基于低秩适配微调大语言模型构建断纸信息知识图谱

打开文本图片集

中图分类号：TS736+.4 文献标识码：A DOI： 10.11980/j.issn.0254-508X.2026.04.024

Construction of a Web Break Information Knowledge Graph Based on Large Language Models Fine-tuned with Low-rank Adaptation

ZENG Qingyu HONG Mengna*LI Jigeng （State KeyLab of Advanced Papermaking and Paper-based Materials，South China Universityof Technology， Guangzhou，Guangdong Province，510640） （*E-mail： femnhong@scut. edu.cn）

Abstract：Toachevepreisepredictionadkowedgesuctuingofbreakfultsduringtepapeakinproc，tisdyfocsedn thecharacteristicsofwebbreakdataandtheimpactofpapermakingequipment.Bycombininglow-rankadaptation（oRA）andchan-ofthoughtprotinggngtustissarchsematicallmpadengtsurpameertf ing（PEFT）stategisogelgagedels，cuingLAeighecopodowandaatio（DRA）fuddaptei ingandamplifingieractivations（A3），andqantiedlowrankadaptation（QLoRA）BasedontheoptiallLRAfine-undlargln guagemodelsatG-iosructedbakiforatioowledgegahaptedtotitsrbesndai thepapermakingomainTheresultsindicatedtatheLRAalgorthchievedtheestcomprehensivepeformanceintheentitycogition and relation extraction tasks for web break information，reaching a recall rate of 100% and an F1 score of 92.31% . Its performance significantlyutperformedthenon-fine-tuned ChatGLM3-6Bandothermainstreamlargelanguage models （suchasiFLYTEKSparkMAXand Qwen2. 5-7B）.

Keywords： large language models; knowledge graph; web break fault; parameter-efficient fine-tuning

纸是人们日常生活中不可或缺的必需品，造纸工业也是我国重要的支柱产业之一。（剩余13970字）