AdoCleanCode/libri-10h-mfa

该数据集是一个语音数据集，包含音频文件及其对应的转录文本、单词和音素标注。音频的采样率为16000Hz。每个样本包含唯一的id和说话者id，以及音频文件路径、转录文本、单词列表（包含单词及其起止时间）和音素列表（包含音素及其起止时间）。数据集仅包含训练集，共有8176个样本。

This dataset is a speech dataset containing audio files along with their corresponding transcripts, word-level, and phoneme-level annotations. The audio has a sampling rate of 16000Hz. Each sample includes a unique id, speaker id, audio file path, transcript, list of words (with their start and end times), and list of phonemes (with their start and end times). The dataset only contains a training split with 8176 samples.