Jam Master Jay, RIP

I'm blacking out for the rest of the day in remembrance of Jason Mizell, aka Jam Master Jay.

Jam Master Jay, the D.J. who provided beats and scratches to the rap group Run-DMC's groundbreaking records, was shot and killed in a recording studio in Queens on Wednesday night. [...]

For most of its history, rap has been criticized for promoting violence, and several rappers who sang the praises of the gangster life, including Tupac Shakur and the Notorious B.I.G., were murdered. But Run-DMC and Jam Master Jay, all middle-class natives of Hollis, Queens, a mile or so from where Mr. Mizell was shot, created rap with a social conscience, urging listeners (between boasts) to stay in school, fight prejudice and respect one another. [NYTimes via Adam Curry]

I saw Run D.M.C. in '93, part of the 19-Naughty-3 tour. One act cancelled at the last second and both of the headline acts stank, with Naughty by Nature practically boo'd off the stage. Run D.M.C. took up the slack, coming back for a second high-energy performance. They "rocked the house" for over two hours.

Run D.M.C. turned an otherwise miserable concert into one that I've always remembered fondly.

3:14:26 PM | Comments:

Airport security leads to topless checkpoint

A French tourist got so fed up with having her chest wanded by airport security in the USA that she took off her shirt and bra to demonstrate her bomb-and-boxcutter-free chestular region. The airport was closed for 10 minutes. Under the USAPATRIOT Act, she faces up to three years in jail. Link (German-English translation here: Link) [via Boing Boing Blog]

This would be funny if it weren't so friggin' sad. America has become such a fearful place that even a half-naked woman threatens us.

When I lived in Europe, I made a concious effort not to blend in. I was a proud American and rarely wanted to conceal that fact.

These days I dream of blending in somewhere and never looking back.

2:36:05 PM | Comments:

POPFile, Part II

I got POPFile working, at another user's suggestion I exported some mail folders as .csv text files instead of individual .msg binary files. So far I've put 5500 messages into the corpus, with about 5% being known spam and half of the remainder coming from 16 mailing lists.

I'm keeping all of my Outlook rules in place until I am confident in POPFile's classifications. I've added two rules for POPFile, for spam and a mailing list that I just joined. The new list is a good test of how quickly POPFile can be taught. The intial corpus was just 15 messages, so far it has correctly classified 3 out of 5 new messages.

One problem I see with teaching POPFile is that the web interface only allows for negative reinforcement, ie: this message is classified wrong, it should be this. For a small corpus, my gut feeling is that positive reinforcement would be more beneficial. There's probably a tipping point where that sort of feedback loop would have a negative affect on accuracy, but that is something for a math genius to figure out.

POPFile's author will be on TechTV today at 19:00 Eastern.

I wish I'd known about these types of filtering programs years ago. From 1998 until 2001, I would receive 10,000 messages on a normal day and several times that on really bad days. I needed over 100 Outlook rules to manage the chaos and focus my attention on the 5% that mattered to me. ifile was first released in late 1996.

1:16:30 PM | Comments: | Topics: bayesian spam 

POPFile

Decided to give POPFile a whirl tonight. Exported a few thousand messages from Outlook 2002, upgraded to the latest version of Perl, uninstalled SpamNet.

The program that builds the corpus doesn't seem to like me. It acts like it is importing my messages but the corpus file winds up empty. I suspect that it has problems with Outlook's binary message format, in spite of the documentation saying otherwise. Tried the latest version from CVS, no difference.

I filed a bug report, we'll see what happens...

1:24:00 AM | Comments: | Topics: bayesian spam