Fine-grained assessment of greenspace satisfaction at regional scale using content analysis of social media and machine learning

Abstract

Assessing perceptions of green spaces is of considerable interest to developers aiming for sustainable urbanization. However, there are numerous challenges facing the development of a rapid, effective, and fine-grained method to assess large-scale greenspace perception. Survey-based studies of perception yielded detailed assessments of green spaces but lacked regional comparisons. The few big-data-based studies of greenspace perception lacked fine-grained explorations. Therefore, we used content analysis to interpret perception in two ways: perceived frequency and perceived satisfaction, including overall park satisfaction and satisfaction with individual landscape features. We analyzed social media posts about urban parks in Beijing, China. A structured lexicon was developed to capture detailed landscape features, and machine learning was employed to assess satisfaction levels. Both of these techniques performed well in interpreting greenspace satisfaction from volunteered textual comments. A detailed study of 50 parks demonstrated that overall park satisfaction was positive. Additionally, individual landscape features were more influential than frequency of landscape features in affecting satisfaction. Our framework confirmed the potential of online comments as complementary to traditional surveys in assessing greenspace perception, while enhancing our understanding of this perception on a regional scale. Practically, this study can facilitate sustainable policy-making regarding urban green spaces, specifically through offering a structured landscape-feature lexicon, rapid regional comparison of various parks, and an emphasis on quality rather than quantity of landscape features.