DOCTOR OF PHILOSOPHY
Mohan Raj Chanthran( Award Conferred 25 June 2025)
Title of Thesis:‘ Document-Level Zero-Shot Relation Extraction from Malaysian English News Article.’
This thesis develops a method to predict unseen relation between entities in Malaysian English news articles. Due to the lack of suitable datasets and tools, the research creates the MEN-Dataset and develops MEN-spaCy, a specialised Named Entity Recognition( NER) model for Malaysian English. The final part of the work introduces a Relation Extraction framework that identifies unseen relations from input documents or news articles. The development of the dataset, Named Entity Recognition, and Relation Extraction approach creates a foundation for advancing more NLP-related work in Malaysian English.
Supervisor: Associate Professor Soon Lay Ki Associate Supervisor: Dr Huey Fang Ong External Supervisor: Dr Bhawani Selvaretnam
Yin Yin Low( Award Conferred 19 February 2025)
Title of Thesis:‘ Adversarial Machine Learning for Emotional Privacy.’
This thesis explores leveraging adversarial machine learning to safeguard emotional privacy in the era of pervasive social media. It proposes a method that introduces subtle modifications, or " adversarial perturbations ", to video-based emotion recognition, making emotions more difficult to detect while preserving natural content. The study examines the applicability of these techniques across multimodal systems combining text, images, and video. A novel architecture for universal adversarial attacks is presented, with evaluations demonstrating its effectiveness in maintaining privacy, robustness, and transferability. This study emphasizes the potential of machine learning in responsibly safeguarding privacy in real-world affective computing applications.
Supervisor: Professor Raphael Phan Associate Supervisor: Dr Arghya Pal Associate Supervisor: Dr Xiaojun Chang
Cheng Chi Qin( Award Conferred 30 April 2025)
Title of Thesis:‘ l2Match: Optimizing Subgraph Isomorphism for Efficient Small Query Matching on Labelled Graphs.’
This research addresses Subgraph Isomorphism, proposing four optimization techniques within a filter-and-verification framework to reduce search space. Filtering maps query graph vertices to potential data graph candidates, pruning invalid options. Verification recursively explores solutions, with recent studies suggesting stronger constraints and heuristic approaches. A new algorithm, l2Match, integrates Label-Pair Indexing, Backward Candidate Pruning, and Jump-and-Redo methods for small queries, outperforming existing solutions. Real-world and synthetic dataset evaluations confirm l2Match ' s superiority, highlighting its advantages over Constraint Programming. However, label-based indexing efficacy varies with datasets. Applying optimizations to existing algorithms enhances query completion times.
Supervisor: Professor Wong Kok Sheik Associate Supervisor: Associate Professor Soon Lay Ki
G R A D U A T I O N C E R E M O N Y 41