An Agent-Based Framework for Developing Multiple Conceptual Models from Extended User Stories

Researchers have proposed that generating conceptual models automatically from user stories might be useful for agile software development. However, user stories in their current form have insufficient information to generate conceptual models that show dependencies between user stories, such as process models and state machine diagrams. Using Behavior-Driven Development (BDD) scenarios, we extend the current format of the user stories and demonstrate that such extended user stories can generate multiple conceptual models, including models that help understanding and analyzing dependencies between user stories. Using an agent-based framework, we first map the concepts of the extended user stories to agent concepts. Then using this concept mapping, we standardize the structured representation of user stories extended with BDD scenarios. We then propose mapping rules and algorithms that can be applied to the structured representation to create multiple inter-related conceptual models: use case models, domain models, process models, and state machines. We demonstrate using a proof-of-concept example that these mapping rules and algorithms can be used to generate such conceptual models, given that certain assumptions hold. Our research sets up the path to instantiate the proposed algorithms with programming codes based on Natural Language Processing (NLP) techniques to develop tools that can automatically generate multiple conceptual models from the extended user stories