Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Boundary Prompting: Elastic Urban Region Representation via Graph-based Spatial Tokenization

Haojia Zhu, Jiahui Jin, Dong Kan, Rouxi Shen, Ruize Wang, Xiangguo Sun, Jinghui Zhang

TL;DR

This work tackles the rigidity of fixed urban region boundaries by proposing Boundary Prompting-based Urban Region Representation Framework (BPURF), which defines elastic regions via boundary prompts. BPURF builds a spatial token dictionary to encode urban entities as tokens within a graph, and introduces a region token set representation with multi-channel, subgraph-level message passing to produce adaptable region embeddings. A fast token-set extraction strategy enables online region construction during training and prompting, making the approach scalable for dynamic regions. Experiments across multiple cities and tasks demonstrate superior performance and efficiency for elastic region representations, highlighting strong generalization to unseen regions and robustness to varying region granularities. The framework offers practical benefits for urban planning and policy analysis by enabling task-specific, boundary-driven region representations without retraining for new region definitions.

Abstract

Urban region representation is essential for various applications such as urban planning, resource allocation, and policy development. Traditional methods rely on fixed, predefined region boundaries, which fail to capture the dynamic and complex nature of real-world urban areas. In this paper, we propose the Boundary Prompting Urban Region Representation Framework (BPURF), a novel approach that allows for elastic urban region definitions. BPURF comprises two key components: (1) A spatial token dictionary, where urban entities are treated as tokens and integrated into a unified token graph, and (2) a region token set representation model which utilize token aggregation and a multi-channel model to embed token sets corresponding to region boundaries. Additionally, we propose fast token set extraction strategy to enable online token set extraction during training and prompting. This framework enables the definition of urban regions through boundary prompting, supporting varying region boundaries and adapting to different tasks. Extensive experiments demonstrate the effectiveness of BPURF in capturing the complex characteristics of urban regions.

Boundary Prompting: Elastic Urban Region Representation via Graph-based Spatial Tokenization

TL;DR

This work tackles the rigidity of fixed urban region boundaries by proposing Boundary Prompting-based Urban Region Representation Framework (BPURF), which defines elastic regions via boundary prompts. BPURF builds a spatial token dictionary to encode urban entities as tokens within a graph, and introduces a region token set representation with multi-channel, subgraph-level message passing to produce adaptable region embeddings. A fast token-set extraction strategy enables online region construction during training and prompting, making the approach scalable for dynamic regions. Experiments across multiple cities and tasks demonstrate superior performance and efficiency for elastic region representations, highlighting strong generalization to unseen regions and robustness to varying region granularities. The framework offers practical benefits for urban planning and policy analysis by enabling task-specific, boundary-driven region representations without retraining for new region definitions.

Abstract

Urban region representation is essential for various applications such as urban planning, resource allocation, and policy development. Traditional methods rely on fixed, predefined region boundaries, which fail to capture the dynamic and complex nature of real-world urban areas. In this paper, we propose the Boundary Prompting Urban Region Representation Framework (BPURF), a novel approach that allows for elastic urban region definitions. BPURF comprises two key components: (1) A spatial token dictionary, where urban entities are treated as tokens and integrated into a unified token graph, and (2) a region token set representation model which utilize token aggregation and a multi-channel model to embed token sets corresponding to region boundaries. Additionally, we propose fast token set extraction strategy to enable online token set extraction during training and prompting. This framework enables the definition of urban regions through boundary prompting, supporting varying region boundaries and adapting to different tasks. Extensive experiments demonstrate the effectiveness of BPURF in capturing the complex characteristics of urban regions.

Paper Structure

This paper contains 42 sections, 22 equations, 8 figures, 7 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Boundary Prompting Framework
  4. Graph-based Spatial Tokenization
  5. Data Schema
  6. Spatial Token Dictionary and Graph Construction
  7. Region Representation from Token Set
  8. Token Embedding and Aggregation
  9. Message Passing for Token Set Embedding
  10. Subgraph Level Message Passing
  11. Multi-Channel Interaction Modeling
  12. Training Loss
  13. Fast Token Set Extraction for Prompting
  14. Experiments
  15. Experimental Setup
  16. ...and 27 more sections

Figures (8)

  • Figure 1: Elastic Region Representation with Boundary Prompting.
  • Figure 2: Comparison between boundaries of URR and EURR
  • Figure 3: The core idea of BPURF is to aggregate diverse spatial entities into a unified graph, where each entity is represented as a token. A region token set representation model is then trained to map these subgraphs to region embeddings. When a boundary prompt is provided, the fast token set extraction algorithm efficiently retrieves the relevant tokens and represents them using the trained model.
  • Figure 4: Average generation time of each region subgraph.
  • Figure 5: Results of ablation studies in NYC.
  • ...and 3 more figures