[Spambayes] How low can you go?
tameyer at ihug.co.nz
Mon Dec 15 19:35:21 EST 2003
> How about the "or something" of checking it in and keying it
> with an x-ified config option? Does it mess with the code
> all over the place or is it a relatively isolated change to
> the source?
Relatively isolated; it's all in classifier.py, anyway. I'll do this. If
you (or anyone else) would like to eyeball it to check that there's no
obvious error anywhere, that would be great (it did appear to work for me,
though, and I used it for a couple of months.
Note that anyone wanting to use it will need to retrain from scratch to get
full benefit, and that it does slow down classifying somewhat. Tim also
suggested that increasing the number of tokens used would be beneficial, and
I certainly found that the case (from 150 to 600, I think).
> Does it include Gary's scoring change?
I wasn't paying enough attention to the earlier messages: is this the change
that means that only the strongest of the two unigrams and one bigram is
used? If so, then yes, it includes that.
More information about the Spambayes