The master's thesis addresses the challenge of automatically generating graphical user interfaces based on user requirements or application descriptions. The study compares four prompting methods for large language models, using generated application descriptions from the Rico dataset as input and creating HTML/CSS prototypes for each method. The methods were evaluated automatically using quantitative metrics and qualitatively through an expert-oriented online questionnaire. The instruction-based prompting method proved to be the most effective, achieving high consistency with requirements, functionality, and aesthetics. The results also highlight the importance of the application description, as it significantly influences the performance of each method. The thesis provides practical guidelines for improving automated GUI prototyping in business applications.
|
