Speaker consistency loss and step-wise optimization for semi-supervised joint training of TTS and ASR using unpaired text data

Re-assign community