ggfox00000/stt-covost25-test-fr

Common Voice FR — read + spontaneous + CoVoST 2 test mirror数据集是一个结合了三种不同语音资源的镜像数据集，用于法语多范式WER/STT基准测试。数据集包含三种配置：read（默认）、spontaneous和covost2。read配置来源于Mozilla Common Voice 25.0 FR，包含16,149个朗读语音样本；spontaneous配置来源于Mozilla Spontaneous Speech corpus FR，包含152个自发回答语音样本；covost2配置来源于facebook/covost2，包含14,760个朗读语音样本，并带有英文翻译。数据集的主要目的是测量时间鲁棒性、范式鲁棒性和翻译任务。

The Common Voice FR — read + spontaneous + CoVoST 2 test mirror dataset is a combined mirror of three different speech resources for French multi-paradigm WER/STT benchmarking. The dataset includes three configurations: read (default), spontaneous, and covost2. The read configuration is sourced from Mozilla Common Voice 25.0 FR, containing 16,149 read speech samples. The spontaneous configuration is sourced from Mozilla Spontaneous Speech corpus FR, containing 152 spontaneous response speech samples. The covost2 configuration is sourced from facebook/covost2, containing 14,760 read speech samples with English translations. The primary purpose of the dataset is to measure temporal robustness, paradigm robustness, and translation tasks.