Taxo-KG-Bench

The benchmark contains 6 datasets which are generated from 3 automatically constructed taxonomy (MSCG, SEMusic, SEMedical) and 2 open knowledge graph (ReVerb, OPIEC). For each dagtaset, entity-concept pairs (`cg_pairs.*.txt`) and entity-relation-entity triples (`oie_triples.*.txt`) are splitted into train, dev, test sets.

本基准包含6个数据集，这些数据集均源自3个自动构建的分类体系（taxonomy）（MSCG、SEMusic、SEMedical）以及2个开放知识图谱（open knowledge graph）（ReVerb、OPIEC）。针对每个数据集，其所含的实体-概念对（`cg_pairs.*.txt`）与实体-关系-实体三元组（`oie_triples.*.txt`）均被划分为训练集、验证集与测试集。