Wikidata KGQA Dataset

该数据集是从Mintaka Wikidata问题库中精心挑选的120个问题，专注于五种现实生活中的问题类型，以生成自然语言到知识图谱查询的任务。这些问题被分为五种类型：多跳、比较、是非、通用和交叠。为了考虑大型语言模型（LLMs）的概率性质，每个问题都执行了三次。该数据集的规模为120个问题，任务定位于知识图谱问答。

This dataset includes 120 carefully curated questions sourced from the Mintaka Wikidata question repository, tailored for the natural language-to-knowledge graph query generation task and focusing on five real-world question categories. These questions are divided into five types: multi-hop, comparative, yes/no, general, and overlapping. Given the probabilistic nature of Large Language Models (LLMs), each question was run three times. This dataset, which totals 120 questions, is targeted at knowledge graph-based question answering tasks.