SEP-28k

该数据集对结巴言语进行了详细标注，涵盖了四种类型的结巴：阻塞、插话、单词重复和声音重复。此外，数据集仅保留了包含老年人和病理语音的片段，并依据数据集的注释进行了筛选。该数据集大约包含2.5小时的声音资料，其任务是进行语音匿名化处理。

This dataset comprehensively annotates stuttered speech, covering four types of stuttering disfluencies: speech blocking, interjections, word repetitions, and sound repetitions. Furthermore, the dataset only retains segments featuring speech from elderly speakers and pathological speech, and was filtered in accordance with the dataset's own annotations. This dataset contains approximately 2.5 hours of audio data, and its target task is speech anonymization.