跨社交网络大数据下的用户建模研究

Among the 4V characteristics of social media big data, “variety” is key to value mining. Variety has two further interpretations as multi-modal and multi-source. While multi-modal has been extensively studied in the “small” data era, it is of vital importance to study multi-source in this project. In the context of social media, multi-source reflects the heterogeneous UGC in different Online Social Networks (OSN). This project aims to exploit the cross-OSN data for comprehensive user modeling. The challenges lie in four-fold: (1) the association between cross-OSN content data is implicit and complex; (2) the contradiction between the dynamic cross-OSN behaviors and the static user demographic attributes; (3) the cross-OSN behaviors are redundant and incompatible in user interest modeling; and (4) the sequential characteristic of cross-OSN footprint is critical for user modeling. Correspondingly, this project will focus on two basic scientific issues as cross-OSN content data integration and cross-OSN user data integration. Specific tasks include: (1) cross-OSN content data association mining, (2) cross-OSN user demographic attribute modeling, (3) cross-OSN user interest modeling, and (4) cross-OSN sequential behavior modeling. The research outcome will provide the fundamental theories and key technique support for multi-source social media big data fusion, and promote the applications in social media-based multimedia information awareness, service and regulatory.

社交媒体大数据的“4V”特征中，以多源性为代表的“Variety”是挖掘大数据价值的关键。社交媒体大数据中的多源性体现在不同社交网络产生的内容上，跨社交网络的用户建模是分析和应用多源社交媒体大数据的重要体现。相关研究处于起步阶段，主要研究难点体现在：（1）跨社交网络内容数据的关联是隐式的；（2）跨社交网络用户行为是动态的，而用户人口属性是静态的；（3）跨社交网络用户行为在反映用户兴趣属性时是冗余、甚至矛盾的；（4）跨社交网络用户行为的轨迹具有很强的时序性。本项目提出围绕跨社交网络内容数据整合和用户数据整合两个关键科学问题，拟研究如下四个研究内容：（1）跨社交网络内容数据关联挖掘，（2）跨社交网络用户人口属性建模，（3）跨社交网络用户兴趣属性建模，（4）跨社交网络序列行为建模。本项目的研究成果将为社交媒体的多源数据融合提供理论和技术基础，推动基于社交媒体的信息感知、服务和监管等应用的发展。

社交媒体大数据的“4V”特征中，以多源性为代表的“Variety”是挖掘大数据价值的关键。社交媒体大数据中的多源性体现在不同社交网络产生的内容上,跨社交网络的用户建模是分析和应用多源社交媒体大数据的重要体现。相关研究处于起步阶段，主要研究难点体现在：（1）跨社交网络内容数据的关联是隐式的；（2）跨社交网络用户行为是动态的，而用户人口属性是静态的；（3）跨社交网络用户行为在反映用户兴趣属性时是冗余、甚至矛盾的；（4）跨社交网络用户行为的轨迹具有很强的时序性。本项目提出围绕跨社交网络内容数据整合和用户数据整合两个关键科学问题，研究了如下四个研究内容：（1）跨社交网络内容数据关联挖掘，（2）跨社交网络用户人口属性建模，（3）跨社交网络用户兴趣属性建模，（4）跨社交网络序列行为建模。项目研究成果为社交媒体的多源数据融合提供理论和技术基础，有望推动基于社交网媒体的信息感知、服务和监管等应用的发展。