Assessing Park Satisfaction from Google Maps Reviews: Novel Evidence from Multimodal Text–Image Analysis

Parks are essential to urban well-being, making park satisfaction crucial for sustainable city development. Traditional survey-based approaches to understand sentiment towards parks among residents are often costly, time-consuming, and limited in scale. Recent social media–based studies have scaled such research but predominantly focus on text and frequently overlook visual information and the joint effects of text–image representations. This study presents an automated multimodal framework using crowdsourced reviews from Google Maps to model park satisfaction by integrating textual and visual features. Using Singapore as a case study, we analysed 76,869 textual reviews and 184,322 images associated with them. The results show that multimodal models are more useful than text-only approaches, with textual sentiment, emotional attributes, and image temporal characteristics identified as the most influential factors. These findings highlight the importance of multimodal analysis for advancing park research and informing planning and policy practices.