We invite the scientific community to test a search engine we have developed for the biomedical literature (see http://pubmed.ict.griffith.edu.au). The aim of this Article-based PubMed Search Engine is to find only those publications that are most relevant to any particular article.

By extracting keywords from a paper's title and abstract, the search engine reveals every related study indexed in PubMed up to 26 July 2017 — typically as many as 60 publications per paper. We then refine the results by asking researchers to score them for relevance, which takes just a few minutes. All such evaluations will be curated into a benchmark data set that can be downloaded and distributed for free.

The web server has attracted several thousand visitors from more than 50 countries since its launch in July. Preliminary data indicate an overall success rate of 80% in identifying relevant articles.