Search papers, labs, and topics across Lattice.
1
0
1
Despite superior head prediction accuracy, learned heads fail to consistently improve constituency parsing, especially when evaluated on punctuation, challenging the assumption that better headedness directly translates to better parsing.
