[spambayes-dev] Pickle vs DB inconsistencies
Meyer, Tony
T.A.Meyer at massey.ac.nz
Fri Jun 13 14:28:42 EDT 2003
> Anybody have a clue WTF is going on here?
Not much of one, but:
> I'm running a
> several-days-old CVS spambayes, so I'll try "cvs up" first.
The only semi-recent thing I can think of that might effect this sort of
thing are Mark's changes to the DB classifier. The two main points,
IIRC, were that the classifier now doesn't cache hapaxes and stores a
list of changed words so that not all tokens are saved. It's possible
that you have a cvs that has the first changes he checked in (which were
buggy, IIRC), and not the fix. Apart from trying current cvs, you might
also try checking out cvs from the 28th of May or before, which doesn't
have these changes, and see if that fixes it.
> And then I guess I'll start picking through the DB and pickle
> files manually to see if those differences are visible that way.
> But I have no idea what that will tell me ...
BTW you can use the DBImpExp script to do this if you want to (to
convert to text/pickles/db).
What happens if you only train on a single message? Do you get the same
result?
=Tony Meyer
More information about the spambayes-dev
mailing list