Speech processing Universal PERformance Benchmark (SUPERB)

SUPERB是一个用于评估预训练模型在语音处理任务上通用性能的基准数据集，由台湾大学创建。该数据集包含10个任务，涵盖语音的内容、说话人、语义和副语言特征等方面。数据集的创建过程遵循了语音社区的常规评估协议，并使用了公开可用的数据集。SUPERB旨在解决语音处理领域缺乏统一基准的问题，推动语音处理技术的研究和应用。

SUPERB is a benchmark dataset for evaluating the general performance of pre-trained models on speech processing tasks, developed by National Taiwan University. This dataset encompasses 10 tasks covering areas including speech content, speaker identity, semantics, and paralinguistic features. The construction of the dataset adheres to the standard evaluation protocols of the speech community, and makes use of publicly available datasets. SUPERB aims to resolve the shortage of unified benchmarks in the field of speech processing, and advance the research and application of speech processing technologies.