The paper (http://www.cs.cmu.edu/~nlao/publication/2014.kdd.pdf) mentions the extracted knowledge base is about 38 times larger than DeepDive's, the largest previous comparable system.