A Hybrid Explainable Product Recommendation System Using TF-IDF and SBERT

Author

Sa Kaung Min Htet, Hmway Hmway Tar

Abstract

 Traditional content-based product recommendation systems often rely on lexical matching techniques such as TF-IDF and cosine similarity. While these methods provide transparent and interpretable recommendations, they struggle with the vocabulary mismatch problem—failing to capture conceptual relevance when queries and product descriptions use different wording. To overcome this limitation, the system proposes a hybrid explainable recommendation framework that leverages Sentence-BERT (SBERT) for deep semantic encoding. By integrating SBERT, the system transforms query and product text into dense contextual embedding vectors, effectively capturing fine-grained semantic intent, synonyms, and implicit conceptual alignment that lexical methods miss. The proposed architecture represents each product through a dual-representation strategy: TF-IDF for explicit keyword precision and SBERT for rich contextual semantics. Normalized scores from both representations are dynamically fused via a weighted hybrid ranking function, constrained by structured metadata parameters (category, brand, and price). Experimental evaluation demonstrates that incorporating SBERT significantly enhances recall and semantic search capabilities while retaining the system's core transparency.

Keywords

content-based filtering, recommender systems, TF-IDF, SBERT

Full Text:

Download Paper PDF

References

[1] G. Salton and C. Buckley, "Term-weighting approaches in automatic text retrieval," Inf. Process. Manage., vol. 24, no. 5, pp. 513–523, 1988.

[2] G. Salton, A. Wong, and C. S. Yang, "A vector space model for automatic indexing," Commun. ACM, vol. 18, no. 11, pp. 613–620, 1975.

[3] F. Ricci, L. Rokach, and B. Shapira, Eds., Recommender Systems Handbook, 2nd ed. New York, NY, USA: Springer, 2015.

[4] C. D. Manning, P. Raghavan, and H. Schütze, Introduction to Information Retrieval. Cambridge, U.K.: Cambridge Univ. Press, 2008.

[5] F. Pedregosa et al., "Scikit-learn: Machine learning in Python," J. Mach. Learn. Res., vol. 12, pp. 2825–2830, 2011.

[6] J. Weizenbaum, "ELIZA — a computer program for the study of natural language communication between man and machine," Commun. ACM, vol. 9, no. 1, pp. 36–45, 1966.

[7] J. L. Herlocker, J. A. Konstan, L. G. Terveen, and J. T. Riedl, "Evaluating collaborative filtering recommender systems," ACM Trans. Inf. Syst., vol. 22, no. 1, pp. 5–53, 2004.

[8] W. B. Croft, D. Metzler, and T. Strohman, Search Engines: Information Retrieval in Practice. Boston, MA, USA: Addison-Wesley, 2010.

[9] K. Järvelin and J. Kekäläinen, "Cumulated gain-based evaluation of IR techniques," ACM Trans. Inf. Syst., vol. 20, no. 4, pp. 422–446, 2002.

[10] The pandas development team, "pandas-dev/pandas: Pandas," Zenodo, 2020.

[11] E. Rahm and H. H. Do, "Data cleaning: Problems and current approaches," IEEE Data Eng. Bull., vol. 23, no. 4, pp. 3–13, 2000.

[12] Hmway Hmway Tar,” Incorporating Ontology-based SBERT for Text Documents Clustering” 2025 6th International Conference on Advanced Information Technologies (ICAIT)

Share your valuable work from Social Media Buttons