Paper ID: 2204.12070

Symlink: A New Dataset for Scientific Symbol-Description Linking

Viet Dac Lai, Amir Pouran Ben Veyseh, Franck Dernoncourt, Thien Huu Nguyen

Mathematical symbols and descriptions appear in various forms across document section boundaries without explicit markup. In this paper, we present a new large-scale dataset that emphasizes extracting symbols and descriptions in scientific documents. Symlink annotates scientific papers of 5 different domains (i.e., computer science, biology, physics, mathematics, and economics). Our experiments on Symlink demonstrate the challenges of the symbol-description linking task for existing models and call for further research effort in this area. We will publicly release Symlink to facilitate future research.

Submitted: Apr 26, 2022