Compression of CTC-Trained Acoustic Models by Dynamic Frame-Wise Distillation or Segment-Wise N-Best Hypotheses Imitation.

Haisong Ding, Kai Chen, Qiang Huo

Browse the full Interspeech paper archive.