LEMAS-Project/LEMAS-Dataset

该数据集是LEMAS（大规模可扩展多语言音频套件）的一部分，包含大规模训练集（超过15万小时）和精选评估集（每种语言100个话语），覆盖10种语言，所有数据都具有单词级对齐。数据集包含音频文件、原始转录、对齐文本和单词级时间戳等信息。统计信息显示，数据集包含意大利语、法语、越南语、葡萄牙语、德语、印尼语、西班牙语、俄语、英语和中文等多种语言，每种语言都有详细的总时长、话语数量、平均时长等统计信息。

This dataset is part of LEMAS (Large-scale Extensible Multilingual Audio Suite). It contains a large-scale training set (150k+ hours) and a curated evaluation set (100 utterances per language) covering 10 languages, all with word-level alignment. The dataset includes audio files, original transcriptions, aligned text, and word-level timestamps. Statistics show that the dataset includes multiple languages such as Italian, French, Vietnamese, Portuguese, German, Indonesian, Spanish, Russian, English, and Chinese, with detailed statistics on total duration, number of utterances, average duration, etc. for each language.