arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.09168eess.AScs.SD
收藏

长序列数据的高斯核自注意及其在基于CTC的语音识别中的应用

Gaussian Kernelized Self-Attention for Long Sequence Data and Its Application to CTC-based Speech Recognition

Sony Corporation(索尼公司);Johns Hopkins University(约翰斯·霍普金斯大学)

Yosuke Kashiwagi, Emiru Tsunoo, Shinji Watanabe

摘要:

基于自注意(SA)的模型由于其灵活的上下文建模能力,最近在混合和端到端自动语音识别(ASR)系统中取得了显著的性能改进。然而,还已知的是,当将SA应用于长序列数据时,准确度降低。这主要是由于推理和训练数据之间的长度不匹配,因为训练数据通常被划分为短段以进行有效训练。为了减轻这种不匹配,我们提出了一种新的架构,这是高斯内核的一个变体,它本身就是一个平移不变内核。首先,我们从数学上证明了具有查询和键的共享权重参数的自关注等价于归一化核函数。通过将该核函数替换为所提出的高斯核,该架构变得完全平移不变,并且使用帧索引技术嵌入相对位置信息。提出的高斯核SA应用于连接主义时态分类(CTC)的ASR。使用自发日语语料库(CSJ)和TEDLIUM 3基准的实验评估表明,所提出的SA在准确性方面实现了显着提高(例如,从24.0%WER到CSJ中的6.0%)。

英文摘要:

Self-attention (SA) based models have recently achieved significant performance improvements in hybrid and end-to-end automatic speech recognition (ASR) systems owing to their flexible context modeling capability. However, it is also known that the accuracy degrades when applying SA to long sequence data. This is mainly due to the length mismatch between the inference and training data because the training data are usually divided into short segments for efficient training. To mitigate this mismatch, we propose a new architecture, which is a variant of the Gaussian kernel, which itself is a shift-invariant kernel. First, we mathematically demonstrate that self-attention with shared weight parameters for queries and keys is equivalent to a normalized kernel function. By replacing this kernel function with the proposed Gaussian kernel, the architecture becomes completely shift-invariant with the relative position information embedded using a frame indexing technique. The proposed Gaussian kernelized SA was applied to connectionist temporal classification (CTC) based ASR. An experimental evaluation with the Corpus of Spontaneous Japanese (CSJ) and TEDLIUM 3 benchmarks shows that the proposed SA achieves a significant improvement in accuracy (e.g., from 24.0% WER to 6.0% in CSJ) in long sequence data without any windowing techniques.

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。

↑