Machine learning tools for peer review

Shanghai 1

George Garrity reasons that most people underestimate the amount of work that goes into the process. “The publisher distributes your content, they polish it, they make sure there’s an archival version, but they also provide all the necessary quality control, and this is typically done by peer review,” he said.

The peer review process is essential for checking that valid arguments and conclusions are present, with appropriate priority, provenance and originality. However, it can be costly and very time-consuming, thus there is great interest in automating as much of the process as possible.

Hoping to do just that, a suite of tools from NamesforLife allows processing of a raw manuscript in mere minutes, validating facts, structure, terminology and cited resources, and annotating any “red flags”. The automation can then extend to the peer review stage, cross-checking the intended submission with a pool of some 40,000 documents in order to identify candidate reviewers based on relevant publication records.

The process removes selection bias, screens for conflicts of interest, and tracks ongoing reviewer performance. What’s more, it keeps up-to-date contact information for reviewers, and constructs a compelling email to send to the reviewer to encourage their participation.

How to ensure content quality in a world of overwhelming scientific complexity