bogotune bugs?

[David Relson]

On Sun, 20 Mar 2005 00:29:12 +0000 (UTC)
Jason Lunz wrote:

> does bogotune have known bugs? I'm using debian sarge, and I've tried
> both 0.94.0 and 0.94.1. No matter what I do, it says:
> 
> [mudd](0) % bogotune -n ham -s spam
> Calculating initial x value...
> Error: /home/lunz/.bogofilter/wordlist.db is not a directory.
> bogotune: cannot find bogofilter directory.
> You must specify a directory on the command line, in the config file,
> or by using the BOGOFILTER_DIR or HOME environment variables.
> Program aborting.
> Initial x value is 0.541073
> Too few high-scoring non-spams in this data set.
> [snip]
> 
> there's lots of other output, but i'm not inclined to trust it or try to
> figure out what it means if bogotune can't even seem to read the
> wordlist.
> 
> What are the error messages supposed to mean? Of course
> "/home/lunz/.bogofilter/wordlist.db is not a directory". It's not
> supposed to be.
> 
> I've tried different combinations of using BOGOFILTER_DIR, -d, giving
> the name of either wordlist.db itself or the directory it's in, and
> nothing works.
> 
> any ideas?
> 
> Jason

Hi Jason,

Bogotune is rather particular about how it's run.  It tries a wide
range of possible values of bogofilters scoring parameters and needs
quite a bit of input data in order to do produce meaningful results.

There are minimum message counts for the wordlist as well as the ham
and spam used in tuning.  There should be at least 2000 ham and 2000
spam in wordlist.db and a different 2000 ham and 2000 spam for the
tuning messages.

It does a lot of work with these messages, ultimately testing several
thousand possible parameter sets, which means each message gets scored
thousands of times.  To get all this done in a reasonable amount of
time, i.e. hours rather than days, it keeps all the data in RAM.

I just started my standard test case and it's running properly.

Try running it as "bogotune -vv -C -d directory/ -n ham -s spam" and
let us know how it goes.

HTH,

David


_______________________________________________
Bogofilter mailing list
Bogofilter at bogofilter.org
http://www.bogofilter.org/mailman/listinfo/bogofilter