PD&CFT
收藏资源简介:
我们发布的首个中文阅读理解数据集,包括人民日报和儿童童话(PD&CFT)。我们希望这将加速未来机器理解研究的过程。
We have released the first Chinese reading comprehension dataset, which includes the People's Daily and Children's Fairy Tales (PD&CFT). We hope this will accelerate the progress of future research in machine comprehension.
数据集概述
数据集名称
- PD&CFT: A Chinese Reading Comprehension Dataset
数据集内容
- 包含两个子数据集:
- People Daily (PD)
- 包含训练、验证和测试文件:pd.train, pd.valid, pd.test
- Childrens Fairy Tale (CFT)
- 包含两个测试集:cft.test.auto(自动生成)和cft.test.human(人工筛选)
- People Daily (PD)
数据集统计
| 数据集 | 查询数量 | 文档最大词数 | 查询最大词数 | 文档平均词数 | 查询平均词数 | 词汇量 |
|---|---|---|---|---|---|---|
| PD-train | 870,710 | 618 | 502 | 379 | 38 | 248,160 |
| PD-valid | 3,000 | 536 | 153 | 425 | 38 | - |
| PD-test | 3,000 | 634 | 265 | 410 | 41 | - |
| CFT-auto | 1,646 | 318 | 83 | 122 | 20 | - |
| CFT-human | 1,953 | 414 | 92 | 153 | 20 | - |
数据格式
-
每条数据包含多个句子,格式为:
sentence_id ||| sentence
-
最后一个句子包含查询和答案,格式为:
sentence_id ||| Query ||| Answer
数据集下载
- 可通过GitHub下载:People Daily & Childrens Fairy Tale (PD&CFT)
许可证
- 数据集使用CC-BY-SA-4.0许可证。
引用信息
-
引用格式:
@InProceedings{cui-etal-2016-consensus, title = {Consensus Attention-based Neural Networks for Chinese Reading Comprehension}, author = {Cui, Yiming and Liu, Ting and Chen, Zhipeng and Wang, Shijin and Hu, Guoping}, booktitle = {Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers}, year = {2016}, address = {Osaka, Japan}, pages = {1777--1786}, }




