Exploring Urban Park Audiosphere Through User-Generated Content Network

Abstract

Urban park soundscapes are a critical yet underexplored component of the visitor experience, intimately linked to well-being and ecological quality. This study harnesses user-generated online reviews to investigate how park users perceive and describe the acoustic environment of a major urban park (Szczytnicki Park in Wrocław, Poland). We collected 1,000 Google reviews and applied advanced text mining techniques – including multilingual sentence embedding (SBERT) and hierarchical clustering – to unravel prevalent themes. A semantic network analysis was then conducted to map connections between comments, using network science metrics and community detection to reveal both cohesive clusters and cross-theme linkages. The results identified seven distinct thematic clusters of park perceptions, among which a prominent cluster centered on soundscape experiences emerged. Reviews in this cluster consistently highlighted tranquility, natural sounds (birdsong, flowing water), and relief from urban noise, indicating a shared appreciation of the park as an auditory refuge. The network topology further showed that these soundscape-centered comments form a tightly knit sub-network that is strongly interwoven with broader themes of relaxation and enjoyment. Key network metrics, such as a high density and short average path length within the soundscape community, underscore the internal consistency of these auditory perceptions, while connections to other clusters demonstrate that positive soundscape experiences amplify overall park satisfaction. This research offers a data-driven perspective on urban soundscapes: by integrating text mining and network analytics, we effectively use public perceptions as ecological indicators of soundscape quality and park restorative value.

Keywords: ecoacoustics, soundscape, Natural language processing, Urban parks, network science, ecoinformatics