Interaction-Grounded Semantic Graph Refinement for LLM-based Recommendation
2025 | IEEE Access
Abstract
Large language models (LLMs) have recently demonstrated remarkable potential for
recommendation by reframing it as a text-generation task. Recent LLM-based
approaches apply GNNs to capture higher-order collaborative patterns from
interaction graphs but handle textual data separately, failing to extract
higher-order semantic patterns from text. One potential solution is to construct
semantic graphs to better capture such semantic relationships. However, naively
selecting top-k connections by profile similarity introduces significant
challenges: 1) preference gap between textual similarity and actual user
behavior, and 2) semantic distortion where identical attributes carry different
meanings for users versus items. To overcome these issues, we propose an
Interaction-Grounded Semantic Recommender (IGSRec), which constructs an
interaction-grounded semantic graph by aligning profile-based connections with
observed interactions. IGSRec employs an LLM-based profile generator, constructs
a top-k semantic graph, then refines it using a learnable scoring function that
identifies relevant semantic neighborhoods conditioned on user-item
interactions. Through dual-graph propagation over both the refined semantic and
interaction graphs, IGSRec captures higher-order semantic and collaborative
patterns. Experiments on Amazon Review benchmarks demonstrate state-of-the-art
performance in both direct and sequential recommendation tasks. Our code and
data are publicly available at https://github.com/oseoko/IGSRec/
Review & Notes
(Detailed review and notes will be added here.)
Resources
← Back to Paper List
|