The Observatorium – The structure of news: topic monitoring in online media with mutual information – David M.S. Rodrigues, In European Conference on Complex Systems 2010 – ECCS10; Lisbon 2010, Sept. 13-17 Setptember 2010. Slides
Abstract:Large, real time text classification systems are becoming a popular topic. We present a method for automatically extracting correlated news from online media using a dynamic similarity graph and use the variation of information as a measure to identify topics, lifespan and key terms. The presented method has the advantage of requiring no human intervention or training and having no pre-assigned categories because they emerge from the dynamics of the generated network.
If you could not attend my talk at ECCS’10 and still are a bit curious, here are paper and slides for the talk.