情报杂志2025，Vol.44Issue(5)：130-138,9.DOI:10.3969/j.issn.1002-1965.2025.05.016
以模治模:基于生成式人工智能的失真健康信息识别
Using Models to Treat Models:Distorted Health Information Recognition Based on Generative Artificial Intelligence
摘要
Abstract
[Research purpose]Based on the large language model,a hierarchical and step-by-step fine-grained distorted health infor-mation recognition method is proposed to alleviate the harm caused by distorted health information in the context of generative artificial in-telligence.[Research method]Key information elements in the health text are extracted and integrated to form a concise and clear claim text;Then,the claim text is decomposed into multiple verifiable sub claims,and each sub claim is verified using the built-in knowledge and search engine of the big model.Based on the verification results of the sub claims,the distortion degree of the original health text is classified;Finally,the effectiveness of the method is verified through experimental evaluation.[Research result/conclusion]The experi-mental results show that the method proposed in this research is significantly superior to CNN,RNN,ChatGPT-Zero-Shot and other methods in terms of accuracy,precision,and recall.At the same time,it can effectively alleviate the problems of key point omission and factual illusion caused by directly identifying distorted information using large models.This research proposes a new path for identifying distorted health information in the AIGC era,which is of great significance for the subsequent governance of distorted information.关键词
失真健康信息/信息治理/信息检测/生成式人工智能/大语言模型Key words
distorted health information/information governance/information detection/AIGC/large language model分类
预防医学引用本文复制引用
张君冬,刘江峰,邓景鹏,黄奇,刘艳华,李娜..以模治模:基于生成式人工智能的失真健康信息识别[J].情报杂志,2025,44(5):130-138,9.基金项目
江苏省研究生科研与实践创新计划项目"图模驱动的在线医疗健康智慧问答服务研究"(编号:KYCX24_0107) （编号:KYCX24_0107）
江苏高校哲学社会科学研究重大项目"中医古籍文献预训练模型构建及其应用研究"(编号:2023SJZD084)研究成果. （编号:2023SJZD084）
