I think I'll actually review the paper. Because I think it'll make a great use-case for the argument that you can't automated-sentiment-analysis your way to a cogent political assessment of entire populations. No matter how hard you want to.