CONFERENCE / ICCAIS-2026

Online product reviews strongly influence customer buying decisions. With the rapid growth of e-commerce plat- forms, a huge number of reviews are generated every day, making it difficult for users to read and analyze them manually. These reviews are often unstructured and written in multiple languages, which further increases the complexity. To solve this problem, this paper proposes a deep learning-based multilingual product review sentiment analysis system using DistilBERT. The system accepts two types of inputs: manually pasted customer reviews and product links, from which relevant reviews are analyzed. It automatically detects the language of each review and translates non-English content into English to maintain consistent analysis. A robust preprocessing process removes unnecessary text and noise, improving overall accuracy. The DistilBERT model classifies reviews into Positive, Neutral, and Negative sentiments and provides a clear percentage distribution along with a final product recommendation verdict. Batch processing is used to reduce response time and improve performance. Experimental results show that the proposed system delivers accurate and context- aware sentiment analysis, outperforming traditional machine learning approaches and making it suitable for real- world e- commerce applications.