Leveraging the Twitter Based Natural Language Processing (NLP) in Sentiment Analysis of Urban and Rural Views

This study examines sentiment differences in rural and urban communities' Twitter discourse on clean energy (solar, wind, electric vehicles). Using a geocoded Twitter corpus from January–June 2024, we apply NLP-based sentiment analysis and comparative statistical testing to explore whether regional sentiment diverges. We collected 1.2 million tweets geolocated to the U.S., classified as rural or urban based on U.S. Census tract codes. Both lexicon-based (VADER) and transformer-based (RoBERTa) models are used. Results indicate urban tweets express more positive sentiment (mean sentiment score = 0.18) compared to rural (mean = 0.10), and variance analysis confirms differences are statistically significant (p < 0.001). Subtopic analysis reveals rural skepticism around cost and infrastructure, while urban tweets emphasize climate action and innovation. These findings have policy implications for targeting cleanenergy communication regionally. We discuss limitations and propose future directions involving deeper topic modeling and multilingual expansion.