Chris describes this as an AI-systems class. He'll cover e.g. the theory behind expectation maximization (actually, a (good) TA did that lecture this year), bu

Chris describes this as an AI-systems class. He'll cover e.g. the theory behind expectation maximization (actually, a (good) TA did that lecture this year), but isn't afraid to talk about special casing stuff, and one reliable way to get extra credit is to implement the core algorithm correctly and then come up with performance-improving hacks based on insight into the quirks of English. This makes it a slightly unusual class; if you are actually interested in building a functional NLP product, this practicality is a good thing.