项目名称: 社会化多媒体中的事件分析研究
项目编号: No.61502139
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 计算机科学学科
项目作者: 刘学亮
作者单位: 合肥工业大学
项目金额: 21万元
中文摘要: 随着社会化多媒体的快速发展,大量媒体数据被共享到互联网上。挖掘社会化多媒体信息之间的潜在关联,高效地进行社会媒体组织和管理成为一个具有重要现实意义的科学问题。近年来,事件作为一种非常有效的信息管理方式,得到了学术界越来越多的关注。目前互联网上面存在不同模态数据的共享平台,而记录事件的相关文档被不同的用户,以不同的格式共享到这些平台上。如何综合利用这些多源异质数据,进行事件的检测和摘要,为本课题的核心研究目标。主要研究内容包括:对社交共享平台的社会化多媒体数据进行特征学习和建模;从多源异质数据流中进行事件的检测;以及对检测到的事件进行多模态的摘要表示等。本课题提出以事件为中心的社会化多媒体管理和检索的新思路。本课题的实施将对社会化多媒体内容分析研究起到推动作用,并为新一代网络社会化多媒体服务与管理提供核心算法。
中文关键词: 社会化多媒体;特征学习;事件检测;事件呈现
英文摘要: With the rapid development of social multimedia, a lot of multimedia content is shared online. With the aim of improving the multimedia management and retrieval system, how to efficiently mine the relations among all kinds of social multimedia still remains an open and challenging problem. In recent years, event, as one of the most effective way to organize media data, has drawn more and more attention in research community. Actually, there are different modality data in social media share sites, and the event-related documents are captured in different format and then shared in different social web service. The proposal aims at detecting and illustrating social events from the multi-source heterogeneous data. Our roadmap is generally described as follows. We first study how to formulate the social media by ubiquitous model, and then investigate detecting events from multi-source data streams; finally exploit summarizing the detected events by multimodal data. This project presents a new roadmap for the social multimedia management and retrieval in the perspective of event, which paves a way to new methods and technologies in next-generation social media management services.
英文关键词: Social Multimedia;Feature Learning;Event Detection;Event Illustration