计算机科学

计算机科学 ›› 2025, Vol. 52 ›› Issue (12): 224-230.doi: 10.11896/jsjkx.250600140

• 人工智能 • 上一篇    下一篇

高低资源下思维链增强的药物不良反应关系抽取

李浩, 杨雨濛, 赵博扬, 郑朴琪, 林鸿飞   

  1. 大连理工大学计算机科学与技术学院 辽宁 大连 116081
  • 收稿日期:2025-06-20 修回日期:2025-09-23 出版日期:2025-12-15 发布日期:2025-12-09
  • 通讯作者: 林鸿飞(hflin@dlut.edu.cn)
  • 作者简介:(dutlihao@163.com)
  • 基金资助:
    国家自然科学基金(62302076,62276043)

Adverse Drug Reaction Relationship Extraction Based on Chain of Thought Enhancement UnderHigh and Low Resources

LI Hao, YANG Yumeng, ZHAO Boyang, ZHENG Puqi, LIN Hongfei   

  1. School of Computer Science and Technology, Dalian University of Technology, Dalian, Liaoning 116081, China
  • Received:2025-06-20 Revised:2025-09-23 Published:2025-12-15 Online:2025-12-09
  • About author:LI Hao,born in 2001,postgraduate.His main research interests include natural language processing,large language models and information extraction.
    LIN Hongfei,born in 1962,Ph.D,professor,Ph.D supervisor.His main research interests include natural language processing,affective computing and computational humor,information retrieval and recommendation systems.
  • Supported by:
    This work was supported by the National Natural Science Foundation of China(62302076,62276043).

PDF (PC)

7

摘要： 药物不良反应(Adverse Drug Reactions,ADRs)是指正常剂量的药物用于预防、诊断、治疗疾病或调节生理机能时出现有害的、与用药目的无关的反应,这些反应是预期之外的,会严重影响患者的身体状态与健康情况。为了从社交媒体数据与医药文献数据中发现药物不良反应关系三元组,从而为患者和医疗系统提供警示作用,并为药物研发机构提供药物迭代的参考,结合生成式大语言模型分别针对有标注数据低资源和高资源的情况提出两个框架:思维链提示下GPT的药物不良反应关系抽取ADR-GPT框架,以及思维链增强下微调的药物不良反应关系抽取CADRE-LLM框架。分别在Twitter和Pubmed数据集中评估了两个框架的性能结果,其中CADRE-LLM相比于之前最先进的模型,F1值分别提高1.51个百分点和1.74个百分点;低资源下的ADR-GPT框架也取得了优秀的效果,在Pubmed数据集中超越了Qwen2.5模型的全监督微调。最后,通过消融实验证明了两个框架中各模块的有效性。

关键词: 关系抽取, 药物不良反应, 大语言模型, 思维链, 低资源

Abstract: ADRs refer to harmful and unintended responses that occur when drugs are administered at normal doses for prevention,diagnosis,treatment of diseases,or regulation of physiological functions.These reactions are unexpected and can significantly impact patients’ health and physical condition.To identify ADR-related relational triples from social media and biomedical literature data—providing early warnings for patients and healthcare systems,as well as references for pharmaceutical research and drug development－this paper proposes two frameworks based on generative large language models for both low-resource and high-resource annotated data scenarios:the ADR-GPT framework,which extracts ADR relations using Chain-of-Thought promp-ting,and the CADRE-LLM framework,which leverages Chain-of-Thought-enhanced fine-tuning.It evaluates the performance of these two frameworks on the Twitter and PubMed datasets.CADRE-LLM achieves F1 score improvements of 1.51 percentage points and 1.74 percentage points over previous state-of-the-art models on the respective datasets.The low-resource ADR-GPT framework also demonstrates strong performance,outperforming fully supervised fine-tuned Qwen2.5 on the PubMed dataset.Ablation studies further validate the effectiveness of each module within the two proposed frameworks.

Key words: Relation extraction, Adverse drug reactions, Large language models, Chain of Thought(COT), Low resource

中图分类号: 

  • TP311.13
[1]NATIONAL CENTER FOR ADR MONITORING.Annual report of national adverse drug reaction monitoring(2023) [J].Chinese Journal of Viral Diseases,2023,13(4):245-251.
[2]LI H,QIU Y Z,LIN H F.Adverse drug reaction detection with multi-feature enhancement for social media [C]//Proceedings of the 23rd Chinese National Conference on Computational Linguistics.2024:515-525.
[3]ZHANG T,LIN H,REN Y,et al.Adversarial transfer network with bilinear attention for the detection of adverse drug reactions from social media [J].Applied Soft Computing,2021,106:107358.
[4]ZHANG T,LIN H,REN Y,et al.Adverse drug reaction detection via a multihop self-attention mechanism [J].BMC Bioinformatics,2019,20(1):1-11.
[5]ZHENG S,WANG F,BAO H,et al.Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme[C]//Procee-dings of the 55th Annual Meeting of the Association for Computational Linguistics.2017:1227-1236.
[6]REN F,ZHANG L,ZHAO X,et al.A simple but effective bidirectional framework for relational triple extraction [C]//Proceedings of the 15th ACM Int Conf on Web Search and Data Mining.New York:ACM,2022:824-832.
[7]WANG Y,YU B,ZHANG Y,et al.TPLinker:Single-stage joint extraction of entities and relations through token pair linking [C]//Proceedings of the 28th Int Conf on Computational Linguistics.Stroudsburg,PA:ACL,2020:1572-1582.
[8]SHANG Y,HUANG H,MAO X.Onerel:Joint entity and relation extraction with one module in one step [C]//Proceedings of the 36th AAAI Conf on Artificial Intelligence.Palo Alto,CA:AAAI,2022:11285-11293.
[9]YE H,ZHANG N,DENG S,et al.Contrastive triple extraction with generative transformer [C]//Proceedings of the 35th AAAI Conf on Artificial Intelligence.Palo Alto,CA:AAAI,2021:14257-14265.
[10]DEVLIN J,CHANG M,LEE K,et al.BERT:Pre-training ofdeep bidirectional transformers for language understanding [C]//Proceedings of the 2019 Conf of the North American Chapter of the ACL:Human Language Technologies.Stroudsburg,PA:ACL,2019:4171-4186.
[11]WANG X,ZHOU W,ZU C,et al.InstructUIE:Multi-task instruction tuning for unified information extraction [J].arXiv:2304.08085,2023.
[12]BAI J,BAI S,CHU Y,et al.Qwen technical report [J].arXiv:2309.16609,2023.
[13]LUO L,NING J,ZHAO Y,et al.Taiyi:A bilingual fine-tuned large language model for diverse biomedical tasks [J].Journal of the American Medical Informatics Association,2024,31(9):1865-1874.
[14]WADHWA S,AMIR S,WALLACE B.Revisiting relation ex-traction in the era of large language models [C]//Proceedings of the 61st Annual Meeting of the ACL.Stroudsburg,PA:ACL,2023:15566-15577.
[15]KOJIMA T,GU S,REID M,et al.Large language models are zero-shot reasoners [J].Advances in Neural Information Processing Systems,2022,35:22199-22213.
[16]SHUM K,DIAO S,ZHANG T.Automatic prompt augmenta-tion and selection with chain-of-thought from labeled data [C]//Findings of EMNLP 2023.Stroudsburg,PA:ACL,2023:12113-12139.
[17]REIMERS N,GUREVYCH I.Sentence-BERT:Sentence em-beddings using Siamese BERT-networks [C]//Proceedings of EMNLP-IJCNLP 2019.Stroudsburg,PA:ACL,2019:3982-3992.
[18]DETTMERS T,PAGNONI A,HOLTZMAN A,et al.QLoRA:Efficient finetuning of quantized LLMs [J].Advances in Neural Information Processing Systems,2023,36:10088-10115.
[19]ALVARO N,MIYAO Y,COLLIER N.TwiMed:Twitter andPubMed comparable corpus of drugs,diseases,symptoms,and their relations [J].JMIR Public Health and Surveillance,2017,3(2):e6396.
[20]WEI Z,SU J,WANG Y,et al.A Novel Cascade Binary Tagging Framework for Relational Triple Extraction[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.2020:1476-1488.
[21]ACHIAM J,ADLER S,AGARWAL S,et al.GPT-4 technicalreport [J].arXiv:2303.08774,2023.
[22]GRATTAFIORI A,DUBEY A,JAUHRI A,et al.The LLaMA 3 herd of models [J].arXiv:2407.21783,2024.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发