Automated Translation of UML Class Diagrams
Bessaac Gambe, Anitta Thomas · 2024
The translation of diagrams into natural language is useful in several scenarios, and one such scenario is improving the accessibility of diagrams for visually impaired users. Manual translation of diagrams into natural language can be cumbersome. Still, off-the-shelf tools that support automated translation of diagrams into natural languages are limited. In this research, a prototype is implemented and evaluated to realise automated translation of Unified Modeling Language (UML) class diagrams—a type of UML diagram used in computing—into English, specifically within the teaching and learning context. The prototype takes a Scalable Vector Graphics(SVG) image of the class diagram and produces two types of translation: the literal translation, which generates text describing the diagram in terms of the graphical objects and spatial relationships between these objects, and the second form of translation, which generates text describing the model represented in the diagram. A preliminary assessment of the prototype on a limited set of real-world class diagrams showed that it generates accurate node translations. The accuracy of link translations in class diagrams still needs significant improvements, which will be addressed in the next iteration of prototype development.