Extraction of optimal synthesis conditions from scientific literature using a knowledge graph

Abstract

Autonomous experiments for material synthesis have been developing with increasing speed. With the expanding search space for accessible materials, the efficient collection of prior knowledge, particularly synthesis conditions, before autonomous experiments is crucial in reducing the number of trials. In this study, we developed a workflow that systematically extracts synthesis conditions from scientific literature. We constructed a knowledge graph to test the simple hypothesis that significant correlations exist between the physical properties and synthesis conditions that appear nearby in a text. The performance of this scheme was demonstrated by extracting the appropriate thin-film synthesis conditions for conductive Nb-doped TiO2. The proposed methodology is expected to accelerate autonomous material synthesis experiments.

Keywords

• Natural language processing
• knowledge graph
• autonomous experiments
• thin-film deposition
• material synthesis