RAG mixed retrieval method and device

By dynamically adjusting vector and keyword retrieval weights, combined with multi-dimensional feature scoring and domain-specific parameter adjustment, the problems of query adaptability and semantic recognition of RAG technology in professional fields have been solved, achieving efficient retrieval and generation results.

CN121501944AActive Publication Date: 2026-02-10NANJING DOLPHIN INTELLIGENT TECH CO LTD
6 Cites 3 Cited by

Patent Information

Application Number
CN202511580922.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-10-31
Publication Date
2026-02-10
Estimated Expiration
2045-10-31

AI Technical Summary

Technical Problem

Existing RAG technology faces challenges in professional applications, including poor query adaptability due to fixed-weight fusion strategies, insufficient single-dimensional similarity evaluation, and insufficient adaptability of general models, which affect the quality and reliability of generated content.

Method used

By employing dynamically adjusted vector retrieval weights and keyword retrieval weights, combined with a multi-dimensional feature scoring strategy, including keyword matching degree, semantic similarity, document structure features, and positional weights, and adjusting parameters through a domain dictionary and a pre-trained model, we can achieve accurate recognition of professional terms and domain-specific semantic patterns.

Benefits of technology

It improves the retrieval accuracy of query texts and the relevance of response texts, overcomes the limitations of a single scoring dimension, and enhances semantic understanding capabilities and the reliability of generated content in professional fields.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides an RAG mixed retrieval method and device, and relates to the technical field of information retrieval, and the method comprises the following steps: obtaining a query text and a term library, and calculating to obtain a vector retrieval weight and a keyword retrieval weight according to the query text and the term library; obtaining a document library, and performing vector retrieval on the document library according to the query text and the vector retrieval weight to obtain a first candidate document set; performing keyword retrieval on the document library according to the query text and the keyword retrieval weight to obtain a second candidate document set; merging the first candidate document set and the second candidate document set to obtain a fused document set; and calculating, screening and sorting comprehensive feature scores of the documents in the fused document set to obtain a final document sequence, and inputting the final document sequence into the large language model to obtain an answer text. According to the method, the limitation of a single scoring dimension in professional field retrieval is overcome, and the correlation between the answer text and the query text is improved through multi-dimensional feature joint evaluation.
Need to check novelty before this filing date? Find Prior Art