Semi-automated Refinement of Semantic Models
The dissertation by Dr.-Ing. Alexander Paulus deals with the annotation of data sets using semantic technologies. In it, he describes a new approach with a focus on humans in the annotation process and presents support systems based on machine learning methods.
We asked Alexander about his dissertation:
What was the context of your dissertation? What projects or other factors particularly influenced your dissertation?
My dissertation is the result of my research at RWTH Aachen University and the University of Wuppertal. My colleagues and I recognized early on that the use of semantic technologies to describe data depends largely on the usability of the available applications. Especially for data exchange in new approaches such as dataspaces, the semantic description of data is an essential component for improving the findability and uniform interpretation of data sets. Due to the complexity of the task at hand, i.e. the creation of semantic models for different data sets, many users from the industrial environment are overwhelmed by the manual creation of such a model.
In the ESKAPE research project, we therefore developed an initial prototype for a new platform for semantic data annotation that is limited to the essential functionalities. This was then further developed in the bergisch.smart_mobility project at the BUW and published as an independent tool called “PLASMA” at the end of the project.
How does your work contribute to the field of research?
As part of my research, in addition to the continuous improvement of PLASMA, I have been working on the development of assistance systems. These aim to support human modelers in their work. In addition to the already existing fully automated approaches, my work and the associated publications also describe a new approach that actively involves the modeler in the process. To support this, ML-generated recommendations are intended to reduce the workload for the modeler. At the same time, we promote the enrichment of semantic models with information that is only known to the modeler but is essential, such as reference values or measurement inaccuracies, through hints in the PLASMA modeling interface.
What does the future hold for you and the topic?
The use of semantic technologies continues to advance. More and more people will have to provide data in the future, for example for digital product passports. If this data is then to be exchanged in semantically annotated form within the framework of dataspaces, employees in SMEs will also have no choice but to create semantic models. My research should also contribute to the usability of data in the future. PLASMA and the associated recommendation systems are being improved in ongoing projects and integrated into more existing systems.