Analyzing the Perceptual Structure of Scented Cosmetics via BERTopic and Ego-network Analysis

This study explored consumer perceptions of scented cosmetics, focusing on their role in self-expression and personal image construction. Consumer review data were collected from the fragrance and diffuser category of Olive Young, a leading domestic health and beauty (H&B) retailer in South Korea, using Python-based web crawling. The collection period spanned from January 2020 to October 2025, targeting the top 100 products by sales ranking. After removing duplicates and low-quality entries from 44,991 raw records, 44,815 valid reviews were retained for analysis. Text mining methods—including TF-IDF keyword analysis, BERTopic modeling, and ego-network analysis—were applied to identify key consumer perceptions and their structural relationships. The analysis revealed that consumer perception is organized around three dimensions: Functional Attributes, Purchase Context, and Scent Profile. A total of 19 meaningful topics were identified through BERTopic, and ego-network analysis identified ‘Longevity’ as a Key Evaluation Criterion, which consumers perceive not merely as duration but as a sensory experience of lingering scent on skin, significantly associated with satisfaction. These findings offer foundational data for product planning and communication strategies. Academically, this study visualizes latent consumer evaluation structures using text mining; practically, it underscores the importance of product development centered on residual scent and strategies tailored to gift-giving contexts. Limitations include the single-channel scope of data, and future research should expand to multiple platforms and examine causal relationships more rigorously.