City Harmony: Exploring Shared Urban Characteristics
Introduction
Understanding how people feel, think, and perceive their environment, known as the Spatial Triad 1 , is important.
It's essential for policymakers and urban planners as they strive to create spaces that enhance human well-being and quality of life 2 .
How can we utilize the understanding of how people see, think, and feel their surrounding environment (referred to as locality characteristics)? Here are several practical implications:
- Public Health: Understanding environmental impacts on health
- Urban Planning: Guiding city development and resources
- Policy Making: Formulating sustainable policies
- Tourism and Travel: Enriching experiences with similar destinations
Can we use online text descriptions to understand locality characteristics?
Descriptions of things on Earth found online describe the physical space and how humans feel and think about space.
Online Text With Locations
Let's take a look at several online text descriptions in Minneapolis!
Toggle the point to see descriptions. Descriptions are Airbnb listing descriptions from publicly available Inside Airbnb 3 platform
By toggling the Airbnb listing in Marcy-Holmes as an example, we can see that the text descriptions often mention thoughts and feelings about the neighborhoods.
Approach
How can we leverage those online text descriptions to capture locality characteristics?
We develop an approach to automatically learn shared characteristics from text descriptions using machine learning methods 4 .
With the proposed approach, we can find cities with similar characteristics by grouping the locality embeddings together.
"Embeddings" in this context refer to the abstract numerical vectors that represent the characteristics of localities learned from machine learning methods.
Urban Characteristics
The map presents the results of the proposed approach, which learned shared urban characteristics from 687,006 Airbnb and Redfin listing descriptions across 176 global cities.
Let's explore cities with similar urban characteristics!
The findings indicate that New York and Boston are identified as similar cities (i.e., same group ID '0'), aligning with evidence from expert-curated data 5
Toggle polygons to see the city name and group ID. Cities within the same group ID are considered to have similar characteristics.
What cities have similar urban characteristics to Minneapolis and Saint Paul - group ID '11'?
The results show that Washington, D.C. shares similar characteristics with Minneapolis and Saint Paul - same group ID '11'.
San Francisco is another city identified as similar to Minneapolis and Saint Paul - same group ID '11'.
The result also shows that Hong Kong and Taipei share similar locality characteristics - group ID '12'.
For European cities, Milan, Florence, and Rome belong to the same group ID '18', meaning they have shared locality characteristics from online text.
Feel free to explore more cities with similar urban characteristics yourself!
Toggle polygons to see the city name and group ID. Cities within the same group ID are considered to have similar characteristics.
Enhancing Locality Analysis
The proposed approach can also process localities at the neighborhood level (instead of the city level) to identify shared locality characteristics of neighborhoods!
Are you curious about the details of the proposed approach for capturing locality characteristics from the online text? Check out our paper published at the ACM SIGSPATIAL 2023 Student Research Competition below.
