智能化新闻核查技术:算法、逻辑与局限

Automatic Fake News Detection:Algorithm,Logic and Limitation

近年来,为解决"后真相时代"假新闻泛滥的问题,计算机界对智能化的新闻核查技术进行了大量研究。然而,由于文理的藩篱,新闻界却缺乏对智能化新闻核查技术的深入探讨,更未能前瞻性地思考智能化新闻核查技术介入媒介生态的后果。本文希望打破新闻传播学与计算机学科的隔阂,围绕智能化新闻核查主流算法——"基于内容模型的算法"和"基于社会情境的算法",从技术路径出发,分析其技术原理、运行逻辑及与新闻界接洽所产生的价值风险。并在此基础上,跳出具体技术模型,从更为宏观的层面上反思纯粹的技术手段解决作为复杂的社会现象的虚假新闻的有效性,以及虚假新闻的复杂性及其新的表现形式对智能算法核查新闻的有效性带来的挑战。

In recent years,computer scientists have devoted a lot of efforts to develop fake news detection system for the purpose of dealing with the widespread fake news in the post-truth era.However,due to the cognitive barriers between arts and sciences,scholars of journalism haven’t paid enough attention to the emerging technology or the consequences it brings to media ecology.This paper intends to bridge the gap between journalism and computer science,and take a technical perspective to analyze the mechanism,logic and value of content-model-based algorithm and socialcontext-based algorithm,which are two mainstream algorithms of fake news detection system.Then we jump out of details of specific technical model to a macro level to review the effectiveness of preventing fake news dissemination,which is complex social phenomenon,with computer technology.Especially,we analyze the limitation of automatic news detection system when considering the complexity and new form of fake news.