AdaptGOT: A Pre-trained Model for Adaptive Contextual POI Representation Learning

2025-06-21

AI Summary

To address challenges in POI representation learning—including insufficient multi-context modeling, weak generalizability, and poor task adaptability—this paper proposes AdaptGOT, an adaptive pre-training framework. Methodologically, it introduces (1) a novel tri-source context modeling framework integrating Geographic, co-occurrence, and Textual (GOT) signals; (2) a hybrid neighborhood sampling strategy combining KNN, density-, importance-, and class-aware criteria; and (3) an attention-enhanced Mixture-of-Experts (MoE) encoder-decoder architecture regularized by Jensen–Shannon divergence to enforce topological consistency across heterogeneous contexts. Evaluated on multiple real-world datasets and downstream tasks—including POI recommendation and classification—AdaptGOT consistently outperforms state-of-the-art methods, demonstrating superior generalization capability and context-adaptive representation learning.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Search and Optimization: Learning to Search
• Machine Learning: Mixture of Experts (MoE)

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Currently, considerable strides have been achieved in Point-of-Interest (POI) embedding methodologies, driven by the emergence of novel POI tasks like recommendation and classification. Despite the success of task-specific, end-to-end models in POI embedding, several challenges remain. These include the need for more effective multi-context sampling strategies, insufficient exploration of multiple POI contexts, limited versatility, and inadequate generalization. To address these issues, we propose the AdaptGOT model, which integrates both the (Adapt)ive representation learning technique and the Geographical-Co-Occurrence-Text (GOT) representation with a particular emphasis on Geographical location, Co-Occurrence and Textual information. The AdaptGOT model comprises three key components: (1) contextual neighborhood generation, which integrates advanced mixed sampling techniques such as KNN, density-based, importance-based, and category-aware strategies to capture complex contextual neighborhoods; (2) an advanced GOT representation enhanced by an attention mechanism, designed to derive high-quality, customized representations and efficiently capture complex interrelations between POIs; and (3) the MoE-based adaptive encoder-decoder architecture, which ensures topological consistency and enriches contextual representation by minimizing Jensen-Shannon divergence across varying contexts. Experiments on two real-world datasets and multiple POI tasks substantiate the superior performance of the proposed AdaptGOT model.

Problem

• Improves multi-context sampling for POI representation
• Enhances exploration of geographical, co-occurrence, and textual POI contexts
• Addresses versatility and generalization in POI embedding models

Innovation

• Adaptive representation learning with GOT
• Advanced mixed sampling for contextual neighborhoods
• MoE-based adaptive encoder-decoder architecture