Advancing Acoustic-to-Word CTC Model With Attention and Mixed-Units | IEEE Journals & Magazine | IEEE Xplore