Does the linked data web need OWL/RIF/SWRL etc?

A slight rephrasing of the question posed at ISWC (which was does the linked data web need ontologies - but the definition of ontology was kept sufficiently vague as to make the question somewhat meaningless IMHO)...

"Does the linked data web need expressive ontology languages such as OWL, RIF, SWRL etc."

Personally I'd say we could probably get by without them, but would certainly make the linked data web a better place if they were adopted. It's all very well inserting owl:sameAs between instances, but to get full interoperability between datasets such as (say) Freebase and DBpedia you really do need OWL to do it properly.


I'm not sure if my view on this is unusual, but I think it's worth noting the value of the RDFS and OWL vocabularies as expressive ways to describe the expected shape of data. This information can be used in all sorts of ways short of reasoning. Specifically, I use OWL and RDFS all the time to:

  • Drive user interfaces
    • Provide auto-complete capabilities
    • Auto-generate forms, facets, ...
    • Guide input (e.g. pick lists based on owl:oneOf constructs)
  • Drive client-side validation (yes, this assumes a closed world)
  • Drive server-side validation / integrity constraint checking (yes, this assumes a closed world)

I'm sure there's more, too. But the overall point is that these vocabularies are extremely useful (way more useful than things like XML Schema in my opinion) for simply describing data and relationships.

The fact that in many circumstances I can use these very same schemas and ontologies to reason over my data and derive new data and new classifications that can further simplify users' experiences is simply gravy for me.


PS I should add that rules is a different beast for me. I think rules are incredibly useful, though sometimes opaque to create. I haven't thought a lot about how they interact with The Semantic Web (Linked Data Web), but in general I think the best path forward for rules (& reasoning) is often to make it a "local thing" that happens in the context of a particular application.

I'd make a distinction between OWL and rule-based systems. Both will have a place, but I think rule systems are easier for developers to get to grips with.

I also think that these expressive languages are not necessarily needed as far as we talk about public websites and/or huge public datasets. But in smaller projects, targeted at specific domains, these languages together with reasoning machanisms can provide great features.

I like the idea of "RDFS plus" that Allemang and Hendler present in their excellent "Semantic Web for the Working Ontologist" - ie RDFS plus a few bits of OWL that cover most of the commonly used relationships.

A little bit of OWL-powered inference can be very useful, eg subclasses, symmetric properties, inverse properties etc.

So yes, I think we do need OWL.

(And sorry, I don't know enough about RIF or SWRL to comment on those)