mort666/cv_corpus_v22

该数据集是Mozilla Common Voice Corpus 22的非官方版本，从项目网站https://commonvoice.mozilla.org/下载并转换而来。目前正在转换为parquet格式以便于使用。数据集包含多种语言，适用于自动语音识别任务。数据集特征包括client_id、path、音频（采样率为48000）、句子、赞成票、反对票、年龄、性别、口音、地区、片段和变体。数据集大小在100B到1T之间，包含多种语言配置和分割（训练、验证、测试、其他、无效）。

This dataset is an unofficial version of the Mozilla Common Voice Corpus 22, downloaded and converted from the projects website https://commonvoice.mozilla.org/. It is currently being converted to parquet format for convenience. The dataset includes a wide range of languages and is designed for automatic speech recognition tasks. The datasets features include client_id, path, audio (with a sampling rate of 48000), sentence, up_votes, down_votes, age, gender, accent, locale, segment, and variant. The dataset is categorized by size as between 100B and 1T, and it includes multiple configurations for different languages and splits (train, validation, test, other, invalidated).