Exploring Database Normalization Effects on SQL Generation
Not on the map. Why?
Abstract (arXiv)
Schema design, particularly normalization, is a critical yet often overlooked factor in natural language to SQL (NL2SQL) systems. Most prior research evaluates models on fixed schemas, overlooking the influence of design on performance. We present the first systematic study of schema normalization's impact, evaluating eight leading large language models on synthetic and real-world datasets with varied normalization levels. We construct controlled synthetic datasets with formal normalization (1NF-3NF) and real academic paper datasets with practical schemes. Our results show that denormalized schemas offer high accuracy on simple retrieval queries, even with cost-effective models in zero-shot settings. In contrast, normalized schemas (2NF/3NF) introduce challenges such as errors in base table selection and join type prediction; however, these issues are substantially mitigated by providing few-shot examples. For aggregation queries, normalized schemas yielded better performance, mainly due to their robustness against the data duplication and NULL value issues that cause errors in denormalized schemas. These findings suggest that the optimal schema design for NL2SQL applications depends on the types of queries to be supported. Our study demonstrates the importance of considering schema design when developing NL2SQL interfaces and integrating adaptive schema selection for real-world scenarios.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 25 of its 47 references that are in the map. First posted on arXiv 2 Oct 2025.
Builds on
25 of 47 references.
References · 25
- 2018Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Taskcited by 1,120
- 2017Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learningcited by 679
- 2023Can LLM Already Serve as A Database Interface? A BIg Bench for Large-Scale Database Grounded Text-to-SQLscited by 667
- 2023DIN-SQL: Decomposed In-Context Learning of Text-to-SQL with Self-Correctioncited by 465
- 2023Text-to-SQL Empowered by Large Language Models: A Benchmark Evaluationcited by 378
- 2023StructGPT: A General Framework for Large Language Model to Reason over Structured Datacited by 339
- 2024Spider 2.0: Evaluating Language Models on Real-World Enterprise Text-to-SQL Workflowscited by 223
- 2023RESDSQL: Decoupling Schema Linking and Skeleton Parsing for Text-to-SQLcited by 191
- 2024CodeS: Towards Building Open-source Language Models for Text-to-SQLcited by 183
- 2023C3: Zero-shot Text-to-SQL with ChatGPTcited by 159
- 2023MAC-SQL: A Multi-Agent Collaborative Framework for Text-to-SQLcited by 158
- 2020Semantic Evaluation for Text-to-SQL with Distilled Test Suitescited by 123
13 more
- 2024Next-Generation Database Interfaces: A Survey of LLM-based Text-to-SQLcited by 118
- 2021KaggleDBQA: Realistic Evaluation of Text-to-SQL Parserscited by 105
- 2024MCS-SQL: Leveraging Multiple Prompts and Multiple-Choice Selection For Text-to-SQL Generationcited by 103
- 2024The Dawn of Natural Language to SQL: Are We Fully Ready?cited by 78
- 2023ACT-SQL: In-Context Learning for Text-to-SQL with Automatically-Generated Chain-of-Thoughtcited by 65
- 2024Before Generation, Align it! A Novel and Effective Strategy for Mitigating Hallucinations in Text-to-SQL Generationcited by 49
- 2024DTS-SQL: Decomposed Text-to-SQL with Small Large Language Modelscited by 45
- 2023ScienceBenchmark: A Complex Real-World Benchmark for Evaluating Natural Language to SQL Systemscited by 42
- 2023Symbol-LLM: Towards Foundational Symbol-centric Interface For Large Language Modelscited by 38
- 2024PURPLE: Making a Large Language Model a Better SQL Writercited by 34
- 2022Semantic Enhanced Text-to-SQL Parsing via Iteratively Learning Schema Linking Graphcited by 22
- 2023Interactive Text-to-SQL Generation via Editable Step-by-Step Explanationscited by 19
- 2024Improving Retrieval-augmented Text-to-SQL with AST-based Ranking and Schema Pruningcited by 16
Built on by
Citing papers per year: 2025: 2.
2 papers.