arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2202.00673cs.LGcs.AIcs.CLcs.HCcs.SDeess.AS
收藏

可视化自动语音识别--更好理解的手段?

Visualizing Automatic Speech Recognition -- Means for a Better Understanding?

Fraunhofer AISEC(弗劳恩霍夫应用与集成安全研究所);Technical University Munich(慕尼黑工业大学)

Karla Markert, Romain Parracone, Mykhailo Kulakov, Philip Sperl, Ching-Yu Kao, Konstantin Böttinger

摘要:

自动语音识别(ASR)在模仿人类语音处理方面不断改进。然而,ASR的功能在很大程度上仍然被它们所基于的深度神经网络(DNN)的复杂结构所混淆。在本文中,我们展示了所谓的归因方法,即我们从图像识别中导入并适当地适应处理音频数据的方法,可以帮助澄清ASR的工作。以ASR的端到端模型DeepSpeech为例,我们展示了这些技术如何帮助可视化输入的哪些特征对确定输出最有影响。我们专注于三种可视化技术:逐层相关传播(LRP),显着性地图和Shapley加法解释(SHAP)。我们比较了这些方法,并讨论了潜在的进一步应用,如在检测对抗性的例子。

英文摘要:

Automatic speech recognition (ASR) is improving ever more at mimicking human speech processing. The functioning of ASR, however, remains to a large extent obfuscated by the complex structure of the deep neural networks (DNNs) they are based on. In this paper, we show how so-called attribution methods, that we import from image recognition and suitably adapt to handle audio data, can help to clarify the working of ASR. Taking DeepSpeech, an end-to-end model for ASR, as a case study, we show how these techniques help to visualize which features of the input are the most influential in determining the output. We focus on three visualization techniques: Layer-wise Relevance Propagation (LRP), Saliency Maps, and Shapley Additive Explanations (SHAP). We compare these methods and discuss potential further applications, such as in the detection of adversarial examples.

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。

↑