Fetching the paper…

STEMM: Self-learning with Speech-text Manifold Mixup for Speech Translation · Around