Enhanced Ecommerce Search and Recommendation Engine

Links

Dataset

Google News Vectors

🚀 Overview

This project implements a hybrid ecommerce search and recommendation engine combining:

• BM25 for term-based ranking
• Fuzzy Matching for handling misspellings and approximate search
• Semantic Embeddings for context-aware search and recommendations
• Multilingual Search support for diverse user queries
• Spell Correction using SymSpell
• Fuzzy Relevance Scoring & Semantic Similarity to improve product retrieval accuracy

✨ Features

• Hybrid Search Engine: Combines BM25, fuzzy matching, and semantic search
• Multilingual Search: Users can search in different languages
• Spell Correction: Uses SymSpell to correct user queries
• Fuzzy Matching: Helps find relevant products even with typos
• Semantic Similarity: Embeddings help retrieve products based on meaning rather than just keywords
• Personalized Recommendations: Context-aware product recommendations based on search history

🛠 Technologies Used

• BM25 (for ranking documents)
• Fuzzy Matching (handling misspellings and variations)
• SymSpell (spell correction)
• Sentence Transformers (semantic embeddings)
• FastAPI / Flask (for API development)

🏆 Future Improvements

• 🔍 Improve search relevance with user behavior tracking
• 📊 Implement real-time analytics for search trends
• 🌍 Expand multilingual capabilities with additional embeddings