Incorporating User Inputs for Improved JSON Schema Inference

JSON Schemas, as descriptive JSON files, define the expected structure of other JSON data, serving as a valuable resource for both developers and programs. They play a crucial role in data validation, testing, and maintaining data consistency. Since creating JSON Schemas can be challenging, it is common to derive schemas from example data. In this research, we focus on the introduction of user inputs during the inference process with the goal of reducing ambiguity and allow an algorithm to make, otherwise inconclusive, speculations from the sample data. We describe numerous strategies for utilising JSON Schema features based on sample JSON files and how they were implemented into a Kotlin program. We evaluated the program on five distinct real world sample JSON datasets from which the results showed it is able to infer complex patterns.