Create new DanNet ontology

[simongray]

The few DanNet relations that could not be represented using the GWA relations will need to be represented inside a new OWL file with a fitting namespaces. This file can probably be generated using one of these Clojure OWL libraries:

• https://github.com/phillord/tawny-owl
• https://gitlab.com/sam_russell/porta-owl-sqwrl

If that is not ideal, then writing them by hand is also an option.

Furthermore, Ontolex OWL (and other) files will also need to be referenced inside Apache Jena instance. I will also likely have so set up some kind of Inference for opposite relations.

[simongray]

The schemas are available here (and I think they are OWL?): https://github.com/globalwordnet/schemas

Now I need to find out how to load the existing schemas. Jena has some documentation on this, but I need to figure out the difference beween the Ontology API (haven't read the documentation) and the inference API (already read it). And what kind of inference we will be needing... or whether is a better idea to preprocess rather than infer.

[simongray]

Inferencing was handled as part of #8, so this issue now solely pertains to creating and adding a new DanNet schema.

[simongray]

It seems that I have to choose between either Ontolex (wn-lemon-1.1.ttl) or valuable OWL info (wn-simple-1.1.ttl) such as owl:inverseOf which I need for inference. In that light, I will either need to

• convince the GWA people to add information such as owl:inverseOf to their fairly sparse Ontolex schemas
• or create my own very elaborate DanNet schema which - at the very least - lists equivalent and inverse relations for every relation in the GWA set.
  • in case the latter becomes necessary, I will need to re-namespace the GWA triples I'm creating in the process.

[simongray]

GWA have suggested that I contribute to their schemas, although they need to get back to me on the formalities. If they are OK with a PR adding in owl:inverseOf to their 1.1 schemas, most of the issues are effectively resolved (although I still wish the OWL statements were much more elaborate).

[simongray]

An MVP of a DanNet concepts ontology was added in 4e1e1cb in the Turtle format.

[simongray]

I made an issue about the missing owl:inverseOf relations and they agreed to put this in the 1.2 release, provided I make the necessary changes: globalwordnet/schemas#53 (comment)

The relevant PR: globalwordnet/schemas#54 (comment)

[simongray]

The owl:inverseOf relations are included as of 1c8aa9e in a tentative version 1.2 of the GWA schema.

I have also made a new, separate PR for inclusion in the 1.2 branch of the GWA schema: globalwordnet/schemas#57

---

Now just a few remaining DanNet-specific things need to be added to the DanNet schema allowing this issue to finally be closed.

[simongray]

TODO schema definitions:

• hypernym_ortho (in use, but not defined in the old DanNet schema)
• connotation (e.g. "negative", don't I use this data right now, but it's in the old DanNet RDF)
• sex (... in the old DanNet RDF) Does not seem to be in use

[simongray]

Ignoring connotation for now as this has been pulled into #16.

Work on version 1.2 of the GWA schema is ongoing. This will eventually allow the bootstrap code to create the wn:partOfSpeech triple only, inferring the ontolex:partOfSpeech from it. See: #17 for more.