CCD V2

该数据集名为CCD V2，包含了来自印度、美国、印度尼西亚、越南、巴西、墨西哥和菲律宾等不同地区的5,567位独特发言人的语音数据。其中包含了26,467段视频录音，总计354小时的非脚本自然回应和319小时对陀思妥耶夫斯基作品《白痴》的朗读。此外，该数据集还包含了七个自我标记的属性，如年龄、性别、语言、残疾和肤色等级，这允许分析在各类人口统计属性中，自动语音识别系统性能的差异。规模上，该数据集涉及5,567位发言人，26,467段录音，总计354小时的录音内容。该数据集的任务是评估自动语音识别系统的公平性。

This dataset, named CCD V2, contains speech data from 5,567 unique speakers across diverse regions including India, the United States, Indonesia, Vietnam, Brazil, Mexico, and the Philippines. It comprises 26,467 video recordings, totaling 354 hours of unscripted spontaneous responses and 319 hours of recitations of Fyodor Dostoevsky's novel *The Idiot*. Additionally, the dataset includes seven self-annotated attributes such as age, gender, language, disability, and skin color grade, which enable analysis of performance disparities of automatic speech recognition systems across various demographic characteristics. In terms of scale, this dataset covers 5,567 speakers, 26,467 recordings, with a total of 354 hours of audio content. The core task of this dataset is to evaluate the fairness of automatic speech recognition systems.