semantic-systems.org / resources
We build tools, resources, and ontologies that improve trustworthiness of intelligent systems — and methods for evaluating them. This page tracks the resources we've published.
more about the group →A human-centric approach to evaluating ontology quality, so that ontologies built into AI systems support factually correct, unbiased behaviour. Funded by the FWF Austrian Science Fund.
A visual notation for annotating machine learning pipelines, with a core ontology, a risk-focused extension, a draw.io library and a worked tutorial covering PyTorch workflows.
An ontology for describing machine learning systems on the semantic web — their components, data and provenance — so ML pipelines can be represented as knowledge graphs.
Architecture diagrams accompanying our paper on semantic-based data augmentation, comparing baseline, embedding- and clustering-based strategies for improving ML predictions.
An ontology for explainability in cyber-physical and sensor systems, extending the W3C SOSA/SSN vocabularies. Developed as part of the FFG-funded SENSE project.
An ontology for representing and synthesising causal knowledge, connecting expert domain knowledge with data-driven causal discovery in cyber-physical systems.
An ontology for object-centric event data, plus a translator that converts XES event logs into OCEDO-compliant RDF. Developed with the University of Rome and Utrecht University.
A common data format ontology for representing football match and event data, published as JSON-LD, RDF/XML, N-Triples and Turtle with a WebVOWL visualisation.