Towards Online End-to-end Transformer Automatic Speech Recognition

TL;DR

This paper introduces an online end-to-end Transformer-based speech recognition system that incorporates a context-aware inheritance mechanism and monotonic chunkwise attention, enabling real-time processing with improved accuracy.

Contribution

It presents a novel online Transformer ASR system with a context-aware inheritance mechanism and MoChA-based decoding, advancing real-time speech recognition capabilities.

Findings

Outperforms conventional chunkwise approaches on WSJ and AISHELL-1 datasets.

Effectively encodes global linguistic and speaker attributes in online processing.

Demonstrates the feasibility of online Transformer ASR with improved accuracy.

Abstract

The Transformer self-attention network has recently shown promising performance as an alternative to recurrent neural networks in end-to-end (E2E) automatic speech recognition (ASR) systems. However, Transformer has a drawback in that the entire input sequence is required to compute self-attention. We have proposed a block processing method for the Transformer encoder by introducing a context-aware inheritance mechanism. An additional context embedding vector handed over from the previously processed block helps to encode not only local acoustic information but also global linguistic, channel, and speaker attributes. In this paper, we extend it towards an entire online E2E ASR system by introducing an online decoding process inspired by monotonic chunkwise attention (MoChA) into the Transformer decoder. Our novel MoChA training and inference algorithms exploit the unique properties of Transformer, whose attentions are not always monotonic or peaky, and have multiple heads and residual connections of the decoder layers. Evaluations of the Wall Street Journal (WSJ) and AISHELL-1 show that our proposed online Transformer decoder outperforms conventional chunkwise approaches.