sdertsr6 wrote:Is this re-training really necessary after the initial training exercise?.

thanks

Absolutely, because spammers expect you to learn to handle their messages, so they change the format frequently.

annew wrote:

sdertsr6 wrote:Is this re-training really necessary after the initial training exercise?.

thanks

Absolutely, because spammers expect you to learn to handle their messages, so they change the format frequently.

Wouldn't the same thing be accomplished by simply tagging all new incoming spam messages as spam after the initial training on the spam folder? Seems like it would be unnecessary to keep a folder for training and then to re-train on that folder in which all messages have already been identified as spam.

undoIT wrote:Wouldn't the same thing be accomplished by simply tagging all new incoming spam messages as spam after the initial training on the spam folder? Seems like it would be unnecessary to keep a folder for training and then to re-train on that folder in which all messages have already been identified as spam.

Yes, this is the question I raised the other day. The impression given is that Bogofilter requires initial training so that it has a representative sample of the e-mail you receive and can learn what you consider is spam.

After that you should only need to indicate ham or spam by exception.

Comprehensive retraining should not be necessary and in my experience is not necessary.

Well, Bogofilter is definitely not worthless. I trained it with about 100 emails of spam to get started, and I've been marking any new emails that it didn't catch as spam since then. I've only trained it for ham with the emails it accidentally filtered. It is now catching a lot of the spam and hasn't filtered any valid email for the past two days

I do have a couple of suggestions though. I really think it would be a good idea to add a window to the Spam Wizard that explains what needs to be done to get Bogofilter working, for the sake of newbies and people who are not familiar with Kmail / Bogofilter. The window could read something like this:

Final Window:

You have now completed the spam wizard for Bogofilter. It is necessary to train Bogofilter to recognize spam by using the "Spam" button. Make sure to check the spam folder regularly in the beginning. If Bogofilter has accidentally filtered any good emails, mark those using the "Ham" button. The more you train Bogofilter to identify spam and ham, the more accurate it will become.

Also, I think the Spam Handling filter should mark the messages as Unread by default, rather than Read. This makes it easy to scan for new messages that have been filtered as spam by Bogofilter, identify any good email messages that were mistakenly filtered and then mark them as ham. Otherwise, the important messages may get lost.

undoIT wrote:Well, Bogofilter is definitely not worthless. I trained it with about 100 emails of spam to get started, and I've been marking any new emails that it didn't catch as spam since then. I've only trained it for ham with the emails it accidentally filtered. It is now catching a lot of the spam and hasn't filtered any valid email for the past two days

This echoes my experience, once initial training is done Bogofilter does a very good job.

undoIT wrote:Also, I think the Spam Handling filter should mark the messages as Unread by default, rather than Read. This makes it easy to scan for new messages that have been filtered as spam by Bogofilter, identify any good email messages that were mistakenly filtered and then mark them as ham. Otherwise, the important messages may get lost.

Not sure I follow this, perhaps my system is configured differently. On my system none of the incoming mail is marked as Read - including all mails identified as Spam (which are automatically put in the Spam folder for ease of identification).

Have you made changes to the default spam/ham filters?

@undoIT: I don't know whether Bogofilter will be the same in the new version of KMail, but I think that it's worth filing a KMail bug/wish, making your suggestion for the end of the wizard. Improvements to documentation are always worth suggesting, and your comments look fair and reasonable to me.

I notice that Bogofilter (for Kmail, KDE v.4.4.3, Mandriva 2010 (free)) doesn't work. Period.

I have trained it using over 150 Spam emails. The "Spam" button on Kmail does transfer emails to a Junk folder, but only manually. Bogofilter does not find spam and transfer it automatically.

Any suggestions?

Please post a screenshot of the filters you have installed. Another filter may be preventing them from running correctly.

El Viejito wrote:I have trained it using over 150 Spam emails.

Training needs to see very many emails, both spam and ham. You need to set up folders into which you can copy a representative sample of your 'good' emails, as well as showing it the 'bad' ones, so that it can create tokens in the corpus. What's more, spam creators are very inventive. They know they are fighting filters, so re-training at regular intervals with new samples is necessary, so keep using your training folders for this purpose.

The "Spam" button on Kmail does transfer emails to a Junk folder, but only manually. Bogofilter does not find spam and transfer it automatically.

Any suggestions?

As bcooksley said, let us see your filters. We will need a screenshot of the Advanced tab for the bogofilter one, too.

Just to say training is nececessary and a continous exercise. Just 'junk' emails that appear to have got through the spam filter. As you apply training the filter like this you will subsequently find less and less training is required.