JF17K-3

该数据集是从JF17K中提取的一个子集，仅包含3元关系事实。此外，事实数据被按照JF17K原始的划分方式，分为训练集、验证集和测试集。该数据集的规模是3元关系，任务目标是链接预测。

This dataset is a subset extracted from JF17K, which exclusively contains ternary relational facts. Moreover, the factual data is split into training, validation, and test sets in accordance with the original partitioning scheme of JF17K. This dataset focuses on ternary relational data, with the task objective being link prediction.

搜集汇总

背景与挑战

背景概述

JF17K-3是一个N元关系知识库数据集，用于知识图谱补全和链接预测任务，在实验中表现出较高的预测准确率（MRR达0.732）。该数据集是论文《Generalizing Tensor Decomposition for N-ary Relational Knowledge Base》的核心实验数据之一。