DOI: 10.26466/opusjsr.1933468 ISSN: 2791-9781

Analysis of satisfaction perceptions through online comments: The case of Artvin

Muhammed Çağrı Aksu, Sılanur Soyar, Emine Nisa İlkutlu, Işıl Çiftçi, Berra Uludere
Show PDF Cite
Social media and online review platforms provide valuable data for monitoring public perceptions of local environments, services, and everyday experiences. This study examines the sentiment orientation, temporal variation, and thematic structure of online comments related to Artvin through a multi-platform dataset collected from Facebook, Twitter/X, TikTok, Google Maps, Instagram, and YouTube. Artvin was selected as the research context because its natural and cultural attractions, mountainous geography, transportation and infrastructure conditions, environmental transformation processes, and everyday service experiences generate a multidimensional local digital discourse. After meaningless and duplicate content was removed, the comments were classified as negative, neutral, and positive using a BERT-based Turkish sentiment analysis model, and very short comments were additionally filtered at the topic modeling stage. Monthly sentiment scores were analyzed through time series and change point analyses, while BERTopic outputs were interpreted by grouping representative terms and comments under higher-order themes. The findings indicate that negative comments were more dominant in the digital discourse related to Artvin and that a statistically significant structural break occurred in February 2023. Topic modeling showed that negative discourse mainly focused on artificial or automatically generated content, transportation and infrastructure problems, local administration and political discussions, dam/expropriation processes, environmental impacts, and economic difficulties. Positive discourse clustered around gratitude, good wishes, natural beauty, scenery appreciation, and cultural values. The study shows that multi-platform social media analysis can support local public perception monitoring and contribute to data-driven regional decision-making processes.

More from our Archive