@

초록

Large language models (LLMs) are revolutionizing technologies and industries, excelling at processing and analyzing massive amounts of textual data. Against this backdrop, this study explores consumer perceptions and evaluations of fashion brandsand products by applying an LLM to analyze actual consumer comments. Moving beyond traditional text-mining approaches such as LDA and TF-IDF, this research employs OpenAI’s text-embedding-3-small model to generate semantic embeddings of consumer reviews collected from a leading Korean athleisure brand. Using K-means clustering and silhouette scores, this study identifies seven meaningful topics: (1) product appreciation, (2) brand impressions, (3) product sizing, (4) product color, (5) ease of use for exercise, (6) wearing comfort, and (7) various overall evaluations. These topics reveal that consumers’ evaluations extend beyond purely functional benefits to include subtle emotional and aesthetic responses, suggesting the salience of psychological and experiential factors in fashion purchases. This study demonstrates that LLM-based embedding techniques can effectively structure qualitative consumer opinions into interpretable themes, offering both methodological advancements and theoretical insights, and it provides practical implications for fashion brands seeking to better understand and respond to the nuanced drivers of consumer satisfaction.

키워드