|
|
|
| LLM-Based Intelligent Association Between Protocol Templates and Object Models |
| ZHANG Yusheng1， XU Yonghui1， ZHOU Yuqi2， DU Jiang2， WEI Changan1 |
| 1. School of Electronics and Information Engineering， Harbin Institute of Technology， Harbin 150001， Heilongjiang， China； 2. Norinco Group Test and Measuring Institute， Huayin 714200， Shaanxi China |
|
|
|
|
Abstract To bridge the semantic gap resulting from the transmission of external products via heterogeneous communication protocols, in conjunction with the operation of the joint test platform utilizing the built-in standard object model (SDO) subscription-publish mechanism, this paper introduces an intelligent association method and system for protocol template-object models founded on large language models (LLMs). The process took the XML protocol template and interface description model as input, and generated the intermediate representation through structured preprocessing. The local knowledge base was constructed under the retrieval, augmentation, and generation (RAG) framework to uniformly store SDO definitions, historical association pairs, and proprietary corpora. And through the “dual-channel index,” combined with sparse keyword matching and dense semantic vector retrieval, highly recalled candidates for protocol elements and SDO attributes were generated. Subsequently, a lightweight, high-performance large-scale reasoning model was adopted to perform semantic disambiguation and consistency verification on candidate pairs, and the optimal match was output alongside the prompt-word norms and rule constraints. For “strange inputs” such as abbreviations, pinyin, and mixed Chinese-English writing, a multi-agent diversion parsing strategy was introduced, significantly enhancing the robustness against non-standard expressions. The system ultimately automatically generated a standardized XML association relationship list, supporting traceable evidence fragment backlinks and threshold filtering to avoid low-correlation strong matching. The prototype software integrated the full-process modules of file parsing, knowledge retrieval, intelligent matching, and result export. In verifying typical protocol templates and object model scenarios, it demonstrates robust alignment capabilities and engineering availability across languages and naming systems, providing an efficient and scalable semantic mapping path for rapid access to external resources by the joint test platform.
|
|
Received: 10 December 2025
Published: 13 January 2026
|
|
|
|
|
|
| [1] |
LYU Tao, CHEN Jing, XUE Shanye. Joint Analysis of Product Indicators Based on Multi-Source and Multi-Modal Data in Discrete Manufacturing[J]. Air & Space Defense, 2024, 7(5): 110-119. |
| [2] |
ZHANG Tao, ZHANG Long, ZHANG Guoqun, JIA Guifen, SUN Junge. Review of Research on Software-Defined System Technologies[J]. Air & Space Defense, 2024, 7(2): 1-7. |
|
|
|
|
2025, Vol. 8 
