Pretext Tasks selection for multitask self-supervised speech representation learning