Non-Autoregressive Transformer ASR with CTC-Enhanced Decoder Input
This work addresses the trade-off between speed and accuracy in speech recognition for real-time applications, representing an incremental improvement over existing non-autoregressive methods.
The paper tackles the accuracy degradation in non-autoregressive transformer ASR models by proposing a CTC-enhanced decoder input, achieving 50x faster decoding than autoregressive baselines with only 0.0 to 0.3 absolute CER degradation on Aishell datasets.
Non-autoregressive (NAR) transformer models have achieved significantly inference speedup but at the cost of inferior accuracy compared to autoregressive (AR) models in automatic speech recognition (ASR). Most of the NAR transformers take a fixed-length sequence filled with MASK tokens or a redundant sequence copied from encoder states as decoder input, they cannot provide efficient target-side information thus leading to accuracy degradation. To address this problem, we propose a CTC-enhanced NAR transformer, which generates target sequence by refining predictions of the CTC module. Experimental results show that our method outperforms all previous NAR counterparts and achieves 50x faster decoding speed than a strong AR baseline with only 0.0 ~ 0.3 absolute CER degradation on Aishell-1 and Aishell-2 datasets.