User Profiling across Multiple Social Networks

User profiling, which aims to infer users' unobservable information based on observable information such as individual's behavior or utterances, is the basis for many applications, such as personalized recommendation, and expert finding. Traditional user profiling conducted with traditional medium, such as document records, is always hindered by the limited data sources. Recent years, the proliferation of social media has opened new opportunities for user profiling. Moreover, as different social networks provide different services, increasing number of people are involved in multiple social networks. Different aspects can be revealed by different social networks. Therefore, to comprehensively learn users' profiles, it is time to shift from a single social network to multiple social networks.

In a sense, user profiling across multiple social networks in nature can be treated as multi-source learning. Furthermore, according to the nature of users' attributes, user profiling across multiple social networks can be both framed by a multi-source mono-task learning scheme or a multi-source multi-task learning scheme. For example, users' gender or volunteerism tendency can be learned by the multi-source mono-task learning scheme, where only one binary classification (task) is involved. When it comes to learning users' interests, which usually involves a set of binary classifications (tasks), it should be appropriate to frame it in the multi-source multi-task learning scheme. Consequently, we first proposed a multi-source mono-task learning scheme for user profiling on multiple social networks, and applied it in a practical scenario: volunteerism tendency prediction. Sequentially, we moved from the mono-task scenario to the multi-task context, proposed a multi-source multi-task learning scheme, and applied it to the application of user interest inference.

Privacy Preserving in Social Media

The boom of social networks has given rise to a large volume of user-generated contents (UGCs), most of which are freely and publicly available. The potential of using the rich set of UGCs to study people's personal attributes and personalized applications has been widely validated. Despite its value, UGCs can also place users at high privacy risks, which thus far remains largely untapped. Privacy is defined as the individual's ability to control what information is disclosed, to whom, when and under what circumstances. As people and information both play significant roles, privacy has been elaborated as a boundary regulation process, where individuals regulate interaction with others by altering the openness degree of themselves to others. Therefore, we aim to reduce users' privacy risks on social networks by answering the question of Who Can See What.