bigbio/flambe

FlaMBe数据集旨在从生物医学文本中提取程序性知识，特别关注学术论文中描述的单细胞研究方法。该数据集包含55篇全文文章和1,195篇摘要的注释，覆盖了近710,000个标记，并以其对组织/细胞类型、软件工具和计算方法的全面命名实体识别（NER）和消歧（NED）而著称。此外，数据集还链接了相关知识库中的实体标识符，并注释了近400个工作流关系。

FlaMBe is a dataset aimed at procedural knowledge extraction from biomedical texts, particularly focusing on single cell research methodologies described in academic papers. It includes annotations from 55 full-text articles and 1,195 abstracts, covering nearly 710,000 tokens, and is distinguished by its comprehensive named entity recognition (NER) and disambiguation (NED) for tissue/cell types, software tools, and computational methods. This dataset, to our knowledge, is the largest of its kind for tissue/cell types, links entities to identifiers in relevant knowledge bases and annotates nearly 400 workflow relations between tool-context pairs.

原始信息汇总

数据集概述

基本信息

• 名称: FlaMBe
• 语言: 英语
• 许可证: CC-BY-4.0
• 多语言性: 单语
• 主页: https://github.com/ylaboratory/flambe

数据集描述

• 目标: 从生物医学文本中提取程序性知识，特别关注单细胞研究方法论
• 内容: 包含来自55篇全文文章和1,195篇摘要的标注，覆盖约710,000个令牌
• 特点: 全面的命名实体识别（NER）和消歧（NED），涵盖组织/细胞类型、软件工具和计算方法
• 规模: 据我们所知，是同类数据集中最大的，涉及组织/细胞类型
• 关联: 将实体链接到相关知识库，并标注近400个工具-上下文对的流程关系

任务类型

• 命名实体识别（NER）
• 命名实体消歧（NED）

引用信息

@inproceedings{ author = {Dannenfelser, Ruth and Zhong, Jeffrey and Zhang, Ran and Yao, Vicky}, title = {Into the Single Cell Multiverse: an End-to-End Dataset for Procedural Knowledge Extraction in Biomedical Texts}, publisher = {Advances in Neural Information Processing Systems}, volume = {36}, year = {2024}, url = {https://proceedings.neurips.cc/paper_files/paper/2023/file/23e3d86c9a19d0caf2ec997e73dfcfbd-Paper-Datasets_and_Benchmarks.pdf}, }