SQL Query Generation from Natural Language

Author: Yufan Zhang

Introduction

The project is about generating SQL queries from natural language questions. The dataset used in this project is the Flights dataset, which is a dataset of flights and airports. The dataset contains 25 tables, including airports, airlines, flights, etc. The goal of this project is to generate SQL queries from natural language questions in three approaches:

1. Fine-tuned a pre-trained T5 model on the Flights dataset.
2. Train a T5 model from scratch on the Flights dataset.
3. Prompting & In-context Learning with Gemma 1.1 2B model.

Project Structure

The project structure is as follows:

.
├── README.md
├── .env  # Environment file to store your HuggingFace Access Token
├── checkpoints
├── data
│   ├── alignment.txt
│   ├── dev.nl
│   ├── dev.sql
│   ├── flight_database.db
│   ├── flight_database.schema
│   ├── test.nl
│   ├── train.nl
│   └── train.sql
├── dataset
│   ├── __init__.py
│   └── sql_dataset.py
├── options
│   ├── __init__.py
│   ├── prompting_options.py
│   └── t5_options.py
├── prompting.py
├── requirements.txt
├── results
│   ├── records
│   └── queries
├── t5.py
└── utils
    ├── __init__.py
    ├── args.py
    ├── data.py
    ├── evaluation.py
    ├── prompting_utils.py
    └── t5_utils.py

How to Run

Development Environment Setup

To set up the development environment, you can run the following command:

conda create -n text2sql python=3.10
conda activate text2sql
pip install -r requirements.txt

Fine-tuned T5 Model

To run the fine-tuned T5 model, you can run the following command:

python3 t5.py --finetune

Train T5 Model from Scratch

To train the T5 model from scratch, you can run the following command:

python3 t5.py

Prompting & In-context Learning with Gemma 1.1 2B Model

To run the prompting & in-context learning with Gemma 1.1 2B model, you can run the following command:

python3 prompting.py

Acknowledgement

This project is a part of the course project for the course CS 5740: Natural Language Processing (2024 Spring) at Cornell Tech.