Seite 1 von 1

Learning Filter

BeitragVerfasst: 3. Mär 2010, 14:31
von Rein
I hope the administrators don't mind, but I start a new topic,
and hope it collects all messages about this filter in the future.

I discovered that a new "way of learning" is to learn things like:
T*contact S*contact and even T*livingtale (my company name).
It appears in spam as non-spam.
Got no idea where it comes from, but it pollutes the non-spam filter, and the spamfilter.
Any suggestions?

Re: Learning Filter

BeitragVerfasst: 3. Mär 2010, 22:46
von michel

S* means the Learning Filter found the word in the mail's subject. T* relates to the To header field and so on. These prefixes do not pollute the filter, they help to identify spam more accurately actually.


Re: Learning Filter

BeitragVerfasst: 4. Mär 2010, 08:48
von Rein
Thank you!

Then: how can I make thta those words are always ignored?
No spam, no non-spam?

Re: Learning Filter

BeitragVerfasst: 4. Mär 2010, 18:38
von michel

Simply use the Training Area more offen to improve the Learning Filter. That's all.


neutral zone within

BeitragVerfasst: 4. Mär 2010, 21:29
von Rein
Okay, I don't understand why that should be so, but I believe you.
But but .... in Dutch we say: no 'but' telling ahhhh

I need some confirmation...
the spam which has those words and are at this point seen as non-spam...
If I tell the learning filter that it is spam, then I learn it that those words are really spam.
But I don't want to do that. The message is spam, but only those words are non-spam.
If that's true, it means that I should always control the non-spam words to see if I can learn it to be spam.
Complicated, true.
Otherwise, if you should say: don't bother, because only the quantity is what is counting, than it should be different.
If not so: I think it would be best if there was a neutral zone in the learning filter, which 'deletes' the words when they appear in the spam/nonspam part.

Maybe a stupid idea, but stupid idea sometimes are leading to great inventions....