Challenges with hasSelf Relation

[H_Borgelt]

Hello Jean Baptise,

thanks again for implementing the general class axiom, as of now it works beautifully, however I have encountered another challenge that I hope you can help me with. I tried to implement an ontology which shall use an n-ary structure (http://ontologydesignpatterns.org/wiki/Submissions:N-Ary_Relation_Pattern_%28OWL_2%29) for describing relations between chemical entities. However, I ran into the problem that I could not find the correct file format for loading relations with complex “hasSelf” relations into owlready2. Which should look like the relation depicted below. Here the circle filled with Me1_Sub is a class with should point with the relation has_Me1_en to an individual S_Ni_ind. As this relation shall be generated automatically an additional loop from the individual S_Ni_Ind to itself must be generated for further inferring of axioms.


In Protégé’s OWL Manchester syntax, this would look like this:
 

However, when I try to load an ontologie in an rdf/xml or owl/xml file format into python using owlready2’s onto.load() I get a relation looking like this:

  for rdf xml or like this:

  for owl/xml.
I first found out that this happened after saving the ontology and reopening it in protégé, where the R1_meEn Object-Property had been changed to a class like structure, however I think this only happened due to faulty syntax after saving the ontology.
I tried narrowing the error down as much as possible and think it is a parsing/reading error, where the “hasSelf” relation of the owl2 syntax is not correctly interpreted. As I am not familiar how the conversion process from owl2 file to object-oriented-ontology in python works, I am sorry to tell you that I could not narrow it down further. I also checked the "as_rdflib_graph()" function, to check whether the quadstore is un affected, however this also did not work.

I have used the code given below, with the files you can find attached to this mail in order to produce the error. I made sure to have updated owlready2 to 0.40 and use rdflib in version 6.2.0.


Short_Example_for_Issue_before_loading.rdf
Short_Example_for_Issue_after_saving.rdf
Short_Example_for_Issue_after_serialisation.ttl


Code:
import owlready2
import rdflib
from ontopy import World
from owlready2 import GeneralClassAxiom

owlready2.reasoning.JAVA_MEMORY = 1000
owlready2.JAVA_EXE = 'C:/Users/xxx/Desktop/Protege-5.5.0/jre/bin/java.exe'
world_3 = World()
onto = world_3.get_ontology("C:/Users/xxx/Short_Example_for_Issue_before_loading.rdf")
onto.name = 'onto'
onto.load()

gcas = list(onto.general_class_axioms())
classes = list(onto.classes())
individuals = list(onto.individuals())
object_properties = list(onto.object_properties())

classes_2 = list(onto.classes())
onto.save("C:/Users/xxx/Short_Example_for_Issue_after_saving.rdf", format = "rdfxml")
onto.save("C:/Users/xxx/Short_Example_for_Issue_after_saving.owl")
graph = world_3.as_rdflib_graph()
graph.serialize(destination="C:/Users/xxx/Short_Example_for_Issue_after_serialisation.ttl", format="ttl")

I don’t know if I have done any mistake myself, by loading in the wrong syntax or by misusing a structure/syntax not meant to be used in such a way. Therefore, I am grateful for any support you can give me. Also, feel free to contact me when you have any questions or if and when I can be of any assistants.

Best regards
Hendrik Borgelt

[Jiba]

Hello,

I've just fixed two bugs on hasSelf relation, in the development version of Owlready. The first one was in the OWL/XML parser (you'll need to recompile the optimized C module if you are using it), and the second one in the loading of hasSelf relations from the quadstore (they are actually "data" triples and not "object" triples).

It now appears as follows:

[Short_Example_for_Issue_before_loading.Me_Class, Short_Example_for_Issue_before_loading.has__Me1__en.some(OneOf([Short_Example_for_Issue_before_loading.S_Ni_ind]) & Short_Example_for_Issue_before_loading.R1_meEn.has_self())]

Jiba