Listen Attentively, and Spell Once: Whole Sentence Generation via a Non-Autoregressive Architecture for Low-Latency Speech Recognition
This work addresses the problem of slow inference times in speech recognition systems, which limits practical applications, by introducing a non-autoregressive architecture that significantly reduces latency while maintaining competitive accuracy.
The authors tackled the high inference latency of autoregressive speech recognition models by proposing LASO, a non-autoregressive end-to-end system that predicts tokens in parallel without beam-search, achieving a character error rate of 6.4% and reducing average inference latency to 21 ms, which is 1/50 of the state-of-the-art autoregressive transformer model.
Although attention based end-to-end models have achieved promising performance in speech recognition, the multi-pass forward computation in beam-search increases inference time cost, which limits their practical applications. To address this issue, we propose a non-autoregressive end-to-end speech recognition system called LASO (listen attentively, and spell once). Because of the non-autoregressive property, LASO predicts a textual token in the sequence without the dependence on other tokens. Without beam-search, the one-pass propagation much reduces inference time cost of LASO. And because the model is based on the attention based feedforward structure, the computation can be implemented in parallel efficiently. We conduct experiments on publicly available Chinese dataset AISHELL-1. LASO achieves a character error rate of 6.4%, which outperforms the state-of-the-art autoregressive transformer model (6.7%). The average inference latency is 21 ms, which is 1/50 of the autoregressive transformer model.