Express COB as a formal schema

[cmungall]

We frequently have requests for:

• general modeling patterns using COB
• visualizing how COB classes should connect together
• validating COB in a way that is stricter and captures intuitive connections more than OWA OWL reasoning

We should do this, and do it in a machine actionable way. Hand drawn diagrams are always nice but they become stale, they are ambiguous, cannot be used for validation or for machine guided data entry

There are a few candidate systems

• template-oriented
  1. dosdp
  2. robot template
  3. ottr
• schema / validation
  1. UML
  2. json-schema
  3. LinkML
  4. SHACL
  5. ShEx

Before expressing preference we should first outline criteria, for example:

• produces diagrams useful to non OWL experts
• can be used to valid RDF aboxes
• works off of "native" ABoxes
• can express examples
• can assist in generation of axioms (e.g. via UI or spreadsheet compilation)
• ...

The part about native ABoxes is nuanced yet important. @balhoff and I have experience in using ShEx to model highly connected OBO-compliant ABox graphs. Many schema languages have type theory that is at odds with an OBO realist type theory, OBO lacks a schema and instead rdf:types range over massive TBoxes. This has implications for balancing the criteria above.

[matentzn]

How complex would this schema be? Cant we just build one to start working, and later change the primary representation language?

[wdduncan]

@cmungall Would you be doing this in LinkML and then translating to ShEx, SHACL, Json-schema, etc.?