New SPACE framework boosts fairness for long-tail POIs in recommendation systems

Researchers have developed a new framework called SPACE (Supply- and Physics-Aware Conditional Embedding generation) to address fairness issues in next point-of-interest (POI) recommendation systems. These systems often favor popular locations, leaving less-known POIs under-exposed. SPACE aims to improve exposure for long-tail POIs by generating virtual user data that respects user constraints and POI supply limits. The framework involves stages for community inference, unbalanced optimal-transport allocation, and constraint-guided latent diffusion to create user embeddings. This generated data can then be used to train existing recommender models without architectural changes, as demonstrated by experiments showing improved provider fairness and maintained recommendation accuracy. AI

IMPACT Improves fairness in location-based services by ensuring less popular points of interest receive adequate exposure.

RANK_REASON Academic paper detailing a new framework for POI recommendation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.IR (Information Retrieval) →

• alphaXiv
• arXiv
• CatalyzeX
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• ScienceCast

• paper
• product
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →