[Spambayes] Spam Clues: Missing the obvious spam

[Fran ç ois Granger]

Le 22/04/04 8:36, « Katz, Amir » <Amir_Katz at bmc.com> a écrit :

> I got this mail with the subject "how big is your p**is" (w/o the asterisks)
> and it scored only 87% and thus went into the 'unsure' folder. If I were
> SpamBayes (:-) I would give it 110% immediately just because of the P word.
> Any comments?

Results from my database below:
(for my database "small" is more important than "penis" ;-)

> ------------------------------------------------------------
> Combined Score: 87% (0.874382)
> 
> Internal ham score (*H*): 0.251129
> Internal spam score (*S*): 0.999893

Spam probability: 99.29% (0.992941904799).
Clues: (29)
 
Word Probability Times in ham Times in spam
*H* 0.000652 - -
*S* 0.986536 - -
x-mailer:internet mail service (5.5.2653.19) 0.091837 2 0
re: 0.252689 5 2
sent: 0.347699 6 4
content-type:text/plain 0.363413 117 86
to: 0.36388 11 8
from: 0.379662 9 7
message----- 0.388487 5 4
url:www 0.603918 33 65
solution 0.640935 6 14
sender:none 0.64432 62 145
your 0.649876 30 72
url: 0.677069 11 30
charset:windows-1252 0.713699 3 10
url:us 0.774425 1 5
content-type:multipart/alternative 0.80156 11 58
content-type:text/html 0.824829 16 98
1:14 0.844828 0 1
enlargement: 0.844828 0 1
size: 0.844828 0 1
to:no real name:2**2 0.844828 0 1
url:1v3 0.844828 0 1
url:b 0.844828 0 1
url:vcrx 0.844828 0 1
penis 0.934783 0 3
small 0.958716 0 5
subject:your 0.965116 0 6
to:2**3 0.969799 0 7