4 ms·
Both here and Mollom's homepage make a big deal about accuracy (though they call it efficiency), but only talk about spam accuracy. If spam is really 90% of th
by tedunangst 16y ago
Both here and Mollom's homepage make a big deal about accuracy (though they call it efficiency), but only talk about spam accuracy. If spam is really 90% of the inputs, it's pretty easy to have skewed stats, so I'm curious (suspicious) what their accuracy is for legit messages. How many false positives?
i.e., a classifier as simple as "return (random() % 10000 < 5)" will mark 99.95% of spam as spam.
- nowarninglabel 16y agoAs a heavy user of Mollom, I can say anecdotally that it performs well, but I don't have stats for you, would be cool if Dries could provide some.