>-----Original Message-----
>From: Keith C. Ivey [mailto:[EMAIL PROTECTED]
>Sent: Wednesday, May 19, 2004 9:00 AM
>To: [EMAIL PROTECTED]
>Subject: RE: Bayes Poison detection
>
>
>Jonas Eckerman <[EMAIL PROTECTED]> wrote:
>> On Mon, 17 May 2004 14:07:50 -0400, Chris Santerre wrote:
>>
>> > SARE had done some experiments with a custom eval that would
>> >also check for 20-50 words without "tie in" words like : a,
>> >at, is, the, of, on, that, have, had, and,.......
>>
>> If someone does decide to pick this up, I really hope it'll be
>> in it's own separate rule-set as it will be heavily language
>> dependant.
>
>Good point. Note, though, that for the rules like this I've
>seen, words are defined as matching /^[a-z]+$/. For most (but
>not all) non-English text, you're unlikely to get through many
>words without using a non-ASCII letter.
>
>--
>Keith C. Ivey <[EMAIL PROTECTED]>
>Washington, DC
>
Well I had my ninjas confused. It was ninja 'C' ('C' for "I bet I can get it
to run on a 'C'ommadore 64!" :o)
I've been talking to him about it. He has it working for him, but it is kind
of a complicated thing. It has more tweaking to do, and he is going to
revisit it. However I don't believe this type of rule is going to be for
your average user. This one is a little more complicated.
And even ninja 'C' agrees that the bayes poison is actually bayes fodder.
All it does it create a larger Bayes DB. It doesn't seem to really poison
anything. I think sometimes bayes fodder gets blamed for normal DB shift
which happens. Heh, isn't that a bumper sticker? "Shift happens!"
And for those thinking on scanning by a larger scale, we tend to ignore
bayes entirely. Great tool, but not the panacea it was hyped as.
So it is a question of more computations, lanquage custimization, and S/O
ratings that are just 'OK'. Signs still point to, "It just ain't worth it."
--Chris (The puck throwing ninja!)
