I actually hadn't seen that one (I did know of the "Buffalo buffalo" family of sentences). Regarding observation of the phenomenon, in a way I first observed t

I actually hadn't seen that one (I did know of the "Buffalo buffalo" family of sentences). Regarding observation of the phenomenon, in a way I first observed the inverse of it. My first proofs of concept didn't learn probabilities, only structure. (One might say all edges had the same probability.) Yet aesthetically they performed just as well as versions which included probability in the algorithm.