KAMEL

该数据集是基于KB Wikidata的知识构建的问答数据集，涵盖了234种被视为工具的关系。它不仅包含了一个黄金训练集，还包含了一个由ChatGPT生成的合成训练集。规模上，数据集以234种关系作为工具，任务定位于基于知识库的问答。This question answering (QA) dataset is built upon the knowledge from Wikidata, a prominent knowledge base (KB). It encompasses 234 categories of relations utilized as tools. The dataset comprises both a gold standard training set and a synthetic training set generated by ChatGPT. In terms of task design, the dataset employs these 234 relations as core tools, with the primary task being knowledge base-based question answering (KBQA).