Example-Driven User Intent Discovery: Empowering Users to Cross the SQL Barrier Through Query by Example

Abstract

Relational databases require complex SQL queries for data retrieval, creating a significant barrier for nontechnical users. This paper implements an Example-Driven User Intent Discovery system based on the SQUID approach using the Django web framework and IMDb dataset. The architecture comprises three layers: raw data ingestion, normalized relational schema, and an abduction-ready database containing derived semantic relationships. Users provide example entities (such as actors), and the system discovers shared semantic properties, applies abductive reasoning to infer query intent, and retrieves relevant results automatically. Experimental evaluation demonstrates 85% intent discovery accuracy and 78% result precision, with users reporting 4.1/5 satisfaction. The system successfully bridges the gap between complex database querying and user-friendly data exploration.