Hello, Sean.

You wrote 20.04.2004 @ 15:21 
  using mailer 

SC> Hi All

SC> Can  anyone  recommend  what  I'm  supposed  to  put in the "Number of
SC> ranking  tokens"  configuration  ?  Also  ..  should  I  "Use enhanced
SC> evaluation" ? Any tips welcome !

Number of runking tokens is the number of tokens which will be selected from
the dictionary of a letter to be tested which will be actually used to "judge"
or "rank" the letter. First, the letter is divided to the tokens. Then this
tokens is being sorted by the "interesting" (the difference between actual
token's rank value (0.01 for non-spam tokens, 0.99 for spam, usually some middle
value) and neutral rank which is 0.5), and the N most "interesting" tokens
participate in letter's grade evaluation. N - is exactly "number of ranking
tokens". Paul Graham recommends to use 15.

"Use enhanced evaluation" - again, if the tokens are very expressive, and actual
number of tokens with maximum "interesting" is more than defined number of
ranking tokens, the filter will take all such tokens (i.e. it exceed this
value).

-- 
Sincerely,
 Alexey.
Using TB 2.10.01 on WinXP Pro SP1 (2600), spelling by ORFO2002 (CSAPI) 
..with Kaspersky Antivirus Plugin (ver 3.5 Gold) & antispam filter BayesIt! 0.5.3

   mailto:[EMAIL PROTECTED]


________________________________________________________
 Current beta is (none) | 'Using TBBETA' information:
http://www.silverstones.com/thebat/TBUDLInfo.html
IMPORTANT: To register as a Beta tester, use this link first -
http://www.ritlabs.com/en/partners/testers/

Reply via email to