I have recently been obsessing over the predictive abilities of Bayes theorem and its application to everyday computing. I have seen it work as an extremely effective and adaptive spam filter. I have even see it being used to determine whether or not a piece of text was written by a male or female. I am convinced that it can and should be used to adapt internet content to users based on their preferences, because it appears to be the most accurate and statistical way to predict the future based on the past. In the next few weeks I am going to be outlining my challenges, frustrations, and successes while attempting to apply a Bayesian filter to a dynamic content delivery system based on user preferences. I am not sure whether or not this has been done before because I don't know how last.fm or stumbleupon work, but I believe they probably only use general genre information and counting, because their content never appears very targeted.
permalink