SciClaim
收藏arXiv2021-09-22 更新2024-06-21 收录
下载链接：
https://github.com/siftech/SciClaim
下载链接
链接失效反馈官方服务：
资源简介：
SciClaim数据集是由SIFT和Northeastern University联合创建，专注于科学声明的细粒度知识图谱构建。该数据集包含12,738个标注，涵盖了社会与行为科学、PubMed和CORD19论文中的901个句子。数据集通过创新的图注释模式，不仅包括粗粒度的实体跨度作为节点和关系作为边，还包含了细粒度的属性，用以修改实体及其关系。SciClaim数据集旨在捕捉实验变量间的因果、比较、预测、统计和比例关联，以及它们的限定、子类型和证据，适用于科学声明的细粒度信息提取和知识图谱构建。
The SciClaim dataset was jointly created by SIFT and Northeastern University, focusing on fine-grained knowledge graph construction for scientific claims. This dataset contains 12,738 annotations, covering 901 sentences from social and behavioral sciences, PubMed and CORD19 papers. Adopting an innovative graph annotation schema, it not only includes coarse-grained entity spans as nodes and relationships as edges, but also incorporates fine-grained attributes to modify both entities and their corresponding relationships. The SciClaim dataset aims to capture causal, comparative, predictive, statistical and proportional associations between experimental variables, as well as their qualifications, subtypes and supporting evidence, making it suitable for fine-grained information extraction and knowledge graph construction for scientific claims.
提供机构：
SIFT, Minneapolis, MN, USA ♥Northeastern University, Boston, MA, USA
创建时间：
2021-09-22
