Automated Construction of 3D Model Knowledge Graphs and Knowledge Extraction using Large Language Models

In the field of design, the construction of a knowledge graph for 3D product models is crucial for obtaining design norms and successful cases. Traditional knowledge extraction methods rely on industry experts to manually define rules, which suffer from low recall rates, poor robustness, and difficult maintenance. To address these issues, this paper proposes an automatic construction system for a comprehensive knowledge graph of 3D product models. By leveraging large model technology, it simplifies automated semantic understanding and representation, significantly reducing the reliance on manual labor in traditional 3D product model knowledge extraction methods. It is the first to complete the automatic extraction of 3D model knowledge based on large language model technology. Additionally, this paper presents a workflow agent module for 3D model knowledge extraction. This module is plug-and-play without the need for additional training and supports further fine-tuning of the large language model to enhance the accuracy of knowledge extraction, demonstrating good generalization. Experiments show that this method is more efficient than manual extraction, providing an efficient and automated solution for constructing 3D model knowledge graphs.