Ontology management

Adding axioms to ontologies

The method insert_annotations allows adding new axioms in the form \(C \sqsubseteq \exists R.D\). Entities \(C\), \(R\) and \(D\) must be stored in a .tsv file.

For example, let’s say we have an ontology called MyOntology.owl and we want to add new axioms with a relation http://has_annotation.

The axiom information must be stored in an annotations file with the following format:

http://prefix1/class1        http://prefix3/class3
http://prefix2/class2            http://prefix4/class4           http://prefix5/class5

Then to add that information to the ontology we use the following instructions:

from mowl.ontology.extend import insert_annotations
annotation_data_1 = ("annots.tsv", "http://has_annotation", True)
annotations = [annotation_data_1] # There  could be more than 1 annotations file.
insert_annotations("MyOntology.owl", annotations, out_file = None)

The annotations will be added to the ontology and since out_file = None, the input ontology will be overwritten.

Note

Notice that the variable annotation_document_1 has three elements. The first is the path of the annotations document, the second is the label of the relation for all the annotations and the third is a parameter indicating if the annotation is directed or not; in the case it is set to False, the axiom will be added in both directions (\(C \sqsubseteq \exists R.D\) and \(D \sqsubseteq \exists R.C\)).

In our example, the axioms inserted in the ontology will be the following in XML/OWL format:

       <Class rdf:about="http://prefix1/class1">
     <rdfs:subClassOf>
         <Restriction>
             <onProperty rdf:resource="http:///has_annotation"/>
             <someValuesFrom rdf:resource="http://prefix2/class2"/>
         </Restriction>
     </rdfs:subClassOf>
 </Class>

<Class rdf:about="http:///prefix2/class2">
     <rdfs:subClassOf>
         <Restriction>
             <onProperty rdf:resource="http:///has_annotation"/>
             <someValuesFrom rdf:resource="http://prefix4/class4"/>
         </Restriction>
     </rdfs:subClassOf>
 </Class>

<Class rdf:about="http:///prefix2/class2">
     <rdfs:subClassOf>
         <Restriction>
             <onProperty rdf:resource="http:///has_annotation"/>
             <someValuesFrom rdf:resource="http://prefix5/class5"/>
         </Restriction>
     </rdfs:subClassOf>
 </Class>

Creating ontology from triples

To transform a triples from a .tsv file into a .owl, we can do using the create_from_triples method. As before, an input triple (h,r,t) will be inserted as axioms of the form \(H \sqsubseteq \exists R.T\).

Let’s assume we have a triples file called my_triples_file.tsv of the following form:

http://mowl/class1    http://mowl/relation1    http://mowl/class2
http://mowl/class2    http://mowl/relation4    http://mowl/class3
http://mowl/class5    http://mowl/relation2    http://mowl/class2
http://mowl/class1    http://mowl/relation1    http://mowl/class3

To create an ontology from those triples we would do:

from mowl.ontology.create import create_from_triples

triples_file = "my_triples_file.tsv"
out_file = "my_new_ontology.owl"

create_from_triples(triples_file, out_file)

In case we have a simpler triples file like the following:

class1    class2
class2    class3
class5    class2
class1    class3

we can create an ontology assuming all the triples will have the same relation and also inputting a prefix for all the classes:

from mowl.ontology.create import create_from_triples

triples_file = "simpler_triples_file.tsv"
out_file = "my_new_ontology.owl"
prefix = "http://mowl/"
relation = "http://mowl/relation"

create_from_triples(triples_file,
                    out_file,
                    relation_name = relation,
                    bidirectional = True,
                    head_prefix=prefix,
                    tail_prefix=prefix)

Note

The bidirectional parameter indicates whether the graph will be directed or undirected.

\(\mathcal{EL}\) normalization

The \(\mathcal{EL}\) language is part of the Description Logics family. Concept descriptions in \(\mathcal{EL}\) can be expressed in the following normal forms:

\[\begin{split}\begin{align} C &\sqsubseteq D & (\text{GCI 0}) \\ C_1 \sqcap C_2 &\sqsubseteq D & (\text{GCI 1}) \\ C &\sqsubseteq \exists R. D & (\text{GCI 2})\\ \exists R. C &\sqsubseteq D & (\text{GCI 3}) \end{align}\end{split}\]

Hint

GCI stands for General Concept Inclusion

The bottom concept can exist in the right side of GCIs 0,1,3 only, which can be considered as special cases and extend the normal forms to include the following:

\[\begin{split}\begin{align} C &\sqsubseteq \bot & (\text{GCI BOT 0}) \\ C_1 \sqcap C_2 &\sqsubseteq \bot & (\text{GCI BOT 1}) \\ \exists R. C &\sqsubseteq \bot & (\text{GCI BOT 3}) \end{align}\end{split}\]

We rely on JCEL to provide \(\mathcal{EL}\) normalization by wrapping into the mOWL’s ELNormalizer

from mowl.datasets.builtin import FamilyDataset
from mowl.ontology.normalize import ELNormalizer, GCI

ontology = FamilyDataset().ontology
normalizer = ELNormalizer()
gcis = normalizer.normalize(ontology)

The resulting variable gcis is a dictionary of the form:

Key	Value
“gci0”	list of GCI0
“gci1”	list of GCI1
“gci2”	list of GCI2
“gci3”	list of GCI3
“gci0_bot”	list of GCI0_BOT
“gci1_bot”	list of GCI1_BOT
“gci3_bot”	list of GCI3_BOT