Biographical

Biographical 数据集是首个针对关系抽取的半监督学习数据集，由沃尔弗汉普顿大学的研究团队开发。该数据集通过自动匹配维基百科文章中的句子与Pantheon和Wikidata等来源的结构化数据来构建，包含十种在数字人文领域重要的关系类别。数据集的创建过程中利用了维基百科文章的结构和强大的命名实体识别技术，以较高的精度匹配信息。Biographical 数据集主要用于训练数字人文和历史领域的神经网络模型进行关系抽取，但也适用于其他研究目的。

The Biographical Dataset is the first semi-supervised learning dataset dedicated to relation extraction, developed by a research team at the University of Wolverhampton. Constructed by automatically matching sentences from Wikipedia articles with structured data from sources including Pantheon and Wikidata, this dataset encompasses ten critical relational categories in the digital humanities domain. During its development, the dataset leverages the structural properties of Wikipedia articles and robust named entity recognition (NER) techniques to match information with high precision. It is primarily intended for training neural network models for relation extraction in the digital humanities and history fields, though it is also applicable for other research purposes.