Peringkasan Ulasan Produk Rumah Tangga Menggunakan BERtopic dan Maximal Marginal Relevance

Abstract

Product reviews are opinions or assessments from customers regarding products they have purchased and used. In the online shopping process, product reviews are crucial because they provide information that can convice potential customers. Household appliances rank fourth among the most searched and purchased products, leading to an increase in reviews, making it difficult for potential customers to obtain clear information. This research aims to summarize product reviews to convey key information concisely. The reviews were sourced exclusively from Shopee Indonesia online. IndoBERT was used for sentiment classification, followed by topic modeling using BERTopic to identify hidden topics. Summarization was done using MMR by selecting sentences that represent relevant topics. The research results show the f1-score values for positive reviews as follows : food processor: 0.229167, humidifier: 0.156566, vacuum cleaner: 0.363359. Meanwhile, for negative reviews, the f1-score values are : food processor: 0.804781, humidifier: 0.505703, vacuum cleaner: 0.659280.