Listen and Fill in the Missing Letters: Non-Autoregressive Transformer for Speech Recognition

TL;DR

This paper introduces non-autoregressive transformer models for speech recognition that enable faster inference while maintaining high accuracy, outperforming traditional systems and matching state-of-the-art autoregressive models.

Contribution

The paper proposes two non-autoregressive transformer frameworks for ASR, demonstrating effective training and inference strategies with significant speed improvements.

Findings

Outperforms Kaldi ASR on Aishell benchmark

Matches state-of-the-art autoregressive transformer performance

Achieves 7x inference speedup

Abstract

Recently very deep transformers have outperformed conventional bi-directional long short-term memory networks by a large margin in speech recognition. However, to put it into production usage, inference computation cost is still a serious concern in real scenarios. In this paper, we study two different non-autoregressive transformer structure for automatic speech recognition (ASR): A-CMLM and A-FMLM. During training, for both frameworks, input tokens fed to the decoder are randomly replaced by special mask tokens. The network is required to predict the tokens corresponding to those mask tokens by taking both unmasked context and input speech into consideration. During inference, we start from all mask tokens and the network iteratively predicts missing tokens based on partial results. We show that this framework can support different decoding strategies, including traditional left-to-right. A new decoding strategy is proposed as an example, which starts from the easiest predictions to the most difficult ones. Results on Mandarin (Aishell) and Japanese (CSJ) ASR benchmarks show the possibility to train such a non-autoregressive network for ASR. Especially in Aishell, the proposed method outperformed the Kaldi ASR system and it matches the performance of the state-of-the-art autoregressive transformer with 7x speedup. Pretrained models and code will be made available after publication.