Natural Language Verbalization of SQL Query Results for Human-AI Interaction with Databases

Abstract

AI-native systems are becoming core components of modern data-centric infrastructures, where artificial intelligence is embedded directly into data access, processing, and interpretation workflows. Despite advances in large language models, most research focuses on natural language to SQL generation, while the inverse problem - semantic interpretation and verbalization of SQL query results - remains underexplored, limiting accessibility for non-technical users. This paper proposes an AI-native framework for natural language verbalization of SQL query results in multimodal human-AI interaction systems. The system integrates schemaaware reasoning, query validation, metadata enrichment, and large language model-based semantic interpretation, with optional voice-enabled interaction via ASR and TTS technologies. The system is implemented as a Python desktop application using PyQt6 and SQLite. Evaluation on 550 SQL queries demonstrates improved interpretability and context-aware explanations, enhancing human-AI interaction in database systems.