John Hardin writes: > On Thu, 13 Nov 2008, Karsten Bräckelmann wrote: > > > You do realize that "similar text" won't necessarily remove this from > > the auto-generated patterns, right? > > > > That said -- a really good, diverse ham corpus is crucial for this type > > of generated rules. "Similar" style text might prevent FPs in the > > future, so it is a good idea to share them, if you can. > > Justin: should I make a faked ham corpus of just the "corporate > disclaimers" from the fraud corpus?
yep, please do. the more ham, the merrier ;) --j.
