ウエダ　マユミ   Ueda Mayumi   上田　真由美    所属   追手門学院大学  経営学部 経営学科    職種   教授
発表年月日	2025/10/23
発表テーマ	Considering the Expansion of Product Feature Analysis Methods' Applicability in Review Videos
会議名	The 21st International Conference on Advanced Data Mining and Applications 2025 (ADMA 2025)
学会区分	国際学会
発表形式	ポスター
単独共同区分	共同
国名	日本
開催地名	Kyoto
開催期間	2025/10/22～2025/10/25
発表者・共同発表者	Aiko Kobayashi, Fumiya Yamaguchi, Asaka Lan Cheng, Mayumi Ueda and Shinsuke Nakajima
概要	To enhance the efficiency of acquiring product information in online shopping environments, this study addresses the challenge of analyzing product features from user-generated review videos. We propose a method for automatically extracting product feature information from video subtitles and user comments, employing a document classification approach based on a pre-trained BERT model. Recognizing that the applicability of BERT-based classifiers in the context of review video analysis has not been thoroughly examined, this research focuses on the cooking appliances domain as a representative case study. We investigate strategies for constructing robust, product-specific classifiers capable of accurately categorizing textual content across a wide variety of product types within this category. The proposed approach aims to improve both the accuracy and generalizability of product feature extraction, thereby contributing to the development of more effective review video analysis systems for e-commerce applications.