arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 1705.04400cs.CL
收藏

减少生产语音模型中的偏差

Reducing Bias in Production Speech Models

Baidu Silicon Valley AI Lab(百度硅谷人工智能实验室)
Eric Battenberg, Rewon Child, Adam Coates, Christopher Fougner, Yashesh Gaur, Jiaji Huang, Heewoo Jun, Ajay Kannan, Markus Kliegl, Atul Kumar, Hairong Liu, Vina… 展开作者

Eric Battenberg, Rewon Child, Adam Coates, Christopher Fougner, Yashesh Gaur, Jiaji Huang, Heewoo Jun, Ajay Kannan, Markus Kliegl, Atul Kumar, Hairong Liu, Vinay Rao, Sanjeev Satheesh, David Seetapun, Anuroop Sriram, Zhenyao Zhu

摘要:

用端到端的深度学习系统取代手工设计的管道,在语音和对象识别等应用中取得了很好的效果。然而,生产系统的因果关系和延迟约束使端到端语音模型回到了欠拟合状态,并暴露了模型中的偏差,我们发现这些偏差无法通过“扩大规模”来克服,即,在更多数据上训练更大的模型。在这项工作中,我们系统地识别和解决偏差的来源,减少出错率高达20%,同时保持实用的部署。我们通过利用改进的神经架构进行流推理,解决优化问题,并采用增加音频和标签建模多功能性的策略来实现这一目标。

英文摘要:

Replacing hand-engineered pipelines with end-to-end deep learning systems has enabled strong results in applications like speech and object recognition. However, the causality and latency constraints of production systems put end-to-end speech models back into the underfitting regime and expose biases in the model that we show cannot be overcome by "scaling up", i.e., training bigger models on more data. In this work we systematically identify and address sources of bias, reducing error rates by up to 20% while remaining practical for deployment. We achieve this by utilizing improved neural architectures for streaming inference, solving optimization issues, and employing strategies that increase audio and label modelling versatility.

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。

↑