基于词抽象化的虚假新闻检测方法  

Fake news detection based on word abstraction

作　　者：李润东 汪旦丁 王政嘉 刘晟 孙一凡 吴娅 曹娟[1,2] 盛强 程皓楠 LI Rundong;WANG Danding;WANG Zhengjia;LIU Sheng;SUN Yifan;WU Ya;CAO Juan;SHENG Qiang;CHENG Haonan(Institute of Computing Technology,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China;State Key Laboratory of Media Convergence and Communication,Communication University of China,Beijing 100024,China)

机构地区：[1]中国科学院计算技术研究所,北京100190 [2]中国科学院大学,北京100049 [3]中国传媒大学媒体融合与传播国家重点实验室,北京100024

出　　处：《信息传播研究》2025年第1期9-19,共11页Information and Communication Research

基　　金：“媒体融合与传播国家重点实验室(中国传媒大学)”开放课题(SKLMCC2022KF001)。

摘　　要：在社交媒体新闻时代,假新闻泛滥问题日益严重,对假新闻的自动化检测对维护社会稳定意义重大。虽然各种假新闻变化莫测,在字面描述上往往差别很大,但其更深层面的内涵,即其造假意图和目的却可能十分相似。当前的自动化假新闻检测方法忽略了由于深层内涵相同而表现出的假新闻的规律性。基于词抽象化的虚假新闻检测方法通过挖掘真假新闻在深层内涵上的差别,能够有效从字面表述上和内在含义上对虚假新闻进行识别。实验结果表明,本文提出方法的假新闻检测性能明显优于已有方法。In the era of social media news,the proliferation of fake news has become an increasingly pressing issue.The automated detection of fake news is of paramount importance for maintaining societal stability.Although various manifestations of fake news can differ significantly in their literal descriptions,their deeper connotations—namely,the intent and purpose behind their fabrication—may bear similarities.Current automated methods for detecting fake news often overlook the patterns that emerge from these shared underlying meanings.Our fake news detection method based on word abstraction effectively identifies fake news by uncovering the distinctions in deeper connotations between real and fake news.Experimental results demonstrate that our approach significantly outperforms existing detection methods.

关 键 词：假新闻检测 抽象语义 义原 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]