Product Review & Sentiment Analysis Using Natural Language Processing: RSR ANALYZE

This project started from a simple observation: most of the best NLP models in research are not actually usable by most real-world systems. Our goal was to close that gap — at least partially — by building a system that is accurate enough to be useful, fast enough to be practical, and flexible enough to handle the kind of messy multilingual text that real Indian users produce every day. RSR Analyze demonstrates that traditional ML models, when combined with thoughtful preprocessing and feature engineering, can still be highly competitive. LinearSVC with TF-IDF + Bigram features achieves 88% accuracy at under 5ms per inference — a combination that transformers simply cannot match at that compute level. The multilingual evaluation results also show that handling Hinglish is genuinely feasible without a heavy multilingual model, provided the preprocessing is done right. We see this as meaningful progress toward making sentiment analysis tools accessible for Indian-language e-commerce platforms. That said, we are aware of the trade-offs. Deep context understanding, sarcasm detection, and aspect-level sentiment analysis remain open challenges. Our future work will explore hybrid architectures — combining a fast TF-IDF backbone with lightweight contextual modules — to see if we can push accuracy higher without sacrificing the efficiency that makes this system practical. As a final thought: building this project gave us hands-on experience with the full ML pipeline, from raw data to a deployed API. That practical experience was, in many ways, the most valuable outcome.

[1]       Pang, B., Lee, L., & Vaithyanathan, S. (2002). Thumbs up? Sentiment classification using machine learning techniques. Proceedings of EMNLP 2002, 79–86. [2]        Joshi, A., Prabhu, A., Shrivastava, M., & Varma, V. (2016). Towards sub-word level compositions for sentiment analysis of Hindi-English code mixed data. Proceedings of COLING 2016, 2482–2491. [3]       Yadav, S., Chakravarthi, B. R., Bansal, P., & Kumar, V. (2020). Sentiment analysis of Hindi-English code mixed data. Proceedings of the Workshop on Tools & Resources to Empower People with Reading Difficulties (READ), 37–44. [4]        Pedregosa, F., et al. (2011). Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12, 2825–2830. [5]        Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. Proceedings of NAACL-HLT 2019, 4171–4186.