A double-engine government affair question and answer method based on large model fine tuning and RAG retrieval

Abstract

The application discloses a kind of based on big model fine-tuning and RAG retrieval dual-engine government affair question and answer method, including steps: S1, constructs government affair knowledge base, government affair knowledge base includes structured policy data, unstructured document and government affair knowledge graph;S2, fine-tuning is carried out to basic big model using government affair field data, obtains government affair fine-tuning big model;S3, receives the natural language question input by user;S4, natural language question is carried out intent recognition and entity extraction, generates search strategy;S5, according to detection strategy and parallel or sequentially trigger RGA retrieval engine and government affair fine-tuning big model engine;The beneficial effects of the present application: the present application constructs the composite knowledge base including structured policy data, unstructured document and government affair knowledge graph, combines "fine-tuning big model+RGA retrieval" dual-engine collaborative architecture, both play the powerful semantic understanding and generation capability of big model, and also introduce authoritative knowledge source for fact constraint by RGA technology.

Images