MESSAGE
| DATE | 2004-01-26 |
| FROM | Mike Richardson - NYLXS PRESIDENT
|
| SUBJECT | Subject: [hangout] What are Bayesain spam filters? and Re: [wwwac] The End of Spam as
|
On Mon, 26 Jan 2004, Michael L. Richardson wrote:
> On Mon, 2004-01-26 at 02:25, John T. Haller wrote:
> > hens wrote:
> >
> > > Interesting article in the NYTimes this Sunday on the meaning of all that
> > > gibberish in your Inbox lately:
> > >
> > > http://www.nytimes.com/2004/01/25/weekinreview/25john.html
> > >
> > >
> > > Comments? Are spam filters working?
> >
> > Short answer: yes, if you install them. Bayesian spam filters work
> > well, after just a small amount of training. After a lot of training,
> > they work nearly perfectly... on the order of 99.9% effectiveness over
> > 4000 emails.
> >
> > For those unfamiliar, Bayesian spam filters analyze the words and code
> > appearing in each email and assign scores to them, based on messages you
> > have told it are spam and not spam. It learns as you feed it more spam
> > and more legit email. Most importantly, it learns your personal real
> > email and spam email. There are numerous implementations. I've found
> > SpamBayes to be one of the best. I use it in proxy form. But there's
> > also an easy-to-use Outlook plugin.
> >
> > Unfortunately, Bayesian filters aren't available in any major email
> > server. Most of the tools are client-side (as the scoring system works
> > best for an individual user... and may not work as well across a
> > domain). In addition, you still end up downloading and storing the
> > email until the filter is run, so you're still paying for it. You're
> > just saved the time of reading it.
> >
> > If at all interested in how it works, I recommend reading:
> > http://www.paulgraham.com/spam.html
> > http://www.paulgraham.com/better.html
> >
> > Regards,
> > John
>
____________________________
NYLXS: New Yorker Free Software Users Scene
Fair Use -
because it's either fair use or useless....
NYLXS is a trademark of NYLXS, Inc
|
|
