Flipkart Product Semantic Search with Transformers

This project implements a semantic search system for e-commerce products using pre-trained transformer embeddings.
The system allows searching Flipkart product descriptions based on the meaning of user queries, not just keyword matching.

Features

• Uses the Flipkart products dataset from Kaggle.
• Employs the sentence-transformers library to generate embeddings.
• Performs semantic similarity search using cosine similarity.
• Returns the top-k most relevant products for any input query.

Dataset

The dataset used in this project is the Flipkart products dataset from Kaggle.

Requirements

• Python 3.7+
• pandas
• sentence-transformers
• torch

You can install the dependencies with:

pip install pandas sentence-transformers torch

Usage

1. Load the dataset.
2. Clean and preprocess product data.
3. Encode product descriptions using a pre-trained SentenceTransformer model.
4. Use the search_products(query, top_k) function to find products semantically similar to the query.

Example:

search_products("wireless bluetooth headphones", top_k=5)

Next Steps

• Integrate with a web app using Gradio or Streamlit for an interactive search experience.
• Fine-tune transformer models using e-commerce data for enhanced accuracy.
• Expand to multi-modal search, including images and reviews.

License

This project is licensed under the MIT License.

---

Made with ❤️ by Özlem