Hierarchy-Transformers/DOID

DOID数据集是一个用于评估层次嵌入模型的混合跳预测数据集集合，基于DOID的包含层次结构（TBox）创建。数据集包含多种配置，如随机负样本和三重样本、硬负样本和三重样本等，每种配置都有训练、验证和测试集。数据集的使用包括评估和训练，评估时建议使用对样本结构，训练时则可以根据模型设计选择不同的样本结构。

This dataset is a collection of Mixed-hop Prediction datasets created from DOIDs subsumption hierarchy (TBox) for evaluating hierarchy embedding models. It includes various configurations such as random negatives with triplets, hard negatives with triplets, etc., each with training, validation, and test sets. The dataset is used for both evaluation and training, with pairs recommended for evaluation and the choice of sample structures for training depending on the model design.