<![CDATA[In e-commerce, product reviews play a crucial role in influencing potential buyers by sharing user experiences and assessing product quality. This is especially important for beauty products, where poor quality can lead to physical harm. Reviews also help increase consumer interest in purchasing. Previous research has shown that product reviews differ in various aspects and content, making it challenging for consumers to quickly analyze them from multiple perspectives. This study applies aspect-based sentiment analysis to beauty product reviews on the Female Daily Network using a combination of BERT and LSTM. The goal is to provide more precise sentiment classification across different aspects, aiding consumers in selecting the best products. Several evaluation scenarios were conducted to assess different aspects of product reviews, including price, packaging, staying power, moisture, and aroma. The F-1 score revealed that the price aspect achieved the highest performance, reaching 100% in a 90%:10% test data scenario. However, the aroma aspect proved the most challenging to analyze, indicating that the model struggles to capture features related to scent effectively under the given evaluation setup.]]>
Copyrights © 2025
