Knowledge Management for Design Development and Maintenance through Text Mining of Product Information

(Constructing Knowledge Management Systems for Automobile Failure with Graph RAG)

Abstract

In the automotive industry, there is a growing demand for handing down the skills of failure analysis these days. However, failure events are phenomena that occur in a chain reaction, making them difficult for beginners to understand. While a Knowledge Graph (KG) that can structure information is effective in describing failure events, understanding KGs themselves is not easy. On the other hand, there is growing anticipation for the use of Graph RAG, a type of Retrieval-Augmented Generation (RAG) technology that utilizes large language models (LLMs) and KGs for knowledge management. However, when using Graph RAG with an existing knowledge graph for automobile failures, several issues arise because it adopts Semantic Parsing-based Method. This study proposes an optimized Graph RAG pipeline for existing knowledge graphs by adopting Information Retrieval-based Method. By using an original Q&A dataset, the ROUGE f1 score of the sentences output by the proposed method showed an average improvement of 157.9% compared to the existing method. This indicates the effectiveness of the proposed method for failure analysis of automobiles.