Talk to Your Data: A Seq2seq Model for Transforming Natural Language Queries into SQL Statements

Abstract

Accessing and interacting with relational databases using Structured Query Language (SQL) can be challenging for users without technical expertise, posing a barrier to effective data management and analysis. This challenge highlights the need for solutions that bridge the gap between natural language and SQL, enabling more intuitive data queries. We propose a novel supervised deep learning model based on the BART-large-CNN architecture for transforming natural language queries into SQL statements to address this issue. This seq2seq model facilitates smoother and more efficient database interactions by allowing users to query data using everyday language. Our methodology includes training the model on a comprehensive dataset of paired natural language queries and corresponding SQL statements. We implement a custom training and evaluation pipeline for performance evaluation using the Hugging Face Transformers library and ROUGE metrics. The experimental results demonstrate that our model accurately translates various natural language queries into SQL statements, offering a robust solution for enhancing database accessibility and user experience.