SciER
收藏资源简介:
SciER是一个用于科学文档中实体和关系提取的数据集,由天普大学创建。该数据集包含106篇手动标注的全文科学出版物,涵盖超过24,000个实体和12,000个关系。数据集内容来自Papers with Code网站,涉及人工智能领域的自然语言处理、机器学习、计算机视觉等主题。创建过程中,数据集采用了细粒度的关系标签集,以捕捉实体在全文中的复杂交互。SciER数据集主要用于评估和开发科学信息提取模型,旨在解决科学文献中实体和关系的精确提取问题。
SciER is a dataset dedicated to entity and relation extraction from scientific documents, developed by Temple University. It comprises 106 manually annotated full scientific publications, encompassing more than 24,000 entities and 12,000 relations. The content of this dataset is sourced from the Papers with Code website, covering topics within artificial intelligence such as natural language processing, machine learning, and computer vision. During its creation, a fine-grained relation label set was employed to capture the intricate interactions between entities throughout the full texts. The SciER dataset is primarily utilized for evaluating and developing scientific information extraction models, with the goal of resolving the challenge of accurate extraction of entities and relations from scientific literature.
SciER 数据集概述
数据集内容
- 实体标注:包含科学文档中的实体标注。
- 关系标注:包含科学文档中的关系标注。
实体类型
- Dataset
- Method
- Task
数据格式
LLM 文件夹
- 文件格式:每行是一个句子,包含以下结构: json { "doc_id": "文档ID", "sentence": "句子文本", "ner": [["实体字符串", "实体类型"]], "rel": [["主体实体", "关系类型", "客体实体"]], "rel_plus": [["主体实体:实体类型", "关系类型", "客体实体:实体类型"]] }
PLM 文件夹
- 文件格式:用于训练监督模型,包含以下结构: json { "doc_key": "文档ID", "sentences": [["句子中的词"]], "ner": [[边界位置, 实体类型]], "relations": [[主体实体边界, 客体实体边界, 关系类型]] }




