[Spambayes] Re: Training oddity/confusion

Tim Peters tim.peters at gmail.com
Mon Jan 17 21:19:09 CET 2005


[Mathew Hendry]
...
> The latest one has another little problem, in that SpamBayes
> didn't pick up the lone key word (OnlineMeds) in the body as a
> token:

It did, but not in the way you expected:

> |||OnlineMeds
...
> All Message Tokens
...
> 'skip:| 10'

SB splits on whitespace, so "|||OnlineMeds" was viewed as one token. 
A token that's "too long" (this one is 13 characters) gets replaced
with a synthesized "skip:" token.


More information about the Spambayes mailing list