Bayesian Filter for spam

[Guest]

Have you guys thought of implementing a Bayesian filter for spam in future
versions of Outlook? It would be handy and more effective than the Junk
controls at the moment which seem like simple block tools.

 

[Diane Poremsky [MVP]]

Which version of Outlook are you using?

--
Diane Poremsky [MVP - Outlook]
Author, Teach Yourself Outlook 2003 in 24 Hours
Coauthor, OneNote 2003 for Windows (Visual QuickStart Guide)
Author, Google and Other Search Engines (Visual QuickStart Guide)



Join OneNote Tips mailing list: http://www.onenote-tips.net/

 

[Vanguardx]

Have you guys thought of implementing a Bayesian filter for spam in
future versions of Outlook? It would be handy and more effective than
the Junk controls at the moment which seem like simple block tools.

Outlook 2003 does use [Microsoft's concept of] a Bayesian filter. They
hide it behind their long-winded description at
http://www.microsoft.com/office/editions/prodinfo/junkmail.mspx:

"... junk e-mail message based on several factors—such as the time the
message was sent and the content and structure of the message. The
filter does not single out any particular sender or type of e-mail
message. Rather, it uses advanced analysis to determine how likely it is
to be thought of by you as a junk e-mail message."

See my other post at
http://groups.google.com/groups?selm=#[email protected].
Of course, that doesn't stop you from implementing someone else's
Bayesian filter, like using SpamPal along with its Bayes plug-in (both
free) or the Bayesian-only plug-in/proxy at
http://spambayes.sourceforge.net.

 

[Guest]

I am using Office 2003 and SpamBayes.

Overall the Office Junk Mail filter is 95% effective; however, it has a big
problem with FALSE POSITIVES (that is, the junk mail filter incorrectly
identifies some email as spam when it is not). This means I MUST review all
of the Junk Email in order to recover any email messages that were
incorrectly marked as spam... sort of defeats the whole purpose of having a
spam filter.

The Junk Mail filter needs some additional means of learning what is NOT
spam; the "accept email from address xyz" is not an acceptable solution.
(email addresses are meaningless and they change).

Note that SpamBayes has an effective learning model; maybe Microsoft should
license that and add that in.

 

[Jeff Stephenson [MSFT]]

Note that anything with an "effective learning model" still requires you to
search your Junk Mail folder for false positives so that you can *teach* it.
Spam filtering will always require user-checking until we get more effective
means of stopping it into the Internet email infrastructure, rather than
just at the receiving end...

 

[Guest]

Actually, that is not a problem.

I used SpamBayes with Outlook 2000; after a month of *training* SpamBayes
was 98% effective at filtering spam, and NEVER had a false positive... which
is the critical point. I could trust SpamBayes would not junk good mail.

Anyway, the key point for me, is that I would like is someway to finetune
how MS filters out junk mail.