Moose Method for Latent Concept Learning in OWL 2 EL Ontologies
August 14, 2026
Moose compiles EL++ TBoxes and ABoxes into Sentential Decision Diagrams (SDDs) to enable differentiable weighted-model-counting. This neuro-symbolic method allows for latent-concept learning in large-scale production ontologies like SNOMED CT while maintaining formal soundness.
HOW THIS AFFECTS YOU
●
researcherThis provides a formal framework for neuro-symbolic learning that handles reasoning shortcuts in ontology settings.
●
healthThis could improve the automated reasoning and classification of large-scale clinical ontologies.