Вариационный иерархический автоэнкодер диалогов

AbstractRecent works have shown that generative data augmentation, where synthetic samples generated from deep generative models complement the training dataset, benefit NLP tasks. In this work, we extend this approach …...