Knowledge Graph Tokenization for Behavior-Aware Generative Next POI Recommendation

2025-09-15

AI Summary

Existing generative POI recommendation methods suffer from two key limitations: (1) conventional tokenization fails to model heterogeneous recommendation signals, leading to semantic information loss; and (2) instruction fine-tuning solely captures POI visit behaviors while ignoring multimodal mobility patterns—such as walking and driving—thereby limiting behavioral understanding. To address these issues, we propose KGTB, a knowledge graph–driven tokenization framework. KGTB constructs a structured knowledge graph from user–POI–behavior triplets and designs a structure-aware tokenizer to preserve heterogeneous semantics. Furthermore, it introduces a novel instruction fine-tuning paradigm for joint prediction of multiple behavioral modalities—including visits, displacements, and interactions—enabling unified behavioral modeling. Extensive experiments on four real-world city-scale datasets demonstrate that KGTB consistently outperforms state-of-the-art methods, validating its effectiveness in fine-grained behavioral perception and generative POI recommendation.

Technology Category

• Natural Language Processing: Generation
• Planning, Routing, and Scheduling: Planning with Language Models
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Generative paradigm, especially powered by Large Language Models (LLMs), has emerged as a new solution to the next point-of-interest (POI) recommendation. Pioneering studies usually adopt a two-stage pipeline, starting with a tokenizer converting POIs into discrete identifiers that can be processed by LLMs, followed by POI behavior prediction tasks to instruction-tune LLM for next POI recommendation. Despite of remarkable progress, they still face two limitations: (1) existing tokenizers struggle to encode heterogeneous signals in the recommendation data, suffering from information loss issue, and (2) previous instruction-tuning tasks only focus on users' POI visit behavior while ignore other behavior types, resulting in insufficient understanding of mobility. To address these limitations, we propose KGTB (Knowledge Graph Tokenization for Behavior-aware generative next POI recommendation). Specifically, KGTB organizes the recommendation data in a knowledge graph (KG) format, of which the structure can seamlessly preserve the heterogeneous information. Then, a KG-based tokenizer is developed to quantize each node into an individual structural ID. This process is supervised by the KG's structure, thus reducing the loss of heterogeneous information. Using generated IDs, KGTB proposes multi-behavior learning that introduces multiple behavior-specific prediction tasks for LLM fine-tuning, e.g., POI, category, and region visit behaviors. Learning on these behavior tasks provides LLMs with comprehensive insights on the target POI visit behavior. Experiments on four real-world city datasets demonstrate the superior performance of KGTB.

Problem

• Improving tokenization to encode heterogeneous recommendation data with less information loss
• Addressing limited behavior understanding by incorporating multiple behavior types in learning
• Enhancing generative next POI recommendation through knowledge graph structural preservation

Innovation

• Knowledge graph tokenizer preserves heterogeneous information
• Multi-behavior learning tasks for LLM fine-tuning
• Structural ID quantization supervised by KG structure