|本期目录/Table of Contents|

[1]王明达,赵宝熙,吴志生,等.基于大语言模型的燃气事故调查报告实体识别*[J].中国安全生产科学技术,2025,21(2):139-145.[doi:10.11731/j.issn.1673-193x.2025.02.018]
　WANG Mingda,ZHAO Baoxi,WU Zhisheng,et al.Entity recognition of gas accident investigation reports based on large language model[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2025,21(2):139-145.[doi:10.11731/j.issn.1673-193x.2025.02.018]
点击复制

基于大语言模型的燃气事故调查报告实体识别*
分享到：

《中国安全生产科学技术》[ISSN:1673-193X/CN:11-5335/TB]

卷:
21
期数:
2025年2期
页码:
139-145
栏目:
职业安全卫生管理与技术
出版日期:
2025-02-28

文章信息/Info

Title:
Entity recognition of gas accident investigation reports based on large language model
文章编号:
1673-193X(2025)-02-0139-07
作者:
王明达; 赵宝熙; 吴志生; 冷高强
(中国石油大学（华东） 机电工程学院，山东 青岛 266580)
Author(s):
WANG Mingda; ZHAO Baoxi; WU Zhisheng; LENG Gaoqiang
(College of Mechanical and Electrical Engineering,China University of Petroleum,Qingdao Shandong 266580，China)
关键词:
燃气事故调查报告; 命名实体识别; 大语言模型; 指令微调; 数据增强
Keywords:
gas accident investigation report;  named entity recognition;  large language model;  instruction fine-tuning;  data enhancement
分类号:
X937
DOI:
10.11731/j.issn.1673-193x.2025.02.018
文献标志码:
A
摘要:
为解决样本稀少对大语言模型(LLM)在燃气事故调查报告中的实体识别精度影响显著的问题，提出1种基于两阶段训练的大语言模型实体识别方法。在数据集构建阶段,LLM根据对话式指令微调模板自动生成燃气事故调查报告数据集，采用简单数据增强(EDA)技术扩充人工标注的关键样本；在模型微调训练阶段，采用低秩适配微调技术对Phi3-mini-128k模型进行微调训练，第1阶段微调训练利用LLM自动标注数据集，在训练基础上利用增强数据集对模型进行第2阶段微调训练。研究结果表明：经过第1阶段微调训练后，Phi3-mini-rq模型的实体识别综合评价指标提高11.01百分点；当EDA增强数据占总数据的50%时，模型第2阶段微调效果最佳，综合评价指标值进一步提升2.49百分点。研究结果可为燃气领域的事故报告自动化处理提供有效技术支持。
Abstract:
In order to solve the problem of the significant impact of sample scarcity on the entity recognition accuracy of large language model (LLM) in gas accident investigation reports,a LLM entity recognition method based on two-stage training was proposed.In the dataset construction stage,LLM automatically generates the dataset of gas accident investigation reports according to the conversational instruction fine-tuning template,and adopts simple data augmentation (EDA) technique to expand manually labeled key paper and then manually annotate it.In the model fine-tuning training stage,the low-rank adaptation fine-tuning technique was adopted to conduct the fine-tuning training on the Phi3-mini-128k model.The first-stage fine-tuning training utilized LLM to automatically annotate the dataset,and the second-stage fine-tuning training wad carried out on the model by using the enhanced dataset on the basis of training.The results show that after the first-stage fine-tuning training,the comprehensive evaluation index of entity recognition of Phi3-mini-rq model is improved by 11.01%.When the EDA enhanced data accounts for 50% of the total data,the second-stage fine-tuning effect of the model is the best,and the value of comprehensive evaluation index is further improved by 2.49%.The research results can provide effective technical support for the automated processing of accident reports in the gas field.

参考文献/References:

［1］ZHANG Y,ZHOU J.A trainable method for extracting Chinese entity names and their relations［C］//Second Chinese Language Processing Workshop.2000:66-72.
［2］俞鸿魁,张华平,刘群,等.基于层叠隐马尔可夫模型的中文命名实体识别［J］.通信学报,2006，27(2):87-94. YU Hongkui,ZHANG Huaping,LIU Qun,et al.Chinese named entity identification using cascaded hidden Markov model ［J］.Journal of Communications,2006，27(2):87-94.
［3］王红,祝寒,林海舟.航空安全事故因果关系抽取方法的研究［J］.计算机工程与应用,2020,56(11)：265-270. WANG Hong,ZHU Han,LIN Haizhou.Research on causality extraction of civil aviation accident［J］.Computer Engineering and Applications,2020,56(11)：265-270.
［4］罗凌,杨志豪,宋雅文，等.基于笔画ELMo和多任务学习的中文电子病历命名实体识别研究［J］.计算机学报,2020,43(10)：1943-1957. LUO Ling,YANG Zhihao,SONG Yanwen,et al.Chinese clinical named entity recognition based on stroke ELMo and multi-task learning ［J］.Chinese Journal of Computers,2020,43(10)：1943-1957.
［5］QIN Y,SHEN G,ZHAO W,et al.A network security entity recognition method based on feature template and CNN-BiLSTM-CRF［J］.Frontiers of Information Technology & Electronic Engineering,2019,20(6)：872-884.
［6］郑立瑞,肖晓霞,邹北骥,等.基于BERT的电子病历命名实体识别［J］.计算机与现代化,2024,(1):87-91. ZHENG Lirui,XIAO Xiaoxia,ZOU Beiji,et al.Named entity recognition in electronic medical record based on BERT ［J］.Computer and Modernization,2024,(1):87-91.
［7］余丹丹,黄洁,党同心,等.基于ALBERT的中文简历命名实体识别［J］.计算机工程与设计,2024,45(1):261-267. YU Dandan,HUANG Jie,DANG Tongxin,et al.Recognition of named entity in Chinese resume based on ALBERT ［J］.Computer Engineering and Design,2024,45(1):261-267.
［8］王明达,张榜,吴志生，等.基于强化学习的城镇燃气事故信息抽取方法［J］.中国安全生产科学技术,2023,19(3)：39-45. WANG Mingda,ZHANG Bang,WU Zhisheng,et al.Information extraction method of urban gas accidents based on reinforcement learning［J］.Journal of Safety Science and Technology,2023,19(3)：39-45.
［9］杨柳.基于文本数据的轨道交通事故致因分析及风险研究［D］.北京交通大学,2021.
［10］LI X,SHI T,LI P,et al.BiLSTM-CRF model for named entity recognition in railway accident and fault analysis report［C］//Proceedings of the Asia-Pacific Conference on Intelligent Medical 2018 & International Conference on Transportation and Traffic Engineering 2018.2018：1-5.
［11］QIN Y,ZENG Y.Research of clinical named entity recognition based on Bi-LSTM-CRF［J］.Journal of Shanghai Jiao Tong University (Science),2018,23(3)：392-397.
［12］关斯琪,董婷婷,万子敬，等.基于BERT-CRF模型的火灾事故案例实体识别研究［J］.消防科学与技术,2023,42(11)：1529-1534. GUAN Siqi,DONG Tingting,WAN Zijing,et al.Fire accident case named entity recognition based on BERT-CRF model［J］.Fire Science and Technology,2023,42(11)：1529-1534.
［13］WANG S H,SUN X F,LI X Y,et al.GPT-NER:named entity recognition via large language models［EB/OL］.(2023-10-07)［2025-01-17］.https://arxiv.org/abs/2304.10428.
［14］MEONI S,DE LA CLERGERIE E,RYFFEL T.Large language models as instructors：a study on multilingual clinical entity extraction［C］//The 22nd Workshop on Biomedical Natural Language Processing and BioNLP Shared Tasks.2023：178-190.
［15］ASKOK D,LIPTON Z C.PromptNER： prompting for fewshot named entity recognition［EB/OL］.(2023-06-20)［2025-01-17］.https://arxiv.org/abs/2305.15444.
［16］WEI J,ZOU K.EDA:easy data augmentation techniques for boosting performance on text classification tasks［EB/OL］.(2019-08-25)［2025-01-17］.https://arxiv.org/abs/1901.11196.
［17］VASWANI A,SHAZEER N,PARMAR N,et al.Attention is all you need［C］//Proceedings of the 31st International Conference on Neural Information Processing Systems.Long Beach:ACM,2017:6000-6010.
［18］ABDIN M,ANEJA J,AWADALLA H,et al.Phi-3 technical report:a highly capable language model locally on your phone［EB/OL］.(2024-08-30)［2025-01-17］.https://arxiv.org/abs/2404.14219.
［19］OOUYANG L,WU J,JIANG X,et al.Training language models to follow instructions with human feedback［J］.Advances In Neural Information Processing Systems,2022,35： 27730-27744.
［20］HOULSBY N,GIURGIU A,JASTRZEBSKI S,et al.Parameter-efficient transfer learning for NLP［C］//International conference on machine learning.PMLR,2019:2790-2799.
［21］LI X L,LIANG P.Prefix-tuning:optimizing continuous prompts for generation［EB/OL］.(2021-01-01)［2025-01-17］.https://arxiv.org/abs/2101.00190.
［22］HU E J,SHEN Y,WALLIS P,et al.Lora:low-rank adaptation of large language models［EB/OL］.(2021-10-16)［2025-01-17］.https://arxiv.org/abs/2106.09685.
［23］LIU X,JI K,FU Y,et al.P-tuning v2:prompt tuning can be comparable to fine-tuning universally across scales and tasks［EB/OL］.(2022-03-15)［2025-01-17］.https://arxiv.org/abs/2110.07602.
［24］广州市南沙区人民政府.南沙区“8·3”燃气管道泄漏起火一般事故调查报告［EB/OL］.(2022-04-21)［2025-01-17］.https://www.gzns.gov.cn/zwgk/zdlyxxgk/aqsc/sgdcbgxx/content/post_8179818.html.

