Wikipedia Multi-hop Search Problems
收藏arXiv2025-09-30 收录
下载链接：
https://ml4ai.github.io/OpenDomainFR/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了369个多跳搜索问题，这些问题源自于6,880篇维基百科文章的一个子集。每个搜索问题通过关系路径连接了一对实体。在创建此数据集时，使用了维基化过程来提取和规范化命名实体，并且通过在三个句子窗口内共同出现的实体来诱导关系。数据集的规模为369个搜索问题，涉及6,880篇文章，其任务是对实体间进行多跳关系搜索。
This dataset contains 369 multi-hop search questions derived from a subset of 6,880 Wikipedia articles. Each search question connects a pair of entities via a relational path. During the dataset construction, the wikification process was employed to extract and normalize named entities, with relational links induced based on entities co-occurring within a three-sentence window. The primary task of this dataset is to conduct multi-hop relational search between entities.
提供机构：
ML4AI
