massive false negatives

[David Relson]

At 12:53 PM 5/5/03, Dan Stromberg wrote:

>Ever since I upgraded to 0.11.1.3, I've been getting a Lot of false
>negatives.  In fact, I'm not sure Anything is getting filtered out.
>
>I recently recreated my db from known spam and ham, but that didn't
>appear to help.
>
>I did:
>
>/dcs/packages/bogofilter/bin/bogofilter -v < /tmp/bad
>X-Bogosity: No, tests=bogofilter, spamicity=0.000001, version=0.11.1.3
>
>...with what I consider fairly obvious spam (the "banned cd" thing).  Is
>that a strange spamicity.  The man page didn't appear to say.
>
>I'm attaching -vvv output, which I'm not sure what to do with.

Dan,

I need some more information.  As part of bogofilter-0.12, bogofilter 
understands a new file format, known as the msg-coutn format.  A msg-count 
file contains the .MSG_COUNT values from spamlist.db and goodlist.db and 
all the tokens from the message, along with their spam and ham 
counts.  Given your spam message in the .mc format, I can run bogofilter 
and reproduce your numbers to learn more about what is happening.

Run the attached script to convert the spam message to msg-count format, 
i.e. "bogolex.sh < spam > spam.mc", then gzip the output file and send it 
to me.

David
-------------- next part --------------
A non-text attachment was scrubbed...
Name: bogolex.sh
Type: application/octet-stream
Size: 359 bytes
Desc: not available
URL: <http://www.bogofilter.org/pipermail/bogofilter/attachments/20030505/31722f66/attachment.obj>