Making recommenders tell the truth.
Recommender systems decide, quietly, what billions of people see each day. I study how to make them faithful to what users actually want — not just to whatever maximises click-through — and how modern language models can help align those two goals.
The triad my WSDM'25 paper aligns.
My recent work treats user, item and review representations as three points in a shared embedding space — and asks the model to pull them together through a contrastive objective.
Hover any node to see which edges light up: every pair is a loss term the model tries to satisfy.
Research topics
Review-aware & Contrastive Recommendation
How can free-text reviews — the only place users actually say why they liked something — drive a recommender? I design contrastive objectives that align user, item and review representations so that language is a signal, not garnish.
LLMs for Recommendation & IR
Large language models are tempting black boxes for ranking. I'm interested in where they genuinely add value (reasoning over sparse interactions, cold-start, explanation) versus where they just add cost and hallucination.
Fairness & Risks in Recommenders
Who gets recommended, and who doesn't? I look at fairness measures for both users and items in implicit-feedback settings, and at the risks that popular algorithms quietly amplify.
Text & Data Mining
Classification, clustering, named entity recognition, keyword extraction, intent analysis — the unglamorous plumbing that makes everything else work. I have a soft spot for Vietnamese NLP problems.
Publications
Google Scholar →2025
-
WSDM 2025New
A Contrastive Framework with User, Item and Review Alignment for Recommendation
The 18th ACM International Conference on Web Search and Data Mining.
2024
-
KES 2024
Collaborative Fair-is-Better Filtering for Implicit Feedback
International Conference on Knowledge-Based and Intelligent Information & Engineering Systems.
Author list marked with underline indicates me. Links will be updated as camera-ready versions and code are released.
Advisors & Collaborators
Yuan Fang
Associate Professor, SMU School of Computing and Information Systems.
Hady W. Lauw
Associate Professor, SMU School of Computing and Information Systems.