The rapid growth of internet usage has significantly transformed consumer behavior, particularly in the skincare industry, where online reviews play a crucial role in influencing purchasing decisions. This study focuses on sentiment analysis of beauty product reviews from the Female Daily website, specifically analyzing four aspects: scent, packaging, price, and product quality. The scope includes both translated and untranslated reviews to examine the impact of language preprocessing on sentiment prediction. The problem addressed is the lack of effective sentiment classification methods tailored to Indonesian beauty product reviews. The objective is to evaluate the performance of Naive Bayes variants (GaussianNB, MultinomialNB, and BernoulliNB) combined with Word2Vec embedding models (Skip- Gram and CBOW) to classify sentiments accurately. The novelty of this research lies in the comparative analysis of translation impact and Word2Vec model variants on sentiment classification performance. The experimental results show that BernoulliNB with Skip-Gram achieves the highest performance on translated data, with Accuracy of 93%, Precision of 91%, Recall of 93%, and F1-Score of 90%, while GaussianNB with Skip-Gram performs best on untranslated data, achieving Accuracy of 83% and F1-Score of 87%. Across all experiments, the Skip-Gram model consistently outperforms CBOW by up to 14% in Recall and 11% in F1-Score, indicating its superiority in capturing contextual sentiment. These findings emphasize the importance of selecting appropriate models based on dataset characteristics. Future work may explore hybrid deep learning methods and larger, more diverse datasets to further improve sentiment analysis accuracy in the beauty product domain. These insights are valuable for platforms handling multilingual or translated reviews, ensuring sentiment classification systems remain accurate across diverse linguistic inputs.