리뷰에서의 특징 추출을 통한 특징 기반 뷰티 상품 검색 시스템

초록

In online shopping for beauty products, where preferences vary depending on individual skin types and tastes, reviews play a significant role in making purchase decisions. Recently, reviews have rapidly accumulated, significantly increasing the amount of information; however, this has made it difficult to find desired information. This study proposes a system that automatically extracts the key features of beauty products from reviews and visualizes the sentiment polarity of those features. The proposed system utilizes KoBERT, a Korean pre-trained language model, to perform binary sentiment classification and keyword extraction, enabling feature-based searches. The proposed system helps customers easily search for products using positively evaluated keywords and instantly identify the strengths and weaknesses of each product.

키워드