Re: testing corpus?

On 2026-06-26 at 14:48:18 UTC-0400 (Fri, 26 Jun 2026 18:48:18 +0000)
Rick Wesson <[email protected]>
is rumored to have said:

> Hello,
>
> I've been doing evaluations of various clustering algorthims. I'd like 
> to do some testing to see if I can improve SA speed by adding some 
> clustering capabilities. Is there a goto corpus that you use for 
> testing and baseline.
>
> I've attached a draft of some of my research using malware corpus and 
> I'd like to begin working on doing the same with spam.

As far as I know, there's no shareable corpus of email that is useful 
for spam research. The problems are that there is a very loose consensus 
on what is or is not spam, what spam and non-spam people get has huge 
variability, and personally identifying information can be a significant 
clue in identifying spam.

The corpora that I've used for research have all been built from the 
mail flows of the people backing the research.



-- 
  Bill Cole
  [email protected] or [email protected]
  (AKA @[email protected] and many *@billmail.scconsult.com 
addresses)
  Please keep discussion mailing list replies *on-list*
  Not Currently Available For Hire