Dynamic-KGQA
收藏资源简介:
Dynamic-KGQA是一个由华盛顿大学研发的动态问答数据集生成框架,旨在从知识图谱中生成适应性的问答数据集。该框架能够生成新的数据集变体,同时保持底层分布不变,确保公平和可复现的评价。此外,它还支持细粒度的数据集特性控制,可以生成特定领域和主题的问答数据集。Dynamic-KGQA通过构建紧凑、语义连贯的子图,促进知识图谱问答模型的训练和评估,提高了模型利用结构化知识的能力。
Dynamic-KGQA is a dynamic question answering dataset generation framework developed by the University of Washington, intended to generate adaptive question answering datasets from knowledge graphs. This framework can create novel dataset variants while maintaining the underlying data distribution, thus ensuring fair and reproducible evaluations. Moreover, it supports fine-grained control over dataset characteristics, enabling the generation of question answering datasets for specific domains and topics. Dynamic-KGQA facilitates the training and evaluation of knowledge graph question answering models by constructing compact and semantically coherent subgraphs, thereby enhancing the models' ability to leverage structured knowledge.




