RxnScribe: A Sequence Generation Model for Reaction Diagram Parsing.

Reaction diagram parsing is the task of extracting reactionschemesfrom a diagram in the chemistry literature. The reaction diagramscan be arbitrarily complex; thus, robustly parsing them into structureddata is an open challenge. In this paper, we present RxnScribe, amachine learning model for parsing reaction diagrams of varying styles.We formulate this structured prediction task with a sequence generationapproach, which condenses the traditional pipeline into an end-to-endmodel. We train RxnScribe on a dataset of 1378 diagrams and evaluateit with cross validation, achieving an 80.0% soft match F1 score,with significant improvements over previous models. Our code and dataare publicly available at https://github.com/thomas0809/RxnScribe.