Sentence-Level Sentiment-Based Topic Modeling of Airbnb Reviews: Identifying Spatial Disparities in Commercial Areas of Gyeongsangnam-do and Proposing Regional Revitalization Strategies

Key Points

• The aim is to develop a new diagnostic framework that integrates sentiment analysis with spatial data to address regional economic revitalization.
• Analyzed commercial districts in Changwon and Geoje using sentiment analysis of Airbnb reviews.
• Visualized spatial structures and accommodation density with GeoPandas, DBSCAN, and KDE.
• Utilized KcELECTRA for sentiment classification and BERTopic for topic analysis.
• Found that Changwon's commercial sentiment is focused on accessibility and service, whereas Geoje emphasizes nature and relaxation.
• Identified significant differences in sentiment topic distribution based on commercial structure and accommodation density.
• Distinguished satisfaction and inconvenience factors, suggesting differentiated marketing strategies.

Abstract