I'm looking for some thoughts from you natural-language processing people out there. For my undergraduate NLP/CL class this fall, I'm thinking about assigning final projects in a new way. Instead of having students design their own projects, I'd provide a small set of classic NLP papers, say five or six. Each student would replicate the result in the paper, and if inspired, extend it a bit.
Now for the question: What papers do you think would be appropriate for such an exercise. To qualify, a paper should introduce a relatively simple method, one reimplementable in a couple of weeks; operate over available data or data that can be easily constructed; be amenable to a quantitative evaluation; and generate a satisfying feeling upon completion.