Requirements engineering is a big part of software engineering and consumes a lot of time. We propose a novel approach of automatically creating software domain models from textual requirements specifications using semantic annotation. Natural language processing (NLP) has progressed much in the last years and the usage of NLP tools for automatic annotation shows promising results. We use thematic roles (see Fillmore) to explicitly denote the semantic relations in a sentence.

Semantic annotations also maintain the connection between textual artifacts and their corresponding model elements. Therefore changes in the domain model can be fed back to the textual specification. Additionally, changes in the textual specification can be analyzed and their impact towards the software model can be assessed.

---

---