Jade Franklin
This work presents a novel method of using web ontology language description logic semantics to match knowledge graph entities to ontology concepts. The two main contributions of the method are: (1) It outputs a score from 0.0 to 1.0 describing the relative degree to which the knowledge graph instance structure within the graph matches the structure defined by the ontology concept, and (2) It uses a novel, GPU-capable algorithm, in which constraints are modeled as a directed acyclic graph of dependencies that are solved from leaf to root. The algorithm was implemented and tested on the GPU and evaluated against an Apache Fuseki triplestore running equivalent SPARQL queries, and achieved performance improvements of several orders of magnitude.
Links:
- Final paper: https://drive.google.com/file/d/15ZrDhcL2RWg4dlMur9ADHW7tXYXRGvAv/
- Final presentation (slides): https://docs.google.com/presentation/d/1lU2XVyPDe6tdljUASn5ECGxoeuNd7bvCtBhQk7-aeXc/
- Final presentation (video): https://youtu.be/qImTojBYG-8
- github repository: https://github.com/frankj-rpi/Scored-semantic-alignment-via-OWL-DL-semantics-on-the-GPU-/
