面向社交媒体的讽刺检测研究综述  

Comprehensive review of sarcasm detection for social media

作　　者：余本功[1,2] 李晨越 Yu Bengong;Li Chenyue(School of Management,Hefei University of Technology,Hefei 230009,China;Key Laboratory of Process Optimization&Intelligent Decision-Making(Ministry of Education),Hefei University of Technology,Hefei 230009,China)

机构地区：[1]合肥工业大学管理学院,合肥230009 [2]合肥工业大学过程优化与智能决策教育部重点实验室,合肥230009

出　　处：《计算机应用研究》2025年第4期961-974,共14页Application Research of Computers

基　　金：国家自然科学基金资助项目(72071061)。

摘　　要：近年,网络社交媒体充斥着丰富的自由言论,而检测其中的讽刺语义作为一种特殊的情感分析引发了越来越多的关注。为了更好地研究面向社交媒体的讽刺检测,回顾了传统方法和基于深度学习方法的演进过程,并从文本讽刺检测和多模态讽刺检测两种角度来系统分析其发展趋势。首先,对讽刺检测的数据集进行了归纳;其次,探讨句子级、附加语境、知识和辅助任务的文本讽刺检测,详细阐述注意力机制、预训练模型、图神经网络和量子神经网络在多模态讽刺信息融合中起到的作用;接下来展望该领域的应用前景;最后总结当前研究现状与挑战,并结合近期大语言模型的发展提出讽刺检测可能的研究方向,为未来研究人员创新讽刺检测方法提供了参考和帮助。In recent years,online social media platforms have become saturated with a diverse range of free expression,lea-ding to heightened attention on the detection of sarcastic semantics as a specialized form of sentiment analysis.To enhance the study of sarcasm detection in social media contexts,this paper reviewed the evolution of both traditional methodologies and deep learning techniques,systematically analyzing their developmental trends from two perspectives:textual sarcasm detection and multimodal sarcasm detection.Firstly,this paper compiled and summarized the datasets utilized for sarcasm detection.Next,this paper investigated sentence-level detection,contextual augmentation,knowledge integration,and auxiliary tasks within textual sarcasm detection.This paper elaborated on the contributions of attention mechanisms,pre-trained models,graph neural networks,and quantum neural networks in the fusion of multimodal sarcastic information.Subsequently,this paper explored the potential applications within this domain.Finally,this paper encapsulated the current state of research,identified existing challenges,and proposed potential research directions for sarcasm detection,incorporating recent advancements in large language models to provide valuable guidance for future researchers aiming to innovate detection methodologies.

关 键 词：讽刺检测 社交媒体 多模态 研究综述 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]