Hello Peter,
>a quick question about initial training: i have a 4k spam corpus, 1
>mail per file. They all contains X-clapf headers. Will this have any
>negative effects on the token database?
>
> Will clapf learn its header prefixes as spam tokens?
No, clapf uses only a very limited set of the email headers. The
X-clapf-... field is not among thm.
Just a reminder: you should really use ham emails as well during the
training, not only spam. Giving only spam emails to clapf for learning
would result definite disaster.
Best regards,
Janos