Structured PubMed: 23.2M Biomedical Abstracts Now Searchable by Section

A team led by Chia-Hsuan Chang from Yale University has introduced Structured PubMed, the largest corpus of section-labeled biomedical abstracts compiled from the complete PubMed database, totaling over 23.2 million records. The dataset is split into two subsets: 5.9 million author-structured abstracts parsed from official XML files, and 17.2 million originally unstructured abstracts that were automatically labeled using a verbatim-extraction Large Language Model pipeline. All records are harmonized under a unified five-section schema (Introduction, Methods, Results, Discussion, Conclusion) and mapped to original PubMed IDs, publication types, and dates.

This resource addresses a major bottleneck in biomedical literature processing: the vast number of unstructured abstracts that hinder downstream text-mining and retrieval applications. The dataset can be used to train sentence-classification models, benchmark text-segmentation architectures, and perform large-scale section-specific information extraction at an unprecedented scale. By making every abstract's structure explicit, Structured PubMed enables more precise knowledge synthesis, better search capabilities, and supports the development of AI systems that can navigate biomedical literature section by section. The data and code are publicly available, opening new avenues for NLP research in healthcare and life sciences.