ChatGPT在网络健康信息鉴别中的应用探讨  

Discussion on the Application of ChatGPT in Online Health Information Identification

作　　者：李雅静 孙佳佳 LI Yajing;SUN Jiajia

机构地区：[1]武汉大学信息管理学院

出　　处：《图书馆论坛》2023年第6期85-92,共8页Library Tribune

摘　　要：文章探索ChatGPT在网络健康信息鉴别中的应用,探讨其优缺点;使用ChatGPT对网络健康信息进行真伪鉴别,评估其性能,并与专家鉴别结果进行对比分析。实验结果显示:ChatGPT的F1值达0.80,具有较好的鉴别效果;ChatGPT生成的解释文本与专家的解释文本存在较高的语义相似度,但文本长度较短,存在科学依据不够详细、逻辑错误等问题。结果表明:ChatGPT可以辅助网络健康信息的鉴别任务,提高识别效率,但仍需要人工干预,避免出现误导用户现象。This paper explores the application of ChatGPT in the identification of online health information and discusses its advantages and disadvantages.It uses ChatGPT to authenticate online health information,evaluates its performance,and compares the performance with the results obtained by experts.The experimental results suggest that ChatGPT has an F1 score of 0.80,which shows a good identification effect;the explanation text generated by ChatGPT has a high semantic similarity with that of the experts,but the text length is short and there are problems such as lack of detailed scientific basis and logical errors.Therefore,ChatGPT can assist with the task of identifying online health information to improve the identification efficiency,but human intervention is still needed to avoid misleading users.

关 键 词：ChatGPT 网络健康信息 鉴别方法 人工智能 

分 类 号：G250.7[文化科学—图书馆学] R197.1[医药卫生—卫生事业管理] G203[医药卫生—公共卫生与预防医学] R-05