SenseFace: Towards a Context-Aware Social Network Framework

Social networks offer multimedia information consumption and sharing with millions of people around the world. The backbone of social networks stems from the advancements of different Web 2.0 technologies. Moreover, smartphones equipped with high speed Internet access have added ubiquitousness to Internet-based services. This gives rise to the fact that many people consume diversified services on a daily basis and have developed an association with different communities of interest via these services. However, a person only accesses a subset of these services at a given time either to consume certain information or to share information with a community of interest (COI), depending on the context. This poses three important research questions that need to be addressed: 1) "How do we dynamically capture user context from heterogeneous sources?" 2) "Which services are found to be interesting by a subject in a given context?" and 3) "How do we define the COI in a given context?" In this thesis, we propose detailed models and algorithms to answer the above three questions in the social network perspective. First, we propose algorithms to dynamically extract the services and social ties of an individual from the global heterogeneous social network space, for example, the Internet, which is called a personal social network (PSN). Second, we propose a framework named SenseFace, which employs a novel ubiquitous stack to extract user context from two sources: body sensor network (BSN) and the PSN. The BSN helps us in capturing different vital physical information, body activity and ambient information to infer user context while the PSN offers user context in the form of events, real-time messages, to-do lists, location and so on. Third, we present detailed models and algorithms to dynamically map a user context with a subset of relevant services and a subset of social network members. Finally, we present a prototype addressing the detailed design and implementation of the framework. The effectiveness of the framework has been validated by conducting both usability and run time performance testing