New paper “Self-training with Two-phase Self-augmentation for Few-shot Dialogue Generation” is out. [Paper][Code]