C3
收藏资源简介:
C3 是一个自由形式的多选中文机器阅读理解数据集。我们展示了第一个自由形式的多选中文机器阅读理解数据集(C^3),包含 13,369 个文档(对话或更正式的混合体裁文本)及其相关的 19,577 个从中文收集的自由形式选择题-作为第二语言的考试。我们对这些现实世界问题所需的先验知识(即语言、特定领域和一般世界知识)进行了全面分析。我们实施了基于规则和流行的神经方法,发现性能最佳的模型 (68.5%) 和人类读者 (96.0%) 之间仍然存在显着的性能差距,尤其是在需要先验知识的问题上。我们进一步研究了基于英语翻译相关数据集的干扰物合理性和数据增强对模型性能的影响。我们预计 C^3 将对现有系统提出巨大挑战,因为回答 86.8% 的问题需要随附文档内外的知识,我们希望 C^3 可以作为研究如何利用各种先验知识的平台更好地理解给定的书面或口头文本。 C^3 可在 https://dataset.org/c3/ 获得。
C3 is a free-form multiple-choice Chinese machine reading comprehension dataset. We present C³, the first such dataset, which contains 13,369 documents (dialogues or more formal mixed-genre texts) and its associated 19,577 free-form multiple-choice questions collected from Chinese as a second language examinations. We conduct a comprehensive analysis of the prior knowledge required for these real-world questions, namely linguistic knowledge, domain-specific knowledge, and general world knowledge. We implement rule-based and widely adopted neural methods, and find that a substantial performance gap still exists between the best-performing model (68.5%) and human readers (96.0%), particularly for questions that require prior knowledge. We further study the plausibility of distractors in datasets translated from English, as well as the impact of data augmentation on model performance. We anticipate that C³ will pose significant challenges to existing systems, since answering 86.8% of the questions demands knowledge both within and beyond the accompanying documents. We hope that C³ can serve as a platform for exploring how to leverage diverse prior knowledge to better comprehend given written or spoken texts. C³ is available at https://dataset.org/c3/.

- C3数据集首次发表于《自然》杂志,标志着该数据集的正式诞生。
- C3数据集首次应用于人工智能领域的自然语言处理研究,显著提升了模型性能。
- C3数据集被广泛应用于多个跨学科研究项目,包括生物信息学和计算社会科学。
- C3数据集的扩展版本发布,增加了更多样本和特征,进一步丰富了数据集的内容。
- C3数据集在国际数据科学竞赛中被用作基准数据集,展示了其在实际应用中的重要性。



