ADI-17

该数据集名为ADI-17，包含了来自YouTube的17种阿拉伯语方言的音频数据。总计拥有3033小时的训练数据以及58小时的开发和测试数据。然而，该数据集在各个方言的语音数量和数据小时数上存在不平衡。音频剪辑的时长从1秒到26分钟不等，并且数据集被分为短时、中时和长时三个时长段。ADI-17的数据规模为3033小时的训练数据和58小时的开发和测试数据，其任务是进行阿拉伯语方言的识别。

The dataset is named ADI-17, which encompasses audio data of 17 distinct Arabic dialects sourced from YouTube. It has a total of 3033 hours of training data, alongside 58 hours of development and test data. However, this dataset exhibits imbalance in both the quantity of speech samples and total audio duration across different dialects. The duration of individual audio clips ranges from 1 second to 26 minutes, and the dataset is partitioned into three duration categories: short-duration, medium-duration, and long-duration. The data scale of ADI-17 is 3033 hours of training data and 58 hours of development and test data, with its core task being Arabic dialect identification.