School of
Information Technology and Electrical Engineering

Speaker: Jingwei Ma
Seminar Date: Thu, 23/11/2017 - 10:00
Venue: 78-631; Data Science MMLab
Host: Prof Xue Li; Dr Mingyang Zhong

Seminar Type:  PhD Confirmation Seminar


Micro-video, as a new type of Internet media, has shown a rapid growth trend in social media. Nowadays, social media has evolved into one of the most important channel to share micro-videos. The sheer volume of micro-videos available in social networks often undermines user’s capability to choose the micro-videos that best fit their interests. Recommendation appear as a natural solution to this problem. Traditional video recommendation systems always produce relevant videos based on video content and user profiles, without exploring any social relations and video contextual semantics.

We demonstrate the roadmap to complete this research: Micro-Video Recommendation on Social Media. Firstly, a micro-video recommendation framework with content perception will be proposed, that considers multi-modal contextual information for providing recommendations fitting users' interests best. Specifically, multi-modal features such as textual and visual features of micro-videos are extracted using word embedding and recurrent neural network, and then the framework is designed based on a non-fully concatenating neural network in which the latent genre of micro-videos have been captured. Secondly, we enhance our framework by applying a social supervised latent genre model that considers users' social relationship and multiple interests for learning the latent genre of micro-videos in a supervised manner. Thirdly, a cross-platform instant interest aware micro-video recommendation model will be proposed that captures users' interests and the change of their interests inferred by the contents of micro-videos in finer dimensions from micro-video platforms and users' interests from social media. After collecting a large-scale real world dataset, extensive experiments on real-world dataset will be carried out to validate the effectiveness and the efficiency of our proposed method compared with state-of-the-art approaches.


Jingwei Ma received his B.S Degree in Information Technology from the Beijing University of Technology in 2014, and M.S Degree in Computer Science from the University of Queensland in 2016. Currently, he is doing PhD in at the School of Information Technology and Electrical Engineering, University of Queensland, under the supervision of Prof Xue Li and Dr Mingyang Zhong. His research interests include recommender systems, social media analysis and machine learning.