HOVER
收藏资源简介:
HOVER是一个用于多跳证据提取和事实验证的数据集,由北卡罗来纳大学教堂山分校创建。该数据集包含26,000个需要从多达四个英文维基百科文章中提取证据的声明,并涉及多种形状的推理图。数据集的创建过程分为三个阶段,通过训练有素的众包工作者进行问题重写和实体替换。HOVER数据集旨在推动复杂多跳推理在信息检索和验证领域的研究,特别是在处理长距离依赖关系和多文档信息整合方面。
HOVER is a dataset for multi-hop evidence extraction and fact verification, developed by the University of North Carolina at Chapel Hill. This dataset contains 26,000 claims that require extracting supporting evidence from up to four English Wikipedia articles, and covers reasoning graphs of various structures. The creation process of the HOVER dataset is divided into three stages, with trained crowd workers undertaking question rewriting and entity replacement. The HOVER dataset aims to advance research on complex multi-hop reasoning in the fields of information retrieval and fact verification, particularly in addressing long-distance dependencies and multi-document information integration.

- 1HoVer: A Dataset for Many-Hop Fact Extraction And Claim Verification北卡罗来纳大学教堂山分校 · 2020年



