XEUS

该数据集是一个多语言编码器基础的语音识别模型，该模型在约100万小时的未标记音频上进行预训练，覆盖了4057种语言。此外，该模型还融入了去混响训练，以增强对各种声学环境的鲁棒性。其任务是语音识别。This dataset is a multilingual encoder-based speech recognition model. The model is pre-trained on approximately 1 million hours of unlabeled audio, covering 4057 languages. Additionally, the model incorporates dereverberation training to enhance its robustness against various acoustic environments. Its target task is speech recognition.