Publication Type
PhD Dissertation
Version
publishedVersion
Publication Date
5-2025
Abstract
Understanding user preferences remains a central challenge in recommender systems due to their inherently complex, unstructured, and multi-faceted nature, exacerbated by the sparsity of user interaction data. Traditional approaches often compress user interests into a single latent vector, overlooking the fact that user preferences are typically shaped by multiple underlying factors that differ across individuals. These latent drivers are not directly observable and must be discovered through unsupervised modeling, further complicated by limited historical interactions per user.
This dissertation addresses these challenges by introducing a principled framework for multiinterest modeling, which disentangles user behaviors into multiple latent factors to better reflect the diversity of user preferences. The contributions are organized into two main directions: (i) modeling multiple user interests from interaction data through iterative and compositional clustering strategies, and integrating interest sharing among users; and (ii) enhancing multi-interest user representations with side information, such as textual content, via cross-modal interest alignment. These innovations enable more accurate and granular user modeling, significantly improving recommendation performance across various real-world benchmarks.
A key novelty of this work lies in its systematic treatment of disentangled preference learning, where each model is designed to reveal interpretable and semantically meaningful user interest factors. This not only improves predictive performance but also supports model transparency: offering insights behind latent representations of user interest factors. Furthermore, the methods developed in this thesis contribute broadly to the fields of representation learning and interpretable AI, which could be applied for other kinds of sparse and heterogeneous data.
Keywords
Multi-interest modeling, disentangled representation learning, recommender systems
Degree Awarded
PhD in Computer Science
Discipline
Artificial Intelligence and Robotics
Supervisor(s)
LAUW, Hady Wirawan
First Page
1
Last Page
165
Publisher
Singapore Management University
City or Country
Singapore
Citation
TRAN, Nhu Thuat.
Disentangling user preferences towards self-interpretable recommender systems. (2025). 1-165.
Available at: https://ink.library.smu.edu.sg/etd_coll/772
Copyright Owner and License
Author
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.