On 2 Jul 2003 at 17:47, Chris Newman wrote:
> Note that similar issues apply to the $Spam keyword but I believe they
> can be corrected. As the definition of the term "spam" is
> controversial, I recommend using the term "junk" which can more easily
> be defined as we wish. I would use four keywords: $Junk, $NoJunk,
> $AutoJunk, $AutoNoJunk. "Junk" is defined as an explicit end-user
> indication they do not wish to read similar messages in the future.
> "AutoJunk" is an automated system's guess that the end user might
> consider that message Junk. As $AutoJunk and $AutoNoJunk are
> error-prone, an MUA MUST provide the ability to view $AutoJunk
> messages with reasonable convenience.
Looking at the popularity of Bayesian filtering at the moment, wouldn't it
be better to have "Junk", "NoJunk", "AutoJunk", "AutoNoJunk" and
"AutoMaybeJunk"? I know that the Bayesian port that is being
developed for my mailer at the moment distinguishes between things
that the Bayesian classifier has definitely decided are junk, things that it
has decided are definitely not junk, and things where it's not sure but
thinks that they *might* be junk. Or is there a combination in Chris's
flags that would provide this shade of meaning that I have missed? The
"maybe junk" setting is actually quite important because it typically
represents stuff that is on the border between spam and not-spam (I
passionately hate the term "ham" for this), so there's a significant
training impact that can be made by filing it separately.
Also, could someone please explain why all these flags are "$xxx"?
Where did the $ convention come from? I can't find any reference to it
in RFC3501... I don't have any particular beef with it, but I'd be curious
to know the rationale.
> This allows the combination "$NoJunk" and "$AutoJunk" which indicates
> the automated system made a false-positive (very useful information to
> retain).
Certainly no argument here. Being able to store this kind of information
is quite important.
Cheers!
-- David --
------------------ David Harris -+- Pegasus Mail ----------------------
Box 5451, Dunedin, New Zealand | e-mail: [EMAIL PROTECTED]
Phone: +64 3 453-6880 | Fax: +64 3 453-6612
In a Maltese guidebook:
"Although every possible care has been taken, I do not accept
responsibility for inoccurancies."
