Multi-head Monotonic Chunkwise Attention For Online Speech Recognition

Abstract

The attention mechanism of the Listen, Attend and Spell (LAS) model requires the whole input sequence to calculate the attention context and thus is not suitable for online speech recognition. To deal with this problem, we propose multi-head monotonic chunk-wise attention (MTH-MoChA), an improved version of MoChA. MTH-MoChA splits the input sequence into small chunks and computes multi-head attentions over the chunks. We also explore useful training strategies such as LSTM pooling, minimum world error rate training and SpecAugment to further improve the performance of MTH-MoChA. Experiments on AISHELL-1 data show that the proposed model, along with the training strategies, improve the character error rate (CER) of MoChA from 8.96% to 7.68% on test set. On another 18000 hours in-car speech data set, MTH-MoChA obtains 7.28% CER, which is significantly better than a state-of-the-art hybrid system.

References in corpus (6)

• SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition
• Sequence Transduction with Recurrent Neural Networks
• A Comparative Study on Transformer vs RNN in Speech Applications
• Online and Linear-Time Attention by Enforcing Monotonic Alignments
• Lingvo: a Modular and Scalable Framework for Sequence-to-Sequence Modeling
• On the Choice of Modeling Unit for Sequence-to-Sequence Speech Recognition

Cited by in corpus (4)

• Thank you for Attention: A survey on Attention-based Artificial Neural Networks for Automatic Speech Recognition
• Transformer-based Online Speech Recognition with Decoder-end Adaptive Computation Steps
• Multi-QuartzNet: Multi-Resolution Convolution for Speech Recognition with Multi-Layer Feature Fusion
• Decoupling recognition and transcription in Mandarin ASR