(From left to right): Prof. CHEN Changwen, Chair Professor of Visual Computing; JIANG Yiyang; YANG Zhanpeng; Dr HUANG Xiao, Assistant Professor.
We are pleased to announce that two of our graduates, JIANG Yiyang and YANG Zhanpeng, achieved outstanding results in the Final Year Project Competition 2023-2024, organised by the IEEE (Hong Kong) Computational Intelligence Chapter. This competition is open to all tertiary institutions in Hong Kong, with participants submitting their final-year undergraduate projects for evaluation.
JIANG Yiyang won the Champion award for his project, "Video Moment Retrieval Using Deep Learning Models." This project enhances video moment retrieval (VMR) by integrating large language models (LLMs) to address challenges in generating continuous outputs for inter-frame relations, such as salience scores and inter-frame embeddings. His approach refines inter-concept relations in multimodal embeddings, enabling effective retrieval without relying on textual training. “Winning this award has been an incredibly rewarding experience that has taught me valuable lessons about independent research, exploration, and the application of academic research to real-world scenarios,” Yiyang said. He extends his gratitude to his supervisor, Prof. CHEN Changwen, Chair Professor of Visual Computing, as well as Prof. LI Qing, Chair Professor of Data Science and Head, Prof. WEI Xiaoyong, Visiting Professor of COMP, for their guidance and invaluable support.
YANG Zhanpeng received the Merit award for his project, "Generating Attributes for Knowledge Graphs with ChatGPT - Distant Supervision for Sentence Classification", supervised by Dr HUANG Xiao, Assistant Professor. This project explores sentence classification using distant supervision methods to aid knowledge graph construction. It classifies sentences into categories like definitions and examples, utilising ChatGPT to generate noisy labels for training. The project shows promising results in improving classification accuracy. Zhanpeng supplemented that future work could focus on improving data diversity and classification robustness, and addressing ethical considerations, such as cultural differences and information integrity. This research contributes to automating knowledge graph enrichment, with potential applications in educational technology and content organisation.
