# Next-Generation Database Interfaces: A Survey of LLM-based Text-to-SQL

**Authors:** Zijin Hong, Zheng Yuan, Qinggang Zhang, Hao Chen, Junnan Dong, Feiran Huang, Xiao Huang  
**arXiv:** [2406.08426](https://arxiv.org/abs/2406.08426)  
**Date:** 2024-06-12  
**Summary page:** https://www.summarizepaper.com/en/paper/2406.08426v5/

## Key points

- Authors Zijin Hong, Zheng Yuan, Qinggang Zhang, Hao Chen, Junnan Dong, Feiran Huang, and Xiao Huang focus on challenges and advancements in generating accurate SQL from natural language questions
- Traditional text-to-SQL systems combine human engineering with deep neural networks for progress
- Pre-trained language models (PLMs) with limited parameter sizes often produce incorrect SQL queries as databases and user queries become more intricate
- Large language models (LLMs) are a promising solution due to enhanced capabilities in natural language understanding as model scale increases
- LLM-based solutions present unique opportunities for improving text-to-SQL research
- The paper provides a comprehensive review of existing LLM-based text-to-SQL studies, technical challenges involved, evolutionary process in the field, datasets and metrics for evaluation
- Recent advances in LLM-based text-to-SQL approaches are systematically analyzed with benefits and potential drawbacks highlighted
- Key findings are summarized along with remaining challenges discussed; future research directions suggested to enhance accuracy and efficiency of generating SQL queries from natural language inputs

## Summary

In their paper titled "Next-Generation Database Interfaces: A Survey of LLM-based Text-to-SQL," authors Zijin Hong, Zheng Yuan, Qinggang Zhang, Hao Chen, Junnan Dong, Feiran Huang, and Xiao Huang delve into the challenges and advancements in generating accurate SQL from users' natural language questions. The complexity arises from the need to understand user queries, comprehend database schemas, and generate SQL queries effectively. Traditional text-to-SQL systems have made significant progress by combining human engineering with deep neural networks. However, as databases and user queries become more intricate, pre-trained language models (PLMs) with limited parameter sizes often produce incorrect SQL queries. This limitation necessitates the development of more sophisticated optimization methods tailored to address these challenges. Large language models (LLMs) have emerged as a promising solution due to their enhanced capabilities in natural language understanding as model scale increases. The integration of LLM-based solutions presents unique opportunities for improving text-to-SQL research. In their survey, the authors provide a comprehensive review of existing LLM-based text-to-SQL studies. They offer an overview of the technical challenges involved in text-to-SQL processes and discuss the evolutionary process of this field. Additionally, they introduce datasets and metrics designed to evaluate the performance of text-to-SQL systems. The paper systematically analyzes recent advances in LLM-based text-to-SQL approaches, highlighting the benefits and potential drawbacks of these methods. The authors also summarize key findings and discuss remaining challenges in the field. They suggest future research directions that could further enhance the accuracy and efficiency of generating SQL queries from natural language inputs. Overall, this survey contributes valuable insights to the ongoing efforts to improve text-to-SQL systems using large language models.