Analyzing Public Perceptions of Sustainable Electric Mobility Using Natural Language Processing: The Case of TOGG on X

Social media platforms generate substantial volumes of unstructured textual data that can provide timely insights into public perceptions of emerging technologies. This study applies an integrated natural language processing framework to analyses Turkish-language discourse concerning TOGG, Türkiye’s domestic electric vehicle brand. The dataset comprises 10,584 posts published on the social media platform X between 1 January and 31 July 2026. The analytical framework combines BERT-based sentiment classification, bigram-based term frequency–inverse document frequency analysis, Latent Dirichlet Allocation (LDA) topic modelling, keyword co-occurrence analysis, and density visualization. The sentiment analysis classified 48.4% of the posts as positive, 30.4% as negative, and 21.2% as neutral. Sentiment-specific TF-IDF analysis showed that positive discourse emphasized domestic production, national brand identity, purchase interest, financing support, and charging infrastructure, whereas negative discourse concentrated on software problems, charging-related difficulties, spare-parts availability, and after-sales services. LDA identified three principal themes: domestic production, brand identity, and purchase intention; electric vehicle markets, sales, and taxation; and vehicle technology, charging infrastructure, and after-sales services. Keyword co-occurrence and density analyzes further revealed semantic relationships among the technological, economic, symbolic, and ownership-related dimensions of the discourse. The findings demonstrate how transformer-based sentiment classification, probabilistic topic modelling, and network-based text analysis can generate complementary insights from domain-specific social media data. The study provides an integrated analytical workflow for examining online discourse surrounding electric vehicles and other emerging technologies.