|
About:
SpamProbe is a spam detection program that uses a
Bayesian analysis of the frequencies of terms used
in the email. Because it filters email based on
content rather than on general rules, it easily
adapts itself to the types of email that each
individual user normally receives.
Author:
Brian Burton [contact developer]
Homepage:
http://spamprobe.sourceforge.net/
Tar/GZ:
http://prdownloads.sourceforge.net/[..]spamprobe/spamprobe-1.4b.tar.gz?download
Trove categories:
[change]
Dependencies:
[change]
Berkeley DB (required)
Libungif (recommended)
PBL ISAM (recommended)
[download links]
|
|
» Rating:
8.56/10.00
(Rank N/A)
» Vitality: 0.01% (Rank 3631)
» Popularity: 2.40% (Rank 2014)
 
(click to enlarge graphs)
Record hits: 26,579
URL hits: 11,506
Subscribers: 56
|
|
Branches
Articles referencing this project
Comments
[»]
Update from version 1.0a to 1.2a not useful
by tmstaedt - Dec 7th 2005 13:31:51
Hallo,
spamprobe 1.2a somehow blows it completely for me. Bad, bad results, while
1.0a was and is pretty good, albeit somewhat slow.
Regards,
Thomas Mittelstaedt
tmstaedt @ t-mittelstaedt . de
[reply]
[top]
[»]
Update from version 1.0a to 1.2a *very* useful
by tBB - Dec 8th 2005 01:35:24
Hi,
completely different experiences over here. At least when the PBL DB is
used SpamProbe belongs to the fastest Bayesian classifiers. The accuracy
in 1.2a is excellent as well. Even better in 1.3.x btw.
Regards, Nico
[reply]
[top]
[»]
Re: Update from version 1.0a to 1.2a *very* useful
by tmstaedt - Dec 20th 2005 11:57:56
Nico,
Well, I made a mistake when building spamprobe. My database is Berkeley db
and I somehow built it with that new hash feature enabled.
Here is the configure command, that built it correctly, I assume:
LDFLAGS=-L/usr/local/lib CPPFLAGS=-I/usr/local/include/db3/ ./configure
--enable-cdb --with-db=/usr/local/lib/
Regards,
> Hi,
> completely different experiences over
> here. At least when the PBL DB is used
> SpamProbe belongs to the fastest
> Bayesian classifiers. The accuracy in
> 1.2a is excellent as well. Even better
> in 1.3.x btw.
>
> Regards, Nico
[reply]
[top]
[»]
Re: Update from version 1.0a to 1.2a *very* useful, still blows it!!
by tmstaedt - Dec 20th 2005 12:04:43
Even with the correct build, when I run spamprobe find-good on my latest
spam mails, I get totally miserable results compared to 1.0a! Strange! Do
I have to retrain this new version with train-spam?
> Nico,
>
> Well, I made a mistake when building
> spamprobe. My database is Berkeley db
> and I somehow built it with that new
> hash feature enabled.
> Here is the configure command, that
> built it correctly, I assume:
>
> LDFLAGS=-L/usr/local/lib
> CPPFLAGS=-I/usr/local/include/db3/
> ./configure --enable-cdb
> --with-db=/usr/local/lib/
>
> Regards,
>
>
> % Hi,
> % completely different experiences over
> % here. At least when the PBL DB is
> used
> % SpamProbe belongs to the fastest
> % Bayesian classifiers. The accuracy in
> % 1.2a is excellent as well. Even
> better
> % in 1.3.x btw.
> %
> % Regards, Nico
>
>
>
[reply]
[top]
[»]
Re: Update from version 1.0a to 1.2a *very* useful, still blows it!!
by Brian Burton - Dec 20th 2005 12:16:30
> Even with the correct build, when I run
> spamprobe find-good on my latest spam
> mails, I get totally miserable results
> compared to 1.0a! Strange! Do I have to
> retrain this new version with
> train-spam?
1.2 uses a different parser than 1.0 so there would be a benefit to
updating your database using 1.2a to get the new terms found by the new
parser. I don't know what "totally miserable" means to you but if you are
saying that 1.2a isn't finding any spams then clearly there is a problem
with your database. 1.2 is at least as accurate as 1.0 in all of my
tests.
[reply]
[top]
|
