Unified Speech-Text Pre-training for Speech Translation and Recognition

39270

Knowledge Triples

Source Entity	Relationship	Target Entity
Joint Pre-training Method	uses	Multi Task Learning Method
Multi Task Learning Method	includes	Self-Supervised Task Task
Multi Task Learning Method	includes	Supervised Text-to-Text Task Task
Multi Task Learning Method	integrates	Speech Recognition Task
Speech Recognition Task	comparesWith	wav2vec 2.0 Method