Bogofiler with a specified wordlist.db
tanderso at oac-design.com
Thu Apr 6 18:11:37 EDT 2006
> the SA is used to weed out email based on their overall structure (the internal
> bayseian is not used at all) . If the email gets passed that, it then gets bogofiltered in a
> procmail recipie (instead of different wordlist, i use different parameters for certain
> users) with the -u option on. I have not done a retraining in years , just a monthly
> maintain run every month to get rid of tokens more than 95 days old.
Do you find that SA is more efficient (uses less resources) than
Bogofilter? Not on my system it isn't. Assuming it isn't on yours
either, and since you say that you use Bogofilter as the filter of last
resort (either it's spam here or it's ham), then what do you gain from
using SA at all? If Bogofilter is more authoritative, you're not
benefitting in the least by putting a less efficient, less authoritative
filter in front of it.
Also, how can you possibly maintain a valid database if you're using the
-u option and not training errors? It seems to me that you're probably
filtering many more FPs than you realize.
> And things are just fine, users have only complained once this year about email
> getting through and have had only 3 false positived so far this year.
That you know about. Do you personally review all 25 users' spam
folders for false positives?
More information about the Bogofilter