Math Formula Question Answering Dataset (MFQuAD)

MFQuAD是由IBM研究院欧洲分部创建的数据集，专注于数学标识符描述阅读任务。该数据集包含7508个标注的标识符出现，源自100篇维基百科文章，旨在通过提取或生成公式标识符的描述来理解文档内容。MFQuAD支持现代机器学习和自然语言处理技术，适用于数学信息检索、数学内容的无障碍访问、自动化特征工程和文档理解等领域。数据集的创建过程中，采用了保守的上下文选择策略，确保描述的完整性和准确性。

MFQuAD is a dataset created by IBM Research Europe, focusing on the mathematical identifier description reading task. It contains 7508 annotated occurrences of identifiers derived from 100 Wikipedia articles, and aims to enhance document understanding by extracting or generating descriptions for mathematical formula identifiers. MFQuAD supports modern machine learning and natural language processing technologies, and is applicable to domains including mathematical information retrieval, accessibility of mathematical content, automated feature engineering, and document understanding. During the dataset creation process, a conservative context selection strategy was adopted to ensure the completeness and accuracy of the descriptions.