arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2005.10463cs.SDcs.CLeess.AS
收藏

基于transformer的端到端语音识别的简化自注意

Simplified Self-Attention for Transformer-based End-to-End Speech Recognition

Northwestern Polytechnical University(西北工业大学);School of Computer Science(计算机学院);Alibaba DAMO Academy(阿里巴巴达摩院)

Haoneng Luo, Shiliang Zhang, Ming Lei, Lei Xie

摘要:

Transformer模型由于其在建模长期依赖关系方面的优势,已被引入到端到端语音识别中,在各种任务上具有最先进的性能。然而,这种改进通常是通过使用非常大的神经网络来获得的。Transformer模型主要包括位置前馈层和自注意(SAN)层两个子模块。在本文中,为了降低模型的复杂性,同时保持良好的性能,我们提出了一个简化的自我注意(SSAN)层,使用FSMN内存块,而不是投影层,以形成查询和关键向量的转换为基础的端到端语音识别。我们评估了基于SSAN和传统的基于SAN的Transformers器上的公共AISHELL-1,内部1000小时和20,000小时的大规模普通话任务。结果表明,我们提出的基于SSAN的Transformer模型可以实现超过20%的相对减少模型参数和6.7%的相对CER减少AISHELL-1任务。令人印象深刻的是,我们的模型在20,000小时的大规模任务中没有表现出识别性能的损失。

英文摘要:

Transformer models have been introduced into end-to-end speech recognition with state-of-the-art performance on various tasks owing to their superiority in modeling long-term dependencies. However, such improvements are usually obtained through the use of very large neural networks. Transformer models mainly include two submodules - position-wise feedforward layers and self-attention (SAN) layers. In this paper, to reduce the model complexity while maintaining good performance, we propose a simplified self-attention (SSAN) layer which employs FSMN memory block instead of projection layers to form query and key vectors for transformer-based end-to-end speech recognition. We evaluate the SSAN-based and the conventional SAN-based transformers on the public AISHELL-1, internal 1000-hour and 20,000-hour large-scale Mandarin tasks. Results show that our proposed SSAN-based transformer model can achieve over 20% relative reduction in model parameters and 6.7% relative CER reduction on the AISHELL-1 task. With impressively 20% parameter reduction, our model shows no loss of recognition performance on the 20,000-hour large-scale task.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