Reproduce “User Memory Reasoning for Conversational Recommendation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hu Xu, Seungwhan Moon, Honglei Liu, Bing Liu, Pararth Shah, Bing Liu, Philip S. Yu
This research introduces a user memory graph for conversational recommendation and proposes the User Memory Graph Reasoner (UMGR) model, designed to dynamically manage and reason over user preferences. The approach, evaluated on a new dataset, achieved a 71.03% success rate, outperforming existing baselines in various recommendation and dialog policy metrics.
Key insights
- Dynamically updating the user memory graph is essential for effective conversational recommendation, allowing the system to adapt to evolving user preferences.
- Relational Graph Convolutional Networks (R-GCNs) can effectively reason over heterogeneous user memory graphs, supporting both dialog act prediction and entity ranking in an open space policy.
- The structured user memory graph inherently supports zero-shot reasoning for new users and items by learning a reasoning process independent of specific user or item identities.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.