Recently, I found that there are two evolution plugins: evolution-plugin-bogofilter and evolution-plugin-spamassassin. Both try to tackle the same problem - help to analyze spam by getting manual training from the UI. Additionally SpamAssassin includes a lot of default regex rules, while bogofilter fully relies on baysian word filters.

There is some help to decide which to choose from here ¹. Somehow bogofilter was already installed on my system, so I sticked with it.

Now I found out, that it actually helps to use the “Mark as Junk” and “Mark as not Junk” buttons in the context menu of mails in evolution. It is used to train the filters and list the content as good or bad. This also has to be done for mails which are not SPAM to explicitly add these to the filter. Of course, one should watch out about confidential data in the filter, as all words in the mails are then stored unencrypted in a BerkeleyDB in the user home directory. All entries and the occurances as good/bad can be shown with:

bogoutil -d ~/.bogofilter/wordlist.db | less

or a single word can be queried using

bogoutil -w .bogofilter/wordlist.db Guten
                                 spam   good
Guten                               0     53

which shows that German SPAM is rather rare in my mailbox.

I also found out, that adding a mail to one, as well as to the other does not revert the first decision unfortunately and adds the entries to spam and good equally.

One also has to enable the Junk filter globally in the Preferences as well as on a per account basis:

I am very happy that this functionality is built in. Quite sad that I have to train it myself, but one can also mark a lot of mails at once, so this is fine. So I will just train the filter a little bit further and see how well this works out and update the post accordingly.