[3K5-IS-2b] Machine learning

[3K5-IS-2b-03] Information Extraction of ORR Catalyst for Fuel Cell from Scientific Literature

〇HEIN HTET¹, AMGAD AHMED ALI IBRAHIM¹, YUTAKA SASAKI ², RYOJI ASAHI¹ (1. Institute of Innovation for Future Society, Nagoya University, Nagoya, Japan, 2. Computational Intelligence Laboratory, Toyota Technological Institute, Nagoya, Japan)

キーワード：Natural Language Processing (NLP), Information Extraction, Fuel Cell Catalysts, Oxygen Reduction Reaction (ORR)

The development of advanced catalysts for the Oxygen Reduction Reaction (ORR) is critical for improving the performance and efficiency of Polymer Electrolyte Fuel Cells (PEFCs). However, the vast and growing body of scientific literature poses challenges for researchers aiming to identify key insights. This study focuses on the information extraction of ORR catalysts from fuel cell-related literature using a hybrid approach combining manual annotation and automated machine learning techniques. A comprehensive dataset was constructed through the Brat annotation tool, identifying 12 critical entities such as catalyst, support, and value, alongside two relationship types: equivalent and related_to. The annotated data was used to fine-tune the DyGIE++ framework with the pre-trained BERT models. The model demonstrated effective performance in extracting complex material science concepts and their interrelationships. The finding suggests that this automated framework can accelerate catalyst discovery by providing structured, high-quality data for downstream analysis. This research highlights the potential of Natural Language Processing (NLP) in enabling efficient literature mining and fostering advancements in clean energy techniques.