Mozilla Common Voice Mcv-Accent
收藏arXiv2025-09-30 收录
下载链接:
https://tinyurl.com/accent-dataset
下载链接
链接失效反馈官方服务:
资源简介:
该数据集涵盖了五个已知口音(澳大利亚、加拿大、苏格兰、英国、美国)和九个未知口音(非洲、香港、印度、爱尔兰、马来西亚、新西兰、菲律宾、新加坡、威尔士)。此外,该数据集还包括了100小时和620小时的训练分割音频数据,以及17小时的验证和测试数据。其规模分为Mcv-Accent-100(100小时)和Mcv-Accent-600(620小时)。该数据集的任务是自动语音识别(ASR)。
This dataset covers five known accent categories (Australian, Canadian, Scottish, British, and American) and nine unknown accent categories (African, Hong Kong, Indian, Irish, Malaysian, New Zealand, Philippine, Singaporean, and Welsh). Furthermore, it includes 100-hour and 620-hour training split audio data, as well as 17 hours of validation and test data. It has two scale variants: Mcv-Accent-100 (100 hours) and Mcv-Accent-600 (620 hours). The task of this dataset is automatic speech recognition (ASR).
提供机构:
Mozilla



