网络媒体大数据聚合与呈现关键技术研究

With the rapid development of internet and multimedia technology, the volume of data on the web is experiencing an explosive increase. The popular concept "big data" emerges in this context. However, we have to face the dilema of the intelligent management and content understanding of the big internet multimedia data due to its characteristics of "big" volume and heterogeneous resources. In this proposal, we explore the key technologies of big internet multimedia data aggregation and presentation by leveraging hot web event detection. Specifically, we uncover the potential factors of hot web event and make use of time, location, causuality and other context informaton to find the thread of the event.The various factors and information can be aggregated for event analysis by using domain knowledge and social demands. We then build up users' profiles based on the analysis of user'perferences and behaviors. The detected hot web events finally can be presented in a multi-dimensional or multi-layer way. The study includes four aspects: the unified representation of heterogeneous data from internet media, semantics based aggregation over internet media, hot web event detection and media content presentation based on users' profiles. The achievements of this research will surely provide key technologies for the content understanding and intelligent management of big internet media data.

随着互联网和多媒体技术的发展，互联网承载的数据量呈现爆炸式增长，大数据的概念应运而生。本课题针对网络媒体大数据的智能管理与理解的现状，以网络热点事件为切入点，研究媒体大数据内容聚合与呈现的关键技术。具体而言，通过挖掘和发现热点事件的潜在因素，利用因果关系、时间、地点等上下文内容对热点事件的关键因素进行推断和串联，找出网络热点事件的关键点和脉络；利用领域知识构建和社会需求发现，对这些不同因素和主题信息进行聚合；在分析用户的属性、兴趣偏好、行为特征等信息的基础上，建立用户的兴趣模型；并结合用户的兴趣模型立体全方位呈现网络事件内容。研究内容包括网络媒体异构数据的统一表示，网络媒体信息的语义聚合，热点事件信息发现和结合用户建模的媒体内容呈现。本课题的研究成果将为网络媒体大数据的智能管理与理解提供核心算法与关键技术支持。

本课题致力于以网络热点事件为切入点，研究媒体大数据内容聚合与呈现的关键技术。具体而言，通过挖掘和发现热点事件的潜在因素，利用因果关系、时间、地点等上下文内容对热点事件的关键因素进行推断和串联，找出网络热点事件的关键点和脉络；利用领域知识构建和社会需求发现，对这些不同因素和主题信息进行聚合；在分析用户的属性、兴趣偏好、行为特征等信息的基础上，建立用户的兴趣模型；并结合用户的兴趣模型立体全方位呈现网络事件内容。在本项目资助下，课题组在四年的执行期中在聚合与呈现方向的若干关键科学问题上取得研究进展，累计发表高水平论文25篇，其中IEEE/ACM汇刊20篇，1篇论文获得CCF A类推荐论文最佳论文提名奖，授权发明专利3项，并合作研发了大连食品药品监督管理平台。