Speech representation learning: Learning bidirectional encoders with single-view, multi-view, and multi-task methods