🤖 AI Summary
This study addresses the cold-start challenge caused by interaction sparsity in POI recommendation by proposing the LLM-MGCL framework. This approach integrates large language model semantics with geospatial information to construct auxiliary graphs, employing bidirectional contrastive learning to align behavioral, semantic, and spatial views. Such alignment effectively compensates for missing collaborative filtering signals and enhances graph neural network representations. Evaluations on the Yelp dataset demonstrate that LLM-MGCL outperforms LightGCN, achieving improvements of 52.0% in Recall@20 and 64.8% in NDCG@20. These results significantly surpass traditional baselines, validating the effectiveness of combining multimodal knowledge augmentation with contrastive learning to mitigate cold-start issues in location-based recommendation systems.
📝 Abstract
Point-of-interest (POI) recommendation models based on graph neural networks achieve strong performance by propagating collaborative signals over user-item interactions, yet they struggle with the cold-start problem, where items with few or no interactions are not represented. In this paper, we propose LLM-augmented Multi-Graph Contrastive Learning (LLM-MGCL), a multi-graph neural network that uses semantic and spatial information about items to extend the LightGCN backbone with two auxiliary item-item graphs: a semantic graph constructed from sentence embeddings of LLM-generated photo summaries and keywords, and a geographic graph derived from Haversine distances between business locations. Item embeddings are propagated over all three graphs in parallel, fused additively, and aligned across views through a bidirectional InfoNCE contrastive objective that connects behavioral, semantic, and spatial representations of the same items. Experiments on the Yelp Multimodal Recommendation Dataset show that LLM-MGCL outperforms classical collaborative filtering, matrix factorization, and interaction-only graph neural network baselines. It improves Recall@20 by 52.0% and NDCG@20 by 64.8% over LightGCN while performing on par with the strongest contrastive baseline, Self-supervised Graph Learning (SGL), which is also affected by the cold-start problem. An ablation study reveals that the cross-view contrastive alignment (CA) is the primary driver of these gains, with the best performance achieved when all three graphs are combined. Our results suggest that externally grounded, LLM-derived item knowledge can effectively compensate for missing collaborative signal and mitigate the item cold-start problem in POI recommendation.