[3F1-OS-8] OS-8

9:15 AM - 9:30 AM JST(12:15 AM - 12:30 AM UTC)

[3F1-OS-8-02] Development of Organic Synthesis Experimental Information Utilization System Integrating Knowledge Graphs and Generative AI

〇Toshiharu Morishita¹, Haruka Hori¹, Masafumi Tsuyuki², Shotaro Agatsuma², Ryota Noma² (1. Chugai Pharmaceutical Co., Ltd., 2. Hitachi, Ltd.)

Keywords:

Knowledge Graph,GraphRAG,Large Language Model (LLM),Organic Synthesis Experiment

In chemical pharmaceutical process development, researchers spend considerable time searching for information to optimize experimental conditions. This study presents an organic synthesis experimental information utilization system integrating knowledge graphs and generative AI. The system comprises: (1) a specialized ontology for organic synthesis experiments with knowledge graph extraction, (2) an integrated search system enabling cross-domain queries across laboratory notebooks, reports, presentations, and publications, and (3) an experimental condition recommendation engine utilizing multi-agent systems and deep research capabilities. Validation demonstrated knowledge graph extraction accuracy exceeding 95% with response times of several seconds to tens of seconds, receiving positive user feedback. The system significantly enhances efficiency in information retrieval, summarization, and experimental condition proposals, allowing to focus on creative thinking.