OWL
Ontology Web Language
OWL use cases
Modeling
Ontologies are created to fix terms in their meaning. So that confusion about their definition is avoided. Many ontologies are written in the OBO format an easily human readable text format. For example GO is defined in OBO. OWL is a more formal definition of an Ontology. This gives it a lot of powerful features that one can use. While also incurring a steep learning curve.
A simple model is:
:woman a rdf:Class .
:woman rdfs:comment "Woman is person who has two XX chromosomes. Per normal healthy cell." . [[br]
:woman rdfs:subClassOf :human .
:mother :subPropertyOf :human .
:mother rdfs:comment "Every human has at least one Mother. We may not know who she is or if she is alive but you must have had one. Clones are not human ;)" .
:human a rdf:Class .
:human rdfs:comment "Human a vague concept" .
Reasoning
OWL allows reasoning over statement using description logic. That means that one does not need to encode all implied information in your system to be able to get conclusive data.
e.g.
jenny a :woman .
jenny :hasMother mum .
mum a :mother .
In a classical system without reasoning when asking the question select ?s where {?p rdf:Type :woman}; will only answer jenny.
After adding the following OWL axiom.
:mother owl:subClassOf :woman .
The answer would become
jenny and mum.
As the axiom leads to the inference that mum is a :Woman. So the new triple
mum a :woman .
Constraint Validation
Lets add a simple constraint too the Ontology. Being rather conservative.
:man owl:disjointWith :woman .
That means if we had made a mistake filling our triple store and we had added a.
mum a :man .
our OWL logic would raise a contradiction.
A product meant for this is http://clarkparsia.com/pellet/icv
In the end I used pellet 2.0.1 for validating the RDF of UniProt?,org to see if it is valid according to our OWL file.
./pellet.sh consistency " http://purl.uniprot.org/uniprot/P12345" " ftp://ftp.uniprot.org//pub/databases/uniprot/current_release/rdf/core.owl"
Which found many errors. Most where datatype errors. e.g. "2009-11" Year-Month is not a valid ISO8601 standard. I changed these to "P2009Y11M" which is valid period encoding. The OWL now has a comment describing what is meant here and why we use this.
For uniprot.org we now also have a script that checks a 1000 entries for correctness.
#!/bin/bash
cd $PELLETDIR
LIMIT=1000
UNIPROT="http://localhost:8080"
for ac in `curl "$UNIPROT/uniprot/?query=*&format=list&limit=$LIMIT"`; do
echo $ac
`curl "$UNIPROT/uniprot/$ac.rdf" -o $ac.rdf`;
./pellet.sh consistency -v $ac.rdf ~/Documents/workspace/uniprot-rdf/src/core.owl
rm $ac.rdf;
done
