Simplifying SQL with Natural Language

Overview

This project implements a text-to-SQL system that allows users to interact with databases using natural language queries. It leverages transformer-based models like BERT and T5 to convert user queries into accurate SQL statements. The system is integrated with Flask/FastAPI for real-time query execution.

Features

• Converts natural language queries into SQL statements.
• Built with transformer-based architectures like BERT and T5 for natural language understanding and SQL generation.
• Integrated with Flask/FastAPI for real-time query execution.
• Supports complex SQL queries and domain-specific use cases.
• Uses techniques like beam search and attention mechanisms to enhance performance and accuracy.
• Trained on the Spider dataset for a wide range of SQL queries.

Installation

Prerequisites

• Python 3.7+
• Flask or FastAPI
• PyTorch (for model implementation)
• Hugging Face Transformers library

Steps

1. Clone this repository:

   git clone https://github.com/yourusername/simplifying-sql-with-nl.git
   cd simplifying-sql-with-nl

2. Install dependencies:

   pip install -r requirements.txt

3. Run the server:

   For Flask:

   python app.py

   For FastAPI:

   uvicorn app:app --reload

Usage

1. Start the backend server (Flask or FastAPI).

2. Send a POST request to the /query endpoint with a natural language query:

   Example input:

   {
     "query": "What is the total revenue for the last quarter?"
   }

3. The response will be the corresponding SQL query:

   SELECT SUM(revenue) FROM sales WHERE quarter = 'last' AND year = CURRENT_YEAR;

Model Training

• The model is trained on the Spider dataset, which contains diverse SQL queries and schemas.
• Fine-tuned with domain-specific data to improve performance for complex queries.

Technologies Used

• Natural Language Processing (NLP): BERT, T5
• Backend Frameworks: Flask/FastAPI
• Modeling Framework: PyTorch, Hugging Face Transformers
• Dataset: Spider dataset

Contributing

Feel free to open issues or submit pull requests. Contributions are always welcome!

License

This project is licensed under the MIT License - see the LICENSE file for details.