Where available, the right side of paper landing pages contain information about different content types that are mentioned in the paper, which include data sets, methods, key resources, and clinical trials. These mentions increase the discoverability and accessibility of the content, as well as helping users save time and effort in finding the right type of information they need for their research and knowledge discovery. Surfacing the content mentioned in a paper in an easy-to-view manner also helps users determine whether or not the paper is of interest to them. The mentions are a valuable resource to early career researchers as well as the more experienced users.
Datasets and methods mentions have been extracted from the paper by a named entity recognition machine learning model, whereas research resources and clinical trials have been extracted by a heuristic-based algorithmic approach.
The mentions you see in Meta today are the results of our initial machine learning effort to mine data from the full text of papers, and thus are considered as a beta stage. We will be improving our machine learning model and making changes over time. For more background, see our blog post on this effort. We welcome your feedback.