.png){kind=icon}
.png){kind=icon}
Return
A two-stage hybrid-computing framework for generating metro-style tourist maps: integrating a fine-tuned LLM with geospatial optimization
.png){kind=icon}
PRE
AI
Z
D
C
Z
J
T
DOI：10.1080/13658816.2026.2699203.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
A well-designed tourist map can help tourists enjoy a more organized and informative journey. However, existing digital map service platforms lack the ability to automatically generate customized route maps that accommodate diverse tourist demands. To address this, we proposed a two-stage hybrid-computing framework. First, a tourism-specific large language model (LLM) was developed via parameter-efficient fine-tuning on two datasets (i.e. attraction description and route planning Q&A datasets) to generate structured travel recommendations. Then, Voronoi-diagram-based attraction distribution optimization and grid-based path optimization algorithms were implemented to generate metro-style tourist route maps based on the LLM’s recommendations. An experimental evaluation was conducted using objective quantitative metrics (e.g. BLEU-4) and subjective surveys with over 300 participants. Experimental results demonstrated that the fine-tuned LLM outperformed baseline LLMs in recommending attractions and routes, and the generated four-direction metro-style tourist route maps surpassed traditional tourist maps in legibility, recognition of attractions and routes, esthetics, and satisfaction. These findings indicate that the proposed framework can effectively generate customized metro-style tourist route maps with high usability. This study provides a hybrid framework from demand input to schematic map output, advancing the role of cartography in tourism services.
Keywords:
Tourist map
metro-style map
large language models
schematic representation
geospatial optimization
Journal
IF:
5.1
Papers:
2.7K
Citations:
9.3K
Organization
Cited Papers