Analyzing Public Sentiments on Environmental Noise Pollution Using Social Media Data

Authors:

Abstract
Environmental noise pollution is a pervasive, yet frequently under-addressed, modern environmental health hazard that actively degrades public health and urban habitability. Traditional methods of acoustic assessment rely heavily on localized physical decibel measurements and manual cross-sectional surveys, which are constrained by high operational costs, narrow spatial coverage, and a lack of real-time public sentiment capture. This study introduces an advanced environmental computational framework that utilizes large-scale social media data to decode public perceptions, annoyance dynamics, and emotional responses to urban noise pollution. Extracting a comprehensive dataset of 254,120 microblogging posts across multiple geographic nodes, we deployed an optimized Bidirectional Encoder Representations from Transformers (BERT) architecture to categorize sentiments into positive, negative, and neutral dimensions. Natural Language Processing (NLP) workflows further isolated specific sound sources behind public grievances. The results reveal a striking dominance of negative public sentiment ( 62.4% ), with road traffic and construction operations emerging as the primary sources of environmental irritation. Spatial-temporal analytics uncovered distinct spikes in negative sentiment during night-time hours, indicating widespread sleep disruption. This research demonstrates that social media data mining acts as a highly effective, continuous proxy for real-time human acoustic perception, providing environmental health authorities and urban planners with an actionable framework to design data-driven noise mitigation policies and build quieter, more resilient smart cities.