Retractions and other editorial notices (like corrections, errata, comments, etc.) are key mechanisms by which authors and publishers can update the broader community about changes made to existing research.

Communicating these updates effectively in a timely manner is essential to ensuring that newer research properly builds on existing work and doesn’t, for example, inadvertently cite something post-retraction.

In the last few years, a number of systems have been developed to track these notices, from RetractionWatch, which manually curates a list of retracted papers, to Crossref and Pubmed, which take updates from publishers and other sources, and provide that data on their platforms.

But the way that this information is communicated has always been inconsistent, with different publishers taking a variety of approaches to broadcasting their notices. Some will rename the paper to “Retracted”, others will add prefixes or suffixes to the original titles, while others will issue a new notice DOI indicating an existing paper is retracted, without any indication on the original one.

We hypothesized that we could build a system that detects the existence of editorial notices to papers directly from the metadata of papers, like their titles, to build a comprehensive database of retractions, withdrawals, and other notices that is updated daily and supplements notices we already get from Crossref and Pubmed.

You can read how this system works in our latest preprint, which includes a comparison of how our notices compare to those in Crossref and Pubmed; this database of notices we curate is also published here and available freely for anyone to use. We’ll periodically update this on a monthly basis.

Check it out and let us know what you think!