(569am) Natural Language Processing Aided Design of Electrochemical CO2 Reduction Catalysts

Electrochemical CO₂ reduction has the potential to be a feasible pathway toward sustainable production of fuels and chemicals while reducing greenhouse gas emissions. However, the complexities between catalyst parameters such as pH, voltage, catalyst metal and structure make designing an optimal catalyst difficult. Historically chemists and engineers have taken a manual trial and error type approach to designing an optimal catalyst for CO₂ reduction leading to vast amounts of literature data scattered across academic journals. Additionally, to fully comprehend all the available literature on the design of electrocatalysts, researchers would require years of extensive study and analysis. Our goal is to aid in the literature search by extraction of data from the literature using a custom pdf to text parser that feeds into a Natural Language Processing algorithm that accurately extracts catalyst properties. The mined catalyst properties are subsequently input into a database where machine learning techniques such as artificial neural networks are used to unravel the complex interplay between catalyst properties aiding in the discovery of optimal catalyst for CO₂ electroreduction.