User makes e-mail filter super

I get a large amount of unsolicited e-mail from public-relations firms and independent sources pitching all kinds of products and services, with the hope that I will write or talk about them.

Over the years, the hundreds of spams a day slowly have grown into the thousands. I now get about 14,000 spam messages a day.

With that amount of junk e-mail a day, my way of dealing with it has caused some changes, including how I manage my computer.

For example, I used to turn off my computer at the end of the day. Now I have to leave it on all the time because when I turn it off, the process of downloading the thousands of e-mails that accrued during the night took nearly an hour.

And by the time that hour of downloading finished, so much more spam had arrived that it took another half-hour or so for that batch to load.

When that finished, the PC needed more time for the next batch, and so on. Eventually, I'd catch up, with the process taking more than two hours. Then, I had to sift through them. It was just a mess, and I had to do something.

Don't think I didn't already have junk e-mail filters in place; of course I did. But the junk filters that come with mail programs just aren't up to the job when the numbers are so big. I needed something else.

Until recently, there was nothing in sight. I couldn't use challenge-response systems, which ask senders to verify themselves by responding to a text image.

Too many were simply refusing to be bothered with the additional burden, and I was missing out.

White lists were no good for me because so many e-mails were unsolicited. I can't white list someone I don't know.

What I needed was some kind of super junk e-mail filter that would somehow know what was bad and good. They exist and they all use something called Bayesian filtering.

Bayesian spam filtering is a type of statistical analysis that calculates the probability of an e-mail being spam based upon its contents.

But the heart of what makes this work is the person using it.

As e-mail comes in, you train the filter by telling it what is and isn't spam. The more you train it, the better its accuracy in identifying junk e-mail.

The one I use on my Macintosh is SpamSieve. It takes about 300 or so e-mails to effectively train the filter, in my case a cakewalk.

In just a few days, the accuracy became so good that SpamSieve's accuracy in detecting spam on my computer is now at about 99.8 percent. Now I have my computer back.

Well, almost. I still have to download everything so I still have to keep my computer on all the time, but in the morning I see only my real e-mail. It's a thing of beauty.

The other thing I like about SpamSieve is that it integrates with all of the popular e-mail programs like Apple Mail and Entourage.

If you decide to change mail clients, you don't have to retrain SpamSieve; it can apply the same statistics it's accrued to the different e-mail applications.

I went from using Apple Mail back to Entourage and SpamSieve didn't skip a beat. I still had all the filtering in place.

So if spam has you down, I suggest you try an e-mail filter that uses Bayesian filtering.

On the Mac, SpamSieve is my recommendation.

There are several filters that use Bayesian filtering available for Windows and can be found with a simple Google search.

But no matter which one you decide to use, you can be sure that your spam problem will be greatly reduced if not completely eliminated.

SpamSieve is for Mac OS X, sells for $30 and is available at c-command.com./spamsieve.