CoronaCentral

This describes the output file for the CoronaCentral data. The scripts used to create it are hosted in the corona-ml Github repo. The sources for the documents before processing for CoronaCentral are PubMed and CORD-19.

The file is a gzipped JSON document containing one record per document. Each document has at least one of: a PubMed ID, a CORD-19 ID (cord_uid), a DOI or a URL.

The fields that documents should have are:

• pubmed_id: PubMed identifier (optional)
• pmcid: PubMed Central identifier (optional)
• doi: Digital object identifier (optional)
• cord_uid: CORD-19 identifier (optional)
• url: URL
• journal: Journal/preprint server
• publish_year: Year of publication (optional)
• publish_month: Month of publication (optional)
• publish_day: Day of publication (optional)
• title: Title of article
• abstract: Abstract of article (optional)
• is_preprint: Whether the article is a preprint
• topics: Predicted topics for article
• articletypes: Predicted article types for article
• entities: Extracted entities (e.g. drugs) with identifiers and locations within text

Please report issues to the corona-ml Github issues page.