FewCLUE
收藏资源简介:
FewCLUE是首个系统全面的中文小样本学习评测基准,由中文自然语言处理评测团队创建。该数据集包含九个任务,涵盖单句分类、句子对分类及机器阅读理解等多种自然语言理解任务。数据集通过精心设计,旨在以少量标注数据评估模型性能,支持零样本和半监督学习研究。FewCLUE的应用领域广泛,旨在推动中文小样本学习技术的发展,解决实际场景中标注数据稀缺的问题。
FewCLUE is the first systematic and comprehensive Chinese few-shot learning evaluation benchmark, created by the Chinese natural language processing evaluation team. This dataset includes nine tasks, covering a variety of natural language understanding tasks such as single-sentence classification, sentence-pair classification, and machine reading comprehension. It is meticulously designed to evaluate model performance with a small amount of labeled data, supporting zero-shot and semi-supervised learning research. FewCLUE has broad application scenarios, aiming to promote the development of Chinese few-shot learning technologies and address the problem of scarce labeled data in real-world scenarios.




