MarkQA

MarkQA是一个大规模的知识库问答（KBQA）数据集，专注于复杂的数值推理。该数据集由南京大学国家软件新技术重点实验室创建，从1000个种子问题自动扩展到31902个示例。每个问题都配有相应的SPARQL查询、QDMR格式的逐步推理过程和PyQL程序。MarkQA旨在支持NR-KBQA任务，即在KBQA中结合多跳推理和数值推理，解决实际问题中需要精确数值计算的场景。MarkQA is a large-scale knowledge base question answering (KBQA) dataset dedicated to complex numerical reasoning. Developed by the State Key Laboratory for Novel Software Technology at Nanjing University, it was automatically expanded from 1,000 seed questions to 31,902 examples. Each question is paired with its corresponding SPARQL query, step-by-step reasoning process in QDMR format, and PyQL program. MarkQA aims to support the NR-KBQA task, which combines multi-hop reasoning and numerical reasoning within KBQA to tackle real-world scenarios that demand precise numerical computations.

搜集汇总

背景与挑战

背景概述

MarkQA是一个大规模的知识库问答数据集，基于Wikidata构建，包含31,902个问题，每个问题都标注了答案、SPARQL查询、QDMR和PyQL。该数据集是首个专注于复杂数值推理的KBQA数据集，旨在支持高级自然语言处理研究，特别是数值推理任务。