Did you ever wonder how Owlin is able to gather articles from a number of languages and sort them by subject? This is made possible in part by a statistical language model developed by Owlin in cooperation with scientists from Stanford University and Edinburgh University. The 5.6 terabyte model, which brings natural language processing to a new level, is now the subject of a scientific article written by Owlin’s Bas van Ooyen with co-developers Christian Buck of the University of Edinburgh and Kenneth Heafield of Stanford University and presented last week at the Language Resources and Evaluation Conference in Reykjavik Iceland.