Gary Robinson: Gary Robinson's Spam RantsA Weblog containing links and occasional rants on spam.(Untitled)- August 28, 2003 "Dennis K. Berman offers a ray of hope to people frustrated and demoralized by the spread of spam: Spamming could eventually burn itself out by becoming a victim of its own proliferation, he muses." From the ACM.org's summary of a Wall Street Journal article. The summary is worth checking out.http://radio.weblogs.com/0101454/categories/spam/2003/08/28.html#a509 (Untitled)- August 23, 2003 Well I'm back in Maine. I actually got a fair amount of work done on my laptop on the drives to and from New York City. Partly on the drive south, I developed and tested some changes to the chi-square approach to spam detectionand testing so far the changes seem to result in roughly a 5% improvement in accuracy. On the drive back I developed another possible improvement that I hope to find time to test over the weekend.http://radio.weblogs.com/0101454/categories/spam/2003/08/23.html#a501 (Untitled)- August 11, 2003 SpamBayes, which is based on math suggested by me as an improvement to the approach in Paul Graham's seminal "A Plan For Spam" article, and supplemented with a great observation from Tim Peters that the "Unsure" range of spamham classification is very useful, won this comparative testing against PopFile. The test was slashdotted today. John Graham-Cumming, who is responsible for PopFile, Tim Peters and I are all members of ActiveState's Anti-Spam Task Force. I must admit I'm competitive...http://radio.weblogs.com/0101454/categories/spam/2003/08/11.html#a494 |