Research projects in Information Technology

Our research group tries to decipher the rules that govern decision making in social groups, from animals that forage and hunt in groups to humans that work in teams.

A large part of modern life is lived indoors such as in homes, offices, shopping malls, universities, libraries and airports. However, almost all of the existing location-based services (LBS) have been designed only for outdoor space. This is mainly because the global positioning system (GPS) and other positioning technologies cannot accurately identify the locations in indoor venues.

The proposed project aims to develop new methodologies for developing NMT systems between extremely low-resource languages and English. Recent advances in neural machine translation (NMT) are a significant step forward in machine translation capabilities. However, "NMT systems have a steeper learning curve with respect to the amount of training data, resulting in worse quality in low-resource settings".

 Although deep learning has produces state of the art results on many problems, it is a data hungry technology requiring a lot of human supervision in the form of annotated data. Potential PhD topic include learning to learn and meta-learning, active learning, semi-supervised learning, multi-task learning, transfer learning, and learning representations for NLP. Techniques include deep generative models (eg auto-encoders and generative adversarial networks) and reinforcement/imitation learning algorithms for Markov Decision Processes.

The project develops methods to use acoustic data for the identification of animals in the wild and in controlled settings. It is part of a broader effort to build AI-enabled methods to support biodiversity and sustainability research. The initial objective is to use deep learning techniques to perform acoustic species identification in real-time on low-cost sensing devices coupled to cloud-based backends. Ultimately, we are aiming to move to Edge-AI, ie.

This project aims to design effective and intelligent search techniques for large scale social network data. The project expects to advance existing social network search systems in three unique aspects: utilizing the geographical locations of queries and social network data to provide more relevant results; acknowledging and handling inherent uncertainties in the data; and exploiting knowledge graphs to produce intelligent search results. Expected outcomes of this project include a next-generation social network search system and enhanced international collaborations.