ForecastTKGQuestions

ForecastTKGQuestions是一个大规模的时序知识图谱问答基准数据集，由慕尼黑大学创建。该数据集包含三种类型的预测问题：实体预测问题（EPQs）、是未知问题（YUQs）和事实推理问题（FRQs）。数据集基于Integrated Crisis Early Warning System（ICEWS）数据集生成，每个预测问题都标注了时间戳，模型只能使用该时间戳之前的信息进行答案推理。该数据集旨在测试时序知识图谱问答模型的预测能力，特别是在未来推理方面的应用。

ForecastTKGQuestions is a large-scale benchmark dataset for temporal knowledge graph question answering, developed by Ludwig Maximilian University of Munich. This dataset encompasses three types of prediction queries: Entity Prediction Queries (EPQs), Yes-Unknown Queries (YUQs), and Fact Reasoning Queries (FRQs). The dataset is generated based on the Integrated Crisis Early Warning System (ICEWS) dataset. Each prediction query is annotated with a timestamp, and models are only permitted to use information prior to this timestamp for answer inference. This dataset is designed to evaluate the predictive capabilities of temporal knowledge graph question answering models, especially their applications in future-oriented reasoning.

搜集汇总

背景与挑战

背景概述

ForecastTKGQuestions是一个由慕尼黑大学创建的大规模时序知识图谱问答基准数据集，基于ICEWS数据集生成，包含实体预测、是未知问题和事实推理三种问题类型。该数据集的特点是每个问题都标注了时间戳，要求模型仅使用时间戳前的信息进行答案推理，旨在测试时序知识图谱问答模型在未来预测方面的能力。