What is this new classification function?
If you have enabled the Bayes-based spam classifier you will enlarge a data collection about the amount of occurences of words as ham (a wanted entry) or spam (an unwanted entry). With this collection the Bayes-based-service B8 tries to classify a new forum entry as ham or spam by itself. B8 will check any input against its existing list of words. If a word was classfied 30 times as ham and only 7 times as spam, the classifier will assume, that this word is harmless. It does the same weighting for all words, that are known in the list and computes a balance for the whole input.
The purpose is to make it unnecessary to use external spam protection services as i.e. Akismet. It will furthermore be possible to use those services but in terms of the visitors privacy it is IMHO desirable to prevent it where possible. And here it is possible.
Trenne niemals Müll, denn er hat nur eine Silbe!