MeSHup: Corpus for Full Text Biomedical Document Indexing

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5473–5483 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5473 MeSHup: A Corpus for Full Text Biomedical Document Indexing Xindi Wang1,3, Robert E. Mercer1,3, Frank Rudzicz2,3,4 1Department of Computer Science, University of Western Ontario, London, Ontario, Canada 2Department of Computer Science, University of Toronto, Toronto, Ontario, Canada 3Vector Institute, Toronto, Ontario, Canada 4 Unity Health Toronto, Toronto, Ontario, Canada xwang842@uwo.ca, mercer@csd.uwo.ca, frank@cs.toronto.edu Abstract Medical Subject Heading (MeSH) indexing refers to the problem of assigning a given biomedical document with the most relevant labels from an extremely large set of MeSH terms. Currently, the vast number of biomedical articles in the PubMed database are manually annotated by human curators, which is time consuming and costly; therefore, a computational system that can assist the indexing is highly valuable. When developing supervised MeSH indexing systems, the availability of a large-scale annotated text corpus is desirable. A publicly available, large corpus that permits robust evaluation and comparison of various systems is important to the research community. We release a large scale annotated MeSH indexing corpus, MeSHup, which contains 1,342,667 full text articles in English, together with the associated MeSH labels and metadata, authors, and publication venues that are collected from the MEDLINE database. We train an end-to-end model that combines features from documents and their associated labels on our corpus and report the new baseline. Keywords: MeSH Indexing, Multi-label text classification 1. Introduction MEDLINE1 comprises 33 million (as of Nov. 2021) references to journal articles in the life sciences with a concentration on biomedicine, which is the National Library of Medicine’s2 (NLM) premier bibliographic database. It in- cludes textual information (title and abstract) and bibliographic information for articles from academic journals covering various disciplines of the life sciences and biomedicine. PubMed3 is a free search engine that provides free ac- cess to the MEDLINE database. In addition to MEDLINE, PubMed also provides access to the PubMed Central4 (PMC) repository that archives open-access, full-text scholarly articles in biomedical and life sciences journals. All records in the MEDLINE database are indexed with Medical Subject Headings (MeSH)5 – a controlled and hierarchically-organized vocabu- lary produced and maintained by the NLM. As of 2021, there are 29,369 main MeSH head- ings, and each citation is indexed with 13 MeSH terms on average. MeSH headings can be fur- ther qualified by 83 subheadings (also known as qualifiers). In addition, Supplementary Concept 1https://www.nlm.nih.gov/medline/medline_ overview.html 2https://www.nlm.nih.gov 3https://pubmed.ncbi.nlm.nih.gov/about/ 4https://en.wikipedia.org/wiki/PubMed_Central 5https://www.nlm.nih.gov/mesh/meshhome.html Records (SCRs) refer to specific chemical sub- stances in the MEDLINE records. MeSH indexing, a process that annotates doc- uments with concepts from established seman- tic taxonomies and ontologies, is important for biomedical text classification and information retrieval. MEDLINE citations are indexed by human annotators who read the full text of the article and assign the most relevant MeSH la- bels to the articles. The manual annotation pro- cess ensures the high quality of indexing but, inevitably, the cost can be prohibitive. There has …

Content truncated to 500 words