Pruned RNN-T for fast, memory-efficient ASR training.

Fangjun Kuang, Liyong Guo, Wei Kang, Long Lin, Mingshuang Luo, Zengwei Yao, Daniel Povey

Browse the full Interspeech paper archive.