Thematic Role Based Generation of UML Models from Real World Requirements

Abstract

Model-driven development depends on good initial models. Creating these models by hand is a challenging task, because of complex specification documents and change requests. We propose a new internal representation based on thematic roles, especially designed for (but not limited to) requirements documents. The representation can be generated automatically out of annotated real-world specification text and can be used to generate UML models based on graph rewriting rules.

Synchronizing domain models with natural language specifications

“…Our implementation uses Gelhausen's [2] approach to derive a UML model from textual specifications. He uses a graph as an intermediate representation of the text.…”

“…This way, the domain model becomes a software model (first platform independent, then platform specific) and finally executable software. Gelhausen presented an approach to generate UML domain models directly from textual specifications [2], [3]. Following his approach, the first step of model driven development can be automated.…”

Enhancing model transformation synthesis using natural language processing

“…Since requirements statements are usually expressed in some form of natural language, the main challenge in the approaches has been the problem of linguistic analysis or natural language processing (NLP) [10,14,22]. The richness of natural language means that the same essential requirement may be expressed by diverse alternative texts.…”