From Prose to Prototype: Synthesising Executable UML Models from Natural Language

This paper presents a vision for a development tool that provides automated support for synthesising UML models from requirements text expressed in natural language. This approach aims to simplify the process of analysis - i.e. moving from written (and spoken) descriptions of the functionality of a system and a domain to an executable specification of that system. The contribution focuses on the AI techniques used to transform natural language into structural and dynamic UML models. Moreover, we envision a 'human-in-the-loop' approach where an interactive conversational component is used based on machine learning of the system under construction and corpora of external natural language texts and UML models. To illustrate the approach, we present a tool prototype. As a scoping, this approach targets data-intensive systems rather than control-intensive (embedded) systems.