I don't think the language has any effect on Bayes training. These spams are
now hitting BAYES_99 on my gateway, based on auto-learned spams that scored
with local and network rules (I installed Matthias Fuhrmann's rule set
yesterday):
Content analysis details: (32.2 points, 6.0 required)
pts rule name description
---- ---------------------- --------------------------------------------------
0.2 NO_REAL_NAME From: does not include a real name
1.0 X_POLITICAL_SPAM_06 Subject line contains political ube
2.0 TEMP_GERM German political virus-spam
1.0 FROM_M2 relayed from mail2
15 BAYES_99 BODY: Bayesian spam probability is 99 to 100%
[score: 1.0000]
3.0 SPAMCOP_URI_RBL URI's domain appears in sc.surbl.org
[www.NoTurkishmembership.com is blacklisted in]
[SpamCop RBL at sc.surbl.org]
3.0 WS_URI_RBL URI's domain appears in Bill's database at
ws.surbl.org
[www.Euronational.org is blacklisted in SpamCop]
[RBL at ws.surbl.org]
2.0 DCC_CHECK Listed in DCC (http://rhyolite.com/anti-spam/dcc/)
3.0 X_POLITICAL_SPAM_A2 Subject line contains political ube
0.8 MANY_EXCLAMATIONS Subject has many exclamations
1.2 PRIORITY_NO_NAME Message has priority setting, but no X-Mailer
-----Original Message-----
From: Paolo Cravero as2594 [mailto:[EMAIL PROTECTED]
Sent: Friday, June 11, 2004 8:02 AM
To: [EMAIL PROTECTED]
Subject: Re: Flooded by spam in German
Stefan Hornburg wrote:
...
What is annoying is... that Bayes doesn't seem to adapt to German
language! Still hitting BAYES_00 after several trainings.
Paolo
