HERec: Hyperbolic Framework to Break Information Cocoons in Recommender Systems
Researchers propose HERec, a hyperbolic framework designed to balance exploration and exploitation in recommender systems, addressing the problem of information cocoons that limit user exposure to diverse content. The framework introduces hierarchical representation where depth search enables exploitation and breadth search facilitates exploration, overcoming limitations of Euclidean methods in capturing hierarchies and hyperbolic methods lacking semantic understanding. HERec provides a principled mechanism for users to adjust recommendation preferences.
Key facts
- HERec is a hyperbolic framework for recommender systems.
- It balances exploration and exploitation to break information cocoons.
- Euclidean methods struggle to capture hierarchical structures.
- Hyperbolic methods lack semantic understanding of user and item profiles.
- The framework uses depth search for exploitation and breadth search for exploration.
- It allows users to adjust their recommendation preferences.
- The research is published on arXiv with ID 2411.13865.
- The paper was announced on arXiv as a replace-cross type.
Entities
Institutions
- arXiv