u1316/Perfume_Recommendation_System_By_UM_IE_GS

An AI-powered fragrance recommendation system using semantic embeddings and advanced machine learning techniques.

🚀 Features

• —24,000+ Perfumes: Comprehensive database from Fragrantica.com
• —Semantic Search: Natural language perfume descriptions
• —Advanced Filtering: Brand, gender, and diversity controls
• —MMR Algorithm: Maximal Marginal Relevance for diverse recommendations
• —Professional UI: Clean, responsive Gradio interface
• —Real-time Results: Instant recommendations using pre-computed embeddings

🛠️ Setup for HuggingFace Spaces

Required Files

File Structure

your-hf-space/
├── app.py
├── requirements.txt  
├── fra_cleaned.csv
└── fra_cleaned_embeddings_structured.npy

📊 Technical Details

• —Model: all-mpnet-base-v2 (768-dimensional embeddings)
• —Algorithm: Maximal Marginal Relevance (MMR) for diversity
• —Search: Cosine similarity in semantic space
• —Framework: Gradio 4.x with custom CSS styling

🎯 Usage

1. 1.Text Search: Describe your ideal perfume in natural language
2. 2.Perfume Lookup: Enter exact perfume names for similar recommendations
3. 3.Filters: Narrow results by brand, gender, or other criteria
4. 4.Diversity Control: Adjust λ parameter for similarity vs diversity balance

Example Queries

• —"Fresh citrus with vanilla base"
• —"Romantic rose perfume for evening"
• —"Woody oriental with amber"
• —"Tom Ford Black Orchid" (exact name lookup)

📁 Data Sources

• —Dataset: Fragrantica.com perfume database
• —Features: Notes, accords, brands, perfumers, ratings
• —Preprocessing: Text cleaning, standardization, embedding generation

Built with ❤️ using HuggingFace, Gradio