How to align an ontology with COB

[sebastianduesing]

@matentzn suggested I create and post a COB alignment checklist. This is a first draft of that document based on my experience working on OBI's COB transition. The document will be added onto as more ontologies work their way through the process and we get a better sense for which steps are universally helpful and which are more ontology-specific. Other feedback is also welcome.

[matentzn]

@sebastianduesing we are highly dependent on RO integrity checking, especially things like "a continuant can never be part of an occurrent". What is the recommendation to persist such integrity checks right now? In the absence of a solution, can we have BFO and COB side by side if we need these integrity checks?

[sebastianduesing]

@matentzn That's a good question. I'm not too familiar with exactly which RO integrity checks are implemented (or how they're implemented)—if you could point me to where to look for those, I'd be happy to dig in and come back with a more thorough answer. Off the top of my head, though, my first thought would be to do one of the following:

1. Where possible, substitute disjunctions or COB approximates for high-level BFO terms. In the case of the "a continuant can never be part of an occurrent" check, the only COB branch that is a BFO:occurrent is the process branch, so we'd instead be checking that no anatomical entity, characteristic, material entity, immaterial entity, or information content entity are parts of a process. It's admittedly inelegant. This is also an approximation of James' original proposal for how to deal with the general problem of RO domains/ranges in COB, and as far as I'm aware, he no longer believes this approach would work for that general problem. I'm not sure if it would work for this narrower problem, but it did come to mind.
2. Keep a BFO-based version of COB side by side with standard COB, like you suggested, more or less solely for the purpose of running checks. Ideally we'd want to have some workflow in place to check that no changes made to cob.owl caused problems in hypothetical-bfo-cob.owl. My only variation from your proposal would be to treat this as a solution in its own right, haha.

These are just first impressions, and I need to do a bit more research before I put together a better answer. Suggestions from others are absolutely welcome.

[jamesaoverton]

See #119 for some important technical details about COB alignment.