Skip to main navigation Skip to search Skip to main content

Construction and Practice of a Standard LLM System for Electrical Apparatus Based on Residual Network Algorithm Feature Extraction

  • Chaoqun Cheng
  • , Jiaxin You
  • , Weizhuo Zhang
  • , Jiajing Tang
  • , Lingling Li*
  • , Yuhang Wang
  • *Corresponding author for this work
  • Harbin Institute of Technology
  • Xiamen University of Technology
  • Xiamen Key Laboratory of Relay and Control Components Reliability

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Standard retrieval in the electrical apparatus field faces challenges such as the fragmentation of multimodal information and insufficient semantic understanding. This paper proposes a layered processing architecture: First, a keyword indexing system is built based on the TF-IDF and TextRank algorithms to achieve the batch annotation of core terms in standard documents. Then, the Llama3 system is locally deployed, integrating text, image, and video features. The experiments show that, compared to traditional pure text model retrieval, Precision significantly improves when handling moderate-sized datasets. The system’s F1 score reaches 91.0% with a dataset of 50 documents, significantly addressing the semantic gap problem in traditional methods. It provides a practical intelligent solution for industrial standardization.

Original languageEnglish
Title of host publicationThe Proceedings of the 20th Annual Conference of China Electrotechnical Society
EditorsQingxin Yang, Dianguo Xu, Xuerong Ye, Qiuyue Nie, Yueshi Guan
PublisherSpringer Science and Business Media Deutschland GmbH
Pages197-206
Number of pages10
ISBN (Print)9789819583362
DOIs
StatePublished - 2026
Event20th Annual Conference of China Electrotechnical Society, ACCES 2025 - Harbin, China
Duration: 19 Sep 202521 Sep 2025

Publication series

NameLecture Notes in Electrical Engineering
Volume1561 LNEE
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Conference

Conference20th Annual Conference of China Electrotechnical Society, ACCES 2025
Country/TerritoryChina
CityHarbin
Period19/09/2521/09/25

Keywords

  • Electrical apparatus
  • Feature extraction
  • LLM
  • Multimodal retrieval
  • Residual network

Fingerprint

Dive into the research topics of 'Construction and Practice of a Standard LLM System for Electrical Apparatus Based on Residual Network Algorithm Feature Extraction'. Together they form a unique fingerprint.

Cite this