Use of owl:inverseOf without explicitly stating domain and range attaches all object properties to "thing"

[christopheralcock]

Hello,

Thank you very much for your work on this incredibly useful tool, by far the most usable and helpful thing we've found for visualising ttl files as we discuss and develop our new ontology files here at the UK's Parliamentary Digital Service.

Since your upgrade to WebVOWL 1.0.1 an issue has cropped up that has taken all of our object properties away from their stated domains and ranges.
Your beta (0.5.2) visualised the file like this:
http://vowl.visualdataweb.org/webvowl/#iri=https://raw.githubusercontent.com/ukparliament/Physical-Ontology/master/Proposed-Physical-Ontology-with-certain-subclasses-commented.ttl
But 1.0.1 visualises it like this:
http://visualdataweb.de/webvowl/#iri=https://raw.githubusercontent.com/ukparliament/Physical-Ontology/master/Proposed-Physical-Ontology-with-certain-subclasses-commented.ttl (sorry, was previously displaying from local file rather than github)
I think I've diagnosed the problem though:
We decided to add inverse relationships for all object properties to make sparql queries a little easier for our end users.

eg:

:propertyA
	a owl:ObjectProperty ;
	rdfs:domain :classA ;
	rdfs:range :classB ;
.
:propertyB
	owl:inverseOf :propertyA ;
.

Your beta version understood from the example above that :propertyB would have a domain of :classB and a range of :classA. However, your new version seems to see that :propertyB has no domain and no range, so says that it's domain and range are both "thing". It then seems to realise that :propertyA is the inverse of :propertyB and conclude that its domain and range are both "thing" despite it being explicitly stated that they're :classA and :classB respectively.

I assume that this is an OWL2VOWL issue but it's possible that it happens on the WebVOWL side.

Once again, thank you for this tool and please let me know if there's anything I can do to help.

[christopheralcock]

Hello,

Since the beta version's URL now returns an internal server error 500 I've manually updated the ontology to explicitly state the domains and ranges of inverse properties, meaning my example on 1.0.1 presents correctly, so neither of my examples demonstrate the issue anymore. The issue still exists though as far as I'm aware.

[steffen-l]

Thanks for your feedback and for opening up this issue. It likely results from the different OWL2VOWL versions used in the WebVOWL versions you mentioned. We completely refactored the OWL2VOWL converter when we integrated further language constructs of OWL 2. Unfortunately, this refactoring introduced a few issues that we have to take a closer look at when time permits. Note that both versions of OWL2VOWL - the old and teh refactored one - are still in beta status (0.2 and 0.3) and do NOT visualize all ontology constructs correctly at the moment. Fixing all these issues will likely take some time though (due to limited resources for this workstream).

Thank you for reporting the 500 error. This is caused due to some server reorganization and should (hopefully) be fixed soon.

[christopheralcock]

Thanks very much Steffen.

The issue is easy enough for users to get around as soon as they figure out what's happening, so hopefully having this issue open will help with that

[steffen-l]

Thanks, Christopher. I agree, and the issue might indeed be helpful for others. Nevertheless, our ultimate goal is that WebVOWL viosualizes (nearly) all valid OWL ontologies correctly at some point. As development resources are limited, this might take a while though.

[vitalis-wiens]

Hello Christopher , I took a look at this behaviour and it is really on the owl2vowl side.
It's output generates the extra class thing and the connections.

Thanks again for reporting this issue

[handrews]

@steffen-l regarding limited resources: I don't see a contributor's guide in either this repo or WebVOWL. Are you open to pull requests from outside developers? I'm not promising anything yet- I've been using WebVOWL and OWL2VOWL for all of three days! (love it so far :-) But if you're open to PRs it would be great to know how to coordinate with your road map and work.

[steffen-l]

Thanks, @handrews - as stated in VisualDataWeb/WebVOWL#82, we are generally fine with pull requests and will likely accept them (if we don't have any bigger concerns or objections). Just go ahead!

Regarding our current developing plan: Currently, we are focusing on improving WebVOWL. We don't have the time and resources at the moment to work on OWL2VOWL in parallel but will likely come back to it later this year (e.g. in summer or autumn). We know there are a number of issues in OWL2VOWL that need to be fixed, but they have to wait until we get back to it, or until someone else does. Contributions are always appreciated (though changing the OWL2VOWL code often isn't straightforward).