[Tim Peters] > It's sure nice to see someone is still testing ideas! I was mother-in-law-sitting for a few days, so leaving the machine running tests was an easy option :)
> It would be even nicer if we could find a good one <wink>. It would be easier if the with-defaults results weren't so good, of course :) If I find time to really try something out I suppose I ought to start by staring at the mistakes that running with defaults is generating. > Indeed, I had to stare hard to find any difference at all. Me too :) I wondered if maybe I didn't have any of the things, so grepped through the mail, but there were some (mostly from the same domains, which probably means that there were clues being harvested already). [...] > In the cases where you added the canonicalized form (in addition to > retaining the original form), it may have a bad interaction with the > bigram option (which I believe you use), destroying the natural > bigrams. It would be clearer to turn bigrams off in that case. I ran them all with bigrams off, although I do have it on with the classifier I actually use. > But I wouldn't expect it to help anyway. You'd be right :) =Tony.Meyer _______________________________________________ spambayes-dev mailing list [email protected] http://mail.python.org/mailman/listinfo/spambayes-dev
