OpenBookQA

OpenBookQA是一个专为开放式书籍问答设计的新型数据集，由艾伦人工智能研究所创建。该数据集包含约6000个多选题，每个问题都与一个来自1326个基础科学事实的‘开放书籍’相关联。这些问题的目的是测试对这些事实及其在新情境中应用的理解。数据集的创建过程涉及多阶段众包和部分专家筛选，确保问题的质量和难度。OpenBookQA的应用领域主要集中在科学问答，旨在评估和提升机器对科学知识的理解和推理能力。

OpenBookQA is a novel dataset tailored for open-book question answering, developed by the Allen Institute for AI. It comprises roughly 6,000 multiple-choice questions, each linked to an "open book" sourced from 1,326 fundamental science facts. The questions are designed to assess understanding of these facts and their application in novel contexts. The dataset is built through multi-stage crowdsourcing and partial expert filtering to ensure the quality and difficulty of the questions. Its main application domains focus on scientific question answering, aiming to evaluate and enhance machines' understanding of scientific knowledge and their reasoning capabilities.