Text-to-SQL

Abstract

This bachelor's thesis focuses on the analysis and comparison of the performance of smaller language models versus large language models (LLM) in the task of converting text to SQL queries. The main goal was to identify available smaller models, test them on specific datasets for SQL queries, and analyze their success on practical test questions from the subject of database systems. The results show significant differences in the success of these models, with their limitations in understanding and generating complex SQL queries being identified. Overall, this study contributes to a better understanding of the capabilities and limitations of available language models in the context of database interaction without the need for SQL knowledge.