Work-In-Progress: Converting textual software engineering class diagram exercises to UML models

Abstract

Class diagram exercises are an important part in the development of software engineering students in higher computer science education. Generating textual exercises with sample solutions for such courses is time-consuming for educators, especially with multiple courses and different contexts. According to literature, the automatic generation of diagrams from structured text is possible. However, students often do not receive template based exercise texts but descriptions in natural language, which is still not a closed research topic. To address this problem, this paper discusses a model that analyses real exercise texts used for software engineering education, considers each individual sentences components and provides a class diagram. Due to the complexity of natural language, the model does not deliver perfect results so far, but is a great work in progress for the attempt to generate sample solutions for given exercise texts.