Whelk

• What it is — Whelk is an OWL reasoner based on the algorithm implemented by ELK. It represents reasoning state as an immutable functional data structure: adding axioms creates a new state while references to earlier states remain unchanged.

• Key concepts —

  • OWL EL classification is the core reasoning capability; Whelk also provides OWL RL and a subset of SWRL for reasoning over individuals.
  • Its persistent reasoner states support preclassifying and storing a shared ontology TBox, then independently extending it with multiple axiom sets such as ABoxes or rolling back to an earlier state.
  • Documented reasoning features include object-property-assertion materialization, SWRL class and object-property atoms, ABox-oriented OWL RL features, extended Self restrictions for rolification, and limited classification of unions in superclass position, such as inferring a common superclass of the union operands.

• How you’d use it — Use Whelk from application code when submitting many description-logic queries, running queries concurrently without blocking, or branching multiple reasoning workloads from a saved TBox classification. A basic OWL API OWLReasoner implementation is included, but the documented preferred route for its immutable-state features is the Scala API.

• LLM angle — none stated

• Pitfalls & lessons —

  • ELK is much faster for ordinary classification of a single ontology; Whelk targets workloads that benefit from persistent states, parallel queries, or its additional ABox/rule features.
  • Classification involving unions in superclass position is explicitly not guaranteed to be complete.
  • The project is under development and its Scala API is in flux; the OWL API interface is described as basic.

• Verdict — A specialized OWL reasoner for immutable, branchable, concurrent reasoning workloads and selected ABox/SWRL features, rather than a speed-first replacement for ELK on basic single-ontology classification.

Sources consulted

• README.md