While good graphics are considered a critical component of an effective scientific paper, literature search engines have historically been text or citation based.

This week the MIT Technology Review highlighted the first visual search engine for scientific diagrams – Viziometrics. The team behind this pioneering project includes Electrical Engineering Ph.D. student Po-Shen Lee, iSchool professor and eScience Data Science Fellow Jevin West, and eScience Associate Director Bill Howe.

After assembling a searchable database of over 4.8 million figures from 650,000 papers, they trained a machine learning algorithm to search for and recognize five different types of figures and assess how the frequency and type of graphics varied by discipline and journal. The team found that the most successful papers tend to have a higher number of figures.

Learn more in this article from The Economist.