Owlready2 is a module for ontology-oriented programming in Python 3, including an optimized RDF quadstore.

Owlready2 can:

  • Import OWL 2.0 ontologies in NTriples, RDF/XML or OWL/XML format.
  • Export OWL 2.0 ontologies to NTriples or RDF/XML.
  • Manipulates ontology classes, instances and properties transparently, as if they were normal Python objects.
  • Add Python methods to ontology classes.
  • Perform automatic classification of classes and instances, using the HermiT reasoner.
  • Tested up to 100 millions of RDF triples (but can potentially support more).
  • In addition, the quadstore is compatible with the RDFlib Pyton module, which can be used to perform SPARQL queries.

Owlready has been created by Jean-Baptiste Lamy at the LIMICS reseach lab. It is available under the GNU LGPL licence v3. If you use Owlready in scientific works, please cite the following article:

Lamy JB. Owlready: Ontology-oriented programming in Python with automatic classification and high level constructs for biomedical ontologies. Artificial Intelligence In Medicine 2017;80C:11-28

In case of troubles, questions or comments, please use this Forum/Mailing list:

What can I do with Owlready2?

Load an ontology from a local repository, or from Internet:

>>> from owlready2 import *
>>> onto_path.append("/path/to/your/local/ontology/repository")
>>> onto = get_ontology("")
>>> onto.load()

Create new classes in the ontology, possibly mixing OWL constructs and Python methods:

>>> class NonVegetarianPizza(onto.Pizza):
...   equivalent_to = [
...     onto.Pizza
...   & ( onto.has_topping.some(onto.MeatTopping)
...     | onto.has_topping.some(onto.FishTopping)
...     ) ]
...   def eat(self): print("Beurk! I'm vegetarian!")

Access ontology class, and create new instances / individuals:

>>> onto.Pizza
>>> test_pizza = onto.Pizza("test_pizza_owl_identifier")
>>> test_pizza.has_topping = [ onto.CheeseTopping(),
...                            onto.TomatoTopping(),
...                            onto.MeatTopping  () ]

Export to RDF/XML file:


Perform reasoning, and classify instances and classes:

>>> test_pizza.__class__

>>> # Execute HermiT and reparent instances and classes
>>> sync_reasoner()

>>> test_pizza.__class__
Beurk! I'm vegetarian !

For more documentation, look at the doc/ directories in the source.


version 1 - 0.2

  • Fix sync_reasonner and Hermit call under windows (thanks Clare Grasso)

version 1 - 0.3

  • Add warnings
  • Accepts ontologies files that do not ends with '.owl'
  • Fix a bug when loading ontologies including concept without a '#' in their IRI

version 2 - 0.1

  • Full rewrite, including an optimized quadstore

version 2 - 0.2

  • Implement RDFXML parser and generator in Python (no longer use rapper or rdflib)
  • Property chain support
  • Add utility
  • Bugfixes: - Fix breaklines in literal when exporting to NTriples

version 2 - 0.3

  • Add destroy_entity() global function
  • Greatly improve performance for individual creation
  • When searching, allow to use "*" as a jocker for any object
  • Bugfixes: - Fix nested intersections and unions - Fix boolean - Fix bug when removing parent properties - Fix parsing of rdf:ID - Fix multiple loading of the same ontology whose IRI is modified by OWL file, using an ontology alias table - Fix ClassConstruct.subclasses() - Check for properties with multiple incompatible classes (e.g. ObjectProperty and Annotation Property)

version 2 - 0.4

  • Add methods for querying the properties defined for a given individuals, the inverse properties and the relation instances (.get_properties(), .get_inverse_properties() and .get_relations())
  • Add .indirect() method to obtain indirect relations (considering subproperties, transivitity, symmetry and reflexibity)
  • search() now takes into account inheritance and inverse properties
  • search() now accepts 'None' for searching for entities without a given relation
  • Optimize ontology loading by recreating SQL index from scratch
  • Optimize SQL query for transitive quadstore queries, using RECURSIVE Sqlite3 statements
  • Optimize SQL query for obtaining the number of RDF triples (ie len(default_world.graph))
  • Add Artificial Intelligence In Medicine scientific article in doc and Readme
  • Bugfixes: - Fix properties loading when reusing an ontology from a disk-stored quadstore - Fix _inherited_property_value_restrictions() when complement (Not) is involved - Fix restrictions with cardinality - Fix doc on AllDisjoint / AllDifferent