摘要
随着互联网的普及和数字技术的发展,用户生成内容(UGC)数量呈现爆炸式增长。得益于它天然的真实性和多样性,UGC成为分析用户情感、需求及行为的重要数据来源,然而这些特点也为数据处理和分析带来了显著挑战。针对UGC在规模庞大、模态多元和非结构化等方面带来的分析与应用难题,综述UGC分析方法与应用。首先,系统性地总结文本和图像的主流分析方法,包括文本摘要、文本情感分析、图像识别、图像情感分析以及针对多类型数据的多模态分析方法;同时,结合多模态融合面临的异质性与动态性,讨论该领域的关键挑战与应对策略,并展望该领域未来在强化隐私安全和提升适应性方面的潜在趋势;最后,通过对现有成果的整合,为UGC多模态分析技术的创新与应用落地提供思路,并为进一步研究指明方向。
With the widespread of Internet and the advancement of digital technology,the size of User-Generated Content(UGC)has grown explosively.Thanks to its natural authenticity and diversity,UGC has become an important source of data for analyzing users’emotions,needs,and behaviors.However,these characteristics also bring significant challenges to data processing and analysis.Aiming at the analysis and application challenges posed by UGC in terms of large scale,modal diversity,and unstructuredness,methods and applications for analyzing UGC were reviewed.Firstly,the mainstream analysis methods for texts and images were sorted out systematically,including text summarization,text sentiment analysis,image recognition,image sentiment analysis,and multimodal analysis methods for multiple types of data.Meanwhile,by combining the heterogeneity and dynamics faced by multimodal fusion,key challenges and coping strategies in this field were discussed,and potential future trends of this field in enhancing privacy security and improving adaptability were prospected.Finally,integration of the existing results provides ideas for the innovation and application of UGC multimodal analysis technology,as well as indicates the directions for further research.
作者
刘瑜
周晓航
张宁
LIU Yu;ZHOU Xiaohang;ZHANG Ning(Business College,Qingdao University,Qingdao Shandong 266071,China;School of Management,Qingdao City University,Qingdao Shandong 266106,China)
出处
《计算机应用》
北大核心
2025年第S2期14-20,共7页
journal of Computer Applications
基金
山东省自然科学基金资助项目(ZR2022MG022)
教育部人文社会科学基金青年项目(24YJC790248)。
关键词
用户生成内容
文本分析
图片分析
文本摘要
图像识别
情感分析
多模态分析
User-Generated Content(UGC)
text analysis
image analysis
text summarization
image recognition
sentiment analysis
multimodal analysis
