This folder includes codes and data of the paper "OWL2Vec*: Embedding of OWL ontologies", i.e., class membership prediction on HeLis (v1.00), and subsumption prediction on FoodOn (accessed in March 2020, imported ontologies localized) and GO (accessed on 08 August, 2020, with deprecated classes deleted).
- train.csv, valid.csv and test.csv: training, validation and testing axioms extracted from the original ontology (in ontologies.tar.gz).
- xx.train.owl: the ontology after the validation and testing axioms are removed.
- xx.train.projection.ttl: the projection of xx.train.owl.
- classes.txt, individuals.txt: all named classes and individuals (case studies on FoodOn and GO have no individuals.txt).
- annotations.txt, axioms.txt: annotations and declared axioms of xx.train.owl.
- axioms_hermit.txt: all declared axioms plus axioms inferred by HermiT from xx.train.owl.
- inferred_classes.txt: all inferred classes of each testing individual/class except for its ground truth (which is the declared class/super-class to predict).
- The pre-processing can be implemented by OWL JAVA API-based programmes under case_studies/java/: annotations.txt, axioms.txt, axioms_hermit.txt are generated by Ontology_Axioms_Annotations.java; xx.train.projection.ttl is generated by Ontology_Projector.java; train.csv, valid.csv, test.csv, classes.txt, individuals and inferred_classes.txt are generated by ClassAssertion_NormalSplit.java.
$ python OWL2Vec_Plus.py
- See examples with different settings in OWL2Vec_Plus_Run.sh
$ python Baselines.py
-
See examples with different settings in Baseline_Run.sh (e.g., RDF2Vec_Run.sh)
-
For TransR/DistMult/TransE, please train the embeddings by OpenKE, and set the learned embedding file in Baselines.py.
-
For Quantum Embedding, please train the embeddings by its codes or use the embeddings trained by us under qembeddings-27000iters, and then run Quantum_Run.
-
For EL Embedding, please train the embeddings by its own codes, and set the learned embedding file in Baselines.py.