PC Review


Reply
Thread Tools Rate Thread

Re: Spam filtering tools - Bayesian filtering

 
 
Aaron
Guest
Posts: n/a
 
      24th Aug 2003
"David K. Wall" <(E-Mail Removed)> wrote in
news:Xns93E01BC74ADF8dkwwashere@216.168.3.30:

> I was looking around pricelessware.com and noticed that two useful
> Bayesian mail filtering programs aren't included.
>
> http://popfile.sourceforge.net/ POPFile - written in Perl
> http://spambayes.sourceforge.net/ SpamBayes - written in Python


Actually popfile IS included in 2003. Under "email manager" not email
spam tool.

Spambayes is newer probably wasnt around last year.

> Both will run on Windows and *nix. You do not need to install Perl of
> Python to run them. They come with a minimal interpreter for the
> language they use.


Really? I had to download and install python for spambayes to try.

> Here is a review of both:
> http://home.dataparty.no/kristian/reviews/bayesian/ It's not quite
> fair, in a way, since popfile does more than just filter spam, it
> filters email into a number of "buckets", so you can classify your
> mail into more categories than just "spam" and "not-spam". On the
> other hand, Spambayes has an "uncertain" category that can mean fewer
> false positives.


> (I haven't used either of these programs <snip>


I have used popfile and SAproxy for a long while. I'm currently trying
out Spambayes. First impressions is that it's faster to setup since you
can feed ham/spam directly through the webinterface You can do that in
Popfile too, but it's harder.




Aaron
--
Want to learn how to use Winboard and the 150+ free Winboard
Chess engines?Visit http://www.aarontay.per.sg/Winboard/
 
Reply With Quote
 
 
 
 
Iain Cheyne
Guest
Posts: n/a
 
      26th Aug 2003
Aaron <(E-Mail Removed)> wrote in news:Xns93E1AE88F5EB8unknown23
@203.116.1.193:

>> Both will run on Windows and *nix. You do not need to install Perl of
>> Python to run them. They come with a minimal interpreter for the
>> language they use.

>
> Really? I had to download and install python for spambayes to try.


There's an Outlook plugin that doesn't need a separate python installation.
It's great and I use it all the time. I much prefer it to K9 or Popfile.

http://starship.python.net/crew/mhammond/spambayes/

--
Iain Cheyne
Remove the numbers and change "invalid" to "net" to reply.
Please check www.pricelessware.org, the FAQ and Google Groups before
posting in alt.comp.freeware.
 
Reply With Quote
 
Aaron
Guest
Posts: n/a
 
      27th Aug 2003
Iain Cheyne <(E-Mail Removed)> wrote in
news:Xns93E3DB9765B0Fiaincheynenet@195.149.20.147:

> Aaron <(E-Mail Removed)> wrote in
> news:Xns93E1AE88F5EB8unknown23 @203.116.1.193:
>
>>> Both will run on Windows and *nix. You do not need to install Perl
>>> of Python to run them. They come with a minimal interpreter for the
>>> language they use.

>>
>> Really? I had to download and install python for spambayes to try.

>
> There's an Outlook plugin that doesn't need a separate python
> installation.


Oh.. Outlook plugin, that explains it.

It's great and I use it all the time. I much prefer it
> to K9 or Popfile.


Any particular reason? besides it being integreted? The POPfile equalavant
is Outclass btw.

My own experiences with the real spam bayes were not as fortunate, as i was
hit by a corrupt database bug.I'm told they are still tracking/figuring
out.

Well that's to be expected cos spambayes is still in "alpha" ?





> http://starship.python.net/crew/mhammond/spambayes/
>




Aaron
--
Want to learn how to use Winboard and the 150+ free Winboard
Chess engines?Visit http://www.aarontay.per.sg/Winboard/
 
Reply With Quote
 
Jari Lehtonen
Guest
Posts: n/a
 
      27th Aug 2003
On 26 Aug 2003 20:35:15 GMT, Iain Cheyne <(E-Mail Removed)>
wrote:

>There's an Outlook plugin that doesn't need a separate python installation.
>It's great and I use it all the time. I much prefer it to K9 or Popfile.
>
>http://starship.python.net/crew/mhammond/spambayes/


I have been using that for some days now and it shows 100% results
(all spam blocked, good mail saved) after just one single learning
session! It is incredible!

I was trying Ihatespam, but it leked despite is's fine rulesets etc.
Spambayes does not leak.

jari

 
Reply With Quote
 
Iain Cheyne
Guest
Posts: n/a
 
      29th Aug 2003
[Spambayes outlook plugin]

> It's great and I use it all the time. I much prefer it
> > to K9 or Popfile.

>
> Any particular reason? besides it being integreted? The POPfile equalavant
> is Outclass btw.


Apart from the excellent integration, I like the Possible Spam folder.
It kills false positives completely.

Popfile was clunky to use and interfered with the Proxomitron. I might
try Outclass as I liked the way Popfile is able to have several
different buckets, which are not necessarily spam-based.

K9 simply was not accurate enough. I tried rebuilding the database,
but it was still useless.

