arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2106.08960cs.CLcs.SDeess.AS
收藏

用于语音识别的声学编码器协同训练

Collaborative Training of Acoustic Encoders for Speech Recognition

Facebook

Varun Nagaraja, Yangyang Shi, Ganesh Venkatesh, Ozlem Kalinli, Michael L. Seltzer, Vikas Chandra

摘要:

设备上语音识别需要不同大小的训练模型,以部署在具有各种计算预算的设备上。在构建这些不同的模型时,我们可以通过联合训练它们来利用它们之间共享的知识。联合训练也是有效的,因为它减少了训练程序的数据处理操作中的冗余。我们提出了一种方法,用于协同训练不同大小的声学编码器的语音识别。我们使用一个序列传感器设置,其中不同的声学编码器共享一个共同的预测器和连接器模块。声学编码器还通过用于帧级chenone预测的辅助任务以及换能器损耗使用共蒸馏来训练。我们使用LibriSpeech语料库进行实验,并证明协同训练的声学编码器可以提供高达11%的相对改善的字出错率的测试分区。

英文摘要:

On-device speech recognition requires training models of different sizes for deploying on devices with various computational budgets. When building such different models, we can benefit from training them jointly to take advantage of the knowledge shared between them. Joint training is also efficient since it reduces the redundancy in the training procedure's data handling operations. We propose a method for collaboratively training acoustic encoders of different sizes for speech recognition. We use a sequence transducer setup where different acoustic encoders share a common predictor and joiner modules. The acoustic encoders are also trained using co-distillation through an auxiliary task for frame level chenone prediction, along with the transducer loss. We perform experiments using the LibriSpeech corpus and demonstrate that the collaboratively trained acoustic encoders can provide up to a 11% relative improvement in the word error rate on both the test partitions.

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。

↑