This study applies aspect-based sentiment analysis to Vietnamese lipstick reviews collected from an e-commerce platform. Utilizing a dataset of 16,227 reviews, we leverage large language models (LLMs), such as GPT, with zero-shot, one-shot, and five-shot inference strategies to extract product aspects and classify sentiments. The experimental results demonstrate that the five-shot approach achieves high accuracy (ranging from 92.86% to 100%) across various aspects, including price, packaging, texture, scent, and longevity. These findings highlight the potential of LLMs in processing Vietnamese language, particularly for product review analysis in the context of e-commerce.