Discovering entity inheritance relationships in document stores

Abstract

While declaring schemas is not required in most NoSQL stores, explicit schemas are crucial to provide essential capabilities in the development of database applications such as schema visualization, code generation, and data quality improvement. Therefore, some published works and data modeling tools have recently presented reverse-engineering approaches to extract NoSQL schemas from data or code. None of them have addressed the inference of inheritance relationships in which a parent entity type is specialized in one or more child entity types. In this paper, we present a strategy to discover entity inheritance relationships from the set of existing structural variations of an entity type. The algorithm proposed also identifies the discriminator field, if it exists. A validation process has been defined, which is applied to three real datasets.