CO-Fun

CO-Fun是一个专注于德国基金招股说明书中公司外包情况的数据集，由德国人工智能研究中心创建。该数据集包含948个句子，总计5,969个命名实体标注和4,102个关系标注，涉及外包、公司、地点和软件四种实体类型。数据集的创建过程涉及从1,054个公开的基金招股说明书中提取信息，并通过专家标注完成。CO-Fun数据集主要用于支持网络映射过程，通过自然语言处理模型识别实体和提取关系，以揭示金融实体和服务提供商之间的关联，从而帮助发现潜在的网络风险。

CO-Fun is a dataset dedicated to corporate outsourcing scenarios in German fund prospectuses, developed by the German Research Center for Artificial Intelligence. This dataset comprises 948 sentences, with a total of 5,969 named entity annotations and 4,102 relational annotations, covering four entity types: outsourcing, corporation, location, and software. The dataset was constructed by extracting information from 1,054 publicly available fund prospectuses and finalized via expert manual annotation. The CO-Fun dataset is primarily intended to support network mapping workflows: it enables natural language processing models to identify entities and extract relational information, thereby uncovering the associations between financial entities and service providers, and assisting in the detection of potential cyber risks.