Termin in Kalender importieren: iCal (Download)
Vorträge
A Context-Based Approach for Change Propagation in Vitruvius
| Vortragende(r)
|
Josua Eyl
|
| Vortragstyp
|
Bachelorarbeit
|
| Betreuer(in)
|
Raziyeh Dehghani
|
| Vortragssprache
|
Englisch
|
| Vortragsmodus
|
in Präsenz
|
| Kurzfassung
|
Model-driven software development is the process of developing software on different abstraction levels. Therefore, various models describe one system from different viewpoints.
These models, besides describing different parts of the system, have overlaps. Due to these
overlaps, there needs to be an effort to make those models consistent with each other.
For this purpose, a framework called Vitruvius was developed as a view-based software
development tool. A domain-specific language called Reactions was developed to specify
how changes are propagated between models. This language facilitates the propagation of
changes made in one model to other corresponding models. To this moment, there is no
possibility that the Reactions language propagates its changes based on external context data.
In current propagation mechanisms, changes are applied uniformly, without considering
environmental or situational factors. However, in practice, the correct propagation often
depends on context, such as user roles, system state, or domain-specific constraints. Here,
the bachelor’s thesis aims to develop a context-aware change propagation mechanism for
software development.
Therefore, a context meta-model needs to be developed to describe how context can be
defined. The Reactions language must be extended to incorporate this context, enabling
propagation decisions to be made based on it. The developed approach will be evaluated
using a representative use case that demonstrates how context information can influence
and improve change propagation decisions.
|
Fine-tuning vs. Prompting: The Case of Requirements Classification
| Vortragende(r)
|
Maximilian Supp
|
| Vortragstyp
|
Bachelorarbeit
|
| Betreuer(in)
|
Dominik Fuchß
|
| Vortragssprache
|
Deutsch
|
| Vortragsmodus
|
in Präsenz
|
| Kurzfassung
|
Requirements classification is a key task in software development, in which requirements are grouped into different categories, such as functional and non-functional requirements. It helps to prioritize requirements and improve quality of the software. To automate this process, many machine learning approaches have been studied in recent years. With the rise of LLMs, fine-tuning and prompting were evaluated, showing promising results. However, current research does not provide guidelines for when to use which technique. This thesis aims to address this gap by systematically comparing state-of-the-art fine-tuning and prompting approaches on different application scenarios. For this purpose, three realworld scenarios are examined, covering the varying availability of sample data. Based on that, the goal is to investigate how much training data is required to achieve similar or better results with fine-tuning as with prompting techniques. When evaluating only on unseen projects, zero-shot prompting often achieves a similar or even better performance than fine-tuning. If requirements from the same project are available for training, 20 to 40 requirements are enough to outperform prompting with fine-tuning. Depending on the task and fine-tuning approach, labeling 20 requirements manually improved the performance by up to 25 percentage points in the best case and often by at least 10 percentage points. The evaluated approaches showed that fine-tuning always achieves a better performance than prompting if enough training data was provided. This thesis provides guidelines for requirements classification in practice and serves as a foundation for further research regarding this topic.
|
Integrating Human-Related Factors into Change Management
| Vortragende(r)
|
Manuel Odinius
|
| Vortragstyp
|
Bachelorarbeit
|
| Betreuer(in)
|
Raziyeh Dehghani
|
| Vortragssprache
|
Englisch
|
| Vortragsmodus
|
in Präsenz
|
| Kurzfassung
|
Systems are getting more and more complex. To handle the complexity of constructing them models are used.
The systems, such as the models evolve and adapt over time, creating a big information problem, to keep everyone up to date and distributed knowledge consistent.
To address this problem models are correlated with each other using correspondance models, that can be used to create (virtual) single underlying models, that aggregate all information at a single point.
However implementations for this strategy lack information on human-related aspects of these changes.
To solve this problem a metamodel, that can be used to add information on human-related aspects of changes will be presented.
The information on changes can be added through annotations, that preserve the information in different formats, some more machine readable, some more human readable.
This thesis, will discuss some human-related aspects, that are deemed relevant as annotations, to give additional context on changes, that otherwise might go missing.
Further this thesis will discuss a concrete implementation of using the preseneted annotations in the Vitruvius framework, together with a theoretical example showing a possible application of the implementation.
|
Neuen Vortrag erstellen
Hinweise
