BKRAG : A BGE Reranker RAG for similarity analysis of power project requirements
Jun Guo, Bojian Chen, Zhichao Zhao, Jindong He, Shichun Chen, Donglan Hu, Hao Pan · 2024
This paper proposes an innovative method called BKRAG (A BGE Reranker Retrieval Augmented Generation for similarity analysis of power project requirements), which integrates information retrieval techniques and NLP to achieve automated analysis and similarity evaluation of power project requirements. The core of the BKRAG method lies in the utilization of a Rerank model to re-rank the initially retrieved candidate documents, improving their semantic matching degree with user queries, thereby optimizing the results of requirements similarity analysis. In this paper, we elaborate on the construction principles and workflow of the BKRAG method and verify its effectiveness through a series of experiments. Results demonstrate that the BKRAG can significantly improve the retrieval accuracy of power project requirement documents and the performance of requirements similarity analysis. The research findings of this paper not only provide a new solution for the field of power project requirements analysis, but also offer new insights into the cross-application of information retrieval and natural language processing technologies.