Speech recognizers trained on close-talking speech do not generalize to distant speech and the word error rate degradation can be as large as 40% absolute. Most studies focus on tackling distant speech recognition as a separate problem, leaving little effort to adapting close-talking speech recognizers to distant speech. In this work, we review several approaches from a domain adaptation perspective. These approaches, including speech enhancement, multi-condition training, data augmentation, and autoencoders, all involve a transformation of the data between domains. We conduct experiments on the AMI data set, where these approaches can be realized under the same controlled setting. These approaches lead to different amounts of improvement under their respective assumptions. The purpose of this paper is to quantify and characterize the performance gap between the two domains, setting up the basis for studying adaptation of speech recognizers from close-talking speech to distant speech. Our results also have implications for improving distant speech recognition.
Data Augmentation Methods for End-to-end Speech Recognition on Distant-Talk Scenarios
Sony Corporation（索尼公司）；Carnegie Mellon University（卡内基梅隆大学）
虽然端到端自动语音识别（E2 E ASR）在具有大量配对数据的任务中已经实现了很好的性能，但是使E2 E ASR对噪声和低资源条件具有鲁棒性仍然具有挑战性。在这项研究中，我们研究了远程通话场景中E2 E ASR的数据增强方法。E2 E ASR模型在一系列CHiME挑战数据集上进行训练，这些挑战数据集适合于研究对噪声和自发语音的鲁棒性。我们建议使用三种增强方法及其组合：1）使用文本到语音（TTS）数据的数据增强，2）循环一致的生成对抗网络（Cycle-GAN）增强训练映射两个不同的音频特征，干净的语音和嘈杂的录音，以匹配测试条件，以及3）由预训练的ASR模块提供的伪标签增强，用于平滑标签分布。使用CHiME-6/CHiME-4数据集的实验结果表明，每种增强方法都在传统SpecAugment的基础上单独提高了准确性;通过组合这些方法可以获得进一步的改进。我们实现了4.3\%的字错误率（WER）减少，这是更显着的SpecAugment，当我们结合所有三个增强CHiME-6任务。
KnowDis: Knowledge Enhanced Data Augmentation for Event Causality Detection via Distant Supervision
National Laboratory of Pattern Recognition, Institute of AutomationChinese Academy of Sciences（中国科学院自动化研究所模式识别国家实验室）；School of Artificial IntelligenceUniversity of Chinese Academy of Sciences（中国科学院大学人工智能学院）
Data augmentation using generative networks to identify dementia
Department of Computer Science（计算机科学系）；Sheffield Institute for Translational Neuroscience (SITraN)（谢菲尔德转化神经科学研究所（SITraN））；Department of Human Communication Sciences（人类传播科学系）；Department of Neuroscience（神经科学系）；Royal Hallamshire Hospital（皇家哈勒姆郡医院）；Academic Neurology Unit, Royal Hallamshire Hospital（皇家哈勒姆郡医院学术神经科）
VaeDiff-DocRE: End-to-end Data Augmentation Framework for Document-level Relation Extraction
School of Electrical Engineering and Computer Science, The University of Queensland（昆士兰大学电气工程与计算机科学学院）；Department of Data Science and Artificial Intelligence, The Hong Kong Polytechnic University（香港理工大学数据科学与人工智能系）
