온라인 패션 리뷰 데이터를 활용한 속성 기반 감성 분석

다국어 초록 (Multilingual Abstract)

[Purpose] This study aims to classify categories such as women’s clothing, men’s clothing, fashion shoes, and accessories based on online fashion review data. It further seeks to analyze sentiments, such as positive or negative, based on these classifications to identify the importance of product attributes.<BR/>[Methodology] Online fashion review data were collected and categorized into women’s clothing, men’s clothing, fashion shoes, and accessories. Data preprocessing, such as removing duplicate sentences, correcting spacing errors, and eliminating stop words, was conducted. After that, we performed attribute －based labeling for each category and trained and analyzed it using the KcELECTRA model.<BR/>[Findings] The results of the attribute－based classification analysis show that the F1 score for the test data was highest in the order of women’s clothing, fashion shoes, accessories, and men’s clothing. For women’s clothing, men’s clothing, and fashion shoes, price and product composition were identified as important attributes, while for accessories, product composition and design were considered significant.<BR/>[Implications] This study contributes to a deeper understanding of customer experiences through the analysis of online fashion product reviews. Based on the findings, companies can develop strategies to enhance customer satisfaction. In particular, identifying significant attributes for each category through attribute－based sentiment analysis can serve as valuable data for product improvement and marketing strategy development.