I was reading about n-gram smoothing and thinking how you would approach that from an ML point of view. But how would you be ...