Abstract
The rapid growth of global e-commerce has generated large volumes of customer reviews written in multiple
languages, making multilingual feedback analysis increasingly important. Traditional sentiment analysis
systems are often optimized for English and may lose contextual information when processing reviews in other
languages. This paper presents a multilingual sentiment analysis system for e-commerce platforms that
automatically detects the language of customer reviews and classifies them into Positive, Neutral, and Negative
categories. The proposed methodology combines language detection using LangID, translation of non-English
reviews using Deep Translator, contextual representation using multilingual BERT (mBERT), and Random
Forest for final sentiment classification. TextBlob is additionally used to calculate polarity and subjectivity
scores. A Flask-based web application integrates these components into a unified workflow that displays the
original review, detected language, translated text, predicted sentiment, confidence, polarity, and subjectivity.
The system was evaluated on 300 test instances, achieving F1-scores of 0.92 for both Positive and Negative
classes and 0.90 for the Neutral class. The results demonstrate that the proposed hybrid approach provides
reliable multilingual sentiment predictions and supports practical e-commerce customer-feedback analysis.
References
1) Liu, B., Sentiment Analysis and Opinion Mining, Morgan & Claypool Publishers.
2) Jurafsky, D., and Martin, J. H., Speech and Language Processing, Pearson.
3) Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K., “BERT: Pre-training of Deep Bidirectional
Transformers for Language Understanding.”
4) Mąbökela, S., Celik, T., and Rabie, T., “Multilingual Sentiment Analysis for Under-Resourced Languages.”
5) Breiman, L., “Random Forests,” Machine Learning, Vol. 45, pp. 5–32.
6) Bahrawi, N., “Sentiment Analysis Using Random Forest Algorithm.”
7) Junczys-Dowmunt, M., et al., “Marian: Fast Neural Machine Translation in C++.”