James
She,
Department
of
Electronic
and
Computer
Engineering
Hong
Kong
University
of
Science
and
Technology
The information about the social graph of a user (i.e., the social connections and related interactions of a user) is essential for many social computing techniques and applications, such as content recommendation, user profiling and community detection, in many social media and e-commerce platforms. However, this online social information is getting difficult to access due to the user privacy settings, or exclusively accessible by dominating platforms, such as Facebook, Instagram, Tencent Wechat, Pinterests, etc. and their affiliated parties. Without relying on this social graph information alone, our research investigates the possibility and how to learn the user backgrounds, interests and their connections through their shared content (e.g., user-generated pictures, videos, and their liked, shared and commented media content) using some novel machine learning and social computing techniques.
In this talk, I will describe how the joint approaches of multimedia big data analytics and social computing can offer a more accessible and powerful alternative to achieve many social computing techniques with comparable performance and interesting social applications even without the social graphs. Besides sharing our excited findings, the results from this area of our research have also led to some impacts beyond the publications, such as some patent-filing, tech-transfer activities and a startup company founded by our researcher with the related innovations from this research.
With the recent advancements in machine learning and AI technologies, more interesting research challenges and opportunities on data science in multimedia and social computing are induced. Hence, we are able to better understand what content to be recommended or generated at certain timings or locations for emerging social media and multimedia applications in our smart cities and digital societies.