Hi all, just got to know Stardog and very excited. I got two questions:
In Stardog website, it says it is an Enterprise Knowledge Graph solution. The term of Knowledge Graph initially came from Google Knowledge Graph.
Do you distinguish between Knowledge Graph and OWL Ontology when using the term “Enterprise Knowledge Graph”?
It seems Google Knowledge Graph isn’t intended to represent natural language meanings as OWL ontology does, and so it’s simply a store of RDF triples. Stardog natively supports OWL ontology, which supports more fine-grained knowledge representation based on OWL. I am afraid it’s a bit confusing to use the term Knowledge Graph in this context. Could Startdog team shed some light on this? Thanks.
Can an OWL ontology created in Stanford Protege be imported into Stardog seamlessly without any semantic loss?
I’m not sure I really understand yr question. Google has no exclusive monopoly on “knowledge graph”. Since they are building one out of public Web documents, it’s not surprising that it’s built and looks somewhat different from Stardog Knowledge Graph. For example, Stardog Knowledge Graph query language is much more expressive than Google’s. Of course that’s because Google does this for a planetary-scale set of docs. OWL ontologies are part of how Stardog manages Enterprise Knowledge Graphs since we use them for schema alignment and ontology-based data access (Global Local as View source access queries automatically generated from schema mappings etc).
Thanks for the explanation. My real question was that Google Knowledge Graph represents factual knowledge in purely dependency triples, i.e. Obama wasBorn USA, and this triple may be simply stored in a relational database. But in OWL ontology you have to use quantifiers like Some, All, Exactly 1, etc. to enforce restrictions to make the semantics precise. Also, in OWL ontology, you can represent meanings like such sentences:
If John loves Mary, he will buy a present for her. (Condition)
John first entered the room, and after 5 minutes, he left. (temporal sequence)
Even in OWL, you may have rules (underlying description logic) to help represent knowledge extracted from sentences like above. A simply triple can’t express the meaning, and I feel this is what makes OWL ontology different from Google Knowledge Graph.
So I feel OWL ontology is a Knowledge Graph, but we can’t say Knowledge Graph (the original meaning of Google Knowledge Graph) is an ontology. So Stardog is really designed for OWL ontology, and that makes it different from other graph databases. That may also be the reason why Google didn’t use the existing “ontology” term, but chose to use the term Knowledge Graph, not ontology.
How do think of the difference between Google Knowledge Graph, ontology and Stardog Knowledge Graph?