Retrieval Augmented Generated for Earth Fault Simulation
Farid Cenreng, Rona Riantini, Supeno Mardi Susiki Nugroho, Reza Fuad Rachmadi · 2024
Ship electrical troubleshooting remains a challenging process and requires a lot of practice. A simulated learning media has been developed to train students to solve electrical problems. To achieve optimal outcomes, students require training guidance, but the availability of mentor-discussion guidance is constrained by time and resources. It is proposed the use of a chatbot for effective consultation by incorporating a Retrieval-Augmented Generation (RAG) approach. This technique enables the chatbot to pull information from a predefined set of documents or a knowledge base when answering questions. The test results indicated that the model achieved a reach of Faithfulness of 93,095 %, Answer Relevance of 98,299%, and Answer Correctness of 77,093 %. This greatly improves the quality and depth of the answers, offering trainees a richer learning experience.