Panga-Azazia/all-in-one

该数据集包含三个子配置：1. afvoices：包含文本、音频和参与者ID特征，用于语音相关任务，训练集有145,909个样本，测试集有7,711个样本。2. bam-asr-all：包含音频（采样率16kHz）、时长、bam和法语文本特征，用于自动语音识别（ASR）任务，训练集有37,105个样本，测试集有1,461个样本。3. kunkado：包含音频（采样率16kHz）、时长、半标签和修正标签特征，用于语音标注任务，训练集有28,238个样本，测试集有4,900个样本。所有配置都分为训练集和测试集，支持多语言或特定领域应用。This dataset includes three sub-configurations: 1. afvoices: Features text, audio, and participant ID, designed for speech-related tasks, with 145,909 training examples and 7,711 test examples. 2. bam-asr-all: Features audio (sampling rate 16kHz), duration, bam, and French text, intended for automatic speech recognition (ASR) tasks, with 37,105 training examples and 1,461 test examples. 3. kunkado: Features audio (sampling rate 16kHz), duration, semi-label, and corrected-label, used for speech annotation tasks, with 28,238 training examples and 4,900 test examples. All configurations are split into training and test sets, supporting multilingual or domain-specific applications.