Forschungsthemen
[BA] Extending Artemis With a Rule-Based Approach for Automatically Assessing Modeling Tasks
E-Learning systems have long since become an integral part of school and academic education. These systems offer a large number of users the opportunity to work on and submit exercises at the same time. In the field of computer science, these tasks are mainly programming tasks, but also often modeling tasks. To provide consistent yet individual feedback despite the constantly increasing number of users, systems for automatically assessing and correcting submissions for the above-mentioned task types are used. A current example of such a system is Artemis, an interactive open-source learning platform for individual feedback. In addition to programming tasks, multiple-choice, and free-text tasks, it also offers modeling tasks for parts of the Unified Modeling Language and other frequently used model types. There are several approaches for the automatic evaluation of modeling tasks, although unfortunately only a few types of systems are used in everyday university life. Rule-based approaches and evaluation by machine learning seem to be the most common approaches at the moment. Unfortunately, there is currently no system that offers both approaches to its users, despite the fact that both approaches have different advantages and disadvantages. The goal of this work is to extend Artemis with a rule-based approach to make the above approaches available to instructors. Artemis currently supports only semi-automatic, similaritybased machine-learning-based correction. In the first step, both approaches need to be compared. Based on this comparison, a benefit analysis for future teachers will be derived. Afterward, a modular design for integrating the rule-base extension into the system shall be created. Since Artemis is an actively maintained open-source project with regular updates, the design needs to be as non-invasive as possible to ensure compatibility with new releases. Finally, the design is to be implemented in a prototype. For the evaluation of the functionality of the design and the results of the theoretical review, a comparison between the new rule-based rating system and the existing similarity-based system will be performed.
Betreuer: Markus Hamann-:#-#:- Andreas Domanowski