Resonac Develops an Original Generative AI, “Chat Resonac,” That Utilizes Accumulated In-house Materials Including Handwritten Documents

January 18, 2024
Resonac Holdings Corporation

Resonac Corporation (President: Hidehito Takahashi, hereinafter "Resonac") has developed an in-house generative AI system called "Chat Resonac" that can interactively use internal materials such as accumulated data and documents of the former Showa Denko K.K. and the former Hitachi Chemical Co., Ltd. Chat Resonac allows all users to access materials that only experienced employees know about and that inexperienced employees do not even know exist. As a result, it can help engineers develop new materials, such as semiconductors, by merging technologies from the former Showa Denko and the former Hitachi Chemical.

The former Showa Denko and the former Hitachi Chemical, the predecessors of Resonac Holdings Corporation and Resonac Corporation, each had a history of about 100 years and had accumulated more than fifty thousand materials related to materials development and manufacturing throughout their history. Veteran employees can search for and use the materials they need in paper files and databases. However, there are many cases where younger employees don't have access to these materials because they don't know they exist. In addition, there is a risk that these materials will not be used after the veteran employee retires.

To reduce such risk, Resonac developed Chat Resonac, a generative AI system that allows users to interactively use internal materials. Chat Resonac loads files of internal materials into a data storage server without leaking information outside the company. It receives data and feedback on generated responses from experienced employees, thereby improving the accuracy of generated responses. When the system reads handwritten documents and digitizes them with an optical character reader (OCR), it corrects omissions and errors using generative AI technologies. The system can also use data stored in digital lab notebooks.

Chat Resonac mediates between different generations and allows each employee to communicate and use information across the walls between the departments to which each employee belongs or formerly belonged. Employees can use Chat Resonac to search for the necessary information, including knowledge about the composition of materials and data on analysis results, and contact and communicate with other employees with such information.

Mr. Soichiro Takeshita, who is a professional at the Materials Analysis Center, Institute for Polymer Technology, talks about his expectations for the active use of Chat Resonac and his vision for the future: “I am confident that Chat Resonac will help the company to strengthen its technology development capability by enabling employees to share and utilize personal knowledge and know-how accumulated through past experiences.”

Resonac has developed two types of Chat Resonac applications: general purpose Chat Resonac and specialized Chat Resonac. The former handles information that can be shared among all employees. In contrast, the latter holds information that can be shared among employees belonging to specific departments, and there are more than 20 application programs under development as specialized Chat Resonac. In addition, Resonac plans to apply Chat Resonac to various fields, such as support tools for document creation, business efficiency improvement, and career development. Dr. Yoshishige Okuno, Fellow and Head of the Research Center for Computational Science and Informatics, says, "We have entered a new era in which humans can tackle more complex problems while entrusting tasks that AI-based systems can handle to AI-based systems. We will accelerate such reform.”

With the help of computational science and computer science, Resonac will continue to strive to shorten the development time of materials, including semiconductor materials that require rapid development.