Re: Outlook should process rules first, before junk.
From: Vanguardx (see_signature)
Date: 09/21/04
- Next message: Vanguardx: "Re: E-Mail Delay"
- Previous message: Joe: "Really need help with deleting rules."
- In reply to: corpcon: "Re: Outlook should process rules first, before junk."
- Messages sorted by: [ date ] [ thread ]
Date: Tue, 21 Sep 2004 16:00:45 -0500
"corpcon" <corpcon@discussions.microsoft.com>
wrote in news:19E0CB91-CE8E-492A-BCF3-A787EDAA8C69@microsoft.com:
> You may be thinking of previous versions of Outlook that had the
> "Junk" and "Adult" filters. I know of no such thing in 2003.
The OP never identifed WHICH version of Outlook they were using, so any
response regarding any version is a valid response. When you ask for a
tire at the car shop and don't identify the specific car then any tire
they give you satisfies your request, even a tire for a wheelbarrow.
I realize Microsoft wants to hide that they employ Bayesian filtering in
OL2003 but there is little you can configure for that filtering. You
cannot teach the filter before employing it by feeding it sets of good
and bad e-mails (i.e., there is no learn or train mode) to reflect YOUR
e-mail history, you cannot configure word expiration (to reduce the
noise floor for words that fade out of your vocabulary or are not used
for extended periods), it may be more susceptible to bias by using a
flood of inocuous word lists in messages (I'm still looking for proof
that good-word flood lists will negatively bias Bayesian filters *if*
properly configured regarding expiry), and, in fact, there is almost
nothing configurable regarding the Bayesian, er, Junk filter used in
OL2003.
>From http://office.microsoft.com/en-us/assistance/HP052429671033.aspx:
"There are two parts to the Junk E-mail Filter: the Junk E-mail Filter
Lists and the state-of-the-art technology developed by Microsoft
Research that evaluates whether an unread message should be treated as a
junk e-mail message based on several factors, such as the time it was
sent and the content of the message. The filter does not single out any
particular sender or type of e-mail message. The filter is based on the
content of the message in general and uses advanced analysis of the
message structure to determine the probability that it is a junk e-mail
message."
All that fluff just so Microsoft doesn't divulge that they use their own
variation of Bayesian filtering. Geez. I wonder why Microsoft slapped
the article at http://www.mapilab.com/news/0042.html (so the linked
article isn't available anymore). Yeah, we wouldn't want to divulge how
Microsoft did it although there are other open source Bayes filters
available. Unfortunately I never got to see the article before
Microsoft forced its deletion, but articles like
http://support.gfi.com/manuals/en/me10/me10manual-1-12.html indicate
that OL2003's spam filtering is Bayesian and comes with a preset level
of training (so you cannot adapt the filter's statistical weighting
towards YOUR sampling of good and bad messages -until- you have used it
for quite some time).
As noted in GFI's article, "... [Bayesian] spam detection rates of over
99.7% can be achieved with a very low number of false positives!"
However, even one false positive can be one too many. That's why I use
the Quarantine plug-in in SpamPal to give me a plain-text version of
every message that got marked as suspected spam (alas, there is no
expiration of saved plain-text copies of the messages in the quarantine
folder so I had to write a batch file to expire them, and which I gave
to the Quarantine plug-in author to put on his site). I also move
suspect messages (i.e., spam tagged) into my Junk folder in Outlook
which has auto-archiving enabled to permanently delete messages within
it that are over N days old (I usually set it to expire in 3 days
although I'm tottering on just 1 day). They'll go away if I don't
attend to them. Unfortunately auto-archiving or message expiration
isn't available in other e-mail clients, like Outlook Express.
> I've tried SpamPal and hated it. It tagged a lot of very legitimate
> e-mails ... including all that came from Road Runner accounts.
Then don't use the overly aggressive and super vigilante DNSBLs (DNS
blacklists), like SPEWS. If you use SPEWS, expect lots of false
positives. I used to use SORBS but discontinued using that list when
they choose to incorporate the SPEWS list (SpamPal indicates the
interdependency by enabling SPEWS if you select SORBS). I've had 2
false positives by SpamPal since mid-February (whereas I get about 2
false positives per month via the ISP's server-side spam filter, so I
may have to disable server-side spam filtering). I use the following
blacklists: SpamHaus SBL+XBL, Composite Blocking List, ORDB, SpamCop,
NJABL, and blitzed.org (although I might drop this last one). Use the
Quarantine plug-in to capture the content of any spam-tagged e-mails but
which are in plain-text format. If you're unwilling to configure the
product to behave as YOU want regarding your e-mails experience then
don't use that product. Same for anti-virus and anti-spyware products.
-- _________________________________________________________________ ******** Post replies to newsgroup - Share with others ******** Email: lh_811newsATyahooDOTcom and append "=NEWS=" to Subject. _________________________________________________________________
- Next message: Vanguardx: "Re: E-Mail Delay"
- Previous message: Joe: "Really need help with deleting rules."
- In reply to: corpcon: "Re: Outlook should process rules first, before junk."
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
