> On Tue, Mar 3, 2009 at 16:59, Marc Perkel <m...@perkel.com> wrote: > > Yet Another Ninja wrote: > >> On 3/3/2009 5:32 PM, Marc Perkel wrote: > >> > The important point here is that SA needs to evolve beyond the concept > >>> > >>> of using addition to compute scores. Ideally there should be more hard > >>> coded rule combinations or using baysian statistics to find how rule > >>> combinations where the combinations are a more accurate indication than > >>> the > >>> rules themselves. > >>> > >>> Anyhow - just throwing this out there for people to chew on and think > >>> about. > >> > >> What you mean in SA languages is to use more meta rules. > >> > >> > > Yes - meta rules is a step in the right direction and I do know SA has that. > > What I'm asking is - perhaps there is a way to make an automatic meta rule > > generator. Some way to statistically look at rule combinations to generate > > meta rules.
On 03.03.09 17:45, Justin Mason wrote: > yes -- there is at least 1 in the masses subdir, we've used it in the past. > (to generate the ADVANCE_FEE_* ruleset iirc) > > it can be quite dangerous -- if the source rules have any FPs, this can > wind up reinforcing the FPs and making them much more serious. that's > why we haven't used it more widely. otoh it can lower down FP rate. When we do a mass-check we do know the FP rate for each rule, don't we? And we also know the rate of rule combinations. So, we can and will expect some combinations to have positive score, some will apparently have negative score... -- Matus UHLAR - fantomas, uh...@fantomas.sk ; http://www.fantomas.sk/ Warning: I wish NOT to receive e-mail advertising to this address. Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu. I don't have lysdexia. The Dog wouldn't allow that.
