no code implementations • EAMT 2022 • Dongqi Pu, Khalil Sima’an
Existing syntax-enriched neural machine translation (NMT) models work either with the single most-likely unlabeled parse or the set of n-best unlabeled parses coming out of an external parser.