Crowdsourced Intelligence for Inclusive Urban Accessibility
Traditional accessibility surveys are valuable but expensive and geographically limited. This project uses large-scale online reviews to identify where residents experience persistent barriers across everyday urban services.
Technical innovation
The study analyzes more than one million Google Maps reviews using a LoRA-fine-tuned Llama 3 model and socio-spatial regression. It links public accessibility sentiment across restaurants, retail, hotels, and healthcare to demographic and neighborhood conditions.
Practical insight
Negative accessibility experiences remain widespread, and perceptions vary with socioeconomic context. Crowdsourced evidence can help planners, policymakers, and advocates prioritize field audits and target improvements toward more inclusive urban design.