EFRAG:Embedding-FineTuning Retrieval Augmented Generation for QA in power projects
Jun Guo, Jianye Huang, Zhichao Zhao, J H Chen, Yuyou Weng, Guoqing Lin · 2024
This paper introduces an innovative method called EFRAG (Embedding-FineTuning Retrieval Augmented Generation for QA in power projects), which incorporates an external knowledge base as an additional information source to precisely answer questions related to power projects. The core of EFRAG lies in Embedding-FineTuning to enhance the model’s comprehension of user queries and project documents, thereby improving matching capability in power-related QA tasks and assisting the LLM in deriving accurate answers. This paper details the fundamental principles and implementation process of EFRAG and validates its effectiveness through a series of experiments. The experimental results indicate that EFRAG significantly improves the accuracy of QA results. The research findings not only provide a solution for enhancing power-related QA but also offer new insights into empowering the power industry through large language models.