September 27, 2012 Leave a comment
- Here is a very nice survey about machine learning by Pedro Domingos. There is minimal math, instead he focuses on a high-level discussion of how machine learning people think about the world.
- Hal Daume’s blogpost has touched off a very fun discussion of the interaction between natural language processing, machine learning, and social science. In my personal opinion, language technology and social science are on a collision course, whether we like it or not. But I agree with Hal that a lot of the cross-over work that’s been done so far is not very interesting. In my opinion, NLP researchers have to start to engage with social science on a much deeper level. Much of the most interesting and durable NLP — including statistical NLP — has been done by people with a serious linguistics background. For example, my thesis advisor Regina Barzilay has gotten a lot of mileage out of being able to bridge the gap between state-of-the-art statistical methods and linguistically-serious models of discourse. I think we’re rapidly reaching the same point with computational social science. The future belongs to machine learning researchers who can engage seriously with the social science literature — and to social scientists who can understand the machine learning toolbox.