[19p-C601-1~15] 23.1 合同セッションN「インフォマティクス応用」

15:45 〜 16:00英語発表

[19p-C601-9] Question-answering-based approach for mining information from documents using Large Language Models

〇Luca Foppiano^1,2, Guillaume Lambard¹, Toshiyuki Amagasa², Masashi Ishii¹ (1.Data-driven Materials Design Group, CBRM, NIMS, 2.KDE, CCS, Univ. of Tsukuba)

キーワード：

magnetic materials、large language models、machine learning

We present an alternative method for extracting relevant information at document-level using a Question-Answer approach. We exploit large language models (LLM) ability to produce embeddings and chain of queries on extended context.