 About
POPfile...
2005-09-10: Efficiency
After a first quick training phase, I can report the actual efficiency
of this software tool. Strictly no problem here: 97,89% of successful
identification of SPAM with a training on only a little less than
200 messages (655 good words, 14564 spam words).
The result is exactly what was expected and satisfying. Good for
POPfile. It can be recommended on the basis of the detailed evaluation
you'll find on the right (pros/cons).
2005-09-03: First filtering
As should be expected the first filtering/classification left all
messages as unclassified, but after the first reclassification
using the web interface, it corrected itself nicely.
For the moment, the only inconvenience seems to be the lack of
a presentation of messages sorted by probability of being in the
category they were assigned to (it's a neat feature of K9 that saves
a lot of time in everyday management of the mail).
2005-09-02: Installation
The installation of POPfile under Windows is refreshingly simple.
I feared that the Linux origin may be seen in a rough installation
process. Nothing could be more wrong. It's zip. I appreciated the
full recognition of all my email programs (and their automagic configuration):
Bayesian
filtering of SPAMs
For a long time now, I have been advising individual users to protect
themselves with a software based on Bayesian statistical techniques.
My prefered PC-Windows solution for this still goes to K9
from keir.net.
But the program at the origin of K9, POPfile,
now has a specific Windows version. I am going to evaluate it in
order to decide if it should go to my
list of prefered downloads.
Some later news will be shown here.
|
Not far from POPfile...
POPfile:
pros and cons
While evaluating POPfile, here are the positive and negative sides
I could detect. I let you know:
Pros:
- Fully automated installation
- Good filtering efficiency.
- Able to handle more than SPAM (as many categories as you'd like)
- Works with Windows from 98 to XP and with Linux.
Cons:
- The presentation of messages (in order to reclassify them to
avoid/correct errors) does not help in concentrating on the near-the-limit
probabilities, despite that it would help management of large
quantities of email (as in my case).
- Sometimes, the User Interface (UI) of the Windows version will
seem to stop working. The user manual says this is only temporary
and happens when the software is classifying email. My experience
shows it leads to restarting POPfile instead.
Download POPfile.
Paul Graham's plan
A (rather long) paper from Paul Graham about why fight against
SPAM will necesarily use content evaluation and preferably Bayesian
techniques.
I say the same thing for quite some time, but Paul is much clearer
and argumented than I could be.
A
plan for SPAM
Some older news archived at SpamAnti.net
|
POPfile
on SpamAnti.net
