Winnow efficiently trains and operates any number of unique Naive Bayes classifiers on large sets of content. Written in C, runs on any Unix, very high performance, works with very small
training and unbalanced training sets. Winnow powers http://winnowtag.org, an innovative web feed reader that uses smart tags that learn and find the content you want to see, from more sources than you can follow with traditional feed readers. Subscribe, create, and share tags. Add feeds, import feed lists, publish tags as feeds. http://winnowtag.org shows Winnow’s accuracy and performance searching over 3/4 million items updated daily from 8,000 feeds. Winnow works particularly well with Ruby and Ruby on Rails.