Speaker
Description
Abstract:
Implementation of FAIR principles based on a knowledge graph (KG) is an attractive approach that is being pursued by many NFDI consortia, as KGs can naturally represent entities, semantic relationships and provenance. However, practical implementations face several technical and organizational challenges that, if not carefully addressed, can hinder a wide adoption of this methodology.
The main challenges include:
• Designing an ontology that accurately models the scientific domain while remaining reusable, extensible, and interoperable.
• Collecting complete, consistent and semantically rich metadata satisfying ontology constraints and mandatory metadata requirements
• Minimizing manual metadata entry while maintaining high metadata quality.
• Integrating heterogeneous data sources
• Ensuring the consistent assignment, resolution, and linkage of persistent identifiers
• Providing intuitive, user-friendly interfaces for metadata representation, navigation, and editing.
• Supporting metadata validation and quality assurance throughout the metadata lifecycle.
Among these challenges, the generation of RDF data by end users remains one of the biggest hurdle, as it typically requires knowledge of semantic technologies and ontology structures. Another major challenge is the highly variable context of research metadata. Depending on the type of research object, experiment, workflow, or scientific discipline, both the number and the semantics of mandatory metadata fields may change considerably. Consequently, metadata acquisition systems must be sufficiently flexible to adapt dynamically to different metadata contexts while still ensuring semantic consistency and FAIR compliance.
To systematically address these challenges, we have developed the TRIQ,[1] a web-based platform for the interactive creation, validation, browsing, and management of FAIR research metadata. At its core is the TRIQ Core Java module, built on the Apache Jena semantic web library.[2,3] A key feature of TRIQ is its decision tree–based guidance for metadata creation, editing, assignment, and semantic linking. Rather than requiring users to manually identify all relevant metadata fields, the system dynamically derives them by reasoning over the metadata already provided together with formal ontology constraints, such as OWL restriction classes or SHACL shape definitions. Consequently, only metadata fields that are valid and required in the current semantic context are presented to the user.
In our presentation,we also describe the key principles underlying the ontology design and the program architecture of the TRIQ Core Java module. In particular, we demonstrate how defined OWL classes play a central role in enabling ontology-driven, decision tree–based guidance for metadata acquisition.
Keywords: NFDI4Cat, knowledge graph, ontology, metadata
[1]. T. Petrenko et al., Zenodo, 2025. doi: https://doi.org/10.5281/zenodo.14389435
[2]. McBride, B. IEEE Internet Computing, 6(6), 55–59, 2002.
[3]. Apache Jena, https://jena.apache.org/
| background | NFDI4Cat |
|---|