rschf/pubmedqa

PubMedQA是一个从PubMed摘要中收集的生物医学问题回答数据集。该数据集的目的是使用对应的摘要回答研究性的生物医学问题，答案形式为Yes/No/Maybe。数据集包含1k个专家注释的实例（PQA-L）、61.2k个未标注的实例（PQA-U）和211.3k个人工生成的QA实例（PQA-A）。每个PubMedQA实例包括一个问题（现有研究文章标题或由此派生）、一个上下文（对应的PubMed摘要，不包括结论）、一个长答案（摘要的结论，假定回答了研究问题）和一个Yes/No/Maybe的总结性答案。

PubMedQA is a biomedical question answering (QA) dataset collected from PubMed abstracts. The task of PubMedQA is to answer research biomedical questions with yes/no/maybe using the corresponding abstracts. PubMedQA consists of 1k expert-annotated instances (PQA-L), 61.2k unlabeled instances (PQA-U), and 211.3k artificially generated QA instances (PQA-A). Each PubMedQA instance includes a question (either an existing research article title or derived from one), a context (the corresponding PubMed abstract without its conclusion), a long answer (the conclusion of the abstract, presumably answering the research question), and a yes/no/maybe summary answer.