237403 - Bayesian Noise Reduction

Status: RESOLVED WONTFIX

(MailNews Core :: Backend, enhancement)

Description

[Robert Accettura [:raccettura]]

This might be something worth looking into.  

Since spammers are starting to pollute their spam in hopes of drowing bayes with
bogus data.

In any regard, there's some interesting reading on that page, and through linkage.

Comment 1

[Scott MacGregor]

I read about this too and I thought I filed a bug on it but I guess not. The
problem is that website says nothing about how BNR does what it does so we could
mimic it :)

Comment 2

[Robert Accettura [:raccettura]]

Well, there's always the source :-)

But there's also:
http://www.nuclearelephant.com/projects/dspam/BNR%20ACM.pdf
and
http://www.nuclearelephant.com/projects/dspam/BNR%20LNCS.pdf

Comment 3

[José Jeria]

Bug 235225 is the same thing?

Comment 4

[Jack Eidsness]

On or about July 24 2004, a link to a GPL BNR library was added to that page (I
see the comments are from before that;  if the library was already there, I
apologize for the noise).

If this is new, relative to this discussion in March, the doability of this may
have increased significantly since the ticket was submitted, and I think it
would be a fabulous addition.  Adding this quickly would kick outlook right in
the pants, because already over the past year and a half, tbird's spam catching
numbers have dropped off significantly.

Comment 5

[Wayne Mery (:wsmwk)]

(In reply to comment #3)
> Bug 235225 is the same thing?

rkent, are the above and this bug still useful?

Comment 6

[Kent James (:rkent)]

(In reply to comment #5)
> 
> rkent, are the above and this bug still useful?

Probably not. If we were investigating improvements to the algorithms, we would probably look to the literature, not to discussions in bugmail. And we would be more likely to work on making the algorithm more pluggable, and rely on extensions to try specific techniques.