Emotion-semantic interaction network for fake news detection: Perspectives on question and non-question comment semantics

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management Pub Date : 2025-09-09 DOI:10.1016/j.ipm.2025.104391

Zhenhua Tan , Tao Zhang

{"title":"Emotion-semantic interaction network for fake news detection: Perspectives on question and non-question comment semantics","authors":"Zhenhua Tan , Tao Zhang","doi":"10.1016/j.ipm.2025.104391","DOIUrl":null,"url":null,"abstract":"<div><div>Current fake news detectors often overlook the association between news content emotion and comment semantics, especially questioning vs. non-questioning language. We observe that fake news evokes distinct comment patterns: non-sad emotions (e.g., anger, surprise) in fake content drive more questioning semantics (e.g., “Where is the video to support this?”) and suppress non-questioning replies, while sadness shows the opposite trend. Ignoring this emotion-semantic interaction limits detection accuracy. To address this limitation, we propose an Emotion-Semantic Interaction Network (ESIN), which learns a latent association between content emotions and comment semantics from questioning and non-questioning perspectives. Specifically, ESIN incorporates the interaction of content emotion with original comment semantics, as well as the distribution of extracted comment semantic categories (i.e., question and non-question), achieved through our proposed mechanisms called Codebook Initialization and Semantic Quantification, based on cross-attention. The ESIN model is comprehensively evaluated on two widely used fake news datasets, namely RumourEval and WeiBo. The ESIN achieves competitive performance, outperforming baseline models by a significant margin of 5.27% and 4.08% in weighted F1-score (F1.) and accuracy (Acc.) on the RumourEval dataset, and by 2.43% and 2.38% in F1. and Acc. on the WeiBo dataset. The promising result verifies the effectiveness of our ESIN model. Theoretically, this study fills a critical gap by highlighting how content emotions shape comment semantics (questioning vs. non-questioning) as a veracity cue, advancing understanding of user engagement patterns in fake news spread. Practically, ESIN offers actionable strategies for platforms to flag high-risk content and equips policymakers with tools for user training and evidence-based regulations, enhancing misinformation mitigation and public trust in digital information ecosystems.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 2","pages":"Article 104391"},"PeriodicalIF":6.9000,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Information Processing & Management","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0306457325003322","RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}

引用次数: 0

Abstract

Current fake news detectors often overlook the association between news content emotion and comment semantics, especially questioning vs. non-questioning language. We observe that fake news evokes distinct comment patterns: non-sad emotions (e.g., anger, surprise) in fake content drive more questioning semantics (e.g., “Where is the video to support this?”) and suppress non-questioning replies, while sadness shows the opposite trend. Ignoring this emotion-semantic interaction limits detection accuracy. To address this limitation, we propose an Emotion-Semantic Interaction Network (ESIN), which learns a latent association between content emotions and comment semantics from questioning and non-questioning perspectives. Specifically, ESIN incorporates the interaction of content emotion with original comment semantics, as well as the distribution of extracted comment semantic categories (i.e., question and non-question), achieved through our proposed mechanisms called Codebook Initialization and Semantic Quantification, based on cross-attention. The ESIN model is comprehensively evaluated on two widely used fake news datasets, namely RumourEval and WeiBo. The ESIN achieves competitive performance, outperforming baseline models by a significant margin of 5.27% and 4.08% in weighted F1-score (F1.) and accuracy (Acc.) on the RumourEval dataset, and by 2.43% and 2.38% in F1. and Acc. on the WeiBo dataset. The promising result verifies the effectiveness of our ESIN model. Theoretically, this study fills a critical gap by highlighting how content emotions shape comment semantics (questioning vs. non-questioning) as a veracity cue, advancing understanding of user engagement patterns in fake news spread. Practically, ESIN offers actionable strategies for platforms to flag high-risk content and equips policymakers with tools for user training and evidence-based regulations, enhancing misinformation mitigation and public trust in digital information ecosystems.

查看原文本刊更多论文

基于情感语义交互网络的假新闻检测：问题和非问题评论语义的视角

目前的假新闻检测往往忽略了新闻内容情感与评论语义之间的联系，特别是质疑与非质疑语言之间的联系。我们观察到，假新闻引发了不同的评论模式：假内容中的非悲伤情绪（例如，愤怒、惊讶）驱动了更多的质疑语义（例如，“支持这一点的视频在哪里？”）并抑制非质疑性的回答，而悲伤则表现出相反的趋势。忽略这种情感-语义交互限制了检测的准确性。为了解决这一限制，我们提出了一个情感-语义交互网络（ESIN），它从提问和非提问的角度学习内容情感和评论语义之间的潜在关联。具体来说，ESIN结合了内容情感与原始评论语义的交互，以及提取的评论语义类别（即问题和非问题）的分布，这是通过我们提出的基于交叉关注的代码本初始化和语义量化机制实现的。ESIN模型在RumourEval和微博两个广泛使用的假新闻数据集上进行了综合评估。ESIN达到了具有竞争力的性能，在RumourEval数据集上，其加权F1得分（F1.）和准确率（Acc.）分别比基准模型高出5.27%和4.08%，在F1中分别高出2.43%和2.38%。和Acc。在微博数据集上良好的结果验证了ESIN模型的有效性。从理论上讲，本研究填补了一个关键的空白，强调了内容情绪如何塑造评论语义（质疑与非质疑）作为真实性线索，促进了对假新闻传播中用户参与模式的理解。实际上，ESIN为平台提供了可操作的战略，以标记高风险内容，并为决策者提供了用户培训和基于证据的法规工具，从而加强了减少错误信息和公众对数字信息生态系统的信任。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Information Processing & Management 工程技术-计算机：信息系统

CiteScore

17.00

自引率

11.60%

发文量

276

审稿时长

39 days

期刊介绍： Information Processing and Management is dedicated to publishing cutting-edge original research at the convergence of computing and information science. Our scope encompasses theory, methods, and applications across various domains, including advertising, business, health, information science, information technology marketing, and social computing. We aim to cater to the interests of both primary researchers and practitioners by offering an effective platform for the timely dissemination of advanced and topical issues in this interdisciplinary field. The journal places particular emphasis on original research articles, research survey articles, research method articles, and articles addressing critical applications of research. Join us in advancing knowledge and innovation at the intersection of computing and information science.