[Python-Dev] Re: [Python-checkins] python/nondist/sandbox/spambayes GBayes.py,1.7,1.8

Tim Peters tim@zope.com
Wed, 21 Aug 2002 17:05:58 -0400


[Tim]
> the version of this we've got now does update during scoring

[Neil Schemenauer]
> Are you planning to check this into the sandbox?

Update-during-scoring was already in the initial version.  This works with a
Python dict, though (which Barry pickles and unpickles across runs), not
with a persistent database (like ZODB).  Changes to use a ZODB BTree would
be easy, but not yet most interesting to me.  There are many more basic open
questions, like which kinds of tokenization ("feature extraction") do and
don't work.  BTW, that's why the WordInfo records have a .killcount
attribute -- the data will tell us which ways work best.