DoME: An Architecture for Domain Model Evolution at Runtime Using NLP.

In traditional information systems, domain models are represented as database tables with attributes and relationships. Changes in the domain models exist due to system evolution and the emergence of new requirements. In these applications, domain models evolve using CRUD operations requested by users. However, it is necessary to support changes in domain models during the applications’ runtime when new (unforeseen) situations may occur. This work presents an architecture called DoME, which relies on natural language processing (NLP) to allow users to trigger changes in the domain models and self-adaptation techniques to update the models at runtime. It is instantiated in a concrete architecture using a chatbot in Telegram and Transformers Libraries for NLP. The architecture has been preliminary evaluated regarding its assertiveness and user satisfaction, resulting in an 82.55% hit rate and confirming that NL provides good usability and facilitates data manipulation.