Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Toward satisfactory public accessibility: A crowdsourcing approach through online reviews to inclusive urban design

Lingyao Li, Songhua Hu, Yinpei Dai, Min Deng, Parisa Momeni, Gabriel Laverghetta, Lizhou Fan, Zihui Ma, Xi Wang, Siyuan Ma, Jay Ligatti, Libby Hemphill

TL;DR

This study uses Google Maps reviews across the United States and Llama 3 model with the Low-Rank Adaptation technique to analyze public sentiment on accessibility and highlights the potential of crowdsourcing via online reviews for identifying accessibility challenges and providing insights for urban planners.

Abstract

As urban populations grow, the need for accessible urban design has become urgent. Traditional survey methods for assessing public perceptions of accessibility are often limited in scope. Crowdsourcing via online reviews offers a valuable alternative to understanding public perceptions, and advancements in large language models can facilitate their use. This study uses Google Maps reviews across the United States and fine-tunes Llama 3 model with the Low-Rank Adaptation technique to analyze public sentiment on accessibility. At the POI level, most categories -- restaurants, retail, hotels, and healthcare -- show negative sentiments. Socio-spatial analysis reveals that areas with higher proportions of white residents and greater socioeconomic status report more positive sentiment, while areas with more elderly, highly-educated residents exhibit more negative sentiment. Interestingly, no clear link is found between the presence of disabilities and public sentiments. Overall, this study highlights the potential of crowdsourcing for identifying accessibility challenges and providing insights for urban planners.

Toward satisfactory public accessibility: A crowdsourcing approach through online reviews to inclusive urban design

TL;DR

This study uses Google Maps reviews across the United States and Llama 3 model with the Low-Rank Adaptation technique to analyze public sentiment on accessibility and highlights the potential of crowdsourcing via online reviews for identifying accessibility challenges and providing insights for urban planners.

Abstract

As urban populations grow, the need for accessible urban design has become urgent. Traditional survey methods for assessing public perceptions of accessibility are often limited in scope. Crowdsourcing via online reviews offers a valuable alternative to understanding public perceptions, and advancements in large language models can facilitate their use. This study uses Google Maps reviews across the United States and fine-tunes Llama 3 model with the Low-Rank Adaptation technique to analyze public sentiment on accessibility. At the POI level, most categories -- restaurants, retail, hotels, and healthcare -- show negative sentiments. Socio-spatial analysis reveals that areas with higher proportions of white residents and greater socioeconomic status report more positive sentiment, while areas with more elderly, highly-educated residents exhibit more negative sentiment. Interestingly, no clear link is found between the presence of disabilities and public sentiments. Overall, this study highlights the potential of crowdsourcing for identifying accessibility challenges and providing insights for urban planners.
Paper Structure (28 sections, 5 equations, 7 figures, 4 tables)

This paper contains 28 sections, 5 equations, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Accessibility in urban planning
  4. Crowdsourcing to investigate accessibility
  5. Crowdsourcing to study urban environments
  6. Active crowdsourcing
  7. Passive crowdsourcing
  8. Data and Methods
  9. Data preparation
  10. Data annotation
  11. Attitude classification
  12. Sentiment models
  13. TF-IDF baseline models
  14. Large language models (LLMs)
  15. Performance measures
  16. ...and 13 more sections

Figures (7)

  • Figure 1: An illustrative framework to conduct the data analysis.
  • Figure 2: Classification performance of candidate models measured by Precision, Recall, F1-score, and Accuracy.
  • Figure 3: Descriptive results for sentiment by POI. (a) Number of POIs by types. (b) Distribution of sentiment. The x-axis represents the weighted sentiment by averaging all sentiments associated with a POI, while the y-axis shows the POI count.
  • Figure 4: Semantic analysis using LSVA based on POI types. The x-axis represents the salience using equation \ref{['euqation: salience']}, while the y-axis represents the valence using equation \ref{['euqation: valence']}.
  • Figure 5: Spatial distribution of accessibility sentiment.
  • ...and 2 more figures