EL>> what about that new statistical spam filter(using bassian alg)
EL>> did anyone try it out?

If you mean bayesian filters like bogofilter, I'm using it (bogofilter) at
home and at work and so do a number of my colleaugues and the results are
rather satisfying. The downside of the filter is that it requires training
- it took me about a week to train it to level where I was satisfied,
though it started to give tolerable results immediately, because I had
large arvhives of both legitimate mail and spam (don't ask me why I was
archiving spam, I don't really know, but it came handy).
For now, it has no more than one-two false negatives per day (daily spam
amount is about 50 to 100 items, daily mail amount about 400-500 mails)
and usually not more than one false positive per day, which usually
is non-routine mail (i.e., not from lists or people you are
communicating with frequently) and something that 'looks like' spam (e.g.,
has a lot of HTML, 'spam' words, etc.). This means you still have to skim
over the spam folder about once a day or to employ some additional
mechanism. But I find the results of using bogofilter highly satisfying.

However, this kind of filter seems to be rather for personal use - because
if I would pass our sales/marketing communication through my database as
it is now, a lot of legitimate mails would end up in 'spam' bin, since
they talk exactly about the same things that many spams talk - prices,
sales, discounts, "commercial opportunity of your life"s, etc. So
statistics of different person to what should be considered spam is very
different and this filter has to be personally tuned. This makes it not
very suitable to company-wide deployment, for example - though I see that
on the bogofilter list people are talking about this and may come to some
good solution.
-- 
[EMAIL PROTECTED]      \/  There shall be counsels taken
Stanislav Malyshev      /\  Stronger than Morgul-spells
phone +972-50-624945    /\              JRRT LotR.
whois:!SM8333




=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]

Reply via email to