malaysia-ai/fleurs-r-neucodec-all-languages
收藏资源简介:
该数据集包含102种语言的FLEURS-R元数据、源音频和预计算的NeuCodec令牌。具体布局包括:data/目录以Parquet格式存储训练和开发元数据;audio/目录按语言和分割组织源FLEURS-R音频存档;neucodec/目录按语言、分割和处理等级组织NeuCodec令牌存档。每个Parquet行包含语言、分割、源音频路径、NeuCodec令牌路径、转录文本、归一化文本、说话者性别和样本计数。源数据集为google/fleurs-r,遵循CC BY 4.0许可证,适用于多语言文本到语音和自动语音识别任务。
This dataset contains FLEURS-R metadata, source audio, and precomputed NeuCodec tokens for 102 locales. The layout includes: data/ with train and development metadata in Parquet format; audio/ with source FLEURS-R audio archives organized by locale and split; neucodec/ with NeuCodec token archives organized by locale, split, and processing rank. Each Parquet row includes the locale, split, source audio path, NeuCodec token path, transcript, normalized text, speaker gender, and sample count. The source dataset is google/fleurs-r, distributed under the CC BY 4.0 license, and it is designed for multilingual text-to-speech and automatic speech recognition tasks.




