CoBERT: Self-Supervised Speech Representation Learning Through Code Representation Learning | AMiner