Sign In

Communications of the ACM

ACM TechNews

Atlas of Biomedical Literature Could Help Track Down Fabricated Studies

View as: Print Mobile App Share:
An atlas groups together papers from distinct fields into large, color-coded bundles.

To create the atlas, Kobak’s team downloaded the abstracts of nearly 21 million English-language articles from the PubMed search engine.

Credit: González-Márquez et al

An atlas of biomedical literature furnishes a "bird's-eye" map of the relationships among nearly 21 million English-language research papers, which "should prove quite useful in looking at high-level trends," according to Kevin Boyack at research consulting company SciTech Strategies.

Researchers at Germany's University of Tübingen (TU) and information cartography company Nomic AI downloaded abstracts of the articles from the PubMed search engine, then sorted them by similarity using the PubMedBERT artificial intelligence (AI) large language model.

The model clustered similar publications into color-coded "neighborhoods," and then the researchers explored trends such as gender gap variation across the literature. They also underscored nearly 12,000 papers labeled as retracted on PubMed that often formed "islands."

TU's Dmitry Kobak suggested this approach could help identify other papers suspected of fabrication.

From Science
View Full Article


Abstracts Copyright © 2023 SmithBucklin, Washington, D.C., USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account