WIKES (Wiki Entity Summarization Benchmark)

WIKES是一个综合性的知识图谱实体摘要基准数据集，由都灵理工大学等机构创建。该数据集包含约500个种子节点，通过随机漫步方法从Wikidata和Wikipedia中提取，以保持真实世界知识图谱的复杂性。WIKES利用Wikipedia摘要自动生成高质量、无偏见的实体摘要，无需人工标注，适用于多个领域，旨在解决现有数据集规模小、依赖人工标注和忽视图结构信息的问题。

WIKES is a comprehensive knowledge graph entity summarization benchmark dataset developed by institutions including Polytechnic University of Turin. This dataset contains approximately 500 seed nodes, which are extracted from Wikidata and Wikipedia via random walk methods to preserve the complexity of real-world knowledge graphs. WIKES automatically generates high-quality, unbiased entity summaries using Wikipedia abstracts without requiring manual annotation, and is applicable across multiple domains. It aims to address the shortcomings of existing datasets, such as small scale, dependence on manual annotation, and neglect of graph structural information.