Development of Artificial Intelligence that Automatically builds first-principle models from the literature

2022 Fiscal Year Final Research Report

• Project/Area Number: 21K18849
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Review Section: Medium-sized Section 27:Chemical engineering and related fields
• Research Institution: Kyoto University
• Principal Investigator: KANO Manabu 京都大学, 情報学研究科, 教授 (30263114)
• Co-Investigator(Kenkyū-buntansha): 加藤 祥太 京都大学, 情報学研究科, 助教 (60883402)
• Project Period (FY): 2021-07-09 – 2023-03-31
• Keywords: 人工知能 / 第一原理モデル / 物理モデル / 自然言語処理 / 数式処理 / 化学工学 / プロセスシステム工学

Outline of Final Research Achievements

To optimize the design and operation of a manufacturing process, a first-principle model (physical model) of the process plays a crucial role. However, building such a model requires an enormous effort. The ultimate goal of this research is to realize an artificial intelligence (AI) that automatically builds a first-principle model by collecting relevant literature, understanding the meaning of variables and formulas in the literature, and extracting and editing information necessary for model construction, given a target process. We developed several elemental technologies, including a method for extracting variables and their definitions and a method for determining the equivalence of variables and formulas, and also we developed a prototype of AI.

Free Research Field

プロセスシステム工学

Academic Significance and Societal Importance of the Research Achievements

様々な能力を持つ人工知能（AI）が開発され，活用されている．しかし，製造業で決定的に重要でありながらその構築が困難な第一原理モデルを自動的に構築してくれるAIを開発しようという試みは本研究が世界初である．自然言語処理や数式処理の新しい方法論を開発するという点で学術的意義が大きいのみならず，第一原理モデル自動構築AIが完成すれば，製造業の生産性を劇的に向上させることができるため，その社会的意義も計り知れない．