MLSea Resource Artifacts

This web page provides the artifacts related to our ESWC 2024 publication entitled
“MLSea: A Semantic Layer for Discoverable Machine Learning”.
The resource includes:

  • The Machine Learning Sailor Ontology (MLSO) documentation.

  • The GitHub repository for the Machine Learning Sailor Ontology and Taxonomies.

  • The Zenodo repository for the RDF snapshots of MLSea-KG.

  • The MLSea-KG SPARQL endpoint.

  • The GitHub repository containing resource code and RML mappings used in the construction of MLSea-KG.

  • MLSeascape, a web application that provides user-friendly access over MLSea-KG.



Artifacts Description

MLSO: A machine learning ontology that reuses and extends state-of-the-art ontologies to describe ML workflows, configurations, experimental results, models, datasets and software implementations.

Description of the figure
Figure 1: MLSO Overview.

MLST: Eights Simple Knowledge Organization System (SKOS) taxonomies of ML-related concepts (e.g., task types, evaluation measures) with a combined total of 4532 SKOS concepts.

MLSea-KG: A declaratively constructed and regularly updated KG with more than 1.44 billion RDF triples of ML experiments, regarding datasets used in ML experiments, tasks, implementations and related hyper-parameters, experiment executions, their configuration settings and evaluation results, code notebooks and repositories, algorithms, publications, models, scientists and practitioners.

Description of the figure
Figure 2: MLSea-KG Construction Process Overview.