
I’ve used other Bayesian filters, and there definitely seem to be limits to them. That said, I find the Thunderbird filter to be less effective than others, in terms of what it lets through (certainly less effective than the previous anti-spam solution I was using with OE). I’ll have to consider this more carefully.