> 1. Why does the third line of yelling account for a negative score? It's > quite common for UCE to have multiply lines in ALLCAPS.
The scores are calculated using a genetic algorithm (GA) that analyzes a huge corpus of spam and nonspam messages, and it probably found some overall benefit in reducing false positives when it made that score. But then, the GA has been greatly improved recently and it changed the score in 2.43. score LINES_OF_YELLING_3 0.315 score LINES_OF_YELLING_2 0.217 score LINES_OF_YELLING 0.212 Scores this low indicate that all three LINES_OF_YELLING rules have become pretty worthless in detecting spam. > 2. I don't think that auto_report (to Vazor) is a good feature because > anything that can be automatically spotted as spam doesn't neccessarily > need to be in any database at all but we risk to submit mail that is not > spam (but uses spam characteristics) to Vazor, which will make it score > even higher next time it will be checked. In General, a DB like Vazor is > only then of any help if its contents are VERIFIED spam, not likely spam. SpamAssassin hasn't had an auto-report feature for quite some time. (was it 2.30 that changed this?) The developers agree with you. -- Michael Moncur mgm at starlingtech.com http://www.starlingtech.com/ "Pity the meek, for they shall inherit the earth." --Don Marquis ------------------------------------------------------- This sf.net email is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ads.sourceforge.net/cgi-bin/redirect.pl?sunm0002en _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
