>- word stemming engine is used - language finnish detected
>- word stemming engine is used - language dutch detected
>The mails are German - can this causing Problems? The language detection?
Yes, this leads in to wrong word stemming - how ever, if the rebuild has taken the same 'wrong' stemming (language) the results will be fine.
I use both the Bayes and the HMM. HMM is more exact - Bayes is fine for very short mails.
>One other question is why it also uses words of the header informations for the bayesian? Is this a mistake in my configuration?
Yes some important tags are used from the header (addresses, subject ..)