ESAformer: Enhanced Self-Attention for Automatic Speech Recognition

Abstract

In this paper, an Enhanced Self-Attention (ESA) module has been put forward for feature extraction. The proposed ESA is integrated with the recursive gated convolution and self-attention mechanism. In particular, the former is used to capture multi-order feature interaction and the latter is for global feature extraction. In addition, the location of interest that is suitable for inserting the ESA is also worth being explored. In this paper, the ESA is embedded into the encoder layer of the Transformer network for automatic speech recognition (ASR) tasks, and this newly proposed model is named ESAformer. The effectiveness of the ESAformer has been validated using three datasets, that are Aishell-1, HKUST and WSJ. Experimental results show that, compared with the Transformer network, 0.8% CER, 1.2% CER and 0.7%/0.4% WER, improvement for these three mentioned datasets, respectively, can be achieved.