can't train bogofilter

[David Relson]

Hello Renato,

There are several problems with what you're doing.


The "-N" option increases ham counts in the wordlist and "-s"
decreases spam counts in the word list. Using "-Ns" is the proper flag
combination for reclassifying a spam message as ham, i.e for correcting
a false positive.

Use "-nS" when the message has been classified as ham, but should
have been classified as spam.

Bogofilter's unsure classification indicates that bogofilter can't
determine if the message is ham or spam.  So, when bogofilter
classifies a message as Unsure, the message's words are not entered in
the wordlist. To train an unsure message as spam, use "-S".

The flag combinations "-Ns" and "-Sn" will not change the wordlist
size.  As they are classification corrections, the words in the
messages being used are already in the wordlist.  These flag
combinations change counts for the already existing wordlist
entries.

 Training bogofilter does not change the input file(s) used.  You need
to run those messages through bogofilter again to see how the
classification will change.

You can use bogoutil to see a token's counts.  Command 

	bogoutil -p <path to wordlist> word1 word2 etc

will give you the spam and ham counts for word1, word2, etc

Hope this helps!

David

Messages c On Sun, 10 Jun 2012 12:40:55 +0200 renato wrote:

> Hello, I have a mailbox folder with spam mails that bogofilter has
> incorrectly filed as "Unsure" or "Ham", and I'd like to train it. So
> in the directory I gave this command:
> 
> for i in $(ls); do bogofilter -Ns < $i; done
> 
> which takes a while, but after that all messages still have an
> unmodified X-Bogosity header (i.e. it still reads "Unsure" or "Ham") -
> is this the supposed behaviour? If I also give the -p option, the
> message is written to stdout with a correct X-Bogosity header (i.e.
> spam), but my bash-fu is weak and I can't figure out how to correctly
> redirect this output to overwrite the original file - and most
> importantly I don't know if that's really what I want to/should do.
> 
> Also, after that command the file ~/.bogofilter/wordlist.db hasn't
> increased in size.
> 
> What am I missing?
> 
> cheers,
> renato
> _______________________________________________
> Bogofilter mailing list
> Bogofilter at bogofilter.org
> http://www.bogofilter.org/mailman/listinfo/bogofilter