Bayesian Filtering At Its Best!
- 0
- Add a Comment
Speaking as someone who has used his fair share of Open Source programs, the one program that truly sold me on the idea of Open Source software as being uber-cool had to have been POPFile. Up until this point, I had been spending hours and hours working to get my e-mail rules in Outlook to work just right.
Don’t misunderstand me now. E-mail rules are very powerful when they are being properly implemented. On top of that, you can get them to do all sorts of incredible things once you get the hang of them. Still, it was at that point in my computing life that I felt like there had to be a better way to sort my e-mail. One that would adapt to my needs, not the other way around.
Now keep in mind that this was a long time before Thunderbird even hit the mainstream. Besides that, I was just discovering the wonders of Mozilla (and its own e-mail client) at that point too, so my exposure to Open Source was still pretty new. Being that I felt as if I had taken abilities with e-mail rules as far as I was going to be able to get them, I decided to begin looking to the Open Source community for an alternative that would allow me to continue using my standard e-mail client. So it was after much research that I finally discovered Bayesian filtering.
My discovery of Bayesian filtering is actually what led me to POPFile, thus opening my mind to a whole new reality of possibilities. See, at that time, most people wanting to use POPFile were looking to gain some control over the growing spam problem. They had discovered that traditional e-mail filters were all but useless against spam and blacklisting was proving to be a losing battle as well.
While POPFile may have been a great tool for dealing with spam, I discovered that it was trainable in other areas as well. See, instead of wasting time with a dozen sets of rules for my e-mail client, I figured out that I could teach POPFile specific classifications for specific folders. Allow me to explain.
With POPFile, you set up what are referred to as ‘buckets.’ These buckets are for e-mails that are classified by a certain desire you may have for them. For instance, let’s say you have a bucket called ’spam.’ OK, so you have some incoming e-mail and one of them is what you consider to be spam. Since POPFile was just installed, it does not understand what you consider to be junk mail yet. Therefore, you will have to train it to meet your needs - much like in Thunderbird or Mozilla mail.
So you jump into the POPFile UI, find the reference to that specific e-mail and classify it as spam from there. OK, now let’s say you have an e-mail from ‘Uncle Aunty.’ Now obviously there is no way this program can possibly understand that you want this e-mail to go into the family folder, right? Wrong. You can actually teach POPFile how to distinguish your own family members from other e-mail that you receive. This is also the case with multiple accounts under one identity.
Now here is where it gets really cool. After a week or so of ‘training’ your POPFile install, this amazing program can actually begin to successfully ‘guess’ who you consider to be part of your family, thus putting them into an assigned folder! Yes, it will examine the content of the e-mail and make an executive decision with the information you have taught it. This can also be done with work versus home contacts, plus a zillion other possibilities. Folks, the sky is the limit. But you must remember, it will take time… that, and it only does what you teach it to do.
OK, now for the real question. How hard is this to set up? The install is easy. Configuration, however, was a little tricky back then unless you were a computer power user. Since you really need to have some understanding of setting up a your mail settings and other things of this sort, you may find that you are spending a bit of time with the documentation. This is OK, though. I just took a look and it appears that the help area has been greatly improved since last time I visited. With this in mind, I would check it out.
Today we take Bayesian filtering for granted thanks to Thunderbird and other similar e-mail clients with that sort of spam filtering capability. Unlike these other clients, however, POPFile gives us that little bit extra beyond the traditional e-mail filtering and allows us to do just about anything you can possibly imagine with your e-mail correspondence.
