ASAPP-ASR: Multistream CNN and Self-Attentive SRU for SOTA Speech Recognition
This work improves speech recognition accuracy for applications like voice assistants, though it is incremental as it builds on existing hybrid ASR frameworks.
The paper tackled speech recognition by introducing a multistream CNN for acoustic modeling and a self-attentive SRU for language modeling, achieving state-of-the-art word error rates of 1.75% on test-clean and 4.46% on test-other on the LibriSpeech corpus.
In this paper we present state-of-the-art (SOTA) performance on the LibriSpeech corpus with two novel neural network architectures, a multistream CNN for acoustic modeling and a self-attentive simple recurrent unit (SRU) for language modeling. In the hybrid ASR framework, the multistream CNN acoustic model processes an input of speech frames in multiple parallel pipelines where each stream has a unique dilation rate for diversity. Trained with the SpecAugment data augmentation method, it achieves relative word error rate (WER) improvements of 4% on test-clean and 14% on test-other. We further improve the performance via N-best rescoring using a 24-layer self-attentive SRU language model, achieving WERs of 1.75% on test-clean and 4.46% on test-other.