Learning Hierarchical Representations for Expressive Speaking Style in End-to-End Speech Synthesis | IEEE Conference Publication | IEEE Xplore