--
Iain Cheyne
 
Reply With Quote
 
Aaron
Guest
Posts: n/a
 
      29th Aug 2003
(E-Mail Removed) (Iain Cheyne) wrote in
news:(E-Mail Removed):

> [Spambayes outlook plugin]
>
>> It's great and I use it all the time. I much prefer it
>> > to K9 or Popfile.

>>
>> Any particular reason? besides it being integreted? The POPfile
>> equalavant is Outclass btw.

>
> Apart from the excellent integration, I like the Possible Spam folder.
> It kills false positives completely.


Yes, that's what sets spambayes apart from other bayesian filters.
However, does this mean you trust it enough just to delete the mails
tagged by spambayes as spam without looking? If not, it's kind of
pointless, since you will be looking at 2 folders instead of one.

My first attempt at using the "real" spambayes (sans outlook plugin) was
not succuessful because i ran into a bug that corrupted the database, i
was told they are still tracking the bug.

Oh well, it's listed as still being in prerelease, that's why... I'll
give it another try when they release another version


> Popfile was clunky to use and interfered with the Proxomitron.


Not that i noticed. Simple enough to change the port that popfile works
on.



Aaron
--
Want to learn how to use Winboard and the 150+ free Winboard
Chess engines?Visit http://www.aarontay.per.sg/Winboard/
 
Reply With Quote
 
siDetRaked
Guest
Posts: n/a
 
      29th Aug 2003
On 29 Aug 2003 01:18:17 -0700,
(E-Mail Removed) (Iain Cheyne) wrote:

>Popfile was clunky to use and interfered with the Proxomitron.


Both Popfile and Prox use port 8080. I changed Prox to port 8081 and
everything works fine.

Cheers!

--
siDetRaked
since 1938


 
Reply With Quote
 
Iain Cheyne
Guest
Posts: n/a
 
      29th Aug 2003
>>> Any particular reason? besides it being integreted? The POPfile
>>> equalavant is Outclass btw.

>>
>> Apart from the excellent integration, I like the Possible Spam
>> folder. It kills false positives completely.

>
> Yes, that's what sets spambayes apart from other bayesian filters.
> However, does this mean you trust it enough just to delete the mails
> tagged by spambayes as spam without looking? If not, it's kind of
> pointless, since you will be looking at 2 folders instead of one.


I do trust it. There is an ini file you can edit to hide unread message
notifications. Unfortunately, Microsoft will not allow developers to
remove the new mail envelope from the systray.

>> Popfile was clunky to use and interfered with the Proxomitron.


> Not that i noticed. Simple enough to change the port that popfile
> works on.


I did that, but it was worth pointing out.

--
Iain Cheyne
Please check www.pricelessware.org, the FAQ and Google Groups before
posting in alt.comp.freeware.
 
Reply With Quote
 
John Fitzsimons
Guest
Posts: n/a
 
      30th Aug 2003
On 29 Aug 2003 01:18:17 -0700,
(E-Mail Removed) (Iain Cheyne) wrote:

< snip >

>Popfile was clunky to use and interfered with the Proxomitron.


No problems here. They just need to use different ports.

>I might
>try Outclass as I liked the way Popfile is able to have several
>different buckets, which are not necessarily spam-based.


>K9 simply was not accurate enough. I tried rebuilding the database,
>but it was still useless.


I found K9 hugely inaccurate (compared to PopFile) for quite a while.
A recent release however changed the way it worked to the way PopFile
works. I now get 99%+ accuracy with K9.

Regards, John.

--
****************************************************
,-._|\ (A.C.F FAQ) http://clients.net2000.com.au/~johnf/faq.html
/ Oz \ John Fitzsimons - Melbourne, Australia.
\_,--.x/ http://www.aspects.org.au/index.htm
v http://clients.net2000.com.au/~johnf/
 
Reply With Quote
 
jack horsfield
Guest
Posts: n/a
 
      30th Aug 2003
On 29 Aug 2003 19:10:35 GMT, Iain Cheyne <(E-Mail Removed)>
wrote:


>I do trust it. There is an ini file you can edit to hide unread message
>notifications. Unfortunately, Microsoft will not allow developers to
>remove the new mail envelope from the systray.
>


somewhere in the spambayes docs you will find an ini file setting:


delete_as_spam_message_state:Read


that removes the new mail envelope for you.


FWIW, i have been using spambayes since the end of May. i now have
1800 spam messages and spambayes is working well. it has never put a
good message into spam, but it has put a number into Possible. these
days i trust it.


jack
We are certainly moving with alacrity to the future
 
Reply With Quote
 
 
 
Reply

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Outlook 2003 Junk filter not filtering SPAM, filtering items from =?Utf-8?B?S3lsZSBPcnI=?= Microsoft Outlook Discussion 0 26th Apr 2006 08:13 PM
spam filtering Tony Stewart Microsoft Outlook 1 3rd Sep 2004 03:58 PM
Bayesian SPAM filtering Jim Adams Microsoft Dot NET 0 23rd Feb 2004 05:32 PM
spam filtering Matt Microsoft Outlook Discussion 0 26th Dec 2003 01:43 AM
Spam filtering Ian Pendlebury Microsoft Outlook 2 27th Oct 2003 05:43 PM


Features
 

Advertising
 

Newsgroups
 


All times are GMT +1. The time now is 12:20 AM.