Popfile update #2

Accuracy of the bayesian spam filter is now up to 80% after training on 900 messages (two days worth of traffic). The downside is I spent at least 45 minutes each day training the filter, and the application slows down as you feed it more mail. That’s not acceptable beyond a couple more days; hopefully the accuracy will continue to improve (especially on the false positives) so that I can cut back on the training.