ABSTRACT This study presents a human-centric technique for social video expansion based on semantic processing and graph analysis. The objective is to increase metadata of an online video and to explore related information, thereby facilitating user browsing activities. To analyze the semantic meaning of a video, shots and scenes are firstly extracted from the video on the server side. Subsequently, this study uses annotations along with ConceptNet to establish the underlying framework. Detailed metadata, including visual objects and audio events among the predefined categories, are indexed by using the proposed method. Furthermore, relevant online media associated with each category are also analyzed to enrich the existing content. With the above-mentioned information, users can easily browse and search the content according to the link analysis and its complementary knowledge. Experiments on a video dataset are conducted for evaluation. The results show that our system can achieve satisfactory performance, thereby demonstrating the feasibility of the proposed idea.

中文翻译：

---

基于语义处理和图形分析的社交多媒体网站概念索引与扩展

摘要 本研究提出了一种基于语义处理和图形分析的以人为中心的社交视频扩展技术。目的是增加在线视频的元数据，探索相关信息，从而方便用户浏览活动。为了分析视频的语义，首先在服务器端从视频中提取镜头和场景。随后，本研究使用注释和 ConceptNet 来建立底层框架。详细的元数据，包括预定义类别中的视觉对象和音频事件，使用所提出的方法进行索引。此外，还分析了与每个类别相关联的相关在线媒体以丰富现有内容。有了上述信息，用户可以根据链接分析及其补充知识轻松浏览和搜索内容。对视频数据集进行实验以进行评估。结果表明，我们的系统可以达到令人满意的性能，从而证明了所提出想法的可行性。