Add a --include-indirect option to materialize

[cmungall]

From slack:

Hi all, I have a question about the use of the materialize command. Let's consider the following axioms:

car SubclassOf has part some wheel
convertible SubclassOf car

I would like to assert the inferred axiom:

convertible SubclassOf has part some wheel

My comprehension is that I should use the materialize command like this: robot materialize -i input.owl --reasoner ELK --term IRI:convertible -o output.owl

However, I don't see any difference between input and output (Robot v1.9.6)

[cmungall]

My response:

Materialize still produces an ontology that is in some sense non-redundant. It lacks the analog of --include-indirect in reason

However, we should clarify the intended semantics. the words "direct" and "redundant" are often used imprecisely

consider:

car SubclassOf has part some wheel
convertible SubclassOf car
wheel SubClassOf has part some hubcap
Transitive(has part)

visualized as:

Would you also expect to see convertible has-part hubcap?

This is in some sense the most straightforward scenario and is the output of relation-graph

If you want the most specific relationships materialized, currently this is only handled by OAK:

• https://github.com/INCATools/ontology-access-kit/releases/tag/v0.6.4
• Add option to relationships for walking up subclass hierarchy INCATools/ontology-access-kit#739

runoak -i car.db relationships .all --non-redundant-entailed

subject	predicate	object	subject_label	predicate_label	object_label
EXAMPLE:car	BFO:0000051	EXAMPLE:wheel	car	has part	wheel
EXAMPLE:convertible	BFO:0000051	EXAMPLE:wheel	convertible	has part	wheel
EXAMPLE:convertible	rdfs:subClassOf	EXAMPLE:car	convertible	None	car
EXAMPLE:wheel	BFO:0000051	EXAMPLE:hubcap	wheel	has part	hubcap

runoak -i car.db relationships .all --include-entailed

subject	predicate	object	subject_label	predicate_label	object_label
EXAMPLE:car	BFO:0000051	EXAMPLE:hubcap	car	has part	hubcap
EXAMPLE:car	BFO:0000051	EXAMPLE:wheel	car	has part	wheel
EXAMPLE:car	rdfs:subClassOf	EXAMPLE:car	car	None	car
EXAMPLE:convertible	BFO:0000051	EXAMPLE:hubcap	convertible	has part	hubcap
EXAMPLE:convertible	BFO:0000051	EXAMPLE:wheel	convertible	has part	wheel
EXAMPLE:convertible	rdfs:subClassOf	EXAMPLE:car	convertible	None	car
EXAMPLE:convertible	rdfs:subClassOf	EXAMPLE:convertible	convertible	None	convertible
EXAMPLE:hubcap	rdfs:subClassOf	EXAMPLE:hubcap	hubcap	None	hubcap
EXAMPLE:wheel	BFO:0000051	EXAMPLE:hubcap	wheel	has part	hubcap
EXAMPLE:wheel	rdfs:subClassOf	EXAMPLE:wheel	wheel	None	wheel

[cmungall]

Note: it's OK to close this as out of scope, as robot doesn't have a very graphy view of the universe, but I wanted to capture the discussion where it won't be swallowed by the slack memory vortex

[cmungall]

An alternative is the plain reason command with emr as the reasoner:

robot reason --include-indirect true --reasoner emr -i car.owl -o car2.owl

this includes things like 'car has-part hubcap`

[souzadevinicius]

I want to start a bit with robot development, so I decided to pick this issue.
However, it didn't work, the key is this:

Set<OWLClassExpression> sces = emr.getSuperClassExpressions(c, !includeIndirect);

@cmungall you know how better how emr works. Can you help me with that?

[pfabry]

An alternative is the plain reason command with emr as the reasoner:

robot reason --include-indirect true --reasoner emr -i car.owl -o car2.owl

this includes things like 'car has-part hubcap`

Unfortunately, this adds also a lot of "noise" such as 'convertible subClassOf owl:thing'

[balhoff]

One option would be to use relation-graph and then merge the output using ROBOT:

relation-graph --ontology-file input.owl --mode OWL --output-file relations.nt --property 'http://purl.obolibrary.org/obo/RO_0002162' --output-subclasses false

or leave off the property argument if you want all properties materialized.