Order-AgnosticCrossEntropyforNon-AutoregressiveMachineTranslationCunxiaoDu1ZhaopengTu2JingJiang1AbstractAnumberofrecenteffortshaveexploredwaystoimprovetheNATmodels’abilitytohandlemultimodality.One...
MixedCrossEntropyLossforNeuralMachineTranslationHaoranLi1WeiLu1AbstractSourceDieGeburtenrategehtweiterzurück.one-hotmodelgroundpredictiontruthInneuralmachinetranslation,Crossentropy(CE)isthestanda...
AlignedCrossEntropyforNon-AutoregressiveMachineTranslationMarjanGhazvininejad1VladimirKarpukhin1LukeZettlemoyer1OmerLevy1AbstractTargetYittastesprettygoodthoughNon-autoregressivemachinetranslationm...