Knowledge-Enhanced Large Language Model-Based Assistance Training System for Subway Maintenance Personnel
Da Cheng Yang, Hongbo Wang, Shanzhong Shao, Shutian Liu · 2024
To address the various challenges faced in training urban rail transit system maintenance personnel, this paper proposes a solution for developing a training system for subway maintenance personnel using knowledge graphs and a Retrieval-Augmented Generation (RAG)-enhanced large language model. The approach involves first creating a fine-tuning dataset from subway maintenance technical documents to fine-tune the large language model. This fine-tuned model then assists in constructing a subway maintenance knowledge graph. Concurrently, a vector database of subway maintenance knowledge is established. Finally, a question-answering system leveraging both the knowledge graph and the vector database as external knowledge sources is developed to support the training of subway maintenance personnel. Results demonstrate that this system can effectively enhance the learning efficiency of maintenance staff.