[15p-S2_204-1~18] 23.1 Joint Session N "Informatics"

3:15 PM - 3:30 PM JST(6:15 AM - 6:30 AM UTC)

[15p-S2_204-7] Extraction of compound synthesis methods from paper in material science

〇Hironori Nakaoka¹, Inui Takashi² (1.Sumitomo Metal Mining, 2.University of Tsukuba)

Keywords:

Text Mining,Natural Language Processing

In the field of materials research, research results are published in text media such as papers and patents, and a vast amount of knowledge and data has been accumulated. However, because text is unstructured, it is currently not being fully utilized. In this study, we investigated methods for extracting synthesis methods from battery material literature and structuring the text data. We trained extraction models based on Bidirectional Encoder Representations from Transformers (BERT) and Generative Pre-trained Transformer (GPT), and both models extracted synthesis methods with practical accuracy.