selected publications blog posting Parallel WaveNet: Fast High-Fidelity Speech Synthesis. arXiv (Cornell University). 2017