A method for retrieving and enhancing generation of a large model for historical building knowledge

By using a retrieval-enhanced intelligent processing system for historical buildings, which combines structured knowledge graphs and semantic retrieval, the system addresses the issues of insufficient professional knowledge coverage and information structuring in the field of historical buildings by general-purpose large language models. It enables verifiable generation and cross-modal reasoning, thereby improving the professionalism and accuracy of historical building protection and restoration.

CN122198113APending Publication Date: 2026-06-12GUANGXI UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUANGXI UNIV
Filing Date
2026-02-25
Publication Date
2026-06-12

AI Technical Summary

Technical Problem

The application of existing general-purpose large language models in the field of historical buildings has problems such as insufficient coverage of professional knowledge, the generation of results being prone to illusions, low degree of information structuring, and difficulty in unifying the modeling of multimodal and spatiotemporal information, which cannot meet the scientific and compliant requirements for the protection and restoration of historical buildings.

Method used

The system employs a retrieval-enhanced generation-based intelligent processing system for historical buildings. Through data acquisition and preprocessing, knowledge extraction and graph construction, semantic retrieval-enhanced generation, and structured auxiliary generation modules, combined with structured knowledge graphs and semantic retrieval, it achieves structured expression of knowledge in the architectural field, verifiable generation of facts, and cross-modal semantic reasoning.

Benefits of technology

It enhances the professionalism, accuracy, and feasibility of the model in the protection, restoration, and digital application of historical buildings, realizes the verification, structuring, and semantic integration of professional knowledge, supports cross-modal and spatiotemporal reasoning, and has the ability to automatically update knowledge.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122198113A_ABST
    Figure CN122198113A_ABST
Patent Text Reader

Abstract

The application relates to the field of artificial intelligence, in particular to a history building intelligent processing system and method based on retrieval enhancement generation, which comprises a data acquisition and preprocessing module, a knowledge extraction and graph construction module, a semantic retrieval enhancement generation module, a structured auxiliary generation module and an application layer module; the data acquisition and preprocessing module acquires original building information and pre-processes the original building information to obtain standard text; the knowledge extraction and graph construction module extracts building entities and semantic relationships thereof and constructs a history building knowledge graph; the semantic retrieval enhancement generation module carries out fragmentation and vectorization on the standard text, generates semantic vectors, establishes vector indexes, and generates features of an input building image; the structured auxiliary generation module introduces structured constraints from the knowledge graph, controls a generated structure, realizes controllable generation output, and the application layer module accesses an image recognition submodule to realize building style recognition based on visual features and cross-modal reasoning.
Need to check novelty before this filing date? Find Prior Art