Authors: Ankit Kumar, Bhuvaneshwaran V, Maheshwaran C, Thulasidass M
Abstract: Modern e-commerce platforms contain millions of products, making it increasingly difficult for customers to discover products that match their individual requirements. Traditional recommendation systems primarily depend on collaborative filtering, product popularity, or keyword-based search. Although these techniques are useful, they may perform poorly when users have limited interaction histories, when products are newly introduced, or when customers express their requirements using natural language.This paper presents ShopMind AI, an intelligent e-commerce recommendation and semantic search framework designed to combine multiple sources of information within a unified per-sonalization architecture. The proposed framework integrates user preferences, historical interactions, product attributes, behavioral signals, semantic product representations, and con-textual search information. A hybrid recommendation strategy combines collaborative signals, content-based similarity, pop-ularity information, and semantic similarity. In addition, a transformer-based semantic representation layer enables users to search for products using natural-language descriptions rather than relying exclusively on exact keyword matching. The architecture also supports product comparison, price-related information, personalized recommendations, explain-able recommendation signals, and scalable vector-based re-trieval. The paper describes the system architecture, data-processing pipeline, recommendation model, semantic-search mechanism, evaluation methodology, scalability considerations, and potential future improvements. The proposed framework is intended to provide a practical foundation for developing more intelligent and user-centered e-commerce applications.
DOI: https://doi.org/10.5281/zenodo.23263005
