Leveraging large language models for tourism research based on 5D framework: A collaborative analysis of tourist sentiments and spatial features

https://doi.org/10.1016/j.tourman.2024.105115

Abstract

Experience-oriented travel models have posed new demands for optimizing urban environments to promote tourism development. This study introduced a natural language classification and scoring method to explore the relationship between tourism experiences and spatial characteristics. We found that online textual data can infer and represent physical spatial features. Our findings include: (1) Tourists perceive density from moving objects, with threshold effects caused by their temporal instability. (2) Ecological and cultural-technological tourism models have varied dependencies on transportation facilities. (3) Central areas dominated by artificial functions and landscapes require more natural planning approaches to enhance the tourist experience. (4) Accessibility perceptions are influenced by driving time and proximity to the city center, rather than walking duration or the actual distance. (5) The development of a dual-network policy for buses and subways is crucial to enhance the travel experience. Our study provides evidence-based recommendations for urban renewal to improve tourism experiences.

📍 Background

With the rise of experience-oriented tourism models, optimizing urban environments has become a significant demand for promoting tourism development. The quality of the urban environment directly affects tourists' experiences and satisfaction, influencing the sustainable development of the tourism industry. Although tourists' evaluations of the urban environment often encompass factors such as transportation convenience, environmental comfort, and cultural atmosphere, there is currently a lack of sufficient spatiotemporal data to track the characteristics of the built environment, making it difficult to propose targeted interventions for spatial optimization.

🎯 Research Objective

This study aims to explore the relationship between tourism experiences and spatial characteristics by introducing a natural language classification and scoring method. Using the widely applied 5D framework (Density, Diversity, Design, Destination Accessibility, and Distance to Transit), the study analyzes the relationship between tourist sentiments and urban spatial features in online hotel reviews. The goal is to extract environmental features and tourist perceptions from hotel reviews and provide evidence-based recommendations for urban renewal to enhance tourism experiences.

🧰 Methodology

• Data Collection: The study focuses on Shenzhen as the research area, collecting hotel review data from Ctrip and geographic information system (GIS) data to analyze tourists' sentiments in hotel reviews and construct regression models with urban spatial characteristics.
• Application of 5D Framework: The 5D framework was used to quantify emotional values related to space in the reviews and integrate these with physical urban characteristics, such as building coverage, street networks, and public transport density, for multidimensional analysis.
• Model Analysis: The study uses XGBoost and Multiscale Geographically Weighted Regression (MGWR) models to explore the nonlinear relationships between virtual tourist perceptions and physical spatial features.

📊 Research Results

1. Perception of Density: Tourists’ perception of density is influenced more by dynamic factors like pedestrian and vehicular flow rather than static structures. The perception of density is positively correlated with congestion in city center areas.
2. Perception of Diversity: Different tourism development models, such as ecological and cultural tourism, have varying dependencies on transportation infrastructure. Central areas require optimized public transport systems, while suburban areas need more self-driving transportation options.
3. Design Perception: Central urban areas, dominated by artificial environments, lack integration of natural landscapes, leading to a lower design experience for tourists. Increasing greenery and improving architectural design can significantly enhance the overall experience.
4. Destination Accessibility: Tourists are more concerned with driving time and proximity to the city center rather than walking duration or actual geographic distance when evaluating destination accessibility.
5. Transit System: The study confirms that a dual-network policy combining buses and subways is crucial for enhancing the tourist travel experience, especially by improving transportation networks from city centers to suburban areas.

🌱 Research Contributions

• This study integrates large language models and natural language processing technologies to provide an innovative perspective on spatial feature quantification in urban planning.
• A novel method for classifying hotel review texts using the 5D framework is proposed, overcoming the limitations of simple keyword extraction methods, achieving more accurate sentiment analysis and spatial data integration.
• The study provides urban planners with practical recommendations for optimizing urban spatial layouts based on tourist feedback, particularly in areas like transportation, environmental design, and spatial diversity.

🏙 Practical Implications

This research offers data-driven insights for urban tourism planning, promoting the idea of optimizing urban spaces from the perspective of tourists. The findings not only help improve the quality of urban tourism environments but also provide a reference for other cities in the process of tourism recovery post-pandemic, especially for enhancing visitor satisfaction through spatial optimization.

Fig. 4. Collaborative analysis: (a) Density-related environmental features and perceived density; and (b) Diversity-related environmental features and perceived diversity.