相似文献/References:

[1]牛飞,钟少波,刘楠,等.一种改进的灾害新闻3要素提取方法研究*[J].中国安全生产科学技术,2023,19(2):13.[doi:10.11731/j.issn.1673-193x.2023.02.002]
　NIU Fei,ZHONG Shaobo,LIU Nan,et al.Research on an improved extraction method for three elements of disaster news[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2023,19(2):13.[doi:10.11731/j.issn.1673-193x.2023.02.002]
[2]王明达,张榜,吴志生,等.基于强化学习的城镇燃气事故信息抽取方法[J].中国安全生产科学技术,2023,19(3):39.[doi:10.11731/j.issn.1673-193x.2023.03.006]
　WANG Mingda,ZHANG Bang,WU Zhisheng,et al.Information extraction method of urban gas accidents based on reinforcement learning[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2023,19(2):39.[doi:10.11731/j.issn.1673-193x.2023.03.006]
[3]成全,张双宝.基于深度学习的特征增强式安全事故文本实体识别模型研究*[J].中国安全生产科学技术,2024,20(6):58.[doi:10.11731/j.issn.1673-193x.2024.06.008]
　CHENG Quan,ZHANG Shuangbao.Research on feature-enhanced model for entity recognition of safety accident text based on deep learning[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2024,20(2):58.[doi:10.11731/j.issn.1673-193x.2024.06.008]

备注/Memo

备注/Memo:
收稿日期： 2024-09-21
* 基金项目： 国家自然科学基金项目(52075549)
作者简介： 王明达，博士，讲师，主要研究方向为油气安全工程与安全工程信息化。
更新日期/Last Update: 2025-03-04