Abstract: Abstract: With the widespread adoption of the internet and the advancement of digital technology, the volume 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, but these characteristics also bring significant challenges to data processing and analysis. Aiming at the analysis and application challenges posed by user-generated content in terms of large scale, modal diversity, and unstructuredness, a review of methods and applications for analyzing user-generated content was synthesized, in which the mainstream analysis methods for text and images were systematically sorted out, including text summarization, text sentiment analysis, image recognition, image sentiment analysis, and multimodal analysis methods for multiple types of data. Meanwhile, key challenges and coping strategies were discussed in light of the heterogeneity and dynamics faced by multimodal fusion, and potential future trends in enhancing privacy security and improving adaptability were envisioned. The integration of the existing results can provide ideas for the innovation and application of UGC multimodal analysis technology, as well as indicate the direction for further research.

Key words: Keywords: User-Generated Content (UGC), text analysis, image analysis, text summarization, image recognition, sentiment analysis, multimodal analysis

摘要： 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。得益于其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，但这些特点也为数据处理和分析带来了显著的挑战。针对用户生成内容在规模庞大、模态多元和非结构化等方面带来的分析与应用难题，综述了用户生成内容分析方法与应用，总结了文本、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。与此同时，结合多模态融合面临的异质性与动态性，讨论了关键挑战与应对策略，并展望了未来在强化隐私安全和提升适应性方面的潜在趋势。通过对现有成果的整合，可为UGC多模态分析技术的创新与应用落地提供思路，也为进一步研究指明了方向。

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