20180906 Ontology Change Improvement Call

2018.09.06

Attendees: Marijane White, Mike Conlon, Ralph O'Flinn, Anna Kazprzik, Muhammad Javed

Agenda: https://wiki.duraspace.org/display/VIVO/2018-09-06+Ontology+Improvement+Call

Reviewing Representing Identifiers in VIVO:

• Raised by Christian because the TIB folks want to add some identifiers
• UFL did this a while ago, they just added data properties, but that's not very meaningful
• Not very clear how to add identifiers to VIVO
• Took different approach with ORCiDs; they are validated by the software and interact with ORCiD API. The Cornell team that added it needed to assign them properties, which required the ORCiD to be an owl:Thing. This is incomplete, what kind of thing is it?
• ORCiD best practice is to require validation, Ralph wonders if there should be a flag to call this out. Mike points out that this is heading in the direction of adding provenance.
• Today's discussion is more focused on defining how identifiers can be added in the future.
• It's not easy to add identifiers because some are data properties and some are object properties
• In BFO, ORCiD is an Information Content Entity in the IAO, which is in turn a Generically Dependent Continuant and a Realizable Entity. But realizations of ORCiDs are not very interesting to VIVO, which leads to some of our confusion over what it is.
• Mike proposes that we create vivo:Identifier as a superclass of identifiers
  • Mike has found evidence that IAO decided not to implement this, because they felt it was too general
  • VIVO operates in the domain of scholarship where our use of identifiers are confined to the domain and are well known to those operating in the domain.
  • If we create this, this explanation should be part of the annotations in the ontology
• Anna asks if we should add annotations to assist with validation of ORCiDs, which we definitely could do.
• Mike discovered that the 'denotes' property is missing an inverse, 'denoted by', which we will add. VIVO can implement this as a faux property.
• VIVO has a property, has_value, but Mike is not sure this is the right one to use.
• Some identifiers are URI, in which case the URI and the value are the same, but all three triples should be implemented regardless, otherwise we end up having to treat ORCiD as a special case. This will also help out in SPARQL queries.
• Some identifiers are considered secret, since VIVO cannot protect such information, such identifiers should not be stored in VIVO.
  • Ralph wonders if there should be a flag for secret things, but Mike and Marijane thinks this would encourage people to do things we don't want them to do. This led to a brief discussion about whether the ontology should change for the software or vice versa, ontological realism, pragmatism, etc. Good fodder for a future conference panel.
• This decision to model identifiers as Information Content Entities has consequences for VIVO, which are discussed in detail Mike's write-up, outlining recommendations and a list of identifiers currently in the ontology.
• Everyone interested in this topic is asked to review the document and send their feedback to Mike.