WikiGraphs

WikiGraphs是一个由DeepMind创建的数据集，包含23,522对维基百科文章及其对应的Freebase知识图谱子图。该数据集旨在支持条件文本生成、图形生成和图形表示学习等研究。每个子图平均有38.7个节点和48.3条边，而每篇文章平均包含3,533.8个令牌。此数据集通过自动化的过程从现有的Freebase知识库中提取相关子图，为维基百科文章提供丰富的结构化信息。应用领域包括但不限于机器学习模型在长文本生成和基于图形的文本生成中的性能评估。

WikiGraphs is a dataset created by DeepMind, consisting of 23,522 pairs of Wikipedia articles and their corresponding Freebase knowledge graph subgraphs. This dataset is designed to support research in fields such as conditional text generation, graph generation, and graph representation learning. Each subgraph has an average of 38.7 nodes and 48.3 edges, while each article contains an average of 3,533.8 tokens. The dataset extracts relevant subgraphs from the existing Freebase knowledge base through an automated process, providing rich structured information for Wikipedia articles. Its application areas include, but are not limited to, performance evaluation of machine learning models in long text generation and graph-based text generation.