copenlu/citeworth
收藏资源简介:
CiteWorth数据集是一个用于检测科学文档中引用价值(cite-worthiness)的英文数据集。该数据集基于S2ORC数据集构建,包含了大量从科学文档中提取的纯文本数据,并经过严格的清理和标注。数据集的主要任务是文本分类,具体是判断一个句子是否引用了外部来源。数据集的创建过程在相关论文中有详细描述,且数据集的质量高、具有挑战性,适合用于研究领域适应等问题。
The CiteWorth Dataset is an English-language dataset designed for detecting cite-worthiness in scientific documents. Built upon the S2ORC dataset, it comprises a large quantity of plain text data extracted from scientific documents, which has undergone rigorous cleaning and annotation. The core task of this dataset is text classification, specifically to determine whether a sentence cites an external source. The creation process of the dataset is thoroughly described in the relevant research paper, and the dataset features high quality and challenging characteristics, making it suitable for research on issues such as domain adaptation.
数据集概述:CiteWorth
数据集描述
- 名称:CiteWorth
- 语言:英语
- 许可证:CC-BY-NC-4.0
- 多语言性:单语种
- 大小:1M<n<10M
- 来源数据集:S2ORC(扩展)
- 标签:
- 引用检测
- 引用
- 科学
- 学术文档
- 生物
- 医学
- 计算机科学
- 引用价值
- 任务类别:文本分类
数据集结构
- paper_id:S2ORC论文ID
- section_idx:原始S2ORC数据中章节数组索引
- file_index:S2ORC数据集中论文所属卷
- file_offset:S2ORC论文PDF文件中论文JSON的起始字节偏移
- mag_field_of_study:论文所属领域
- original_text:段落原文
- section_title:段落所属章节标题
- samples:段落中句子数组,每个句子包含:
text:句子清理后的文本label:句子标签,check-worthy(引用价值)或non-check-worthy(非引用价值)original_text:句子原文ref_ids:句子中引用的论文在S2ORC数据集中的ID列表citation_text:句子中的所有引用文本列表
数据集创建
- 来源:S2ORC数据集20200705v1版本
- 许可证:CC By-NC 2.0
- 详细创建过程:参见论文第3节
引用信息
@inproceedings{wright2021citeworth, title={{CiteWorth: Cite-Worthiness Detection for Improved Scientific Document Understanding}}, author={Dustin Wright and Isabelle Augenstein}, booktitle = {Findings of ACL-IJCNLP}, publisher = {Association for Computational Linguistics}, year = 2021 }




