Tacotron-Based Acoustic Model Using Phoneme Alignment for Practical Neural Text-to-Speech Systems | IEEE Conference Publication | IEEE Xplore