WN18RR

WN18RR是评估知识图谱完成和链接预测模型的重要基准数据集。它是从WordNet知识图谱中提取的改进版本，旨在解决原始WN18数据集中的测试集泄漏问题。该数据集包含大约 40,943 个实体和 11 个关系，共有 93,003 个三元组，分为训练集、验证集和测试集。

WN18RR is an important benchmark dataset for evaluating knowledge graph completion and link prediction models. It is an improved version extracted from the WordNet knowledge graph, designed to address the test set leakage issue in the original WN18 dataset. This dataset contains approximately 40,943 entities and 11 relationships, with a total of 93,003 triples, and is divided into training, validation and test sets.

搜集汇总

背景与挑战

背景概述

WN18RR是一个用于知识图谱完成和链接预测评估的关键基准数据集，基于WordNet知识图谱构建，通过改进原始WN18数据集解决了测试集泄漏问题。该数据集包含约40,943个实体、11个关系和93,003个三元组，并提供了训练集、验证集和测试集的划分，适用于模型开发和性能测试。